Selected Papers in Structural Linguistics: Contributions to English and General Linguistics Written in the Years 1928–1978 9783110816631, 9789027931481

Table of contents :
Preface
Acknowledgements
List of Abbreviations
Introduction: A Brief Survey of the life and Work of Bohumil Trnka
SECTION ONE. GENERAL LINGUISTICS
Methode de Comparaison Analytique de Grammaire Comparée Historique (1929)
About Analogy in Structural Linguistics (1936)
General Problems of Structural Linguistics (1943)
Linguistics and the Ideological Structure of the Period (1948)
Zur Erinnerung an August Schleicher (1952)
Prague Structural Linguistics (1958)
A Theory of Proper Names (1958)
On the Linguistic Sign and the Multilevel Organisation of Language (1964)
A Remark Concerning the Linguistic Sign and Communication (1966)
Words, Semantemes, and Sememes (1967)
On Analogy (1968)
SECTION TWO SYNCHRONIC PHONOLOGY
General Laws of the Phonemic Combinations (1936)
On the Combinatory Variants and Neutralisations of Phonemes (1938)
Norwegian Alveolar Consonants (1941)
About Monophonemic Words (1941)
Phonological Foreignisms in Czech (1942)
The Determination of the Phoneme (1954)
On Some Problems of Neutralisation (1958)
On Foreign Phonological Features in Present-day English (1964)
The Phonemic Organisation of Morphemes (1967)
On the Relationship of Phonemes to Sounds (1974)
SECTION THREE. STATISTICAL LINGUISTICS
Quantitative Linguistics (1951)
The Distribution of Vowel Length and its frequency in Czech (1966)
On the Frequency and Distribution of Consonant Clusters in Czech (1972)
SECTION FOUR. HISTORICAL LINGUISTICS: DIACHRONIC PHONOLOGY AND MORPHOLOGY
Some Remarks on the Perfective and Imperfective Aspects in Gothic (1929)
Can Verner’s Law be applied to Modern English? (1934/35)
The Phonological Development of Germanic Vowels (1935/36)
The Phonemic Development of Spirants in English (1938)
Phonological Remarks Concerning Scandinavian Runic Writing (1939)
Some Remarks Concerning Germanic Expressive Gemination (1940)
From Germanic to English. A Chapter from Historical English Phonology (1948)
The Old English Diminutive Suffix — INCEL (1956)
A Phonemic Aspect of the Great Vowelshift (1959)
About Morphonological Analogy (1961)
On the Change of Middle English -erC into -arC in Early New English (1969)
The Old English Vowel System and the Problem of Monophonemes (1975)
SECTION FIVE. SYNCHRONIC MORPHOLOGY, SYNTAX AND STYLE
Analysis and Synthesis in English (1928)
Bemerkungen zur Homonymie (1931)
Some Thoughts on Structural Morphology (1932)
Morphological Oppositions (1958)
Autonomous and Syntagmatic Words (1960)
Principles of Morphological Analysis (1961)
On the Morphological Classification of Words (1962)
On Morphemic Homonymy (1963)
On the Basic Categories of Syntagmatic Morphology (1966)
On Word Order in Structural linguistics (1968)
Conversion in English (1969)
A Few Remarks on Homonymy and Neutralisation (1974)
The Theory of Sign Levels and the Relation of Words to Morphemes (1978)
The Problem of Style (1941)
Afterword
Index of Names
Index of Subjects

Citation preview

Selected Papers in Structural Linguistics

JANUA LINGUARUM Studia Memoriae Nicolai van Wijk Dedicata edenda curat

C. H. van Schooneveld Indiana University

Series Maior 88

Bohumil Trnka

Selected Papers in Structural Linguistics Contributions to English and General Linguistics written in the years 1928-1978 Afterword by Roman Jakobson Vilém Fried (Editor)

M o u t o n Publishers Berlin · N e w York · Amsterdam

Library of Congress Cataloging in Publication Data Trnka, Bohumil, 1 8 9 5 Selected papers in structural linguistics. (Janua linguarum. Series maior ; 88) English, French, and German. 1. Structural linguistics-Collected works. I. Fried, Vilém. II. Title. III. Series. P146.T74 410 81-11035 ISBN 90-279-3148-8 AACR2

© Copyright 1982 by Walter de Gruyter & Co., Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced in any form - by photoprint, microfilm, or any other means - not transmitted nor translated into a machine language without written permission from the publishers. Typesetting: R. James Hall, Harpenden, England. - Printing: Druckerei Hildebrand, Berlin. Binding: Lüderitz & Bauer Buchgewerbe GmbH, Berlin. Printed in Germany.

Preface

The idea of publishing a volume of Professor Bohumil Trnka's linguistic papers was brought up many years ago. From the mid-sixties we, i.e. some of his former students and later close collaborators, began to urge him to consider a re-publication of his contributions to numerous fields of linguistic research, written over a span of many years, in view of the revival of interest in Prague School Linguistics in countries both East and West. This interest was supported not only by the re-publication of many contributions published by the members of the Cercle Linguistique de Prague in its classical pre-war period, and by numerous accounts of the linguistic doctrine of the 'Prague School', but also by the active and permanent interest taken in the advancement of Prague School functional and structural theory, as it has become manifest in the large number of linguistic publications by the follow-up generation of Czech and Slovak linguists, both in Czechoslovakia and abroad, in the past three decades. In our view Bohumil Trnka was not only one of the founders of the Prague Circle and thus a very important representative of its classical period, but in his later writings he also presented himself as the most consistent and unswerving pursuer and developer of its basic theoretical tenets. Professor Trnka at first dismissed the idea of re-publication of his papers saying that he intended to write a separate linguistic monograph in which he wanted to explain specifically his linguistic theory. At the turn of the nineteen-seventies, however, he began to realise that there was little chance of such a monograph being accepted for publication in Prague. In the same period it so happened that several of Trnka's students had taken up positions outside Czechoslovakia. Eventually, soon after his eightieth birthday, Professor Trnka agreed to my suggestion that he consider the preparation of a publication of his selected papers. He prepared the first draft and I was privileged to submit the proposal t o the publisher. I would most gratefully like to acknowledge the immediate, almost enthusiastic response in favour of the proposed volume that we received from Professor C. H. van Schooneveld (Indiana University) and the late Professor Robert Auty (Oxford University), the great life-long admirer of Czech and Slovak scholarship, literature and art. In particular I should like to express my gratitude to

vi

Preface

Professor Roman Jakobson (Harvard) not only for expressing approval of the Trnka volume, but also for his readiness to write an afterword for it. It so happened that the final arrangement is made up of papers that were written over a period of exactly fifty years. I believe that they are a significant testimony to the history of linguistic thought in this century. The organisation of the volume, as it stands, as well as my translation of the articles originally written in Czech, have been fully endorsed by the author himself. I would like t o emphasise Professor Trnka's most active, alert and attentive cooperation in the course of m y editing; this co-operation was in no way handicapped by the necessary lack of direct personal contacts nor by the often rather capricious postal services. For the final arrangement of the volume I received valuable advice from Professor Miroslav Rensky (City University of New York) who also belonged to Bohumil Trnka's 'inner circle'. My editorship would not have been possible without the help I received from many sides. In the first place I would like to acknowledge with gratitude the generous financial aid for the preparation of the manuscript of this volume given by the "Gesellschaft der Freunde der Niederrheinischen Universität, Duisburg". Further I would like to thank my colleagues in the Department of English of Duisburg University Dr. Jürgen Esser and Dr. Axel Hübler for their help in checking bibliographical references, and Ms Christine Klein-Braley for her valuable advice about the adequate use of English. I thank Professor Takahashi and Ms. J.K. Glasser of Hiroshima University for their help in finding bibliographical data of Trnka's work in Japan. Finally I would like to thank Ms Gisela Görtz, Ms Ute Intveen-Theuerkauf and Ms Barbara Hoffmann for their devoted and often exhausting secretarial assistance. All responsibility, however, for mistakes and omissions that may be found in this volume is my own. Duisburg, Federal Germany, January, 1980

Vilém Fried

Acknowledgements

For permission to reprint the following papers grateful acknowledgements are due to the editors and publishers of: Travaux Linguistiques de Prague, published by Academia, Publishing House of the Czechoslovak Academy of Sciences, Prague, for "On the Linguistic Sign and Multilevel Organisation of Language", "On the Basic Categories of Syntagmatic Morphology", "On Word Order in Structural Linguistics"; Prague Studies in Mathematical Linguistics, published by Academia, Prague, for "The Distribution of Vowel Length", and "On the Frequency and Distribution of Clusters in Czech"; Acta Universitatis Carolinae-Philologica, published by the Caroline University, Prague, for "The Phonemic Organisation of Morphemes" and "On the Relationship of Phonemes to Sounds"; Brno Studies in English, published by the J.E. Purkyné University of Brno, for "Conversion in English"; Studii si Cercetâri Lingvistice, published by Academia Republic» Populare Romine, Bucharest, for "A Theory of Proper Names", "Autonomous and Syntagmatic Words", "The Old English Vowel System and the Problem of Monophonemes"; Omagiu lui Iorgu lordati cu prileful împlinirii a 70 de ani, published by Academia Republicii Populare Romîne, for "On Some Problems of Neutralisation"; Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikation, published by Akademie-Verlag, DDR, Berlin, for "Zur Erinnerung an August Schleicher" and "On Analogy (Zum Andenken an F.Slotty)"; Studies in Language and Literature in Honour of Margaret Schlauch, published jointly by Pañstwowe wydawnictwo naukowe, Warsaw, Rüssel & RusselPublishers, New York, for "A Remark Concerning the Linguistic Sign and Communication" ; Prace Filologiczne, published by Pañstwowe wydawnictwo naukowe, Warsaw, for "On Morphemic Homonymy";

vili

Acknowledgements

In Honour of Daniel Jones, published by Longman, London, for "On Foreign Phonological Features in Present-day English"; Mélanges de Linguistique et de Philologie: Fernand Mossé in memoriam, published by Didier Erudition, Paris, for "A Phonemic Aspect of the Great Vowel-Shift"; Lingua, published by North-Holland Publishing Company, Amsterdam, for "On the Morphological Classification of Words"; Poetica, published by Shubun International Co., Tokyo, Japan, for "A Few Remarks on Homonymy and Neutralisation".

Contents

Preface Acknowledgements List of Abbreviations Introduction: A Brief Survey of the Life and Work of Bohumil Trnka, by Vilém Fried SECTION ONE

1

GENERAL LINGUISTICS

Méthode de Comparaison Analytique de Grammaire Comparée Historique (1929) About Analogy in Structural Linguistics (1936) General Problems of Structural Linguistics (1943) linguistics and the Ideological Structure of the Period (1948) Zur Erinnerung an August Schleicher (1952) Prague Structural Linguistics (1958) A Theory of Proper Names (1958) On the Linguistic Sign and the Multilevel Organisation of Language (1964) A Remark Concerning the Linguistic Sign and Communication (1966) Words, Semantemes, and Sememes (1967) On Analogy (1968)

SECTION TWO

ν vii xii

23 29 32 49 61 70 81 86 94 97 102

SYNCHRONIC PHONOLOGY

General Laws of the Phonemic Combinations (1936) On the Combinatory Variants and Neutralisations of Phonemes (1938)

113 119

Contents

χ Norwegian Alveolar Consonants (1941) About Monophonemic Words (1941) Phonological Foreignisms in Czech (1942) The Determination of the Phoneme (1954) On Some Problems of Neutralisation (1958) On Foreign Phonological Features in Present-day English (1964) The Phonemic Organisation of Morphemes (1967) On the Relationship of Phonemes to Sounds (1974)

SECTION THREE

125 128 132 144 149 156 162 165

STATISTICAL LINGUISTICS

Quantitative Linguistics (1951) The Distribution of Vowel Length and its frequency in Czech (1966) On the Frequency and Distribution of Consonant Clusters in Czech (1972)

171 187 195

SECTION FOUR HISTORICAL LINGUISTICS : DIACHRONIC PHONOLOGY AND MORPHOLOGY Some Remarks on the Perfective and Imperfective Aspects in Gothic (1929) Can Verner's Law be applied to Modern English? (1934/35) The Phonological Development of Germanic Vowels (1935/36) The Phonemic Development of Spirants in English (1938) Phonological Remarks Concerning Scandinavian Runic Writing (1939) Some Remarks Concerning Germanic Expressive Gemination (1940) From Germanic to English. A Chapter from Historical English Phonology (1948) The Old English Diminutive Suffix - INCEL (1956) A Phonemic Aspect of the Great Vowelshift (1959) Ab out Morphonological Analogy (1961) On the Change of Middle English -erC into -arC in Early New English (1969) The Old English Vowel System and the Problem of Monophonemes (1975)

205 210 219 224 232 236 245 253 258 263 276 281

Contents

xi

SECTION FIVE SYNCHRONIC MORPHOLOGY, SYNTAX AND STYLE Analysis and Synthesis in English (1928) Bemerkungen zur Homonymie (1931) Some Thoughts on Structural Morphology (1932) Morphological Oppositions (1958) Autonomous and Syntagmatic Words (1960) Principles of Morphological Analysis (1961) On the Morphological Classification of Words (1962) On Morphemic Homonymy (1963) On the Basic Categories of Syntagmatic Morphology (1966) On Word Order in Structural Linguistics (1968) Conversion in English ( 1969) A Few Remarks on Homonymy and Neutralisation (1974) The Theory of Sign Levels and the Relation of Words to Morphemes (1978) The Problem of Style (1941)

287 295 299 303 317 320 333 336 340 345 350 356 363 364

Afterword by Roman Jakobson

383

Index of Names Index of Subjects

385 389

List of Abbreviations

CMF

Casopis pro modernifìlologii [Journal of Modern Philology], Prague, (Founded 1911. From vol. 54 (1972) Supplement to PP)

PBB

Paul Brauns Beiträge ζ. Geschichte der deutschen Sprache und Literatur

PP

Philologica Pragensia. Prague

PSRL

Prague School Reader in Linguistics, edited by Josef Vachek. Bloomington, Indiana University Press, 1964

SaS

Slovo a Slovesnost [Word and Verbal Art], Prague. (Founded 1935)

SW

Selected Writings of Roman Jakobson. Vol. 1 , 2 , 3 , 4 , 5. The Hague, Mouton (1971, 1971,1981,1966,1980.)

TCLP

Travaux du Cercle Linguistique de Prague, Prague. (Vols. 1 -VIII, 1929-1939)

TLP

Travaux Linguistiques de Prague, Prague. (Vols. 1 - 4 , 1964-1971)

VILÉM FRIED

Introduction: A Brief Survey of the Life and Work of Bohumil Trnka

The impact of Prague School linguistics on the twentieth century development of linguistic theory has been generally acknowledged. In all the many accounts of the activities of the Cercle Linguistique de Prague the name of Bohumil Trnka is given prominence. He was one of the founding members of the Circle in 1926, a member of its steering committee and its honorary secretary practically right to the end of the Circle's formal existence at the beginning of the 1950's. Some of Trnka's earlier contributions have been included in the numerous anthologies and readers of modern linguistics that have been published in many countries.1 The Prague School Reader2 contains three articles by Trnka from the classical period of the Prague Circle and the joint article 'Prague School Structural Linguistics' of 1956, whose chief author was Trnka.3 However, this due attention to Trnka's early scholarly work does not do justice to his continuous and extensive linguistic research from the early twenties up to the present.4 Trnka's oeuvre must be looked upon as a significant and highly representative component of what is often referred to as "Prague School Linguistics", a term which extends to Czech and Slovak linguistic contributions published in the period after the Circle had formally ceased to exist and which is meant to reflect a continuous scholarly tradition initiated in the foundation year of the Circle.5 Not only the above mentioned accounts of the Prague School theory and the numerous anthologies have contributed to the widespread information about Prague School linguistics. Mention should also be made of the various volumes of selected papers by former members of the Prague Circle or by younger Czech linguists published so far.6 It appears to be justified that a volume of papers by Bohumil Trnka should be added to this list, the more so as this linguist has adhered to the basic tenets of Prague structural and functional linguistics most consistently in his publications over more than half a century. He has, in my view, applied Prague School linguistic theory to a wide range of linguistic topics in which he brought forward ever new empirical evidence in support of his fundamental linguistic principles.

2

Introduction

Bohumil Trnka was born June 3, 1895 in the vicinity of a small town in Southern Bohemia. His studies in the secondary school (a classical 'gymnasium') brought him to Prague. In 1913 he began his studies at the philosophical faculty of the Czech University of Prague, called the Caroline University of Prague. He took an early interest in languages, already in the upper forms of the secondary school. He was, as he has told me,7 fascinated by questions of why certain Czech expressions and grammatical forms had evidently changed in the course of their history when he compared them with the forms he had learned in other languages, like Latin, Greek, French, Serbocroatian, Russian, German. So also in his undergraduate years his linguistic interests were chiefly and almost solely geared towards comparative language history as this was then the time-honoured tradition in the philologies. His synchronistic interests in language concentrated solely on learning modern languages. As he wanted to pursue his philological studies 'ad fontes', he eagerly studied the older stages of English, German and of the Slavic languages. But he also sought for the psychological and sociological explanations that accounted for the differences existing among genetically non-related languages. He therefore studied thoroughly the works of H. Steinthal, A. Schleicher, H. Paul, W. Wundt, 0 . Jespersen, and many others. In his university years, which were almost entirely overshadowed by the hardships of the First World War, at the end of which the Czechs and Slovaks regained their national independence in the new republic of Czechoslovakia, Trnka was chiefly attracted by the personalities of Professor Vilém Mathesius, the first holder of the chair of English studies at Prague University, of Josef Janko, professor of Germanic philology, by the comparative philologist Oldrich Hujer, and in particular by Josef Zubaty, Professor of Indoeuropean philology and Sanskrit and a scholar of world-wide reputation. 8 Trnka attended Zubaty's lectures on syntax and felt as impressed by him as by Mathesius, because Zubaty, too, was a militant modernist in linguistic thinking who tried to break the fetters of the traditional positivist methods of Neogrammarian linguistics. Another scholar to whom Trnka felt very indebted was Karel Skála, who published under the pseudonym of Ch. Rocher. He was not attached to the University, but seems to have exercised great influence especially on the younger generation of Czech Romance scholars. Rocher was well-known because of his textbooks of modern Romance languages, which were systematically based on Latin. Trnka assisted Skála in his Romance and general linguistic research. Under his guidance Trnka studied 'exotic' living languages, like Malay, Swahili, Turkish, Hungarian, Arabic and others. In 1918 Trnka sat for his final state examinations in Czech and German (English was then not a 'school subject'). In 1920 he was awarded the Ph. D. degree in Germanic philology and English philology on account of his doctoral thesis 'The origin of the Germanic weak verb conjugations', written in German and presented in 1919. From this follows

Introduction

3

that he had by then narrowed his scholarly interests to the domains of Czech, English and Germanic philology. Although Trnka had studied a considerable number of modern and dead languages, although he had acquired a thorough knowledge of the historical development of the European languages, and though he read widely about the history of European linguistics and also studied works in the sciences adjoining with linguistics, he felt that all these studies lacked some unifying principle aiming at a specific goal in linguistic research. From 1922 Trnka established close personal contacts with Vilém Mathesius, who introduced him to his functional principle in linguistic analysis, which opened up entirely new ways in linguistic methodology. Almost from the beginning of his scientific career, i.e., already before 1914, Mathesius had discarded the established and more or less generally accepted doctrines of the Junggrammatiker that were preoccupied with historical phonology and morphology. Mathesius focussed his research on the problems of synchronic syntax. In his linguistic analysis he was mainly interested in the grammatical and semantic functions of words and utterances, observing a concrete language as a functional system which differed from that of another language, related or not (let us say Modern Czech from Modern English) because of certain functional features or because it was similar to it in the domain of other features. It was under the influence of Mathesius that Trnka directed his research towards syntactic problems, while still concentrating on earlier stages of the history of English. He wrote the

monograph "The Syntactic Characteristics of the Language of Anglosaxon Poetry ' (Trnka 1925); it was accepted as his 'habilitatio' by the Caroline University and led to his appointment as professor of English philology and older English literature in 1925. Trnka held this position until 1970, in which year he officially retired, but he continued lecturing almost up to his eightieth birthday. The treatise mentioned, together with an extensive article (Trnka 1924) and a second monograph on the syntax of the verb brought him international repute (Trnka 1930). 9 Trnka analysed OE syntactic phenomena from the point of view of the functional and semantic use of the morphologically determined parts of speech (nouns, adjectives, pronouns, etc.), and from that of the syntactic relations like subject and predicate, adnominal modification, the syntax of cases, and especially the syntactic relations of the verb such as the sequence of tenses, mood, etc., in constant confrontation with the older Germanic languages. These early syntactic treatises already revealed Trnka's method of approach. He followed Mathesius's unifying concept of language as a system of language phenomena and of their function in the system; but Mathesius did not investigate the laws to which these phenomena conform, he was merely interested in their individual communicative functions in the act of communication. Mathesius was already thoroughly acquainted with the principles of Saussurean structuralist concepts

4

Introduction

from the early nineteen-twenties, and also later he always had an open mind for new trends in linguistic thinking, e.g., for Russian structural linguistics, which he quickly adapted to the needs of his own linguistic research. Contrary to Mathesius, Trnka's point of departure was the rigorous application of the idea of the systemacity of all language phenomena; he saw the main goal of linguistic science in the discovery of the historical development of the laws operating in the system of language as a whole and prone to changes in space and time. These scientific tendencies are already apparent in Trnka's early treatises although his linguistic concepts had not yet fully matured. In his approach syntactic analysis, which he later preferred to call morphological syntax, remained one of the ultimate goals of his linguistic research in accordance with the theory of the taxonomy of hierarchical linguistic levels, from the lower level to the higher ones. Trnka was convinced that he could not tackle syntactic problems fully before he had solved all outstanding problems arising from his structural and functional analysis of the units of the lower level of the language system. He accepted in principle one of the Saussurean dichotomies, i.e. the distinction of the synchronistic and diachronistic (historical) analysis of a given language as opposed to the traditional concepts of the Junggrammatiker who, in the words of H. Paul, recognised only historical research as the valid and only legitimate scientific endeavour in linguistic science, and fully subscribed to Mathesius's view that the scientific status of the linguistic investigation of contemporary (living) languages was at least equal to that of historical linguistics. He nevertheless refused to draw a strict line of separation between the two approaches in the study of language. He has always adhered to the view that not only synchronic analysis enables us to discover the language system and its underlying laws, but that also the diachronic approach fully contributes to this goal as long as it is based on structuralist principles (Trnka 1933): language never ceases to be a system of structures of its signs (phonemes, words, sentences) and of the relations existing between them, as well as of the relations existing between the units on the various linguistic levels (phonology, morphology, syntax, etc.). Time constantly affects all levels and all components of language. The language historian should not merely amass detailed linguistic data as the prestructuralists did, but he must, on the basis of these data (even if they are undocumented or only poorly documented), conceive the linguistic system of a given period in the history of a language as a coherently functioning whole. Trnka's later contributions to historical linguistics, many of which are included in this volume, are evident proof of one of the characteristic features of Prague School structural and functional linguistics, i.e. the claim that language must be necessarily studied synchronically as well as diachronically if we endeavour to grasp fully its systemic and structural build-up. The language system, in fact, is never static, therefore the researcher must never lose sight

Introduction

5

of its basically dynamic character. Trnka often and repeatedly argued against the atomising positivist methodological concepts of pre-structuralist language historians. Vachek's statement that the 'actual dividing line separating the functionalist and structuralist approach from the traditional Neogrammarian has never been one distinguishing between the synchronistic and the diachronistic study of the language but rather one drawing a distinct line between the systematising and functionalist analysis on the one hand and the atomising approach on the other' 1 0 applies to Trnka's linguistic theory more than to the work of any other member of the Prague School. While Trnka's method of analysis of the syntax of earlier stages in the history of English in the above mentioned monographs may still bear many marks of traditional historical philology, his later structural and functional reappraisals of important events in the history of the Germanic languages such as the development of the Germanic vowel system, the consonant changes (reflected in Grimm's Law and in Verner's Law), the Modern English great vowelshift of phonological and morphological analogy, and many others cannot be separated from Trnka's overall view of language as a functional system of structures. We have already mentioned Trnka's very active participation in the work of the Prague Linguistic Circle; it brought him into close personal contacts with Roman Jakobson and N.S. Trubetzkoy as well as with other internationally renowned linguists. He had a significant share in the preparation of the various collective theses presented by the l'école de Prague at several international linguistic gatherings in the decade preceding the Second World War.11 Trnka took part in the International Congresses of Linguists in the Hague (1928), Geneva (1930), Rome (1933), and Copenhagen (1936). He took, of course, an active share in the first International Phonological Conference, held in Prague in 1930, and in the preparation of its Proceedings (cf. volume IV of TCLP). This collective venture was very much the work of the committee of the Prague Circle, in preparation of the second International Congress of Linguists in Geneva. Trnka was also among those that represented the Prague School concepts at the International Congresses of Phonetic Sciences held in London in 1935, and in Ghent in 1938 as well as at the First International Congress of Slavicists in Prague in 1929, for which the Prague Circle had prepared its well-known collective Theses which were published in the first volume of TCLP (1929). Trnka's international contacts came to a sudden standstill because of the outbreak of the Second World War which he spent more or less unmolested in his Nazi-occupied native country. He was able to pursue his research in the course of which his structuralist linguistic theory matured further. After 1945 Trnka no longer participated in international linguistic gatherings directly. He was appointed rapporteur of the Section for Morphology and prepared in writing his "answer" for the 6th International

6

Introduction

Congress of Linguists in Paris, but did not attend in person (Trnka 1948). At the same congress, he was appointed secretary of the committee for linguistic statistics (Trnka 1950).12 Mathesius (1936) declared that of all the linguistic explorations of the Prague School, phonology was "our main cri de guerre" in international linguistic circles in the 1930's. So it was, indeed. The pioneering work and progress in phonological research before the Second World War has generally been ascribed to Prague School Linguistics, with special reference, of course, to the contributions made by N.S. Trubetzkoy and R. Jakobson. However, one of the most significant contributions to phonological research at that time came from B. Trnka. In 1935 he published his monographs Phonological Analysis of Present-Day Standard English (Trnka 1935) which immediately brought him international recognition; it was the first attempt of an almost complete systematic description of the phonology of a living language based on minute statistical research. The work appeared four years before N.S. Trubetzkoy's "Grundzüge der Phonologie" and it stood at the beginning of the intensive study of phonology and its enormous expansion both in Europe and in America in the following decades. Proof of its long-lasting value is its re-edition in a considerably revised version in Japan in 1966. It can well be said that phonology was the linguistic domain in which Trnka was able to elaborate, refine and advance his linguistic concepts and methodology within a consistent linguistic theory. The logic of his arguments, supported by an impressive knowledge of extensive linguistic data, was so convincing that he was soon looked upon as one of the most outstanding Czech linguists both at home and abroad. In phonological analysis he developed and applied his outstanding faculty of sober and coherent reasoning and abstraction. After 1945 he extrapolated his linguistic concepts equally consistently to the analysis of the linguistic levels of morphology and also partly of syntax, as numerous articles included in this volume demonstrate. For many years Trnka was planning to write a similar monograph of a systematic description of Present-day English morphology. The three volumes of his university textbook, written in Czech, of a synchronistic description of Present-day Standard English may be thought of as the first attempt at realising this plan (especially the second volume).13 However, even his first papers on structural morphology published before 1938 already clearly indicate his line of approach in this domain. In the years after the Second World War he also often returned to problems of phonological theory attempting to define anew some of his concepts (e.g., that of phonemic contrast). Trnka, as he has often stated himself, was very much indebted for his theoretical concepts about language to his teacher Vilém Mathesius, especially for the concepts of linguistic function, system and structure. But while

Introduction

7

Mathesius (like, e.g., Ch. Bally) seemed to prefer to observe language experience and spoke merely of linguistic trends or tendencies, did not believe in linguistic laws that permit no exceptions and was more interested in finding increasingly well-defined functional and semantic distinctions of linguistic forms, Trnka introduced more methodological rigour into the theoretical concept of the language system. "The aim of linguistic research is the discovery of the laws which can be conceived by the logically irrefutable analysis of the general interrelationship and completeness of facts without evasion and concessions. Even if a correction of our formulations appears necessary, I think — in agreement with M. Grammont and R. Jakobson — that a law that needs complementation is more useful than no law" (Trnka 1965). These words, I believe, reflect the basic tenet of Trnka's scientific view. Linguistic research has to be based on a well-defined unified theory, or better, must be an edifice of well-defined theoretical concepts which must be constantly verified or modified in the light of our new linguistic knowledge of empirical data. This rigour of scientific method in a structuralist and functional linguistic theory was already manifest in Trnka's pre-war work and further matured in the course of the following years. Instead of trying to trace the stages in the development of this theory based on an account of Trnka's successive writing, I shall attempt to present a coherent picture of this theory and his concept of linguistic science.14 Trnka's approach to language and linguistics is based on the principles of a functional theory of language as a semiotic system within the theoretical framework of structuralist linguistics. Structural linguistics presupposes that any entity only exists in relation to other entities and that there is no relation that does not entail this existence of entities. Nor is there an absolute disjunction of entities. Language is a system and structure of binary relations of the type Ά stat pro B\ in which the entity A is in sign relation to the entity B\ they appear on hierarchically graded planes (levels), it is their co-operation that makes intersubjective communication about any matter possible. A is the sign (signifier, signifiant, realiser) of the signified entity B, the entity Β is the sign of the entity C, C is the sign of the entity D. In a language system: the phoneme ¡P¡ is the sign of the word /W/, the W is the sign of the sentence ¡S/, and the 5 is the sign of the utterance \U\. The hierarchical system of linguistic signs is manifest in the multilevel organisation of language of at least four hierarchically arranged and co-existent planes (levels): phonology, morphology, syntax, super-syntax (utterance level).15 These levels are to a certain degree autonomous within the overall system of language. Their hierarchy is not constituted by the higher order of complexity: each 'lower' level realises by means of its own entities (units) a 'higher' level so that the levels are both 'réalisé' and 'réalisant'. The co-

8

Introduction

operation of the four co-existent levels of linguistic relationships is indispensable for the normal, non-pathological human communication by means of speech. No single plane makes sense (has a communicative function) without the others. The incorporation of the utterance level among the four basic levels of analysis (involving a particular speaker/hearer and a particular extralinguistic situation) which obeys its own functional systemic and structural laws, makes the Saussurean dichotomy of'langue'/'parole' superfluous; the utterance constitutes a complex linguistic sign by means of its whole semantic and relational content. 16 The sign relations have the following logical features: (a) Ρ and W cannot exchange places: it cannot be said that, if it is true that the phoneme is the sign of the word, the word is the sign of the phoneme. In the relation of the identity Ά is identical with B\ the two entities can be exchanged and we can say that 'B is identical with .4'; (b) the sign relations are both asymmetrical and transitive, as e.g. the relation 'part versus whole': if it is true that the phonemes determine the word and the words determine the sentence, it is also true that the phonemes determine the sentence (syntactic phonology; cf. Trubetzkoy's notion 'morphonology'); and if it is true that the sentence determines the utterance, it is also true that the phonemes determine the utterance. As the language levels constitute a firmly linked hierarchical system, this transitivity is called the realisation; thus the phonemes realise words, indirectly the sentence (via the level of morphology) and indirectly also the utterance (by means of the two lower levels). The levels of the linguistic system are not serial components which succeed one after the other, but are in indirect relation to the whole system, and by the way of it, also its components. The basic unit (or chief entity) of the morphological level is the 'word' which signifies the sentence. It is not the morpheme which we are able to determine only on the ground of the secondary morphemic segmentation of the word into smaller semantic components. The morpheme is the smallest meaningful component of the word, but - like the phoneme — it is bound to the entity of the word and is not displaceable in the word. The morpheme is, therefore, an ancillary semantic component on the morphological level, it has lower status than the word, which realises the sentence directly. The analysis of this sublevel which may be called 'morphemology' or 'morphemics' is a component of morphological analysis, especially in relation to the description of the word stems. A similar parallel existence to that between morphemology and morphology can be observed in the structural interpretation of the functional distinctive features by which phonemes are realised and which as entities of the phonological level are also void of meaning. The functional co-operation of the units of every linguistic level (and their sublevels) results from the systemic oppositions of their units and their linear contrastivity according to the logical differentiation 'either/or' and 'this and

Introduction

9

that'. In linguistics we speak of the co-operation of units along two axes: the paradigmatic (or opposition) axis and the syntagmatic (or contrasting) axis. Along the paradigmatic axis the units of the linguistic plane form a system of oppositions, whereas on the syntagmatic axis the units simultaneously form structures by means of actually existing contrasts and confrontations within such structures. The laws of contrasts differ from the systemic laws of oppositions on each plane. Both axes exist in the systems of all natural languages and are of equal importance. The syntagmatic axis is not only the distributor of the paradigmatic factors, but its laws also affect the functions of these factors (cf. the neutralisation which restricts the phonological oppositions). The planes of the linguistic signs are mutually correspondent. Each plane has its autonomy, which results from the fact that even the mutual relationship between its own units, and not only their relationship to other levels, is meaningful. E.g. the phonemes, as units of the phonological plane, realise words and morphemes and through them they also realise higher planes, but simultaneously each phoneme partakes in oppositions and contrasts with other phonemes in accordance with the phonemic laws of the analysed language. Similarly, words as units of the morphological plane, which realise the sentence and through it the utterance, take part in oppositions and contrasts with other words in a given language. The basic word-classes (parts of speech) in different languages are constituted by different totals of morphological oppositions in which the word of a particular language can participate. The syntactic plane, the unit of which is the sentence, is characterised by oppositions and contrasts among the constituent relationships of the sentence, such as subject — predicate, predicate - attribute, subject — object, etc. The autonomy of the linguistic levels can also be observed in the course of the historical development of a language, evidence of which inter alia seems to be the varying tempo of changes (e.g. on the phonological plane when compared with the morphological plane). The autonomy of the levels cannot be interfered with by the impact of the higher plane on the immediately preceding lower one; e.g. morphological analogy cannot change, decrease or increase the number of phonemes, nor cancel their neutralisation in a given language. The search for a universe governed by laws in the diversity of phenomena is a legitimate objective of any science, but it must be borne in mind that the results of our discoveries cannot be projected into the world of phenomena and create a false dichotomy of a body of phenomena versus its law discovered (or, to be discovered) by investigators. Whereas natural laws are nomothetic and do not presuppose the concept of systematic arrangement, the normothetic laws operating in language imply varying degrees of validity, some of them extending to all or most languages, others applying only to a limited number of them. Every language is a complex organisation of signs of different

10

Introduction

degrees of validity, totally distinct from the mechanical interplay of natural laws in complex physical phenomena (Trnka 1964). Linguistic laws are not only the result of logical reasoning but also emanate from the ideological function, in the light of which language appears as a system of means aiming at a specific goal, namely that of communication, in space and time. This functional goal (means-ends model of language description in a functional theory of language) is decisive for the choice of language means. It conceives language as a dynamic system and humanizes it. The search for abstract models of artificial or natural languages is the concern of mathematical linguistics employing algebraic methods in abstracting entities from the concepts of a language system and its teleology; it relinquishes the tenets of structuralist linguistics, it may, however, exploit some of its findings. Linguistics, the science of languages which serve the goal of intersubjective communication, also fundamentally differs from logic, the laws of which are of universal validity aiming at the determination of whether true or false without constraints by space and time. For its own goals of research structural linguistics is indebted to the findings of numerous sciences whose objectives of research is Man, his organism, his social relations, ecological conditions, and his psychology, as well as to the knowledge of logic and mathematics; however, in acknowledging this debt linguistics must not allow itself to be dominated by the objectives, methods and procedures of these disciplines. Linguistic science is a homogeneous science having its own methodology of research, and its own specific object: the analysis of linguistic signs. Trnka's seemingly highly abstract edifice of a uniform structural and functional linguistic theory has never led him to deductions which were not simultaneously firmly grounded on empirical data collected from natural languages. Trnka has investigated in depth the language material collected from Modern Czech, Modern English, the old Germanic languages, Old English, Middle English and Early New English. In the course of this research work Trnka has contributed specifically to the progress of English language studies, both diachronic and synchronic, of Germanic philology and of Czech language studies. One of the most convincing proofs for the significance of empirical data in Trnka's method of linguistic research is the fact that he was instrumental in the advancement of quantitative linguistics the findings of which he considered highly significant for new qualitative conclusions about the systemic and structural features of concrete languages. That is why a special section in this volume has been reserved for this linguistic topic. As the articles included in this section chiefly interpret Czech language data, we easily detect in them the Czech linguist who always considered it to be one of his specific research tasks to contribute to a better knowledge and also more efficient functioning of his native tongue. In this respect Trnka

Introduction

11

has always been the typical Czech intellectual to whom the cultivation of the national tongue was always an integral part of nationhood and of the competitive maturity of the national culture. After the foundation of the Czechoslovak Republic in 1918, Trnka took an active part in several projects designed to aid the standardisation of Present-day Czech and its development as an efficient, sensitive and diversified tool of cultural and scientific communication. These efforts of the Prague Linguistic Circle and of Czech and Slovak linguistics in general are by now quite renowned. This brings us to a further important trait in Trnka's scholarly work. In spite of his constant pre-occupation with the conceptual perfection of linguistic theory, he did not neglect or underestimate the importance of the possible practical applications of the findings of structural linguistic research. In fact he first recognised the validity of phonemic concepts when he elaborated his own system of Czech shorthand. He has never ceased to be interested in the further exploitation of linguistic knowledge in this domain (Trnka 1937a). He actively participated in the discussions about Czech spelling reform which in the post-war years bore practical results (spelling manuals, school textbooks, etc.). He was also always interested in the practical problems of foreign language teaching and contemplated the importance of structural linguistics for foreign language teaching methodics (Trnka 1937b). 17 Trnka is the author of several language textbooks for Czech learners of English (Trnka 1926, 1927, 1928,1937e), of Danish (Trnka 1937c), of Dutch (Trnka 1939), of Swedish (Trnka 1953a), of Norwegian (Trnka 1958b); he also published a textbook of Czech for foreign students (Trnka 1937d). Finally brief mention should be made of the fact that Trnka is the author of numerous university textbooks on Old and Middle English, on Modern English, on the history of English literature from Beowulf to Shakespeare, and many more. His permanent interest in linguistic theory and thus in general linguistics is manifest from the large number of his writings which were intended as contributions to the problems of language universals, to linguistic typology (especially to morphological typology and to analysis and synthesis in language structure). In these papers Trnka made abundant use of data drawn from many European and non-European languages. However, even in the articles that were not directly intended as contributions to general linguistics Trnka also discussed issues of general linguistic import. To name just a few recurrent topics: analogy, neutralisation of oppositions, homonymy, classification of words, languages in contact (this refers to one of his favourite themes: the stuctural impact of language foreignisms on the receiving language), etc. For the same reason Trnka was also almost permanently pre-occupied with the history of linguistics in pre-structuralist times and in the philosophical implications of language theory. He was widely read in the treatises of nineteenth century philology and in language philosophy, as well as in the history

12

Introduction

of ideas in general. In arguing against the views and concepts of the Junggrammatiker, especially those expressed by H. Paul, Trnka elaborated the theoretical and philosophical principles for which modern structural and functional linguistics appeared to him to be scientifically more rewarding. Trnka also has an intimate knowledge of the work of Otto Jespersen, but he was always critical of Jespersen's theoretical views as they did not seem to him to be sufficiently grounded on a systemic approach. Among nineteenthcentury linguists he was particularly attracted by A. Schleicher because of his rather unconventional linguistic concepts. Most of Trnka's contributions to language typology, language universals and to the development of linguistic thinking have been included in this volume. While loyal to the tenets of the Prague Linguistic Circle and consistent in their application, Trnka never hesitated to be critical of some of the tenets and to seek for re-formulations of established concepts. Holding his own positions, based on a rigorous, highly logical edifice of a functional theory of language, he never admitted that this theory should reflect the system of language as a means in itself. He was rather sceptical about the concept of the immanency of the language system, which disregarded the relevance of the socio-communicative context; in Trnka's writings we hardly find any mention of the necessity to apply the notion of the therapeutic character of phonological and other changes to the diachronic study of language. Trnka was very appreciative of the strictly logical edifice of Hjelmslevian glossematics, but he refuted its deductive method of the algebraic calculus because it has been brought into language analysis from without, being independent of any linguistic reality and divorced from it (Trnka et al. 1958a, Trnka 1967). While agreeing to many tenets of American descriptivism, Trnka remained rather critical of some of them because the methodological approach appeared to him too mechanistic and far too much divorced from any attempt at semantic analysis which was of course inherent in Trnka's basic concept of the communicative function of language. In some private discussions Trnka showed a remarkably reserved attitude towards more recent developments in structural linguistics, in particular towards transformational and generative grammar. This is understandable: Trnka has always adhered to a taxonomic approach in language description and language analysis, moreover he has always upheld the primacy of the functional concept of language in the light of which language never ceases to be a social phenomenon answering to the needs of human communication, hence the inclusion of the suprasyntactic level (the level of utterance) on par within the basic network with the other levels, systems and structures and their underlying laws. 18 His linguistic theory was never focussed on innate language competence. He is, however, aware of the importance of language internalisation. Language cannot be judged only from the standpoint of intellectual values and relations. For a

Introduction

13

language system to be a language it must be internalised. Only to the extent to which it is internalised, and thus also becomes an efficient means of expressiveness and therefore of part of Man himself, does it become a dynamic system capable of change in time and space (Trnka 1943). Trnka conceives language primarily as a semiotic system that is an overt one and which is open to direct observation and analysis. As a structural linguist adhering to the functional principles of the Prague School he has always regarded the concept of the linguistic sign as a reliable basis in modern linguistic research (Trnka 1961). On the foregoing pages I have tried to outline how Trnka's concepts of language and linguistics are representative of the tenets of Prague School structuralism and how consistently, over a period of many years, he further developed the basic theoretical linguistic notions first formulated collectively in the nineteen-thirties by the Cercle Linguistique de Prague. Although a young scholar then, he had already achieved national and international renown. He was elected a regular member of the Royal Bohemian Society of Sciences (1930), he became a member of the Czech Academy of Sciences and Arts (1940), he became a member of the Philological Society in London in 1934 (in 1973 he was elected its honorary member), etc. Trnka has always been open-minded to the views of others and he dislikes any kind of dogmatism, be it political or ideological. When, in 1950, under political pressure, the decision was made to disband the Prague linguistic Circle, Trnka opposed this move as it seemed to him in disagreement with the ideas for which the Prague Circle was internationally respected and by which its members had established a progressive tradition in linguistic science. A few years later new scientific societies were founded under the auspices of the Czechoslovak Academy of Sciences, founded in 1950; among them the Linguistic Association (Jazykovèdné sdruzení) and the Circle of Modem Philologists (Krüh modernich filologû). The former claimed to be the successor of the Prague Circle because several former members of the Prague Circle had joined this organisation, although some of them had publically renounced the structuralist linguistic ideas a few years before. Trnka disapproved of this kind of disloyalty. But he accepted the membership and later on the chairmanship of the Circle of Modern Philologists in 1955 and was for many years its inspiring organiser. He is now honorary president of the Circle. One of Trnka's first moves was the foundation of a Work Group for Functional Linguistics, in whose meetings Professor Trnka still participates. His idea was to organise regular discussions in which the basic tenets of the Prague Linguistic Circle should be elaborated further. It is a relatively small group of younger Czech linguists (Anglicists, Romanists, Slavicists, etc., lecturers as well as research workers in the institutes of the Academy of Sciences and students) who present their papers for discussion, or to whom Professor Trnka expounds the

14

Introduction

results of his latest linguistic reflections. The Group for Functional Linguistics was one of the means by which he wanted to keep the Prague School tradition alive. 19 The most striking traits of Professor Trnka's character, i.e., scholarly tenacity and consistency, strict rationality of arguments and loyalty both in matters of science and in his stand as a citizen, could not have been better phrased than Professor Jakobson did in his afterword written for this volume. In conclusion, a few editorial remarks. The selection of papers to be included in this volume was first proposed by the editor, alternatives were then suggested by the author, and after some discussion the arrangement, as it stands, has found the author's approval. The papers in this volume embrace a period of fifty years. The earliest article appeared in 1928, the most recent one was written in 1978. The volume should thus bear testimony of a significant era in the development of structural linguistics represented by the work of one of its protagonists, and contribute to a better knowledge of the history of modern linguistics. For the same reason the articles in each of the five sections, into which the volume has been divided, follow in historical sequence; with one exception, the last article in section 5 on stylistics, as its topic holds rather an exceptional position among the linguistic themes that form the core of Trnka's research. The division of the fifty papers into five sections is a reflection of the author's diversity of scholarly interests. The titles of the sections speak for themselves and require no further explanation. The papers written originally in English, French and German as well as the Czech articles translated before by various hands were taken over practically unchanged. Only absolutely necessary linguistic improvements were carried out. The English translations of the articles published originally in Czech try to accommodate as much as possible to the terminological uses and habitual ways of stylistic encodings which Trnka employed in his English-written publications. A conscious adaptation to the style of more recent linguistic publications written in English has not been attempted. The reader of this volume will therefore encounter a considerable variation in the use of terminology (e.g., phonology, phonemic, phonological, phonemic; plane, level, dimension; morphology, morphemics, morphological, morphemic; morphological basis or base, superbasis, exponent; and many more). In accordance with the above-mentioned historical principle in the editorial conception of this volume, the included papers were not shortened in order to avoid repetition. The editor is quite aware that there are many overlappings and that rather frequent repetitions occur in a number of articles that have been selected for inclusion in this volume. Where cuts or changes and revisions, as they have been indicated in the footnotes informing about the publication history of each article, were made, they took place on the direct suggestion of Professor Trnka himself or were approved by him. This also applies to the variations in terminological use.

Introduction

15

The bibliographical references throughout the volume have been verified, if necessary corrected and complemented as far as it has been possible. IPA symbols were used in the phonetic transcriptions except in Czech or Slovak words, in which the original spellings with diacritics have been retained. [Duisburg, January 1980]

NOTES 1. Cf. the bibliographical references inter alia in Vachek (1966), Fried (1972), Fried (1978). 2. Vachek (1964). 3. Cf. its reprint in this volume in section 1, p. 70. 4. Cf. the bibliographies of Trnka's writings compiled by J. Nosek in the years 1965a, 1971a, 1975a, 1980a. I would have liked to include a bibliography of Trnka's writings in this volume. But for reasons beyond our control and very much to my regret I could not approach my friend and former colleague Professor Nosek of Prague University for permission t o reprint his bibliographies of Trnka's work in this volume. I should like to mention one omission: 'Vilém Mathesius'. English translation of the obituary, originally published in CMF 29, 1946, by V. Honza in T.A. Sebeok ed., (1966). 2: 4 7 4 - 4 8 9 . J. Nosek was also so far the only linguist to publish a critical appraisal of Professor Trnka's oeuvre on the occasion of his birthday jubilees (Nosek 1965b, 1970, 1971b, 1975b, 1975c, 1980b). A Festschrift including a tabula gratulatoria was presented to Professor Trnka on the occasion of his seventieth birthday with contributions by scholars from many countries: 'Professor Bohumil Trnka Septuagenarian', PP 8 (CMF47), 1965, (2) and (3): 1 1 3 - 3 8 0 . The papers of volume 11 of Prague Studies in English, Caroline University Prague (1965), were also published in his honour on the same occasion. 5. Cf. a discussion of the concept 'Prague School Linguistics' in Fried (1978), Vachek (1979). 6. Cf. V. Mathesius (1947); V. Mathesius (1975); R. Jakobson (1966, 1971); Β. Havránek (1963); J. Mukarovsk^ ( 1 9 6 6 , 1 9 7 1 ) [mention is here only made of the English editions (1977)] ; J. Vachek (1976); J. Krámsk* (1976); V. Skalicka (1979). The list is not complete. 7. My observations are based on personal notes which professor Trnka has kindly mailed to me as a source of information for this introduction. 8. Cf. Mathesius (1931). 9. S. Potter (1965) "In some way this essay was one of the best Trnka ever wrote. It was praised by Otto Jespersen, Wilhelm van der Gaaf, Karl Jost, Fritz Karpf and other scholars of international repute". There are two editions of the Japanese translation of this book. The editor and translator, Professor Shizuka Saito (Fukui University) relates in his preface to the book edition the following story: "I visited Professor Edward Sapir at the University of Chicago on the 28th of April, 1931. I asked him about recent

16

10.

11. 12.

13. 14. 15.

16.

17. 18.

19.

