155 42 181MB
English Pages [374] Year 2000
Japanese Phonology A Functional Approach
Tsutomu Akamatsu
2000 LINCOM EUROPA
in this series 01 Karen Ebert
A Grammar of Athpare
02 Stefan G eorg
Marphatan Thakali
03 Tsutomu A kamatsu
Japanese Phonetics.
Theory and Practice 04 D.N.S. Bhat & M.S. Ningomba 05 SlEW-YUE KlLLINGLEY
Manipuri Grammar Learning to Read Pinyin Romanization and its Equivalent in Wade-Giles:
A Practical Course for Students of Chinese 07 Barbara Niederer
Les langues Hmong-Mjen (Miao-Yao).
Phonologie historique 12 Duck-Young Lee
Korean Phonology.
A Principle-based Approach 25 Heinrich W erner
Probleme der Wortbildung in den Jenissej-Sprachen
26 T sutomu A kamatsu
Japanese Phonology.
A Functional Approach 27 John Newman & A nand V. Raman
Chinese Historical Phonology
32 X iaonong Sean Z hu
Shanghai Tonetics
forthcoming: 10 Frank F. Li
The Rise and Development of Verb Compounding in Chinese:
A Historical and Grammaticalization Perspective with Cross-Linguistic Implications 21 G eoffrey Haig
Turkish syntax.
A functional-typological approach 22 Margaret Mian Yan
Introduction to Chinese Dialectology
24 Barbara Niederer
Le pa-hng de TSn Trinh
28 Ming Chao Gui
Yunnanese and Kunming Chinese:
A Study of the Language Communities, the Phonological Systems, and the Phonological Developments 29 T imur KocaoGlu
A Grammar of Karay
30 Yavar Deghani
A Grammar of Iranian Azari including comparisons with Persian
33 Nikolai Vakhtin
The Old Sirenik Language
34 Lalnunthangi Chhangte
Mizo Syntax
36 K.K. Luke & O.T. Nancarrow
A Reference Grammar of Cantonese
Published by LINCOM EUROPA 2000.
All correspondence concerning LINCOM Studies in Asian Linguistics should be addressed to:
LINCOM EUROPA Freibadstr. 3 D-81543 Muenchen [email protected] http://home.t-online.de/home/LINCOM.EUROPA
All rights reserved, including the rights of translation into any foreign language. No part of this book may be reproduced in any way without the permission of the publisher.
Printed in E C.
Die Deutsche Bibliothek - CIP Cataloguing-in-Publication-Data A catalogue record for this publication is available from Die Deutsche Bibliothek (http://www.ddb.de)
Printed on chlorine-free paper
Contents
vii
JAPANESE PHONOLOGY A Functional Approach
Contents
Acknowledgements..........................................................................................................xvii List of Tables ................................................................................................................... xix Preface ............................................................................................................................. xxi
Part I ................................................................................................................................. 1 Introduction : What is a language?....................................................................................3 Chapter 1.1 1.2 1.3 1.4
1What is functional phonology?......................................................................... 5 Functional linguistics......................................................................................... 5 Dynamic synchrony........................................................................................... 5 Phonetics and phonology ..................................................................................7 Functional phonology ....................................................................................... 7
Chapter 2.1 2.2 2.3 2.4 2.5 2.6 2.7
2 Functions of phonicsubstance....................................................................... 9 General remarks ..................................................................................................9 The indexical function....................................................................................... 9 The contrastive function ................................................................................... 9 The delimitative function.................................................................................. 10 The expressive function ...................................................................................10 The distinctive function.................................................................................... 11 Physical reality and linguistic function ........................................................... 12
Chapter 3.1 3.2 3.3
3 Distinctive units ........................................................................................ 15 Distinctive function and distinctiveu n its.........................................................15 Discreteness........................................................................................................15 Types of distinctive units .................................................................................16 3.3.1 The phoneme..................................................................................... 17
Japanese Phonology: A Functional Approach
viii
3.3.2 3.3.3
The archiphoneme ............................................................................ 19 The relevant feature............................................................................20
Chapter 4.1 4.2 4.3
4 Opposition ................................................................................................... 25 General remarks ................................................................................................25 What is opposition? What is contrast?.............................................................. 25 Opposition between distinctive units ................................................................25
Chapter 5.1 5.2 5.3 5.4 5.5 5.6 5.7
5 Phonological opposition................................................................................27 General remarks ................................................................................................27 Phonic difference............................................................................................... 27 Simple opposition and multiple opposition .................................................... 28 Exclusive opposition and non-exclusive opposition........................................ 29 Correlative opposition..................................................................................... 30 Neutralizable opposition and constant opposition ........................................ 31 Functional yield of a phonological opposition...............................................32
Chapter 6.1 6.2 6.3
6 Neutralization, neutralizable opposition, archiphoneme............................35 General rem arks................................................................................................35 Neutralization and neutralizable opposition ...................................................35 The archiphoneme............................................................................................ 37
Chapter 7.1 7.2 7.3
7 The commutation te s t...................................................................................41 The commutation test, its place in functional phonology.................................41 What is a commutative series?.........................................................................42 Does a commutative series consist of minimal multiplets or near-minimal multiplets?............................................................................... 43 7.4 Are minimal multiplets and near-minimal multiplets equally valid in performing the commutation te st?...............................................................44 7.5 What is the criterion whereby to accept or reject near-minimal multiplets?..........................................................................................................44 7.6 Why multiplets, not minimal pairs, for the commutation test?........................ 46 7.7 The necessity of operating with a number of commutative series...................47 7.8 'Maximally differentiative commutative series' and 'non-maximally differentiative commutative series'.................................................................. 47 7.9 Implications of non-maximally differentiative commutative series................. 49 7.10 No multiplet should contain a potential pause.................................................. 50 7.11 All multiplets of a given commutative series should share an identical suprasegmental context ............................................................. 51
Contents
7.12 7.13
IX
No multiplets should contain phonic manifestation associated with the expressive function ................................................................................... 53 A few examples of commutative series............................................................53
Part I I ................................................................................................................................ 59 General remarks..................................................................................................................61 Chapter 8 Identifying the vowel phonemesof Japanese............................................ 63 8.1 Preliminary remarks ........................................................................................ 63 8.2 Presenting commutative series with a view to eliciting the Japanese vowel phonemes................................................................................................ 63 8.3 Eliciting the Japanese vowel phonemes ........................................................65 8.3.1 Remarks on Commutative series 1 .......................................................65 8.3.2 Remarks on Commutative series 2 ...................................................66 8.3.3 Remarks on Commutative series 3 ...................................................66 8.3.4 Remarks on Commutative series 4, 5,6 and 7 ....................................67 8.3.5 Voiceless vowel segments [j] and [rq], etc.......................................... 68 8.4 Defining the Japanese vowel phonemes in terms of relevant features........ 68 8.5 The structure of the Japanese vowel phoneme system.....................................69 8.6 Realizations of the Japanese vowel phonemes................................................. 70 8.7 Are there instances of neutralization of oppositions between the vowel phonemes of Japanese? .................................................. 71 Chapter 9 Identifying the consonant phonemes of Japanese ................................... 75 9.1 Presenting some commutative series with a view to eliciting the consonant phonemes of Japanese................................................................75 9.2 Remarks on Commutative series 1 to 5 ........................................................... 82 9.3 Remarks on Commutative series 6 to 1 0 ......................................................... 83 9.4 Defining the Japanese consonant phonemes in terms of relevant features............................................................................................................... 84 9.5 The structure of the Japanese consonant phoneme system ........................... 86 9.6 Realizations of the Japanese consonant phonemes....................................... 88 9.7 Justificatory explanations of the relevant features in terms of which the Japanese consonant phonemes have been defined.................................... 90 9.8 Explanation about some of the Japanese consonant phonemes established........................................................................................................ 93 9.8.1 Identifying !\1 and /d/ .......................................................................... 93
Japanese Phonology: A Functional Approach
X
9.8.2 9.8.3 9.8.4 9.8.5 9.8.6 9.8.7 9.8.8 9.8.9 9.8.10
Identifying /ts/, Izl, I cq/ and /?/ as single phonemes............................94 Identifying /ts/ as "apical affricate".................................................... 94 Identifying /cq/ as "palatal affricate".................................................... 95 Identifying /?/ as "voiced hush"......................................................... 95 Identifying/q/ as "voiceless hush".................................................... 95 Identifying Izl as "voiced hiss".............................................................95 Identifying 1)1 as "palatal spirant".........................................................96 Identifying /uj/ as "dorsal spirant" ..................................................... 97 Identifying /a/ as "dorsal spirant".........................................................99
Chapter 10 Establishing Igl, Ig'l, /g/, /g’/, Ikl and Ik'I (in the minority speech) and IXJ, /X7, IYI and /Y7 (in the majority speech)..................................105 10.1 Preliminary remarks .........................................................................................105 10.2 Variation in the occurrence of [g], [g’j, [g] and [g’j in Japanese ...................105 10.3 What phonetic factor(s) might cause free variation between [g] and [g] or between [g?] and [g’j ? ............................................ 106 10.4 The minority speech and the majority speech............................................... 107 10.5 Establishing /g/, /g7, Igl and Ig'l in the minority speech.................................108 10.5.1 Partial reproduction of Commutative series 1 to 1 0 ........................ 109 10.5.2 Defining /g/, /g7, IgJ and Ig'l in terms of relevant features.............. I ll 10.5.3 The non-occurrence of [g] and [g’ ] in word-initial context in the minority speech and the majority speech............................ 112 10.5.4 Complexity of the occurrence of [g] or [g], or [g’ ] or [g’j, in word-medial context in the minority speech and its phonological implications................................................... 113 10.5.5 Pairs of Japanese words conducive to establishing /g/ and IgJ (and /g7 and Ig'l) ............................................................ 114 10.5.6 Consequence of the establishment of /g/, /gV, Igl and Ig'l on phonological notation............................................................... 115 10.6 Identifying IXJ, /X7, IYI and IY'1 in the majority speech ............................. 116 10.6.1 The pattern of occurrence of [g], [g’j, [g] and [g’j in the majority speech ..................................................................... 116 10.6.2 Setting up commutative series conducive to establishing IXJ, IX'I, IYI and IY' / ..........................................................................117 10.6.3 Defining IXJ, IX'I, IYI and IY'1 in terms of relevant features.......... 121 10.6.4 Correlative structure of IXJ, IX'I, IYI and /Y 7 .................................123 10.6.5 Realizations of 1X1, IX'I, IYI an d /Y 7............................................... 123 10.7 Consequence of the establishment of IXJ, IX'I, IYI and /YV on phonological notation..................................................................................124
Contents
xi
10.8
IXJ, IX'I, IYI and IY'1 in the majority speech and Igl, /g7, /g/ and /g7
10.9
in the minority speech.................................................................................... 125 The phonological implication of IXJ, IX'I, IYI and IY'1 in the majority speech ............................................................................................................ 125
Chapter 11 Neutralization of oppositions between the consonant phonemes of Japanese ..............................................................................129 11.1 Preliminary remarks ........................................................................................ 129 11.2 Two major groups of neutralizable consonant oppositions in Japanese..... 129 11.3 The neutralization of Ipl -Ip 'I, /b /- /b ’/,/k /-/k ’/,/g /-/g ’/,/m/-/mV, /n /-/n ’/,/g /-/g ’/, a n d /r/-/r’/ ........................................................................ 130 11.4 The neutralization of /XJ - /ts/?, Id! - /z/?, /cp/ - /t/?, /?/ - /z/?, Id - /s/?, /?/-/$/?, and I q l- l h ll......................................................................................134 11.5 Is I I I - 1)1 neutralized in some contexts?...........................................................137 11.6 Is An/ - /uj/ neutralized in some contexts?........................................................ 138 11.7 A summary of the non-occurrence of some consonant phonemes before some vowel phonemes in Japanese................................................... 140 11.8 The neutralization of /m /- /m’/ -I n i-I n '/ - /g/- /gV, the neutralization of Ipl - p7 - /b/ - lb’ I - N - Id/ - M - Ik'l -1%) - Ig'l- / $ / - / ? / - 1)1- M - Id - /?/ - I s l - l z l - /ts/ - /cp/ - Id -Ic'l - lv)l and the neutralization of /p /-/p 7 - /b /- /b 7 - /t/- /d /- /k /- /k 7 - /g /- /g 7 - /$ /- /? /- /j/- /h /- /e / - Id - Is/ - I z l - /ts/ - /cp/ - Id -Ic'l - Iml - /m7 - Ini - In'I - /g/ - /g7 - /ig/ ... 140 11.8.1 Preliminary remarks..........................................................................140 11.8.2 Commutative series 1 to 18 (Group 1)............................................. 142 11.8.3 Commutativeseries 19 to 36 (Group 2 ) .............................................. 147 11.8.4 Commutative series 37 to 43(Group 3) ...........................................154 11.8.5 Commutative series 44 to 52(Group 4) ...........................................161 11.8.6 Commutative series 53 to 57(Group 5) ...........................................167 11.9 A summary of instances of neutralization of oppositions between consonant phonemes of Japanese (in the minority speech) ....... 170 11.10 The neutralization of IXJ - IX'I and of IYI - IY '1 ......................................... 172 11.11 Other instances of neutralization of oppositions between consonant phonemes in Japanese (in the majority speech) ........ 175 Chapter 12 The commutation test reviewed................................................................ 179 12.1 What has the commutation test achieved in Chapters 8, 9, 10 and 11?....... 179 12.2 Inventories of all the distinctive units of the second articulation of Japanese and their phonological contents ..................................................180
xii
Japanese Phonology: A Functional Approach
12.2.1 The inventory of the distinctive units of the second articulation of Japanese (in the minority speech) ............................................. 181 12.2.2 The inventory of the distinctive units of the second articulation of Japanese (in the majority speech)............................................. 182 12.2.3 The main differences between the phonological systems of the minority speech and the majority speech.............................. 184 12.2.4 The minority speech and the majority speech ............................... 184 Chapter 13 Combinabilities of the phonemes and archiphonemes of Japanese ......187 13.1 Generalizable principles governing combinabilities of the phonemes and the archiphonemes of Japanese.............................................................187 13.2 Four main types of combinabilities of the phonemes and the archiphonemes in Japanese .............................................................188 13.2.1 The succession of two or more vowel phonemes, i.e. Type I ........188 13.2.2 The succession of a consonant phoneme/an archiphoneme + a vowel phoneme, i.e. Type I I ....................................................... 190 13.2.3 The succession of an archiphoneme + a consonant phoneme/ an archiphoneme, i.e. Type i n ...........................................................194 13.2.4 The succession of two archiphonemes + a consonant phoneme, i.e. Type I V .........................................................................................197 Chapter 14 Aspects of dynamic synchrony in Japanese phonematics...................... 199 14.1 Prelimary remarks..............................................................................................199 14.2 The question of /a/ in the conservative dialect and /iq/ in the innovating dialect................................................................................. 200 14.3 The question of /g/, /g’/, /g/, /g7, DtJ and /k7 (in the minority speech) and /X/, /X’/, /Y/ and /Y7 (in the majority speech).........................................201 14.3.1 Introductory remarks..........................................................................201 14.3.2 Phonological catalysis........................................................................ 201 14.3.3 Does Igl - /g/ (or /g7 - /g’/) minimally distinguish pairs of Japanese words? ........................................................................ 202 14.3.4 Why do /g/, /g/, /g7 and /g7 continue to survive in the minority speech?.................................................................. 204 14.3.5 Why has the majority speech beengaining ground?.................... 206 Chapter 15.1 15.2 15.3
15 The delimitative function........................................................................ Preliminary remarks ...................................................................................... The delimitative function fulfilled by [g] or [g’j ............................................ The delimitative function fulfilled by a stressed falling melody..................