Introduction trends in the development of linguistics and about what he thought to be the most notable recent book in the field of English philology. The very book which he promptly mentioned was 'On the Syntax of the English Verb from Caxton t o Dryden* b y t h e Czech scholar Professor Bohumil Trnka". Professor Sapir seemed t o be very impressed with this book. "As there has been no book so far which describes the process and phenomena of the development of English precisely and concisely as this one", Professor Saito decided to translate it into Japanese. This translation Ά Survey of the History of the Verb Syntax' was published for the first time in English Studies ( 1 9 3 4 - 3 5 ) . [This reference could not be further verified. - V.F.] "In response to the readers' enthusiastic demands" he decided to re-edit his translation in book form in Tokyo in 1956. Cf. Vachek (1966), Vachek (1977), Vachek (1979). Vachek (1979) quotes in support of this view two very early papers by R. Jakobson and B. Trnka, both published in 1929. I find it unfortunate that in the many admirable and deserving contributions published by Professor Vachek about Prague School Linguistics so far a more fully documented acknowledgement of Trnka's contributions is as yet missing, in particular of those pertaining to linguistic theory and methodology as well as those on historical linguistics, etc. which Trnka published after 1950. Cf. Vachek (1966), chp. 1. After the political changes in Czechoslovakia in 1948 Professor Trnka was no longer permitted to travel freely abroad in order to attend international gatherings in linguistics. However, in the 1960's he was able to accept invitations for lectures in London, Erlangen, Berlin (GDR). He was also able t o keep up an extensive international correspondence with scholars in numerous countries and to publish articles in many countries. Cf Trnka (1953b), Trnka (1954), Trnka (1956). These observations draw on the personal notes I received from Professor Trnka in t h e course of 1979. Trnka does not exclude the possible recognition of a larger number of linguistic levels. He occasionally spoke of a 'stylistic level' as the highest level in linguistic description which must account for all facts that play a role in the organisation of the verbal utterance. He admits the possibility of a level of lexicology, but as long as the structural features of linguistic meaning have not been satisfactorily dealt with, a structural analysis of the lexicon does not seem to him feasible. Trnka's concept of at least four basic levels in linguistic analysis (he often refers to the existence of sublevels) seems to me to be merely a methodological restriction because of the present state of functional linguistic research as Trnka perceives it. The concept of morphology (the 'word' level) holds a central position in Trnka's linguistic theory. Professor Trnka informs m e that his later logical arguments, especially in support of the concept of the 'sign planes', which was one of the basic tenets of the Prague Linguistic Circle, was a reflection of B. Russell's notion of 'classes'. Cf. Fried (1965). More recently J. Lyons (1977) re-emphasised the significance of the functional concept of Prague School Linguistics: "One of the most important characteristics of the Prague School which, in the heyday of structuralism distinguished it most strikingly from other schools of structural linguistics, was its emphasis on functionalism . . . their concern with the way language-systems are designed, as it were, to perform their communicative functions", (p.506) (Cf. also p. 249). Trnka's loyal adherence to the heritage of the Prague Linguistic Circle has not met

Introduction

17

with general public approval. In spite of several and repeated efforts which Trnka's friends and junior colleagues made in the years after 1955 proposing him for membership in the new Czechoslovak Academy of Sciences for which he seemed to have every claim as a widely internationally and nationally recognised scholar, the proposal has never been acceded to by the Czechoslovak government. The only official award bestowed upon him was the 'Golden Plaque of Josef Dobrovsky' "for the advancement of the social sciences" presented to him by the president of the Czechoslovak Academy of Sciences on the occasion of his seventy-fifth birthday.

REFERENCES Fried, V. (1965). "The Prague School and Foreign Language Teaching". Prague Studies in English, Università Karlova. 11: 15-32. Fried, V. ed., (1972). The Prague School of Linguistics and Language Teaching. London. Fried, V. (1978). "On the History of the 'Linguistic Circle of Prague' and the 'Prague School of Linguistics' ". Historiographia Linguistica 5: 289—304. Havránek, Β. (1963). Studie o spisovném jazyce. [Studies in the Standard Language] Prague. Jakobson, R. (1929). "Remarques sur l'évolution phonologique de russe comparée à celle des autres langues slaves". TCLP 2. (1971, 1971, 1976). Selected Writings Vol 1, 2,4. The Hague: Mouton. Krámsky, J. (1976). Papers in General Linguistics. Janua Linguarum Minor, 209. The Hague: Mouton. Lyons, J. (1977). Semantics Vol. 1, 2. Cambridge. Mathesius, V. (1931). "Die Persönlichkeit Josef Zubatys". Prager Rundschau 1: 2 3 9 247. Reprinted in Sebeok (1966), 2: 7 8 - 8 6 . (1936). "Deset let Prazského linguistického krouzku". SaS 2: 37-145. English translation "Ten Years of the Prague Linguistic Circle" in Vachek (1966), Appendix 2: 137-151. (1947). CeStina a obecny jazykozpyt. [Czech and General Linguistics] Prague. A shortened reprint was published in Frankfurt/M (1979) in Specimens Philologiae Slavicae, Bd. 22 (eds. O. Horbatch.G. Freidhof). (1975). A Functional Analysis of Present-Day English on a General Linguistic Basis. Prague. - Original Czech version: Obsahovy rozbor souiasné anglictiny na zâkladé obecnè linguistickém. (J. Vachek, ed.) Prague, 1961. Mukafovsky, J. (1966). Studie ζ estetiky. [Studies in Aesthetics J. Prague. (1971). Cestami poetiky a estetiky. [Along the Paths of Poetics and Aesthetics]. Prague. (1977). The Word and Verbal Art. Foreword by R. Wellek. Transi, and ed. by J. Burbank and P. Steiner. New Haven. Structure, Sign and Function. Transi, and ed. by J. Burbank and P. Steiner, New Haven. Nosek, J. (1965 a). "Soupis uverejnènych prací univ. profesora Dr. Bohumila Trnky" [Bibliography of Professor Trnka's Published Work]. Prague Studies in English, Università Karlova. 11: 6 9 - 8 3 . (1971 a). "Bibliography of Professor Bohumil Trnka's Work Published in the Years 1965-1970". Prague Studies in English, Univerzita Karlova. 14: 143-145.

18

Introduction (1975 a). "Bibliography of Professor Bohumil Trnka's Work Published in the Yeats 1 9 7 0 - 1 9 7 5 " PP 18: 1 6 0 - 1 6 1 .

(1965 b). " O vëdeckém dile univ. prof. Dr. Bohumila T r n k y " [ " T h e Scientific Work of Professor Bohumil T r n k a " ] . VIII. roèenka Kruhu modernich filologû pri Ceskoslovenské akademii vèd za rok 1964 [The 8th Annual Report of the Circle of Modern Philologists associated with the Czechoslovak Academy of Sciences for t h e Year 1 9 6 4 ] . Prague, p. 6 - 9 . (1970). "Univ. Prof. Dr. Bohumil Trnka, DrSc., pètasedmdesàtilety". [Professor Bohumil T r n k a ' s seventy-five years o l d ] . CMF 5 2 : 2 0 9 - 1 1 . (1971 b). " T h e Life and Work of Professor Bohumil Trnka (On t h e Occasion of his Seventy-fifth Birthday)". PP 14: 3 9 - 4 1 . (1975 b). "Professor T r n k a : Eightieth Birthday". Prague Studies in English, Univerzita Karlova. 16: 7 - 9 . (1975c). "Professor B o h u m i l Trnka, Octogenarian". PP. 18: 159. (1980a). " A Supplement of the Bibliography of Professor Bohumil Trnka's Writings". P P 2 3 : 169. (1980b). ' T h e Eighty-fifth Birthday Anniversary of Professor Bohumil T r n k a " . PP. 23: 168. Potter, S. (1965). "English Phrasal Verbs". PP 8 : 2 8 6 - 2 8 9 . Sebeok, Th. A. ed., (1966). Portraits of Linguists, Vols. 1, 2. Bloomington, London. Skalicka, V. (1979). Typologische Studien. Hg. P. H a r t m a n n . Braunschweig/Wiesbaden. Trnka, B. (1924). "Prispëvky k syntaktickému a frazeologickému vyvoji slovesa TO H A V E " [Contributions t o the Syntactic and Phraseological Development of t h e Verb t o H A V E ] . Prague Studies in English, Università Karlova. 1: 1 - 3 5 . (1925). Syntaktická Charakteristika feci anglosaskych památek básnickych. [Syntactic Characteristics of Anglosaxon P o e t r y ] . Prague Studies in English, Università Karlova. 1, 2. (1926, 1927, 1928). Ucebnice jazyka anglického pro stredni ikoly / , II, III. [Textbook of English for Secondary Schools]. Joint a u t h o r : S. Potter. Prague. (Several revised editions until 1939). (1929). " S o m e Remarks o n the Phonological Structure of English". Xenia Pragensia (Festschrift für J. Janko). Prague. (1930). On the Syntax of the English Verb from Caxton to Dryden. TCLP 3. Kraus Reprint, Liechtenstein, 1974. Japanese translation, edition and preface by Professor Shizuka Saito ["Survey of the History of Verb S y n t a x " ] . T o k y o , 1956. (1933). "Synchronie a diachronie ν s t r u k t u r n í m j a z y k o z p y t u " . [Synchrony and diachrony in Structural Linguistics.] CMF 2 0 : 6 2 - 6 6 . (1935). A Phonological Analysis of Present-Day Standard English. Prague Studies in English, Università Karlova. 1 , 5 . Revised new edition (edited by T. Kanegiyo and T. Koizumi) T o k y o 1966. (1937 a). Pokus o vèdeckou apraktickou reformu tèsnopisu. [For a Scientific and Practical R e f o r m of S h o r t h a n d ] . Università Karlova, Prague. (1937 b). " V y z n a m f u n c n i h o jazykozpytu pro vyucovánf modernich jazykfl". [The importance of Functional Linguistics for Modern Language Teaching]. SaS 3: 2 3 6 - 2 4 1 . (1937 c). Ucebnice dánitiny se slovnikem [ T e x t b o o k of Danish w i t h a Dictionary.] Joint author M. Lesna. Prague. Revised Edition 1948. (1937 d). The English Visitor to Czechoslovakia. J o i n t author F.P. Marchant. Prague.

Introduction

19

(1937 e). Cech mezi Anglicany. [The Czech among the English] (in English). Prague. 1939). Ucebnice holandstiny se slovnikem. [Textbook of Dutch with a Dictionary). Joint author L.J. Guittart. Prague. Revised edition 1948. (1943). "Obecné otázky strukturálního jazykozpytu". [General Problems of Structural Linguistics.] English translation in this volume in Section 1, p. 32. (1948). "Peut-on poser une définition universellement valable des domaines respectifs de la morphologie et de la syntaxe?" Rapports sur les questions théoretiques et practiques mises à l'ordre du jour du Sixième congrès international des linguistes. Paris, p. 19-24. (1950). A Tentative Bibliography. (With an introduction by Marcel Cohen). Publication of the Committee of Linguistic Statistics, International Permanent Committee of Linguists. Utrecht-Brussels. (1953 a). Ucebnice svédstiny. [A Textbook of Swedish.] (mimeographed). Prague. ('1953 b, 3 1970). Rozbor nynefèl spisovné anglictiny. I. Rozbor fonologick}>. [Analysis of Present-Day Standard English, I. Phonological Analysis] (mimeographed). Prague. ( ' 1 9 5 4 , 2 1 9 7 2 revised). Rozbor nynèjSi spisovné anglictiny, II. morfologie slovnich druhu (césti feci) a tvoreni slow. [Analysis of Present-Day Standard English, II. Morphology of the Parts of Speech and Wordformation.] (mimeographed). Prague. (1956). Rozbor nynèjsi spisovné anglictiny III. Syntaxe jména a jmennych tvarû slovesnych. [Analysis of Present-Day English, 3. Syntax of the Noun and of the Nominal Verb Forms. ] (mimeographed). Prague. Trnka, B. et al. (1958 a). "Prague Structural Linguistics". Reprinted in this volume in Section 1, p. 70. Trnka, B. (1958 b). Ucebnice spisovné norstiny (bokmâl) a úvod do nové norstiny. [Textbook of Standard Norwegian (BokmSl), An Introduction to Modern Norwegian] (mimeographed). Prague. (1961). "Reply to a question". Zeichen und System der Sprache Vol. /., Berlin (GDR), 142-145. (1964). "On the Linguistic Sign and the Multilevel Organisation of Language". Reprinted in this volume in Section 1, p. 86. (1965). "Dëkovny projev predneseny 2. cervna 1965". [A Vote of Thanks Proposed on June 2, 1965]. VIII. rocenka Kruhu modernich filologä pfi Ceskoslovenské akademii véd za rok 1964. [8th Annual Report of the Circle of Modern Philologists associated with the Czechoslovak Academy of Sciences for 1964]. Prague, p. 10-11. (1967). "Hjelmslevova teorie jazykové analyzy". [Hjelmslev's Theory of Language Analysis] ÓMF49: 24-29. Vachek, J. ed., (1964). A Prague School Reader in Linguistics. Bloomington. (1966). The Linguistic School of Prague. Bloomington, London. (1976). Selected Writings in English and General Linguistics. Prague. (1977). "N.S. Trubetzkoy and the Prague Linguistic School". Wiener Slavistisches Jahrbuch. 33: 103-109. (1979). "The Heritage of the Prague School to Modern Linguistic Research (with Special Reference to English Studies)". Zeitschrift f . Anglistik und Amerikanistik 27: 5 2 - 6 1 .

SECTION ONE

General Linguistics

Méthode de Comparaison Analytique et Grammaire Comparée Historique

Dans la conception de la linguistique qui est celle des néogrammairiens, et que l'on peut considérer comme codifiée par H. Paul, 1 on n'admet l'emploi, pour les recherches linguistiques, que de la seule méthode historique, qui étudie les conditions chronologiques des faits de langue. Car on est sur le terrain historique dès qu'on va au-delà de la simple constatation des faits de langue, et que l'on vise à en déterminer l'enchaînement interne, autre nom pour l'enchaînement historique. La linguistique doit tâcher de fixer la succession des faits de langue non seulement dans l'examen des textes appartenant aux divers stades d'évolution d'une même langue, mais aussi dans la comparaison des langues et dialectes apparentés: il s'agait non pas tant de relever des ressemblances que de reconstruire le fonds commun initial d'où les dits dialectes et langues seraient sortis par l'effect de "lois". Cette conception, qui triompha entre 1870 et 1880 et fut d'abord appliquée dans la grammaire comparée des langues indo-européennes, apporta à la linguistique des résultats inattendus, et elle s'implanta aussi dans l'étude scientifique d'autres familles de langues telles que les familles sémitique, ougro-finnoise et bantoue. Seule était considérée comme étude scientifique, même dans le domaine des langues non indo-européennes, une comparaison génétique limitée principalement à la phonétique (et à la morphologie comme phonétique appliquée), alors que la description exacte des langues et dialectes contemporains sans appareil historique était tenue pour une simple besogne préparatoire ou pour un travail imparfaitement historique. Les néogrammairiens condamnaient la comparaison analytique de langues non apparentées entre elles. Actuellement, après la publication du "Cours de Linguistique Générale" de Saussure, 2 il est facile d'apercevoir les faiblesses du raisonnement des néogrammairiens. En face de la méthode historique, ou, selon la terminologie de Saussure, diachronique, on a la méthode de comparaison analytique ou synchronique. En regard de la surestimation faite par les néogrammairiens

Originally published in TCLP 1, 1929, pp. 3 3 - 3 8 . Reprinted in PSRL, pp. 6 8 - 7 4 .

24

General Linguistics

de la méthode historico-génétique, on pourrait, aussi légitimement et aussi partialement, dresser l'affirmation suivante, à savoir qu'expliquer un fait de langue à la lumière de la tradition comme un équivalent successif d'une forme plus ancienne n'en constitue pas encore la pleine explication, et que toute étude historique, dès qu'elle ne se borne pas à simplement constater la régularité des changements linguistiques et qu'elle passe à leur explication profonde, se trouve déjà sur le terrain de la grammaire synchronique, car les changements d'éléments linguistiques donnés dans une phase d'évolution donnée doivent être expliqués par l'action d'autres faits synchroniques. C'est la recherche diachronique, et non l'étude synchronique, qui constitue une simple étude préparatoire, un classement chronologique des faits servant d'introduction à l'examen profond, et synchronique, des systèmes d'expression qu'étaient les phases d'évolution antérieures de la langue. La grammaire analytique admet en outre la possibilité de la comparaison d'états de langue entre langues apparentées de loin seulement ou pas du tout apparentées, et permet ainsi de constater des tendances linguistiques et des catégories grammaticales plus générales, constatations que l'on n'aurait pu faire par la grammaire historique des langues d'un même groupe. Bien que la linguistique actuelle penche dans l'ensemble vers la méthode de comparaison analytique notamment dans les recherches de syntaxe et de sémasiologje, il y a lieu d'insister sur ce point que les deux méthodes se complètent l'une l'autre pour l'étude des phases de développement des langues apparentées. Si, pour le passé, la méthode diachronique passe avant la méthode synchronique, elle passe après pour l'étude de la langue contemporaine. La méthode diachronique bâtit les formes d'expression surtout sur les faits successifs, et s'occupe des changements différenciant l'expression linguistique à des époques diverses. La méthode synchronique, qui n'a pas besoin de se limiter à des langues apparentées, mais peut comparer entre eux n'importe quels systèmes d'expression, peut aller non seulement de la forme à la signification, mais aussi de la signification (fonction) à la forme, en s'occupant surtout d'analyser finement les aspects fonctionnels de la langue, et en fournissant ainsi de nouveaux matériaux à la méthode diachronique. L'une et l'autre méthodes sont comparatives, recherchent les différences et les ressemblances existant entre des faits de langue; seul les résultats diffèrent, même lorsqu'on opère sur les mêmes matériaux. Différentes, tout en étant complémentaires, sont aussi les lois auxquelles on arrive de l'une et l'autre façons. Les deux méthodes doivent être distinguées avec précision, et leurs résultats ne sauraient être confondus. Par exemple, la comparaison analytique de langues slaves ou germaniques ne saurait aboutir à une reconstruction de formes antérieures initiales, de même que la grammaire historique de ces langues ne saurait aboutir à dresser le tableau synchronique des formes

Méthode de Comparaison Analytique et Grammaire Comparée Historique 25 primitives et reconstruire le système d'expression primitif proto-germanique ou protoslave.3 Quelle est la différence principale entre les deux méthodes? Ce n'est pas, comme le croit de Saussure, le fondateur de l'école de Genève, le temps, éliminé dans l'étude synchronique à l'encontre de l'étude diachronique; le point décisif est le but de l'étude. On emploie la méthode synchronique quand on compare des systèmes linguistiques, que ceux-ci représentent des stades successifs d'une même langue ou des stades de langues apparentées ou non. Ce faisant, on se comporte comme quand on compare deux ou plusieurs tableaux: on note les couleurs, les dessins, les rapports des parties au tout, on relève les ressemblances et les différences, et l'on tend à voir dans les détails qui se reproduisent la manifestation de tendances déterminées. Quand on emploie la méthode historique, la comparaison poursuit un but tout autre: il ne s'agit pas de comparer des systèmes, ou des particularités à l'intérieur de ces systèmes, mais de reconstruire l'image primitive dont procèdent les copies examinées, ou de mettre en parallèle les traits fonctionnellement équivalents de l'original et des copies. La linguistique qui emploie la méthode comparative analytique vise à déterminer les relations réciproques des différents éléments d'un système d'expression donné, c'est-à-dire à constater des relations mutuelles cycliques. Au contraire, la grammaire historique vise à déterminer l'ordre de succession des faits, qu'elle suit dans leur développement linéaire, et elle s'échappe ainsi perpétuellement des cercles en lesquels les faits sont groupés à une époque donnée de l'évolution de la langue. Le temps ne saurait être un caractère distinctif essentiel des deux méthodes puisque la conscience des changements chronologiques (cf. archaïsmes et néologismes) est un facteur psychologique à tout moment de l'évolution de chaque langue. En comparant, par exemple, le latin avec une langue romane, on aurait à peine le droit de parler de comparaison synchronique. Il vaut mieux parler de comparaison analytique en regard de la comparaison historico-génétique. La grammaire historique actuelle ne fait qu'établir, grâce aux documents écrits, la simple succession des faits de langue à partir des stades les plus anciens de la langue jusqu'au stade le plus récent. La substitution à d'anciens faits linguistiques de faits différents ayant les mêmes fonctions ne peut se comprendre au point de vue psychologique que si l'on admet comme loi générale le fait qu'il ne saurait exister dans la subconscience linguistique des sujets parlants d'homonymes absolus, autrement dit deux formes diverses pour la même fonction. Deux formes en vieux tchèque comme reka et reka, qui expriment la même chose [la rivière j ne peuvent se maintenir l'une à côté de l'autre, et l'une des deux disparaît. Lorsqu'une idée ou un rapport avec un même contenu intellectuel et affectif est rendu, pour des raisons de phonétique ou d'emprunt dialectal par deux variantes, on voit apparaître une

26

General Linguistics

tendance à employer chacune des variantes avec une nuance de sens différente ou à en supprimer une. Lorsqui'il en est ainsi, la variante la plus ancienne, ayant cessée de constituer un élément du système de la langue de concert avec les autres éléments, devient un fait de la grammaire historique. Ces faits successifs, qu'ils soient de caractère phonétique, morphologique ou syntaxique, qui expriment, au moins en apparence, la même fonction, forment l'objet de la grammaire diachronique, qui n'est au fond rien d'autre qu'une grammaire de différenciation, puisqu'elle examine les changements linguistiques sur la base du stade d'évolution le plus ancien qui soit accessible. Bien que les lois phonétiques actuelles ne constituent que des indications dans la complexité des phénomènes que comporte le mot comme terme de phrase et placé au premier plan de l'intérêt du linguiste, elles sont en somme, dans leur secteur, indépendantes des autres faits linguistiques, car les sons constituent des relations directes de la forme à la signification, en d'autres termes ils ont leurs fonctions spécifiques propres, ce sont des phonèmes chargés d'une fonction différente de celle du mot comme unité de la phrase. 4 Il est curieux de constater que la linguistique ait aperçu si tardivement le rapport des éléments phonétiques à leur fonction à l'intérieur d'un système linguistique, alors que les exceptions aux lois phonétiques, dites changements analogiques, étaient expliquées visiblement par l'action de facteurs significatifs de la phrase sur les sons qui le constituent. La tâche de la nouvelle linguistique est donc de suivre les changements phonétiques non seulement au point de vue de la phonétique pure, mais aussi dans leurs rapports avec les fonctions qui forment une unité d'expression. Si la grammaire historique suivait les changements phonétiques dans leur mouvement linéaire sans égard au système, la nouvelle grammaire historique fonctionnelle doit suivre le déplacement des habitudes articulatoires en tenant compte des fonctions des autres sons groupés en cercle, et du mouvement linéaire de certains éléments qui réagit perpétuellement et se déplace en conséquence perpétuellement. Les linguistes russes, tels que MM. Troubetzkoy et Jakobson, soulignent le postulat scientifique de la linguistique fonctionnelle, à savoir que tous les faits phonétiques doivent être expliqués comme la manifestation de certaines tendances fondamentales, qu'il y a lieu de constater pour le système d'une langue ou d'un groupe de langues. Sans nier la légitimité de ce point de vue, qui marque un progrès important sur la conception de simples lois des changements phonétiques isolés dans l'esprit des néo-grammairiens et un pas fait vers la conception de lois plus larges, je suis loin de penser que l'on puisse rendre compte de tous les changements phonétiques, unilatéralement, par la réaction de tout le système phonologique. Outre les changements phonétiques qui sont produits par cette réaction de tout ou partie du système phonétique fonctionnel sur une modification déterminée, il y a lieu de reconnaître, pour

Méthode de Comparaison Analytique et Grammaire Comparée Historique

27

une période donnée d'évolution, des changements phonétiques, où les facteurs fonctionnels ne jouent qu'un rôle de second plan, ou même qu'un rôle tout passif. L'appréciation de la valeur relative des deux méthodes complémentaires, l'analytique et l'historique, tient à l'objectif finad poursuivi par l'étude linguistique. S'il s'agit d'établir le caractère propre d'une langue donnée, de rendre compte de la langue dans sa particularité concrète comme système de moyens d'expression et de communication propre à une période donnée de son développement dont il s'agit de rendre compte, la méthode essentielle doit être nécessairement celle qui est orientée statiquement et qui conçoit l'histoire de la langue comme un total de périodes particulières d'évolution à comparer en qualité de périodes indépendantes. Mais si le linguiste est orienté historiquement, s'il se propose de rendre compte de la succession des faits de langue par où les différentes périodes d'évolution se distinguent les unes des autres plutôt que de comparer des systèmes dans leur intégralité concrète, il relie les faits linguistiques se succédant sans interruption en un système reposant sur "l'axe du temps", système dans lequel les faits sont classés suivant des lois abstraites et qui est aussi abstrait que les schémas des biologistes représentant par exemple le développement de la circulation du sang chez les divers animaux sans tenir aucun compte des fonctions des autres organes de l'animal. Suivant la valeur relative accordée à l'une ou l'autre méthode, on aurait une esquisse toute différente de l'histoire par exemple des langues germaniques du point de vue analytique et du point de vue historico-génétique; l'appréciation de la valeur des résultats des deux modes de comparaison dépend des dispositions psychologiques de tel ou tel peuple. Mais, dans une étude partielle de la langue, il faut employer les deux méthodes complémentaires: plus on a de coupes transversales et longitudinales, plus on se rapproche de la compréhension d'une langue donnée et de ses phases évolutives. Le réseau le plus serré de coupes transversales est fourni par l'étude des langues des générations successives, le réseau le plus serré de coupes longitudinales est fourni par celle du plus grand nombre possible de faits linguistiques depuis les stades les plus reculés jusqu'au stade contemporain. Attachée à l'analyse des langues contemporaines, l'école de Genève renonce, du fait de son attitude d'indifférence à l'égard de l'analyse des phénomènes synchroniques des phases antérieures, aux possibilités linguistiques de la comparaison analytique de plusieurs systèmes linguistiques possédant un ensemble de faits déterminés communs et parlés par les générations successives, sans lequel la compréhension de la langue contemporaine elle-même est sommaire et non définitive.

28

General

Linguistics

NOTES 1. Paul, H. ( 5 1920). Prinzipien der Sprachgeschichte. Halle. 2. Saussure, F. de ( 2 1923). Cours de Linguistique Générale. Genève. 3. C'est la langue contemporaine qui est le plus accessible à l'étude linguistique, puisqu'on peut la connaître directement, c.-à-d. par l'oreille. Les états de langue passés ne nous sont accessibles que par l'oeil, par l'intermédiaire de l'écriture, qui conserve par la force de la tradition une orthographe ancienne voilant les changements phonétiques ou morphologiques, alors que la langue vivante pouvait avoir déjà les nouvelles formes. C'est presque une loi que le fait suivant, qui se produit dans l'étude de l'évolution d'une langue: lorsque les documents écrits font défaut pour une certaine époque, on en est réduit aux résultats que l'on peut obtenir par la comparaison des phases chronologiquement voisines. Lorsqu'il s'est conservé de nombreux monuments écrits, témoins d'une forte tradition littéraire, c'est à son tour celle-ci - qui ne peut être interrompue que par une révolution - qui s'interpose entre les faits et nous, sous forme soit d'une orthographe fixée (comme en français et en anglais), soit d'une morphologie et d'un vocabulaire fixés (comme en sanscrit). La connaissance d'un stade de développement d'une langue donnée dans toute la plénitude des faits est d'autant plus fragile qu'on a moins de documents et qu'il faut plus de constructions hypothétiques. Je pense aux tessons de poteries antiques: plus ceux-ci sont volumineux ou nombreux, plus la reconstruction du vase entier est sûre. De même, dans l'analyse de la langue d'un monument isolé, il faut compléter les traits conservés par des traits reconstruits, à distinguer scrupuleusement des faits atteints directement par l'analyse. C'est seulement par l'union complémentaire des résultats des études de comparaison analytique avec les résultats reconstruits par les recherches historico-génétiques que l'on peut approcher la réalité linguistique de la langue primitive reconstruite. 4. Je suis arrivé â la notion de phonologie dans mes réflexions relatives aux problèmes de linguistique générale avant d'avoir eu entendu parler des recherches phonologiques des savants russes, et ce par une double voie: d'une part, j'y ai été conduit par une tentative de constitution d'un système de sténographie tchèque plus parfait, système devant reposer sur un examen précis de la fonction des sons du tchèque (itérations et combinaisons), d'autre part j'y ai été amené en cherchant â l'étude phonétique un pendant dans l'étude du contenu de ces sons.

About Analogy in Structural Linguistics

Analogy is often discussed by the Neogrammarians as a psychological factor responsible for the creation of new morphological signs following a certain pattern or representing a deviation from a certain form which is phonologically expected; the latter then constitute formations which strive for the general use against the synonymic "regularly developed" forms in the speech community. Both of the rivalling forms may co-exist in the historical development of languages for some time, but very often the new analogical formations become prevalent and wholly obliterate the latter, if they do not happen to be protected by spelling or by associations of sociolinguistic character. When dealing with the diachronic substitution of one phoneme by another by way of analogy we must be aware of the fact that such a change is not perpetrated by an element of the phonological level of language, but by a synonymous morpheme realised by one phoneme or a combination of two or more phonemes; hence analogy is a morphological factor which operates on the mo rphological plane that functions between the phonological and syntactic plane. Whereas the diachronic change of, e.g., Old Czech a> è and later > e (cf. dusa ->· duSè, later on -»• duse [soul] ) was purely phonemic, as it took place only if preceded by a "soft" consonant, the analogical change of /' > u in Modern Czech (cf. pisi -*• písu [I write, am writing], vázi vázu [I b i n d , a m b i n d i n g ] , piji ->• piju [I drink,

am drinking]

) c o n s i s t s in t h e

substitution of the older morpheme -i by the synonymous morpheme -u. Morphological (or, better, morphemic) analogy must be regarded therefore as a factor whose function is to simplify the phonemic realisation of morphemic oppositions. It cannot interfere either with morphological or phonological structural rules of any linguistic system and even if it may result in the substitution of several phonemes in a morpheme (cf. mohu mû'zu

Originally published under the Czech title "O analogii ν strukturálni'm jazykozpytu" in SaS 2, 1936, pp. 2 2 1 - 2 2 2 . Revised. (Translated by V. F.)

30

General Linguistics

[I can], mohou ->• mäzou [they can], on analogy of mâzés [thou canst], mäze [/he/ can], etc.) no structural change is effected by this process. Similarly the origin of the accusative sing, cloveka [Man], oràce [the ploughman], etc. is an analogy to the syntactically transferred genitive sing, on the syntactic plane. From the static point of view, the ending -e is fixed by the phonological as well as by the morphological structure and also by the syntactic structure, and the diachronic changes which brought it about are the result of a functioning towards a certain goal (end) which may be achieved on every language plane in accordance with the principle of the least resistance. From the point of view of structural linguistics analogy is only a means -end towards the realisation of morphemic relationships, or towards the diachronic mutations of these relations (re-morphologisation, etc.), and the task of structural morphology to which R. Jakobson has significantly contributed in his pioneering paper about Russian cases1 is to discover these relations. The following remark may be of some interest to the students of historical linguistics. Analogy as a morphological phenomenon operates only with phonemes, not with allophones or in other words, allophones cannot be transformed into phonemes by the process of morphological analogy. While analogy causes a change of a -*• e in the forms predu, predou [I, they spin] instead of pradu, pradou (according to predes [thou spinest], the allophone [«] in Czech and many other languages can never be found in the position which would qualify /η/ as a phoneme opposable to /«/. Cf. okénko (little window] (diminutive, nom. sing.), okénka (gen. sing, and nom. acc. pl.), okénku (dat. sing.), etc. versus okének (gen. pl.). Similarly the phoneme t / i / / in the adjective léiebny [sanitary, medical] cannot be replaced by the allophone [d^ ] which occurs in the basic noun lêcba [le.dgba] [cure, medical treatment], whereas the phoneme /si in prosebny [precatory] is often pronounced as the phoneme /z/ on the analogy of prosba [prozba] [request]. On the other hand, the occurrence of analogical forms such as papes (gen. sing.)pape (dat. sing.) in 10th and 11th century OE manuscripts due to morphological analogy shows us that both œ (pœp nom. sing.) and a functioned as phonemes in this period. The impossibility of analogical phonological mutation also applies to positions of neutralisation. In Czech, e.g. the correlation of voice (as t/d, s/z, s/ζ (J13), etc.) is neutralised at the end of words or in front of a paired phoneme as in let [flight] vs. led [ice], vez [drive!] vs. ves [village], vèzt [to drive] vs. vest [to lead], kresba [krezba] [drawing]. This position cannot be held by another allophone (e.g., d instead of t in the word led ¡let! according to the nom. pl. and the indirect cases), i.e. by means of the impact of the morphological plane on the phonological one. In positions where no neutralisation occurs one member of the correlation may suppress

About Anabgy

in Structural

Linguistics

the other on analogy of the form in which neutralisation is realised by the voiced member, as krezeb [of the drawings] instead of kreseb (according to [krezba] ; written kresba)', cf. also [vet'me], according to [vet'}, [vet'te], written ved'me, ved', ved'te [let's lead, lead!, lead! pl.], These observations prove the fundamental importance of the distinction of phonemes and allophones without which the description of a linguistic system would be incomplete, or, at least, difficult to understand; they represent a further test which helps us to distinguish the two phenomena even in past stages of language systems.

NOTES 1. Jakobson, R. (1936). "Beitrag zur allgemeinen Kasuslehre. Gesamtbedeutung der russischen Kasus". TCLP 6: 2 4 0 - 2 8 8 . Reprinted in 2: 2 3 - 7 1 .

31

General Problems of Structural Linguistics

Nineteenth-century linguistics considered genetic comparison the only fully valid method. Putting aside problems of the cultural function of language, it considered language an organism which grew and which could be understood only in terms of its natural development; or it discussed language as the intellectual product of the group from the standpoint of individual and group psychology. Together with its historicism, linguistics embraced the individualist, atomistic trend as over the static view in vogue in the 18th century, which emphasised the mind and the general validity of the facts gathered by the mind, taking interest primarily in the agreements rather than the differences among individuals. The 19th century linguist thus saw linguistic history as a mass of isolated, gradual changes which continually transformed language; but he never analysed language structurally as a system of signs, thus embracing language as a functional whole. He tried, to be sure, to explain changes in terms of various causes, but frequently he was satisfied with post hoc, ergo propter hoc, and the precedence of linguistic phenomena was confused with their explanation. 1 It is interesting to note that atomism in historical linguistics was accepted even by Ferdinand de Saussure who, in the first decade of this century — following the trends outlined by Brentano, Husserl, Russell, Höffding, and others, which provided a new view of language — was the first linguist to pose the question of the object of linguistic science. Answering his own question in his well-known posthumous Cours de linguistique générale (1916), he became one of the founders of structural linguistics. The Swiss linguist drew a sharp line between the synchronic and the diachronic study of language: only synchronic phenomena are contained within a system, while diachronic changes destroy the system. But the line which de Saussure drew does not

Originally published in Czech under the title: "Obecné otâzky strukturàlnîho jazykozpytu". SaS 9, 1943, pp. 5 7 - 6 8 . On the suggestion of P.L. Garvin (Georgetown University) translated into English by Philip H. Smith, mimeographed in USA 1962. No publisher and place stated. Reprinted from the English translation.

General Problems of Structural Linguistics

33

actually exist. Never, at any point in its development, does language cease to be a system of functions; for even when changes occur all of its elements remain well-ordered, change almost simultaneously, and each exists in relation to the others. Shifts in the system occur constantly, although to varying extents and at varying rates. In supporting the position that even in language the form of being is not the opposite of the form of action, the Prague school of linguistics has, since its inception, directed the attention of structural linguists to the fact that language at a given period represents not only a system whose components are maintained in a relatively unchanging form by the fixed mutual relations of their elements, but that it is at the same time a system in which complementarity (synonymity) and the capacity of some components to replace others (homonymy) allow for change and thus enable it to function adequately through time. Thus language always contains latent possibilities for mutation, i.e., the substitution of new norms for old, and thus the capacity for constant innovations.2 The historical linguist who wishes to understand the movement of a language as a functional whole must necessarily use structural methods. The historical continuity of a language is accessible to us only by comparing the stages through which it is maintained. The larger the number of such stages, i.e., the shorter the historical continua between individual cross sections, the more reliable are the results of the examination of the past. The fewer the cross sections of which we have knowledge, and the longer the time intervals between them, the further we are from being able to establish the actual history of the continuum. An absolute continuum, whose limits would be zero time intervals between linguistic stages, can never be attained, since it assumes an infinity of stages. The difference which we observe between older and more recent language systems are linguistic changes. Historical grammar is thus based on the differential grammar of linguistic systems stratified in time just as the comparison of dialects (or languages) ordered in space is a differential grammar of its kind. Dialect comparison and historical comparison share a common theoretical problem: the difficulty of delimiting from each other the systems which form the basis of comparison. The extent and quality of the structural features which set off dialect systems in space are not directly given any more than they are for isolating individual periods in time. If we attempt an analysis and project all the facts of a given stage on an apparently synchronic plane, we find that we are standing not on a flat surface but on the surface of a sphere, the horizon of which retreats the further our gaze wanders from the point in time which we took as the basis of our historical observation. There are no lines which can be drawn across a progressive continuum dividing it into individual, comparable cross sections, and the one which we have chosen never represents a perfect synchrony of the language under investigation, which includes, for example, the synchronic

34

General Linguistics

differences in the speech of generations, and the "functional dialects" of social classes. Hence the difficulties of structural periodisation, which parallel the difficulties of the structural grouping of dialects. If we ascertain the outstanding structural features of a language system of the past we will select largely those (like the Czech sound shifts of the twelfth through fourteenth centuries) which attract attention because of their contrast with the language at previous and subsequent periods; but their apparent dominance for that period will vanish, as it were, before our eyes, since it will be seen that in the continuous chain of mutations these changes were not the dominant ones in the system of Czech of the time. We must of course specify what features we consider structurally dominant. They are those which are the condition for the existence of a large number of other structural elements of the system, according to the logical formula "If A, then B, C, D", etc. The establishment of a hierarchy of dominance, which is one of the objectives of structural linguistics, is based on a quantitative comparison: the more functional oppositions depend on a primary one, the more dominant this primary opposition becomes. A distinction must be made between this dominance and the characteristic nature of the phenomena, which can be observed without reference to structural analysis. The present periodisation of the history of language is based either on such characteristic phenomena or on points of view which are external to the shifting time continuum, whether the point of view be negative (the lack of written monuments) or positive (the cultural, religious, or political trends with which the language is continually involved in a necessary symbiosis and which have their own structure). These external structures intersect the linguistic continuum and, precisely because of their heterogeneity, provide practically useful points of reference and dividing points, although from the standpoint of the linguistic system they are completely external. Internal, structurally determined periodisation still remains a problem: the language of any period is a system of norms of varying time depth, and to consider the duration of a relatively dominant norm as identical with the length of a given historical period (which would be thus defined by it) always means taking the norm out of context — a norm which is definable only in terms of the dynamics of the entire system. It would appear that the structural point of view forces us not to divide the time continuum of language into segments but to treat it as a single process. In speaking of the movement of language in time and space it may be added that we can characterise the standard language as opposed to the dialects in terms of its extension in space. As a result of its special functions the standard language develops a capacity for much greater expansion in time and space than the dialects, although this is not accomplished without some loss in another direction. In its spatial expansion it attempts to embrace an ever larger territory; it seeks to turn linguistic systems with which it was

General Problems of Structural Linguistics

35

originally on a parallel functional level into dialects; and it does this until it comes into conflict with another standard language which makes the same demands on the so-called transitional dialects. In addition to its spatial expansion a standard language also expands in time: this is seen in its striving for greater monumentality and time-stability. Its most important tool is writing. In this technological discovery the standard language gained a particularly valuable ally, since the writing system transforms the standard language into a special type of institution, more resistant to the erosion of time and space. Language, which was originally an instrument of understanding and expression, becomes more and more the vehicle of political, religious, social, and other cultural aspirations, i.e., of essentially non-linguistic tendencies. It also affects dialectal systems (against which it makes its expansionist demands), being a common source of new material for the dialects, although its profferings need not be and often are not accepted by all dialects. I will not dwell on these problems, which were discussed by B. Havránek,3 and, from the standpoint of orthographic norms, by J. Vachek;4 I, too, have touched on them, in discussing the price which the standard language must pay for its expansion, in the Mathesius' Festschrift. 5 In this connection I should like to add that another ally of the standard language in its temporal expansion may be acoustic literature — in which I include all technical media for capturing the spoken word, such as phonograph and tape recordings and sound films — since they provide coming generations with a direct acoustic record of the language, thus reinforcing the efforts of the standard language toward time-stability. This will become necessary, furthermore, since the linguistic system will be under increasing external pressure as a result of the development of air transport, which promotes contact with members of other linguistic groups. Let us return to the subject. Linguistic history is, as we have shown, based on the structural comparison of cross sections forming systems. It follows from this that the changes, i.e., the differences between cross sections, must be evaluated from a structural standpoint. 6 In phonology, therefore, we speak of the phonologisation of phonemic oppositions in a given position in the word, the loss of neutralisation (deneutralisation), rephonologisation (e.g., the transformation of a correlative opposition into an opposition of a different type and conversely), the regrouping of phonemes, etc. Both aspects of language - "langue" and "parole" - are represented in the movement of the system; one cannot say that a change in one aspect is the cause of a change in the other, or that deviations in the "parole" — (i.e., the manifestation of speech by physical persons) is the forerunner of a shift in norms, as seems to be the position taken by N. van Wijk7 and by A.W. de Groot. 8 A linguistic system, even when it is changing, appears only in concrete expressions, but both aspects are sides of the same coin. They are, to be sure,

36

General Linguistics

interrelated, but they are not in contrast with one another, nor do they stand in a cause-and-effect relationship. The historical linguist who wishes to lay claim to scientific accuracy must look elsewhere for his laws, i.e., for the constant relations between the older and newer stages of a language. Some linguists, as e.g. V. Brandal,9 recognise this constancy only in phonology, but not in the area of meaning. The assumption of such a dualism is clearly incorrect, and the linguist must lay claim to scientific accuracy throughout the field, he must attempt to achieve regular correspondences even on the level of meaning, as he has achieved the methodologically important concept of the "sound-laws without exception". 10 All the more must we recognise regular correspondences in functional phonology, in phonemics;if we find a phonemic mutation (i.e., a new sound law) we must expect, assuming of course that we have interpreted it correctly from its concrete manifestations, that all cases will follow this law without exception. We can in Czech, for example, discover the law of neutralisation of the opposition of voice, which occurred a certain time after the loss of fers in word-final before pause, and which remains to this very day. This functional law must apply without exception in all cases in which a consonant participating in the opposition of voice occurs in this position. This neutralisation is undoubtedly tied up with the entire phonemic system of Czech, and must have produced certain reactions within this system. For every functional loss, such as neutralisation, dephonologisation, etc., there must be a functional compensation and conversely, be the compensation quantitative, qualitative, or both. In linguistic comparison the old and new linguistic systems must be treated as two sides of a complex equation: if something is added to one side a corresponding loss must occur to the other side for the equation to remain in balance. The truly structural history of a given language (e.g., of its phonological system) must present the entire complex range of hierarchically ordered laws, each of which has a definite but different duration of function. 11 The regular correspondences which we demand in comparing the older and newer stages of a language must be assumed even in comparing linguistic systems in space, without regard for their time relations. The Neogrammarian school considered this procedure unscientific (particularly in the field of phonological comparison), but functional linguistics, and particularly functional phonology, have demonstrated the correctness and fruitfulness of this functional comparison, which will bring the more complete results, the finer the network of systems under comparison. The question of whether we emphasise functional correspondences against a background of differences, or functional differences against a background of agreement among the individual languages, will be decided by the purpose of the investigation. If we emphasise the surprising agreements against the background of many differences (e.g., the agreements among various Balkan languages), we arrive

General Problems of Structural Linguistics

37

at the concept of a linguistic area. This concept, by the way, is also based on certain cultural, geographic, and political features, since there can be no doubt that structural agreements will be found between other languages which may be geographically far apart. This would certainly be shown by an atlas of phonemic and other structural isoglosses, the compilation of which is an important objective of linguistics. Equally great and deserving a project would be the compilation of an atlas of historical isoglosses, showing in advance and time-depth of individual functional phenomena in linguistic development, just as geographical isoglosses indicate spatial diffusion. The concept of temporal isoglosses which we have set up here is one of the consequences of the structural view of language. The older linguists were unable to arrive at this concept since they believed either in the theory of slow, imperceptible phonetic changes, or were of the opinion that a change once carried out (e.g., the shift of the stops in Germanic), continues. Let us now consider in more detail the relationship between phonetic and phonemic changes. Phonemic mutations are very frequently announced by definite changes in articulation, but this is not always the case. If, for example, at a certain linguistic stage an allophone becomes phonemic this new phoneme need not be manifested differently from what was the allphone. The Middle-English allophones [ν], [ζ], [o] were probably pronounced the same before and after they became separated phonemes, a result of the loss of final -e (before 1400). It would be equally easy to imagine that the pronunciation of the Czech allophone [g] (as in the word kdo [gdo] [who] ) would remain the same even if this allophone of the phoneme ¡k¡ became a separate phoneme. The only indication of its phonologisation probably would be an increased effort to eliminate variations of the type [lokâl] : [logâl], [place of entertainment], since \g\ would find itself in opposition to [A:]. This might lead to a meaning difference: logâl would represent a popular place, and lokàl a place for more elegant guests. At the present time the articulation difference between [fc] and [g] is not capable of carrying this distinction. If we see that phonemic mutations need not be accompanied by phonetic changes we must ask whether phonetic change must not always be supported by phonemic changes. We can answer this question affirmatively. We thus include allophones and those articulations of phonemes which I have called modifications (the voiced [z] in Danish, where there is no distinctive voiced-voiceless opposition, as a modification of the phoneme /s/, as over the Czech [g], which is an allophone in a language in which voicing is a distinctive feature) in the study of functional phonology, since we arrive at it through functional evaluation, considering them as elements which are not capable of distinguishing words intellectually. Even allophones (and modifications) have a certain function: they signal the following sound, or rather fuse the word into a whole. If we therefore recognise allophones and modifications

38

General Linguistics

as products of functional evaluation we arrive at the thesis that all changes in articulation — provided they are not purely physiological — must correspond to certain phonemic shifts, while, on the other hand, phonemic mutations need not be accompanied by changes in articulation. But other matters must be pointed out as well. Some phonemicists, such as A.W. de Groot, completely eliminate from historical phonology those changes which do not concern the phonemic inventory, or the appearance or disappearance of phonemic oppositions. For him phonological mutation does not include such phenomena as rhotacism in Latin ( Venesis> Veneris), since the consonants js¡ and ¡r¡ were phonemes in Latin even before this change; nor, in his opinion does it include the substitution of quality for quantity in later Vulgar Latin. On the other hand he considers as sound-laws all regular changes in "word forms", and believes that these changes always precede phonemic changes. Narrowing the scope of phonemics to a mere repertory of phonemic oppositions, as de Groot does, and the refusal to accept the study of combinations of phonemes as a part of phonemics, are surely unjustified here. It is clear that the ordering of phonemes is also governed by certain laws, which do not belong to "parole", and that we must therefore speak not only of the repertory of phonemes and of phonemic oppositions but of the repertory of phonemic combinations. The change from Venesis to Veneris must be viewed as the neutralisation of the phonemic opposition ¡s¡ :/r¡ in a certain intervocalic position, since it eliminates one of the members of the s:r opposition in that position. Similarly the Old Czech umlaut a>e is a phonological mutation, although it does not expand the inventory of Old Czech phonemes; it represents the neutralisation of the /a/:/ë/ opposition which occurs in certain positions (word-finally after soft consonants, and word-medially after soft consonants provided a soft vowel follows in the next syllable) and which is represented by the vowel [é]. And these mutations, which can be called mutations in the phonemic sequence, as opposed to mutations of the phonemic inventory, also occur by bounds. 1 2 A linguistic system is composed of several large sub-systems which intersect and cooperate smoothly, without overlapping or friction. These systems are the phonological, morphological, and syntactic. Although they are autonomous, neither makes any sense without the rest. It may be said that phonemes (or combinations of phonemes) are potential words and words potential sentences, or rather that morphological oppositions (such as the opposition singular:plural) are manifested by sequences (combinations) of phonemes, while syntactic oppositions are accomplished by sequences of words which in turn are on the morphological level. Despite this functional intersection the system of syntactic oppositions (such as the opposition subject.predicate) is in itself independent of the system of morphological oppositions, and this is in turn independent of the system of phonological

General Problems of Structural Linguistics

39

oppositions. No phonological (or syntactic) change can explain such a matter as the disappearance of cases in Bulgarian, the loss of the dual or of the aorists and the imperfect in Czech, the disappearance of subjunctives in the Scandinavian languages, the syncretism of dative with accusative in English, or the appearance of a new compound perfect in the Slavic languages. The area affected by phonological laws (which in the historical perspective are mutations) is only the manifestation of morphological oppositions; this is, however, something else than the oppositions themselves. If a phonemic mutation endangers the realisation of a morphological opposition, and if the complementarity of the resources of the language (i.e., the capacity of these resources to support one another, without being inactivated by the loss of one of them) is sufficient, analogy will find another way to manifest this opposition. Frequently the loss of the means of expressing a morphological opposition is connected with its homonymy. 13 The autonomy of the subsystems of a language explains the phenomenon which I have mentioned previously: the lack of proof that the mutation of phonemes, of combinations of phonemes, or of phonemic oppositions can be caused directly by morphological analogy. Two examples will recall my earlier observation. If in Czech forms of the same word exist side by side containing the allophones [«] and [77] of the phoneme /«/ (cf. [skler\ka], [skler\ky], etc. vs: [sklenek], [sklence] [glass] we should expect one of the two allophones to be eliminated by analogy (as occurs, for example, in the opposition /s/:/z/, leading to forms like \prozeb] on the analogy of \prozba] [request] ). But analogy cannot work here because [n] and [77] are allophones of the same phoneme and such an analogical change would make both allophones into phonemes. Another proof is the well-known phenomenon that in Czech and other languages, (e.g., Dutch) there is never a loss of the neutralisation of the voice in consonants in word-final by the analogy with inflected forms. If we do not wish to explain the lack of analogical levelling in this case and others as a matter of "chance" we must conclude that analogy as a morphological factor does not affect the phonemic inventory of the language, but operates only with complete semantic segments of words (with morphemes) formed from the phonemic inventory. 14 Similarly it can be stated that the existence of syntactic relations cannot be endangered either by changes in the morphological oppositions and their combinations or by their manifestations. Autonomy in these three areas permits us to treat each of them separately even in historical investigation. The discovery of structural changes from one linguistic stage to another does not exhaust the work of the historical linguist. It is also necessary to discover the goal which gives sense to the sequence of functional changes and orders them into a single line - their tendency. Without assuming a goal, a tendency, mutations would be isolated changes devoid of sense. Thus, when

40

General Linguistics

we discover a mutation we must ask what this means in terms of the overall movement of the given language. Perhaps I may be permitted to use an example from a field close to me, Verner's Law. From the phonemic standpoint this law deals with the neutralisation of voicing of the three spirants, f,p,x, after an unaccented vowel (or unaccented vowel plus consonant) and the formation of the allophone [z] of the phoneme /si in the same position. This occurred at a time when Germanic still possessed phonemic (movable) stress, and it ceased after the stress became fixed, i.e., dephonologised. We can thus view the disappearance of neutralisation as a compensation for this dephonologisation. But what was the tendency shown in this development? We will see the answer when we observe that a large number of other consonantal changes, which extended deep into the historical period of Old Saxon, Old English, and Old Norse as well, followed a single goal: to change the correlative bundle of the type p :f:v (which is documented, for example, in Modern Greek as well) into a bundle of the type p:b:f, i.e., to dephonologise the voicing of the spirants and phonologise the voicing of the stops. The beginning of this tendency is Verner's Law. This was the first step because with this change the voice opposition among spirants ceased to function in a certain position in the word. The shifting of stress destroyed the effect of this step and, in addition, threw a new roadblock in the way of the tendency: a new correlative pair of spirants s:z, which had to be eliminated. This was done by replacing ζ with R (r), which did away with the voice opposition in this pair. Unlike the individual mutations themselves, tendencies — as we can see — do not work by bounds but rather gradually. Initially they take up only a small space and go into effect cautiously; later, however, they consolidate their position and wage battle with other tendencies crossing their path and keep them from the goal, which they do not always achieve. The roads to the goal are thus varied, and the language is continually faced with the choice of several possibilities, which it anticipates and of which it chooses those which permit the greatest conservation of energy. A tendency brings a certain order into the bounds which a language makes; they therefore do not occur blindly and do not refashion the language mechanically section by section. They are rather like meaningful scenes in a play which never ends. Doubts might be expressed as to whether all the gradual mutations in the Germanic consonant system which we have mentioned — Verner's Law, the change of ζ to r, the change of voiced spirants into stops in certain positions, assimilation, loss of voicing of spirants in word-final, the change of voiceless spirants into voiced in a voiced environment, the loss of [Λ] in a voiced environment, etc. 15 — many of which took place after the breakup of Germanic linguistic unity, can indeed be grouped together as the workings of a single tendency. But are we to attribute to chance the fact that all these mutations work toward the same goal? If we exclude chance we must

General Problems of Structural Linguistics

41

recognise a tendency which tends toward a single goal, which develops, which calls forth other special tendencies, each of which preserves and develops those features of the original tendency which are not incompatible with its own speciality. The presence of these tendencies is what makes linguistic reconstruction possible, and the failure to recognise them would mean undermining the scientific foundations of reconstruction. In speaking of tendencies we must add that we always use the term as opposed to that which is constant. Tendencies presuppose constants, and the determination of constants (in our case, for instance, the stability of the oppositions among the liquids) is just as important as the establishment of tendencies. The concept of absolute changes is inadmissible to us, since thought in general is possible only in terms of relations. The observation of linguistic development leads to the recognition that language cannot be judged only from the standpoint of logical, intellectual values and relations. For a linguistic system to be a language it must be internalised. Only to the extent to which it is internalised, takes on the necessary expressiveness, becomes part of our self, does it become a dynamic system capable of continuing development in time and space. We can experience linguistic internalisation when we learn a foreign language. Even if we become relatively well acquainted with it from the standpoint of its intellectual relations it will remain a mechanical, unchanging system, remote from us, and one on which we have no effect, unless it becomes part of our life in at least some sector, and unless we feel deviations from it to be mistakes. By internalisation a complex intellectual tool becomes a simple organism, which meets the needs of its internal and external environment, into which it sinks numerous roots, as a tree into fertile soil. Without internalisation language would not function as a non-machine-like system of hierarchically ordered signs, since its elements would not have that capacity to intensify and supplant one another, which allows its dynamic movement through time. The theory of internalisation (which is an individual and a collective fact) explains particularly clearly certain aspects of language which have hitherto escaped our understanding. We recognise that internalisation of language does not achieve the same depth at the centre as on the edges; certain elements are on the periphery of internalisation, at a distance from those which have a more profound internalisation and which thus form their own linguistic stream, which is sensed as a uniformly functioning totality. With this concept we can explain the linguistic problem of foreignness (which cannot be dealt with merely in terms of the external origin of foreign features), since the cross section which in a language takes the form of a curve between foreignness and non-foreignness depends precisely on the varying depth of their internalisation. How else are we to explain, for example, the fact that in Czech linguistic consciousness /// is a foreign phoneme, although

42

General Linguistics

it stands in potentially the same relation to the Czech phoneme /v/ as /pi to I b / l Its structural foreignness — which is shown in t h e fact that its voiced opposite does not cause neutralisation of the voicing of preceding paired consonants, and thus does not belong t o any correlative pair — is reflected against the background of domestic phonological features, but this background itself is given b y a deeper level of internalisation. Without such an assumption we could not grasp the concept of a hierarchy of foreignness, since the varying degree of remoteness from the domestic background cannot be measured in terms of the number of deviant features alone. Without assuming internalisation it would also be impossible to imagine how one linguistic system could affect another, how it could not only take over foreign features as deliberate foreignisms, but even consider it desirable t o assimilate them and to bend its own elements to this end. The degree of internalisation of two linguistic systems which meet in the consciousness of speakers determines the type of structural influence exerted by one language on another, of the sort experienced by Rumanian and Hungarian in contact with the Slavic languages, the Lower Franconian dialects of German in contact with French, etc. A deeper degree of internalisation also serves to explain the difference between the traditionalisms in the standard language (such as the phonological traditionalism of /é/ in the standard Czech forms kamének [little stone], dobré [good] and the dynamic dialectal forms corresponding to these (kaminek, dobry). Different degrees of internalisation also determine the varying dynamic movement of the surface of language; it appears that the tendencies and bounds which we observed in the phonemic system of language were prepared within this coherent stratum which interlards the entire system, making of it a truly organic unit. This makes clear how correct is t h e statement that a completely successful phonemic analysis of a language can be made - all other things being equal — only by a linguist whose mother tongue and everyday speech is the language under investigation, since it is only by using his linguistic consciousness that he can perceive the smallest shifts in the tendencies which affect the system and which create its history. 1 6 The linguist who is unable to rely on his experience of an internalised language is, in analysing it, confined largely to quantitative evidence. The system of logical relations in language and its internalisation are two different aspects facing one another. Structural linguistics has emphasised in language the realm of logical relations, forms and categories, without which the objective nature of language cannot be approached. We must be aware, however, that another factor has silently been included in this investigation. In analysing language as a system of pure logical relations and values we would not be completely true to language without recognising the role of internalisation, because, however detailed such an analysis of language is, it

General Problems of Structural Linguistics

43

would lead us only to the establishment of classes with various numbers of members — including classes containing one and classes containing zero members 17 — which are inborn in the system and form its sub-systems. Removed from the system and properly evaluated, or rather numbered, they can then be put back into the system, like parts in a perfect machine. It is only the concept of internalisation which explains how this machine becomes a dynamic system and an adequate tool of the linguistic community, the tool we know and which was the object of our earlier considerations. 18 The internalised totality of language cannot exist by itself, but neither can the system of pure logical relations of language exist without it. This is the reason for the impossibility of foretelling the future state of a language, since to foretell precisely the future of a language as a whole would mean to have internalised it already. The internalisation of language occurs in the individual and in the collective consciousness and therefore it must not be confused with de Saussure's concept of "parole", the individual aspect of manifested speech. It belongs not only to "langue" but to "parole" as well, and its direction and intensity are determined by the experiences and goals of individual and collective life. 19 Although it is always accompanied by affect it must not be identified with the affective aspect of speech, since we can internalise values which are purely intellectual. If language necessarily includes internalisation, the question can be asked whether there is any regular correspondence between a language and the psychology of its speakers. Phenomenology denies any such direct relation; it contradicts W. Dilthey and W. Wundt by teaching that although meaning (Bedeutung) is intended by psychic acts, it is not exhausted by them, but remains in its objective field. It is isolated from them by its direction within a concrete whole, so that it does not coincide with them, although it is not opposed to them as something incomparably foreign, something forever separated from them in a real and unequivocal way. Linguistics and the psychology of speech are thus, according to phenomenology, two separate sciences — from the standpoint of linguistics it may be said that language shapes the individual thinking of its speakers, who internalise it, in some respect differently from the way in which thinking is shaped by another language. This feedback effect of an objective human product on its "mind" is not reserved to speech but belongs to every intellectual product of the group, which continually creates under the pressure of its own products. The forms of speech, however, do not correspond to the forms or levels of the "mind". 2 0 It is interesting that Brandal (cf. note 9) states that the English prepositional system emphasises a much more logical numerical relation than the French system, in which ordinative (geometric) logical relations stand in the foreground, and a greater tendency to abstraction exists; but to place these features in direct correspondence with the psychology of the two

44

General Linguistics

nations is inaccurate and premature. Furthermore, speech may reflect obsolete thinking patterns, since its development does not proceed at the same pace as the development of thought, as Ch. Bally correctly pointed out. 2 1 Who today would, for example, attempt to find a precise correspondence between the division of nouns into three genders in Czech and the psychology of Czech speakers, or to draw conclusions concerning the psychology of Hungarians, who have n o such grammatical category in their language? It is also found that even at primitive levels of development cultural and attitudinal categories do n o t correspond to linguistic categories. In numerous African tribes, for example, all objects are classified with respect to the four cardinal points of the compass, while their languages group nouns according to completely different categories. In other primitive tribes there is a very sharp social and attitudinal differentiation between men and women, but no trace of this is to be found in their languages. Let us ask another question. Does there exist any general tendency common to all languages? We know that all languages of the world have certain laws governing their common structure — in phonology, for example, there is the law o f the incompatibility of phonemic length with phonemic stress, and the law o f a certain oppositional minimum between phonemes in sequence within a word - and that the development of languages must follow these general laws. The broader question, however, cannot be answered affirmatively. If by this general tendency we mean the differentiation of functions, one can point to cases in which development leads to differentiation, but contrary cases can also b e found in which greater homogeneity develops from less. Nor can development toward ever greater abstraction or to ever greater concreteness be mentioned as a general law of linguistic development. One cannot even state that the development of languages is circular, i.e., that isolating languages become agglutinating and these then become inflecting, then isolating, etc. The basic error in this view is that we see in these types three basic stages of a single developing tendency, whereas in actuality they represent three different directions of the organisation of language, or, strictly speaking, of the manifestation of morphological oppositions, which may and do co-exist within one and the same linguistic system. One is not based on another. If the general tendency of linguistic development is thought to be the desire for simplicity one can point, to be sure, to examples of development resulting in greater simplicity (identical with the analytical structure in Jespersen's Progress in Language); 22 but one can also find cases in which development led to greater complexity. Furthermore, the judgment of the simplicity of a language is very subjective: from our standpoint, for example, we cannot say that Chinese, which has a very simple grammatical structure, is a particularly easy language, since we would find difficult precisely its analytical, kaleidoscopic character, with numerous homonyms. All that can be said in general

General Problems of Structural

Linguistics

45

of the development of language is that a language always attempts to be an adequate means of communication and expression for its speakers or, in Br0ndal's words, it is its own optimum. In order to be adequate, language uses mutations, which strive for a certain goal by the way of least resistance. The principle of vis inertiae also applies to linguistic systems: it moves where it finds the least resistance. This explains the fact that in linguistic development the (semantically neutral) phonemic level is more easily subject to change, while the system of morphological and that of syntactic oppositions, which are not so easily suppressed, are much more constant. Evidence for this is in the breakdown of a language into dialects: this occurs first in the phonology, against a solid common background of morphological and syntactic oppositions. Let us in conclusion consider the question of the immanence of a linguistic system. If the system could exist without internalisation, the impulse to the dynamics of such a system would necessarily be outside it. If, however, we recognise internalisation as an objective condition for its functioning as human language, both as "langue" and as "parole", the immanence of the system is established and the question of its self-initiated mobility clarified. A linguistic system conceived monistically as a system of relations, and the internalisation of this system in the collective and the individual consciousness of the speakers, together strive toward a common goal, although they are unrelated elements. The historical development of semantic systems depends on the interweaving of these two unrelated elements.

NOTES 1. Kant formulated the law of causality thus: "Everything which occurs presupposes something from which it follows according to some rule". 2. The name "mutation" was first used in biology, whence it was borrowed into the terminology of phonology. Hugo de Vries used the term to denote variations which, in contrast to the slow changes observed by Darwin, occur by bounds. 3. Havránek, Β., Weingart, M. (1932). Spisovná cestina a jazykovd kultura [Standard Czech and the Cultivation of Standard Language]. Prague. See also: Garvin, P.L. (1964). A Prague School Reader in Esthetics, Literary Structure and Style. Georgetown, Washington, D.C. 4. Vachek, J. (1942). "Psany jazyk a pravopis" [Written Language and Orthography]. In: B. Havránek, J. Mukarovsk^, ed., Ctenío jazyce a poesii [Readings about Language and Poetry]. Prague: 229-306. 5. Trnka, B. (1942). "Jazykovjf v^voj a tradicionalismy" [Language Development and Traditionalisms]. CMF 28: 397-402. 6. Paul, H. ( 5 1920). Prinzipien der Sprachgeschichte. Halle, Niemeyer. The concept of system was lacking in the linguistic theory of H. Paul. He could not do without it entirely, of course, so he used a substitute: a cross section through the speech activity of innumerable individuals. It was their interaction which he took to

46

General

Linguistics

be the proper subject of linguistics. What the qualities were which defined this cross section Paul does not say. His Prinzipien thus contributed to the atomistic development of linguistics since, in a period which rejected linguistic philosophy and preferred to deal only with "facts", it was the only work to unify linguistics into a single science. The structural linguist can use the conclusions of older historical grammar only in part; since this older discipline did not have an overall view of the total system, it saw only time sequences of isolated facts. The same is true of dialectology, which organised its material according to the time-depth of linguistic changes, regardless of their relationship to present-day dialects as linguistic systems. 7. Wijk, N. van (1939). "L'étude diachronique des phénomènes phonologiques et extra-phonologiques". TCLP 8: 2 9 7 - 3 1 8 . 8. Groot, A.W. de (1941). De Klankwetten in het licht der structurale taalkunde. Lecture delivered in the Netherlands Phonological Society. Published under the title "Structural Linguistics and Phonetic Law". Lingua 1, 1948, pp. 1 7 5 - 2 0 8 . 9. Brandal, V. (1950). Théorie de prépositions. (Translation of the Danish original 'Praepositionemes theori'. 1940) Copenhagen. 10. The question of whether the mental sciences (Geisteswissenschaften) can, by their very nature, permit the establishment of any laws at all was posed by: Rickert, H. ( 2 1913). Die Grenzen der naturwissenschaftlichen Begriffsbildung. Reacting to the monistic tendencies of the 19th century, outstanding among which was Mach's doctrine of sensation as the foundation common to the natural and mental sciences (a doctrine characteristic of the period of impressionism), and opposing the tendency to introduce the methods of the natural sciences into the mental sciences, Rickert denied the possibility of establishing any laws at all in the latter. This was the old Cartesian dualism in all its strictness: Rickert placed the mental, historical sciences with their individualising method in irreconcilable opposition to the generalising natural sciences, which lack a technological history. No such contrast exists in structural linguistics. We base our studies on experience, and, when we discover a structural system of linguistic phenomena, we demand that the individual cases follow laws absolutely. 11. If we believe that we can arrive at the history of a language by comparing individual cross sections, i.e., various linguistic systems, we cannot agree with J.M. Korinek that "for the exact solution of the problem of so-called linguistic development structural linguistics is not adequate". Korinek, J.M. (1942). "O jazykovém vyvoji" (About the Development of Languagej. CMF 28: 3 7 1 - 7 5 . Structuralism on the contrary lends precision and completeness to the examination of the linguistic facts of the past. 12. De Groot's idea of what belongs in the "linguistic system", like Brtfndal's, is based on de Saussure's concept, cf. Saussure, F. de ( 2 1923). Cours de linguistique générale. Geneva. Brandal includes in the concept of phonemic system only the repertory of phonemic oppositions, while combinations of phonemes are considered part of "parole". Syntax for him is also a matter of "parole", in contrast to morphology, which he includes under "langue". The relationship between syntax and morphology is treated similarly by A.H. Gardiner. Brandal, V. ( 1932). Morfologi og Syntax. Kopenhagen. Gardiner. A.H. (19321. The Theory of Speech and Language. London. 13. Cf. Gardiner (op. cit. (Note 12) ). Homonymy (multiplicity of meanings) is a fundamental linguistic phenomenon.

General Problems of Structural Linguistics

14.

15.

16.

17.

18.

19.

20.

47

Multiplicity of meanings is inconvenient in a language only when one of the meanings of a word-form is in contrast with another meaning of the "same" word within the same semantic sphere. In these cases it becomes necessary to use synonymous expressions or to eliminate the homonymy by distinguishing the two meanings which are to be placed in opposition by more detailed means (such as the use of the genitive, an adjective, etc.). Trnka, B. (1936). "O analog» ν strukturálnfm jazykozpytu" [On Analogy in Structural Linguistics]. SaS 2: 221 - 2 2 2 . The English translation is included in this volume, Section 1, p. 29. Trnka, B. (1938). "On the Combinatory Variants and the Neutralisation of Phonemes". Proceedings of the Third International Congress of Phonemic Sciences. Ghent. 23-30. Czech version in CMF24: 261-270. Reprinted in this volume. Section 2, p. 119. Trnka, B. (1935). "Fonologicky vyvoj i.e. explosiv ν star^ch germánsk^ch jazyclch" [Phonological Development of IE Plosives in the Old Germanic Languages]. CMF 21: 4 4 - 5 2 . Trnka, B. (1935). "Je prokázán Vernerûv zákon pro novou anglictinu?" [Can Verner's Law be applied to Mod.E.':]. CMF 21: 154-162. The English translation is included in this volume, Section 4, p. 210. Trnka, B. (1937). "Hláskoslovné zákon y ν strukturním jazykozpytu" [Sound Laws in Structural Linguistics]. CMF 23: 385-388. In examining the older stages of a language, therefore, to be successful, the scholar must immerse himself in all the written monuments of the period which he is investigating; he must not be content with the mere mechanical amassing of documents, which in themselves are of little value. He must penetrate the linguistic system of the period as a coherently functioning whole. By thus penetrating the individual stages of a language his reconstruction of stages undocumented or poorly documented with written monuments will take on greater validity. For classes containing zero and classes containing one member cf. "The Logic of Relations" in: Russell, B. (1930). Introduction to Mathematical Philosophy. London. Phonemics is indebted to Russell's logic for many ideas and terms. Without assuming internalisation a logical monistic system of this type would be only an abstract framework with no possibility of real existence. The fact that certain investigators, such as Ch. Bally and V. Mathesius, do not emphasise the concept of linguistic laws is perhaps because they look at language more from the standpoint of its internalisation. We must take this opportunity to point out that we use the word "individual" in two different senses. By individual we mean the smallest group of the collectivity to which it is structurally bound. Individuality, however, also refers to a totality which is unique and non-repeatable. By adding together individuals conceived in the latter way we will never arrive at collectivity, since they exclude the opposition of whole and part, being a purpose unto themselves. Inflexion is not a sign of superiority, nor is agglutination a sign of inferiority in language. The statement of the linguist Byrne that "there must be synthesis of thought where there is synthesis of expression", is not correct. Byrne, J. (1866). On the Science of Language. London. Byrne goes even further, however. Since psychic differences are explained by differences in the climate and soil on which a nation lives, the differences between

48

General Linguistics

languages are to be correlated with geographical differences. This is an old idea, to be found, for example, in Montesquieu, who established the dependence of the constitution and laws of individual nations on the climate in which they live. F.N. Finck does not go as far as Byrne, but he does connect the typical properties of a language with the character features of its speakers. 21. Bally, Ch. (1935). Le Langage et la Vie. Paris. 22. Jespersen, O. ( J 1909). Progress in Language, with Special Reference to English. London.

Linguistics and the Ideological Structure of the Period

It is almost a truism that any science, as a system of ordered knowledge, develops in close connection with the whole ideological structure of its period, and that its goals shift as the whole of this structure moves on. The history of a science can be defined as a history of its effort to reach its own goals; these goals, however, are never fully attained, because the structure of the superordinate science repeatedly imposes new goals upon that effort. Thus some problems remain unsolved, others appear in a new light, or problems entirely new are seen to emerge against the background of the constant features of the old ones. The establishment of the stages of development of individual sciences is done by immanent criteria which are selected by observers from the continuous flow of the happenings within the examined structure, but at the same time are prompted by the structure of the given period. The autonomous development of the given discipline inside the development of sciences taken as one whole marks off that discipline from the other sciences and guarantees that the tendencies operating in other sciences are not introduced into it mechanically but with due subordination to the structural features of that discipline. Whenever one attempted, in the past, to apply the achievements of one science to another in a non-structural manner — as a typical instance of that kind may be mentioned Schleicher's attempts at the application of Darwin's biological theory to linguistics — such attempts were doomed to fail. The history of linguistics, like the history of any other social science, has been conditioned by both the above-mentioned factors, the ideological

Originally delivered as a lecture at the Czech Academy of Siences and Arts in Prague, October 1943. Published under the title "Jazykozpyt a myálenková struktura doby" in SaS 10,1048, pp. 7 3 - 8 0 . Reprint of Vachek's English translation, published as "Appendix III" in J. Vachek (1966) The Linguistic School of Prague, Bloomington, pp. 152-165. (Slightly revised). (Cf. Vachek's reference to the political circumstances in which the lecture took place.) Czech original reprinted in J. Vachek, ed., (1972). O klasickèm obdobt praiské Skoly 1925-1945. (From the classical period of the Prague School 1925-1945. Sources of Czech and Slovak Linguistics.) Vol. 2. Prague, pp. 5 7 - 6 6 .

50

General Linguistics

structure of the period and the autonomous trends of the discipline. The higher speed of the development of linguistics, evidenced in the earlier part of the nineteenth century, is connected with the more violent development of research of the period building on the rationalist foundations laid by the eighteenth century. If linguistics is considered globally, as a whole of which the reconstructive Indo-European linguistics constitutes only one of the sections, it cannot be regarded as an exclusively Romanticist achievement; it should be realised that the formation of many important concepts, employed b y linguists even now, goes back t o earlier periods. It should be noted that the very historicism in linguistics, so typical of the nineteenth century, was growing out of strong rationalist roots. One can point out, for instance, that Josef Dobrovsk^, one of the founders of Slavic philology, was a typical rationalist of the Illuminati t y p e ; there was hardly any trace in him of the Romanticist trends of the younger Czech generation led by Josef Jungmann. After all, even the latter generation still revealed many Illuminati features, and Ugro-Finnic linguistics, which was being born in Hungary at about t h e same time, was also building on rationalist traditions. Admittedly, the Romanticist approach considerably broadened the linguistic horizons and intensified the interest in linguistic research. Furthermore, this approach prepared the ground for the flourishing of linguistics by removing from the culture and literacy documents of the past periods the stigma of barbarism, fought intensely b y the eighteenth century. In this way, even languages and dialects lacking classical elegance came to be regarded as objects worthy of study; there the Romanticist approach emphasised individuality as against generality, and t o classicism was opposed national specificity, reflected mainly by language, its phonic and morphological levels. But all this broadening and intensification of the study of language, which had originated in the early part of the nineteenth century, was based on firm rationalist foundations which were to become even stronger in the latter part of the same century. The positivist period was to preserve, as a Romanticist heritage, only the atomistic observation of the individual and his psychology, coupled with aversion f r o m logically semantic evaluation. On the other hand, it was to eliminate the Romanticist admixtures that, until that time, had affected the study of comparative Indo-European linguistics, and, besides, it was to add another important methodological step, namely, the distinction of changes purely phonetic from those caused by a factor called by the Aristotelian term of analogy. For a period characterised by biological and materialist tendencies it was only natural to regard sound changes as reflections of basic, blindly operating factors of the physiological kind, while the operation of morphological analogy was evaluated as a merely negative factor, violating,

Linguistics and the Ideological Structure of the Period

51

by subsequent psychological intervention, the absolute regularity ("Ausnahmslosigkeit") of physiological processes. Classified as another factor apparently violating the regularity of sound-laws was the taking over of loans in which the native language obtained such words as had not been subjected to the processes of its sound changes (on the other hand, however, such words yielded valuable chronological cues). These three simple principles established by the diachronistic Neogrammarian theory of language — the absolute regularity of sound changes, the operation of analogy, and the taking over of loans — yielded a basis satisfactory for the classification into periods, as well as for the stratification, of richly preserved language materials, and for finding out their exact genetic relationships by using the method of genetic reconstruction. The fruitful pioneering effort of the German grammarians of the 1870's is certainly admirable, but today their achievement can and must be subjected to criticism. The Neogrammarians atomised facts of language, in full conformity with the bias of their period, tearing them out of their structural connections, although it is exactly these connections to which the facts of language owe their existence as signs. The sound correspondences, found in Neogrammarian linguistic analyses, admittedly inform about the exact time sequences of the examined phonic facts but disregard the changing structures of subsequent stages, so that the established facts cannot be duly placed into the systems to which they belong. This shortcoming is clearly seen by anyone studying, for instance, Noreen's Vârt Sprâk. 1 The distinguished Swedish linguist, who can be regarded as one of the precursors of modern linguistic thinking, enumerates in exhaustive detail all sounds represented both in Standard Swedish and in Swedish dialects, but his list of sounds has no intrinsic value for a scholar personally ignorant of these dialects, because no information is given on the structural position of these sounds in the systems of the dialects concerned. Similarly, grammars of dialects presenting diachronistic correspondences of sounds can give no idea of what the present-day structure of such dialects looks like, not to speak of the lack of structural description of their development. Linguists of the earlier periods never drew methodological conclusions from the fact that speech sounds have no real existence outside their own system of language, and they used to place on the same level elements structurally very different (and sometimes did not even realise the existence of such elements). The misrepresentation of the system of language, now seen in their writings, was not obvious to them, of course, for the simple reason that the object of their research had been the genetic comparison of related languages, which, structurally too, are very closely similar to one another. It was exactly the concentration of research upon diachronistic similarities of the compared stages of languages and the tendency to deduce such

52

General Linguistics

similarities from the parent language that prevented these linguists from analysing structures of languages as wholes. It is clear that the above-characterised type of research could not give rise to structural linguistics, which could only arise from considerations of nonhistorical comparison of totally different systems of language. Besides, the non-structural bias of the Neogrammarian approach was necessarily implied by the very theory of this approach. This is clearly seen from H. Paul's formulation of the task of linguistics, given in his book Prinzipien der Sprachgeschichte (the very title of the book reflects the ideology of the period of its origin).2 In Paul's opinion, the task of the linguist is to ascertain the sequence of changes of the speaking activity of all individuals using a language; as, in practice, this ideal goal cannot be reached, one should establish an average (Durchschnitt) of the speaking activities of a greater number of individuals in their mutual relations. Is it, however, possible to identify a present-day Czech word with a word pronounced centuries ago by John Hus or J.A. Comenius without making use of the term "immanent structure"? Obviously, linguistics cannot claim the exactness demanded for it by H. Paul to make it, as it is in his opinion, the most exact of the historical sciences, if the linguist's view is narrowed down to what is individual and cannot be repeated, having no history. The theoretical contradiction, which cannot be bridged by Paul's postulate of establishing the average of the speaking activities, is seen to undermine the foundation of historical linguistics, and it was only logically consistent if, in the Neogrammarian school itself, voices came to be heard denying any possibility of historical comparison. 3 Paul's views were in full agreement with the structure of science of his period. Perhaps it will not be too far from the truth to say that Paul's views on the psychological basis of sound changes already reflect the impressionism of the years to come, the impressionism anticipated also, for instance, by the German physicist Ernst Mach4 with his theory of sensations forming a basis common to both social and natural sciences. Steinthal's and, later on, Wundt's effort to account for facts of language in terms of collective psychology is a step which, despite some differences, does not take us too far ahead of Paul's linguistic theory. The former, despite the motto "Völkerpsychologie" written on its banner, 5 is hardly based in essence, on anything but the individual psychology of sensations and images. In Steinthal's case one has to do with Herbart's associationist psychology, in Wundt's case with his own apperceptionist psychology. The chief difference between Steinthal and Wundt on the one hand and Paul on the other was that the latter, as a linguist, openly stressed the individualist basis, common also to the former two scholars, and that, unlike them, led by his common sense, he wanted to obtain support for linguistic research not only from psychology but from other disciplines as well.

Linguistics and the Ideological Structure of the Period

53

It is thus clear that the nineteenth century German linguists worked out their theories in close contact with the then contemporary currents of German thinking in other sections of the structure of science. This connection of linguistics and other social sciences could be traced back to the times of Leibnitz, Herder, and Kant (let us recall that Humboldt's general linguistic theory had been shaped under the influence of Herder's and Kant's philosophical systems). Similarly it could be observed that British, American, and, to a degree, Scandinavian linguistics reflected the influence o f philosophical views of Bentham and Spencer. One should, however, devote some attention to French linguistics, which, too, in the nineteenth century kept on developing in contact with the native currents of thinking, even if it did not shrink from accepting concrete results and methods of German research work. Good evidence of this is the linguistic theory of Antoine Meillet. 6 In its essence, Meillet's linguistic theory points to the old French ideal of a general grammar, aimed at b y philosophers and linguists of the seventeenth and eighteenth centuries (it is symptomatic that even in the turbulent times of the French Revolution one published a general grammar intended for mankind as a whole), but it chooses a different method and a different starting-point. Meillet, unlike Leibnitz, does not base his conception on aprioristic philosophical and logical considerations, but, in agreement with the historical bias of his century, concentrates his observation on the changes of a language, phonetic as well as morphological, and by their mutual comparison intends to discover the laws that overstep the limits of one language family and are concerned with the whole of mankind and with every individual whose psychophysical nature they express. In other words, the French comparativist tries to find one general formula and one psychophysical reason for any phonetic and other type of change. Changes or droppings of intervocalic consonants, for instance, are explained as due to general principles of articulation — here the promising development of French phonetics was giving Meillet's theories ample support — and the simplification of complicated morphology is again accounted for b y the psychological and logical tendencies of languages to arrive at one and the same kind of form f o r one and the same kind o f function. Physiological, psychological and logical conditions of language, which are common to all people, are conceived as unifying factors, working in language development incessantly against moments of variable character. The big factor continuously disrupting the unity of language is the structure o f the society speaking the given language and concretising, after its own regularities, one of the general potentialities given b y the language. Language changes incessantly, but its changes, as Meillet emphasises, must be sudden, because otherwise mutual understanding, which is the main task of language, might be jeopardised. Changes of language correspond, in part, to historical conditions, to shifts in the social situations due to foreign invasions,

54

General Linguistics

colonisation, changes in population, and various cultural currents. In the long run, language unity is rooted in the homogeneity of civilisation, and dissociation of that unity is connected with big historical events. All these views of Meillet, which can be outlined only very briefly, are clearly based not only on the general French cultural tradition but also on the contemporary currents of French sociology and philosophy. Meillet connects changes in language (which is considered by us as an autonomous system of oppositions sui generis), in an absolute, straightforward manner, with changes of structures essentially alien to the system of language. The lack of the concept of the system of language où tout se tient prevented Meillet from being completely successful in looking for his unifying principle, but his undisputed merit lies in his effort to emancipate linguistics from purely genetic considerations. The same can be said about Grammont's theory of sound changes, in which the French phonetician is trying to arrive at a unifying view of historical kinetics of most varied languages.7 With the stimulating philosophical activity of the pre-war years is also closely connected the linguistic theory of the Geneva scholar Ferdinand de Saussure, whose place in the history of linguistics is comparable to that held in philosophy by F. Brentano with his pioneering work. In his well-known work, 8 the Swiss comparativist posed, in the first place, the question of what exactly is the object to be examined by linguistics. He posed that question in a period whose most distinguished thinkers had already given up dogmatic positivism and atomising psychologism, and answered it in what is now termed the structuralist manner. At that time, the whole structure of science was found to be slowly changing. The associationist and apperceptionist psychology was being replaced, in Germany too, by "Denkpsychologie", examining bigger and higher contexts of psychic processes. Also the new logic was coming into its own; unlike the logic of J.S. Mill, it was grounded not in the psychology of sensations or in psychological experience but in relations and system. In philosophy, the positivist era was being outdone by emphasis laid on purposeful activity, whether in the form of Bergson 's philosophy of élan vital, or of American pragmatism, English neorealism or, finally, the logistically oriented phenomenology. Saussure's merit consists in his having emancipated linguistics from the danger of isolation, impending on it in these days of new scientific thinking, and in having placed it on a new logically structuralist basis by replacing the atomising realism and naive empiricism of the older school with the concept of function and system of hierarchically ordered signs. Linguistics — and, later on, aesthetics — have been definitely getting rid of the psychological bias and finding an ally in the modern relational logic worked out by Bertrand Russell; modern logic takes on the part of the leading science in the structure of sciences of today, and even permeates

Linguistics and the Ideological Structure of the Period

55

psychology itself. Linguistics may entrust itself to this current without any fears for its autonomy, formerly menaced by psychologism, because relational logic and allied currents of thought teach it to rest on its own basis and to start from the minimum number of further unanalysable prerequisites - in short, provides it, just as it does, for instance, psychology — only with a model of its procedure. How fruitful the structuralist procedure is, resting on that basis, is evidenced by the functional analysis undertaken, with the cooperation of Prague linguists, by Professor N.S. Trubetzkoy of the University of Vienna in the domain of the phonic facts of languages. It has become known under the term "phonology". Although the model of modern linguistics as a science dealing with the signs of language is furnished by relational logic, closely related to the philosophy of mathematics, it must be stressed that the object of linguistic research is empirical reality, not the underlying fact of human thinking, and that, thus, the results of linguistic research, dealing with empirical facts, have empirical validity. Language as a system of signs used for definite purposes becomes changed in space and time, and laws arrived at have thus a limited spatial and temporal validity. Even the validity of the so-called general laws of language is necessarily of an empirical character; they cease to have this validity as soon as one finds that for this or that language they do not hold. Today's structuralism takes into account both factors, time in language and space in language, and regards Saussure's identification of structuralism and synchrony as erroneous. The sharp line drawn between synchrony, treating language as a system, and the non-structuralist diachrony, which only violates that system, is clearly fictitious, because language changes continuously, even in the present period. One must not be misled by the apparent absence of changes in the standard language: language is a system at any moment of its development. Unless the structuralist approach is to commit the well-known methodological mistake of a metabasis eis alio genos, it must subject to structuralist analysis also the development of language: facts of language must not change their status in diachronistic research. Equally unacceptable, of course, is the old standpoint that the scientific approach to language is necessarily an exclusively historical one: in doing one's research, one is always free to substitute zero for the time factor and not to delimit the temporal extent of the established structure of language. If we approach this task from the structuralist angle, our results will not contradict those which could be arrived at in taking into consideration the state of language immediately preceding or following the period under examination. In establishing the system of language of the remote past, it is of course indispensable to consider the preceding and the following stages of development, because this consideration includes moments of heuristic

56

General Linguistics

importance for the stage under analysis. On the other hand, in analysing the language of the present-day stage the suspension of the factor of time often proves to be advantageous and sometimes necessary; this should also apply to the comparison with a number of languages of any period. Structural linguistics is, then, not oriented antihistorically as it is still believed in some quarters; on the contrary, it acknowledges the importance of diachronistic research, which is also deepened and made more precise by the structuralist approach. In the domain of historical phonology, which constitutes the most developed branch of structural linguistics, one examines the sound-laws (i.e., the regular historical correspondences of sounds) in their functional relation to the changes of the system taken as a whole, and one tries to find out the functional connections existing between individual mutations which resemble consecutive moves on the chessboard, moves directed toward some goal (which need not necessarily be always reached and can be dropped, in the course of the game, and replaced by some other goal). Thus, phonology does not place the phonic changes in language in a sort of Cartesian opposition, to the so-called changes by analogy, but rather regards these phonetic changes as a result of some sort of analogy, co-operating with the morphological analogy very precisely and without any overlappings, just as the morphological level co-operates with the level of syntax. As a matter of fact, this conception makes more precise the very concept of the absolute validity of sound-laws, because it is found that phonological mutations, as long as they are functionally valid, cannot be violated by interventions of morphological analogy. The absolute regularity of mutations — just as the non-existence of exceptions from morphological laws — does not follow from the blind operation of physiological factors but from the necessity of a purely logical kind. As long as the Old Czech umlaut a>è (resulting in the neutralisation of the opposition a / è in some positions of the word) was in operation, every word had to participate in this mutation just as necessarily and unfailingly as, in the field of morphology, any feminine noun ending in -a had to form its genitive case by the ending -y. Only after this neutralisation had come to be cancelled by another mutation (by the merger of è and e), an analogical a-sound could replace the umlauted e (from é)\for the sake of clarity it should be pointed out that the opposite analogy (replacing a by ê, e) is not concerned with neutralisation, so that it could emerge even before the merger of ë and e. Of the same kind, i.e., unaffected by analogy, is also the absolute validity of the neutralisation of paired voiced consonants in word-final positions before a pause. The fact that both analogies, phonic and morphological, operate side by side, without any mutual interference, can be used as a cue for the establishment of the chronology of mutations. As noted above, the task of structural diachrony is to link individual mutations into their mutual functional connection which they have in view

Linguistics and the Ideological Structure of the Period

57

of their common goal. The history of language may thus be compared to a never-ending game of chess in which each of the moves is given by the total situation of the hierarchically ordered signs. The inner division of this infinite game into a series of stages is based on criteria immanent to the game but, at the same time, relevant for the observer at a given time, and it varies not only according to the position in time occupied by the observer but also according to the length of the period comprised by the observer establishing such stages. The establishment of stages in the past can thus never be definite (and the same is true about the classification of sciences), because one of the factors, the observer, is a variable factor in this process. The above commentary will have shown with sufficient clearness the distinction between the earlier and structuralist linguistics in the very domain which has ever been in the foreground of Neogrammarian interest. It would be possible to discuss other domains of language, such as morphology, syntax, and semantics, and to follow in them, too, the radical change of orientation, but lack of space forbids this. These other domains, too, would reveal again the same parallelism with the structure of the science of the period, just as does an effort to ascertain structural oppositions with the purpose of establishing the system, one would further ascertain departure from the psychologistic approach, which, in these sections of the analysis of language, was denounced as inadequate already by K. Vossler, and the stress laid on the objectivising procedure, the most outstanding model for which is supplied by relational logic. It may not be superfluous to emphasise, here too, that despite the connections existing between the development of linguistics and that of broader science groups — documented, for example, by the parallelisms found between American pragmatism and Sapir's theory of language - linguistics is developing independently in so far that it has to reshape the external influences inside its own scientific sphere if it is to discover new treasures of "facts" and problems. Any science has (and must have) a certain expansiveness and, at the same time, the ability to go through a crisis involving its fundamental concepts. For this reason, the history of new achievements in separate sciences is most complex. A graphic example of such complexity is supplied by the development of the concept of the phoneme from its emergence with Baudouin de Courtenay until our days: this development shows how its psychologistic definitions have given way to definitions worded in purely logical terms. The phoneme, the basic operational unit of phonology, is no longer conceived as a psychological bond conditioning the unity of differently articulated sounds, but as a member of the inventory of language means which are capable of distinguishing the items of the intellectual vocabulary of the given language. It was only in this deeper conception, emancipated from psychologistic admixtures, that the phoneme could become a fruitful linguistic concept.

58

General Linguistics

Speaking about the system one should realise that this concept cannot exist without another important factor which, for want of a better term, we propose to call language internalisation. One may use here the Latin dictum, Expellas hominem logica, tarnen usque recurret. Structuralism does not exclude this factor permeating all levels of the language system; on the contrary, this factor is included in the structuralist approach by the emphasis laid on the teleological aspect of the speech act. Concretely one can refer here to the work of Ch. Bally who systematically studied the affective means of language, basing his theory not on the old psychology, but on the theory of language outlined by his teacher F. de Saussure.9 Herein lies the great merit of Bally's work. Like Saussure, he admits that language is a system of superindividual, intellectual values, but he studies the emotional aspect of language, an aspect which he terms stylistic. Rousseau's countryman, Bally, may have been here, like other contemporary French linguists, under the influence of Bergson's philosophy. We cannot discuss here his views in detail. Suffice it to say that our concept of language internalisation has a broader meaning because it includes the internalising not only of affective but also of intellectual elements of language. This broader conception of our term seems to be indispensable for linguistics. It is only by being internalised that the logically structured mechanism of language, for the most part very complex, becomes a functional totality whose adequacy can be satisfactorily tested by a non-mediated reality of language consciousness. Any deviation from this non-mediated reality, constituting a system of values, is identified as a mistake. Unequal degrees of language experience give rise to zones of different depths, which also have different structural developments. Differences in internalising language are responsible, for instance, for the different developments of the dialect and the standard language, or, for that matter, of domestic and foreign elements of the vocabulary. Without considering language internalisation one could not imagine how a certain language system could, in different degrees, influence other systems, or why that system could not only take over some items of some other language in the quality of intentional foreignisms, structurally differentiated from the native elements in the background, but even, on other occasions, regard as preferable a complete assimilation of such elements. The very concept of function would be unclear if one did not consider the factor of language internalisation which, however, should not be taken in the psychological but in the purely linguistic sense. Both factors, the system of language and language internalisation, must necessarily be regarded as mutually related counterparts, because language lacking internalisation would be no more than an unchanging system of relations with no possibilities of development; and, on the contrary, without the values of logical order, language

Linguistics and the Ideological Structure of the Period

59

would not only be equal to an accumulation of phonic utterances. The existence of their fundamental polar relation lies at the basis not only of Bally's conception but also of that of his German forerunner Ph. Wegener whose Untersuchungen may be classified, incidentally, as one of the best attempts at general linguistic analysis written in the nineteenth century. 10 Finally, the said polar relation can account for alternate swings to "classicist" or "romanticist" trends in periods of cultural history (it is, however, necessary to stress the fact that the never-recurring concrete historical periods, denoted by these terms, are structures of very complex character). Both factors are, of course, always present in language. Some scholars are more inclined to analyse the former of the two aspects of language and lay particular stress on the postulate of absolute regularity of language phenomena, while others, such as Ch. Bally and V. Mathesius, prefer to observe language experience and speak only of language trends or tendencies. Naturally, it cannot be denied that by its teleological essence language is a system of intellectually motivated diacritical values and that, thus, also the examination of the other factor, that is, language experience, must be done against the background of the results obtained by the analysis of those values. The present-day structure of science has been shifted — as is shown above all by linguistics — in the direction of structurally objectivising orientation, in which the noetic model begins to be supplied by logic. This shift is noticeable also in other sciences, social as well as natural, and so things appear to be tending to an era of a certain mental balance, such as used to characterise the eighteenth century. This time, of course, the balance is turning upon a different axis, and also its relation to the preceding centuries is different. We do not discuss here the speed of penetration of the new orientation in individual disciplines or the concrete formation of sciences and their presentday tasks — all this would involve taking into account factors of very complex character. The progress of the structure of science is best implemented in peaceful cultural communities whose linguists or aestheticians or physicists can draw inspiration from their philosopher and sociologist (and other) colleagues, and, the other way round, whose philosophers and sociologists can be stimulated by their linguist, aesthetician, physicist (and other) colleagues who are able to keep pace with the latest developments in their sciences, who can experience and assimilate the results of these developments, and so can exercise leadership in pioneering research work. But even in the absence of such mutual support of local disciplines the structure of a science can profitably develop because a more highly developed sector, whether a science or an art, can substitute for the shortcomings of other sectors. Finally, one should not forget the strength obtained by any cultural community from the contact and solidarity with sectors of the structure of science proper to mankind taken as a whole.

60

General Linguistics

NOTES 1. Noreen, A. (1912). V&rt Sprtk, Sweden. 2. Paul, H. ( 5 1920). Prinzipien der Sprachgeschichte. Halle, Niemeyer. 3. Schmitt-Erlangen, Α. (1936). "Die Schallgebärden der Sprache" Wörter und Sachen 17: 5 7 - 9 8 . 4. Mach, E. ( 5 1906). Die Analyse der Empfindungen und das Verhältnis des Physischen zum Psychischen. Jena. 5. Cf. Steinthal, H. and Lazarus, M. (1860). "Einleitung". Zeitschrift für Völkerpsychologie und Sprachwissenschaft 1: 1 - 7 3 . 6. Cf. Meillet, A. (1921, 1938). Linguistique historique et linguistique générale. Vols. 1, 2. Paris. 7. Cf. Grammont, M. (1933). Traité de phonétique. Paris. 8. Saussure. I". de ( 2 1923). Cours de linguistique générale. Genève. 9. Bally, Ch. (1935). ¿ e Langage et la Vie. Paris. 10. Wegener, Ph. (1885). Untersuchungen über die Grundlagen des Sprachlebens, Halle.

Zur Erinnerung an August Schleicher

In der Prager Universitätsbibliothek befindet sich ein interessantes Schriftchen, das, soviel ich weiß, in völlige Vergessenheit versunken ist. Es ist ein tschechisch geschriebener Brief Odevreny list cizozemského linguisty cechoslovanovi [Offenes Schreiben eines ausländischen Linguisten an einen tschechischen Slawen], welcher den Haupttitel "O spisovnej cestìnè" [Über die tschechische Schriftsprache] trägt. Es wurde in Bonn geschrieben und in Leipzig bei Breitkopf und Härtel im Jahre 1849 gedruckt. Was ist der Inhalt dieses Schreibens? Der Verfasser, ein deutscher Sprachforscher, preist das Tschechische wegen seiner Altertümlichkeit und seiner wahren Prosodie, welche es der Vollkommenheit der klassischen Sprachen nahebringt und sogar über das Polnische und Russische stellt, bedauert aber, daß diese vorzüglichen Eigenschaften sich in der Schriftsprache weniger auswirken als im Munde des Volkes: Es gebe nämlich in jener weniger grammatische Fälle und Endungen, auf denen die freie Wortfolge, Genauigkeit und Schönheit der Sprache beruhe. Es sei aber ein übliches Gesetz, daß die Sprachen derjenigen Völker, welche in Literatur und Volkskultur vorgeschritten sind, sich mehr abschleifen und schneller ihre ehemalige Schönheit verlieren als die Sprachen der weniger gebildeten Völker. Das Tschechische habe eine seltsame Stellung inne, weil es in seiner Entwicklung durch fremde Einflüsse gehemmt wurde, und man müsse darum fragen, wie seine kostbaren Eigentümlichkeiten geschützt werden sollen. Im grossen und ganzen müsse man sagen, da/3 die Fürsorge der Patrioten für die tschechische Schriftsprache auf gutem Wege ist; er empfiehlt aber, sie mit dem zu bereichern und zu verschönern, was älter und überhaupt besser in mährischen Dialekten sei. Der Verfasser wisse freilich, da/3 keine Sprache in ihrer Entwicklung stehenbleiben könne; aber das Tschechische sei eine unterdrückte Sprache, und es sei die Pflicht der tschechischen Schriftsteller, sie von diesem Verfall zu befreien, ihr eine angemessene und richtige Form zu geben und so zur Heilung der

Originally published in Zeitschrift für Phonetik und allgemeine Berlin, No. 6, 1952, pp. 1 3 4 - 1 4 2 . (Partly Revised).