207 207 207 208
Contents
xiii
Chapter 16 The indexical function ............................................................................ 211 Chapter 17 The expressive function..............................................................................215 Chapter 18.1 18.2 18.3 18.4 18.5 18.6 18.7
18 Moraic unit ................................................................................................219 Preliminary remarks ........................................................................................ 219 The moraic unit and the sub-moraic unit.........................................................219 'Inherent moraic unit' and 'moraic unit'......................................................... 220 An illustration of moraic units as phonological un its................................. 220 Types of moraic unit........................................................................................ 221 Sub-moraic units in a moraic u n it................................................................ 222 The boundary of a moraic unit: where does a moraic unit start and end?.................................................................................................. 222 18.8 The moraic unit as revealed in the phonological behaviour of native speakers of Japanese.................................................................... 223 18.9 The relevance of the moraic unit and the sub-moraic unit to the commutation te s t............................................................................... 225 18.10 Demonstration of the commutability between moraic units and between sub-moraic units....................................................................... 225 18.11 The characterization of IN/ both as "nasal" and as a moraic u nit................ 227 18.12 The characterization of IQI both as "non-nasal" and as a moraic u n it....... 229 18.13 The characterization of ICI both as "consonantal" and as a moraicunit .... 229 18.14 Potential commutability between different types of moraic unit, depending on the contexts ......................................................................... 229 18.15 The archiphonemes INI, IQI and ICI and traditional phonological analyses of Japanese.................... 230 18.16 Some implications of the concept of the moraic unit; is [i] or [in] elided or not in Japanese? and do we have e.g. [ii] or [i:] in Japanese?.... 230 18.17 The relevance of the concept of the syllable in Japanese phonology....... 233 18.18 Dichotomy between a vowel phoneme and a consonant phoneme and between a moraic unit and a sub-moraic u n it...................................... 234 18.19 An inventory of the moraic units and the sub-moraic units in the minority speech and the majority speech in Japanese ..................... 235 18.20 The round-up of the distinctive units of the second articulation of Japanese ................................................................................................... 236
Chapter 19 Accent ................................................................................................... 239 19.1 The contrastive function briefly revisited................................................... 239 19.2 Accent defined............................................................................................... 239
xiv
Japanese Phonology: A Functional Approach
19.3 19.4 19.5 19.6 19.7 19.8
Accent is a discrete unit which is either presentor absent........................... 240 Accent is not a distinctive u n it.................................................................... 240 The location of accent may fulfil the distinctivefunction..............................241 The inherent function of accent is contrastive................................................. 241 Accent in Japanese...........................................................................................241 How differently is accent in Japanese manifested from that in some other languages? .......................................................................................... 242 19.9 Is Japanese a tone language because accent is manifested through pitch? ............................................................................................. 243 19.10 Is the syllable relevant to accentual patterns in Japanese?......................... 243 19.11 Accentual patterns in Japanese................................................................... 244 19.11.1 Traditional classification of accentual patterns in Japanese........ 244 19.11.2 Accentual notation ......................................................................... 248 19.12 Can all of Variants 1 to 8 be properly considered as accentual patterns in Japanese? ................................................................................................... 251 19.12.1 Theoretical problems arising from Variants 1 and 2 ....................... 251 19.12.2 Theoretical problems arising from Variants 3 to 5 ......................... 252 19.12.3 Proposed solutions to problems arising from Variants 1 to 5 ........ 252 19.12.4 How are the different accentual patterns (reclassifiable as Variants 1 to 8) traditionally ascribed to individual Japanese words? ..... 255 19.13 Neutralization of oppositions between syntagmatic relations (i.e. contrasts)............................................................................................... 256 19.13.1 The nature of neutralization of oppositions between variants..... 257 19.13.2 Instances of neutralization of oppositions between variants in Japanese....................................................................................... 257 19.13.3 Cases of non-occurrence of all eight variants in certain contexts in Japanese....................................................... 263 19.14 Accentual patterns of composite words in Japanese ................................. 268 19.15 Alternative accentual patterns of Japanese w ords..................................... 270 19.16 Accentual patterns and dynamic synchrony................................................. 274 19.17 The pitch status of a moraic unit consisting in a sequence of a voiceless consonant segment and a voiceless vowel segment............ 281 19.18 The pitch status of a moraic unit consisting in a voiceless consonant segment ...................................................................................... 285 19.19 Phonological notation of Japanese passages .............................................. 288
Conclusion
293
Contents
xv
Appendix 1 (Japanese consonant phonemes (minority speech))............................... 297 Appendix 2 (Japanese consonant phonemes (majority speech))............................... 299 Notes................................................................................................................................. 301 References ...................................................................................................................... 337 Index
349
Acknowledgements
X V II
Acknowledgements
My deepest indebtedness and gratitude in writing this book go to Andre Martinet (1908-1999) who had been the guiding star in my pursuit of general linguistics ever since my arrival in Paris in 1964 up to his recent death. Martinet's scholarship, perspicacity, integrity and realism in his practice of linguistics never failed to inspire me all those years in my own linguistic research. It is through Martinet himself that I have come to know and be convinced with functional linguistics. I had been aware for some time until his death that one of his wishes was that I should one day produce a work on a phonology of Japanese from a functional point of view. The present book is my modest exercise in responding to his wish. My only regret in publishing the book at this date is that he is no longer here to see it. Martinet's unexpected death occurred only a few months before the completion of the manuscript, which was delayed largely by the fact that I was faced with a good number of theoretical problems which I had to resolve one by one as the writing progressed. Martinet would no doubt have found in my book what he would have agreed or disagreed with on various theoretical points of my phonological analysis of Japanese. My hope is that I have not misunderstood or misrepresented too much his teaching on functional linguistics, in particular functional phonology. I wish to mention two Japanese scholars in particular who have been extremely kind, helpful and prompt in responding in espitolary form to my questions about some aspects of Japanese. They are Shin Kawakami and Isamu Abe. I consider myself to be very lucky to have these Japanese friends who have given me expert assistance unstintingly and readily from across the oceans. My wife, Maryvonne, has once again, as for my previous books, involved herself in going through the whole of the manuscript with a fine comb, thus sparing me accidental misspellings, illogicalities in my arguments, and improving various aspects of presentation, and in discussing with me a number of specific points of theory as I found myself confronted by them. In her capacity as a fellow functionalist, she has been enormously helpful to me in all these tasks. She has also sorted out many technological difficulties while I was preparing the camera-ready copy of the manuscript on our home computer. I am infinitely grateful to her for all her multifaceted and unstinting assistance. Last but not least, I thank the publisher, Lincom Europa, who bore with me for a considerable period of time till I was ready to submit the camera-ready manuscript of this book well beyond the original deadline .
T. A.
List of Tables
xix
List of Tables 1.
The Japanese consonant phonemes and their relevant features ....................84-85
2.
The Japanese consonant phonemes with their phonological contents and their contexts of validity..............................................................................85-86
3.
The Japanese consonant phonemes with their realizations and their contexts of validity..................................................................................... 88
4.
The Japanese vowel phonemes (minority speech)................................................ 181
5.
The Japanese consonant phonemes (minority speech).............................. 181-182
6.
The Japanese archiphonemes (minority speech)..................................................182
7.
The Japanese vowel phonemes (majority speech).................................................. 182
8.
The Japanese consonant phonemes (majority speech) ...................................... 183
9.