Sprachwissenschaft,

62

General Linguistics

Krankheit, welche sie befiel, beizutragen. Sie werden keinen unbekannten Weg betreten, weil ihnen schon die Griechen mit gutem Beispiel vorangegangen seien, welche ihre Muttersprache von verunzierenden türkischen Wörtern gereinigt und ihr besonders die schönen vollen Formen der antiken Sprache zurückgegeben hätten. Auch das Tschechische solle sich seiner alten Schönheit annähern, und dies sei nur möglich, wenn die volleren Formen der mährischen Dialekte, welche in einiger Hinsicht sogar altertümlicher als die slowakischen Mundarten seien, ihm einverleibt würden. Der Verfasser habe selbst das "Mährische" kennengelernt, es mit dem Alttschechischen, Slowakischen und anderen slawischen Sprachen verglichen und führe die Formen an, welche seiner Meinung nach die tschechische Schriftsprache bereichern würden. In diesem so reformierten Tschechisch ist das ganze offene Schreiben des deutschen Linguisten absichtlich abgefaßt. Durch sein Bestreben, die mährischen Formen in die tschechische Schriftsprache einzuführen, nähert er sich einigen mährischen Schriftstellern der dreißiger Jahre, aber seine Motive sind ganz verschiedenen Charakters. Es handelt sich bei ihm nicht um den lokalen Patriotismus, wie es bei Fr. D. Trnka und V. Zák der Fall war, sondern um die praktische Durchführung seiner Sprachtheorie, um die Anwendung seiner eigenen Kriterien der sprachlichen Schönheit an eine lebende Sprache. Seine Bemühungen hatten natürlich kein Aussicht auf Erfolg, und sein Schreiben ist daher gänzlich in Vergessenheit geraten. Es ist jedoch ein kostbares Beweisstück des Interesses eines ausländischen Linguisten für das Tschechische, das er gut beherrscht hat:es ist vielleicht überhaupt die erste tschechisch gedruckte, von einem ausländischen Nichttschechen geschriebene, Schrift. Der Verfasser dieser Schrift ist August Schleicher, der hervorragende Sprachforscher und der beste deutsche Kenner des Slawischen und Litauischen in seiner Zeit, der in den Jahren 1850—1857 an der Prager Universität gelehrt hat. Der tschechische Freund, dem er das Büchlein von Bonn her zuschickte, war Alois Vanicek, Vorgänger von Professor Josef Zubaty auf dem Lehrstuhl für Sanskrit und vergleichende Sprachwissenschaft. Das offene Schreiben bezeugt, da/3 Schleicher seinem Ruf an die Prager Universität im Jahre 1850 aus wahrem wissenschaftlichen Interesse Folge leistete und da/3 er seine Professur nicht für eine Anfangs- oder Übergangsstation in seiner weiteren akademischen Laufbahn hielt. Gerade jetzt verdient Schleicher diese — man könnte sagen — "Jubiläums" Erinnerung an seine akademische Tätigkeit in Prag, und zwar nicht nur, weil er gern in seinem Prager Milieu verkehrte, sondern auch weil er durch seine Altkirchenslawische Grammatik, die er vom Standpunkte der vergleichenden Sprachwissenschaft in Prag geschrieben hat, der Begründer der deutschen Slawistik geworden ist.1 Tschechisch hat Schleicher schon während seines frühen Aufenthaltes in den tschechischen Ländern gelernt. Im stürmischen Jahre 1848, und zwar kurz

Zur Erinnerung an August Schleicher

63

nach der Februarrevolution, hielt er sich in Paris als Korrespondent einer deutschen Zeitung auf, von da begab er sich nach Kremsier (Kroméríz), wo der erste österreichische Reichstag zusammentraf. Nach dessen Auflösung am 7. März 1848 reiste er nach Prag ab, wo ihm Jan Pravoslav Koubek, Professor der tschechischen Sprache und Literatur an der Prager Universität, einen Juristen Alois Vanicek, der soeben von Czernowitz nach Prag zurückkehrte, als Lehrer der tschechischen Sprache empfahl. Schleicher hat sich damals in Prag nur zwei Monate aufgehalten. Er wollte hier länger verweilen, aber die österreichische Polizei, der der tschechisch lernende und sprechende Deutsche verdächtig war, vorenthielt ihm Geldsendungen aus Deutschland, so daß er genötigt war, Prag schon gegen Ende April zu verlassen und sich nach Bonn zu begeben, wo er sich vor drei Jahren (1846) habilitiert hatte. Die Summe (40 Gulden), die er zur Rückreise über Berlin brauchte, lieh ihm Vanicek. Obgleich der erste Prager Aufenthalt Schleichers verhältnismäßig kurz war, genügte er ihm, um das Tschechische vollkommen zu beherrschen. Es zeugt davon nicht nur sein Offenes Schreiben, sondern auch seine tschechisch geschriebene Abhandlung über den Infinitiv und das Supinum in den slawischen Sprachen, die er gleichfalls von Bonn nach Prag sandte und die in Casopis ceského Musea abgedruckt ist.2 An die Prager Universität berufen, hat Schleicher seinen tschechischen Freunden bald persönlich begegnen sollen. Die ersten zwei Semester hielt er Vorlesungen auf dem Gebiet der klassischen Philologie, aber schon im Jahre 1851 wurde es ihm erlaubt, statt dessen das Fach "Vergleichende Sprachwissenschaft und Sanskrit" zu vertreten, da klassische Philologie inzwischen auch von Georg Curtius vorgetragen wurde, der ein halbes Jahr vor der Ankunft Schleichers einen Ruf nach Prag erhalten hatte. Die politische Atmosphäre Prags war nach der mißglückten Revolution im Jahre 1848 sehr bedrückend. Die strenge österreichische Zensur überwachte jeden Ausdruck freieren Denkens und zensierte sogar die griechischen und römischen Klassiker. Die Polizei verdächtigte immer den Reichsdeutschen, Protestanten und Hegelianer, welcher über neue freiere Gedanken gern diskutierte und die Gesellschaft von Tschechen aufsuchte. Schleicher befreundete sich mit tschechischen Gelehrten wie P.J. Safarik, J.E. Vocel, F.J. Celakovsky, F.B. Mikovec und V. Hanka und hielt in der philologischen Sektion der Königlichen Böhmischen Gesellschaft der Wissenschaften, in der sich das tschechische wissenschaftliche Leben vor der Gründung der Tschechischen Akademie konzentrierte, dreimal einen Vortrag. Am 6. April 1853 hielt er einen Vortrag über die litauische Sprache mit besonderer Berücksichtigung des Slawischen,3 am 8. Januar 1855 über das Futurum im Slawischen und Germanischen und am 21. Januar 1856 über ein ähnliches Thema. 4 In den beiden letzten Vorträgen verteidigte er seine, jetzt verworfene These über eine engere Verwandtschaft des Slawischen und Germanischen, welche er besonders in der futurischen Bedeutung des Präsens der Perfektiva

64

General Linguistics

in beiden Sprachzweigen bestätigt sah, und meinte gegen die Ansicht Bopps, daß die Germanen sich nicht früher, sondern gleichzeitig mit den Slawen von der gemeinsamen indoeuropäischen Ursprache losgelöst hätten. Der slawischen und speziell tschechischen Grammatik widmete er Abhandlungen auch in den

Sitzungsberichten der Wiener Akademie und in der Zeitschrift für österreichische Gymnasien.5 Seine vorzügliche Kenntnis des Tschechischen bezeugte er auch durch seine Übersetzung der Einschaltungen im Mahäbhärata über die Sintflut und über Naia und Damajanti,6 von der man trotz der ungünstigen Kritik Martin Hattalas auch jetzt noch sagen kann, da/3 sie sprachlich so frisch ist, daß es möglich wäre, sie mit nur geringer Modernisierung auch noch heute herauszugeben. Die wissenschaftliche Arbeit Pavel Josef Safariks schätzte Schleicher hoch; er übernahm auch dessen damalige Theorie über den Ursprung des alten Kirchenslawischen und brachte sie in seiner "Formenlehre der kirchenslavischen Grammatik" zur Geltung. Einen bedeutenden Platz hatte Schleicher den slawischen Sprachen schon in seinen früheren u n d auch späteren Werken zuerkannt. 7 Er begrenzte aber seinen Umgang nur auf philologisch interessierte Gebildete. Er debattierte gern über politische Fragen an Bierabenden in Gasthäusern, manchmal so offen, da/3 sein tschechischer Nachbar am Tisch — aus Furcht vor einem feindlichen Ohre — ihn heimlich am Ärmel ziehen mußte. Durch sein verständnisvolles Benehmen gewann er — obwohl er weder ein guter Redner noch eine besonders anziehende Persönlichkeit war — auch die Sympathien der litauischen Bauern, mit denen er aus gemeinsamer Schüssel aß, als er im preußischen Litauen im heißen Sommer und Herbst (im Jahre 1852) Material für sein "Handbuch der litauischen Sprache" sammelte. 8 Über den unveröffentlichten Teil seines handschriftlichen Materials, welchen er der Prager Universitätsbibliothek schenkte, hat später Josef Zubaty einen ausführlichen Bericht im Vestnik der Königl. Böhm. Gesellschaft der Wissenschaften gegeben. 9 Während seines Aufenthaltes in Litauen, der selbst seinen abgehärteten Körper stark in Anspruch nahm, erinnerte er sich in seinen Briefen gern an Prag und besonders, wie er scherzhaft Vanicek schrieb, an die vollen Fleischtöpfe im " T e m p e l " und an das Bier bei Novotny (Spálená ulice). 10 Sein Aufenthalt in Prag, den er sich mit Musik und Botanik angenehm machte — seine Wohnung in der Chotkova Straße bei der Üjezder Kaserne und dann in der Gürtlergasse war mit kostbaren Pflanzen überfüllt —, wurde Schleicher nur durch das Verhalten einiger Gegner verbittert (wie später in Jena). Schleicher selbst hat vielleicht diese Ungunst nicht nur durch seine Geradheit und seine Strenge gegen die moralischen und Fachunvollkommenheiten seiner Kollegen und Schüler, aber auch durch seine gewisse mit Sarkasmus gemischte Zugeknöpftheit gegen die Leute, denen er nicht gänzlich trauen konnte, teilweise verursacht. Seine Gegner haben seinen Prager Aufenthalt vollkommen getrübt. Auf eine Denunziation hin stattete ihm die

Zur Erinnerung an A ugust Schleicher

65

Polizei einen — nicht ganz unerwarteten — Besuch ab und deportierte ihn nach der Konfiskation eines vor Jahren geschriebenen Briefes nach Wien. Nach einigen Monaten wurde er nach wiederholten Interventionen seiner Freunde, besonders Georg Curtius, freigelassen; aber das grobe Unrecht berührte ihn schmerzlich. Vergeblich ersuchte ihn sein Gönner, Graf Leo von Thun, der ihm durch seine Vermittlung eine Unterstützung der Wiener Akademie (600 Gulden) verschafft und damit seine Reise nach Litauen ermöglicht hatte, das Unrecht zu vergessen; der verbitterte Schleicher konnte nicht vergessen. Seine ganze Umgebung wurde ihm widerwärtig, und endlich nahm er lieber eine schlecht bezahlte Stellung eines Honorarprofessors in Jena an, als daß er in Pragbleiben wollte. Aber auch in dieser alten Universitätsstadt, die nicht so weit von seiner Geburtsstadt Meiningen und der Stadt seiner Kindheit Sonneberg auf dem südlichen Abhang des Thüringer Waldes entfernt war, blieb er seinen slawischen und litauischen Studien treu. Neben der Übersetzung der litauischen Märchen, Sprichwörter, Rätsel und Lieder 11 veröffentlichte er "Chr. Donaleitis'litauischeDichtungen", und den Überresten des ausgestorbenen Polabischen widmete er die seiner Meinung nach beste Studie "Laut- und Formenlehre der polabischen Sprache " welche erst nach seinem Tode von der Petersburger Akademie der Wissenschaften, deren korrespondierendes Mitglied er war, veröffentlicht wurde. 12 Er starb im Jahre 1868 in seinem 47. Lebensjahr an Lungenentzündung, einer Krankheit, die ihn schon einmal in Prag befallen hatte. Sein Nachfolger auf seinem Prager Lehrstuhl war vom Jahre 1860 an Alfred Ludwig. Es bleibt nur übrig, die Sprachtheorie Schleichers, soweit sie sich in seinem offenen Schreiben über die tschechische Schriftsprache widerspiegelt, zu behandeln. Es ist allgemein bekannt, daß sie nicht ohne innere Widersprüche war. Obwohl Schleicher — wie vorher Bopp — schon in seiner Habilitationsschrift über den Zetacismus die Leibnitzsche Ansicht betonte, da/3 das Wort aus der Bedeutung und der Beziehung bestehe, widmete er seine Aufmerksamkeit fast ausschließlich der Sprachform. 13 Das Kriterium der sprachlichen Vollkommenheit war ihm nicht die relative Fähigkeit der Sprache, Gedanken und Gefühle ausdrücken zu können, sondern die Schönheit der Sprache, und diese sah er einseitig in ihrer Fähigkeit, volle morphologische Formen zu bilden. An den vollen, üppigen Sprachformen vergnügte sich Schleicher wie an verschiedenen Blumen, Pflanzen und anderen Naturorganismen. Die Sprachgeschichte war für Schleicher nur Geschichte der Entartung der ursprünglichen Vollkommenheit, einer Entartung, die er für ein zwangsläufiges Ergebnis der fortschreitenden Kultur hielt, so daß sich die ursprüngliche Schönheit der Sprache nur in den kulturell rückständigen Sprachgemeinschaften erhielt. Ist aber diese Ansicht nicht im Widerspruch mit seinem Hegelianismus? Als Hegelianer mußte er nämlich glauben, daß sich alles fortwährend so entwickelt, daß die höhere Geistesstufe alle

66

General Linguistics

vorhergehenden, minder vollkommenen Stufen enthält. Dieser Widerspruch war aber nach Schleicher nur augenscheinlich, weil die Sprachentwicklung schon in der Vorgeschichte vollendet war. Der Sprachaufbau ist also älter als die Geschichte und kann darum keine Vervollkommnung, sondern nur eine, allerdings gesetzmäßige, ständige Entartung in der historischen Periode durchmachen. In diesem Sinne haben die Sprachen der historischen Periode keine wahre Geschichte, und es ist die Aufgabe der vergleichenden Sprachwissenschaft, zur Vollkommenheit der ursprünglichen Sprache vorzudringen. Die Verschiedenheiten des sprachlichen Aufbaus, welche das Verhältnis zwischen Bedeutung und Beziehung betreffen, existieren und stellen auch jetzt verschiedene Sprachen auf ungleiche Ebenen der Sprachvollkommenheit. Das Chinesische, welches gar keine Differenzierung beider kennt, steht nach Schleicher niedriger als die agglutinierenden Sprachen, in denen beide Elemente nebeneinanderstehen; diese sind aber weniger vollkommen als die semitischen und indoeuropäischen Sprachen, in denen eine höhere Worteinheit durch die Synthese beider entsteht. Die Sprachwissenschaft oder die Glottik, wie sie Schleicher nannte, ist nach seiner Theorie eine Naturwissenschaft und der Linguist ein Naturforscher, der alle Sprachorganismen durchforscht, während der Philologe nur seltsame, auserlesene Pflanzen pflegt. Wir könnten aus seiner Theorie schließen, daß zwischen den Dialekten, welche sich spontan entwickeln, und den Schriftsprachen, in denen die Eingriffe des menschlichen Willens sichtbar sind, ein unüberbrückbarer Grundunterschied sei; aber Schleicher zieht meiner Kenntnis nach keinen solchen Schluß expressis verbis. Er war kein konsequenter Denker. Seine theoretischen Auslegungen sind zwar immer klar entworfen, stellen aber nur Hauptlinien dar; alles Problematische wird zur Seite geschoben oder verschwiegen. Wir könnten auch z. B. fragen, in welchem Verhältnis die Sprache zum Sprechen in seiner Theorie steht. Man könnte aus seinen Sprachansichten folgern, daß auch das Sprechen, welches eine individuelle Aktualisation der Sprache als Norm darstellt, ein Naturprozeß ist; aber auch diesen Schluß, welcher die Sprachfunktion verunstalten würde, zieht er nicht. Im Gegenteil, in seinem Buche "Die deutsche Sprache", welches er für ein breiteres Lesepublikum schrieb, äußert er die Meinung, daß das Sprechen ein lautliches Denken ist, eine Ansicht, welche in Deutschland zur Annäherung oder sogar Vereinigung von Sprachwissenschaft und Psychologie führte und welche sich mit seiner naturwissenschaftlichen Konzeption nicht verträgt. 14 Der innere Widerspruch seiner Sprachtheorie spiegelt sich auch in dem offenen Schreiben wider. Die ganze komplizierte funktionelle Problematik der Schriftsprache scheint er zu übersehen und. beachtet auch nicht das Moment der langen historischen Tradition der tschechischen Schriftsprache. Es ist selbstverständlich, daß das Tschechische nicht im geringsten bereichert würde, wenn es ältere vollere Formen für die in der Schriftsprache usuellen und durch die Tradition hergebrachten Formen

Zur Erinnerung an A ugust Schleicher

67

einsetzte oder wenn es sogar das alttschechische harte vom dentalen 1_ (wie heute noch das Polnische oder Russische) unterscheiden wollte. Die Schriftsprache kann natürlich für ihre Zwecke neue Worte und Formen allerlei Ursprungs annehmen (ein interessanter Beleg ist die estnische Schriftsprache, die in den Jahren 1917—1926 eine große Zahl gänzlich künstlicher Stämme übernommen hat); aber alle diese neuen Elemente müssen eine funktionelle Berechtigung haben. Das Büchlein Schleichers hatte und konnte keine Wirkung auf Tschechen haben: weder VaniCek noch ein anderer tschechischer Freund Schleichers hat es je erwähnt, und es ist wahrscheinlich, daß es von ihnen für eine Spielerei eines ausländischen sympathischen jungen Gelehrten gehalten wurde. Die kleine Schrift O spisovnej cestiné ist nicht das einzige offene Schreiben August Schleichers. Vierzehn Jahre später hat er mit einem solchen Schreiben seinen Kollegen an der Jenaer Universität Ernst Haeckel überrascht. Dieses, Schreiben, das in mehreren Ausgaben erschien, ist die bekannte "Darwinsche Theorie und die Sprachwissenschaft ",1S in welcher er die Darwinschen biologischen Termini Gruppe, Art, Abart und Variation und den struggle for life auf das "Leben" der Sprache unmittelbar übertrug. Die Sprache war nunmehr für Schleicher ein biologisches Phänomen, das die Rassen deutlicher kennzeichnete als z. B. ihre Hautfarbe oder Schädellänge; in seinem Vortrag "Über die Bedeutung der Sprache für die Naturgeschichte des Menschen "1β verlegte er den Ursprung der Sprache im Sinne Hegels in die Entwicklungsperiode, die zwischen der Periode der körperlichen Entwicklung und der historischen Periode der Menschheit liegt. Alle diese Ansichten, welche schon von H. Steinthal und anderen Sprachforschern (in Böhmen von Martin Hattala) bestritten wurden, sind jetzt nur vom historischen Interesse. Denn das Bestreben, die Sprache vom biologischen Standpunkt aus zu behandeln, ist ebenso ein Irrtum, wie der Versuch, sie als einen Überbau der wirtschaftlichen Basis zu betrachten. Jede Sprache oder Sprachstufe ist ein System objektiver Tatsachen sui generis und kann nie als eine physiologische, biologische, psychologische oder logische Erscheinung betrachtet werden. Man kann auch keine einzige Sprache oder Sprachstufe als einen Maßstab der Vollkommenheit für andere Sprachen oder Sprachstufen nehmen. Die Bedeutung Schleichers für die Sprachforschung beruht nicht in seiner linguistischen Metaphysik, sondern in seiner positiven linguistischen Forschung, die er jedoch durch die Philosophie unterbauen wollte. 17 Man kann nicht sagen, daß er diachronische Lautgesetze im Sinne der strengen junggrammatischen Methode hätte entdecken können, aber der linguistischen Tatsachen wußte er sich in aller Vollständigkeit und Genauigkeit leicht zu bemächtigen. Er war darum ein idealer Sammler sprachlichen Materials, das er klar und übersichtlich zu bearbeiten verstand. Als ein unermüdlicher Forscher und Lehrer, zog er auch in Prag eifrige Schüler an sich und hatte auch einen feinen

68

General Linguistics

Sinn für die L ü c k e n u n d Forderungen der praktischen E n t w i c k l u n g der S p r a c h w i s s e n s c h a f t . A b e r wir g e d e n k e n seiner auch als Gründer der w i s s e n s c h a f t l i c h e n d e u t s c h e n Slawistik und als ersten Vertreter der vergleichenden Sprachwissenschaft an der Prager Universität. Die Grundlage, die er zur d e u t s c h e n Slawistik gelegt h a t , entsprang seinen t i e f e n S y m p a t h i e n zu d e n slawischen V ö l k e r n und b e s o n d e r s z u m t s c h e c h i s c h e n V o l k e , nicht d e m Streben n a c h den sprachlichen V o r a u s s e t z u n g e n für eine politische E x p a n s i o n nach Osten. Seine rein w i s s e n s c h a f t l i c h e Persönlichkeit erfüllt uns m i t der H o f f n u n g , daß a u c h in Z u k u n f t die Gelehrten beider europäischer N a t i o n e n z u s a m m e n k o m m e n k ö n n e n , u m im Sinne der n e u e n Zeit für die w e i t e r e E n t w i c k l u n g Europas z u s a m m e n z u a r b e i t e n .

ANMERKUNGEN 1. Schleicher, A. (1852). Die Formenlehre der kirchenslawischen Grammatik, erklärend und vergleichend dargestellt. Bonn, Wien, Prag. 2. Schleicher, A. (1849). "O infinitive a supinu w jazyku slowanském". Uber den Infinitiv und das Supinum in der slawischen Sprache. Casopis Ceského Museo (CCM). [Zeitschrift des böhmischen Museums]. 23, Bd. 3, 1 5 3 - 5 7 . Die deutsche Version (s.o., Titel) erschien in: Schleicher, A. (1850). Die Sprachen Europas in systematischer Übersicht. Bonn. S. 2 5 6 - 2 5 9 . Vgl. auch (7). 3. Schleicher, A. (1853). "O jazyku litevském zvlásté ohledem na slovansky". Über die litauische Sprache mit besonderer Berücksichtigung des Slawischen. CCM [Zeitschrift des böhmischen Museums]. Jgg. 2: 3 2 0 - 3 3 4 . 4. Schleicher, A. (1855). "Das Futurum im Deutschen und Slawischen". Zeitschrift für vergleichende Sprachforschung. 4: 1 8 7 - 9 7 . 5. Schleicher, A. (1852). "Briefe an den Sekretär über die Erfolge einer nach Litauen unternommenen wissenschaftlichen Reise". Sitzungsberichte der Kaiserl. Akademie der Wissenschaften. Wien. 9: 5 24 - 5 5 8. Schleicher, A. (1850 u. 1851). "Über böhmische Grammatik". Zeitschrift ßrdie österreichischen Gymnasien. 1: 7 2 5 - 7 5 6 ; 2: 2 6 9 - 3 1 0 . Schleicher, A. (1851). "Über die wechselseitige Einwirkung von Böhmisch und Deutsch". Archiv für das Studium der neueren Sprachen und Literaturen. 9 (Jg. 6): 3 8 - 4 2 . 6. Schleicher, A. (Übersetzer zusammen mit F. X. Sohaj) (1851). " P o t o p a " [Sintflut] [Teile aus dem Sanskritepos Mahabharata zum ersten Mal ins Tschechische übersetzt]. CCM [Zeitschrift des böhmischen Museums]. 25, Bd. 1, 1 1 7 - 1 2 0 . Schleicher, A. (Übersetzer) (1851). "Nal a Damajanti". CCM [Zeitschrift des böhmischen Museums], 25, Bd. 1, 1 2 1 - 1 2 3 , Bd. 2, 8 5 - 1 0 1 , Bd. 3 , 6 2 - 8 4 , Bd. 4, 6 2 - 9 3 . 7. Schleicher, A. (1850). Die Sprachen Europas in systematischer Übersicht. (Linguistische Untersuchungen 2). Bonn. Schleicher, A. ( 1 8 6 1 - 1 8 6 2 ) . Compendium der vergleichenden Grammatik der indogermanischen Sprachen. W e i m a r - B ö h l a u . 3 1 8 6 6 (berichtigt, vermehrt und umgearbeitet); 3 1870; 4 1876.

Zur Erinnerung an August Schleicher

69

8. Schleicher, A. (1856-1857). Handbuch der litauischen Sprache. 1. Litauische Grammatik; 2. Litauisches Lesebuch und Glossar. Prag. 9. Zubaty, J. (1901). "Zu Schleichers litauischen Studien" Vêstnik královské ceské spolecnosti nauk. [Mitteilungen der Königlichen böhmischen Gesellschaft der Wissenschaften]. 10. Vanicek, A. (1869). "Erinnerungen an Prof. Dr. August Schleicher in Prag". Bohemia. Vgl. auch Ed. Novotny (Januar 1869) "August Schleicher", Svétozor, 31-35. 11. Schleicher, A. (1865). Chr. Donaleitis' litauische Dichtungen. St. Petersburg. 1865. 12. Schleicher, A. (1871). Laut und Formenlehre der polabischen Sprache. St. Petersburg (Leningrad) - Riga. 13. Verbürg, P.A. (1950). "The Background to the Linguistic Concepts of Bopp". Lingua. 2: 4 3 8 - 6 8 . 14. Schleicher, A. (1860). Die Deutsche Sprache. Stuttgart 1 1860,' 2 1869 (verbessert und vermehrt, herausgegeben von J. Schmidt); 3 1874, "1879; 5 1888. [ Cf. also: Trnka, B. (1941). 'K otázce stylu' [The Problem of Style]. SaS 7: 6 1 - 7 2 . (The English translation is included in this volume, Section V, p. 364. Trnka explains there more explicitly some of Schleicher's views. V.F.) 15. Schleicher, A. (1863). Die Darwinsche Theorie und die Sprachwissenschaft. Offenes Schreiben an Herrn Dr. Ernst Häckel, a. o. Professor der Zoologie und Direktor des zoologischen Museums an der Universität Jena. Weimar. Englische Übersetzung "Darwinism tested by the science of language" (transi, by V. W. Bibbers). London, 1869. 16. Schleicher, A. (1865). Über die Bedeutung der Sprache für die Naturgeschichte des Menschen. Weimar. 17. In der Vorrede zu seinem Handbuch der litauischen Sprache (8) hat er selbst positivistisch klingende Worte geschrieben: "Beide werke (d. h., Obermeyers litauische Grammatik 1791 und Curtius' griechische Schulgrammatik) haben im großen und ganzen dieselbe behandlung und anordnung des stoffes und zwar diejenige, welche nach meiner meinung die einzig verständige und zweckdienliche ist; es wird nähmlich alles philosphische wesen ferne gehalten . . . Ich kenne nichts, was mir unerquicklicher wäre, als ein philosophisch sein sollendes wesen in der grammatik . . . allgemeinere gesichtspunkte gehören in eine grammatik wenigstens nicht, falls überhaupt die sogenannte philosophische grammatik zur zeit etwas berechtigtes wäre". Über den Hegelianismus Schleichers vgl. Streitberg, W. (1897). "Schleichers Auffassung von der Stellung der Sprachwissenschaft". Indogermanische Forschungen. 7: 360 ff.

Prague Structural Linguistics

The starting point of linguistic work as done by the Prague linguists is the assumption that its true objective is the analysis of speech utterances of all kinds, both spoken and written. This subject matter of linguistics, as any other raw material of physical, psychological and other phenomena, can be grasped and comprehended only in terms of the verifiable laws which govern them. In contradistinction to the nomothetic, mechanical operation of natural laws, the validity of linguistic laws is normothetic, restricted to definite periods of time and to definite bodies of utterances. If formulated and presented in grammar, they may exert a stabilising influence upon the speech of the community by strengthening the stability and uniformity of its linguistic norms, as prescribed by the grammarians. The normothetic character of linguistic laws does not exclude, of course, the validity of some of them for several languages or even for all languages of the world in all their historically accessible periods of development. Despite many differences all languages must have some laws in common (e.g., the general law of minimal phonological contrast, according to which some consonantal clusters and vocalic combinations are excluded from all languages of the world), and one of the true goals of linguistics is to discover and formulate them. Structuralism may be defined as the trend of linguistics which is concerned with analysing relationships between the segments of a language, conceived as a hierarchically arranged whole. The question may be asked whether the

Originally published in PP 1, 1958, pp. 38-40. The authorship was explained in a footnote: "The article by B. Trnka and others is an English adaptation of the Russian paper published under the title 'Kdiskussii po voprosam strukturalizma' [Discussion of Problems of Structuralism] in Voprosy jazkoznan 'ija, Moskwa, No. 3, 1957, pp. 4 4 - 5 2 . It represents the view of a group of Prague linguists associated with a special section of the Krüh modernich filologi» (Circle of Modern Philologists] : B. Trnka, J. Vachek, P. Trost, St. Lyer, V. Polák, O. Duchácek, J. Krámsky, J. Nosek, M. Rensky, V. Hoíejsí, Ζ. Wittoch, L. Dusková." Reprinted in PSRL 468-485. Spanish translation 'La lingüistica structural de Circulo de Praga. In Joan A. Argrate, ed., (1972) El circulo de Praga, Barcelona, pp. 13-29. (It was generally understood that Β. Trnka was the leading author of this presentation of Prague School concepts. V.F.)

Prague Structural

Linguistics

71

segments or the relationships are primary, but this problem cannot be solved at the present stage of our knowledge, at least not by linguists alone. It is clear, however, that both relators and relations are co-existing and correlated entities which cannot appear separately from each other. The relations of a segment are recognised by its properties, and as every segment is constituted by its properties, the elimination of these results in the elimination of the segment itself, not in the discovery of a "substratum" — a concept which modern science makes superfluous. In other words, the structural linguist conceives linguistic reality as a system of sign events, i.e., as a system of linguistic correlates to extra-linguistic reality. The word "structuralism" is used to designate various trends in modern linguistics which came into existence between both world wars, but, apart from the school of Geneva, those associated with the Cercle Linguistique de Prague, the Cercle Linguistique de Copenhagen and the name of Leonard Bloomfield are regarded as the most typical. From the historical viewpoint, these three currents of structural linguistics have at least two features in common: divergence from the Neogrammarian methods which tended to the psychologisation and atomisation of linguistic reality, and a tendency to establish linguistics, looked upon by the older school as a conglomerate of psychology, physiology, sociology and other disciplines, as an independent science based on the concept of linguistic sign. Otherwise they differ considerably from one another in their principles and procedures, and it is therefore advisable to use a special designation for each of them, viz. functional linguistics (V. Mathesius's term) for the linguistic school of Prague, glossematics for Hjelmslevian linguistics, and descriptive linguistics for the Bloomfieldian trend. We cannot devote our attention here to other outstanding currents of structural linguistics deviating from the above-mentioned schools in different ways, and the following remarks will thus be mainly restricted to these three. Hjelmslev's glossematics introduces into linguistics the deductive method of the algebraic calculus and declares itself to be independent of any linguistic reality. His theory does not purport to be a system of hypotheses which would be found true or untrue by reference to factual linguistic evidence, but it tends to serve in analysing a given "text" by means of a strategy of assumptions which should be as few and as general as possible in order to fulfil conditions of applicability to the greatest number of linguistic data. Let us quote his own words: "This calculus, which is deduced from the established definition independently of all experience, provides the tools for describing or comprehending a given text and the language on which it is constructed. Linguistic theory cannot be verified (confirmed or invalidated) by reference to such existing texts and languages. It can be controlled only by tests to show whether the calculation is self-consistent and exhaustive". It must be

72

General

Linguistics

pointed out that the correctness o f such a linguistic theory rests not only on the truth o f calculation, but also on the veracity of the general linguistic assumptions which enter into the calculus. The latter in Hjelmslev's calculus are, however, neither self-evident nor acceptable to all structural linguists (cf. his sharply-drawn dichotomy text—system, contents—expression, the forms of contents—the forms of expressions; his view of sign as representing both "contents" and "expression"), and his logically consistent and conceptually well-arranged theory appears to be divorced from, and inadequate for, linguistic reality. The Prague linguists have not been influenced by Hjelmslev's glossematics, and especially strong reservations are made by them concerning his theory of the phoneme as a mere "taxeme", the identity of which is thought b y Hjelmslev as consisting only in the identity of its distribution in words. Hjelmslev seems to regard both relevant (or, distinctive) and irrelevant features o f a phoneme as a "substance", and while mixing the functional features o f a phoneme with its functionless elements, he builds up an artificial wall between the sound and the phoneme. The Prague School takes into consideration not only the distribution of phonemes in words, but also their relevant features, the superposition of which determines the identity of a particular phoneme. Thus the English sounds \ph\ and [p] in paper are phonetic implementations of the phoneme /p/, because both of them consist o f the superposition of the same relevant (distinctive) features, i.e., bilabiality, plosion, oralness and voicelessness, aspiration being an irrelevant phenomenon determined mechanically by stress. On the other hand, Old Indian Iphl is a phoneme, not a positional variant o f [ρ], because its aspiration is not determined by its environment or its position. If two phonemes (e.g. /h/ and /17/ in English) are in complementary distribution with each other and there is no environment in which both of them occur, their ever-present and potential capacity for differentiating words (cf. pen-fen-ten-den) remains unutilised, so that the only test o f their phonemic status is their differences of make-up in terms o f relevant features. Strictly speaking, the true function o f phonemes is not keeping the meanings of words (or morphemes) from each other (cf. [nait] night-knight, [frendz] friend's-friends), but only distinguishing phonemes between each other. Various trends of descriptive linguistics appear to be in agreement with the linguistic theory o f the Prague group on some points, but owing to their behaviouristic basis they differ from the latter both in terminology, definitions and details of the procedure of linguistic analysis. Both schools agree in rejecting the psychological approach to phonology and in regarding the phoneme as a unit of the phonological plane of language, but while the Prague group stresses the analysis o f the phoneme into the relevant features which constitute it, the Bloomfieldians (Bloch, Trager, Harris, Hockett) seem to lay the emphasis on its distributional features in words or in utterances. As

Prague Structural Linguistics

73

to the semantic criteria, which most American linguists exclude from their definitions of the phoneme, Prague phonologists lay stress on the capacity of phonemes to distinguish words and morphemes (cf. Russian tocka-docka, burit'.buril) and consequently on the morpheme and word boundaries as important factors, the disregard of which might lead to unwarrantable conclusions, e.g., in the phonological evaluation of affricates, in the formulations of neutralisation rules and in ascertaining phonological foreignisms adopted by a given language. On the other hand, Prague linguists are aware of the fact that phonemes as such have no meaning — the word hand is something different from the mere sequence of the phonemes /Λ/, / grevis), whereas the analogical process is characterised by the passivity of the old form which does not play any role in it, and, later on, tends to come into disuse. It may be objected against Hermann's psychological conception of both linguistic processes that some blends did discard the old words (cf. rendere and grevis in Vulgar Latin) if the meaning of the blend and one of the old pair of words is the same or at least very similar, and reversely, if they come to be associated with different meanings. Both analogy and blending can be used for the formation of new lexical units. The main difference between blending and analogy consists therefore in the simple fact that the former process involves only two closely associated words, whereas the latter is based on proportional models. 12. The first linguist concerned with the problem of the regularity of analogical innovations was perhaps: Löwe, R. (1891). "Die Ausnahmslosigkeit sämtlicher Spracherneuerungen". Zeitschrift des Vereins für Volkskunde 1: 56 ff. 13. Horn, W. (1923). Sprachkörper und Sprachfunktion. Halle (Saale).

SECTION TWO

Synchronic Phonology

General Laws of Phonemic Combinations

In the paper "Zur allgemeinen Theorie der phonologischen Vokalsysteme" Professor N.S. Trubetzkoy formulated phonological laws which apply to all linguistic systems.1 According to these general laws, the phonological polytony necessarily involves the presence of the phonological quantity of vowels, which presupposes the existence of the timbre oppositions of vowel phonemes, based on the fundamental vocalic opposition of sonority. These laws stating the hierarchy of phonological factors in languages may be reduced, as Professor R. Jakobson shows in one of his treatises,2 to either of these logical formulas: If there is a, ft also exists. If a exists, b is missing. All linguistic laws that cannot stand this test, e.g. M. Grammont's laws of dissimilation, are therefore only of a limited validity. Trubetzkoy's laws refer to phonological relations which are abstracted from the given linguistic systems. Now the question arises if the combinations of phonemes in the same morpheme are also regulated by rules of general applicability or not. This question may be answered in the affirmative. In examining the combination and iteration of English phonemes in monomorphemic words consisting of not more than two syllables, the present writer tried to give, according to the method suggested by Trubetzkoy, the special rules restricting phonemic combinations in English.3 One of those rules, by which it is stated that phonemes differentiated by a mark of correlation never combine in the same morpheme, applies to all languages and may be regarded — at least provisionally — as a general law admitting of no exceptions. Thus we do not find such consonantal combinations as \pf, fp, tO, 6t] and \pb, td, kg], etc. in those languages where the correlations of plosion or voice exist, e.g., in Anglo—Saxon and Modern English. Similarly the combinations \ph-p, p-ph, th-t, t-th], etc. do not occur in Sanscrit and Old Greek,4 where aspiration is a correlative mark. In German, Czech, Magyar

Originally published in TCLP 6, 1936, pp. 5 7 - 6 2 . Reprinted in PSRL, pp. 294-300. (Partly revised.)

114

Synchronic Phonology

and other languages in which the quantity of vowels is phonological, the combinations of two vowels differentiated only by the correlative mark of quantity (e.g. [ii:], [uu:], etc.) are non-existent. The same is true in regard to the combinations of stressed and unstressed vowels in Russian. In Modern French the nasalised vowels are never combined with the corresponding oral ones. Nor can palatalisation, which is an important correlative mark of consonants in Russian, be the only mark of differentiation between two neighbouring phonemes and consequently such combinations as [ii', dd', t't, nn ', η 'η], etc. do not occur in this language. The assumption of this and another correlation, which may be called post-alveolar, for Czech (cf. [t : f,d : d,n : ή] and [s :s,z :z,c :c,r :r\) would also account for the non-occurrence of the monomorphemic combinations tt, tt, nn, nit; ss, ss, cc, cc, rr, etc. This rule (which may be expressed by the logical formula; if there is p, it must be neither followed nor preceded by pl in the same morpheme) may be called the law of the minimal phonological contrast. The combinations of the marked phoneme with the corresponding unmarked one are below the limit of the minimal contrast unless they occur at the suture of two morphemes where they are tolerated (cf. bathed [beidd] in English). If, however, the morphological suture is shifted from its original position or entirely disappears (cf. OE dœt 9e > ME datte) we see that the law of the minimal contrast comes into force automatically through the assimilation (or dissimilation) of both sounds. On the other hand, the existence of the combinations of two phonemes differentiated only by one mark of correlation, signalises them as members of two different morphemes (cf. let'-te, sed'—te in Czech). 5 If this is true, the law of the minimal phonological contrast is an important negative test of the correlative character of phonemes. According to Trubetzkoy's definition, 6 the phonological correlation is constituted by at least two pairs of phonemes not connected by another correlative mark. Thus the plosion correlation in English is guaranteed by the fact that there are at least two pairs of phonemes \p¡f\ and [f/0] which are not otherwise correlated to each other. On the same degree of relationship, therefore, we could place the nasal opposition o{m :p (or b),n : t (d), and 17 : k (g) in English, because none of these pairs has another correlative mark in common. If we put, however, both series of phonemes to our test of the minimal phonological contrast, we may see that the existence of such combinations as mp{mb), nt{nd), vk(r¡g) shows a different degree of relationship in the latter series, the phonological difference between m and p(b), η and t(d), η and k(g) being so wide as to allow the members of the same pair of phonological opposition to combine with each other. It is necessary, therefore, to distinguish this kind of phonemic relationship, which may be called

General Laws of Phonemic Combinations

115

parallelism, from that of correlation which represents such a close affinity that it deprives the members of the same pair of the capacity of being contrasted, as individual phonemes, in monomorphemic combinations. Phonemic groups consisting of parallel phonemes constitute therefore the minimal contrast occurring in languages, whereas the maximal degree is represented by the combinations of vowel + consonant. In Germanic, Slavonic and other languages intermediary degrees exist in addition to the minimal and maximal contrast (cf. such combinations as sp, st, sk, ks, si, pr, pi, tr, ti, pt, db, kt in Czech), which are totally missing in Chinese and some other linguistic systems. The observation of the combinative capacity of phonemes suggests the classification of so-called disjunct phonemes which have hitherto defied phonological analysis. At the present stage of phonological investigation, disjunct phonemes are defined to be all those that do not take part in correlations, although it is not doubted that the phonological relationship between phonemes \p, t, is different from e.g. [r] : [/]. It is true that even the former phonemes are neither parallel nor correlative to each other, but the process of their articulation is analogical and takes place symmetrically on the common basis of their relevant features, so that it results in an acoustic similarity. With regard to this fact we are entitled to speak provisionally of symmetrical phonemes on the one hand, and of purely disjunct ones on the other. The phonological symmetry is constituted by the existence of at least three disjunct (i.e. uncorrelated and unparalleled) phonemes of analogical articulation which may be brought into the proportion ρ : t= t : k. Thus the phonemes Ipl, /1\ and ¡k¡ are symmetrical both in English and Russian, whereas the nasals, which form the series of phonemes /ml, /«/, /17/ in English and m/rfi, njñ in Russian, are symmetrical only in the former language, th and h being correlated to m and n, respectively, in the latter. It goes without saying that symmetrical phonemes may take part either in parallelisms or ' correlations, so that symmetry represents a basic relationship underlying those of parallelism and correlation. For example, in Czech two of the series of symmetrical phonemes p, v, 7 t and k are the members of the nasal parallelism (cf. pi/m, ti/η) and voice correlation (cf. p¡b, t¡d), while the correlation of palatalisation refers to the phoneme t and its corresponding d and η (cf. t/t\ d/d\ n¡nr). The phonemes ν and k are outside any correlation or parallelism in native words, |f\ and [g] occurring as phonemes only in words of foreign, or onomatopoetic origin. It is the symmetry of phonemes which helps to account for the fact that the Czech language found it easy enough to adopt these two sounds, whereas the phonemic function /η/ could not be transferred into Czech on account of the non-correlative character of nasals in this linguistic system.8 Let us return to the main subject of this paper, the combination of phonemes. It may be assumed as a working hypothesis that the combination

116

Synchronic

Phonology

of phonemes as well as their position in words, are regulated by strict special rules, the cogency of which is no less than that of other phonological factors. 9 Unlike some phonological laws (e.g. that of voice neutralisation in Czech) these rules, however, do not extend to foreign words which may contain monomorphemic combinations not occurring in native ones, at least not in the same position. For example, in English the monomorphemic combinations of symmetrical phonemes p i and kt are found only in words of foreign, especially Latin, origin, but their existence in the language of adoption is undoubtedly due to the fact that the dimorphemic combinations pt, kt, bd, gd, etc. (e.g. in dipped [dipt], licked [likt], sobbed [sobd], begged [begd] ) occur in the same position as do the monomorphemic pt and kt in foreign words (e.g. sept, sect). If there were not the native dimorphemic combinations of this kind in English, we should probably expect that English speakers would likely pronounce except, sect, etc. always without t, simplifying the consonantal combinations in a similar way as they actually do in the positions which are without any analogy in native words (cf. ptarmigan [ita.migan], mnemonic [ni'monic]). The adoption of unusual combinations (or even foreign simple phonemes) is made possible, as we may see, by the actualisation of the potential functional elements of the language, which might not otherwise be utilised. Thus the monomorphemic character of the combinations pt, kt and mn, which is the foreign feature of the words mentioned, is within the range of the possibilities of the English linguistic system where new monomorphemic combinations may always come into existence by the loss of morphological suture. Similarly the functional value of the Czech variant [#] in words of foreign origin and character is in conformity with the voice correlation of other symmetrical phonemes whereas the neutralisation of this correlation is necessarily without exception in both native and foreign words. The comparison of monomorphemic combinations in various languages entitles us to risk formulating the general tendencies of consonantal combinations as follows: ( 1 ) The existence of phonologically parallel combinations (e.g. mb, nd, •r¡g in English) presupposes the presence of the maximal degree of phonemic contrast, i.e. that of vowel + consonant, or consonant + vowel. (2) The combinations of symmetrical phonemes (e.g. pt, kt, mn in Czech) presupposes the existence of the combinations of disjunct phonemes (e.g. pr, pi, ps, pi, br, bl, etc. in Czech). (3) The occurrence of consonantal combinations at the beginning, or end, of words involves the possibility of their use between vowels, but not vice versa (cf. pi, kl, kw as against mp, mb,lp, lb in English).

General Laws of Phonemic

Combinations

117

A f e w words m a y be said about the question relative to the number o f phonemes o f which the word consists. Although languages differ a good deal in this respect and interesting differences may be also stated in regard t o the various grammatical classes o f words, one feature probably c o m m o n to all languages may be stated, viz. that only vowel phonemes can constitute independent words by themselves, without any combination with other phonemes (cf. E. eye [ai], owe [ou], Czech a, i). This rule is so general that it seems worthless for phonological analysis, but it is not entirely without importance, because it reveals the specific character o f vowels as against consonants which can form independent words only in combination with vowels or other consonants (cf. tm [thorn] ,pln [full] ,prst [finger] in Czech). NOTES 1. Trubetzkoy, N. (1929). "Zur allgemeinen Theorie der phonologischen Vokalsysteme", RCLP 1: 3 9 - 6 7 . Reprinted in PSRL, pp. 108-142. 2. Jakobson, R. (1929). "Remarques sur l'évolution phonologique du russe". TCLP 2. Reprinted in SW 1: 7 - 1 1 6 . 3. Trnka, B. (1935). A Phonological Analysis of Present-day Standard English. Prague Studies in English 5. Prague, cf. p. 30 ff. Revised new edition, (1966). Kanekiyo, T. and Koizumi, T., eds., Tokyo. Cf. p. 36 ff. 4. p-ph in such Greek words as Sappho may be accounted for as the geminated, or long, phoneme Iph/. 5. Jakobson, R. (1932). "Zur Struktur des russischen Verbums". In Charisteria Guilelmo Mathesio. . . oblata. Prague, p. 7 4 - 8 4 . Reprinted in PSRL, pp. 347-357, and in SW 2: 3 - 1 5 . 6. Trubetzkoy, N. (1931). "Die phonologischen Systeme". TCLP A: 9 6 - 1 1 6 . 7. [v] may be regarded as a voiced labio-dental explosive in Czech before voiced consonants excepting [b] and [m], 8. The assumption of parallelism and symmetry as special cases of phonological opposition leads us to the conclusion that the sonority oppositions of vowels (Öffnungsgradgegensätze in Trubetzkoy's terminology) may be also interpreted as a series of symmetrical phonemes (e.g. i-e-ae and U-Ο-Λin English). The timbre oppositions (Eigentongegensätze) are, on the other hand, either parallelisms (cf. '//«, ello, œlhl in English; i'//u, ello, in Czech) or form the groups of phonemes interlinked by two correlations, if there are at least three terms of these oppositions. For example, the German system of short vowel phonemes i-ii-u, e-ö—o, a consists of the labial and velar correlations (i : ti, e : ö, ü-B, ö : o) in two triangular groups:

This interpretation, based on Trubetzkoy's theory of consonantal phonemes,

118

Synchronic Phonology

cf. Note 6, tries to remove an inconsistency in his conception of phonological correlation. Whereas he assumes that the correlative series consists of the pairs of phonemes differentiated by a mark of correlation, his theory of vowel phonemes, cf. note 1, involves the assumption that the triads of vocalic phonemes of the same correlation may exist in addition to conelative pairs (cf. ί : ü : u, e-ö : o in German). What would then be the basic phonemes (aichiphoneme) according to his theory? - The analysis of the vowel oppositions given here is in agreement with the fact that the phonemic combinations iti, ïm, uü, eö, öe etc. do not exist, whereas ui, iu, eo etc. may occur, in the same language. 9. The importance of phoneme combinations for the structural analysis of languages was emphasised by: Mathesius, V. (1929). "La structure phonologique du lexique du tchèque moderne". TCLP 1: 67-84. Reprinted inPSRL, pp. 156-182.

O n The Combinatory Variants and Neutralisations of Phonemes

One of the most notable contributions of phonemics to general linguistics is a clear distinction between variants and phonemes. Based upon the notion of phonemic oppositions as fundamental units o f the linguistic system, phonemics draws a definite line between sounds which may be used to distinguish words in a given language, and those which are not. According to the "Projet de terminologie phonologique standardisée" 1 sound in the given style of a language may be either a term of phonemic opposition, i.e. a phoneme, capable of keeping words apart, or a member o f the combinatory variation of such a term. For example, the velar nasal i? is a phoneme in English, because its function is to distinguish the word king [fc/η] from kin [kin] in the same style of the language, whereas in Czech it is a combinatory variant of the phoneme /«/ exclusively occurring before k and g where the dental η is never pronounced in normal speech. Distinction between phonemes and combinatory variants, far from being a whim of modern linguistics, is a matter of linguistic reality. It is not only confirmed by the linguistic consciousness o f speakers which comes to the fore in the systems o f spelling, but also by the fact that one variant of a phoneme is excluded from influence upon another variant o f the same phoneme by way o f morphological analogy. 2 If for example, η and η are two variants of the phoneme /«/ in Czech and other Slavonic languages, the dental nasal cannot discard the velar 77 in the inflected forms o f the same word, and vice versa. Thus strànka [page] has 17 in all cases except the dative and locative strànce and the plural genitive strânek, both of which have n, but in spite of seven forms with η (stránk-y,-u,-ou,-ám,-ách,-ami) there is not a single instance of the analogical transference of η to the forms strànce and strânek. Similarly in the same language the mechanical regulation o f / - v , k-g, c[is] -dz, c[f/]-dz [¿3 ] , which is invariably determined by the position o f the sounds

Originally published in the Proceedings of the Third International Congress of Phonetic

Sciences. Ghent, Phonetic Laboratory, 1938, pp. 23-30. (Slightly abridged.)

120

Synchronic

Phonology

in words, is never disturbed by morphological analogy. The absence of analogy formation in Czech is the more striking, since the levelling of phonemic alternations of this kind is otherwise very common, cf. [svadeb\ (pi. gen.) on the analogy of svadba [wedding], instead of svateb\krezeb (pi. gen.) on the analogy of krezba [drawing] (spelt kresba, kreseb). This remarkable fact, of which many more examples could be quoted from Slavonic languages owing to their developed system of declension and conjugation, can be also illustrated by instances from the older stages of Germanic languages in which the original morphological system had been better preserved than it is now. In most English dialects of the period lying between the tenth and fifteenth centuries the voiceless spirants/,/» and s occurred at the beginning of prefixes or stem syllables, at the end of words and before voiceless consonants, whereas the corresponding voiced spirants ν, 9, and ζ were limited to positions between voiced sounds, both vowels and consonants. The consequence was that the spirants/and v,p and 9, s and ζ interchanged mechanically in the forms of the same words according to their positions without any interference by morphological analogy (cf. wtdf : wulfes [-ves], dëap : dèapes [-¡fes], hüs : hüses [-zes] ). If we realise the fact that interference with the variants of the same phoneme by morphological analogy in the forms of the same word (or in the different words of the same morphological family) would mean their mutation into phonemes, we are entitled to say that no phonemes can come into existence by analogy formation, and consequently that their number in the development of a linguistic system can neither be increased nor decreased in this way. It is only due to phonemic mutations by which variants change into phonemes and phonemes become variants. The statement of this fact is a material aid for linguists in establishing a more precise relative chronology of morphological formations in the historical development of a language than was possible up to now. If, for example, we know that s and ζ were combinatory variants of the same phoneme in an early period of Old Germanic — ζ occurring after an unstressed vowel (+ a consonant) and s in all other positions — we may safely conclude that the levelling analogy in such forms as those reflected in Old English rison (instead of *riron from *riz-) could not take place at the stage of the development of Old Germanic, but only at the later one when the variant ζ was phonologised owing to the shift of stress to the first stem or prefix syllable of words. Similarly we are able to state with cogency of logical reasoning that in English such forms as beliefs (instead of believes), he rose (-z, instead of -s), wolfish instead of wolvish, elfish instead of elvish, etc. were capable of existence only after the phonologisation of voiced spirants, which took place about 1400 in consequence of the loss of the final -e. On the other hand, an ascertained fact of analogy formation is a safe guarantee for the supposition that the sound

On The Combinatory

Variants and Neutralisations of Phonemes

121

discarded in this way was not a combinatory variant of that by which it was displaced. Thus Old English vowels œ and a must be both valued as phonemes and not as variants of one phoneme, because we find safe instances of analogical transference of both sounds in the forms of the same word (cf. glades instead of glœdes, etc.). More instances could be easily met with in the historical development of languages, but these few illustrations, I believe, are sufficient to show the importance of functional difference between phonemes and combinatory variants for historical grammar. What is the phonemic cause of the non-intervention of morphological analogy? Before we try to account for this remarkable fact which has hitherto escaped the notice of linguists, we must first point out that the phonetic realisation of combinatory variants is dependent on their phonetic neighbourhood, that is to say, combinatory variants are mostly realised byproducts of a complete assimilation of the place or other relevant features of the articulatory complex representing a phoneme. For example, in Czech the velar nasal η, which occurs only before k and g as the secondary variant of n, is to be regarded, even from the static point of view, as a result of assimilation of the articulatory place of the primary variant n. IH the same language the secondary variants g, dz and dz of the phonemes k, c and c, respectively, appear before the marked members of voice opposition, cf. kdo \gdo] [who], leckdo [ledzgdo] [anybody], lécba [le.dzba] [cure, healing]. In Estonian the voiceless variety of h occurs initially, finally and in the neighbourhood of voiceless sounds, whereas the voiced ungeminated h appears between voiced sounds. The voice of Old English spirants is also due to bilateral assimilation with the neighbouring voiced sounds, e.g.,/?// lœfan, lœfde. In the early stage of development of Old Germanic the variants b, d and g of the phonemes /v/, /&/ and ¡yI were restricted to positions after nasals which exercised an assimilative influence upon the voiced spirants in regard to occlusion. The same occurs in Modern Greek. In all these cases the assimilation of either place or some other relevant feature of articulation is a complete one from the viewpoint of the given linguistic system and this is an important characteristic of variation. If assimilation is incomplete, that is, if it refers to a phonetic articulatory element which is not functionally relevant in a given linguistic system, e.g., voice in Danish (cf. viswise) or a partial velarisation of palatals in Czech (cf. sáñky [sa. jifci] [toboggan] as against sâné [sa:pe] [sledge], we are entitled to speak only of what may be called combinatory modification of a phoneme. What are the structural causes of the rise of variants? It is evident that the phonetic assimilation by which combinatory variants are realised, cannot be responsible for them. Even if we abstract from differences in languages, the physiological standpoint would fail to account for the fact that assimilation did not take place under similar physiological conditions in the same language.

122

Synchronic Phonology

In Modern Czech the consonant /, which may be said to be articulated in the same place as η or d, is not subject to the assimilation of the articulatory place to the following k, as is the case with n, although the velar / might be expected by phoneticians in this position. Both phonemes, η and /, have therefore a different power of resistance to physiological assimilation in Modern Czech, and this cannot be accounted for as due to phonetic reasons. The untenability of the phonetic viewpoint is also shown by the different resistance of η and « [ p ] against place assimilation to the following k, because we might expect the complete place assimilation of the latter consonant rather than that of the former from this point of view. Functional factors which enable the rise of the phonemic variants are to be found in the dissociation of the relevant features of which the related phonemes consist. If we observe the Czech phoneme /«/ on the one hand, and /// on the other, we may see that the position of both consonants in the phonemic structure of Czech is entirely different. Whereas the former phoneme is a term of the phonemic opposition m/n, in which /«/ differs from /ml only by one relevant feature, viz. the place of articulation, the latter phoneme cannot constitute such an opposition with any other. In the former case the place of articulation can be dissociated from the other relevant features of the phoneme and the consequence is that the power of resistance of the dental nasal against physiological assimilation is weakened and its phonetic shift may occur. In the latter case the place of articulation cannot be dissociated from the other relevant features of the phoneme /// so that a complete place assimilation to that of the following velar is not possible. If therefore the velar / changed into the dental I in 15th century Czech, it was subject to this change even before velars. This theory may be confirmed by other instances. The rise of the variant / in Czech of the phoneme /v/ finds it explanation in voice neutralisation of the phonemes lb/ and /d/, which relevantly differ from /v/ only by the place of articulation. The variant g of the phoneme ¡k¡ in Czech and Dutch is due to the same structural context. The rise of voiced spirants ν and 3 as variants of / a n d p in Old English is accounted for by the fact that the voiceless explosives ρ and t, which formed opposition of plosion with the corresponding voiceless spirants \ p l f , t/p], were at the same time members of voice opposition \p/b, t/d]. Voiceless spirants were therefore less resistant to the assimilative influence of the voice of the neighbouring sounds, and new variations f-v, p-8 (and s-z) could come into existence. If we realise that variation represents a kind of reservoir from which the language may draw for its structural needs in the course of its development — as was really the case in the latter development of English voiced spirants - we may say that the triangular systems of consonant oppositions ip/b/f, t/d/p, k/g/h in Old English) are less stable in the development of languages than the quadrangular ones. 3

On The Combinatory

Variants and Neutralisations of Phonemes

123

If we see now, on the one hand, that combinatory variation between two sounds consists in a phonetic difference which is phonemic between others in the same style of the language, and on the other, that analogy which we speak about is a morphological fact, we may be able to answer the question why analogy cannot interfere with variation in the words of the same morphological family. As a realisation of morphological oppositions, analogy works with the whole forms and the whole phonemes as their constituent units, while the relevant features of sounds are entirely below its grasp. We may see it clearly when a new analogy formation supplants an older one. In Czech the marked member of voice opposition d in the word svadebni (instead of svatebni) [bridal] was transferred from another word of the same morphological family, svadba [wedding], as a whole phoneme member of the morpheme, just as ζ [3] in muzu [lean] (instead of Standard Czech mohu) supplanted h on analogy of mùzes [thou canst], müze [(he) can], etc. If this is true and if we realise that phonological mutations consist either in phonologisations or dephonologisations, 4 we may not only infer that morphological analogy cannot increase the number of phonemes, but also that it cannot oppose — directly at least — their mutations for the time during which they are valid in the development of the linguistic system, that is to say, before they are discarded by other phonemic mutations. It is therefore a task of historical phonemics to state not only the existence of the phonemic law, but also its supersession by another, so that the structural nexus of phonemic mutations may be ascertained by the inductive method. 5 If we hold that morphological analogy cannot increase the number of phonemes in a linguistic system, we do not deny that new sounds may arise by phonemic processes in the sphere of special morphological formations. For example, the long syllabic I and r in Slovak evidently owe their origin to the lengthening of stem syllables, e.g., in forming diminutives, cf. vlk [wolf] — vlcok [vl:tfok] [a little wolf], as pruh [strip] — prùzok [belt] ,prach [dust] — prúsek [speck of dust], sma [roe] — sm [sr.«] (pi. gen.), etc. In this and other similar cases which may be found in Ukrainian, Old Indian and other languages two things must be considered. First, it is clear that this kind of grammatical formation clearly differs from the case of morphological analogy quoted above, because it is not a result of an analogical transference from one form of the same, or related, word to another, as is the case with, e.g., svadebsvadba. Secondly, the sound which owes its origin to such a transference of the opposition mark, e.g., length of vowels to the syllabic / or r in Slovak, and occurs only in this special formation, has a function similar to that of a combinatory variant. We might call it morphological variation to keep it distinct from alternation of phonemes in their relevant positions (cf. Czech ruka [hand] : ruce [sing, dat.] ). To my knowledge no analogical transference of the long syllabic / or r exists in Slovak.

124

Synchronic

Phonology

Combinatory variation is similar to another phonemic factor, neutralisation of oppositions. Just as it is impossible to keep words distinct by the combinatory variants of the same phoneme, neither can the neutralisation of a phonemic opposition be functionally used for this purpose (cf. let [flight] : led [let] [ice] in Czech). The functional difference between them is merely that the lack of distinctive function of the former is general, whereas that of neutralisation is restricted to certain positions in words. All that we have said above of variation also applies to neutralisation. Like variation it is realised by phonetic assimilation, so that new sounds may arise in this way that do not otherwise occur in the language. In Italian the neutralisation of the articulatory place of the nasals m and η is regulated by the following consonants appearing before labials, [n] before dentals and [η] before velar consonants. In Dutch and English the place opposition of ηι/η/η before labials and velars is also realised by nasals, the articulatory place of which is fixed by that of the following consonants. In Gothic the neutralisation of voice of spirants at the end of words is realised by unvoiced consonants. In Old Germanic the neutralisation of voice opposition of the spirants f/v, p/b and χ / γ after an unstressed vowel (+ a consonant) was realised by the marked members of voice correlation owing to the preceding voiced sounds. This instance is especially interesting, because it shows that the phonetic product of suppressed oppositions need not be identical with the unmarked member of the correlations even in "free" positions. Like variation, neutralisation cannot be phonologised by morphological analogy. In Czech the voice of consonants b, d, d, ζ, ζ and h cannot be transferred as a distinctive mark from inflected forms to those in which they have a final position through analogy. Even if we did pronounce d at the end of such words as led, e.g., when dictating to pupils in the class, this pronunciation would not be a result of morphological analogy and would not infringe the functional rule of neutralisation. It would be only regarded as an intentional reference to the "correct" use of letters in spelling. NOTES 1. Cercle Linguistique de Prague (1931). "Projet de terminologie phonologique standardisée". TCLP4: 3 0 9 - 3 2 3 . 2. Trnka, B. ( 1 9 3 6 ) . "O analogii ν strukturálním jazykozpytu" [About Analogy in Structural Linguistics]. SaS 2: 2 2 1 - 2 2 2 . (Reprinted in this volume in Section 1, p. 29.) 3. Trubetzkoy, N. ( 1 9 3 1 ) . "Die phonologischen Systeme". TCLP 4: 9 6 - 1 1 6 . 4. Jakobson, R. ( 1 9 3 1 ) . "Prinzipien der historischen Phonologie". TCLP 4: 2 4 7 - 2 6 7 . French version in SW 1: 2 0 2 - 2 2 0 . 5. The structural importance of a phonemic law for the language in a given period o f its development is, perhaps, in direct proportion with the greater or lesser extension of its functional validity.

Norwegian Alveolar Consonants

In the eleventh volume of the Norsk Tidskrift for Sprogvidenskap we find an interesting account by Hans Vogt1 as to the relationship of functional speech variants within one language system; his general conclusions are based upon the phonological analysis of the Norwegian alveolar sounds't, 'd, 'η, 7.2 The usage of these consonants, so characteristic for Norwegian, is determined by various styles of pronunciation which show even idiolectal variations: in careful pronunciation we observe according to Vogt's survey a maximum of consonantal clusters of the type rt, rd,rn,rl and minimum of alveolare't, 'd, 'η, 7 (instead of these clusters), in careless speech Vogt observes a minimum of clusters and a maximum of alveolare. Usage oscillates between these two extremes, but even in the most standard-like style alveolare occur and also in the most careless pronunciation some clusters remain without replacement by alveolare. Hence the problem, whether alveolare are stylistic variants of consonant clusters or phonemes in the Norwegian linguistic system. C. Borgström3 evaluates them as independent phonemes as far as they appear within words or at their ends, but as phonological equivalents of the clusters rt, rd, etc. as far as they appear at the beginning of words in sentence contexts. Cf. han er tynn [he is thin]. Vogt analyses alveolars only in isolated words and states: (1) Norwegian words are not determined by the opposition of alveolar consonants versus corresponding consonant clusters, with the rare exceptions rd¡'d, and rt¡ 't. (2) All these consonantal clusters may always be substituted by alveolars under certain conditions without a change in the lexical meaning of words, but not the other way round: the alveolars 'n and 7 cannot always be used instead of the clusters rn and rl. The relationship rnj'n and rlj'l is, therefore, not mechanical. But even the bilateral interchange of 't\'d for rt (rd), rt (rd) for Y, ( tf)4 is no argument, according to Vogt, against the evaluation of these alveolars as phonemes. Vogt's general rule

Originally published in Czech under the title "K norskym alveolärem". CMFT1, 1941, pp. 170-72. (Translated by V.F.)

126

Synchronic

Phonology

reads: every two implementations, the occurrence of which cannot be determined by simple rules of a purely linguistic character, must be evaluated as two distinct phonemes or groups of phonemes, be they interchangeable or not. Potentially these phonemes are diacritics, but it does not matter whether a few words, only distinguished by them, occur in the language. For example, the sound't may be considered as a special phoneme without regard to the fact that only rare instances of its distinctive faculty may be found (cf. Kart with rt means [map] but [unripe and sour fruit] w i t h ' t instead of rt). Vogt is right in saying that the evaluation of a sound as a phoneme does not depend on the actual occurrence of a minimal number of pairs of words that are distinctive by this sound. The distinctive faculty, however, is a necessary quality of the phoneme and this is manifested by the fact that the sound actually distinguishes the same segments of words of the same functional speech. In English, for example, two related sounds θ and 8 are phonemes, because they help to distinguish words like thin and this, tooth and booth etc. and they would be phonemes even if the archaic words they and sooth, which are distinguished from thigh and soothe [s«:9] by the opposition θ versus 3, disappeared completely. We may also agree with Vogt when he states that morphology cannot guarantee a correct phonological interpretation. If, for example, in Norwegian we have snär [mase, and fem., fast] and sna't (written snart) the neuter of snär [fast], just as stör [mase, and fem., great] and stört (the neuter of stor) we cannot conclude t h a t ' t is phonologically lr¡ + ¡ti. The phonological evaluation must be based in principle only on phonetic and phonological features. In the above-mentioned case the relationship between the forms snar and snart must be explained by alternation of r¡ 't which functions as the morphological opposition utrum versus neutrum. Alveolars are phonemes in Norwegian, but only in so far — and herein I differ from Vogt — as we analyse colloquial (spoken) Norwegian. In this functional speech variant of riksmâl the alveolar consonants appear in the same sound neighbourhoods as the consonantal clusters m, rt, and they are not interchangeable for alveolars (e.g. arne [scars] : 5(e)'ne [words] besides drene). From the structural diachronic point of view the matter is evident: the phonological change rn > 'n, rt >'t, etc., to which colloquial (spoken) riksmâl has adjusted, took place a long time ago and therefore the present linguistic system can now admit new clusters rt, rd, m,rl which results from later sound changes or by analogy, or as the case may be, from foreign loans or from the intake of literary words. In written standard riksmâl the alveolars are still only allophones (as Ί in the cluster 'si-) or stylistic ones. Both speech variants influence each other, i.e., the spoken colloquial language takes over from the standard language words with consonant clusters, whereas the standard language takes over from the colloquial language words with alveolars. s Under the impact of language cultivation, which preserves the

Norwegian Alveolar Consonants

127

old tradition, based primarily on writing, the differences m\ 'n, rtl't, etc. (in spite of words taken over like kart [unripe fruit], which has lost its colloquial character) are only stylistic variants (or allophones). The alveolare could become real phonemes in standard riksmâl only after its norm has shifted under the pressure of extralinguistic factors and the stylistic difference m/'n, rtl't, etc. has been abolished, i.e., after it has been phonologised, as it is now the case in the spoken (colloquial) language. The circumstances, however, are more complex, but the fact is that the stylistic variation m/'n, etc. is maintained only by the existence of the standard norm and as long as this norm is not brought in line with the system of the colloquial language, it is more acceptable to consider the occasional phonological function rnl'n, etc. in the standard language as a feature that is alien to this literary style. Because of the complexity of the language scene in Norway the attainment of this stage in riksmâl appears unlikely, but not quite impossible. From what has been said so far follows that the practical rule, by which Vogt complements Trubetzkoy's rules must be tested 6 by analysing further languages, as Vogt himself emphasises. As his formulation is only based on an instance from Norwegian phonology, it calls for caution. As far as I can judge, in the evaluation of Norwegian alveolar consonants, it is sufficient to apply the criterion of the oppositions rtl't, rd/'d, etc. in the otherwise identical phonetic surroundings, and the criterion of the strict distinction of normed speech as opposed to speech more or less lacking this norm. Both speech varieties complement each other, but differ in evaluating attitude and dynamism. Whereas the 'natural' language develops according to phonological laws of development, the dynamism of the standard language is determined by the shifting norm, which results from numerous extralinguistic factors (e.g. psychological ones), by which the standard language approaches the 'natural' language or moves away from it, but never merges with it.

NOTES 1. Vogt, H. (1939). "Some Remarks on Norwegian Phonetics". Norsk Tidskrift for Sprogvidenskap 11: 1 3 6 - 1 4 4 . 2. The alveolar s is evidently a phoneme in Norwegian and that is why Vogt does not discuss it in his article. 3. Borgström, C. (1938). "Zur Phonologie der norwegischen Schriftsprache". Norsk Tidskrift for Sprogvidenskap 9: 2 5 0 - 2 7 3 . 4. In the word ordentlig [o'n't'li] [orderly, honest j the 't cannot even be substituted by rt, but such instances that result from assimilation or may be interpreted as phonological neutralisation are not considered by Vogt. 5. The occurrence of Ί in the word skole [school] is striking, but also / occurs. 6. Trubetzkoy, N. (' 1939). Grundzüge der Phonologie. TCLP 7.

About Monophonemic Words

In his article of 19391 J.M. Korinek concludes that every independent word represents at least one prosodie unit, even in those cases in which its lexical form is composed of only one consonant. What from the phonic point of view is the unexceptional condition for the existence of an independent word, is, according to Korinek, its form as a prosodie unit, i.e., its syllabic character (in languages that 'count syllables') or its mora character (in languages that 'count morae'). Every independent word includes at least one prosodeme, either a syllable or a mora.2 As the nucleus of a syllable may be any sound, Korinek concludes that the monophonemic forms may be also composed of consonants. In support he quotes Slovak 'lexical forms' s (in the sentences bielu s ['white' coffee with cream] : smerom k [in direction to] ; pismeno "s" [the letter "s"] ,plsmeno "t" [the letter " t " ] , etc.;isiel na ζ [he went to the 'loo'] ; dostanel h [you will get nothing]. All these examples, Korinek declares, contradict my theory, which says that consonants — in opposition to vowels — never form independent monophonemic words by themselves.3 I fully agree with Korinek's thesis that every independent word may always represent a prosodie unit; it was a good thing that he has pointed out this fact. But from this thesis it does not logically and necessarily follow that in fact independent monophonemic words consisting of one consonant actually occur, because Korinek's conclusion is affected from outside by the question about the permissible number of phonemes in a word which has not been included in the premises. My thesis and KoHnek's statement are not at logical variance. While KoHnek discusses the question of the word as a prosodie unit, my thesis contains the empirical finding of what seem to be the data arising from the present state of phonological research that the shortest independent words may consist either of a vowel or of a group of any two sounds. This

Originally published in Czech under the title " O slovech monophonématickych", CM F 2 7 , 1 9 4 1 , pp. 2 5 7 - 2 6 0 . (Slightly revised. Translated by V.F.)

About Monophonemic

Words

129

rule about the minimal number of phonemes of the word is based, 4 as I have explicitly stated, on empiric findings: if at a later time in the analysis of some language consonants with the undisputed function of monophonemic words have been discovered, the rule in its general implication will be invalidated. 5 Even if this is so, the special rules as to the minimal number of consonantal and vowel phonemes that can be stated for every language in particular are not affected. For Czech, e.g., the special rule applies which says that the vowels function as independent words (cf. the names for the vowel letters α, ι, u, etc.), whereas the consonants only form independent words (others than interjections) when they appear in three-member clusters bib [idiot], hrd [proud]. The examples quoted by Korinek against the validity of my rule are not fully conclusive because they disagree with the phonological linguistic system. The consonants pronounced on their own (e.g., the letters

, , ) are not lexical units in the Slovak or Czech linguistic system, but only phonic reproductions of isolated sound letters the linguistic names of which are pé, em, es. The syllabic k (smerom k [in direction to] ), s (bielu s [white coffee with cream] ), ζ (isiel na ζ [he went to the loo] ) are again non-lexical units of words or units of close lexical combinations which also lack the phonological structure of independent words. They cannot be put on the same level with English clippings, as e.g., biz (= business), pram (= perambulator), quads (= quadruplets) which have the opposite effect of what they have in Slovak, as they are a normal type of word, or the anglicised versions of long foreign words. Such a phonation may be compared to Czech words of the type tedok (= Ceská dopravrtí kancelàr) [internationally used acronymic label for the official Czech travel agency = Czech Travel Agency], but even these cases differ from their English counterparts because they are phonologically normal lexical forms of the vocabulary and can develop in the course of time as any other independent words. Korinek's wrong interpretation of these phonations of word segments also results from the fact that the syllabic consonants never appear at the beginning of words or of the root syllables (cf. rve [he scuffles] : krve [blood, gen. sing.], rie [he neighs] : νrze [it creaks], Ize [he lies] : plze [molluscs], Iká [he moans] : vlka [wolf, gen. sing.], Isti [trick, gen. sing.] : pis ti [felt, gen. sing.] ,mlha [fog] \msta [revenge]. If I maintain that in Czech we have no independent words consisting of one consonant, I thus refer to the phonological reality. This is, I think, in no way contradicted by Korinek's data. I agree with Korinek when he states that every consonant may be syllabic, but this does not explain anything about the phonological structure of a word in many languages. If the vowels are characterised by their ability to form independent words 6 in contrast with the consonants which by themselves are only capable of constituting 'dependent' words (cf. Czech s [with], ζ [from], k [to],

130

Sychronic Phonology

ν [in], Hung. s(= 'es [and]), I consider this capacity merely a characteristic feature, but not a fundamental one, which makes them differ from the actual consonantal phonemes.7 A definition of the two classes of phonemes must pay attention to their actual phonological structure and define them by means of their phonological relevant properties which make them differ as members of a phonological inventory. I, therefore, consider as most appropriate the definition formulated by R. Jakobson: vowels are phonemes that participate in the opposition of gradual aperture (Czech i-e-a, u-o-a), consonants are phonemes which do not participate in this opposition. I find less acceptable Jakobson's second definition, which is also quoted by Trubetzkoy, 8 and to which also Korinek9 refers: vowels are phonemes which function in their basic variants or as unmarked members of the syllabic correlation (corrélation de syllabisation) as the nuclei of syllables, consonants are phonemes which function in their basic allophones, or as marked members of the syllabic correlation are non-syllabic, as this definition is based on the syllable the phonological existence of which is not yet clear and requires further research even after Korinek's pointed observations. It does not exist, according to Trubetzkoy, in some languages from the phonological point of view, but its existence is also denied from the phonetic point of view, cf. the view of G. Panconzelli-Calzia who considers it a fiction on the part of the psychologically orientated linguists.10

NOTES 1. Koíínek, J.M. (1939). "Zur Frage der monophonematischen Wortgebilde". Linguistica Slovaca 1(2): 9ff. 2. Trubetzkoy, N. (' 1939). Grundzüge der Phonologie. TCLP 7, p. 166. 3. Trnka, B. (1936). "General Laws of Phonemic Combinations". TCLP 6: 5 7 - 6 2 . (Reprinted in this volume Section 2, p. 113.) 4. Cf. page 117 in this volume. The rule, as I have formulated it, is perhaps of an earlier date than the one formulated by: Hjelmslev, L. (1935). "On the Principles of Phonematics". In Proceedings of the Second International Congress of Sciences, Cambridge, p. 4 9 - 5 4 . 5. The examples quoted by Frei from Chinese must be phonologically scrutinised: Frei, H. (1936). Monosyllabisme et polysyllabisme dans les emprunts linguistiques. Tokyo. Cf. also: Trnka, Β. (1940). "O soucasném stavu bádáni o fonologii" [About the present state of phonological research.) SaS 6: 1 6 4 - 1 7 0 , 2 0 3 - 2 1 5 . 6. On p. 20 Korinek says: "unrichtig führt Trnka unter seinen Beispielen unabhängige monophonematische Wörter neben engl, eye, owe auch tschech. a und / a n " . This criticism is not justified because besides eye, owe I quote as further examples Czech a, / as names of the vowel letters < a > , < i > . I refer to the last paragraph of my article on p. 117 in this volume. Under the label 'independent words' I include words which have their own lexical meanings and are the opposite of what Chinese grammar calls 'empty words'. Korinek's description of dependent and

About Monophonemic Words

131

independent words does not deviate from accepted views and I have no objections. The examples which he quotes in support of the fact that one and the same word may have in the context of the sentence either the function of an independent word or that of a dependent word may be increased by a large number of instances which are quite common in English and Germanic languages. Kofinek does not mention the possibility of interpreting some of those instances as homonymie expressions. 7. Herein I differ from Hjemslev in Note 4. Against Trubetzkoy's criticism I would like to point out that his interjections rr\ s! (cf. Note 2, p. 83) are only abnormal 'Lautgebärde' (sound gestures). (From Hjelmslev's point of view it would be possible to consider these s! rr\ as vowels, but this only presents a shifting of the problem, not its solution.) 8. Tnibetzkoy op. cit. Note 2. (Cf. pp. 1 6 8 - 1 6 9 , 169n.) 9. Kofinek, J.M. (1939). "Zur Definition des Phonems". Acta Linguistica, Copenhagen, 1: 9 0 - 9 4 . 10. Panconzelli-Calzia, G. (1924). Die experimentelle Phonetik in ihrer Anwendung auf die Sprachwissenschaft, Berlin, p. 119.

Phonological Foreignisms in Czech

Foreign words are fitted into the phonological and morphological system of the receiving language by the very fact of borrowing. The phonological revaluation, which takes place necessarily and automatically, may be characterised in three ways: (i) phonologisation which implies that the combinatory variants of foreign phonemes are substituted by the phonemes of the receiving language ; (ii) dephonologisation, if combinatory variants of the receiving language replace the phonological oppositions of the source language; (iii) rephonologisation, which includes all other cases. All the three instances of substitutions, labelled according to the terms used by R. Jakobson for historical phonological changes, are significant for foreign borrowings in Modern Czech.1 I shall chiefly use proper names as my data in order to isolate as much as possible the necessary phonological substitution from other, not quite necessary changes which affect borrowings in the general vocabulary. Phonologisation occurs in foreign proper names less often than the other two shiftings. In the Czech phonological conception the English occasional variant of /«/ in front of /// in New York is a phoneme /p/. Similarly the palatalised occlusives in Tewkesbury and Dewey are revaluated as phonemes It/ and / -f-). But even today /// is not so fully integrated in the Czech linguistic system (as e.g. /s/), as it lacks a correlation relationship with /v/. The incomplete correlation relationship is evident from the fact that before jv¡ neutralisation of voice in the preceding consonant does not occur, cf. svou [instrum. of svüj = reflex, pron his] : zvou [they invite], as is the case in front of paired voiced consonants, cf. zpivá [spi'. va] [he sings], zbyvá [zbi.va: ] [it remains]. The voice opposition of the sound [/] in words with notional meanings is still a foreign feature, or a feature of expressivity, even if it appears in the most colloquial types of words. The difference between /// and Igl is evident from the fact that /// also occurs in all dialects in its phonemic function the perception of which is not fully native, whereas Igl as a term of the oppositional function occurs only in the standard language (perhaps with the exception of the East Moravian dialects). 8 The co-existence of /// and /gl is continued by the existence of the fully integrated and native Czech phonemes /v/ and /k/. If a phoneme of the source language includes only distinctive properties foreign to the arsenal of phonemic properties of the receiving language, such a phoneme cannot be taken over by the receiving linguistic system and must be substituted by another phoneme which has some of the distinctive features in common with those of the foreign sound. Thus in Russian, the dental spirant / d / was replaced in Greek loanwords by the labiodental spirant ///, which itself was not (and still is not) an integrated phoneme, but felt, however, to be less foreign than d, owing to the whole series of native voice oppositions, such as z/s, d/t, v/f In Danish, which does not possess phonological voice (/// vs. /v/ is an exceptional pair), the phoneme /z/ of loanwords is substituted by the Danish unvoiced consonant /s/. This substitution cannot be explained by stating that the Danes would find the pronunciation of the voiced fricative [z] physiologically difficult. They are quite capable of pronouncing the voiced consonants efficiently if they speak French or English, and they also occasionally pronounce a semivoiced [z] in native words between vowels, without becoming aware of the acoustic difference of which non-native Danish speakers are aware. The Czech speaker articulates also without any difficulty the English bilabial jw/, but this articulation lacks any function in the Czech phonological system. 9 Our arguments so far have clearly demonstrated that every phonological

138

Synchronic

Phonology

foreignness — if it is to be integrated as a foreignism in the linguistic system of the receiving language — must be linked to some of its native distinctive properties and that the foreign phonemes are subjected to the historical development of the language on a different level than the fully integrated native structural phonological elements. It seems to be a principle that the inventory of phonemes which are at the disposal of a linguistic system cannot be enlarged through direct impact of foreign words which represent an external factor. The foreign phonemes may be absorbed in the structural stream of the phonological development only if the new phonemes originate by means of phonemic mutation in the linguistic system and if these phonemes are identical with the foreign ones. As a result of this mutation the words which contained foreign structural elements will become native words and their assimilation will then be complete. This process might be exemplified by the phonologisation of the voiced fricatives jvj, /9/ and /z/ in English in the latter part of the 14th century and by the resulting full assimilation of words of French origin in which the voice opposition of fricatives s-z and f-v was until then a foreign feature. The preparation of the full assimilation of the phoneme /// in Czech was the loss of the hard and soft jers and the resulting neutralisation of voice in consonantal pairs at the end of words. However /// would have become a structurally fully integrated phoneme, which could also be used as alternant, if a n a t i v e / h a d appeared in consequence of some kind of phonemic mutation. 1 0 We do not want to deny the possibility that the mutation of the phonological system may aim at assimilating foreign elements, as was the case in Hungarian which in pursuing this aim differs from Hungarian-cognate languages; we maintain that this aim can only be achieved by means of phonemic mutations within the linguistic system itself, not by way of the external impact of borrowings. The same may be said about the inventory of phonological combinations and the positional occurrence of phonological oppositions. Unusual groups of phonemes or their unusual positions in words remain outside the system and structure of the receiving language which often replaces them by combinations that are at its disposal. The validity of this thesis may again be doubted and reference may be made to consonantal combinations as ksp in the proper name Shakespeare, d-¡ in the word dzungle [dzmgle] [jungle], dzin [d^in] [gin ],dzem [d-^em] [jam], which appear only in foreign words as phonological features of their foreignism. The possibility of taking over these unusual clusters into Czech lies within their occurrence at the juncture of two morphemes (cf. k spolecnosti [to society] ) or of two lexical words. The foreign element thus only consists of the morphological inseparability of these clusters. The heteromorphological phonological clusters which are inherent in the language have merely a mediating role. If a language does not even have such consonantal clusters, a decomposition of the foreign clusters

Phonological Foreignisms in Czech

139

must necessarily take place. In languages in which consonantal clusters do not occur, or hardly ever occur, and the consonant always combines with a vowel, the consonantal clusters in borrowed words automatically change into syllables (cf. Japanese Kirisuto = Christ). This fact is also proved by the embarrassment of German and English native speakers when faced with Czech or Croat proper names which contain a sonorant r between consonants. Similar changes occur in the position of phonemes. In English the initial position of the phonemes /v/, jz¡ and / 3 / is a mark of foreignness (cf. vine, valley, vane, zone, Xerxes, four), it is made possible because these phonemes occur at the beginning of word-medial and word-final syllables in words of purely native origin (cf. leaving, lazy, leisure) so that their occurrence at the beginning of initial-word syllables is not perceived as a deviation, but only as the admissible extension of the validity of the laws of phonological word structures. On the other hand it is not possible in English to have foreign words beginning with the velar /rj/ because such a foreignism has no foundation in the phonological structure of English words: no syllable of the same morpheme begins with this phoneme (cf. ring-ing, Birming-ham). The phonemic mutations 11 bring about changes in the inventory of the phonological combinations and in the positions of phonemes in the word structures. Neither can the possibility of intentional assimilation of such foreignisms be excluded. But this direct impact on the phonological structure of native words is not feasible. Unless native clusters, with which the foreign clusters may fuse, exist in the receiving language or emerge there, they will be evaluated as foreign phonological and morphological features which the receiving language tolerates in contrast to its fully integrated structural elements. This thesis is also confirmed by the transfer of foreign length (vowel quantity) in unusual Czech positions. An unusual position of long vowels in Czech words is their position in front of homomorphemic clusters of two or more consonants (with the exception of st, st', f t ' , zd, zd\ d', mj\, tr) and in this position the length of vowels in the Czech pronunciation of foreign words oscillates or is abolished. The name of the Hungarian writer Kálmán sounds strange to the Czech speaker because of the vowel quantity in front of the cluster Im which is instinctively disregarded in the context of Czech utterances. Similarly the proper names Cambridge, Auckland, Wales, Shaftesbury, Walpole, Wakefield, are often pronounced in Czech with a short vowel, even though the Czech speaker knows that the vowels are long in the stressed syllables of these words in the original pronunciation. In numerous cases (Darwin,Marlowe,Hertford·, York;Irving,Perth,Byrd, Guernsey) Czech or German spelling rules may influence the pronunciation, but the short pronunciation of the vowels is certainly also strengthened by their positions. In the Czech pronunciation of the above mentioned two-syllabic words yet another positional limitation of Czech length plays a role.12 In

140

Synchronic Phonology

Czech there appear only short vowels in the first syllable of morphological indivisible two-syllable and more-syllabic words (or better, stems) — with some rare exceptions - (cf. palub-a [deck], seker-a [axe], tetel-it se [to shiver], holoub-é [squab] ). Both these limitations of the occurrence of vowel length do not represent a phonological neutralisation in the proper sense of this term and cannot be compared with the neutralisation of length in Slovak13 or of the neutralisation ili:, u/u: before / and final m in Czech. In both cases there exists only a non-occurrence of long vowels which is justified structurally by the fact that words of the type park are, almost without exception, borrowings, and further by the fact that stems of Czech words, with the exception of expressive words and of terminological expressions, (which are also partly of expressive origin) are all one-syllabic.14 If they consist of two stem syllables (cf. holub [pigeon], the second syllable may be long, if followed by a suffix (holoub-ek [squab], andèl [angel] : andilek [little angel] ; cf. also jelimän-ek [simpleton] ). Apart from in prefixes, long vowels and the diphthong ou appear (a) in monosyllabic words (mostly nouns, adjectives, and verbs), (b) in the syllable followed by a derivational or grammatical suffix (never in the first stem syllable, after which a second stem-syllable follows). Other positions of vowel length in foreign words are therefore in no way linked with Czech vowel length positions and they are perceived as something that is linguistically useless; it may, however, be benevolantly preserved, but it also may be dropped without any danger of misunderstanding the word. On the other hand vowel lengthening occurs in foreign words if they contain two flanking consonants, the second of which is m (cf. Bern [proper name] Jam = [d^e:m\ ). This lengthening trend most probably happens so that final -em may differ from the phonological structure of the inflexional -em and may be perceived as a component of the stem syllable. The dynamic character of the linguistic system tries to rid it even of these foreignisms that are intentionally tolerated by the standard language, and if they cannot be used for expressive ends, the system adapts them fully to the phonological and the morphological structure of native words, quite often in conscious contrast to the 'original' sound. The system thus not only avoids unusual or less usual combinations of sounds, which are replaced by other more familiar ones, but it also reduces the basic more-syllabic morphemes to one syllable, to which is added a Czech or Czech-like ending, or it fits the borrowing to the morphological pattern to which it does not actually belong (to so-called folk etymology). In this form foreign words often have a vulgar colouring, but in the course of time these words enter into the colloquial and standard written language of educated Czech speakers because of their expressivity, but of course without displaying phonological vulgarisms like ej instead of < y > [/'. ] in Present-day Czech usage; they suppress the 'original'

Phonological Foreignisms in Czech

141

form of the word (sometimes even of its synonym) or they appear next to it in a somehow different meaning. The analysis of foreign words, which is quite wrongly not fully recognised in linguistic research, discloses not only the phonological and morphological creativity of a language collective, but also indicates the hierarchy of the phonological and morphological features in the native linguistic system. Foreign words are also an important source on which a language can draw for material to meet its communicative and expressive needs, a fact that is now being more and more realised. The external resources of the linguistic material are absolutely necessary for every linguistic system, because the internal word-formative means - word derivation - hand in hand with the native lexical stock would hardly suffice over a long stretch of time. It is quite wrong to reject borrowings only because they are of foreign origin, linguistic purism, the heritage of outdated linguistic theories in the last century, which has become more and more remote to us, rightly emphasised the independent derivational capacity of a language, but it did not consider its needs to acquire ever new, stylistically differentiated and viable material and it also underestimated the capacity to create its own material resources from the foreign phonological material. By stating these well-known facts which can be proved by quoting plentiful data from the historical growth of any European language, we do not want to say that the native word stock should be kept wide open to the absorption of foreign borrowings without paying any attention to the norms of language cultivation which have been spontaneously established by language use in agreement with the internal structural linguistic and the extralinguistic factors. So e.g. Latin and French loans in Germanic languages occurred to a far greater extent in older periods and thus indicated the strong cultural activism of the language communities and their social progress. Later on, when important poetic and literary prose works, which formed the basis of the systematic codification by the grammarians, were produced, the Germanic literary languages became naturally more resistant to new foreign influences than they were in times of their lesser stylistic efficiency. The Czech language went through this phase of development only recently, whereas the Germanic languages had passed through it a long time ago. As long as there did not exist outstanding, generally recognised samples of Modern Czech literary prose and poetry, the prescriptivism of the purists was based on the works of long bygone times, as it was then the fashion, but their attitude was felt to be an oppressive burden for the living language which was spontaneously acquiring its own means for an efficient stylistic differentiation. The use of foreign words is now guided by comparatively stabile norms, it is primarily a matter of style. But loanwords and borrowings will always represent a linguistic process which can neither be excluded nor suppressed.

142

Synchronic Phonology

NOTES 1. Jakobson, R. (1931). "Prinzipien der historischen Phonologie". TCLP4: 2 4 7 - 2 6 7 . Reprinted i n S W 1: 2 0 2 - 2 2 0 . 2. Frinta, A. (1937). "Ceská vyslovnost pûvodu slovanského". [Czech pronunciation of Slavic origin.]. SaS 3: 3 5 - 5 8 . 3. Buben, VI. (1941). "Κ vyslovnosti románskych slov ν cestinë". [The pronunciation of Romance words in Czech]. SaS 7: 1 4 4 - 1 5 4 . 4. Kainz, F. (1941). Psychologie der Sprache Vol. I. Stuttgart. Cf. 224 ff. 5. The combinatory variants are not decisive here. In Czech there exists e.g. [tj] as a variant, yet the foreign phonemes realised by the same sound are replaced by the combinations [rjfc], [τ^] and not by a phonological [ r j ] . 6. Havránek, Β. (1923). "Neasimilování párové souhlásky znëlé a neznélé ν stare cestinë". [The unassimilated voiced and unvoiced consonant pairs in Old Czech]. Slovansky sbornik vènovany F. Pastrnkovi. [Slavonic Papers dedicated to F. Pastrnek], Prague, p. 102. 7. Mathesius, V. (1932). "Cizí slova ze stanoviska synchronického". [Foreign Words from the synchronistic point of view]. CMF 18: 2 3 1 - 2 3 9 . Reprinted in Ceitina a obecny jazykozpyt [Czech and General Linguistics], Prague, 1947, pp. 9 6 - 1 0 9 . Mathesius, V. (1934). "Zur synch ronischen Analyse fremden Sprachguts". Englische Studien 70: 2 1 - 3 5 . (Reprinted in PSRL, pp. 3 9 8 - 4 1 2 . ) Mathesius, V. (1935). "Vyslovnost cizlch slov ν cestinë". [The pronunciation of foreign words in Czech]. SaS 1: 9 6 - 1 0 5 . This study is supplemented by the articles of A. Frinta (op. cit. note 2) and V. Buben (op. cit. note 3). 8. The evaluation of Idzl and ¡d^l as special phonemes within the Czech phonological system is not certain although both sound complexes represent from the phonetic point of view voice oppositions of the phonemes Its/ and /t¡¡ and, moreover, they also occur as combinatory variants, cf. leckdy [ledzgdi] [sometimes], licit [le:t}it] [to cure] : lécba \le:d$ba] [ c u r e ] , d z b á n [d^ba.n] [jug]. L o n g / o : / in notional words as well as long /e:/ in colloquial Czech may be evaluated as phonological foreignisms. 9. Trnka, B. (1940). "Poznámky ke germánské expresivní geminaci". [Some remarks concerning Germanic expressive gemination]. CMF 26: 8 5 - 9 2 . (Reprinted in this volume in Section 4, p. 236.) 10. The creation of new phonemes by way of morphological analogy is impossible. Cf. Trnka, B. (1936). " O analogii ν strukturním jazykozpytu". [About analogy in structural linguistics]. SaS 2: 2 2 1 - 2 2 2 . (Reprinted in this volume in Section 1, p. 29.) 11. The interference of morphological analogy with the inventory of phonological combinations is not feasible. This inventory is neither increased nor decreased by analogy in the course of the historical development of a language. If e.g. a new form deska [board] is coined according to the gen. pi. desk (for older dska-desk), the new form is only possible because the combination of phonemes /dl-lel-lsl-lkl already existed in the Czech phonological system. Even the sequence of phonemes is the result of structural phonological laws existing in the language with which morphological analogy does not interfere. 12. Quantity in Hungarian is not limited in this way in agreement with the entirely different structure of the words, cf. Bártfu,Párizs, Zágráb, Késmárk, Svájc, Zólyom, Stájerország, Kárpátok, Sándor, Nándor, Gárdonyi, Ádám, Kázmer.

Phonological Foreignisms in Czech

143

13. This is also indicated by the alternation of length, cf. Bartàk : Bárta. The alternation of length in stem vowels occurs in words the stem of which is one-syllabic and in some rare cases (e.g.zeman [squire] -zemánek [little squire]) in the second syllable of a more-syllabic stem. In the first syllable of a more-syllabic stem alternation between long vowel and short vowel does not exist in Czech. 14. 'Terminological' words (such as e.g. names of birds, cf. holub [pigeon], sykora [titmouse]), displaying phonological deviations in their word structures, are more like proper names.

The Determination of the Phoneme

Trubetzkoy's Grundzüge 1 will remain for a long time a basic textbook which every linguist who wants to become acquainted with the principles and the methods of phonological research must take account of. The phonologists continuing research in this field feel the necessity to make certain improvements of some of Trubetzkoy's rules if phonology is to develop further as a living linguistic science. One of the sections of Trubetzkoy's book which must be discussed further contains his rules for determining the phoneme. 2 Trubetzkoy formulated 11 rules, four rules for distinguishing phonemes from allophones and seven for evaluating sound complexes. RULE 1 : If two sounds occur in the same position in a word and if they are interchangeable without changing the notional meaning of the word, these two sounds are only facultative allophones of one phoneme. RULK 2: If two sounds appear in the same position of a word and if they are not interchangeable without changing the meaning of the word or without making its meaning indistinct, the two sounds are the phonetic realisations of two different phonemes. RULE 3: Two acoustically and articulatorily related sounds of one language, which appear in the same contiguity of sounds, are considered as allophones of one and the same phoneme. RULE 4: If two sounds, which fulfil the conditions of rule 3, can combine in the position in which one occurs isolatedly by itself, they are considered as t w o different phonemes. Trubetzkoy quotes English [a] and [r] as phonemes

Originally published in Czech under the title "Urcování fonému". Acta Universitatis Carolinae 1954, 7, Philologica et histórica. Prague, pp. 1 6 - 2 2 . Revised. (Translated by V.F.)

The Determination

of the Phoneme

145

although they occur in exclusive positions because they occur in combinations ra and ar, which are in the same position asa , e.g. pra 'fefn-pa 'fekf(a)n. Further rules: The monophonemic evaluation of sound complexes is possible (1) if the elements of such sound combinations do not decompose into two syllables; (ii) if they consist in one gliding movement of speech organs or in its successive alteration ("Abbau"); (iii) if the duration of the articulatory process does not exceed that of one single sound. The sound combinations which fulfil these three conditions, must be evaluated as monophonemes, only (i) if they appear in the position in which the combinations of two phonemes in the same linguistic system are not admissible; (ii) if this causes parallelism in the inventory of phonemes in that language; (iii) if the element of the potentially monophonemic sound combination cannot be interpreted as an allophone of a phoneme in the same language (e.g. ^ 3 ] is an allophone of [ t f ] , cf. lécba [le. dzba] [cure] in Czech). The last rule reads: If between a sound and a sound combination that satisfies the above-mentioned phonetic conditions there is a facultative or positional allophone, whereby the sound combination must be evaluated as the realisation of the phonemic combination, the sound in question is evaluated as the realisation of the phonemic combination (e.g. syllabic r,l,m,n'm German are considered combinations of two phonemes: al, ar, am, an). Trubetzkoy's first three rules are based on the principle ofinterchangeability, which is according to his foregoing criteria the main criterion of a phoneme in comparison with an allophone. The fourth rule, which is a limitation of the third rule and which adds a new diagnostic factor into the determination of phonemes, is generally less cogent if we take into account that the data may also be interpreted differently. But there are defects both in Trubetzkoy's second and third rules. In the former Trubetzkoy uses the inexact negative term ('unkenntlich'), in the latter the moment of articulatory relatedness hardly agrees with Trubetzkoy's basic tenet that the phoneme is prius and its phonetic realisation is something that depends on it. Even the first rule is not quite satisfactory because according to this rule we would have to consider the glottal stop in Czech and some other languages as a phoneme; e.g. in the Czech word pair slúdy [with the limbs] — spûdy [from the loft] the [2] and the [p] are not interchangeable without changing the notional meaning of the words. Even the rules for the phonological diagnosis of sound complexes are not sufficiently exact. The Czech sounds ts (spelled < c > ) and tf < c > which fulfil the conditions for a monophonemic evaluation according to rules 1 - 3 occur in the positions of two-phonemic clusters (e.g. st, ft)·, their monophonemic evaluation does even not necessarily follow from the parallelism in the inventory of Czech phonemes, moreover both sounds (t and s, resp. f ) occur

146

Synchronic

Phonology

also in combinations ts, t f . We, therefore, might doubt whether Czech < c > and < c > should be thought of as monophonemes or as groups of two phonemes. Trubetzkoy's rules cannot be considered definitive, and it seems quite likely that the author himself would have modified them had he been able to continue in his indefatigable research. As far as the monophonemic evaluation of sound complexes goes, the complexity of Trubetzkoy's practical rules results from the lack of a basic criterion for the delimitation of phonemes in the word. Such a criterion is the notion of contrast. Every phoneme occurs in functional contrast with the contiguous phoneme(s) in one and the same word. If this contrast does not exist between the sound elements of a phonic combination, the combination represents the realisation of one phoneme. Thus in the Czech sound ts < c > (or tf ) there does not exist a contrast between t and s (or between t and f ) which exists in the clusters st or ft or in the cluster tf between t and r, this lack of contrast suffices in order to evaluate Czech ts < c > and tf < δ > as monophonemes. Contrastivity is therefore a basic criterion for the phonological evaluation of sound complexes. The resistance to indivisability of these two sounds, as well as their gliding articulation and duration not exceeding that of a simple sound, which are mentioned as criteria by Trubetzkoy, are only the outward phenomena of the lack of contrastivity between the two articulatory components. The contiguous contrast may be bilateral (e.g. tr-rt, kl-lk, ba-ab, etc.) or unilateral (e.g. st, ft ). Every phoneme contrasts with the other phonemes in the word, and all these contrasts contribute to the solidity of the phonemic structure of the word. As for the evaluation of the simple sounds, the following rule which Trubetzkoy has not included must be formulated: the sound the omission of which never changes the notional meaning of the word is not a phoneme. So the sound [2] which may be omitted in Czech without altering the meaning of the word (e.g. after the prepositions k, s, v, if a vowel follows) is not a phoneme. On the other hand i and / are phonemes in Czech because it is not possible to omit them, e.g. in the word zapijel [he drank it down] the vowel i cannot be dropped without changing the meaning of the word: zapel I 'zapjelj [he started a song]. The criterion of the omissibility of a sound by itself does not suffice to determine the phonological validity of a sound. It merely states which sounds cannot be phonemes, but it does not assert that the sound the omission of which in the word realises a new word must be a phoneme. For example, the sounds ρ and η cannot be dropped without changing the meaning of the Czech words potok [brook] and otok [swelling], vtnko [vi:r\ko\ [a little wine] and vlko [lid], but of the two sounds only ρ is a phoneme, η is an allophone. The basic criterion of the phonological diagnosis of sounds can be attained if we posit the question which conditions the English phonemes [A] and [η]

The Determination of the Phoneme

147

fulfil. The criterion of phonological relevance of the two sounds cannot be their oppositional position: both sounds occur in exclusive postions in the word: h appears only in front of stressed vowels (or in some cases also in front of j + stressed vowel), whereas η occurs only after vowels, stressed or unstressed ones (cf. hang, behave, human ; king, singing). If we find Trubetzkoy's view unsatisfactory which interprets them as indirectly distinctive (indirect phonological) oppositions and thus implies the imperfection of the principle of opposition as the basic criterion of phonemes, it is necessary to explain their phonological function by the exclusiveness of their relevant properties. Both phonemes /Λ/ and /η/ do not have in common the total of their relevant features and must therefore be distinct phonemes, even if in the lexical inventory their capacity of distinguishing words is not actually exploited in the same positions, as for example in the case of ¡p\-¡b¡, Iti-fri etc. The relevant features of the phoneme are all the properties that are not determined "from outside" by its contiguity with other phonemes in the word, but inherent in the phoneme itself, distinguishing it from the opposite features of other phonemes. Irrelevant properties are those features of a phoneme which are the reflection of the articulatory movements of the contiguous phonemes in the word. Sounds that have the same total of relevant properties and simultaneously different irrelevant properties are allophones of one phoneme. So the nasals n-η are allophones of the same phoneme in Czech and Slovak because the velar articulation of η links it with the following k org so that the total of relevant features of both sounds n and η is the same. If this is the case, we may define the phoneme as the total of relevant properties which differs from other such totals in the word by contrastivity. From our concept of the phoneme as the total of relevant properties follows that every sound which realises a word by itself (e.g. the Russian i [and] ) is a phoneme. Just as the criterion of identity of the phoneme is the identity of the total of its relevant properties, the condition for the distinctive and opposite capacity of the sound is the independence of its relevant properties on its syntagmatic rules of contiguity in the word. The consequence of our concept of the phoneme is that, in the last analysis, we should be able to determine the inventory of phonemes of any language on the basis of the principle of relevant features even without knowing the meaning of words and morphemes as long as we know the delimitation of words and morphemes (inter-word or inter-morphemic sutures). In our concept of the phoneme the following rules are implied: (1) The sound form the phonetic components of which are not in mutual contrast and which contains a certain total of relevant properties is a phoneme. (2) The sound form which cannot be omitted or inserted without the change of the intellectual meaning of the word and whose total of relevant properties is not identical with any other such total, is a phoneme. (3) A sound the

148

Synchronic Phonology

omission (or insertion) of which does not change the notional meaning of a word is not a phoneme. A fourth rule by which phonemes are distinguished from allophones must be added. It distinguishes phonemes from 'stylistic variants'. If two sounds are interchanged in certain positions in the same linguistic system in such a way that the intellectual meaning of the word has not been changed, but only its pertinence to a certain style within this linguistic system has been signalled, the two sounds are combinatory (or positional) variants of the same phoneme. The nature of stylistic variants is quite different from that of allophones because the style of speech, which is marked by the pertinence of the speaker to a certain linguistic collective, modifíes the normal phonological realisation of words in the speaker's message.