The Japanese archiphonemes (majority speech).................................................. 183
10. Succession of two identical vowel phonemes ......................................................188 11. Succession of two different vowel phonemes ......................................................189 12. Succession of three or more identical or different vowelphonemes .................... 189 13. Succession of a consonant phoneme/an archiphoneme + a vowel phoneme.......................................................................................... 190-193 14. Occurrence of combinations of a consonant phoneme/an archiphoneme + a vowel phoneme.......................................................................................... 193-194 15. Succession of an archiphoneme + a consonant phoneme/an archiphoneme.......................................................194-196 16. Occurrence of combinations of an archiphoneme + a consonant phoneme/an archiphoneme..................................................... 196-197 17. Succession of two archiphonemes + a consonant phoneme ............................... 197
Preface
xxi
Preface The objective of this book is to present my own phonological analysis of current standard spoken Japanese (of which I am a native speaker) from a functional point of view, in accordance with the principles of functional phonology set out in a previous work of mine, Akamatsu (1992). I am a Japanese male in my late sixties, born and brought up in Tokyo, where I resided until early adulthood, and have since lived in Europe up to the present. My use of the indefinite article in the title of this book, Japanese Phonology: A Functional Approach, implies that other phonological analyses of Japanese attempted by other functionalists are possible and are therefore not to be ruled out. Consequently, other functionalists may or may not agree with me to different measures regarding details of my own analysis. ’Current standard spoken Japanese', which constitutes the object of my phonological analysis presented in this book, is 'non-regional' in that it is practised in widespread, if not necessarily contiguous, areas in Japan, including Japan's metropolis. It is the variety of Japanese pronunciation that is heard nowadays from the mouths of not only the majority, if not quite all, of newsreaders on radio and television in Japan but also great numbers of Japanese residents in many urban localities. I dispense here with any descriptive account of 'current standard spoken Japanese', for which the readers may wish to consult section 0.5 (Standard Japanese pronunciation) in Akamatsu (1997a), or any other works on current Japanese pronunciation written in Japanese or otherwise. A functional phonology of Japanese consists in analyzing the phonic substance of Japanese with a view to identifying and hierarchically classifying the functions that they fulfil in the language. The phrase 'from a functional point of view' is to be taken in all seriousness. There already exist in linguistics literature some analyses of the phonic substance of Japanese, but from points of view other than functional. So far as I know, none appears to have been published that undertakes a phonological analysis of Japanese from a functional point of view. The approach in the analysis of the phonic substance of Japanese presented in this book contrasts with the approach adopted in Akamatsu (1997a) in which the same phonic substance is treated deliberately without any reference to its various functions. The two books by the present author are therefore, so to speak, companion volumes in which the same object matter is analyzed from two different points of view, i.e. phonetic and phonological. The framework in which a phonological analysis of Japanese is conducted in this book is the type of functional linguistics specifically associated with what I call the Functionalist School (formerly often called the Paris School) and essentially based on the teaching of Andre Martinet (1908-1999). This particular type of functional linguistics is not to be facilely equated with the type of functional linguistics professed and practised by the Prague School from 1926 to the present day. Notwithstanding a broad measure of agreement, there persist a good number of disagreements on fundamental issues between the two types of functional linguistics (cf. Akamatsu, forthcoming). The term functional linguistics, occasionally employed in this book, will therefore be in exclusive reference to the Functionalist School. 1 suspect that the majority of the readers of this book are little familiar with the theoretical framework attributable to the Functionalist School, though not a few will already be well acquainted with that associated with the Prague School. For this reason I intentionally devote the whole of Part I (i.e. Chapters 1 to 7) to explaining at some
Introduction
3
Introduction What is a language? Functional phonology, which forms part of functional linguistics, is unlikely to be a type of phonology with which most current linguists, and for that matter the majority of the readers of this book as well, are already well acquainted. For this reason, I believe it would be appropriate for me to provide right at the outset a short - and in no way exhaustive - account of some aspects of functional phonology, so that the readers may the better follow the contents of this book. Part I is designed to be such an account and divides into the Introduction and then Chapters 1 to 7. Part I explains the concepts and methodology on which my phonological analysis of Japanese from a functional point of view is based. It is subsequently in Part II which consists of Chapters 8 to 19 that I shall actually present step by step my phonological analysis of Japanese. It is essential in this book to set the scene by introducing to the readers a functional concept of a language, for all analyses of various aspects of a language from a functional point of view depend fundamentally on the concept of a language as entertained by functionalists. Functional phonology, among other branches of functional linguistics, is compatible with a functional concept of a language. The concept of a language I subscribe to and operate with is couched in the definition of a language which I quote below from Martinet (1989, p. 12).1 Une langue est un instrument de communication selon lequel l'experience humaine s'analyse, differemment dans chaque langue, dans chaque communaute, en unites douees d'un contenu semantique et d'une expression vocale, les monemes; cette expression vocale s'articule a son tour en unites distinctives et successives, les phonemes, en nombre determine dans chaque langue, dont la nature et les rapports mutuels different eux aussi d'une langue a une autre. (’A language is an instrument of communication by means of which human experience is analyzed, differently in every language, in every speech community, into units each of which is endowed with a semantic content and a vocal expression, i.e. monemes. This vocal expression is in turn articulated into distinctive and successive units, i.e. phonemes, which are of a given number in each language and whose nature and mutual relations too differ from language to language [my translation, taken from Akamatsu 1992b, p. 5].’)2 I shall not go into any detailed explanation of Martinet's definition of a language quoted above, for which the interested readers should consult Martinet (1989, pp. 1220) itself with much profit, or, for a concise vicarious explanation, Akamatsu (1992b, pp. 4-7). I need to stress the three defining characteristics of a language that one sees in Martinet's definition above: (1) a language is an instrument of communication; (2) a language is doubly articulated (hence, the first articulation and the second articulation, which together can be referred to as double articulation); and (3) a language is of vocal character. The concept of a language as defined above will be borne in mind throughout my analysis of current standard spoken Japanese from a functional point of view as presented in this book. Purely for convenience sake, I shall hereafter in this book, if and when occasion demands, employ the term signifier for 'expression vocale (vocal expression)' and the term signified for 'contenu semantique (semantic content)'. The minimum units of the first articulation are monemes,3 each of which is endowed with its signified and signifier. They are significant units. On the other hand, the minimum units of the second articulation are distinctive units,4 Of the three above-mentioned defining characteristics of a language, it is not
4
Japanese Phonology: A Functional Approach
difficult to understand and agree on the first and the third characteristics: a language is an instrument of communication (this does not prevent a language being employed for other purposes, but the communicative employment of it is of the utmost and obligatory importance) and a language is first and foremost of vocal character (every language without exception has or had a spoken form, and some if not others have evolved a written form corresponding to it) and consequently of linear character. It goes without saying that were it not for the vocal character of a language there would be no phonetics or phonology of a language. Few would disagree about these two defining characteristics. What some readers may find novel is the defining characteristic known as double articulation, which I believe needs to detain us a while below.5 A language is first articulated into sequences of monemes (each moneme is endowed with a signified and a signifier); this is the stage of analysis of a language known as the first articulation. The signifier (but not the signified) of a moneme is then articulated into a sequence of distinctive units, phonemes;6 this stage of analysis of a language is called the second articulation. The two above-mentioned stages of analysis of a language are referred to as double articulation. The phonological analysis of a language, with which this book is concerned, is centred on facts of the second articulation, but this does not mean that it is divorced from the facts of the first articulation, as the necessity of the commutation test in particular unmistakably shows.
Chapter 1: What is functional phonology?
5
Chapter 1 What is functional phonology? 1.1
Functional linguistics
The primary aim of functional linguistics consists in investigating individual languages in such a way that the various functions fulfilled by elements of the languages are identified and classified. A functional point of view is always at the core and is never lost sight of in functional linguistics. There exist a number of works which one can turn to for a good introduction to functional linguistics. I mention only a few of them, all of which make excellent reading: Martinet (1960), Martinet (1962), Martinet (1965), Martinet (1989) and Martinet (1991). Note that, though Martinet's linguistics is related in certain respects to interwar Prague School linguistics, there exist major divergences, both of theoretical and practical order, between the two streams of 'functional linguistics’, so that it is more appropriate to designate these two streams of 'functional linguistics’ as the Functionalist School and the Prague School, respectively (cf. Akamatsu, forthcoming). It is not irrelevant to point out in this connection that, unlike frequent statements to the contrary ascribable to many a linguist, Martinet has not been 'influenced' by the teaching of the Prague School linguistics, and that it is truer to say that at a certain point of his own linguistic thinking and research, Martinet happened to find that his line of linguistics was along similar lines as those held by the Prague School (cf. Akamatsu, forthcoming).7 1.2
Dynamic synchrony
One special feature of functional linguistics is what is known among functionalists as dynamic synchrony. The distinction between diachrony and synchrony is well known among linguists in general. Saussure (1916, p. 119 = 1928, p. 117) actually introduced the terms 'diachrony' (Fr. 'diachronie') and 'synchrony' (Fr. 'synchronie') in elaborating the distinction between the two methods of linguistic analysis. However, Saussure’s synchrony is what may be paraphrased as 'static synchrony', from which dynamic synchrony should be distinguished. In referring to dynamic synchrony, Martinet (1991, 2.2) writes as follows: II n'est pas impossible, il est m€me recommande, dans une etude synchronique de relever les tendances evolutives de la langue en opposant les usages de differentes generations en presence. On dira dans ce cas qu’il s’agit d'une synchronie dynamique. On parlera de diachronie lorsqu'on confrontera les synchronies dynamiques successives de chaque langue ('It is not impossible, it is even recommended, in a synchronic study, to note the evolutionary tendencies of the language by contrasting with each other usages of different co-existing generations. We shall say in this case that we have to do with a dynamic synchrony. We shall speak of diachrony when successive dynamic synchronies of the same language are compared with each other [my translation, taken from Akamatsu 1992b, p. 148].') It is clear that not only is the concept of dynamic synchrony different from Saussure's concept of synchrony (i.e. static synchrony) but Martinet's concept of diachrony which is based on dynamic synchrony is also different from Saussure's concept of diachrony.
6
Japanese Phonology: A Functional Approach
The nature of dynamic synchrony can even better be seen from another passage quoted from Martinet (1975, p. 9): II peut done Stre indique d'opposer, a l'etude diachronique visant deliberement a comparer differents etats successifs du meme objet d'etude, une synchronique dynamique ou l'attention se concentre, certes, sur un seul et meme etat, mais sans qu'on renonce jamais a relever des variations et a y evaluer le caractere progressif ou recessif de chaque trait ('It may therefore be recommended to oppose to a diachronic study which deliberately aims at comparing different successive stages of the same object of study, a dynamic synchrony in which we concentrate our attention on one and the same stage, but in which we never give up noting variations and evaluating the progressive or recessive nature of each feature [my translation, taken from Akamatsu 1992b, pp. 148-149].') Dynamic synchrony can be practised in various domains of linguistic research, including phonology. The first major work on phonology from the point of view of dynamic synchrony is Martinet (1945), which has subsequently spawned a huge number of works in phonology from the same point of view among functionalists. The term 'dynamic synchrony' itself does not occur in Martinet (1945), but it is of historical interest to note that, so far as I know, the term occurs for the first time at a later date in Martinet (1954, p. 17 = 1965, p. 149). Perhaps one of the best examples of a consideration in terms of dynamic synchrony is the phonological opposition in French between /£/ (as in brin 'blade (of grass)' and /ce/ (as in brim 'brown'). Fewer and fewer French speakers of today (except for those living in certain provincial districts) make this phonological distinction, as the two above-mentioned phonemes merge and both brin and brun, for example, have /g/. The phonological opposition between /§/ and/c6/ represents a recessive phenomenon and the original diversity in the use of this phonological opposition among French speakers may well be replaced in future by uniformity in that lei will have become universal, along with the disappearance of /ce/. As the present book is on functional phonology as applied to Japanese, or alternatively, if one prefers, on Japanese analyzed from the point of view of functional phonology, I myself shall be concerned with certain aspects of spoken Japanese which are most appropriately considered and described in terms of dynamic synchrony (see Chapter 14 and 19.16). I am doing here no more than bring to the readers' attention the concept of dynamic synchrony, which applies to phonology as well as other fields of linguistic research. The readers are invited to take a look at a synopsis of dynamic synchrony (as at the beginning of 1990s) which I presented in Akamatsu (1992b, pp. 147-156) and specifically to consult a good number of works written by other functionalists in the spirit of dynamic synchrony that I referred to in my synopsis. I make no attempt here to identify those works which are substantial in number. However, I make an additional special mention here of Houdebine (1985, pp. 7-36) which was accidentally missed out in Akamatsu (1992b). It goes without saying that writings on dynamic synchrony that have been published at subsequent dates have yet to be mentioned in my synopsis. I therefore make here specific mention of La Linguistique (1996, p. 97, pp. 157-158) and Martin (1998) which postdate 1991. There are no doubt more works on dynamic synchrony I should have cited. I end this section by emphasizing that dynamic synchrony has direct implications for diachronic phonology, so that linguistic descriptions conducted in the spirit of dynamic synchrony at a given time may profitably be compared with those conducted at subsequent stages.
Chapter 1: What is functional phonology?