NOTES 1. Trubetzkoy, N. (' 1939). Grundzüge der Phonologie. TCLP 7. 2. Cf. ibid., pp. 4 1 - 5 9 .

On Some Problems of Neutralisation

Neutralisation is one of the most debatable concepts of structural linguistics. Provisionally, it may be defined as a suppression, under special conditions, of a phonological or a morphological opposition. A few examples from familiar languages will suffice for illustration. On the phonological level, the opposition of the vowel phonemes ejç in French (cf. les-laid, mes-mais) is neutralised in all non-final positions. In Slovak and Czech the "paired" voiced consonants lose their voice finally and before voiceless consonants, while the voiceless consonants acquire it before the voiced "paired" consonants, e.g. led [let] [ice], vèzti [ve:st Ï] [to convey], prosba [prozba] [request] ; in other words, the opposition of voice is neutralised at the end of words and before the terms of the same phonological opposition. In Old Greek the aspiration of consonants is suppressed if another aspirated consonant occurs in the same or the following syllable, cf. trichös [gen. sing.] as against thriks [a hair], [etâphën] [I was buried] as against thäptö [I bury]. In modern Greek the opposition of plosion (p¡f, t/θ, k/X) is neutralised in the neighbourhood of spirants, e.g. ftáno [I attain], extrós [enemy], kátse [sit down]. Instances abound also of the neutralisation of morphological oppositions. In French the opposition of the 2nd person/3rd person is neutralised in the singular, e -g· [ty> Ü ferf] tu cherches, il cherche-, [ty, il ferfc] tu cherchais, il cherchait·, [ty, il a] tu as, il a; [ty, il ] tu es, il est, etc. In Old Scandinavian the same opposition is neutralised in pres. ind. In Lithuanian the opposition of singular/plural is suppressed in the third person, cf. veda [he leads, they lead]. In Old English the 3rd pers. pi. is used for all persons of plural (cf. wë, gë, hie singath [we, you, they sing], sungon [we, you, they sang] ) and in all Germanic languages the opposition of the first/third pers. sing, is neutralised in the perfect. In Old Greek the oppositions of moods are non-

Originally published in Omagiu lui Iorgu Jordan cu prilejul implinirii a 70 de ani. Academia Republic» Populare Romtne, Bucharest, 1958, pp. 8 6 1 - 8 6 6 . (Slightly revised.)

150

Synchronic

Phonology

existent in the imperfect. In Slovak the opposition of genitive/accusative is neutralised with masculine substantives denoting animate beings. In New High German, Dutch and the Scandinavian languages, the opposition of the grammatical gender is suppressed in the plural. Let us first deal with the neutralisation of phonological oppositions. There are three problems to be discussed in this paper, namely (1) which class of phonological oppositons is liable to neutralisation, (2) what is the phonemic character of the sound which occurs in the neutralising position, and (3) what is the cause of neutralisation. As to the first problem, the views of phonologists seem to concur on one point, namely, that only one relevant feature which distinguishes the terms of an opposition can be neutralised, but there is no consensus opinionum as to the properties of this feature. Trubetzkoy 1 rightly emphasises the fact that two given phonemes constituting a neutralisable opposition must have a common basis of relevant features, but his exclusion of localisation oppositions from the list of neutralisable features is hardly tenable. As a matter of fact, the localisation oppositions are neutralised in many languages. Thus in English, the oppositions of m/n, τη/η and η/η are stripped of their relevancy before labials and velars, and similarly in Czech the opposition of njji is suppressed before palatals (cf. andel [an¡el] [angel], puntík \punci:k] [a dot] ). The change of kt to pt in Rumanian shows conclusively that a neutralisation of k/p took place before t in Old Rumanian. All these facts lead us to the assumption that any relevant feature keeping two phonemes apart in a given linguistic system is neutralisable, and this view is in perfect agreement with the results of the acoustic analysis of consonants undertaken by Roman Jakobson. 2 The relevant features of a phoneme are those which do not depend on its position or its neighbourhood in a word and serve only to distinguish it from the other phonemes of the same linguistic system. What is the phonemic character of the sounds realising the phonemes in their neutralising position? Are t in the cluster tk, and d in dg, realisations of the phonemes t and d, respectively, or is t in tk a realisation of the phoneme IT/, and d in dg a realisation of the phoneme /D/, both ¡TI and /£>/ being different from t and d? In the former case, says André Martinet, who poses this question, 3 no regard is paid to the fact that the difference which exists between t and d is, in neutralisation positions, purely phonetic, since it cannot be used for distinguishing one word from another; in the latter case it is ignored that the objective identity of the phoneme /t/ and /T/ (/d/ and /D/, respectively) is not always without influence on the subjective reactions of the speaker, apart from the fact that this assumption would imply the admission of a great increase in the phonemic inventory of a certain language. Martinet's paper does not offer a solution of this problem and Trubetzkoy's views of it are too closely connected with some debatable points of his

On Some Problems of Neutralisation

151

phonemic theory to be acceptable to all phonologists. His main idea is that the specific relevant features of one of the terms of an opposition are stripped of their relevancy in the neutralising position, so that only those which they have in common remain distinctive. If the opposition t/d in German is suppressed at the end of words, the specific feature differentiating d from t is annulled in this position, and the sound occurring in it is, from the phonological point of view, neither voiced nor voiceless; it is simply a nonnasal dental plosive opposed both to the nasal η and to the non-nasal labial and velar plosives of the German phonological system. This element, the socalled "archiphoneme", is represented by sounds which are, in most cases, phonetically identical with the realisation of one or the other term of the opposition, and only in comparatively rare cases intermediary between its two terms (cf. a sort of semi-palatal consonant which represents the archiphoneme resulting, in Russian, from the suppression of palatalisation before palatalised dentals). The choice of the sound representing an archiphoneme ("Stellvertreter") is either independent or, in some cases, dependent on its neighbourhood (cf. voiced and voiceless consonants as representatives of voice neutralisation in Czech or Russian). If we are to attempt a more realistic solution of the problem, we must consider, first of all, the cases of neutralisation the products of which are identical with the realisation of one of the given oppositive terms. Is ζ in the Czech word prosba \prozba] [request] really a different phonemic unit from ζ in pozemni [pozemjii: ] [terrestrial], as it would appear according to the theory of Trubetzkoy? This question must be answered in the negative, and the proof of it is given by the fact of the analogical transference of the morpheme [proz- ] to the adjective prosebnj> [prozebjii;] [entreating], where ζ appears in a non-neutralising position. Similarly the adjective svatebni [svadebjii:] [wedding] has itsd from the substantive svatba [svadba] [a wedding], where it is an outcome of the neutralisation of t/d (cf. svatj> [svari:] [holy] ). These cases of morphological analogy show conclusively that ζ in zb and d m db are not differentiated phonologically from ζ in zeb and from d in deb, respectively, either of them having the same total of relevant features in both positions. If d in db were really stripped of voice as a relevant feature, it would not have the same total of relevant features as d has in deb, and both d's would be dissociated from each other. If the dental oral plosive can be either voiced /d/ or voiceless Iti in the Czech phonological system, /// or /d/ must be used even in the specified phonemic contexts in which a dental oral plosive, without any additional qualification, is required. The whole process of neutralisation must be régarded, therefore, as consisting in the exclusion of one of the oppositive terms from some specified phonemic contexts. Trubetzkoy's archiphoneme has no place among phonemes as realisers of words, and his distinction between the internal and external

152

Synchronic Phonology

choices o f "Stellvertreter" does not seem to be fully justified in view o f the fact that all o f them are basically dependent on contextual phonemic factors. If Trubetzkoy refrained from identifying the phonological product o f neutralisation with a phoneme (or a variant o f the phoneme), it is because he disregarded the necessity of the identification o f phonemes and laid stress only on their capacity o f distinguishing words. In fact, the principle o f phonemic distinctiveness, on which Trubetzkoy's phonological theory is based, must be complemented by that o f identification in order that a full understanding o f phonemic processes be achieved. Any solution to this problem would be at least incomplete i f no regard were paid to the latter principle. Let us proceed to the cases of neutralisation in which the neutralised terms o f a phonological opposition are represented by sounds which do not otherwise exist in a given linguistic system. After what was said above it may be clear that the most obvious explanation o f this kind o f neutralisation is that the special sound which occurs in the neutralising position is used as a variant o f one of the terms representing the opposition in question. Thus rr\ in 'comfort [k&n\fat] is a variant o f the phoneme m which represents all nasal phonemes o f English before labials, V in Czech abych byl [abixbil] [so that I am] is a variant o f the phoneme ch which appears before all other "paired" voiced consonants as a representative o f the opposition h/ch. The variants o f this kind may be called neutralisation variants. Our explanation o f the phonological neutralisation helps us to understand the interesting fact that the same phoneme can be used as a representative o f two oppositions in which it takes part. Thus in Old Germanic the voiceless stops in sp, st, sk represent both the opposition o f plosion [p/f, t/0, k/X] and the voice [f/v, 0/8, X/Y]. In Modern Greek the phonemes [v, 3, ν ] , realised by their neutralisation variants b,d,g, are representatives o f the same two oppositions after nasals. I f we know that the distribution of phonemic oppositions is restricted by neutralisation, we are entitled to ask whether a similar restriction does not also take place on a lower functional level, on that o f relevant phonetic features themselves. This is really a common phenomenon in all linguistic systems. In Czech, where voice is a relevant feature o f a number o f consonantal phonemes, its relevancy is suppressed in combination with their labial and velar localisations, so that the consonants ν and k do not take part in the opposition voiced/voiceless. In Rumanian, Italian and Czech the opposition dental/velar is de-functionalised in combination with nasality, η and η being positional variants o f the same phoneme. In English the distinctive sonority of vowels [i-e-ae; U - Ο - Λ ] and their quantity [i-i: ; u-u:; etc.] lose their relevancy when not combined with stress. In most languages voice is not relevant in combination with nasality. This type o f neutralisation,

On Some Problems of Neutralisation

153

which may be called paradigmatic, is a close parallel to the syntagmatic neutralisation discussed in the preceding paragraphs. Both of them consist in the exclusion of a relevant feature, but while the former applies to relevant features as realisers of phonemes, the latter is concerned with phonemes as syntagmatic components of words. What is the cause of phonological neutralisation? It may be seen, as far as the paradigmatic neutralisation is concerned, in the necessity of restricting a great theoretically possible number of combinations of relevant features to a comparatively small number of phonemes. As stated by Jakobson and Halle,4 there are some necessary restrictions in the overall inventory of distinctive features owing to the fact that some of these (e.g. tension and aspiration) are variants of the same opposition and cannot be simultaneously displayed by most languages as two autonomous oppositions. Apart from Jakobson's universal rule there may be universal rules of exclusion applying to the combinations of autonomous relevant features. Thus voice and nasality seem never t o combine in any language, so that nasal consonants are not split up into voiced and voiceless phonemes, as the corresponding oral stops are. The universality of this and other cases is, of course, difficult to prove and we must content ourselves with the statement that languages differ from each other in their choice of combinations of their relevant features according to their own rules. As to the syntagmatic neutralisation, all cases known to me seem to justify the conclusion that none of them apply to all languages, since the specified conditions governing them differ to a great extent from one language to another. All of them may be regarded as phenomena of language economy, concentrating phonemic relevancy to a restricted number of positions. The neutralisation of morphological oppositions must be kept strictly apart from homonymy. Whereas the former consists in the suppression, under non-phonological conditions, of a morphological opposition, homonymy is the identity of the phonemic realisation of a morphological opposition. For example, the ending-ae in feminae is homonymous, since the morphological opposition of gen. sing./dat. sing, is implemented, with other substantives, by non-homonymous expedients, such as i/o (populi/populo), is/i (mulieris/ mulieri), us I üs (tribus/tribus), etc. Although the morphological opposition gen. sing./dat. sing, of femina has the same phonological implementation, both cases are differentiated in the Latin morphological system in a similar manner, just as the semantic opposition [nait] knight/ [nait\ night remains valid in spite of the absence of any phonemic differentiation. No common semantic element is felt between both words and the speakers would not be even aware of the phonemic identity of these two words if their attention were not drawn by linguists to this fact. On the other hand, the opposition dative/ablative is neutralised in Latin plural (e.g. servis, feminis, trabibus),

154

Synchronic

Phonology

i.e. its function is suspended in combination with the plural. In Slovak the identity of the genitival and accusatival forms (e.g. chlapa [of a fellow], bacu [of a shepherd] ; in plural: chlapóv, baâôv) of masculine animate beings is an outcome of the neutralisation of the opposition gen./acc., whereas in Czech the identity of the genitival and accusatival forms chlapa, oráie [ploughman] is to be regarded as homonymy, both cases being differentiated with substantives which end in-a (vévody [of the duke] : vévodu [the duke, acc.] ). The difference between homonymy and neutralisation is of primary importance both in the field of structural morphology and semantics. In the development of languages the homonymy of words or their morphemes may be, and often is, discarded by various means, whereas the neutralisation of morphological oppositions tends to result, diachronically, in the extinction of the opposition itself. The final extinction of the opposition nominative/ accusative, first neutralised with the substantives of the feminine gender in Old French, is very likely a case in point. All these instances exemplify the paradigmatic type of the morphological neutralisation, which consists in the exclusion of a distinctive feature from combination with other specified features of the same level; e.g. on the phonological level of speech analysis, voice cannot be combined as a relevant feature with relevant nasality to produce consonantal nasals (in all languages, presumably) or with the back localisation of spirants (e.g. in Russian and Polish); on the morphological level, the opposition of genders is excluded from combination with plurality in Dutch, Scandinavian languages and New High German. On both levels neutralisation of this kind is responsible for restrictions in the inventory of functional "forms": on the phonological level, in the inventory of phonemes; in morphology, in the inventory of "grammatical forms" and other morphemes. As in the field of the phonological paradigmatic neutralisation, some cases of the incompatibility of specified morphological oppositions seem to apply to all languages. The universal restriction of the comparison to adjectives which admit semantic opposites (e.g. warm/cold, heavy/light, good/bad, old/young, short ¡long) may be adduced as one of the clearest examples. The paradigmatic type of neutralisation is not isolated in the field of structural morphology and the suppression of morphological oppositions occurs also under syntagmatic conditions in many languages. In some Dravidian languages the opposition singular/plural is neutralised in all syntagmatic positions except in the subject of a sentence. In most European languages the opposition singular/plural is neutralised with substantives used as predicatives. In the sentence my brothers are merchants the plural form merchants does not denote two, or more, individuals, as is the case with the governing substantive brothers, but a merchant without any numerical qualification. The so-called grammatical concord of predicative substantives

On Some Problems of Neutralisation

155

in Latin and many other languages is a case of the syntagmatic neutralisation of the morphological opposition sing./plural, implemented either by singular or plural in accordance with the singular or plural of the governing substantive. A similar case of the neutralisation of sing./plural takes place in negated substantives (e.g. I have no son = I have no sons), which may be put in singular or plural with reference to the singular or plural in the question directed to the speaker. In some other dependent positions of substantives the neutralisation of number is realised either by singular alone (e.g. Czech host s kloboukem na hlavë [boys with their hat (sic!) on their head] ) or by both singular and plural in formal accordance with the number of the governing substantive, cf. a boy with a hat on his head : boys with hats on their heads. As in the case of phonemic oppositions, languages have no special forms which would be neither singular nor plural, and if some specified contexts require a substantive without the numerical qualification, either singular or plural must be employed. In conclusion to this brief exposition of the writer's theory of neutralisation it is to be pointed out that it is extended far beyond the conceptions of Trubetzkoy or Hjelmslev. While Trubetzkoy treats it as a phenomenon of syntagmatic phonology and Hjelmslev's extension of this concept refers only to the cases of the morphological paradigmatic neutralisation, the present article, written in honour of the writer's distinguished colleague of Bucharest University, is meant to show that neutralisation is an important linguistic phenomenon affecting all levels of structural analysis of languages, both "paradigmatic" and "syntagmatic". As structural linguistics extends its range, all these kinds of neutralisation will receive increasing attention.

NOTES 1. Trubetzkoy, N. (1936). "Die Aufhebung der phonologischen Gegensätze", TCLP 6: 2 9 - 4 5 . Cf. here p. 32. (Reprinted inPSRL: 1 8 7 - 2 0 5 . Cf. p. 189.) Trubetzkoy, N. (' 1939). Grundzüge der Phonologie. TCLP 7. (Cf. pp. 2 0 6 - 2 1 6 . ) 2. Jakobson, R. (1941). Kindersprache, Aphasie und allgemeine Lautgesetze. Uppsala. (Reprinted in SW 1: 3 2 8 - 4 0 1 . Jakobson. R. and Halle, M. (1956). Fundamentals of Language. The Hague. (Reprinted in SW 1 : 4 6 4 - 5 0 4 , 7 3 8 - 7 4 2 . ) 3. Martinet, A. (1936). "Neutralisation et archiphonème". TCLP 6: cf. here p. 49. 4. Cf. Note 2, Jakobson, R. and Halle, M.

On Foreign Phonological Features in Present-day English

The Modern English stock of words is divided, from the phonological point of view, into two large groups. One group comprises words implemented by phonological means of purely genuine character (cf. owe, i f , ask, end, low, wine, son, lend, street, strand, utter, fellow, foreign, enough)·, the other consists of words characterised by peculiar phonological features not shared by those of the former group. Most of these words (e.g. vine, vain, zone, general, liquid, examination, idea) are loans borrowed from Latin and French at different periods in the development of English, but the provenance of words is outside the scope of the present investigation, which is concerned only with their present-day phonological structure. It is obvious that neither the former nor the latter group can be identified simply with the genuine and foreign descent of English vocabulary, respectively. Many words of undoubtedly un-English origin, such as pit, street, wine, flower, flour, air, aunt, fruit, letter, saint, liquor, pharaoh, phlegm, phase, piquant, do not exhibit any foreign phonological features, and belong therefore to the former class, even if the spelling or meaning of some of them is still suggestive of their origin (cf. suite : sweet, reign : rain, pharaoh : fellow), whereas some words of English extraction (e.g. vat, vixen, boy) in addition to many words of expressive or onomatopoetic character (e.g.giggle, titter, tehee, blubber, peep, kick, j i f f y , twitter, etc.) must be placed in the latter group. Nor can the relative frequency of words be a reliable criterion for their inclusion in the one or the other group, since many words containing peculiar phonological features may have a frequent occurrence, while some words of purely English phonological structure occur very rarely. If we speak of 'native' and 'foreign' words in preference to 'phonologically unmarked words' and 'phonologically marked words', both terms will be used for brevity's sake only in the sense of their difference in phonological organisation. It is understood, of course, that there

Originally published in volume of papers In Honour of Daniel Jones, London, 1964, pp. 185-190.

On Foreign Phonological Features in Present-day English

157

is no Chinese wall between the two categories in the structural history of the language. In fact, many foreign words lost their foreignisms as a result of the phonological development of English, and joined the group of native words, which constitutes the basic vocabulary of any language.1 There is, perhaps, no language devoid of phonological deviations from the native patterns, since they serve the needs of speakers in giving a sort of relevance, or a peculiar atmosphere, to the words in question by differentiating them from the common stock of phonologically homogeneous, unmarked words. What foreign phonological features are to be found in Present-day Standard English? They are many, and may be classified as follows: (i)

Unusual phonemes, as the Scottish ¡x¡ in loch and the interjection ugh l3 (cf. measure,leisure, vision, occasion, luxurious), which also took place in the juncture of two words (e.g. as you), it must be considered a native phoneme, the marked term of the opposition Iii ~ /3/. (d) The postvocalic position of the phoneme /// : boy, coy, moist, voice, void. (iii) Monomorphemic consonantal clusters: (a) Consisting of two stops or two fricatives, or a stop + a homorganic fricative, e.g. apt, act, strict, smaragd; sphere, aesthete, phthisis, naphtha, pfennig. In addition to these there are some other consonantal clusters such as /v/ + a stop or a fricative, /0/ (or /9/) + a stop or a fricative, and combinations . consisting of two nasals, which do not belong to the inventory of native consonantal combinations. It is noteworthy that the foreignness of some of these clusters consist only in the absence of the morphemic boundary between their consonantal components, since the same clusters appear in bimorphemic words such as leap/t, robb/ed, pick/ed, begg¡ed, cloth/s, lath/s, cloth/es, cup/ful, fìf/th, off'/set, and at the juncture of two contiguous words in a context, e.g. less five, less thin, if then, up to. The consciousness of their phonological foreignness varies, therefore, with the extent to which the speakers are conscious of the separate identity of morphemic component of words. Their foreignness is consequently felt more keenly in their initial position than in the final or medial position. (b) Consisting of three or more homomorphemic consonants, except the clusters /s/ + /pr, tr, kr, kl, kw, mj, pj, tj, kjj ; a nasal + /br; dr, gr, bl, gl, dw, pt, sk, zd/\ \m, p, l, n¡ + /sí/, which must be considered native.5 (c) Consisting of two monophthongs e.g. tiara, kiosk, piano, meander·, chaos, vehicle, bivouac. (iv)

Unusual position of some consonantal clusters: (a) Occurrence of /w/ between a consonant and an unstressed vowel, e.g. sequel, sequence, liquid, colloquy. The retention of /w/ in-ward,-wards after a consonant is obviously due to the spelling and/or to the consciousness of the separate morphemic identity of-ward(s). Cf. conduit [knndit], biscuit [biskit].

On Foreign Phonological Features in Present-day English

159

(b) Occurrence of monomorphemic clusters (with the exception of /nt, nd, st/) after an unstressed vowel. Examples: orange, messenger, matrix, onyx, herald, ribald. Consonantal clusters occur in this position only in a few words of Germanic origin, as in errand, island, husband, thousand, earnest, harvest. The insertion of ¡η/ in nightingale (OE nihtegale) is undoubtedly due to the need of Middle English speakers for expressiveness. (v)

Double occurrence of the same consonantal phoneme in monomorphemic words: Ipl : pip, pepper, poplar, pap, pup, peep, poop, pauper, purple, purpose, paper, pope, pipe, prep, prop, pupil·, Ibi : bib, barb, baby, blab\ /1/ : tilt, tint, tent, tetter, totter, teat, toot, tort, tart, taste, toast, tout, stout, title, tattoo, tomato, trot, tract, treat, trite, twist, twit'Jd/ : dad, dud, dodo, dude-Jkl : kick, cook, cock, cork, cake, coke, click, clique, clock, cloak, crick, crook, creak, crack, crake, conch, conquer·, IgI : gig, gag, gargle, goggle, grog·, Ifl : f e o f f ,fife,fluff,forfeit-, Ivi : valve, velvet, verve, vive; /s/ : cilice, cist, siskin, cess, sense, system, cease, sauce, source, sice, society, slice, sluice, space, sparse, spice, stress -, ¡dzl : judge ·, /r/ : rarity, rearing, roaring, rural·, HI : lily, loll, lull, leal, lilac, lapel; /ml : memory, malmsey, marmalade, marmot, mine, miñosa, moment, murmur, In/ : nun. The phonemes //, w, h, θ, ζ I do not re-occur in the same morpheme, and the double occurrence of /9, / , ? / / , and /J3/ is limited to a small number of words, such as thither (OE t>ider/, church, judge, George, backsheesh. It is remarkable that the re-occurrence of voiceless consonants in the same morpheme is found more frequently than that of the corresponding voiced consonants.3 This peculiar distributional feature of consonants occurs in many words of expressive character (cf. pip, peep, pipit, blab, twitter, etc.). A few English words of Germanic etymology in which the same phoneme re-occurs go back either to originally bimorphemic words (cf. hundred, dread, tight, deed, dead, thither, since, sister), or to reduplicated forms of expressive character (cf. quick, from */kwikwaz/, originally */kwiwaz/ from IE */gwiwos/). As far as I can see, only two English words, the numerals six and nine, are old IE words (monomorphemic?). (vi)

Morphemes consisting of more than two syllables:4 e.g. hospital, paradise, element, pelican, etc.

(vii) Primary stress (a) on the non-initial syllable of monomorphemic words, such as sincere, severe, lapel, (b) on a suffix, e.g. trainee,

160

Synchronic

Phonology

princess, locale, quintet, kitchenette, and (c) on the second or third stem syllable of words derived by suffixes, cf. familiarity, photographer, position, etc. The alternation of stress, conditioned by suffixes or an increase of final syllables, is therefore always a foreign phonological feature in the structure of morphologically related words. Cf. prince : princess, photograph : photographer : photographic, family : familiar : famüiarity, excel : excellent·, convene : convenance. (viii) Full vowels and diphthongs in unstressed syllables: e.g. in asphalt, chaos, epoch, despot, record. (ix)

Unusual consonant alternations: cf. invade : invasion, social : society, patient : patience.

The foreign phonological features cited above characterise a large portion of the English vocabulary. Their foreignness is obviously on different levels according to their varying remoteness from the native phonological patterns on which they rely. Whereas many of them are felt to be almost imperceptible modifications of native structural and distributional possibilities, others serve the purpose of making words more expressive in contrast to those of ordinary phonological organisation, and still others appear unbearably or uselessly foreign, so that they are suppressed or replaced by usual, or less unusual, features (cf. the substitution of a vowel plus a nasal consonant for French nasals, the use of /fc/ for ¡ χ / , the droppings of initial stops in the initial clusters ¡gz-¡ (in Xerxes,xebec,xylophone), ¡pf-,ps-, tm-, dm-,dv-¡ (in Dvorak), ¡kn-, gn-, gm-¡, etc.). The extent of adoption and toleration of foreign phonological features is always dependent on the resources of the adopting language. It is obvious, e.g., that Chinese speakers, whose words never begin with consonantal clusters and never end in a consonant (except a nasal), must replace the clusters in loanwords from English by syllables, whereas the English can retain unusual clusters in Latin words without much difficulty, because they have them at the boundaries of morphemes and words in their own language. The elimination of preconsonantal (or final) ¡r¡ which, never occurring in Southern English pronunciations, is also automatically dropped in loanwords, seems to be the only notable exception. If we hold that foreign phonological features are reflected against the solid basis of the native phonological pattern irrespective of the origin of words in which they occur, the question may be raised concerning the criteria which enable us to state that, e.g., the cluster /s0/ is foreign and the cluster /si/ is native. Such criteria, based on the structural analysis of language, do exist and are confirmed by the popular speech-instinct which clearly distinguishes

On Foreign

Phonological

Features

in Present-day

English

161

a d o p t e d features — even t h o s e o f l o n g - c o n t i n u e d e x i s t e n c e — f r o m the basic structural e l e m e n t s o f t h e language, the latter b e i n g internalised b y all individual speakers o n a deeper level than the former. There is n o danger for a n y language, s o far as it exists, that the a d o p t e d p h o n o l o g i c a l features m i g h t d e s t r o y its balance and bring the germ o f instability o r even disintegration t o its basic p h o n o l o g i c a l pattern. The never-ceasing p h o n o l o g i c a l f e r m e n t a t i o n in t h e history o f languages c a n n o t be b r o u g h t a b o u t b y a d o p t e d foreignisms, b u t o n l y b y the m o v e m e n t o f their o w n structural e l e m e n t s , while the former m a y b e tolerated or assimilated b y t h e m . A s structural linguistics e x t e n d s its range, t h e interaction o f these t w o p h o n o l o g i c a l strata, w h i c h necessarily coexist in every language and are especially characteristic o f English as a d o m i n a n t cultural language, will receive increasing a t t e n t i o n .

NOTES 1. For example, in Middle English dialects in which the corresponding voiced and voiceless fricatives [z] and [s], [v] and \ f ] , [3] and | θ ] belonged to single phonemes, the words containing single [s] or [/] between voiced sounds (cf. mason, basin, coffin, suffer) were foreign. As a result of the phonologisation of voice of fricatives such words lost their foreign phonological feature and became assimilated to the native pattern. The Latinisation of the ending -ive (for older - i f , e.g., in actif, passif) must have taken place after the phonologisation of //v. 2. Trnka, B. (1935). A Phonological Analysis of Present-day Standard English. Prague Studies in English, 5, Prague, p. 9 - 5 5 . (Revised new edition [edited by Tetsuya Kanekiyo, Tamotsu Koizumi], Tokyo, 1966,Chapt. 4, pp. 4 3 - 5 5 . ) 3. It is tempting to conclude that the double occurrence of the same phoneme in IE morphemes served only for expressive purposes, and that its use for other purposes in Greek and Latin (cf. Gr. kakós, pápyros-, L. populus, pòpulus, çaecus, bibo, coquo, quinqué, etc.) may have been due to the influence of other languages. 4. It is noteworthy that most Czech words have monosyllabic or even non-syllabic roots consisting only of one consonant, whereas many English monomorphem«: words consist of two syllables, e.g. father, daughter, river, follow, arrow, barely, etc. Monosyllabism, characteristic of Modern English, is evidently due to: (1) a small number of productive suffixes, some of which consist of a single consonant or a cluster of consonants; (2) the aversion of the language to the use of multiple suffixes and verbal prefixes; (3) the extensive use of prepositions and other 'small' words which are mostly monosyllabic; and (4) the assimilation of long words to the usual pattern of the domestic ones (e.g. perambulator > pram, examination > exam, etc.).

The Phonemic Organisation of Morphemes

The phonological system of a language consists of a certain number of phonological oppositions among the items of the phonemic inventory of which the language disposes. For example, in English, ρ is in opposition to any other phoneme of the linguistic system, and the same is true of t, d,m,s, etc. According to the number of relevant features by which one phoneme of a language is distinguished from another, two main classes of phonemic oppositions are to be distinguished. The distinction between the terms of one or more oppositional pairs consists either in one and only one relevant feature, such as the absence versus the presence of voice (e.g., pv.b,t v. d, k v. g, f ν.ν,θ v. d,s v. z, f V.3 , tf v. dz , in English), occlusion versus friction (e.g., ρ v. / , b v. ν, t v. e,d v. 3), and a relevant localisation (e.g.,ρ v. t , f . v. 0, b v. d,v v. d,p v. k, b v.g, etc.), or in the combination of two (or even more) relevant features such as absence/presence of voice + occlusion/friction, e.g., pv.v,b v. f , t v. d,d v. 0, in English. The terms of oppositions of the latter class may be distinguished from each other even by a combination of two (or more) relevant features one of which never constitutes any oppositional pair of the linguistic system by itself, e.g., ρ v. r,p v. s, bv.r,p v. /, etc. 1 If we know that phonemes are constituted by their relevant features, we can also see that the morphemes of a language are distinguished from each other not only by the phonemic oppositions in which the language takes part (cf. pit v. bit v. fit v. lit v. wit, in English), but also by what the present writer calls phonemic contrasts. The phonemes of which a morpheme consists are in contiguous or discontiguous contrasts to each other, the p h o n e m e / i n the word friend \frend] contrasts not only to r, but also to the other phonemic components of the word. Phonemic contrasts have different degrees of contrastiveness. The maximum contrast is between a vowel and a consonant, while two (or more) consonants forming clusters exhibit various degrees of

Originally published in Acta Universitatis Carolinae- Philologien 5: Phonetica Pragensia 1, 1967, pp. 9 1 - 9 3 .

The Phonemic Organisation of Morphemes

163

contrastiveness. From this point of view, the morpheme is not a mere sequence of phonemes equally contrasted to each other, but a hierarchical structure of different contrastive relationships between its phonological realizators.2 In examining consonant clusters and vowel combinations in a number of languages before the war, I stated the general rule that two phonemes differentiated from each other by one and only one relevant feature other than the place of articulation can never stand in contiguous contrast within the bounds of the same morpheme. So far as I can see, this rule of "the minimal phonemic contrast" allows no exceptions. It helps to account for the exclusion of such clusters as pb bp, td dt,fi>v f , etc. from English morphemes, for the non-occurrence of combinations of nasals and the corresponding oral vowels in French, long and corresponding short vowels in Czech, Slovak, and Magyar, hard and corresponding soft consonants in Russian, etc. All these clusters and vowel combinations which are below the limit of phonemic contrast in those languages are frequently found at the suture of two words or even morphemes (e.g. ίθ, dd in eighth, bathed), but if the suture happens to disappear in the development of a language, the law of minimal contrast begins to operate and the dissimilation or assimilation of both contiguous phonemes take place, e.g., Old English daet θε > Middle English thatte, > that. This law does not apply, as already mentioned, to the combinations of two phonemes distinguished from each other only by their localisation, e.g., pt, kt, tk, mp, mb, tn, nt, etc. which are frequent in Czech and other languages. Many languages restrict their contiguous phonemic contrast only to the sequence consonant + vowel, i.e., they have no homomorphemic clusters. If any consonants (mostly nasals) occur finally in those languages, clusters do come into existence, but they serve only as signals of the morphemic or word boundaries. In other linguistic systems in which homomorphemic clusters are admitted a number of rules are imposed to restrict their formation in greater or less degree. The following rules apply to English morphemes of the native stock: (1) Except for some foreign words, plosives (including affricates) do not combine with each other. The same holds true for fricatives, clusters like s f , f s , s6 occurring only in foreign words and at morphemic boundaries. (2) Fricatives/, ν, Θ, 9, / , 3 do not combine with the corresponding plosives or affricates. (3) The phoneme /ft/ is neither preceded nor followed by any other consonant, except ///. (4) The phonemes /r/ and /w/ are never followed, and the phonemes /9/ and /ji/ are never preceded, by a consonant. For Czech, restrictive rules can be stated as follows: (1) The phonemes /i/ and /t'l, /d/ and ¡d'l, ¡n¡ and /ρ/, /s/ and ///, /ζ/ and /3/, ¡ts¡ and ¡tf¡, ¡r¡ and /r/ never occur in contiguous neighbourhood. (2) The phonemes ¡t'¡, /d'I and Ini are never followed by a consonant. (3) /// is not preceded by a consonant, except

164

Synchronic

Phonology

ρ, b, v. Similar rules apply to the contiguity of vowels and consonants and consonant clusters. In many languages — even in those which have a rich inventory of clusters — the tendency towards avoiding the recurrence of the same consonant phoneme within the bounds of one morpheme is observable. Thus in English, nearly all morphemes of Germanic origin contain different consonantal phonemes, while the discontiguous occurrence of the same consonant phoneme is mostly a feature of their foreign origin or expressive character, cf. cook, baby, cease, judge,people, pope, bib, booby, gig,goggle, kick, lull, pep, etc. There is even a tendency to avoid the discontiguous occurrence of the phonemes which have the same relevant localisation, the words of this type (e.g. tide, tooth, death, five) being much less numerous than those of the ordinary type such as deal, feel, meal, make, come, home, friend, etc.

NOTES 1. Cf. the detailed discussion of the classification of phonemic oppositions in: Trnka, B. (1935). A Phonological Analysis of Present-day English. Prague Studies in English, 5. Prague. Chpt. 3. (Revised new edition [edited by Tetsuya Kanekiyo, Tamitsu Koizumi], Tokyo, 1966, Chpt. 3, pp. 12 ff.) Trnka, B. (1962). Rozbor nynèjsf spisovné anglictiny, I [Analysis of Present-day Standard English, I]. University textbook. Mimeographed, Prague, pp. 1 9 - 2 8 ^ 2. The theory of phonological contrasts helps us to distinguish phonemes from phonemic combinations. If there is a contrast between the segments of sound complexes such as st and /f in Czech, these complexes must be assessed as biphonemic clusters. On the other hand both Its/ and / f / / are phonemes, as their segments t-s and t-f are not contrastive to each other. Cf. op. cit. Note 1 : Trnka (1966). pp. 6 - 7 .

On the Relationship of Phonemes to Sounds

The 19th century linguist who saw linguistic history as a mass of isolated, gradual changes which continually transformed language never analysed language structurally as a functional whole. He often tried, to be sure, to explain changes in terms of various causes, but his finding of causes was often confused with the finding of specific conditions under which the change in question took place, and frequently he was satisfied with post hoc, propter hoc, i.e., with the examination of the temporal precedence of linguistic phenomena. Ferdinand de Saussure's theory of a sharp line between the synchronic and diachronic study of language brought no help to students of historical grammar, as it implied the statement that diachronic phenomena functioned only as destroyers of the well-ordered synchronic systems. In supporting the position that even in language the form of being is not the opposite of the form of action, the Prague school of linguistics has, since its inception, directed the attention of linguists to the fact that language at any period of its historical development represents a structural whole, the components of which must be compared and analysed in synchronic functional relationship to the whole cross-sections made in the historical continuity of language. Structural historical grammar was thus based on the functional comparison of successive linguistic systems stratified in time, and the clash between the diachronic and the synchronic grammar of a language, felt by the Neogrammarians, disappeared since both were concerned with the structural analysis oflanguage. In diachronic phonology, therefore, we speak of phonemic mutations including both paradigmatic changes (such as phonologisation of allophones or dephonologisation of phonemes, neutralisation or deneutralisation of phonological oppositions, and their various transformations) and syntagmatic regroupings of phonemes in words and morphemes. The question arises as to the relationship of phonemic mutations to

Originally published in Acta Universitatis Carolinae-Philologica, 1974, pp. 7 - 9 .

Phonetica Pragensia 4,

166

Synchronic Phonology

phonetic changes. It is obvious that some of the mutations that took place in the structural movement of a language may have been accompanied by phonetic changes, but one cannot say that any phonemic mutation necessarily brings about a phonetic alteration, or that the latter is always the forerunner of a phonemic mutation, as seems to have been the position taken e.g. by N. van Wijk and A.W. de Groot. 1 We can easily see, e.g., that the phonologisation of Old English and Middle English variants v, 9, and ζ which took place in all English dialects about 1400 had no connection with any phonetic change of the fricatives in question. We hold that sounds are realizators of phonemes, the smallest entities capable of distinguishing words, and the bond between them and the phonemes as their 'réalisés' leads us to the conclusion that their physiological alterations may have some influence on phonemic mutations, but that it is mainly the phonological system of a language which makes sounds subservient for its structural purposes and thus determines their changes. It must be borne in mind, however, that the bond of relationship between sounds and phonemes does not constitute a causal nexus between both, although it determines the specific status and development of sounds, deviating from that to which they are subject outside this relationship. A parallel case may be adduced from the historical development of the human written language. Pictures used as optical signs for intersubjective communication had a totally different development in the subsequent stages of human civilisation from that of the pictural reproductions of the same things and living beings outside their use of linguistic 'réalisants'. Employed for this purpose they had to adjust themselves to different conditions of writing materials, as well as to various changes in their 'signifiés'. In spoken language with its developed hierarchy of levels sounds are not, to be sure, direct realizators of words or morphemes, but only those of phonemes, so that the work of phoneticians and phonologists is done on the closely related terrain, and their collaboration is necessary for the advancement of linguistic sciences. If we are aware of phonemes as 'réalisés' and sounds as 'réalisants' with no cause-and-effect relationship between both, the linguists interested in the diachronic study of languages may ask which causes are responsible for obviously regular sound-and-phoneme correspondences in the subsequent stages of a language. As mentioned above, it is usually the morphological level which has a structural impact on the phonological level and its phonetic realisation, and as the collaboration of the levels does not take place in terms of causal relationship, we come to the conclusion that mutations must be viewed in terms of teleological character, due to the nature of language as an adequate internalised tool of intersubjective communication. In order to be adequate, language undergoes mutations which strive for a certain goal by the way of least resistance. This explains the fact that in linguistic develop-

On the Relationship of Phonemes to Sounds

167

ment the semantically neutral phonemic level is relatively easily subject to change, while the system of morphological and that of syntactic oppositions, which are not so easily suppressed or created, are much more constant. Evidence for this is in the breakdown of a language into dialects: this occurs first in its phonology and morphonology against a relatively solid background of its morphological and syntactic oppositions.

NOTES 1. Wijk, N. van (1939). "L'étude diachronique des phénomènes phonologiques et extraphonologiques". TCLP 8: 297-318. Groot, A.W. de (1941). "De klankwetten in het licht dei structurelle taalkunde".

SECTION THREE

Statistical Linguistics

Quantitative Linguistics

In present-day linguistics ever more attention is being given to the analysis of the quantitative data of linguistic events. The conviction is growing that we cannot explicitly determine linguistic reality if we do not also look at the quantum which is its necessary constitutive component. We must take account of this component in linguistic theory itself because the quantum idea is already inherent in the very definitions of the fundamental linguistic oppositions which, to a certain extent, are determined by quantitative considerations. We also admit that a quantitative fact may appear as a qualitative factor (so e.g. the repetition of words may be a certain qualitative feature of a poem) or that in the course of time a quantum may change into a quality. No concrete linguistic fact can be understood without quantum concepts ; if we want to understand a phenomenon fully, we must also analyse its quantitative relations. Neither is it without significance nor incidental, if e.g., one particular phonological opposition occurs in a linguistic system once, twice, or a hundred times. Linguistic science could never ignore the need for the characteristics of linguistic phenomena, but it was often satisfied — if it was not a case of specialised interests - with inexact statements like "frequent", "rare", "often", "exceptionally", etc.; such statements are quite useful but they merely provide inexact information about the occurrence relations of the described data. It is the task of a new branch of linguistics to provide the numerical characteristics and to define more precisely the quantitative relations. This branch is usually called linguostatistics or quantitative linguistics. This new direction in linguistic research pursues further tasks too. It tries to elaborate new methods that should aid linguistic research by providing means of control and of verification of older findings, but it should also lead linguistics onto the paths of hitherto unknown qualitative relations which

Originally published in Czech under the title "Kvantitativni linguistika" in ÍMF 34, 1951, pp. 6 6 - 7 4 . Slightly abridged and revised. (Translated by V.F.)

172

Statistical Linguistics

could be confirmed subsequently by qualitative analysis. This controlling and heuristic value of statistical analysis results from the fact that we observe a discrepancy between the numerical relations that we expect on the basis of our present knowledge, and the actual figures. If we are faced with such a discrepancy we must explain it and revise the hitherto valid results of research. Only by achieving full agreement between quantitative and qualitative analysis will an all-round understanding of linguistic reality be guaranteed. Statistical procedures as a heuristic method are employed in linguistic research in the same way as in the social sciences when a complex or heterogeneous situation, which cannot be analysed only qualitatively, is being investigated. This is the case when analysing linguistic utterances which comprise phonological, morphological, syntactic and stylistic phenomena in very complex mutual relations and with latent tendencies so that they can be analysed only by these methods. Just as the social sciences, linguistics too does not investigate quantum for quantum sake, but always employs it solely in the analysis of relations and of problems of a qualitative character. We must, however, be aware of the difference existing between these two methods of approach. Whereas one method, qualitative linguistics, differentiates from the quantitative point of view only between the zero and non-zero occurrence of linguistic relations, i.e., between their existence or non-existence, the other method, quantitative linguistics, aims at the exact analysis of their occurrences. Linguistic scientists already made frequent use of numerical findings for specific purposes of their research, e.g., to solve certain phonological and morphological problems in some old literary monuments, metric problems, problems of literary history (dating of texts), stylistic problems, etc. I recall here A. Beer's studies about verbal aspect in Gothic. 1 At present quantitative linguistics is trying to establish its own method of research and looking for its own problems although researchers are aware that linguostatistics is only a branch of linguistics and that every linguist must pay attention to both linguistic tenets. Is such an endeavour not premature? It may seem that it would be advisable to wait and hold back in such an experimental approach until all problems of qualitative linguistics analysis have been solved. In order to analyse linguistic elements quantitatively, qualitative linguistics must first state them. On the other hand it is also evident that the recognition of qualitative relations is infinite, and that quantitative analysis cannot wait for a solution of all linguistic problems by qualitative means. Scientific research is in a permanent flux that is subject to social factors, and the relative progress in the various sections of science is mutually compensatory. Quantitative research is fully integrated in the development of modern linguistics. Phonology, structural morphology and syntax embrace living trends of thought which are characteristic for geometry;e.g., the phonological analysis of lexical structure. Quantitative linguistics is also linked with such ways

Quantitative Linguistics

173

of thought as they are most genuinely applied in arithmetic and algebra. 0 . Jespersen already wanted to demonstrate syntactic relations by means of mathematical formulae, similarly Baudouin de Courtenay and also Ferdinand de Saussure started from mathematics. Linguostatistic research is being carried out at present in several places simultaneously, both in Europe and in the United States. Let me point to some of its immediate concrete tasks. In principle there are two: lexical analysis and analysis of the morphological inventory of a language, and the analysis of utterances (texts). Both approaches differ in their goals and results. The former has as its starting point the lexical and morphological inventory of a language at a given period ; it investigates the quantitative relations between phonological and morphological constituents that the language uses for its lexical and morphemic resources taken as a whole. The results of this analysis are invariable under the ideal assumption of the material completeness and of methodic relevance for the linguistic system of a given period as this research excludes variable factors which result from the choice and frequency of words in contexts. Its significance rests upon the fact that the quantitative findings supplement the results of the qualitative analysis of texts. We may find out the exact quantity of the employment of the different phonemes, of phonemic combinations and of phonological oppositions in various positions in the word and even state the numerical correlations among them; we may statistically investigate the productivity of the various phonological means that are at the disposal of a linguistic system, and try to formulate their hierarchy of frequency. The delimitation of qualitative and quantitative analysis is quite sharp in this field of research: whereas the qualitative analysis of functional constituents (e.g. phonemes) aims at stocktaking and at finding out whether they appear in certain positions or are excluded from some other positions, quantitative analysis does not only aim at stating their existence, but also at investigating their exact numerical occurrence in the whole vocabulary of a language. Such investigations already exist, but a consistent frequency research which would counterbalance qualitative analysis has so far not been attempted for any language. For some time to come we may still have to be satisfied with partial investigations in view of the topical needs of research and of the possibility of comparing several languages which makes these results more interesting and more significant. Such broadly planned research projects that meet these requirements can only be carried out as collective undertakings for every individual language. What should be the methods employed in such projects? We shall try to describe them, but limit our observations only to phonology. The whole language material is to be divided into monosyllabic, disyllabic, trisyllabic words, etc. and then into groups according to the number and sequence of their phonemic constituents. 2 For example, monosyllabic and monophonemic

174

Statistical

Linguistics

Czech words have been classified into types V, VC, CV; VCC, CVC, CCV, CrC, VCCC, CVCC, CCVC, CrCC, CCrC, etc. In such a classification of the lexical inventory in which each type and each word structure within the given type has its definite place, it is not difficult to find out the exact degree of the employment of the various types, as well as that of simple phonemes, phonemic combinations and of phonological oppositions in their various relevant and also neutralised positions in relation to the various types and to the total of all types. This method complies with the requirement that a statistical investigation may include all functional constituents that have already been stated by phonology and that it may also help some phonologists in discovering phonemic data that have hitherto escaped their attention. By means of exact numbers linguostatistical research answers the question how many single phonemes and individual phonemic combinations or how many opposition pairs of phonemes in various positions of the word are necessary to compose the total lexical stock of a given language at a given period. The difficulties of such research are greater than we would expect. Let me point to two problems: the problem of homonyms and the problem of the morphemic composition of words. Homonyms (homophones) are lexical structures which are identical as to their phonological composition, but entirely different in their meanings: e.g. Czech Dán [Dane] -dán [given], plot [fence] -plod [pronounced with ¡ti, fruit], ν ode [dat. sing, water] vodè [participle to 'lead'], byl [pronounced /bil/: he was] -bil [past tense 'he beat']. Should we evaluate the homonyms Dán—dân, plot—plod as a single lexical item with two different meanings or as two lexical items? The answer is: the purpose of the statistical investigation is decisive. For the purpose of purely phonological statistics we shall not go wrong in not differentiating the semantics of identical phonological structures because we only count phonemes. If, however, we count morphemes or count words as lexical units and not merely as phonological structures, we must distinguish the homonyms because they are different words from the lexical and morphological point of view. A word is not only determined by its form or by its meaning, but by the firm link that exists between its form and its meaning. E.g. the Czech zámek [castle] with the meaning [a certain type of building] is a different word from zámek [lock] with the meaning 'a device for opening and closing the door'. Both words have no semantic elements in common, just as plot-plod, let [flight] -led [ice]. If the meanings of a word are linked with one semantic element we identify the realisers of such meanings as one word: e.g., dráha [railway] [a means of communication] and dráha [track] [path] represent one word because the basic semantic element is in both instances the same, it forms the basis for the identification of the word. On the other hand lexical forms which have the same meaning but even slightly different phonological structures, which I call 'diaphones', represent one and

Quantitative Linguistics

175

the same lexical unit, e.g. Czech ctyfka-ctyrka [the numerical four], péro-pero [feather], dvére-dvefe [door]. (The phonemic oppositions are Irl vs. Ir/, /e/ vs. le:/). These simple reflections point to the problem of the substance of the word. In order to distinguish two (or even more) diaphones of one lexical (or one grammatical unit) from the different word units it is necessary to find out whether the former have an identical function and whether the variation of their phonological realisation does never result in any change of their function. Another question that was highlighted by statistical research is the identity of morphemes. Czech words like väd-ce [leader] râd-ce [adviser] are easily segmented into morphemes, but the morphological segmentation of other words seems to be difficult. Do we have in Czech words like talir [plate], halir [penny], milir [pile] two stem-syllables or may they be segmented into two morphemes? The word r'tif [knight] may be segmented into ryt + ir because -if is felt as a suffix deriving agentive nouns (e.g. malif [painter] ), even though the basic morpheme ryt- does not exist in Czech. Although in the above-mentioned words -ir has not an agentive meaning, we hesitate to evaluate these words as morphologically indivisable words, because basic morphemes are almost always monosyllabic in Czech and because the form -ir re-occurs as a morpheme in numerous nouns. The problem of how to identify morphemes is of fundamental importance and deserves particular attention and is, of course, of special interest to linguostatistics. Another problem which should be investigated in the interest of linguostatistics is the problem of the delimitation of the single word and the multiword group. There are not yet many works dealing with lexical statistics, but they already indicate how important quantitative differences are for the comparison of characteristic features of languages. If we compare Czech and English we can observe one characteristic difference: the lexical types VC, CV, CVC, etc. are far more numerous in Czech, but they are less frequently employed than in English which seems to be more economical in this respect. In the analysis of the morphological segmentation of the word it is a surprising fact that English, inspite of its well-known tendency towards monosyllabic words, has a considerably higher number of disyllabic root morphemes, whereas Czech, which displays the tendency to longer words, as it has already been pointed out, has almost always monosyllabic stem morphemes (cf. water : vod-a\ widow : vdov-a\sofa : po-hov-ka\ fellow : sou-druh) and only exceptionally disyllabic ones {holub [holub], the first stem vowel being always short). Correspondences and differences between European languages, which are already apparent from some studies by V. Mathesius, J. Vachek, J. Krámsky,etc. 3 prove that the results of such a systematic analysis will complement our knowledge about the degree of structural relatedness of these languages.