1.3
7
Phonetics and phonology
Functionalists believe that the distinction between phonetics and phonology is essential and valid and that it is based on the criterion of function. (This distinction has, as is well known, been contested or rejected by those linguists who do not operate with the criterion of function, hence non-functionalists.) Here and below, phonology should of course be understood in the sense of functional phonetics or, if one prefers, functional phonology, despite the tautology (tautology as explained below). The object matter of both phonetics and phonology is the same, the phonic substance of languages. However, while phonetics is concerned with phonic substance per se without regard to its functions in individual languages, phonology is concerned with phonic substance with regard to functions which the phonic substance of a given language fulfils. It is therefore the two points of view vis-a-vis the phonic substance of languages that set phonetics and phonology apart from each other, not different domains of discipline nor different objects of study. It is not wide of the mark to say that phonetics can be pursued on its own, but phonology needs phonetics as the basis, plus the criterion of function. Akamatsu (1997a) was written as a work on Japanese phonetics by deliberately setting aside considerations of the functions of the phonic substance of standard Japanese, while the present book has been written, on the basis of such information as is contained in Akamatsu (1997a), by investigating the functions that are fulfilled by the phonic substance of standard Japanese. The readers may wish to consult a conveniently short explanation, buttressed by some examples from a few different languages, of the distinction between phonetics and phonology which is found in Akamatsu 1992b, section 2.1). Any serious linguist will not stop at phonetics but will also concern himself with phonology.8 1.4
Functional phonology
Functional phonology is a part of functional linguistics. In this book as well as other writings of mine, by functional phonology is meant the type of phonology which forms part of functional linguistics attributable to Andre Martinet (1908—1999)9 and being practised by his associates of many nationalities. I wish to draw the readers' attention to the fact that recent times have been seeing a number of works published which are said to be on 'functional linguistics' or 'functional grammar' but have in fact nothing to do with the type of functional linguistics that is associated with Martinet and his associates' linguistic research. So far as functionalists are concerned, the term 'phonology' is equivalent to 'functional phonetics’ (cf. Martinet 1949) and consequently the term 'functional phonology' is tautological. If I nevertheless insist on using the term 'functional phonology', this is simply for the purpose of making it certain to the readers in general that phonology as practised by functionalists is never to be confused with other types of phonology currently being practised by non-functionalists (cf. Akamatsu 1992b, section 2.2).10 As this book is addressed to functionalists and non-functionalists alike, I feel that it is justified to go on employing the term 'functional phonology' in order to preclude any possible terminological misapprehension on the part of the readers. The term 'functional phonology' (and its equivalent in other languages) has actually previously been used in e.g. Martin (1983), Akamatsu (1992b) and Martin (1997), which are books, and also here and there in some academic papers written by functionalists. What then is functional phonology concerned with? I take the liberty of reiterating what I have already said in an earlier book, as the passage to be quoted seems to
8
Japanese Phonology: A Functional Approach
answer the question in a succinct manner (Akamatsu 1992b, p. 11). Functional phonology, which forms part of functional linguistics, is concerned with identifying, describing and classifying various functions of the phonic substance of a given language. Various functions of phonic substance present themselves in a hierarchy from the point of view of their respective contributions to the functioning of a language. In other words, they make varying contributions to a language. This does not mean that only some of the functions are reckoned with while others are ignored. All functions find their proper place relative to each other in functional phonology. It seems important to further reiterate the short footnote {op. cit., loc. cit.) appended to the above quoted passage, as this would obviate, or correct, widespread misunderstanding held by a number of non-functionalists. It should be emphasized [...] that functionalists do not retain, as worthy of their phonological study, just the distinctive function to the exclusion of other functions, as frequently misunderstood by non-functionalists. Apart from Akamatsu (1992b), a number of works in which functional phonology is expounded of course exist, and the readers are invited to consult inter alia Martinet (1945), Martinet (1947), Martinet (1949), Martinet (1955, First Part, i.e. pp. 11-195), Martinet (1956, pp, 11-47), Martinet (1960, Chap. 3), Martinet (1965, Chaps II-VIII), Martinet (1989, Chap. 4), Fran?ois (1968, pp. 190-227), Martin (1983), Martin (1985), Martin (1997), Builles (1998, Chap. 6), etc. The framework of double articulation is relevant to the identification of the two sub-areas of functional phonology, i.e. phonematics and prosody. Phonic substance which conforms to the second articulation will be investigated in the rubric of phonematics: La phonematique traite de l'analyse de l'enonce en phonemes, du classement de ces phonemes et de l'examen de leurs combinaisons pour former les signifiants de la langue (’Phonematics is concerned with the analysis of utterances into phonemes, the classification of these phonemes and the examination of their combinabilities in forming the signifiers of a language [my translation].') (Martinet 1960, 3.5) while phonic substance that eludes the second articulation is examined in the rubric of prosody, or, as Martinet himself puts it: Les faits linguistiques qui ne se conforment pas a l'articulation en phonemes sont souvent dits « supra-segmentaux » et forment un chapitre intitule prosodie... ('The linguistic facts which elude an articulation into phonemes are often called 'suprasegmentals' and constitute a chapter entitled prosody... [my translation].') (Martinet 1960, 1.15) The units of the second articulation, to be treated in phonematics, are distinctive units. The toneme, one particular type of suprasegmental (or prosodic) element, though treated in prosody, is also a distinctive unit, while the rest such as accent, intonation and speech melody are not distinctive units. In this book as well as in other writings of mine, I adhere to the dichotomy between phonematics and prosody. My phonological analysis of Japanese (Chapter 8 onward) will encompass both phonematics (Chapters 8 to 18) and prosody (Chapter 19).
Chapter 2: Functions of phonic substance
9
Chapter 2 Functions of phonic substance 2.1
General remarks
I have emphasized enough in the foregoing parts of this book the crucial importance in phonology of investigating the functions that are fulfilled by the phonic substance of a given language. It is high time that we should see what these functions are. I have provided an exposition on 'functions of phonic elements' in Akamatsu (1992b, Chapter 3, i.e. pp. 17-24), which the readers may wish to have a look at. I make another attempt below to explain some selected functions. 2.2
The indexical function
This is a function whereby a phonic element provides information to the listener about the speaker, i.e. the geographical provenance, the socio-economic status, the sex, the age, etc. of the speaker; these are the speaker's indexes, so to speak. The phonic elements which fulfil the indexical function lie outside the speaker's choice. For example, [r] (apico-alveolar trill) or [r] (apico-alveolar tap) in English (in normal conversational discourse) suggests that the speaker may be Scots or possibly Welsh. A lengthened [a:] or [a-] in an accented syllable as in porridge may suggest that the speaker is American or Canadian. In Chukchi, a language spoken in eastern Siberia, men pronounce [tf] while women and children pronounce [ts] for the same words. The difference between [tf] and [ts] therefore serves in Chukchi to indicate the difference in the speaker's sex, and consequently fulfils the indexical function. A point about terminology calls for a brief remark. What I call the indexical function corresponds to what the Prague School calls 'Kundgabefunktion' or 'Ausdrucksfunktion' (cf. Biihler 1920, Biihler 1934; also Trubetzkoy 1939, p. 19). It corresponds to what the Prague School also calls 'function of expression' (cf. Vachek 1966, p. 34) or 'expressive function1(cf. Vachek 1966, p. 36). The last two terms should not be confused with the expressive function which I, following Martinet (1960, 3.1), employ with a different sense and which I shall explain below (2.5). Note that somatic features are not candidates for the indexical function even though it is true that, for example, a relatively high speaking voice pitch is characteristic of women and young children and a relatively low speaking voice pitch is characteristic of men. This difference is of a somatic nature. 2.3
The contrastive function
This is a function whereby a phonic element helps the listener to analyze utterances into series of significant units, i.e. linguistic units consisting in the union of a vocal expression and a semantic content (cf. Martinet 1960, 3.1). (The term 'significant' as in the expression 'significant unit' above is employed in a technical sense, i.e. 'having a meaning'.) The contrastive function corresponds to what the Prague School calls the 'culminative function’ (cf. Trubetzkoy 1939, pp. 29-30). For example, accent in Spanish, Turkish, Italian, Estonian, etc. serves this purpose. In such languages an incidence of accent which characterizes a polysyllabic word indicates the occurrence of successive significant units in utterances. In other words, a series of incidences of accent in such languages will help the listener to analyze utterances into a series of significant units. This is what is meant when Martinet (1965, p. 146), among other writings of his, says
10
Japanese Phonology: A Functional Approach
that Taccent sert essentiellement a individualiser les unites semantiques dans la chaine parlee (’accent serves essentially to individualize the semantic units in the speech chain' [my translation]'). That this kind of analysis of utterances may or may not be carried out for the whole of each utterance is of lesser importance. What is important is the very fact that the contrastive function is fulfilled by accent in such languages. The choice of the term 'contrastive' (cf. 'contrast') should be particularly noted. The concept and term of 'contrast' refers to the successive relation between linguistic units of the same type, this type of relation being known as syntagmatic relation. For example, the syntagmatic relation between the three sequential significant units in the English utterance Black sheep bleep is contrastive. The syntagmatic relevance of the contrastive function exemplified above by accent in Spanish, Turkish, Italian, Estonian, etc. should be obvious. The concept and term of 'contrast' should be strictly distinguished from the concept and term of 'opposition' which we shall see further below (4.2). In the meantime, it should be stressed that 'contrast' and 'opposition', or 'contrastive' and 'oppositional', are non-interchangeable concepts in functional linguistics. 2.4
The delimitative function
This is also called the demarcative function (cf. Martinet 1960, 3.1) and corresponds to what the Prague School calls the 'delimitative function' (cf. Trubetzkoy 1939, p. 29). This is a function which derives from the contrastive function. The delimitative function is a function whereby a phonic element indicates the boundary between successive significant units. For example, accent in Estonian (but not in e.g. Spanish) fulfils the delimitative function, as accent in Estonian falls in principle on the first syllable of a polysyllabic word, with the result that the boundary between words occurs just before the accented syllable. Accent in Turkish which falls in principle on the last syllable of a polysyllabic word fulfils the same function in that the boundary between words occurs just after the accented syllable. Accent in Polish which falls in principle on the penultimate (i.e. the last but one) syllable shows the boundary to be just after that unaccented syllable which follows the accented syllable. Accent in Estonian, Turkish, Polish, etc. fulfils the delimitative function because accent in these languages has a fixed place, be it the first, last or penultimate syllable within a polysyllabic word, and this is why accent in such languages is often called a 'fixed accent'. Such is of course not the case with accent in Spanish, English, Russian, etc. in which the location of accent in a polysyllabic word, once the segmental structure of the word is identified, shows itself to vary from word to word. This is why accent in these languages is often called a 'free accent'. I said that the delimitative function derives from the contrastive function. To turn the matter backwards, a phonic element that fulfils the delimitative function fulfils also, in fact primordially and necessarily, the contrastive function. 2.5
The expressive function
This is a function whereby a phonic element expresses the speaker's state of mind or mood or attitude, real or feigned, to the listener without resorting to the framework of double articulation (cf. Martinet 1960, 3.1). The concept of double articulation has already been explained in the Introduction in connection with Martinet's definition of a language. The expressive function is fulfilled without the addition of a word or words to achieve the above-mentioned end. For example, if an English speaker says an enormous tree with an overlong [n::: ] and/or [o:::] - in such phonetic notation the symbols : , : : : , etc. signify varying degrees of length of the phonetic segment indicated before the symbols - in pronouncing the word enormous, he is expressing his surprise,
Chapter 2: Functions of phonic substance
11
admiration, incredulity, emphasis or whatever other states of mind, instead of, but in order to express the same state of mind, saying e.g. an extra o rd in a rily (or unbelievably, unusually, very very, etc.) big tree, by employing some additional word(s). The expressive function is also fulfilled in such pronunciation as [gabei] grey (by throwing in [a]) instead of [gjei] in e.g. grey sky frequently resorted to by weather forecasters on British television. The expressive function is a function that the Prague School does not mention. This is because the framework of double articulation, which is an essential factor in defining the expressive function, is extraneous to Prague School linguistics; in fact it is absent in all types of linguistics other than the linguistics of the Functionalist School. For this reason the expressive function should not be confused with what the Prague School calls the 'appellative function' (cf. Btihler 1920, Biihler 1934) - 'Auslosungsfunktion' or 'Appellfunktion' as Biihler in his respective works calls it - or the 'function of appeal' (cf. Vachek 1966, p. 34) or the 'conative function' (cf. Isacenko 1964). Nor should it be confused with 'function of expression' (cf. Vachek 1966, p. 34) or 'expressive function' (cf. Vachek 1966, p. 36) which actually corresponds to 'Kundgabefunktion' or 'Ausdrucksfunktion' mentioned by the Prague School. 2.6
The distinctive function
This may also be called the 'oppositive function', the 'differentiative function', etc. The distinctive function is a function whereby distinction between different significant units is made possible (cf. Martinet 1960, 3.1). For example, the difference between the initial phonological segments in e.g. pit, bit, mitt, knit, sit, lit, etc. in English serves to distinguish these significant units from each other. For this reason, these phonological segments, known as phonemes, are considered to have a distinctive function. In Mandarin Chinese, which is a tone language, the difference between the melodic shapes which co-occur with the segmental phonological elements, also serve to differentiate between words which are segmentally identical. In Mandarin Chinese it is thanks to the different melodic shapes that, for example, the word pronounced [pa] and meaning 'eight', the word pronounced [pa] and meaning '(to) uproot', the word pronounced [pa] and meaning '(to) hold' and the word pronounced [pa] and meaning 'harrow' are distinguished from each other. The four mutually different Mandarin Chinese words are written in four different Chinese characters. The four different melodic shapes, once identified as linguistic elements with a distinctive function are known as tonemes (or tones) in Mandarin Chinese. The four tonemes are traditionally known as tone 1 (indicated above by * ), tone 2 ( ' ), tone 3 ( ”) and tone 4 ( “ ), which one can alternatively call toneme 1, toneme 2, toneme 3 and toneme 4 if one prefers to employ the term 'toneme' (with the ending -eme) rather than 'tone'. The four tonemes can be, preferably in my opinion, indicated as, say, III, 121, /3/ and /4/, in phonological notation, in which a pair of oblique lines are conventionally employed to indicate the phonological units endowed with a distinctive function. Of the various functions mentioned and explained above, it is the distinctive function which is the most important, indeed indispensable, function,11 for the functioning of a language. Communication via a language is only possible if the messages to be transmitted by the speaker to the listener can be properly distinguished from each other, and this is best achieved by employing linguistic elements that fulfil the distinctive function. Identification, description and classification of these linguistic elements is of paramount importance in functional phonology. This is why a substantial part of this book is devoted to the analysis of linguistic units which are endowed with the distinctive function. This is only natural because the distinctive function stands, among the various functions, hierarchically at the top in terms of its contribution to
12
Japanese Phonology: A Functional Approach
language communication. Note that this does not mean that the other functions are disregarded, for they too find their respective proper place in the hierarchy of functions, in accordance with the relative degrees of importance of their contributions to language communication. In the foregoing account of the functions of phonic substance, I have not mentioned all the functions that the functionalist may consider as being fulfilled by the phonic substance of languages. Three of the functions mentioned and explained above, namely the distinctive function, the contrastive function and the expressive function (Martinet 1960, 3.1), stand out as particularly important. 2.7
Physical reality and linguistic function
The linguistic identity of any phonic substance is determined strictly and criterially in terms of the functional role it plays in each of different individual languages, and not merely on the basis of the physical nature of the phonic substance per se. Consequently, there is no necessary correspondence between physical reality and linguistic function. This principle of the utmost importance in functional phonology can be explained with the help of a few examples below. Our first example is length (or quantity, as alternatively called) of a phonic element. Vowel segments are varied in length; they may be long or short, or even half-long or half-short, etc. In Czech, for example, there are pairs of long vowel segments and short vowel segments, i.e. [i:] and [i], [e:] and [e], [a:] and [a], [o:] and [o] and [u:] and [u] (which Czech orthography indicates by means of \ and i, e and e, a and a, 6 and o, and u and u, respectively). The distinction between such a set of long vowel segments and the corresponding set of short vowel segments can keep pairs of Czech words apart, so that the length of Czech vowel segments functions distinctively. This is why one can conceive, phonologically, of the distinction between /i:/ and /i/, /e:/ and /e/, /a:/ and /a/, /o:/ and /o/ and /u:/ and /u/ in Czech. Segments like /i:/, /i/, /e:/, /e/, /a:/, /a/, /o:/, /o/, /u:/ and /u/ in Czech are phonemes which are conventionally symbolized with the use of oblique lines (i.e. //). On the other hand, segments like [i:], [i], [e:], [e], [a:], [a], [o:], [o], [u:] and [u] are phonetic segments, which are conventionally symbolized with the use of square brackets (i.e. [ ]). However, in Italian, for example, vowel segments are long when occurring in accented syllables but short when occurring in unaccented syllables, so that we can say that in this language, accent is manifested with (greater) length of the vowel segments. In other words, length (of vowel segments) functions contrastively in Italian, as accent in this language (or any other languages that have accent) fulfils the contrastive function. Note that Italian has a system of vowel phonemes in which length has no distinctive function, unlike in Czech. In English, (greater) length of vowel segments (and/or of consonant segments) may be so used that if the speaker wishes to express a particular state of mind (surprise, admiration, disgust, etc.), he can do this by resorting to (greater) length of vowel segments (and/or consonant segments), as we have seen in 2.5 in the pronunciation of an enormous tree with [n::: ] and/or [o:::] in the word enormous. (Greater) length is here functioning expressively. From the above three examples about the use of length drawn from three different languages, it will be easily seen that the same phonic substance, the same physical reality, i.e. the length of vowel segments in the above examples, corresponds to (three) different linguistic functions. Martinet (1960, 3.3) illustrates, with more examples from different languages (Czech, Italian, English), that the same phonic element functions differently; the readers are invited to consult Martinet's illustrations. I myself, in Akamatsu (1992b, pp. 13-15 e/ passim), refer to the fact of mutual independence between physical reality and linguistic function by taking the example of the glottal plosive ([?]) which functions distinctively in Arabic, delimitatively in German, and expressively in French, but which has no linguistic function in Japanese.
Chapter 2: Functions of phonic substance
13
The principle that there is no necessary correspondence between physical reality and linguistic function can be shown to apply in one and the same language as well. On the one hand, one and the same phonic element may correspond to different distinctive units in one and the same language, so that for example in English, fm] (voiced bilabial nasal) in cam [kaem] and the same phonetic segment in camp [kaemp] are not realizations of one and the same distinctive unit, for [m] of [kaem] is a realization of the phoneme which we notate by /m/ while [m] of [kae mp] is a realization of the archiphoneme which we notate by /m-n-g/. (The concepts of the phoneme, the archiphoneme and the realization will be explained in 3.3.1 and 3.3.2 onward.) Conversely, different phonic elements may correspond to one and the same distinctive unit in one and the same language. For example, in French, [p] (voiceless bilabial non nasal plosive) and [m ] (voiceless bilabial nasal) are different realizations of one and the same phoneme which we notate by /p/ in French phonology. Martinet (1969b, p. 129) refers to the different pronunciations [kapm a] and [kargma ] for ca m p em en t 'encampment', where both pronunciations correspond phonologically to /kapma/, as [p] of [kapma] and [rt]] of [karrjma] are both realizations of /p/. Note that [rg ] here is not a realization of /m/. In the English examples above, note specifically that [m] of [kaemp] is not a realization of /m/. The reason why [m] of [karri] and [m] of [kaemp], the same physical reality, correspond to different linguistic functions is not that they fulfil different functions, for both [m]s fulfil the same function, i.e. the distinctive function. What should be seen is that the two [m]'s do not have the same range of 'distinctive possibility' (i.e. what Trubetzkoy 1939, pp. 69, calls 'phonologische Giiltigkeit', i.e. phonological validity). In the French example above, the reason why [rq ] of [karqma] and [p] of [kapm a], different physical realities, correspond to the same linguistic function is that they enjoy the identical distinctive capacity in its opposability to other distinctive units in that same context in French. Were it not for the criterion of function, [m] of [kaem] cam and [m] of [kaemp] camp in English would be regarded, on the strength of their identical or near-identical physical reality, as assignable to one and the same distinctive unit, /m/, in English, and one might well hesitate about how to assign [rq ] of [karama] (to /m/ or /p/? and for what reason?) while having no difficulty assigning [p] of [kapma] to /p/.12 The principle of there being no necessary correspondence between physical reality and linguistic function, as illustrated above, has the implications, among others, that in identifying the distinctive units, the functionalist does not base his phonological analysis criterially on phonetic similarity and distribution (complementary or otherwise). It is the commutation test (explained in Chapter 7) that the functionalist requires for that purpose as well as other purposes, and the commutation test is fundamentally based on the concept of opposition.
Chapter 3: Distinctive units
15
Chapter 3 Distinctive units 3.1
Distinctive function and distinctive units
One of the functions, in fact the most important function, when we identify the various functions fulfilled by the phonic substance of a given language, is, as I have said, the distinctive function. I have already mentioned 'distinctive', 'distinctive unit' and 'distinctive function' in the foregoing on several occasions. It is by virtue of the distinctive function fulfilled by phonemes that monemes are kept apart from each other. (Strictly speaking, one should also mention in this connection archiphonemes, tonemes and architonemes as well.) Thus, it is thanks to the distinctive function fulfilled by the initial phonological segments in the signifiers of, for instance, the English monemes pet, jet, set, net, bet, Met., etc. that these monemes are kept apart from each other. In the above examples, the initial phonological segments that function distinctively are what we identify as the phonemes /p/, /cg/,/s/,/n/,/b/,/m/, etc. during the course of our phonological analysis of English. The signifiers of these monemes are /pet/, /d$et/, /set/, /net/, /bet/, /met/, etc. It will have been noticed that these signifiers are identical except for the initial phonemes. Likewise, an archiphoneme fulfils the distinctive function. For example, in the English monemes cast, cask and castle, it is thanks to the distinctive function fulfilled by the final phonological segments that these monemes are kept apart from each other. The phonological segments in question will be identified as the archiphoneme /t-d/, the archiphoneme /k-g/ and the phoneme III, in these three respective monemes. The signifiers of these monemes are, respectively, /kas t-d/, /kas k-g/ and /kasl/. (For my use of a space or spaces round the symbol for an archiphoneme in phonological notation, see my explanation in 6.3.) Apart from /t-d/, /k-g/ and 111, the remainder of the signifiers of these monemes is identical, i.e. /kas/. A relevant feature too fulfils the distinctive function, though somewhat indirectly, so to speak. For example, the English monemes pit /pit/, tit /tit/ and kit /kit/ are kept apart from each other thanks to the distinctive function fulfilled by /p/, N and /k/ in the initial position in the three monemes. It should be added that if the difference between /p/, /t/ and /k/ is directly responsible for the differentiation between these monemes, it is the distinctive function fulfilled by the relevant features "bilabial" (of Ip/), "apical" (of /t/) and "dorsal" (of fkJ) that makes possible the difference between /p/, IxJ and /k/. The contribution made by these relevant features for the differentiation between the monemes in question may be said to be, as it were, at a remove, but nevertheless they do fulfil the distinctive function. Just like the phoneme, the archiphoneme and the relevant feature, the toneme and the architoneme (both are identifiable in terms of a relevant feature(s)) also fulfil the distinctive function. My own application of the term and concept of 'distinctive unit' to any phonological unit which is endowed with the distinctive function is more extensive than most other functionalists' in that they tend to employ the term 'distinctive unit' as pertaining only to the distinctive units of the second articulation (i.e. phonemes and archiphonemes) as opposed to significant units which are the units of the first articulation. This point will be taken up again and explained in 3.3. 3.2
Discreteness Any type of distinctive unit is a discrete unit, not a non-discrete unit. (Discrete =
16
Japanese Phonology: A Functional Approach