176

Statistical

Linguistics

How far statistics seems relevant for historical research may be exemplified as follows: in the Czech word stock unvoiced consonants occur far more frequently than the corresponding voiced ones (p appears in a greater number of words than b). The only exception is h which is more productive than ch. This interesting fact cannot be explained from the synchronic point of view because the phonological (not phonetic) relationship /Λ/ vs. /chi is exactly the same as the relations /Z>/ vs. Ipl, /z/ vs. /s/, etc. But this is not an incidental occurrence; it leads us to the conclusion that a different time factor is reflected in the pair h/ch from that in the other pairs. Indeed the pair h/ch only arose in the course of the development in Czech; its date of origin is much later than that of the other pairs: g changed to h which entered into correlation with the older ch and thus established the youngest correlation pair ch/h. The history of this pair is known, but what is interesting to observe is that it left a trace in the quantitative relations of the present-day language. The minor lexical productivity of the pair /// and fei (in which If I is more frequent) is also a trace of its more recent origin. In Russian the 'hard' consonants are lexically more productive than the corresponding 'soft' ones. The only exception is /'which is more productive than ' h a r d ' f . Trubetzkoy explains it by the fact that the opposition l'I i is supressed only before e whereas other palatal oppositions are suppressed also in front of some consonants. 4 But this may not be the full explanation. The rather surprising frequency relation of the two phonemes is more likely a reflection of the varying depth in time which is so far unknown to us. If such a quantitative analysis of a present-day language seems to disclose traces in its development, we may expect that more hitherto unknown facts may be disclosed by quantitative samplings of older stages of languages, not only in the phonological sphere. Linguostatistics, however, does not only investigate the word stock. Far from it. More interesting and maybe even more important is the quantitative analysis of linguistic texts in every respect, phonologically, morphologically, lexically, and syntactically. In principle this analysis pursues two aims: (a) determining the contextual frequency of the various constituents, (b) determining their periodicity, i.e., stating the pauses by which the constituents in question are separated in a running text. Forexample, we may try to determine the occurrence frequency of the phonemes in texts of a certain length, or try to find out in which intervals (waves) vocalic quantities, monosyllabic lexical structures, sonorant consonants, etc. occur. The results of such an analysis — unlike those of the analysis of the vocabulary — will vary according to the size of the investigated texts. The aim is to exclude variable quantitative elements and to determine the occurrence constant, i.e., the frequency norm. The analysis must therefore be based on texts of such a size that the numeric results would remain constant or vary

Quantitative Linguistics

177

only slightly. The threshold of the approximate constancy varies from language to language and can only be found out on the basis of actual investigations. If we want to find out the order of frequency of the single phonemes which make out the Czech norm, it may suffice to investigate texts of the length of about 100 000 words, i.e., texts comprising about half a million phonemes. The results for less extensive texts vary as the two frequency counts of sounds by F. Hradilik and J. Sedlaöek have shown. 5 Hradilik investigated running texts of 10 000 words and found the following declining order of Czech consonants: njn, t\t\ v, s, I, r/r, did', k, m, p. j, z, b, h, ts, 3, ch, f , t¡. Sedlácek's investigation was based on more extensive samples, and therefore differs: s is in the third place, whereas ν occurs in the sixth, k precedes d¡d\ and the last five consonants with the lowest frequency manifest the declining order: 3,f,ts,ch, t f . Hradilik's and Sedlácek's investigations that were made for shorthand purposes do not comply with the exact requirements of phonological statistics, but they clearly indicate that the different results are due to the varying size of the texts which had been investigated. That is why it would be necessary to find the exact frequency norm on the basis of more extensive material, approximately 50 000—100 000 contextual words. The rough estimate that this number of words suffices for the determination of the quantitative hierarchy of Czech phonemes is based on the observation that only a small number of words which constantly reoccur and to a certain extent are common to all texts (representing 80—90% of the lexical material) is decisive for stating the frequency of phonemes and of the more common phonemic clusters. (According to F.W. Kaeding 1 100 words make up 87% of the contextual material consisting of 11 million running words 6 .) The phonemic structure and the frequency of these words is therefore important. The material of 10 000 words may probably be only the basis for finding the norm of the proportion of consonants and vowels which is in Czech 7 : 5 , and for finding the quantitative norm of the syllable. In my analysis of Czech texts of 10 000 words I counted 30.7% disyllabic words, 25.43% monosyllabic words, 21.65% trisyllabic words, 13.11% tetrasyllable words, 4.09% pentasyllable words and 3.89% non-syllabic words (prepositions k [to j, s [with], ζ [out of], ν [in] ). The order then is as follows: disyllabic words (or better lexical structures), monosyllabic words, trisyllabic words, tetrasyllable words, pentasyllable words, non-syllabic words. Lexical structures made up of six, seven, eight, nine syllables represented only 0.89% of all contextual words. The non-syllabic, monosyllabic, and disyllabic words made up 60%, together with the trisyllabic words 81.6%, with the tetrasyllable words 94.78% of the total material. This order seems to be the norm in the Czech linguistic system. The material of 10 000 contextual words may be therefore sufficient for fixing the frequency order of the most frequent lexical types. In my investigation the declining order goes as follows:

178

Statistical Linguistics

CV (na [on] ; almost 10%), CVCV (nese [he carries] ; 921), F (566), CCVCV (právo [law] ; 518); CVCCV (vadny [defective] ; 457), CVC(tak [so] ; 420), C (v [in] ; 397), CVCVC (potok [brook] ; 324), etc. The words with open syllables are evidently in the majority. The contextual material of 10 000 words may be also sufficient for stating the frequency of the occurrence of consonantal items. The total frequency of the phonemes s///z/3/ is higher than the total frequency of the phonemes tlt'/d/d'/, this however, is again higher than the frequency of w/n/· Then follow the other phonemes. The order of frequency of vowel phonemes will not considerably change within more extensive texts as the corresponding results in the counts of Hradilik and SedláSek indicate: e,i οτα,ο, u,ou. Any vowel occurs more frequently than the most frequent consonant, with the exception of the diphthong ou. The statement of the quantitative norms can be based on a more or less small number of contextual words. On the other hand the statement of the occurrence constancy for the inventory of consonant clusters (e.g. st, zd, pr, etc.) requires a far larger corpus of textual material than 10 000 words. Without doubt, statements about the occurrence norm for all constituents of the language system require assiduous research. Such work will be necessary. We must be able to compare with this norm the statistical findings from certain texts as well as the observations resulting from the quantitative analysis of the whole lexical stock. If we cannot get statements about the occurrence norm we are unable to make generalising statements about the results of our research and cannot maintain that they are valid for the linguistic system as a whole. They are then only valid for the texts which were the material base of our research. In agreement with the Geneva school of linguistics we may say that the discovery of the quantitative norm is within the orbit of the linguistic system, la langue, whereas the quantitative observations of the concrete utterance (text) within the orbit of speech, la parole. The notion of the quantitative frequency norm corresponds to what G.K. Zipf called the balance 7 and W. Bull the natural frequency 8 .1 have reached my conclusions quite independently. The quantitative norm is not exactly observed in all individual utterances (texts), it only represents a trend which the various texts seem to follow and it is a characteristic marker of these texts if they differ from it. Although it is necessary to distinguish linguistic systems also in quantitative linguistics and not to mix a quantitative diachronic approach with a quantitative synchronistic approach, I do not want to exclude the possibility that there exist general laws which regulate the quantitative structures of all languages. It may seem premature to attempt to formulate these laws in the present state of quantitative linguistic research; yet such attempts have already been made. The first attempt to this kind goes back, as far as I know, to the end of the last century, when the French psychologist B. Bourdon

Quantitative Linguistics

179

compared eight European languages including the then less known languages Hungarian, Celtic, and Russian; he observed that these languages display the highest frequency of consonants, whereby the dentals appear three times as often as the labials, the labials twice as often as the velar consonants. 9 This attempt is only of historical interest today just as is J. van Ginneken's interpretation; Ginneken combined frequency with racial differences. 10 Trubetzkoy (note 4) is much more critical as he only refers to trends and he points at the same time to the exceptions in various languages. He maintains that the unmarked phonemes display generally a higher frequency than the corresponding marked phonemes. So, for example, in Czech the voiced members of consonant pairs are rarer than the unvoiced ones, in Russian the 'soft' sounds are rarer than the corresponding 'hard' sounds. Zipf also says this. But there are exceptions to this trend. Trubetzkoy himself points to the fact that Russian 'hard' I occurs more frequently than 'soft' I. In quantitative lexicology quite a few interesting and important observations have been made. The first scholar who tried to formulate laws referring to word frequencies in running texts was the Frenchman J.B. Estoup. 11 This stenographer discovered that in every longer text the total of the frequency of a word and its order of frequency is approximately constant. His law which may be expressed by the formula f.p = C was verified by G.K. Zipf on the basis of several texts, especially on the basis of J. Joyce's well-known novel Ulysses (note 7). In this novel, as M.L. Hanley observed, 12 there are 206 430 contextual words and 29 899 recurrent lexical items; the average contextual occurrence of each word is roughly 8.7. If these individual words are ranked according to frequency in declining order so that the most frequent word appears in the first place (p = 1) and the hapaxlegomena in the last places, the product of the frequency (f) and the rank of frequency (p) in the case of each single lexical item is constant. For example, a word with rank frequency 10 occurs 2653 times and if we multiply its frequency with the frequency rank, 2 653 χ 10, we get the figure 26 530. Fifty most frequent words occur 556 times and according to the same formula 556 χ 50 = 27 800. The hundredth most frequent word occurs 265 times, the thousandth most frequent one 26 times;in most instances the product f.p is again approximately the same (26 500 and 26 000). If the absolute frequency and the rank frequency of each word in Ulysses are entered on the co-ordinates in such a way that the frequencies are entered vertically, the corresponding frequency ranks horizontally, the resulting line will pass at an angle of 45°. Zipf s law also appeared to be valid for the analysis of material that was collected by R.C. Eldridge from the daily American press; it amounted to 43 989 contextual words (running lexical occurrencies) of 6 002 single items; every word appeared on the average 7.3 times. 13 Zipf states that the investigation of texts in other languages (Chinese, Old English, Gothic, etc.) confirmed Estoup's

180

Statistical

Linguistics

law. However, this law does not exactly formulate the occurrence of words in utterances (texts), it only indicates the tendency which appears to work within the broadest outlines; it may differ from language to language. Zipf maintains that the law is valid only for more extensive texts. This fact enables us to attempt its linguistic evaluation. In more extensive texts words of grammatical (structural) character occur frequently, e.g. the prepositions o f , to, at, the pronouns, the articles, auxiliary verbs, conjunctions; i.e. their frequency ranks will not change considerably. If this is the case, we understand why the product of frequency and rank appears constant with these words in larger samples of texts. The same tendency is shown with words of general meanings, as e.g., thing, man, etc. and with words (special terms) which display a higher frequency of occurrence in view of the topic of the text. These words tend to be evenly distributed. Is Estoup's law also applicable to contextual words within a lower frequency of occurrence? These words which occur only a few times or even only once are rather rare. Kaeding discovered in his Häufigkeitswörterbuch, (note 6) which is, as I stated above, based on a count of 11 million words, a considerable number of words which were hapaxlegomena, or occurred only twice or three times whereas three different words der, die, und represented 9.47% of all contextual words, 15 different words one quarter, 66 such words 50% and 320 words 75% of the total of the textual material. Is therefore Estoup's law also valid for words with a very low absolute frequency? I would maintain that it is not. Estoup's law only displays a tendency in the language to establish a balance between grammatical words, words of general meaning and the more frequent words due to the topic on the one hand, and all the other words on the other hand. This tendency is evident only in larger samples of texts, and it may vary according to the typological structure of the language. In Czech, which is a highly synthetic language, and which makes less use of synsemantic words than English, this tendency is more reluctantly apparent than in English. Estoup's law is complemented by another law which was first formulated by Zipf. According to Zipf s law the product of the absolute frequency number raised to the square ( f 2 ) and the number of words with the same frequency (n) is under certain circumstances, which have not been defined, constant. The formula is f 2 - n = C. If 1000 words appear in one and the same text once, 250 words will occur twice, 110 words three times, 63 words four times, etc., always according to the formula f 2 n = C. If we try to interpret this numerical formula linguistically we see that it attempts to express the relationship between the number of different words and their frequencies of occurrence in a larger textual sample, a relationship that indicates a certain compensation between the number of words and their frequencies; a high frequency of words is balanced by a small number of words, and vice versa a low frequency is compensated by a higher number of words. If Zipf s law were universally

Quantitative

Linguistics

181

valid, it would be an important aid as it could serve as a control of statistical work. It would be sufficient if a statistician knew for a larger contextual sample the number of different words which are hapaxlegomena, he could then on the basis of the equation f 2 - n = C in which η is unknown, compute the frequency, i.e. how many words occur in the same text two times, three times, etc. It seems, however, that in this respect Zipfs law was a failure. This means that Zipfs equation itself is not really correct. Hence it is necessary to revise this law, a task that can be solved by mathematicians in collaboration with linguists. Zipf also discussed the periodicity of word repetitions, i.e. the intervals between iterative words. 14 In this context it can be demonstrated how the frequency of one word varies. An experiment carried out on the basis of Kaeding's count showed that in 45 sections of 140 000 words each the frequency of the preposition and prefix zu varied between 974—1950, that of zur between 8 1 - 4 4 6 , that of es between 2 7 8 - 1 2 5 4 , that of endlich between 9 - 1 3 1 , that of zurück between 7 - 1 2 2 , etc. I do not want to discuss further Zipfs law and other formulae that he discovered. I only want to refer to the equation that is stated to express that relationship between the distribution of a word and the distribution of its various meanings (mj = f 2 ) ; accordingly the m number of different meanings of a word and the frequency of a word approximately equals the square of its frequency. Z i p f s remarkable effort to state the quantitative laws underlying utterances (texts) is based on the assumption that the numerical data are in causal relationship with the linguistic form which, as Zipf thinks, is dynamically determined by them. I do not think that this assumption can be upheld in the concept of a static grammar and it is also doubtful for a dynamic grammar. Zipf formulated among others the rule that the frequency number of words in utterances is indirectly proportional to their length, i.e., the longer words are, the rarer they occur. It must be admitted that in all languages there is a tendency that only shorter phonic segments are often repeated, not longer ones. Pronouns, adverbs, auxiliaries, prepositions and conjunctions which must be often repeated because of their grammatical functions - just as inflexional endings — are composed of small phonemic sequences in all languages; longer sequences of such semantically insignificant words would conceal the semantic core of the utterance and thus weaken its communicative and expressive effect. We may observe this tendency towards a certain norm of the size of words in our own stylistic encodings when we choose shorter synonyms to express constituents which do not seem of particular importance for the communication, whereas for expressing ideas that are intended to incite the interest of the listener we choose longer synonyms which because of their forms appear to be more relevant. Of course, every language has its own norm of word size. The norm also differs according to the part of speech to which

182

Statistical Linguistics

the word belongs. The clippings of long foreign words in English which is explained b y Zipf on account of the causation of frequency is not immediately linked with their occurrence frequency in utterances but with the effort to set norms for their length in accordance with the usual size of monomorphemic native words. There is certainly some causal link with the frequency use of the word, but the frequency of its use by itself is not the cause for its clipping. The link, or the relationship, between quantity and quality certainly exists, but it is of a different nature than the relationship between immediate cause and effect. We have discussed statistical linguistics as a branch of contemporary linguistics and shown that it has its own subject, method and problems. In conclusion I want to mention the practical applications of linguostatistical research in various fields of interest: shorthand and the construction of shorthand systems, the construction of languages for international communications, of cipher codes, and — last but not least — foreign language teaching and spelling (e.g. in English, Chinese). These applications of linguostatistics vary according to their special nature. I only want to make a few observations about the significance which statistics has for the teaching of modern languages. 15 Many statistical investigations have been made for this purpose. Some researchers conclude that it would be sufficient to learn words the frequency of which covers about 8—9 tenths of a rather extensive contextual corpus as this would enable the learner to speak, read and write in the foreign language, and that these word frequencies would also determine the order in which they are to be acquired. The number of these absolutely necessary words is rather small. P.B. Diederich found that 1471 words represented 83.5% of the lexical corpus in Latin texts that he had examined. 1 6 Linguostatistics clearly demonstrates that a learner must learn especially these words, if he wants to acquire the foreign language in the shortest possible time, and that he must learn them well by constant repetition. If he makes mistakes in those parts of utterances that permanently re-occur, the periodicity and the absolute number of his errors is higher than if he made mistakes in those parts of utterances that re-occur less frequently. The study of variable constituents of utterances must be planned in a regular and even manner. This planning is guided by the teaching goal and it may be different for grammar school students, or for medical doctors, businessmen, or for future teachers of English in grammar schools. The method of practical foreign language teaching that is founded on statistics is not new, but the recent statistical research carried out by American and Russian educationalists justifies it, as it provides the teachers and foreign language textbook writers with useful aids: cf., e.g., the frequency dictionary by Thorndike-Lorge which gives the frequency of every word in four columns according to the genres of juvenile literature, and in the fifth column it states the total frequency. 1 7 Significant

Quantitative Linguistics

183

are also the statistical investigations of words and forms especially in teaching languages with a more difficult grammar, as e.g. Russian. In spite of the usefulness of such aids word frequency statistics are overvalued in the methodology of modern language teaching and in mother-tongue teaching. Critical comments were published both in the USA and in Europe. No doubt the highly frequented contextual words make up 80—90% of all utterances (texts), but the didactic conclusions based on this fact that the knowledge of this number of words guarantees the understanding of 8—9 tenths of all utterances is not correct. As W. Bull says (note 8), one forgets that these words have several meanings and statistical research should also endeavour to state the frequencies of these meanings, not only the frequency of the common morphological forms because some of the meanings of a frequently used word are very often used, others only rarely; it is, therefore, necessary that a student should learn the words in their most current meanings. But apart from the semantic variations of a word the occurrence of which could be statistically fixed, this conclusion is also not justified from another point of view: a person that knows 50 English words, which according to Ch. Voelker 18 make up half of the lexical corpus in spoken English, does not understand 50% of utterances in spoken English, simply because these words are rarely the carriers of the important semantic constituents of the communication. Similarly in Czech the words byt [be], a [and], ν [in], se [himself], na [on], tam [there], ζ [out o f ] , ie [that], do [into], on [he], s [with], o [about], ktery [who], k [to], tento [this], za [for], i [too], jení [who], svuj [his], mit [have], which appeared in the investigation of J. and H. Korejs, 19 are semantically less significant, because their knowledge does not imply the comprehension of 20% of a Czech context which they help to make up. A context in which the fully semantic words would be omitted and only the words that in their majority are grammatical expressions are left would not be understood; but a reader may understand a context in which these syntactic and semantic fillers were omitted and only the fully semantic words were left. The latter are the carriers of the meaning of the whole utterance, they are its'raison d'être'. The 1471 Latin words which in the context analysed by Diederich represented 83% of the contextual material (note 16) include many fully semantic words, and their number rises disproportionately with higher percentages. R. Michéa maintains that 3 000 English words enable the comprehension of 95%-97% of a text, but that an increase of 1% requires already the total knowledge of 6 000 words. 20 These reflections are not intended to deny the importance of word counts as a means of making language teaching methods more precise, especially when these counts are so advanced as to help us to state exact norms. So far they clearly indicate that the less frequent word material must be studied in smaller doses than the highly frequent words; that no text in a foreign language textbook must

184

Statistical

Linguistics

be overloaded with special terms, but only in a small proportion to the number of words with normal frequency which the author had fixed for its textbook. The main aim of the textbook writer and of language teaching must be the proper choice of topics and of the sentences which correspond to the situations and to the communicative need, and not the teaching of words for the sake of words. The choice of topics and of the conversational sentences is the condition for and determines the choice of lexical items; whereas statistics has only regulating significance. Quantitative linguistic research has already made significant contributions. Until recently it was only a method of linguistic research — figures have often helped to prove linguistic facts and linguistic theory. It will be a manifestation of maturity of modern linguistics if it becomes an established branch of linguistics. But as it was the case with phonetics — the only branch of linguistics that uses experimental methods — it must also be said of linguostatistics that it is only a branch discipline of one science, linguistic science. The statistician, just as the experimental phonetician, must be a linguist in the first place, otherwise he would not be able to direct his analysis towards linguistic problems and he would not be able to evaluate his research results linguistically. He is not concerned in the first place with figures, with probability counts or with statistics itself, but deals with linguistic data and facts. The linguists are able to make use of statistical methods for the advancement of their sciences as other researchers used them for the advancement of social sciences. In our days we stress the importance of group work, of collective research. Research in quantitative linguistics which employs modern technical devices requires more than the efforts of one individual and it is also a highly time-absorbing undertaking of collective work. 2 1 This has been generally acknowledged in international linguistics. The Sixth International Congress of Linguistics which was held in Paris in 1948 took the initiative when, acting on the proposal of Marcel Cohen, it appointed a nine-member committee in support of linguistic statistics. 22

NOTES 1. Beer, A. (1915). Tri Studie o videch slovesného deje ν gotstinê. [Three studies of aspect of verbal action in Gothic]. Prague. 2. Trnka, B. (1949). "K vystavbë fonologické statistiky" [The build-up of phonological statistics! · SaS 11: 5 9 - 6 4 . 3. Mathcsius, V. (1929). "La structure phonologique du lexique du tchèque moderne". TCLP 1: 6 7 - 8 4 . (Reprinted in PSRL, (1964), pp. 1 5 6 - 1 7 6 ) . Mathesius, V. (1931). "Zum Problem der Belastungs- und Kombinationsfähigkeit der Phoneme". TCLP 4: 1 4 5 - 1 5 2 . (Reprinted in PSRL, (1964), pp. 1 7 7 - 1 8 2 . ) Vachek, J. (1940). "Poznámky k fonologii ceského lexika". [Notes on the phonology of the Czech lexis]. Listy filologické. Prague, 60: 3 9 5 - 4 0 2 .

Quantitative

4. 5.

6. 7.

8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.

Linguistics

185

Krámsky, J. (1942). "Pfispëvek k fonologické statistice stare a nové anglíctiny". [Contribution to the Phonological Statistics of OE and ModE]. CMF 28: 376-384. Krámsky, J. (1948). "Fonologické vyuzití samohláskovych fonémat". [Phonological exploitation of vocalic phonemes]. Linguistica Slovaca IV- VI, Bratislava, pp. 39-43. Krámsky, J. (1939). "A study in the Phonology of Modern Persian". Archiv orientâlni, Prague, 11: 6 6 - 8 3 Krámsky, J. (1947). "A Phonological Analysis of Persian Monosyllables". Archiv orientâlni, Prague, 16: 103-134. Trubetzkoy, N. (' 1939). Grundmge der Phonologie. TCLPl. Cf. Chpt.: Zur phonologischen Statistik, pp. 230-41. Holas, A. (1926, 1927). "Naskytání a shlukování hlásek (jich iterace a kombinace) ve slovenstinë a píirovnání s cestinou". [The occurrence and clustering of sounds (their repetitions and combinations) in Slovak and its comparison with Czech]. Tésnopisné Listy [Journal for Shorthand], Prague, 51: 4 3 - 4 9 , 58-62, and 52: 3 - 1 2 . The iteration and combinations of sounds in 10 000 Slovak words of running texts in comparison with Czech. Information of previous counts by Dürich, F. Hradilík and the author. Sedlácek, J. (1917). Zàkladni Studie k ceskému tèsnopisu, [Fundamental Study for the purpose of Czech Shorthand]. MS deposited in the State Shorthand Institute, Prague. Trnka, B. (1932). Pokus o vèdeckou teorii a praktickou reformu tèsnopisu. [An attempt at the theory of shorthand and its practical application to Czech shorthand]. Philosophical Faculty,Caroline Univ. Prague. Kaeding, F.W. (1898). Häufigkeitswörterbuch der deutschen Sprache. Berlin. Zipf, G.K. (1932). Selected Studies in the Principle of Relative Frequency in Language. Cambridge (Mass.). Zipf, G.K. (1935). The Psycho-biology of Language. Boston. Zipf, G.K. (1949). Human Behaviour and the Principle of Least Effort. Reading (Mass.). Bull, W.E. (1949). "Natural Frequency and Word Counts". The Classical Journal 44: 469-484. Bourdon, B. (1892). L 'expression des émotions et des tendences dans langage. Paris. Ginneken, J. van (1935). Ras en Taal. [Race and Language]. Amsterdam. Estoup, J.B. ( 5 1917). Gammes sténographiques. Paris. Hanley, M.L. (1937). Word Index to James Joyce's Ulysses. Madison (Wise.). Eldridge, R.C. (1911). Six Thousand Common English Words. Buffalo. Zipf, G.K. (1945). "The Repetition of Words, Time-Perspective and Semantic Balance". Journal of General Psychology 32: 127-148. Fried, V. (1950). "Linguistická statistika a hospodárnost ν jazykovém vyucování". [ Linguostatistics and the economy in foreign language teaching]. CMF 33: 157-145. Diederich, P.B. (1939). The Frequency of Latin Words and Their Endings. University of Chicago Press. Thorndike, E.L., and Lorge, I. (1944). The Teacher's Word Book of30 000 Words. New York. Voelker, Ch. (1942). "The One Thousand Frequent Spoken Words". The Quarterly Journal of Speech 28: 189-197.

186

Statistical

Linguistics

19. Korejs, J., and Korejsová, H. (1938). Slovník nastch novin. [The Vocabulary of our newspapers], Prague. 20. Michea, R. (1949). "Introduction practique à une statistique du langage". Les langues modernes 43: 29-42, 50-58. 21. A collective undertaking is the investigation of Russian; its chairman is H.H. Josselson (Wayne University) in Detroit; it has the support of the Rockefeller foundation. Its corpus includes novels, short stories, plays and journals, it is prepared in such a way that its results could be used by linguists for various purposes. The main emphasis is on Modern Russian, and so 50% of the materials are chosen from literature written after 1918. Josselson analyses 1 million contextual words selected from samples of 10 000 words of 100 different sources. The work, as it is planned, should be completed in 1952. It will also account for the frequency of derivational types, parts of speech, gender, number, aspect, etc. The word count has since been published (V.F.): Josselson, H.H. (1953). The Russian Word Count. Detroit (Michigan). 22. Trnka, B. (1950)./! Tentative Bibliography. Publication of the Committee on Linguistic Statistics. (C.I.P.L.). With an introduction by M. Cohen, Spectrum, Utrecht, Brussels.

The Distribution of Vowel Length and its Frequency in Czech

Vowel length in Present-day Czech has not been systematically examined in terms of statistical phonology. The following is perhaps the first attempt at summing up the most important results of the author's analysis of its distributional features, and, at the same time, the first account of some of the results obtained by him from his statistical analysis of Czech prose texts in the years 1940-43. The opposition of vowel quantity in contemporary non-dialectal Czech consists of three pairs of vowels i/i [ι:], u/û [ u . ] , and α/ά [α:] *. Cf. the words jistjjist fsure/to e a t ] , syt/sit [sated/to sow], dul/dâl [it blew/pit], hadujhadä [serpent, dat. sing./serpents, gen. plur.], rada !ràda [advice/glad, nom. sing, fem.], vina¡vina¡vinná [guilt/wines/she is guilty] in which it is the only mark of distinction between them. The further two vowel pairs e/é [e:] and o/ó [o:] are not integrated in the Czech phonological system of vowels so fully as the first three pairs. The causes of their incomplete integration cannot be discussed in full length in the present article and it may only be pointed out that é is either a phonological archaism of Standard Czech, e.g., péci [to bake], mile [pleasant, neuter n o m . ] , hole [bare, nom. neut.], kamënek [little stone] or a phonological feature which occurs in foreign, onomatopoetic and emotive words. In fact, we are not far from the truth, if we regard the distinctive role of é as a phonological foreignism which - in spite of its standardised usage in the pronunciation of Standard Czech — does not take an active part in the development of the non-dialectal linguistic system of Czech. Long è maintains its subsistence in Present-day Standard Czech only thanks to the purely phonological analogies ι : i = e : χ, u : ύ = e : χ, and a : ά = e : χ, since it coalesced with ί as early as in the latter part of the 15th century and no new é has come into existence since that time. As for o, which occurs only in a small number of words, its loose integration in the

Originally published in Prague Studies in Mathematical (Slightly revised.)

Linguistics

1, 1966, pp. 11 - 1 6 .

188

Statistical

Linguistics

phonological system of Czech is apparent from its having a purely emotional function in the affected pronunciation of some words (e.g. in Bóíel [Oh, God!],morel [Sea!] as well as from its exclusive occurrence in onomatopoetic words and foreign words, such as lóze [theatre box] versus loie [bed] and chór [choir] versus chorp [sick]. It is also to be noted that the phonetic realisation of both é and ó is slightly different from that of the fully integrated i, ú, and à, and that their length is subject to vacillation even from the phonological point of view, cf. pero-péro [pen] —dest'—dést' [rain], ν nasílózi [in our box], and therefore of subphonemic character in some words. 1 It may be safely stated that the standard language makes more abundant use of vowel quantity than non-standard Czech, the functional load of two vowels, í'and ú, being markedly more extensive in Standard Czech. As to i, its use was extended during the 19th century in consequence of the substitution of i for middle Czech ej < Old Czechs while spoken usage, regarded now as substandard, has kept up ej < Old Czech>> unchanged, so that such words as, e.g., âesky [Czech], byk [bull], my dio [soap], νy Ska [height], vyryt [to dig up] etc. are pronounced with /'in Present-day Standard Czech and with ej in the non-standard language. The substitution of i for ej in Standard Czech has been undoubtedly suggested by the spelling in which y is found to correspond t o y [i] as / [i:] does to i [i]. Similarly the Middle Czech diphthong ou < Old Czech û was replaced by ú at the beginning of words in the 19th century, whereas non-standard Czech continued to keep up ou as a legitimate reflex of Old Czech ú even in this position. Cf. údy-oudy [limbs], úl-oul [beehive], úhor-ouhor [fallow], úroda-ouroda [harvest]. It should be recalled, however, that the initial ú-, which has replaced Middle Czech ouin the refined pronunciation of Modern Czech, seems to win ground even in the speech of non-standard Czech and is exclusively used in words of literary provenance, such as úhoz [musical touch], úhlavní [sworn (enemy)], úder [smash], úpal [sunstroke], úsek [segment], even by less educated speakers. 2 The same is largely true o f y . In view of these facts, however, it is safe to conclude that minimum word pairs differentiated only by word quantity are more numerous in Standard Czech than in the non-standard language. As to the functional load of individual long vowels in Czech that of ó is the minimum one, while / is used more extensively than ά, which again is charged with more functions than é and Ú. The examination of distribution rules of long vowels in Czech leads us to the statement that they are very much the same as those of the short ones, that is to say, the opposition of vowel quantity is phonologically distinctive in any syllable of the word. There are, however, some important exceptions to this rule. First of all, long vowels do not occur before homomorphemic consonant clusters other than st, st', sk, zd, zd\ tr, tr, dr, dr, rt, rt\ mñ, and are also non-occurrent in the first syllable of disyllabic morphemes, cf. haluz

The Distribution of Vowel Length and Its Frequency in Czech

189

[twig], holub [pigeon], topol [poplar], tuleh [seal], biskup [bishop], cypfis [cypress], jestfâb [hawk], papir [paper], beseda [friendly meeting], zelezo [iron], etc. In addition to these distributional restrictions of vowel quantity, the neutralisation of the opposition i/i and u¡ú before / and before the tautosyllabic m is to be mentioned. Both I (j>) and ü (ä) lose their length in these environments in spite of the retention of their 'accents' in spelling, with the result that only short i and u appear in the pronunciation uninfluenced by the spelling. Cf. sije [neck], zabiji[they kill], navíjí [they wind u p ] , fijen [October] \mûj [my], tväj [your], lûj [suet], stüj [stop], stäjte [stop! plur.], üjma [detriment] \prosim [I ask], umim [1 know], svêzim [fresh, instrum.], novym [new, instrum.], znamenim [signal, instrum.] ; dû m [house], synäm [sons, dat. pl.], hochûm [boys, dat. pl.]. As to the nonoccurrence of long vowels before the consonant clusters of the same morpheme, e.g., in kult [cult], rumpàl [windlass], mince [coin], vers [verse], manzel [husband], it is doubtless due to the foreign origin of such words, and the same is true, in addition, of the quantity of the homomorphemic hiatus vowels, for instance in Leo, Laos, faetón, poeta, noetika.3 So far as vocalic length is exluded from the first syllable of disyllabic morphemes, this is no doubt caused by the lack of immediate contact of the syllable with the derivative or inflective suffixes of the word, since the opposition of vocalic quantity asserts itself only in syllables immediately followed by derivational suffixes or by inflective, including zero, morphemes, cf. celàdka [lot, pack], hromádka [little heap], hrách [peas], vrána [crow], zila [vein] ,pise [he writes], rise [realm], nésti [to carry], väle [will], vài [ox]. The sequel of two long syllables in Czech is therefore a reliable signal of their belonging to two different morphemes or words. It is easy to see that the quantity of Czech vowels is in close connection with the morphological modifications of words, i.e., with their derivational and inflexional exponents, and is different, in this respect, from that in Hungarian which is phonologically independent of the agglutinative suffixes of the language and may thus appear in any stem syllable. It functions in Czech most frequently only as a concomitant alternation in the phonological realisation of morphological oppositions, and this subservient function detracts much from its importance as a means of the lexical and grammatical differentiation of words. There are, of course, a great many minimum pairs, such as bit/bit [he is beaten/to beat], dal/dál [he gave/further away], stacl¡stácl [it is enough/he twists round], kul/kul [he hammered/stake], in which it is the only distinction between the particular forms of two different lexical items, but there are few minimal pairs that consist of two lexical items differentiated only by vowel quantity throughout their grammatical forms, such as pas/pás [passport/strip], vada/ vàda [defect/brawl], kladyjklády [assets/beams]. With regard to the prefixes najná [expresses accumulation], za/zá [prefix used in verbal aspect], prijph

190

Statistical

Linguistics

[meaning of addition etc.], and vy/vy [meaning directions upwards, outwards etc.] which take part in quantity alternation according to their forming a verb or a noun; such minimal pairs, as for example, naval/nával [pile up!/ rush], napoj/nápoj [connectî/beverage] ,privallprtval [roll over!/rush], vylov/vylov [catch the fish!/fishing] belong only to the former type, since their verbal members are restricted in Standard Czech only to the forms of the second person singular of the imperative. Unlike prefixes, Czech inflexional suffixes are often distinguished from each other by vowel quantity, cf. oci/ocí [eyes/eyes, gen. pl.], dusi/dusí [soul, dat./soul, instrum.], stolu/ stolu [tables, gen. sing./tables, gen. pl.], veselejveselé [gaily/it is gay], pfátele/pfátelé [friends acc. plur./nom. plur.] ,myslijmysli [mind, ablat. sing./he thinks]. In view of its functions, vowel quantity in Czech is similar to stress in Russian, which is also an important phonological feature firmly embedded in the morphological structure of the language. There is even a deeper relationship between both, and Roman Jakobson is fully justified in stating that the vowel quantity of the Czech and Slovak type on the one hand and stress in Russian on the other are two different manifestations of the intensity opposition, and in regarding them therefore as mutually exclusive. It may be mentioned that vowel quantity in English is of a quite different character, as it is restricted to stressed syllables and seems to be dependent on the syllable as a whole. From what has been stated about long vowels it is evident that their distribution depends largely on the morphological structure of the words, and this is also true of the succession of short and long syllables. As stated above, the sequence of two long stem syllables is impossible, and because long vowels in derivational suffixes are not of frequent occurrence, the succession of two long syllables in words is mainly constituted by long inflexional endings being added to the preceding long stem syllable, cf. pili [diligence, instrum.], pâli [the half, instrum.],péci [care, instrum.], opásá [he girds], láká [he allures], stari [old age], mesicni [monthly]. If a word happens to have a long vowel in both derivational and inflexional morphemes, in addition to a long stem vowel, the sequence of three long syllables will come into existence, e.g., (po)káráni [blaming, verbal η.], kázáni [sermon], kázániéko [a little sermon], vázáni [binding (for skis)], hádáni [quarrel], slézáni [climbing down, verbal noun], pásává [he usually tends cattle], dávává [he usually gives]. There is, however, an apparent tendency in Czech toward the disuse of long vowels in both stem and derivational morphemes of words, the sequence of three long syllables being restricted to relatively few derivational types. It may be pointed out that the rhythmical distribution of vowel quantity in words is much stronger in Slovak than in Czech, the length of vowels being suppressed after the preceding long syllables. This neutralisation of vowel quantity did not take place in Czech. As to words provided with prefixes, we can see that the long

The Distribution of Vowel Length and Its Frequency in Czech

191

vowel of the prefix is usually followed by a short stem syllable, exceptions t o this rule being the prefixes which do not take part in quantity alternation (cf. souëàst [component], and, in addition, words derived by suffixes-efe, -ka, -ko, -átko, -i, -áma, cf. prirùstek [addition], pfikládek [a small example], vyrästek [protuberance], vj>hrâzka [a little threat], nàrûdek [petty n a t i o n ] , zànârti [tarsus], zâtkàma [bottling device]. The tendency toward the rhythmical distribution of vowel length in Czech is apparent also in this type of words and is responsible for the shortening of / and ä in the stem syllable of some of these words, e.g. prirùstek, vyrustek, vyhruzka. We shall turn now to the actual frequencies of occurrence of long vowels (including the diphtong ou), as ascertained in our material of 9 457 running words. In these samples of Czech prose four non-syllabic words, prepositions s, z, k, ν, were represented by 397 occurrences, so that the number of syllabic words was 9 060. The decreasing order of their frequencies was as follows: disyllables, 3 078; monosyllables, 2 601 ; trisyllables 1 827; tetrasyllables, 1 125; pentasyllables, 347;hexasyllables, 66;heptasyllables, 11 ; octosyllable, 1 ; enneasyliable, 1 ; the total of 9 0 6 0 syllabic words in the texts under examination consisted of 20 966 syllables (including a few occurrences of syllabic r and I), so that the average number of syllables for each word was 2· 3. Short vowels were represented by 16 843 occurrences, long ones by 4 123, the ratio being 4 : 1. Their frequency distributions in syllables are indicated by table I. The table reveals that long vowels are represented most frequently in trisyllabic and disyllabic words, the ratio between them and the short ones being 1-30 : 4-17 and 1 -28 : 4-87, respectively, while the smallest relative frequency of long vowels is in monosyllables (1-1 : 11 -9). This remarkable fact is obviously due to the comparatively rare occurrence of long vowels in their inflexional endings, as well as to the high frequencies of monophonemic and biphonemic words, such as a [and], i [ t o o ] , u [at] (word type a); di [or], co [what], do [into], ho [him] , / e [it (acc.)],/'/' [her], ke [ t o ] , ku [ t o ] , mi-my [to m e - w e ] , mu [to h i m ] , ne [ n o ] , po [after], se [enclitic self], ty [ y o u ] , tu [here], ta [that ( f e m . ) ] , ti [these plur. masc.], té [you acc. sing.], vy [you nom. pl.], ze [out o f ] , ze [that] (word type ba)\ od [from] ,ai [let (us)] ,az [until], as [as] (word type ab). The greatest frequency is in the word type bab, as indicated in table II. As seen from the table, long vowels were found only in 51 different monosyllables belonging to six different word types and representing 2 2 0 occurrences. They were not evidenced in seven monosyllabic types, namely a, ab, babbb, bbabb, bbba, bbbab and bbbabb, representing 28 different words in 678 occurrences. Since all the thirteen types of monosyllables used in the texts under discussion consisted of 249 different words (or word forms) in 2 601 occurrences, long vowels were represented only in one fifth

Statistical Linguistics

192

of different monosyllables of a low frequency order (220 occurrences). If a more extensive material than 10 000 words of text were at our disposal, we could undertake a detailed quantitative analysis of monosyllables in point of different types of their morphological organisation, some of them being indeclinable and others inflected by means of zero and non-zero suffixes with or without the alternation of the type den-dne [day, day gen. sing.] but the present material is not large enough to yield higher figures from which reliable conclusions could be drawn. 4 Neither can we utilise for this purpose statistical results obtained by J. Jelinek, J. V. Becka and M. TëSitelovà, since their frequency lists based on the material of 1 623 527 running words are limited to words occurring more than twice and each inflected word is entered only in one of its grammatical "forms". s Table 1 Number of syllables

Words

Number of vowels

all

short

long

short

long

monosyllables disyllables

2 601 6 156

2 381 4 875

220 1 281

trisyllables

5 481

4 174

1 307

tetrasyllables

4 500

3 580

920

2 2 2 1 1 1 1

pentasyllables

1 735

1 425

310

396

328

68

77

62

15

220 498 783 361 284 662 91 139 233 457 12 21 33 114 130 3 4 7 4 20 30 3

hexasy Hables

hep tasy Hables

381 580 295 466 543 165 034 986 892 668 335 326 314 233 217 63 62 59 62 40 42 8

11 11 11 11 5 5

6 6

The Distribution of Vowel Length and Its Frequency in Czech

193

In conclusion, I should like to call attention to G.K. Z i p f s statement that the lengths o f words and their frequencies o f usage are in an inverse relationship. The results o f our investigation seem to confirm the validity o f his law which is, of course, o f statistical character as it does not apply to each individual word o f the text, but to all words of one type as such.

Table II Types

Number of words

Number of their occurrences

Number of long vowels

1. 2. 3. 4. 5. 6.

35 79 54 33 19 1

1 006 420 215 208 73 1

36 (in 108 (in 29 (in 38 (in 8 (in 1 (in

ba bab bbab bba babb bbbab

5 words) 20 words) 13 words) 8 words) 4 words) 1 word)

NOTES * (For reasons of convenience the Czech spelling instead of the PPA symbols have been used in this article. [V.F.] ) 1. Long vowels /, ύ and à in Czech differ phonetically from the corresponding short ones not only by the duration, but also by the intensity of their articulation, the tongue being pressed for I and ύ to a higher position than for i and u, and fora to a lower one than for a. In affected or emotional speech long vowels tend to be overlengthened, and a similar lengthening is observable with a short a, cf. ano. ano! [Yes]. Some speakers lengthen I and η for emphasis instead of the vowel: alle! anno, anno\ [but, yes]. As to é and ó, the phonetic realisation of their length shows some slight deviations from that of the integrated phonemes, /, ύ and ά. 2. Present-day ej for Old Czechs and ou for Old Czech initial ύ have a tinge of vulgarity. In pronouncing ej or ou instead of the refined I or ù in words which belong to the standard language, educated speakers endow their speech with a vulgar and jocular colouring. The use of / for the written é is felt to be nonstandard, but free from vulgarity. It is noteworthy that new words, such as itbor [dress], úloha [task], itsek [segment], keep ύ even in the pronunciation of uneducated persons, and the same is true of the plural suffix -é in pràtelé [friends], ucitelê [teachers], lidé [people]. It may be asserted that the use of ej < Old Czech >\ ou < Old Czech ú-, and i < Old Czech é in non-standard Czech is the most important test of an uninterrupted historical development of the word (or morpheme) in question. 3. Long vowels occur after prefixes ending in a vowel, e.g., in neúcta [disrespect], neúcast [non-participation], neúroda [bad crops],naútociti se [to make many attacks], but even in this position the trend toward the shortening of the hiatus vowel is obvious. Apart from u/ú, however, initial vowels do not appear in the quantity contrast. It is interesting to notice that the names of letters (e, bé, cé, etc.) have long vowels, probably due to the German influence. Adverbs have short endings. Even domû [homewards], dolu [downwards] are usually pronounced with short u, if used in the adverbial function.

194

Statistical Linguistics

4. Small 'grammatical' monosyllables consist at most of four phonemes including, in some of them, one biphonemic cluster, e.g. az [until], uz [already], on [ h e ] , je [is], se [enclitic -self], to [that], nás [us], vás [you, acc. plur.] ; tvâj [your, sing.], tvá [your, fem. nom. sing.], sväj [his, reflexive]. Consonantal clusters give phonemic expressiveness to their words and are utilised therefore for aesthetic purposes in Czech poetry. Hence the interesting fmding of V. Mázlová that their frequency is higher in poetry than in prose. Mázlová, H. (1946). "Jak se projevuje zvuková stránka cestiny ν hláskovych statistikách" [The sound structure of Czech as reflected in sound statistics]. Nase ree [Our T o n g u e ] , Prague, 30: 1 0 1 - 1 1 , 1 4 6 - 1 5 1 . It is worth mentioning that syllabic r and / had zero frequency in the monosyllables occurring in the texts under examination. 5. Jelinek, M., Becka, J.V., and Tésitelová, M. (1960). Frekvence slov, slovnich druhâ a tvarä ν ceském jazyce. [The Frequency of words, word classes, and forms in Czech ]. Prague.

On the Frequency and Distribution of Consonant Clusters in Czech

The Czech language is characterised by a great variety of consonant clusters occurring either within the bounds of a single morpheme or at the junction of two morphemes of a word. The range of the contrastive capability of consonants in Czech was discussed in some previous papers 1 and the present study will be concerned only with the frequency and distribution of clusters, as ascertained by counting their occurrences in 9 , 4 5 7 words of running texts. Four non-syllabic words, namely prepositions k [to], s [with], ζ [out o f ] , ν [in] which yielded 397 occurrences, are not taken into account in the present analysis. With 566 occurrences of four a [and], i [too] ,o [about], u [at], the corpus of all 9 060 syllabic words consisted of 20 966 syllables, including 133 occurrences of those containing syllabic r and /, and 4 words with syllabic φ and p. Let us first examine the distribution and frequency of clusters in monosyllables whose frequency of occurrence was 2 604 in the texts under study. Initial clusters found in seven types of the phonemic organisation of monosyllables (bba, bbab, bbba, bbbab, bbbbab, bbabb, and bbbabb) numbered 496 occurrences, while final clusters occurred only in 82 running words belonging to the types babb, babbb, bbabb and bbbabb. Leaving aside the type brb (3 occurrences) and apart from 566 occurrences of the type a, we may see that 564 clusters found in monosyllables were matched with 2 201 single consonants appearing in types ab, ba, bab, babb, babbb, bbab and bbbab, the ratio between the frequency of occurrence of clusters and that of single consonants being 1 : 3-9. Since the number of different clusterless monosyllables in our texts is almost the same as that of monosyllables containing one or two clusters, we may safely infer that the frequency of occurrence of the former words is four times greater than of the latter monosyllables. On the other hand, it is important to state that the clustering of consonants in monosyllables tends to

Originally published in Prague Studies in Mathematical Linguistics 3, 1972, pp. 1 1 - 1 4 .

Statistical Linguistics

196

impart to them greater phonological and semantic relevance, especially welcome to poetry. In fact, consonant clusters in monosyllabic words do occur more frequently in Czech poetry than in Czech prose. As to the frequency of 75 particular phonologically different clusters found in the basic texts, 2 the present count showed that it ranged from 1 to 72, the most frequent being st (72), r s ( 5 8 ) , p r (56), kd (46), sv (45), pr (42),y* (29),/ton (20), dn (15), sn (14), str (14), zd (13), dv (8), vl (8), etc. Sixty-two clusters filled initial positions, twelve were found at the end of monosyllables, and only one (st) appeared in both initial and final positions. A great majority of clusters were monomorphemic, the only exceptions being -chi, -jz, -mi, -nz, vstf-, vzr, vzt-, zj-, zvl-, ski- (as in sklon [inclination] ) and vl (as in vliv [influence] ), in which -z represents a particular suffix and ν, vz, ζ and s are prefixes. Disyllables showed a higher frequency of occurrence than monosyllables in our basic texts. Cf. the following Table I.