non-continuous; non-discrete = continuous.) The difference between discreteness and non-discreteness may be illustrated by comparing a piano on the one hand and a siren on the other. The notes produced on a piano are discrete in the sense that the notes that a performer produces are graded and are chosen from the eighty-eight specific notes; it is impossible to choose any such note as lies between any two adjacent notes. Unlike a piano, a siren produces non-discrete series of different pitches. On the other hand, a violin or any other string instrument is in principle supposed to be so played as to produce discrete notes but can be so played to deliberately produce a series of nondiscrete notes in order to bring about 'portamento'. Discreteness and non-discreteness can also be illustrated by different gears (1st, 2nd, 3rd, 4th, reverse) which work discretely and the accelerator and the foot brake which work non-discretely on a motor-car. All types of distinctive unit mentioned above - phoneme, archiphoneme, toneme, architoneme, relevant feature - are discrete units in that there is, for instance, no phoneme which lies between two other phonemes. The speaker chooses either this phoneme or that phoneme, and no in-between phoneme. There is~~no continuity between phonemes, and the difference between phonemes is clear-cut. If any English listener happens to find himself hearing what seems to be somewhat like /p/ and somewhat like /b/ in the slurred speech of an inebriated English speaker, the listener must assume and reckon with either /p/ or lb/, not anything like 'a phoneme that lies between /p/ or /b/'. The inebriated person himself will have chosen either /p/ or /b/ anyway. That a distinctive unit is a discrete unit means by the same token that one distinctive unit (say, a phoneme) is chosen instead of another (i.e. another phoneme) at a giveh^joint in the speech chain. Thus, an English speaker chooses /p/ instead of some other phoneme when he says pet /pet/ instead of Is/ (set /set/), InJ (net /net/), Id/ (debt /det/), etc. The listener similarly identifies one phoneme rather than another in comprehending what the speaker says. 3.3
Types of distinctive unit
Two types of distinctive unit can be mentioned first: phonemes and archiphonemes. (I am aware that in his definition of a language, Martinet mentions phonemes only as each and every language has phonemes, but I wish to add archiphonemes in the present discussion.) Both the phoneme and the archiphoneme are basically comparable in that they are both the minimum units of the second articulation, i.e. distinctive units, into which the signifiers of monemes are analyzed. Signifiers of monemes are expressible in terms of, necessarily, the distinctive units, and additionally, where this is the case, other elements like accent and/or tonemes which elude the framework of double articulation. The signifier of a given moneme is indicated by way of a phonological notation which traditionally presents the constituent distinctive units concerned arranged in a given sequential order, together with the indication of accent or a toneme(s) at the appropriate places in the phonological notation. A phonological notation customarily starts with and ends with a pair of diagonal lines. The use of the diagonal lines is meant to indicate that the linguistic elements shown between the diagonal lines are phonological entities (as distinct from phonetic entities). Here are a few examples of phonological notation in which the signifiers of monemes are indicated. I draw some examples from English: /kaet/ for the moneme cat and /'ae m-n-rj ka/ for the moneme anchor, /s p-b it/ for the moneme spit, and /bae m-n-rj k-g/ for the moneme bank. It is to be understood that /kaet/ indicates that the phonemes /k/, /ae/ and Ixl occur in this particular order, /'ae m-n-rj ka / indicates that the phoneme /ae/, the archiphoneme /m-n-rj/, and the phonemes /k/ and /a / occur in this particular order, /s p-b it/ indicates that the phoneme /s/, the archiphoneme /p-b/ and the phonemes hi and Ixl occur in this particular order, and /bae m-n-g k-g/ indicates that the phonemes /b/ and /ae/,
Chapter 3: Distinctive units
17
and the archiphonemes /m-n-g/ and /k-g/ occur in this particular order. What is implied by the use of more than a single symbol, strung with a hyphen(s) for archiphonemes (cf. /m-n-g/, /p-b/ and /k-g/ above) will be explained in 6.3. For the moment it is enough to see that both the phoneme and the archiphoneme enjoy a comparable status in their capacity as distinctive units of the second articulation in the signifiers of monemes. I have mentioned that the phoneme and the archiphoneme are two types of distinctive unit. In fact, two other types of distinctive unit must be mentioned which, however, unlike the phoneme and the archiphoneme, elude the framework of double articulation, and consequently are not distinctive units of the second articulation as the phoneme and the archiphoneme are. They are the toneme and the architoneme. A tone language (e.g. Chinese, Norwegian, Mende, Apache) has, in addition to phonemes, necessarily two or more tonemes, and, depending on individual tone languages, also one or more architonemes. The toneme and the architoneme are among those phonological units often characterized as 'suprasegmental' elements, in opposition to 'segmental' elements like the phoneme and the archiphoneme. Japanese is not a tone language,13 and this is why I shall have no occasion to speak of the toneme or the architoneme in connection with Japanese. Yet another type of distinctive unit is the relevant feature, which has already been mentioned here and there and will continue to be mentioned in what follows and is fully explained in 3.3.3. It is in terms of relevant features that the phoneme, the archiphoneme, the toneme and the architoneme are defined. Some remarks are in order here about the use of the term 'distinctive unit' among functionalists in general. The term 'distinctive unit' (’unite distinctive') seems to be employed by Martinet and most of his associates in opposition to the term 'significant unit' ('unite significative'). Thus, distinctive units are units of the second articulation and are distinctive in their function, and successive, never concomitant, in their occurrence. (In contrast, significant units are units of the first articulation and are significant in their function, and are successive or concomitant, as the case may be, in their occurrence.) This means that phonemes (and archiphonemes as well) are, properly speaking, distinctive units. If so, distinctive units would not include relevant features which are distinctive in their function like phonemes but can be concomitant in their occurrence, unlike phonemes, and this would be at variance with my characterization of the relevant feature as well as the phoneme (and the archiphoneme) as distinctive units. Furthermore, if phonemes (and archiphonemes) are distinctive units of the second articulation and consequently conform to the framework of double articulation, tones and architonemes on the other hand which elude the framework of double articulation would not be distinctive units, though they are endowed with the distinctive function. Besides, those relevant features in terms of which the toneme and the architoneme are definable would also elude the framework of double articulation. Yet it is true that the phoneme, the archiphoneme, the toneme, the architoneme, and the relevant feature all fulfil the distinctive function. Since I wish to call a distinctive unit any linguistic unit which fulfils the distinctive function, I will continue with my own use of the term 'distinctive unit' in the rest of this book (as indeed in the way I have also done in Akamatsu 1992b). 3.3.1
The phoneme
One of the types of distinctive unit, the phoneme, is a minimum unit of the second articulation, as already mentioned. Every language possesses a system of phonemes of its own. The signifiers of monemes, which every language possesses, may consTst of sequences of phonemes (and archiphonemes), e.g. cat /kat/, speak /s p-b i:k/, bank /b ae m-n-g k-g/ in English. The signifiers of some monemes consist of single phonemes (e.g. /a/, the signifier of the French word a 'at, to') or single archiphonemes (e.g. /f-f-v-v’/, the
18
Japanese Phonology: A Functional Approach
signifier of the Russian word B 'in, at'), but an archiphoneme of course presupposes phonemes. (The diacritic mark ’ as in f and v’ above signifies palatalization.) The phoneme is a linguistic unit which is not only distinctive in its function but also successive in its occurrence - in other words, the phonemes are obligatorily linear in nature - in the signifiers of monemes, as we have seen earlier in the examples of the signifiers of some English monemes like cat, anchor, spit and bank. The order in which the phonemes occur in the signifiers of individual monemes is fixed, so that the signifier of cat, for example, is always /kaet/, not /task/ or /aekt/ (which is the signifier of a different moneme, i.e. tack or act), nor /aetk/, /ktse/ or /tkae/ (which are non-occurrent sequences of the three phonemes Ikl, /del and Ixl in English and therefore do not constitute the signifier of any English moneme). As already mentioned in 3.2, the phoneme is a discrete unit, i.e. the phoneme is not amenable to gradation. The distinction between different phonemes is clear-cut so that there is no phoneme that may be considered as lying between two phonemes. A phoneme is, as is also an archiphoneme, a sum of relevant features. (The relevant feature will be explained in detail in 3.3.3.) As Martinet (1965, p. 66) puts it, ...l'identification du phoneme resulte de l'enumeration de ses caracteristiques pertinentes, celles qui assurent la distinction entre ce phoneme et les autres phonemes de la langue. Or, dans le cas du phoneme, l'identification se confond avec la definition ('...the identification of a phoneme results from the enumeration of its relevant features, i.e. those features which ensure the distinction between this phoneme and the other phonemes of a given language. Now, in the case of a phoneme, its identification is equivalent to its definition [my translation].') It is during the course of the commutation test that the phonemes of a given language as well as the relevant features which define the phonemes are identified. The commutation test will be explained with illustrations in detail in Chapter 7. I now illustrate with a few examples the above-mentioned functionalist principle that a phoneme is a sum of relevant features. For example, the English phoneme /n/ is a sum of the relevant features "apical" and "nasal"; this sum can be conveniently expressed as "apical nasal", in the fashion I shall adopt in the rest of this book as I have done elsewhere. In other words, the English phoneme /n/ = "apical nasal", and vice versa. There exists the technical designation, 'phonological content' (cf. Trubetzkoy 1939, p. 59; Akamatsu 1988, pp. 68-70; Akamatsu 1992b, p. 32), which precisely means the sum of the relevant features of a given phoneme (also of an archiphoneme, a toneme, an architoneme). We can therefore say that the phonological content of Ini in English is "apical nasal". A phoneme may not be a sum of relevant features but consist of a single relevant feature. For instance, the English phoneme IV is equal to the relevant feature "lateral". Thus, in English, IV = "lateral", and vice versa. It is true that functionalists customarily operate with the definitional phrase 'a sum of relevant features' in reference to the phoneme (as much as to the archiphoneme, the toneme and the architoneme), and the use of the term 'sum' might be somewhat inappropriate in a case where a phoneme consists of a single relevant feature. This understood, I will continue to employ the phrase 'a sum of relevant features'. All the phonemes (and for that matter, the archiphonemes, the tonemes and architonemes too) of a given language obligatorily differ from each other in respect of their sums of relevant features; in other words, their phonological contents are mutually different, so that all the phonemes (and the archiphonemes, tonemes and architonemes) of a given language are necessarily distinguished from each other. The phoneme is a linguistic reality of phonic nature, but it is not physical, tangible or material in the sense that it is not pronounceable or audible in itself. (The phoneme is neither a phonetic segment nor a group of phonetic segments.) This particular character
Chapter 3: Distinctive units
19
of the phoneme applies to the signifier of a moneme as well, as the signifier consists of a single or a succession of phonemes (with the involvement of an archiphoneme(s), as the case may be) in a given order. In verbal communication by means of a language, what is pronounced or heard is a physical manifestation of single phonemes or sequences of phonemes via the phonic substance to which one generally refers as speech sounds, or simply sounds. In this book, I deliberately employ the term 'phonetic segment(s)' rather than 'sound(s)', and the terms 'vowel segment(s)' and 'consonant segment(s)' rather than 'vowel sound(s)' and 'consonant sound(s)’. For example, the above-mentioned English phoneme InJ is manifested by consonant segments like [n] (voiced apico-alveolar nasal) (cf. net), [n] (voiced apico-dental nasal) (cf. tenth), [ij] (voiceless apico-alveolar nasal) (cf. sniff), [n] (voiced apico-alveolar syllabic nasal) (cf. bidden), etc., and it is these consonant segments, not the phoneme Ini itself, that one pronounces or hears when English is spoken. These manifestations of a phoneme are generally known to functionalists as the phoneme's realizations, manifestations, implementations, actualizations or variants. I shall retain and use the term 'realization' in this sense in the rest of this book, as this term is the most widely used one, though personally I would prefer the term 'manifestation' or 'actualization'. In fact, I employed 'manifestation' in Akamatsu (1992b). What is meant as realizations of a phoneme by functionalists roughly corresponds, physically if not conceptually, to what is meant by 'allophones' of a phoneme by many non functionalists, but in this connection it should be emphasized that so far as functionalists are concerned, a sum of allophones do not constitute a phoneme as they do to many non-functionalists.14 It is both meaningless and impossible to say that a given phoneme has some definite number of realizations. The truth is that there occurs a realization of a given phoneme every time the phoneme occurs in the spoken form of a language. In case the realizations of a phoneme are manifestly different from each other and occur in mutual exclusiveness in different contexts, as the above examples of different realizations of Ini in English will have shown, they are regarded as combinatory variants (Trubetzkoy 1939, p. 44). Combinatory variants - otherwise called contextual variants or positional variants - are those realizations whose occurrences are contextually determined. The overall pattern of occurrences and non-occurrences of combinatory variants of a given phoneme is what is described as 'complementary distribution' (Martinet 1960, end of 3.16). The distinction between a phoneme and its realizations is of crucial importance to functionalists. As will have been seen above, I follow the traditional manner of distinguishing, between a phoneme and its realizations by using a pair of diagonal lines (/ /) for a phoneme and a pair of square brackets ([ ]) for each realization of the phoneme. The distinction between a phoneme and its realizations relates to an aspect of the distinction between phonology and phonetics. I shall dispense in this book with any specific and detailed account of the toneme and the architoneme, as they are irrelevant to Japanese. The interested readers may wish to consult section 4.2.2 (the tone) and section 4.2.4 (the architone) in Akamatsu (1992b, pp. 43-46 and pp. 49-50, respectively). I happen to employ in the present book the term 'toneme' rather than 'tone' and the term 'architoneme' rather than 'architone', but each pair of terms can to all intents and purposes be regarded as interchangeable in my concept and description of them. 3.3.2
The archiphoneme
I am aware that in the foregoing part of this book, the archiphoneme has occasionally been by necessity mentioned and exemplified, if not explained; this has been inevitable in parts of my foregoing exposition of certain other concepts operated
20
Japanese Phonology: A Functional Approach
with in functional phonology. It is in connection with the concept of neutralization of phonological oppositions that the archiphoneme can be best explained. This is why the archiphoneme will be abundantly mentioned in the whole of Chapter 6 in particular in which I talk about neutralization, neutralizable opposition and the archiphoneme itself. All the same, in the present section, I will make some preliminary remarks about the archiphoneme, and will make further remarks on the archiphoneme in 6.3. My remarks about the archiphoneme in both this section and 6.3 are to be considered complementary' and equally important. In the meantime the archiphoneme will continue to be mentioned here and there in the rest of this book. Pending further treatment of the archiphoneme in later parts of the present book, particularly in Chapter 6, I quote below, by way of a preliminary introduction to the archiphoneme, my definitions of the archiphoneme which I have previously provided in a couple of works of mine: The archiphoneme is a distinctive unit whose phonological content is identical with the relevant features common to the member phonemes of a neutralizable opposition, which is distinct from any of these member phonemes and which occurs in the position of neutralization. (Akamatsu 1988, p. 199)) ...the sum of the relevant features common to the member phonemes of a neutralizable (phonological) opposition [...] the common base of the relevant features of the member phonemes of a neutralizable (phonological) opposition. (Akamatsu 1992b, p. 47) It is perhaps appropriate to mention here that, whatever the different definitions of the archiphoneme - and there are - ascribable to functionalists, the Functionalist School espouses and operates with the concept of the archiphoneme in phonological analysis but that, as I noted in Akamatsu (1988, p. 2), the Prague School has abandoned it since about 1939 following its earlier introduction in phonology towards the end of 1920s and its recourse to it during the following decade. For the Prague School's abandonment of the archiphoneme the interested readers may wish to consult e.g. Vachek (1959, p. 110), Vachek (1960, 18), Vachek (1966, p. 62), Szemerenyi (1971, p. 68 fn. 28), Fischer-Jprgensen (1975, p. 30 fn. 12), Akamatsu (1988, pp. 202, 224, 281-286, 335, 341; and nn. 41, 411, 448, 480) and Akamatsu (1992c). Be that as it may, it is obvious that there is an insurmountable divergence between the Functionalist School and the Prague School in analyzing what I understand as the phenomenon of neutralization of phonological oppositions. 3.3.3
The relevant feature
The relevant feature is a linguistic unit which is crucially important to the phoneme and the archiphoneme (also to the toneme and the architoneme) which are only conceivable and definable in terms of relevant features. In fact, we have seen in 3.3.1 that a phoneme is a sum of relevant features. But now it needs to be emphasized that, as Martinet (1965, p. 69) points out, Ce n'est pas le phoneme, mais le trait pertinent qui est l'unite de base de la phonologie ('It is not the phoneme but the relevant feature that is the basic unit of phonology [my translation].') It is indeed easy to forget that the relevant feature is the basic unit of phonology upon which the other types of distinctive unit are founded. This is because one is generally more accustomed to operating with phonological oppositions whose terms
Chapter 3: Distinctive units
21
are phonemes (also archiphonemes, tonemes and architonemes, as the case may be) and only pays attention to phonological oppositions between relevant features when one is concerned with how the phonemes (and tonemes, as the case may be) are distinguished from each other or how neutralization of oppositions between phonemes (or tonemes) is to be accounted for. The widespread pronouncement, ascribable to many, if not all, non-functionalists, that the phoneme is the building block in phonology is extraneous to functionalists. The relevant feature too is endowed with the distinctive function. For example, the relevant feature "nasal" (cf. the English phoneme Ini) serves in that language to distinguish Ini from /t/ and /d/ which are characterized by the relevant feature "non nasal" (and certain other relevant features), as much as the relevant feature "non-nasal" of /t/ and /d/ serves to distinguish them from Ini. It is the distinction between the relevant features "nasal" and "non-nasal" that keeps apart the phonemes /n/ on the one hand and Id and Id/ on the other. Notice that two relevant features, "nasal" and "non nasal", are involved in the above example, and likewise two relevant features, "apical" and "dorsal”, are involved in the opposition between Itl and Ikl in English. In other cases, more than two relevant features are involved. For instance, the differentiation between /ml. Ini and /r)/ in English is made through the distinction between "bilabial" (of /ml), "apical" (of Ini) and "dorsal" (of /g/). Note that the relevant features of a phoneme, for example "dorsal nasal" of Ini in English, are concomitant, not successive, units, despite the convenient expression "dorsal nasal", in which the relevant features concerned are listed successively (due to the inevitable exigency of the writing mode). Note also that I deliberately use a pair of double, not single, quotation marks when indicating any relevant feature or a sum of relevant features. I shall use a pair of single quotation mark for other purposes in my expositions. A relevant feature of a given language is only conceivable and identifiable in opposition to one or more, as the case may be, relevant features of the same language. This implies important characteristics of the relevant feature. A few of them will be mentioned below. (i) A relevant feature is identified during the course of the commutation test performed on the basis of phonetic data drawn from a given language under analysis. The commutation test will be explained with illustrations in detail in Chapter 7. (ii) As follows from what has been said in (i), a relevant feature is conceivable and identifiable in respect of individual languages. There is no such thing as a relevant feature which is universally applicable to all languages. Nor is there any set of relevant features for all languages of the world so that each language chooses some, but not other, relevant features from this set. The relevant feature is language-specific. Even when different languages appear to have a relevant feature whose convenient designation happens to be the same, the identification of the relevant feature can only be made by knowing what other relevant feature(s) it is opposed to. For example, "apical" in Spanish (as in Ini) which is simultaneously opposed to "labial" in Spanish (as in /ml) and "palatal" in Spanish (as in lj\l) should not be confused with "apical" in English (as in Ini) which is simultaneously opposed to "bilabial" in English (as in /ml) and "dorsal" in English (as in /g/). (iii) The relevant feature is based on the concept of opposition, so that there is no relevant feature in any given language that is not opposable to another relevant feature or other relevant features, as the case may be, in the same language. I emphasized this point - which I referred to as 'the relevant feature and its inherent opposability (the functional principle of the relevant feature)' - in Akamatsu (1988, pp. 90-99). (iv) The relevant feature is never binary; in other words, the relevant feature is such that it is impossible to conceive of a given relevant feature that is somehow established for itself and with respect of which one talks about two opposite values, say + (plus) and - (minus). It is essential never to understand, e.g. "nasal" and "non-nasal", in terms
22
Japanese Phonology: A Functional Approach
of "+ nasal" and nasal" wherein there would be a single relevant feature "nasal" which can have the plus value or the minus value. The above-mentioned four characteristics - (i), (ii), (iii), (iv) - of the relevant feature make it absolutely clear that the relevant feature should never been mistaken for, or even worse, equated with, the distinctive feature as employed in certain types of phonology (for example, generative phonology). Conceptually, theoretically and practically, the relevant feature is completely extraneous to the distinctive feature employed in these types of phonology. It must be cautioned in this connection that a number of functionalists do employ the term 'distinctive feature' when they mean what I have been calling the relevant feature above (and will do so hereafter). This said, I should add that, were it not for the current widespread use in other types of phonology of the term 'distinctive feature' that is conceptually completely different from what I refer to as the relevant feature, I would be ready to employ, indeed prefer to employ, the term 'distinctive feature'. Indeed it would be more logical to apply the term 'distinctive feature' to a phonological feature which fulfils the distinctive function and instead to apply the term 'relevant feature' to refer comprehensively to any phonic feature that fulfils any one of the specific linguistic functions, including the distinctive function. See in this connection Akamatsu (1979a, fn. 2 on pp. 109-110), and also Akamatsu (1988, p. 89-90) where I approvingly quote from, inter alia, Marouzeau (1943, p. 173) and Marouzeau (1951, p. 173) and also provide note 160 on p. 447 where I express my agreement, concerning the use of the term 'relevant', with Mounin (1969, p. 207), Buyssens (1972, p. 1096) and Malmberg (1975, p. 120). But now I shall mention two further characteristics of the relevant feature worthy of attention. (v) Whether a relevant feature has bearings on an articulatory property (e.g. "lateral"), an acoustic property (e.g. "sibilant") or a perceptual property ("hiss") is of no importance to functionalists. All that matters is that a designation chosen for a relevant feature of a given language is clearly indicative of its essential differentiative character by virtue of which it is opposed to another or other relevant features in the same language. Such a designation need only be a conventional one and is not intended to be phonetically exhaustively descriptive. (vi) The relevant feature is a sum of multiple indissociable distinctive phonic properties. This concept of the relevant feature is perhaps less easy for some readers to comprehend at first. It has long been enunciated by Martinet, and I suggest that the readers should consult e.g. Martinet (1957, esp., p. 83 = 1965, esp. pp. 138-139) as well as Martinet (1956, 3.12). The readers may also wish to consult some other past relevant expositions; see e.g. Akamatsu (1988, pp. 99-103) for a sub-section entitled 'Internal structure of the relevant feature' and Akamatsu (1992b, pp. 39-43) for a sub-section indicated as (6). Suffice it to say here that, for example, the characteristic of a relevant feature being a sum of multiple indissociable distinctive phonic properties makes it unproblematic that a certain relevant feature, e.g. "voiced" in such English phonemes as lb/, Id/, Igl and /&,/ is physically differently manifested in different contexts. Such diverse physical manifestations of "voiced" in different contexts are extraneous to a timehonoured debate among many phoneticians and phonologists of English as to whether the phonetic property of lenisness (i.e. lack of muscular tension) which is constantly present or the phonetic property of voicedness (i.e. vibration of vocal folds) which is present in some contexts but not in other contexts in realizations of these phonemes should be accorded priority in characterizing these phonemes; that is, should these phonemes be characterized as "voiced" or "lenis"? To functionalists, none of the multiple distinctive phonic properties of a relevant feature (e.g. "voiced") takes precedence over the others. It is the distinctive functioning of a relevant feature as a whole, as a single global functional unit, that is taken crucially into account, irrespective of whether or not some of the multiple distinctive properties may be absent (e.g. voicedness) in some of the contexts in which the relevant feature occurs and which happen to favour the occurrence of such multiple distinctive properties. What is
Chapter 3: Distinctive units
23
of paramount importance is that in all the contexts in which, for example, "voiced" occurs, it is opposed to "voiceless", i.e. /b/ is opposed to /p/, /d/ is opposed to /t/, /g/ is opposed to /k/, and /eg/ is opposed to /tf/ in comparable contexts, despite any variant phonetic manifestations of "voiced" (for that matter of "voiceless"). The relevant feature will again be mentioned in connection with neutralization of phonological oppositions, since neutralization results ultimately from the cancellation of the opposition between relevant features.
Chapter 4: Opposition
25
Chapter 4 Opposition 4.1
General remarks
The concept of 'opposition' is fundamental in functional phonology, as the distinctive function and by the same token the distinctive unit are based on this concept. Different types of distinctive unit, i.e. the phoneme, the archiphoneme, the toneme, the architoneme and the relevant feature, have already been mentioned. The commutation test itself whereby the different types of distinctive units of a language are identified is crucially founded on the concept of 'opposition'. The concept of 'opposition' is also at the core of various other concepts in functional phonology. It should be emphasized that the concept of 'opposition' logically precedes that of 'distinctive unit'; without opposition there would be no distinctive units. 4.2
What is opposition? What is contrast?