Table I Disyllables with 1 cluster

2 clusters

3 clusters

r, 1

no cluster

total

1 375 ( 5 6 1 )

303 (176)

4 (2)

58 (32)

1 388 (376)

3 128 (1 147)

As indicated by the figures in round brackets, the number of phonologically different disyllables is 1 147, while the total frequency of their occurrence is 3 128. The ratio between clusterless disyllables and those with one or two clusters is almost 1 : 2, whereas the ratio between their respective frequencies is 14 : 17. That is to say, clusterless disyllables are more than twice as frequent as disyllables containing one or two clusters. In one-cluster words, 719 clusters filled initial positions, 644 were found between vowels and only 12 occurred at the end of words. The corresponding figures for different one-cluster words are 274, 276, and 9, respectively. Out of the 289 phonologically different clusters which were found in all disyllables occurring in our texts, 29 consisted of three non-syllabic consonants, 6 of four consonants, and all the rest were biphonemic. As t o their positional distribution, four of them (chi, tig, tit, nks, rt) appeared exclusively at the end of disyllables, 4 6 occurred in b o t h initial and medial positions 3 and only two, namely st and t f , were found in all three positions, while 51 clusters occurred exclusively at the beginning of words 4 and 185 only between vowels. Clusters of the most frequent occurrence in disyllables were st (174 occurrences), p r (124), kt (104), pr (65), vj (63), sv (62), str (62),

On the Frequency and Distribution of Consonant Ousters in Czech

197

zn (48), vl (37), kl (34), si (33), vñ (32), mñ (32), St (31), HT(30), sk, zk (30), tk, dk (28), Ik (27), tv (24), zn (22), zb (20), etc. 5 The whole corpus of disyllables represented 30-70 per cent of all running words found in the texts examined. As monosyllables were 25-43 per cent and non-syllabic words 3-89 per cent, the percentage of all running words consisting of less than three syllables was 60-02 per cent. Trisyllables and tetrasyllables provided 21-65 per cent and 13-11 per cent respectively, while the remaining 5-22 per cent was added by pentasyllables (4-09 per cent) and longer words (0-89 per cent). Trisyllables numbered 1 881 occurrences of 1 275 different words. Cf. Table II. Table II

1 clusters

2 clusters

Trisyllables with 3 clusters r, 1, η

861 ( 5 9 8 )

398 (281)

4 4 (32)

4 0 (35)

no clusters

total

574 (329)

1 881 (1 275)

Initial clusters in one-cluster trisyllables were 206 (in 147 different words) and medial ones 6 653 (in 449 different words) whereas the final position was filled by them only in different words. In the whole corpus of trisyllables, 207 phonologically different clusters were found. Some of them occurred exclusively in initial position 7 (24), others (47) both initially and medially 8 , but all the rest appeared only between vowels, except for nt, which was found in medial and final positions, and st, which occurred initially, medially and finally. Clusters of highest frequency of occurrence were st (102), pr (85),/?r(76), vj (59), si (55), tk, dk (51), sk, zk (48), mh (43), ck (40), dñ (38), dn (37), zn (36), st (32), lñ (27), sp, zp (27), bj (26), tf (25), kt (24), js (21), kr (20), str, ztr (20), hi (19), sn (19), bñ (18), tñ (18), bn (17), en (18), srl (16), etc. The distribution and frequency of occurrence of clusters in tetrasyllables are indicated in the following table III. Table III Tetrasyllables with 1 cluster

2 clusters

3 clusters

4 clusters

r, 1, m

no cluster

total

445 (318)

284 ( 2 1 2 )

75 (57)

5 (5)

27(23)

293(200)

1 129 ( 8 1 5 )

198

Statistical

Linguistics

445 occurrences of clusters in one-cluster words were distributed among the initial position (53, in 47 different words), medial (391, in 270 different words), and final positions (1). Out of the total number of 185 phonologically different clusters, 145 clusters occurred exclusively between vowels, 13 only initially 9 , 26 initially and medially 10 and only one cluster (st) appeared in both marginal and medial positions. Clusters st (91 ),dn (52), p r (51), sir (49), pr (47), dn (39), ck (35), mñ (33), en (23), kl (23), It (21), rñ (21), sp (20), ν) (20), ti (19), si (18), ne (18), lñ (17), bl (17),nsk (17), en (16), js (15), etc. were among those of relatively frequent occurrence. Pentasyllables were represented in our texts by 347 running words (269 different words). They consisted of words with one cluster (173 occurrences and 132 different words), with two clusters (61,58), with three clusters (24, 21), with four clusters (3, 3), with syllabic r, 1, m (9, 8) and of clusterless words (77,47). Cf. Table IV. Table IV

1 cluster

2 clusters

Pentasyllables with 3 clusters 4 clusters

173 (132)

61 (58)

24(21)

3(3)

r, 1, m

no cluster

total

9(8)

77 (47)

347 (269)

All 379 clusters found in pentasyllables were distributed rather unequally among initial, medial, and final positions. The figures are 48, 329, and 2, respectively. In all, the corpus of 347 pentasyllables was constituted by 379 clusters, 1 726 vowels, 9 syllabic consonants, and 1315 single consonants. There were 96 phonologically different clusters, out of which 78 consisted of two consonants, 16 of three consonants, and only two of four consonants. Four clusters (dl\ dv, ps, vl) occurred only in initial positions, 80 only between vowels, 11 both in initial and medial positions, 11 and only one cluster (st) appeared in all three positions. Only nine clusters (ck (29), st (23), pr (19), st (15), dn (12), In (12), pr (12), ht (11), vj (10)) were of relatively frequent occurrence. Statistical data concerning clusters in hexasyllables are presented in Table V. Table V

1 cluster

2 clusters

3 clusters

24 (23)

21 (19)

6 (5)

Hexasyllables with 4 clusters 2(2)

r, 1

no cluster

total

1 (1)

11 (10)

65 (60)

On the Frequency and Distribution of Consonant Ousters in Czech

199

The whole corpus of hexasyllables consisted of 390 vowels (incl. 1 syllabic r), 93 clusters, 287 single consonants. There were 9 occurrences of initial clusters, whereas all the rest were found only between vowels. Out of the 39 different clusters, only st and sp occurred both initially and medially, while five (kv, p f , sv, zm, zn) were confined to initial positions and 32 to the vocalic neighbourhood. Six clusters (js (9), st (8), ck (7), vj (6),jd (4), rg (4)) occurred more than 3 times. As may be seen, almost each of 65 hexasyllables occurred only once in the texts under study. The same is true of heptasyllables (Table VI). Table VI

1 cluster

2 clusters

Heptasyllables with 3 clusters 4 clusters

4(4)

5(5)

1(1)

1(1)

m 1(1)

no cluster -

total 11 (11)

Each of the heptasyllables possessed at least one cluster. In all they consisted of 77 vowels including syllabic m in the word (intelektualism), 22 clusters, and 50 single consonants. Out of 13 different clusters, 11 were found in medial positions and only 2 (sp and str) occurred both initially and medially. Eight of the clusters occurred only once, and five of them (sp, st, rsk, kt, str) more than once. There were only two words longer than heptasyllables in the texts examined : nejkvalifikovanëjsî [most qualified] and soukromopodnikatelského [gen. adj. masc.: private enterprising]. In conclusion, the main results obtained by our statistical investigation of clusters in 9,457 words of text can be summarised as follows: (i)

The highest frequency of occurrence of clusters is attained by disyllables (2 033), trisyllables (1 796), and tetrasyllables (1 258). Monosyllables possessed 575, pentasyllables 388, and hexasyllables 93 clusters.

(ii)

As to the ratio between the frequency of occurrences of clusters and single consonants, the most favourable ones for clusters were revealed in trisyllables (possessing 4 007 single consonants) and in disyllables (containing 4 801 single consonants), i.e., the ratios 1 : 2-22 and 1 : 2 4 respectively. The corresponding ratios in hexasyllables, tetrasyllables, pentasyllables, and monosyllables were 1 : 3 , 1 : 3-38, 1 : 3 4 and 1 : 3-84, respectively. With respect to all syllabic words, the ratio is 1 : 2-5, i.e., 40 : 100.

200

Statistical

Linguistics

(iii) The ratio between the frequencies of occurrences of clusters in initial, medial, and final positions is 2 : 3-8 : -15. (iv) The ratio between the frequencies of two-consonant, threeconsonant, and four-consonant clusters is 5-5 : -6 : -038. (Syllabic consonants are not included). (v)

286 running words (3 monosyllables, 233 disyllables, and 5 trisyllables) consisted only of vowels and clusters with the exclusion of single consonants. They represented 3-02 per cent of all words constituting our texts.

(vi) Clusters occurring most frequently in our texts were st (401), pr (305), pr (262), vj (160), kt (147), mn (114), ck (112), si (110), sv (107), dh (103), str (98), st (93), dn (91), vS (88), zn (85), sk, zk (79), tk, dk (79), sp, zp (60), kl (57), In (56), st (53),/i (46), etc. No general rules concerning relationships between the frequency of consonant clusters and the length or other features of the phonemic organisation of words can be established before a morphological analysis of actual texts has been applied to clusters. This paper is the first presentation of quantitative data of the frequency of clusters in Czech prose; the results of the morphological investigation would have to be discussed in another article. NOTES 1. Trnka, B. (1967). "The Phonemic Organisation of Morphemes", Phonetica Pragensia 1, pp. 9 1 - 9 3 . (Reprinted in this volume in Section 2, p. 162.) Trnka, B. (1935). A Phonological Analysis of Present-day Standard English. Prague. Studies in English 5. (Revised New Edition. [T. Kanekiyo, T. Koizumi eds.] Tokyo, 1966.) 2. The consonantal clusters quoted in the present article are given in their usual orthographic renderings as they show the phonological values unequivocally. It should also be remembered that foreign phonemes /// and Ig/ (e.g., in words like nymfa [nymph], Kongo) are never confused with the positional variants of the native phonemes Ivi and /kl in such words as vse [all], vstfic [towards], lev [lion], kde [where], kdyz [when]. 3. bj, br, db, dl, dn, dr, hd, chi, jd, jm, hl, chi, kl, kh, kr, kt, pj, pi, pr, pr, ml, mh, si, sn, sm, sk, zk, sp, spj, zp, st, str, str, s f , tr, tr, tv, vd, vj, vn, vs. vz, vh, zb, zd, zm,'m, zn, zv. 4. cv, cl, ctv, dr, dv, hm, chv, mn, mr, ps, pt, sb, sch, sj, ms, sv, spr, ski, shl, sml, svj, sp, skvj, sk, sp, ir, iv, vb, vh, vi, vr, vs, vi, vzm, vzd, vzr, vzt (vst), vzhl, zc, zh, zj, zl, zr, z f , zhr, zdr, zmh, zpr {spr), ztr (str), zvl. 5. Syllabic r, l, m and η are counted separately from non-syllabic r, l, m and n. Such

On the Frequency and Distribution of Consonant Clusters in Czech

6. 7. 8.

9. 10.

11.

201

words as trh [market], trhl [he tore], mici [he is silent] are not included in our count of consonant clusters. 402 (277) clusters after the first vowel and 251 (172) clusters after the second one. cv, hm, chc, k f , ps, sch, sm, sml, sbl, stv, sv, sv, vpr, vs, vst, vstr, vzd, vzp, zd, zmh, zpl (spi), zvj, zvl, zr. bl, br, ci, dr, d f , hl, hr, kl, kr, ml, pi, pr, pf, si, tr, tr, vl, vr, vj, mh, vh, zb, zl, zn, zm, zh, zr, zr, sk, sn, sh, sp, st, spi, spr (zpr), stl, svj, str, si, v¡, vz, vzd, vzh, kd, kt, tv. pj, si, ski, spi, sv, sp, th, tr, vz, vzd, vzp, vzr, zkr. bl, d f , dv, jm, kl, kr, kr, kt, ml, pi, pr, pr, sk (zk), si, sp (zp), str, si, stf, svj, vs, zd, z3, zm, zn, spr (zpr), zr. kl,kr,mn, pr, pr, sp, str, tr, tr, vi, zn.

SECTION FOUR

Historical Linguistics: Diachronic Phonology and Morphology

Some Remarks on the Perfective and Imperfective Aspects in Gothic

In the well-known essay W. Streitberg 1 tried to show that the distinction between the perfective and imperfective aspects, which constitutes the basis of the verbal system of Slavonic languages, has also been preserved in Gothic. As a result of his comparison of Wulfila's Gothic translation of the Bible with the Greek original, he came to the conclusion that the Gothic verbs compounded with prefixes, such as du-, us-, at-, etc., and especially ga-, are perfective as against the corresponding simple verbs, which he supposed to be imperfective. In addition to the compound verbs, he designated a number of simple verbs, such as briggan, finpan, frapjan, giban, niman, qipan, wairpan, and gaum/an, as perfective. The other verbs were, in his opinion, imperfective, and could be made perfective by means of prefixes, e.g. taujan [to do, to be doing] ,gataujan [to have done]. According to his conception of the Gothic verbal system, each verb was either perfective or imperfective, and this sharp distinction between the two aspects made possible the adequate translation of the Greek aorist (or perfect) and imperfect respectively. It is very hard for a Slavonic philologist to endorse the theory of the eminent German scholar. No period of transition in the system of the Germanic verbal system caused by the supposed loss of verbal aspects has been detected either by Streitberg himself, or by any of his followers, in the history of the Germanic languages, and in modern languages of the Germanic stock (even in English which has developed since the 15th and 16th centuries some periphrastic forms comparable, from the semasiological point of view, to the Slavonic imperfective aspect). In the Slavonic languages one verb usually corresponds to two verbal forms derived regularly from the same root, e.g.,inModernCzechprinésti(perfective),primseti(imperfective) [to bring] ; vyhoditi (perfective), vyhazovati (imperfective) [to throw out] ,prijlti (perfective), pnchâzeti (imperfective) [to come]. If therefore no difference

Originally published in Donum natalicium Schrijnen. 500. (Slightly revised.)

Nijmegen-Utrecht, 1929, p p . 4 9 4 -

206

Historical Linguistics: Diachronic Phonology and Morphology

between the old and the new stages of the same languages may be found in this respect, we may conclude with a good deal of probability that also in Gothic there was no distinction between perfective and imperfective verbs. Streitberg, moreover, had to admit many exceptions and inconsistencies in the Gothic renderings of the Greek aorists and the imperfects. According to A. Beer's computation 2 the exceptions in the verbs prefixed withga-, which Streitberg regarded as the most important exponent of perfectivisation, make up n o less than 66 per cent of the whole number of verbs compounded with ga-. In view of this fact it is clear that the very basis of Streitberg's argumentation is uncertain, even if we accept as a fact that Greek had preserved the distinction between the two aspects down to the 4th century A.D. The fallacy of Streitberg's theory was clearly pointed out by certain eminent philologists, among others by V. Mourek 3 who, after his careful analysis of the Gothic text and its comparison with the Greek original, came to the conclusion that the Gothic prefixes, such as at-, bi-, dis-, in-, us-, had nothing to do with the perfective or imperfective aspect. He attributed, nevertheless, the occasional function of changing imperfective verbs into perfective ones to the prefix ga-, giving it a privileged position among the other Gothic prefixes. The history of the problem was exhaustively treated by A. Beer. In direct opposition to Streitberg's theory he holds the opinion that no Gothic verbal prefix had the power to convert imperfective verbs into perfective ones, their real function being to modify, or emphasise the basic meaning of the simple verb. As to the prefix ga-, Beer is of the opinion that its frequent use was due to the tendency of the translator to imitate — as it is also the case with the other prefixes — the compound verbs in the Greek original, and to the analogy of the past participles. If Beer's negative criticism of Streitberg's hypothesis is undoubtedly justified, no account is taken of the well-known fact that verbs compounded with ga-, or some other prefixes, hardly ever occur after the verbs denoting beginning, such as duginnan, nor of the fact that ga- is prefixed to the past participle. These two facts seem to point to the conclusion that verbal prefixes in Gothic, as well as in the other Germanic languages, had, in fact, something to do with the expression of verbal aspects. In this short article a probable solution of this problem is briefly suggested on the basis of comparison between living Slavonic and Germanic languages, for which Czech and English are chosen respectively. As a result of the analysis of the verbal systems of Czech and English, the word aspect is, as far as I can see, applied to three or four different semasiological series, of which only two are represented in Germanic languages. First the basic meaning of a verbal action may be connected with the additional meaning of its beginning, or its continuation, or its termination. Accordingly as far as the concept of any of the three stages, of which the action consists,

Some Remarks on the Perfective and Imperfective Aspects in Gothic

207

is involved in the verb, verbs are classified as ingressive, continuative and terminative. The ingressive aspect, denoting the beginning of an action, as well as the terminative, expressing the action in its closing moments, are included, sometimes, in the basic meaning of the simple verb itself, e.g. begin, start, increase, fatten, widen, (ingressive) fetch, come, die, finish (terminative), or very often the ingression and the termination of an action are expressed by means of prefixes or adverbs, e.g., stand up, go out, carry through, overthrow, overbalance (terminative), or by verbal constructions e.g., burst out laughing = rozesmâti se (perfective), start crying, burst into tears = rozplakati se (perfective), rozplakávati se (imperfective), live to see = do ζ it i se, doiivati se, come to see = navstiviti, navitëvovati in which the continuative verb is in combination with a verb involving the ingressive or the terminative aspect as the case may be. Secondly, as for the verbs combined with adverbs or prefixes, their terminative aspect results from the meaning of the adverbs or prefixes, which denote a direction, or a point of completion towards which the action tends to proceed. In many cases, however, the continuative verb, which denotes the phase of the action lying between its ingression and termination, is used to express either its ingression or termination, e.g., in such sentences as He already speaks = Jiímluvi (ingressive); I see him going (coming) to Mr. A. = vidim ho jií jiti(= pficházeti) k p. A. (terminative). Slavonic languages coincide, from the functional point of view, in the expression of this first series of aspects with the Germanic languages. The same is true, to a certain degree, in regard to the denoting of the repetition of the same action, whether the verbs are ingressive, continuative or terminative. The third series of aspects (perfective-imperfective) is represented only in the Slavonic languages. The perfective verb denotes an action as a completed fact, in opposition to the imperfective verb which expresses the action as a process, e.g., rekl jsem [I had my say] (perfective) : rikal jsem [I was saying] (imperfective)\privésti [to bring] (perfective): pnvàdeti [to be bringing] (imperfective). As the ingressive-continuative-terminative series of aspects is different from the perfective-imperfective series, the ingressive, terminative, and, exceptionally, continuative verbs may be either perfective or imperfective, e.g.,p/o«fi(imperfective-continuative) [to sail] : vyp/oui/(perfective-ingressive) [to set sail] : vyplouvati (imperfective-ingressive) [to be setting sail] ; doplouti (perfective-terminative) [to reach the coast] : doplouvati (imperfective-terminative) [to be reaching the coast] ; sedè ti (imperfective-continuative) [to sit] : posedeti (perfective-continuative) [*to have done sitting] .4 In Germanic languages, with the notable exception of Modern English, there is no corresponding series of aspects, and consequently, each Germanic ingressive or terminative verb corresponds to two verbs in the Slavonic languages, according to the conception of the action as a completed fact

208

Historical Linguistics: Diachronic Phonology and Morphology

or a process, e.g., the verb bringen in German is equivalent both to prinêsti (perfective) and prináseti (imperfective) in Modern Czech. In English both aspects may be differentiated by using the progressive forms which, however, also imply the actuality of a process as well as its duration. The imperfective aspect is often used in Slavonic languages to emphasise the quality of an action, without any regard to the real actual completion of the action, e.g., kdo to délai? (imperfective) [Who (has) made that?] — Proc jsi spai? (imperfective) [Why have you slept? Why did you sleep?]. The imperfective aspect, therefore, is often employed in expressing contrasted or negated actions, from the effective completion of which the speaker abstracts, the main emphasis lying on the quality of the action itself. In the positive imperative forms, on the other hand, the perfective aspect is in regular use because the main emphasis is, as a rule, put on the result of the prospective action. If, however, the speaker wishes to emphasise the progress, repetition or ingression of the action, the imperfective verb must be used even in this case, e.g., Vstañ (perfective) a choâ\ (imperfective iterative) [Rise up and walk!] ,5 vstávej (imperfective) [Get up! (= Don't be lying)] ; udélej mi to! [Do that for me!] : dëlej to takto! [Do (be doing) it thus!]. Inversely, in the negative imperative the imperfective aspect is regularly employed in the Slavonic languages, because the prohibition most often applies to the action which is in actual progress and has to be stopped, e.g., nesmèj se! (imperfective) [Do not be laughing!]. If an unwished-for action is about to take place, the prohibition of it may be expressed either by means of a perfective or imperfective verb, the perfective aspect denoting, as a rule, only a single prospective act which may be unintentionally made on the part of the hearer, e.g., neslápni (perfective) mi na nohu! [Be careful not to tread on my foot!]. The imperfective imperative neslapej means either a repeated, or a single act made intentionally. Apart from these shades of meaning, the imperfective verbs also imply the intensifying of meaning, in opposition to the perfective, e.g., humanita jiz dlouho resi tuto otázku. [Humanity has tried to solve this problem for a long time]. From what has been said about the meaning of the perfective and imperfective aspect in Slavonic languages, it is clear that the Germanic verb does not display such shades of meaning as are implied in the Slavonic verb. 6 In this respect the Slavonic verb may be compared only to the Greek, in which the aorist and the imperfect correspond partially at least to the Slavonic perfective and imperfective forms. If then Germanic languages had not preserved the distinction between the process and the fact, how was the Greek aorist to be rendered in Gothic? It is evident that Wulfila had only terminative verbs at his disposal and that he substituted them for Greek aorist and perfect forms. The discrimination between the terminative and perfective aspects also throws light, on the one hand, on the inconsistencies of the trans-

Some Remarks on the Perfective and Imperfective Aspects in Gothic

209

lator in rendering Greek tenses and, on the other hand, it accounts for the fact that the verbs such as duginnan hardly ever occur in combination with the infinitive of verbs compounded with prefixes, as well as for the fact that the past participle is very often preceded by the terminative ga- without imparting the perfective aspect to the simple verb.

NOTES 1. Streitberg, W. (1891). "Perfective und imperfective Aktionsart im Germanischen". PBB 16: 70-177. 2. Beer, A. (1915). Tri Studie o videch slovesného dèje ν gotstinè. [Three studies in aspects of verbal action in Gothic]. Prague. 3 Mourek, V.E. (1890). Syntaxis Gotskych pfedbiek. [Syntax of Gothic Prepositions]. Royal Bohemian Society of Sciences. Prague. 4. As a rule, the continuative verbs, which are always imperfective, cannot be converted into the perfective verbs. The Greek aorists, which imply the perfectivecontinuative meaning without involving its ingression or termination, are untranslatable into Czech and other Slavonic languages. 5. Cf. also Vstafi a vezma loie s vé, ¡di do domu svého. [Arise, and take up thy couch, and go into thine house] (imperfective). 6. As the fourth series of aspects in Slavonic languages, the attenuative class, expressing the quantitative distinctions of an action, may be added, e.g., otevrtti [to open] : pootevfiti [to open a little], sméti se [to laugh] : usmàti se (perfective), usmlvati se (imperfective) [to smile], cf. lächeln in German;brdti [to take] : ubrati [to take a little] ; variti [to cook] : navariti [to cook a great quantity of something].

Can Verner's Law be applied to Modern English?

In one of his studies in Danish,1 0. Jespersen explained the Early Modern English changes of the unvoiced spirants f , Θ, s, tf into the voiced spirants ν, d, z, as a consequence of the placement of stress in the sense of the well-known law formulated for Germanic by his fellow-countryman K. Verner:2 if stress did not precede unvoiced aspirants, they changed into voiced ones at the turn from ME to ModE. In contrast with Prim. Gmc for which only uncertain and doubtful proofs are documented (cf. Gmc ga = Lat. com-, gamains = communis), the unstressed syllable in ModE caused in Jespersen's view the voicing of aspirants θ >3 even at the beginning of words: the, that, those, this, these, thou, thee, thy, thine, they, them, their, there, then, than, thither. Jespersen also explains the origin of voiced spirants ν, z, ò at the end of monosyllabic words as a consequence of stress distribution as the words are often unstressed in the sentence, as for example, of, ι/(at present only////, but in the second half of the 16th cent, there is also evidence of lh>l), with, as, his, was, us (/Λ s/ at present only in Scottish pronunciation), ζ in the inflexional morpheme -iz from ME -es (ME stçnes > stçniz > stounz; rises > rîziz; but quince from Fr. coyns, hence from ME hens < hennes, and in foreign words in medial position, especially at the beginning of a stressed stem syllable and at the end of an unstressed prefix, cf. de'sign, re'semble, po'ssess, dis'own, di'scern (vs. disagree with secondary stresses on the prefix), ob'serve, an'xiety (vs. 'anxious), ex'hibit (vs. exhibition), lu'xurious (vs. luxury) (as long as analogy did not operate, as in the prefix mis), in the words ve'xation, re'solve [s, ζ ] , etc.), in the final syllable [-¿¿3], as in 'cabbage, 'cartridge, 'Greenwich, 'ostrich, and occasionally in mid position as in ajar (from ME acherre). This interpretation, which was repeated by Jespersen in his Modern English Grammar,3 was not taken over in the works on the history of English by other scholars. This cautious attitude to

Originally published in Czech under the title "Je prokázán Veinerûv zákon pro novou anglictinu?" CMF 21, 1934/35, pp. 154-161. (Translated byV.F.)

Can Vertier's Law be applied to Modern

English?

211

Jespersen's theory is not justified; nowhere was it explained w h y the older views which explain these changes on the basis of other causes (e.g. assimilation) appear to be more plausible. From the point of view o f principle it must be said that Jespersen's interpretation seems to reflect better the fact that these changes have some common denominator, i.e., that they pertain only to spirants so that the attractive experiment to interpret them more fundamentally on the basis o f Verner's law brings a state o f order into them and explains even those cases which were morphological changes due to analogy as more likely. However, the interesting experiment to prove the validity of Verner's law for the modern period o f the development of one o f the Germanic languages was not successful. The caution in accepting the unstressed position of the preceding syllable as the organic cause for the change o f ModE unvoiced spirants is dictated by a series of important factors. The assumed law does not affect the spirant /// which never changed into voiced ¡3/ (i.e. in the suffix -z's/i); and for phonetic reasons there is no probability for Jespersen's interpretation of the change tf/dz due to Verner's Law. It is further suspicious that the change o f the cluster ks> gz applies only if it was spelled , but not in the case o f , in front of e, i,y, as in the words exceed, except, excise, excite, accept, acception, succeed, successes, etc. which already occur in ME; the interpretation that this is a result o f analogy is most unlikely. Moreover, in order to apply his theory Jespersen wrongly assumes that the letter < χ > was always pronounced Iks/ in ME. In fact < χ > was pronounced in Medieval Latin by native English speakers and in Latin borrowings (with the exception of Greco-Latin loans in the 16th century) as in French, i.e., jgzj at the beginning of words and in the prefix ex- before vowels (cf. Fr and E exact, examination, Xerxes); there is no possible application of Verner's law. The same is true o f written < s > between vowels (cf. reserve, reservation, resignation) as long as the morphological composition of the word in English was not enforced. Hence it is necessary, as will be made clear later, to explain the present pronunciation [fcs], [s] for written ,, rather than the pronunciation [gz], [z]. Written < c > before e, i, y is today hardly ever pronounced [z] even after an unstressed syllable (cf. furnace, palace) and only in the combination sc (discern) there is vacillation. Finally there are also many exceptions to Verner's law even in native words or in words that are felt to be native words (as bodice; purchase) which cannot be explained through analogy. If we ignore foreign words in which the change s > ζ is doubtful, Jespersen's theory can be supported only by the enumeration of a small number o f words such as of, with, as, his, has, was and the inflexional ending (morpheme) Y e) s. A functional approach provides the basis for an attempt to find a new approach to this complex problem. If we want to evaluate the above-mentioned changes of the spirants, it is necessary first to state their function within the

212

Historical Linguistics: Diachronic Phonology and Morphology

structure of ME and to compare it with the function that they have in the later development of English. In both periods the spirants differ in their functions a great deal. In ME the voiced spirants as secondary variants of the unvoiced main variants are positionally complementary: / appeared at the beginning of words which was the position of the greatest differentiation of phonemes as well as in final and medial positions between unvoiced consonants, whereas ν occurred only in medial position before a voiced sound, a vowel or a consonant. If voiced ζ and ν occurred in words of French provenance at the beginning of words, or unvoiced / and s in the middle (before a vowel or a voiced consonant), the phonetic difference f—v, s-z was hybridly phonologised, but this phonologisation did not apply to native words. 4 The words vine, zçl, mason were considered foreign because they contained variants in the function of phonemes, which was a marker of foreignness in ME just like in Czech the difference f/v. The later disappearance of final -e caused the phonologisation of voiced spirants because v, ft, ζ occurred in word-final positions, i.e., in the same positions as the unvoiced spirants, cf. leaf: to leave, smooth (adj.) : to smooth (verb), rice : rise. Foreign words in which the only marker of foreignness was the voice of spirants lost this mark and got very near to being native words, preserving, however, in comparison with English words the special position of the phonemes (z and ν in word initial position,/ and s between voiced sounds, e.g. veal, zeal, offer, mason). Due to the simplification of ss and f f in the 15 th century in medial position of native words and of onomatopoetic words, such as kissen, laffen < lahhens they even lost this feature so that they became fully assimilated to the native words as far as the voiceless spirants in word medial position go. The phonologisation caused new alternations between nom. sg. and pi. (or gen. sg. and pl.), the preterite and the other forms of the strong verbs, nouns and verbs, in the comparison of adjectives etc., but they were simultaneously abolished or later gradually abolished. 6 In Present-day English the alternation exists in a small number of words in the plural (life : lives, bath : baths, house : houses) and in verbs versus nouns of identical or similar composition (the use : to use, grass : to graze). The phonologisation f/v is also responsible for the spelling of the adjectival ending-ive instead of the older -if which disappears in the 16th century being pushed into the background by the Latinised form because -ive permitted more easily the use of derivational suffixes and of comparison morphemes, in particular also of the adverbial -ly. The fact that the suffix -ive was indeed a continuation of the older-if may be recognised from the same stress position and from the same quantity in adjectives derived by it. The phonologisation of voiced spirants proceeded from north to south as may safely be deduced from the loss of final -e occurring already in the oldest Northumbrian monuments of ME literature and later they also disappeared in the central and southern dialects. Chaucer still uses final-e in his verse, but

Can Venter's Law be applied to Modem English?

213

soon after 1400 this-e also becomes silent in London speech. The differentiation 0/ff to be found in some MSS in the northern regions of Central England in the first half of the 14th century is probably linked with the phonologisation of voiced spirants but the a priori possibility may not be excluded that this differentiation still applied to variants (cf., e.g., the distinction of voiced and unvoiced /r/ in some Old Czech MSS by analogy to s (f)—z (3)). The phonologisation of the voiced spirants must be understood as a compensation for the loss of a syllable which was necessary to prevent the occurrence of numerous lexical homophones. Without this consideration the phonemic differentiation between voiced and unvoiced spirants would not have been necessary and the voiced spirants could have become unvoiced spirants after the loss of final -e, as seems to be proved by the words in which the unvoiced spirant appears instead of the expected voiced one (e.g. both, tenth, pith, booth [S and θ ], beneath, to curse as well as the abolition by means of alternation. In monosyllabic words losing stress in sentential context the word final spirants were pronounced voiced in front of a voiced consonant of the following word, unvoiced in front of unvoiced consonants and before pauses. In front of a vowel the voice of the final spirant also depended on the fact whether a glottal stop, which is unvoiced, occurred before the vowel or not. In the 14th century the glottal stop disappeared in English as may be concluded from the fact that especially in the second half numerous instances of the erroneous division of words beginning with an initial vowel can be

found (cf. an atter > a natter, an ekeriäme > a nickname, that other > the tother, etc.) which were caused by the fact that the glottal stop in front of a vowel which signalled the new word was no longer pronounced. The unvoiced spirant in monosyllabic unstressed words thus appeared in the second half of the 14th century only before a pause and before unvoiced spirants. When unvoicing became phonemic, the use of the paired spirants s/z, f/v, θ β conditioned by sentence phonology could not be maintained for ever in English, and it brought about vacillation which was abolished as late as in the 17th century by the generalisation of the use of either the voiced or the unvoiced spirant. In the prepositions of, with, the conjunction as, the pronoun his, the preterite was and the noun Mistress the voiced spirant prevailed because these words were frequently used and the voiced spirants therefore began to appear regularly even in front of a pause; in off (ME of)f was generalised so that of and o f f are now two different words. The vacillation in i f , us was abolished in favour of f , resp. s. The voiced /z/ in the 3rd pers. sing, has, is and in the ending-es may be explained in the same way, but morphological factors played a role in this change, too, namely the ending -(e)d. The morpheme of the regular preterite was phonologically neutralised after the syncope -e1 both as to the voice of

214

Historical Linguistics: Diachronic Phonology and Morphology

the morphological exponent and to the preceding unstressed vowel : after the unvoiced stem consonant appears \t\ (laughed, wrapped, faced), after voiced consonants /d/ (loved, grazed, sobbed) with the exception of t, d after which it is ¡id! (wanted, ended). In the same period the verbal ending -(e)s occurred after unvoiced consonants as /s/ (he sets, laughs, wraps), after voiced sounds /z/ (loves, sobs), with the exception of s, z, , tf after which appeared //z/ (faces, grazes, lashes, matches). Under the influence of the dental morpheme which stood in semantic opposition to the ending -(e)s and agreed with it considerably from the phonological point of view, the voiced spirant replaced the unvoiced spirant even in is [ > iz]. In the preterite was, apart from sentential phonological factors, a tendency for uniformity with rise, rose, is may have played a role. Equally the nominal ending [-¿s] changed into [z] perhaps because the genitival and plural -(e)s was felt to be homonymous with the verbal ending. In the speaker's awareness the genitive, plural and 3rd pers. sing, of the regular present were differentiated only by the positions of the word in the sentence, according to which e.g. loves was the genitive in the syntagma love's strategy, plural in our loves and 3rd pers. sing. pres. in the sentence Peter loves Mary. The position of the form in the sentence was accordingly the morphological exponent, the form itself was homonymous; therefore if -is>-iz changed in the verbal forms, the change was transferred to the other two functions of the same inflexional ending, which, as stated, was distinguished only by means of word order. In other words: if the morphological exponent consisted in all three functions of two (morphological) combinatory means, of which one was common to all three, the tendency to preserve this homonymity when it was to be partially disturbed operated. Moreover, all other related functions were characterised by the insufficiency of this inflexional morpheme so that the morphemic position of the 3rd pers. sing, (ends : end) in the present-tense system was the same as the position of plural or genitive (sg. : pl., gen. : non gen.) in the position of nominal relations. A confirmation of the likelihood of this analysis may also be found in the fact that the alternation f/v in the plural is still preserved in native words which are homonymous in the plural with the 3rd pers. sing, pres. (as calves, halves, shelves, wives) or have no such corresponding verbs (as knives, loaves, wolves, eaves, sheaves), whereas those words which are used secondarily as verbs (dwarf-s; roof-s; hoofs) levelled out the alternation so that their plural is homonymous with the 3rd pers. sing. pres. The exceptions are — apart from some vacillating words — the newly formed plurals semantically distinguished from the older ones (staves—staffs) and those words which had their origin in ModE (e.g. belief, -s). The same tendency may be observed with nouns ending in the spirant s and Θ, but because the development of the vowels a, ö before an unvoiced spirant differed from that before a voiced spirant from the 17th century onwards, there were few

Can Vemer's Law be applied to Modem

English?

215

homonymous plurals with the 3rd pers. sing. pres. and the analogical levelling (supported in Standard English by spelling) happened even in those instances in which we would expect the preservation of voiced spirants. Of the original plurals we still have the form houses [hauziz], homonymous with the 3rd pers. sing, pres., and the plurals with /3/ after diphthongs or long vowels (wreaths, oaths, mouths). If our interpretation is correct, it is easy to explain why in some words which are originally plurals the plural morpheme is pronounced today with ¡s¡ instead with /z/, (cf. bodice (: bodies), truce (= trewes), dice (: dies), invoice (= Fr. envois), trace (= Fr. traits), quince (= Fr. coins)·, in all instances s already lost its association with the plural morpheme in ME times so that it does not share with it the changes of the plural inflexional ending. The same holds good for the genitive morpheme-s in the adverbs hence, thence, whence, once, since. Riches (= Fr. richesse) on the other hand was taken as plural and therefore the change -es > -iz in the second syllable. The change tf > in the native word knowledge was probably caused by mistaken identification of the suffix [/if] for the suffix [à/3 ] written rarely \ similarly we could explain the occurrence of the voiced affricate Id?,I with place names as Greenwich, Woolwich, Bromwich, Sandwich; here [f/] besides [ z. As stated above the instances of the assumed operation of Verner's law in Early Mod E only apply to written < s > (rarely , ) , and to the intervocalic x, xh; the assumption is therefore justified that these cases are of a different nature from those which Jespersen found among native words. In these cases we have no organic sound changes but variant readings of Latin letters. The letter < x > which was generally considered a simple 'sound' was pronounced in the prefix ex- before a vowel [gz] according to the Old French pattern. It occurred in Old French only in Latin words so that its pronunciation must be considered analogous to that of the intervocalic simple < s > (cf. reserver, résigner)? By analogy the letter < x > was also pronounced voiced at the beginning of words as in Present-day French (apart from [fcs] in words borrowed later from Greek), whereas in ModE# at the beginning of words dropped off (e.g. xebec [zL bek] ; cf. mnemonic, psychology, knight, gnat in all these cases the initial letter is 'silent'). Written was pronounced [fa] at the end of the stem syllable (as in ex, vex, fix) and before (or after) consonants. Realising these facts we now ask: why is < x > pronounced [Äs] in English between vowels usually after a stressed syllable instead of the

216

Historical Linguistics: Diachronic Phonology and Morphology

older [gz] (cf. FT. exhibition [egz-] and E. exhibition [eks-\ vs. exhibit [igzi-\? In the linguistic awareness of the learned English humanists < χ > in the stressed prefix ex- was associated with the pronunciation of at the end of the stem syllable; after an unstressed vowel it was associated with the beginning of words if a stressed vowel followed in which position [gz] was maintained. Written < s > in Latin words was pronounced voiced in ME between vowels (and perhaps in contrast to French readings in the middle of words after voiced paired consonants), but otherwise unvoiced. After the phonologisation of the spirants it was necessary to remove the unusual alternations which were an unnecessary ballast and to generalise the voiced or unvoiced pronunciation. In suffixal syllables -sive, -sory, -sity, -sy, -son [s] stabilised in agreement with words in which these suffixes followed after unpaired consonants (cf. evasive, comprehensive, subversive). In the pronunciation of the suffixes-ousy, - 'osity (as jealousy, monstru 'osity, gener'osity) the analogy to adjectives in -ous was also at work. In cases in which at present there is a voiced spirant in front of these suffixes, it may be explained by the influence of French (ME orison, venison) or by analogy to verbs {advisory, [z] according to advise). In words in which there was no awareness of suffixes, such as pansy, quinsy [z] may already have been pronounced in ME as in native words (e.g. cleanse). In some suffixes alternation remains, e.g.-sure, (tonsure : erasure) and -sion (pension : evasion). The suffixes written with as nacioun (later written nation according to Latin) are and always were pronounced with [s]. At the end of the prefixed dis- and trans- we have the pronunciation [s] in front of unvoiced and non-paired voiced consonants (e.g. dislike, dismay, dismount, dislodge, disruptive, translate, transmute). In front of vowels the ME pronunciation of Latin was the same as in Fr. (désavouer; transaction), namely [z]. In Early ModE the pronunciation [s] was stabilised if the stress preceded, for the same reasons as in the case of the prefix ex- (cf. transitive [s] : transition [s, ζ ] , transact, -ion [s, z] ; disorder, dishonour, etc.; also discern, dissolve). The more evident the meaning of the prefixes trans- and dis- in a complex word is, the more stable is its pronunciation with an unvoiced spirant at the end of both prefixes. At the beginning of the stem syllable < s > between vowels and after voiced consonants was read [z] in ME and in Fr.; later, however, in Early ModE there was a tendency to pronounce [s] in compounds in which the morphological juncture seemed evident (e.g. re-sign, prosecute). In cases like design (= Fr. désigner), dessert, resemble (= Fr. ressembler), resent (= Fr. ressentir), po(s)sess (= Fr. posséder) in which there was originally an unvoiced spirant, the change s > ζ may be explained through the association of voice with the following stressed vowel in words in which the morphemic

Can Verner's Law be applied to Modem English?

217

juncture was not evident from the meaning of the whole word. If the alternation of the initial stem syllable occurred only with the support of stress, it is evident that this association has nothing in common with Verner's law. Even here the position of < s >, < ss > (for which often a simple < s > was written) between vowels, or between a vowel and a silent < h > or a voiced paired consonant (as observe : Fr. observer [s] ) was decisive because after a voiced non-paired consonant the initial spirant remains unchanged. In conclusion I want to state that I do not think that Verner's law can apply to Modern English. The items of evidence that seem to prove its validity must be explained as the results of phonemic and morphemic changes which took place in 15th century English, or of the vacillation of the pronunciation of foreign words in this period. In words of Latin provenance the pronunciation of the spirants was guided by the ME pronunciation of Latin as taught in schools which was in this case almost identical with the French school pronunciation. In the course of the 15th century when the influence of French culture in England was on the decline, deviations began to be apparent, especially for morphological reasons, which affect even some old French words that resemble the learned Latin words and change their pronunciation. The agreement of stress and voice which was primarily a consequence of tendencies of a morphological nature could be associated with linguistic awareness in foreign words and it may also have aided the preservation and extension of cases of alternation of the voiced and unvoiced spirant /s/ within the limits allowed by spelling.

NOTES 1. Jespersen, O. (1891). Studier over engelske kasus. Copenhagen. Cf. 178-202. [The final Chapter "Stemmeforhold i deklinationen" was reprinted in English under the title "Voiced and voiceless fricatives in English." In: Jespersen, O. (1933). Linguistica, Copenhagen, pp. 346-383. (V.F.)] 2. Verner, K. (1875). "Eine Ausnahme der ersten Lautverschiebung". KZ (Kuhn's Zeitschrift), vol. 23. 3. Jespersen, O. (' 1909). A Modern English Grammar on Historical Principles 1, Copenhagen, London, pp. 199-208. 4. French words which were fully assimilated in the pronunciation of native speakers have [z] in voiced neighbourhood. Cf. pansy, quinsy. In Present-day English [z] appears also at the end oî Mars, parse. 5. It seems that the simplifications of geminated consonants in English was initiated by the fact that they occurred in the neutralising position after the loss of final -e. 6. Cf. e.g. adj. wolfish 1570 from older wolvish 1430. 7. The loss of e was conditioned by phonetic causes, but even the analogy according to 'called hè > cald Aè'does not explain it. It seems more likely that it was influenced by the endeavour to have the same number of syllables as in the infinitive and present tense forms, cf. sing-sang, have-had, keep-kept. The same

218

Historical Linguistics: Diachronic Phonology and Morphology

. cause probably also brought about the loss of e in the morpheme -es. In ME before the loss of e the phonologisation of the morpheme of the weak preterite followed different rules: t occurred after voiced non-paired and after voiceless paired consonants, d after vowels and paired consonants, cf. kèpte, spilte, lernte, sente (: OE sende), endede. After ν, ζ there appears t in ME because they were then the variants of the non-paired phonemes I f I , /s/, cf. OE te fde > ME lifte; has fde > hadde before the change was effective. The occurrence of the preceding e was not functionally neutralised so that both inflexional endings -d(t) and -ed represented the complementary morphological exponents of the weak preterite. Would it be possible to assume the same for the ending -est This would throw some new light on s after η in forms like quince, thence, hence, whence, once, since and the parallelism of the morphological formation for the plural (resp. gen., 3rd pers. sg. pres.) and the preterite would be prolongued backward into Early ME (see later). This analysis is contradicted by the plurals truce, dice, trace, unless they were already held as singulars from the onset. After the loss of e, -d assimilated to -t after unvoiced consonants and this brought about a new neutralisation of this inflexional ending. 8. The syntactic character of the English plural (its unifying feature) may explain the vacillation of [-iz, -is) in words like porpoise, tortoise. Also business [bizniz] which Jespersen quotes from American pronunciation would belong here. The static transition of the suffix into a simple syllable is rather slow in English as it is conditioned by the meaning of the word itself. 9. It was also evidently influenced by the opposition to the native prefix ess < Lat. ex and the necessity to distinguish both prefixes by means of the voiced spirant according to the native pattern il so : ä ζδ, dis : dizä in the so-called 'liaison'.

The Phonological Development of Germanic Vowels

From my analysis of the phonological development of the Germanic consonantal system 1 it followed that it might be explained on the ground of several basic tendencies. The development of the vowels is not so transparent from the phonological point of view and a structuralist analysis constantly meets with difficulties arising not only from our rather fragmentary knowledge of the vowel system of Prim. Gmc, but also from our deficient knowledge about the relative chronology of the ensuing phonetic changes which makes every phonological interpretation only more or less probable. In this paper I can only try to interpret the most important mutations to which the Germanic vowel system had been exposed. The IE system of vowels in stressed syllables, which we also assume for the Germanic languages, consisted of an uneven number of vowel phonemes which participated in the correlation of length: i e

u o a

I u e δ a

The IE language may have also manifested the correlation of intonation the marked terms of which were the long tonic (trimoric) vowels, the unmarked terms the long toneless (bimoric) vowels. At the time of the operation of Verner's law this phonological system of vowel oppositions — except that of intonation — probably remained unchanged so that it differed from the IE system only in the realm of the consonantal phonemes. Verner's law must be interpreted, from the phonological point of view, as the neutralisation of the voice correlation of the spirants in front of unstressed vowels, a neutralisation which was naturally realised in voiced surroundings by the voiced spirants v, 3,3.

Originally published under the title "Fonologici^ v^voj germánského vokalismu". CMF 22,1935/36, pp. 1 5 5 - 1 5 9 . (Translated b y V.F.)

220

Historical Linguistics: Diachronic Phonology and Morphology

The IE dipthongs ei, oi, ai, eu, ou, au present a difficult problem in a phonological analysis. The results of comparative philology so far allow us to make the following probable conclusion: (i) IE dipthongs were combinations of two phonemes, vowel + i, respec. + u because the first member of these phonemic combinations participated in the correlation of length, i.e., it could be phonologically short or long; (ii) in front of a vowel they evidently decomposed into the original constituents; (iii) in the unstressed syllables its first member alternated with phonetic zero (cf. ei : /, eu : u, also ue : u). The biphonemic character of the IE diphthongs was preserved in Germanic as we may conclude from the fact that in Gothic both diphthongs ai and au decomposed in front of a vowel into their constituents (cf. bai : bajöps; taujan : tawida) and also from their overall development in Prim. Gmc in which the first and second member of the diphthongs changed as in all other positions. An important Prim. Gmc change was the stabilisation of stress which already before the time in which unstressed verbal prefixes came into being, had shifted to the first stem syllable of the word or to the prefix. The unstressed verbal prefixes which we assume for Prim. Gmc, 2 originated at a later time when the verbal adverbs changed into verbal prefixes. Germanic stress, which after it had been fixed on the first syllable, retained only a syntagmatic function, as is the case in Mod. Cz., acquired a morphological function. This function of stress is still preserved in Present-day German and in Present-day English because it enables the speaker to distinguish words belonging to the same morphological family (cf. 'outcast : out'cast), but not words that are morphologically unrelated of the type of Russian 'placu [I cry] : pla'cu [I pay]. Old Norse lost the morphonological function of Gmc stress as a result of the disappearance of its unstressed prefixes in the period before the first written monuments. If Danish and Swedish — in contrast to conservative Icelandic — again comprise this function of stress, we may see here an influence of MHG and NHG from which the former borrowed unstressed prefixes. The fixation of stress in Prim. Gmc resulted in a number of important phonological changes. First of all the neutralisation of voice with spirants disappeared and the combinatory variant [z] of the phoneme /s/ became an independent phoneme. The fixation of stress also extinguished the difference — if it ever existed — between the IE vowel system of stressed syllables and that of unstressed ones, because IE a, or its Germanic successor a, could also occur under stress (cf. *ps 'ter > *fa"per > *fa ïïèr > *'fadêr). Thé fixation of stress stands no doubt in relation to the loss of the intonation correlation which had brought about the reduction of final syllables — characteristic of the latest phase of Common Germanic. In the period when the stress was fixed the changes of IE i and u, which we label a-umlaut, occurred. Already before the operation of this umlaut

The Phonological Development of Germanic Vowels

221

y y y both phonemes were realised either by open sounds, if there were a, a, e λ ( > œ) in the following syllable, or by closed sounds, if they occurred before η + consonant, or if an i followed in the unstressed syllable. Open u became a phoneme when the IE endings -o(m), -a(m) had disappeared so that it occurred in the same position as u. We would have also expected the parallel phonologisation of the variant i, but for reasons hitherto unknown to us 3 — it may be that the language resisted an increase of the vowel inventory by i/e — it did not happen, and in order to preserve the timbre opposition in the language old IE o merged with IE a; thus the old frame of vowel oppositions i—u, e-o ( < IE «), a ( < IE o, a) had been maintained. The change o > a may also be due to the tendency not to increase the number of vowel phonemes. From what has been said so far, it is evident that the Gmc mutations of IE o > a and tf > o stand in some co-existing relationship. If we were to assume that the first of the two mutations is much older than the second we assume less probable phonological changes: in the frame i-u, e-a ( < IE o, a) the phoneme a would have been pushed out of the opposition with e by the new phoneme o ( < IE u), whereby the change IE o > a would not be explained. Following both interpretations, the phonologisation of the combinatory variant μ represents a compensation of the loss of the final IE syllables -e, -o(m), -a(m). In Gothic IE e merged with IE i (cf. Goth, mitán with OE metan), for IE u there is always the alternant it. These changes are most likely secondary ones and presuppose an older stage o f the vowel system that was common to all Germanic languages. If this is so, it is rather difficult to explain phonologically the Goth, changes (Gmc o > Goth, u, Gmc e > Goth. /). In Wulfila's Gothic translation of the Bible the vowels i—u, a occur side by side with the vowels e (written ai) and o (written au); although their occurrence in native words is limited to the position before the consonants r, h, in which almost never i or u occur, it is necessary to evaluate \e\, which appears in reduplication (e.g. saislêp, lailöt) as a phoneme and not as a combinatory variant o f the phoneme ///. In IE there existed the same system of long vowels as of short vowels (F-Û, e - o , a). This state o f affairs remained in Prim. Gmc, but in this period a new phone e1 arose (from IE ei and by some contraction of the reduplication with the stem syllable in the reduplicated preterites); it formed the timbre opposition to δ so that ë 1 < IE ë was pushed into the opposition to a. As this gave rise to the opposition 7 and ê2 versus labial Ü, resp. 5, labialisation took place also with Gmc a. The following system arose:

ι ë2 ë1

u 5 ρ ( < IE I )

222

Historical Linguistics: Diachronic Phonology and Morphology

It differed from the system of the short vowels. The merging of 9 and 0 then changed the system into the scheme corresponding to that of short vowels

which remained unchanged in North and West Gmc. The maximum sonorant partner in these oppositions was later realised as a, but this change did not disturb the scheme of phonological oppositions. In Goth, a simplification of the short vowels took place as in the assumed six-member system, not only ρ merged with o, but also ë 1 with e 1 . The scheme which thus originated (T-w, e - o ) was supplemented by a new phoneme a, which resulted from the Prim. Gmc a before χ (cf. brähta),4 thus Goth, too developed the same system of long vowels as of short ones (7-ü, e-δ, a). In Prim. Gmc after the fixing of stress and the change 2>a there was no difference in the oppositions of stressed and unstressed vowels. Only later as a result of the change e > i in unstressed syllables within a word was the occurrence of short vowels limited to three vowels i, u, a.s The development of the unstressed final syllables is difficult to explain; it seems necessary to assume for the time when final IE e, o(m), a(m), os had got lost intonational differences only for the vowels è 1 , δ ( < IE δ, a) and the diphthongs, not also for the vowels Τ and ü because the development of final syllables in Gmc was determined by the presence of only one correlation, i.e. either by the opposition of length (/ : 7, u : ü) or by the correlation of intonation (e/é¡ δ/