In functional linguistics, the concept and term of 'oppositioi)' should never be equated or confused with those of lcontrasti. They are to be strictly distinguished from each other. The concept of opposition derives from the concept of paradigmatic relation, while the concept of contrast derives from the concept of syntagmatic relation. 'Linguistic units are said to be in paradigmatic relation to each other at a given point in the same context in a linguistic message' (Akamatsu 1992b, p. 110), and the linguistic units which are in paradigmatic relation are said to be 'in opposition to' (note, not 'in contrast with’), or 'opposed to' (not 'contrasted with'), each other.15 For instance, in the linguistic messages I see a red book, I see a black book, l see a green book, etc., the monemes red, black, green, etc. which occur in the same context, are in opposition to each other. The above is an exemplification of the concept of opposition with regard to monemes. The concept of opposition also applies to distinctive units, as I will show in detail in the next section. On the other hand, for example, the monemes I, see, a, red and book, in the first of the messages cited above, are in syntagmatic relation, and consequently 'in contrast with', not 'in opposition to', each other. The same can be said, mutatis mutandis, of the monemes in the two other messages cited above. The concept of contrast also applies to distinctive units. For instance, in the signifier of the moneme red, i.e. /red/, the relation between the three phonemes, /r/, /e/ and /d/, is syntagmatic and the three phonemes are therefore said to be in contrast with each other. One particularly good example of a phonic element that fulfils the contrastive function is accent. For example, accent in English fulfils the contrastive function, and the accented syllable and the unaccented syllable(s) within what can be termed 'accentual unit' are in syntagmatic relation and consequently in contrast with each other. In the word apology [a'po Isd31], for example, in which accent falls on the second syllable, there is contrast between the accented second syllable and the unaccented first, third and fourth syllables. Both the paradigmatic relation and the syntagmatic relation, hence opposition and contrast, are contracted by phonemes, archiphonemes, tonemes and architonemes, while relevant features entertain the paradigmatic relation only. 4.3
Opposition between distinctive units The type of opposition with which we are particularly, if not exclusively, concerned
26
Japanese Phonology: A Functional Approach
in this book is opposition between distinctive units, i.e. between phonemes, between archiphonemes, between phonemes and archiphonemes, or between relevant features. For instance, in the signifters /get/, /pet/, /set/, /vet/, /net/, etc. of the monemes get, pet, set, vet, net, etc., the initial phonemes are in opposition to each other. The difference between /g/, /p/, /s/, /v/, Ini, etc. in the signifiers of the above monemes is responsible for the distinction between these monemes. The replacement of any one of these phonemes by any other will result in the replacement of one moneme by another. The abovementioned phonemes are said to be in opposition to each other, expressible as, say, Igl /p/ - Is/ - /v/ - Ini - ... (where the dash means 'versus' or 'in opposition to'). It is in fact not only between phonemes that opposition can exist. Opposition can exist between phonemes and archiphonemes, or between archiphonemes. Witness, for example, Ini - l \ l - /p-b/ - /t-d/ - /k-g/ in the signifiers /sno:/, /sb:/, Is p-b o:/, Is t-d oil and Is k-g oil of the monemes snore, slaw, spore, store and score. It is important to acknowledge that there can be an opposition between a phoneme(s) and an archiphoneme(s), just as between phonemes or between archiphonemes. This is because both the phoneme and the archiphoneme are the distinctive units of the second articulation. Likewise, opposition can exist between tones, a toneme(s) and an architoneme(s), or architonemes. Opposition involving tonemes and architonemes is extraneous to Japanese, however, as it is not a tone language. But to cite some examples from Mandarin Chinese which is a tone language, there is opposition between four tones, generally known as tone(me) 1, tone(me) 2, tone(me) 3 and tone(me) 4, which I will symbolize as III, 121,13/ and /4/, exemplifiable in the four Chinese words earlier cited, i.e. the word pronounced [pa] and meaning 'eight', the word pronounced [pa] and meaning '(to) uproot', the word pronounced [pa] and meaning '(to) hold' and the word pronounced [pa] and meaning 'harrow'. Furthermore, in the context 'followed by 131', there is opposition, between III and 14/ and the architoneme 12-31, this last being associated with the neutralization of the opposition between 121 and 131 in the said context (for my exemplification of this neutralization of a tonematic opposition, see Akamatsu 1992a and Akamatsu 1992b, pp. 99-101). Relevant features are also in opposition to each other. For example, there is opposition between the relevant features "voiced" and "voiceless" which is involved in the opposition between, for example, /b/ and /p/, two of the phonemes of English, or between the relevant features "bilabial", "apical" and "dorsal" which is involved in the opposition between the three phonemes of English, /m/, Ini and /g/. Note that the relevant feature is not opposable per se to the phoneme, the archiphoneme, the toneme or the architoneme. Even in the case of the opposition between the relevant features "lateral" and "spirant", which opposition one conceives of in connection with the opposition between the phonemes IV and Irl in English, one has in reality to do with the opposition between "lateral" as being identical with IV and "spirant" as being identical with /r/, since IV consists of "lateral" only and Irl "spirant" only.16
Chapter 5: Phonological opposition
27
Chapter 5 Phonological opposition 5.1
General remarks
The type of opposition between distinctive units receives the technical designation of 'phonological opposition', which we shall employ in this book, too. The distinctive units of an individual phonological opposition are the terms of the phonological opposition. It is not the distinctive units by themselves that are of primary importance. This is because phonemes, archiphonemes, tonemes, architonemes and relevant features are only conceivable and identifiable by being terms of phonological oppositions. As Trubetzkoy (1933, p. 238) puts it, Une qualite phonologique n'existe que comme terme d'une opposition phonologique. ('A phonological quality only exists as a term of a phonological opposition [my translation].') This principle is reiterated in Trubetzkoy (1939, p. 85), who directly mentions 'distinctive property' (i.e. 'relevant feature') and 'distinctive opposition' (i.e. 'phonological opposition'), as follows. Es darf nicht vergessen werden, dafi eine distinktive Eigenschaft nur als Glied einer distinktiven Opposition besteht. ('It should never be forgotten that a distinctive property exists only as a term of a distinctive opposition [my translation].') In other words, it is a phonological opposition that logically precedes phonemes, archiphonemes, tonemes, architonemes, or relevant features, and not the other way round. Without the concept of phonological opposition, the concept of none of these distinctive units is possible. All this is unambiguously put by Trubetzkoy (1939, p. 60) when he writes as follows. Man darf ja nie vergessen, daB in der Phonologie die Hauptrolle nicht den Phonemen, sondem den distinktiven Oppositionen zukommt. ('It should never be forgotten that in phonology the principal role is played not by the phonemes but by the distinctive oppositions [my translation]'.) 5.2
Phonic difference
If the distinctive units which are opposed to each other constitute and are the terms of a phonological opposition, what do we understand, and how do we refer to, the fact that different phonetic segments occur in the same context, for example [m] and [f] in the phonetic sequences [mset] and [fse t], or [ph] and [p] in the phonetic sequences [diph] and [dip=], in English? ([ph] is [p] pronounced with the bilabial closure released with a slight aspiration, while [p=] has no such aspiration: the diacritic mark = signifies non aspiration.) Do we talk about a phonetic opposition in such a case? It would seem best to talk about a phonic difference, and not a phonetic opposition.17 Once we have ascertained that the difference between [m] and [f] in the above example leads to the distinction between the monemes mat and fa t in English, we consider this phonic difference to be phonologically relevant and therefore [m] and [f] to be realizations of ImJ and /f/ in English, /m/ and III are therefore opposed to each other and form the
28
Japanese Phonology: A Functional Approach
phonological opposition /m/ - /f/. On the other hand, once we have ascertained that the difference between [ph] and [p=] in English does not lead to the distinction between different monemes - both [diph] and [dip=] are recognized as relating to one and the same moneme dip in English, we regard this phonic difference as phonologically irrelevant and consider [ph] and [p=] as remaining a phonic difference. [ph] and [p=] are in fact different realizations of one and the same phoneme /p/; there is no question of seeing a phonological opposition here. In this book, I continue to employ the term 'phonological opposition' as indicated above, that is, an opposition between distinctive units, and the term ’phonic difference1as any difference between phonic elements (be they segmental or suprasegmental) irrespective of their being phonologically relevant or irrelevant. 5.3
Simple opposition and multiple opposition
It will have been understood from examples such as were earlier adduced in 4.3 that some phonological oppositions (e.g. /g/ - /p/, Id - /p-b/, /p-b/ - /t-d/, "voiced" "voiceless") consist each of two terms, while other phonological oppositions (e.g. /g/ /p/- /s/ - /v/- /m/; Ini - I d - /p-b/ - /t-d/ - /k-g/; /p-b/ - /t-d/ - /k-g/; "bilabial" - "apical" "dorsal") consist each of more than two terms. Thus, it would be convenient to operate with the concepts and designations corresponding to these two types of phonological opposition with regard to the number (either two, or more than two) of the terms of individual phonological oppositions. I have proposed (cf. Akamatsu 1988, p. 51 et p a s sim ) the designations 'simple phonological opposition' (in which case a phonological opposition consists of two terms) and 'multiple phonological opposition' (in which case a phonological opposition consists of more than two terms), or expediently, 'simple opposition' and 'multiple opposition' (see also Akamatsu 1992b, pp. 51-53). Incidentally, what I call 'multiple opposition' should never be confused with what Trubetzkoy (1939, p. 61 et passim) calls 'mehrdimensionale Opposition [multilateral opposition]'; the two types of phonological opposition have nothing to do with each other, conceptually as well as terminologically. Note that both 'bilateral opposition' (i.e. 'eindimensionale Opposition' as Trubetzkoy 1939, p. 61 et passim calls it) and 'multilateral opposition', with which I do not operate but Trubetzkoy does, are simple oppositions. (For my critique of 'bilateral opposition' and 'multilateral opposition', neither of which I operate with, see Akamatsu 1988, pp. 41-52.) The distinction between simple opposition and multiple opposition also applies to relevant features. On the one hand, relevant features like "voiced" and "voiceless" as in /p/ and /b/ in English are inherently opposed to each other only and form a simple opposition. On the other hand, relevant features like "bilabial", "apical" and "dorsal" as in /m/, Ini and /g/ in English, each of which is inherently opposed simultaneously to the other two relevant features, form a multiple opposition. Malayalam (spoken in southern India) has, for nasal consonant phonemes, the relevant features "bilabial", "dental", "alveolar", "retroflex", "palatal" and "velar", each of which is simultaneously opposed to the other five relevant features, and which therefore form a multiple opposition. I have proposed in Akamatsu (1988, pp. 275-276) the concept and designation of 'bioppositional relevant feature' for a relevant feature in a simple opposition, and the concept and designation of 'multi-oppositional relevant feature' for a relevant feature in a multiple opposition. I have subsequently employed them in Akamatsu (1992b, p. 38) too. The dichotomy implied by 'bi-oppositional' and 'multi-oppositional' is extraneous to the phoneme, the archiphoneme, the toneme and the architoneme since all the phonemes and the archiphonemes of a language are by their very nature opposable to each other, as are all the tonemes and the architonemes of a tone language.
Chapter 5: Phonological opposition
5.4
29
Exclusive opposition and non-exclusive opposition
I have earlier indicated (3.3.1, 3.3.2) that the phoneme or the archiphoneme is a sum of relevant features. For example, the phoneme /p/ in English is definable as the sum of the relevant features "voiceless", "bilabial" and "non-nasal", or shortly as "voiceless bilabial non-nasal", and the phoneme /b/ also in English, as "voiced bilabial non-nasal". These two phonemes form the phonological opposition /p/ - /b/ in English. The common base, i.e. the highest common denominator, of the relevant features of /p/ and Ibl is "bilabial non-nasal", which is not found in any other phoneme of English than /p/ and /b/ themselves. The concept and designation whereby to characterize a phonological opposition whose common base is exclusive to the terms of this phonological opposition, i.e. not shared by any other phoneme of the same language, is 'exclusive opposition’. (I coined the terms 'exclusive opposition' and 'non-exclusive opposition, which had not pre-existed, in Akamatsu 1988, p. 58.)18 Thus, /p/ - Ibl in English is an exclusive opposition. Another example of an exclusive opposition in English is /ml ("bilabial nasal") - Ini ("apical nasal") - /g/ ("dorsal nasal"), since the common base of these three phonemes, i.e. "nasal", is exclusive to these three phonemes of English, and so /ml -In i - /g/ in English is an exclusive opposition. On the other hand, the phonological opposition /p/ ("voiceless bilabial non-nasal") —/t/ ("voiceless apical non-nasal") is a non-exclusive opposition, since the common base of these two phonemes, i.e. "voiceless nasal", is also possessed by some other phonemes of English, e.g. /k/ ("voiceless dorsal non-nasal"). To determine whether a given phonological opposition in a given language is an exclusive opposition or a non-exclusive opposition can be a delicate task and requires a careful investigation during the course of the commutation test, in order to confirm whether or not the phenomenon of neutralization of a phonological opposition is involved. (The concept of neutralization will be explained in Chapter 6, esp. 6.2). For example, Iml - /g/ in English, and Ini - /g / also in English, are both non-exclusive oppositions, as the common base of Iml and /g/, i.e. "nasal", is also found in Ini, and the common base of Ini and /g/ is also found in Iml. However, Iml - Ini in English, whose common base is also "nasal", is not a non-exclusive opposition in a certain speech style (i.e. a quick, casual, relaxed speech) if Iml - Ini is neutralized before /f/ or /v/ in such a speech style, in which case the common base is not "nasal" but "non-dorsal nasal", which is not found in any phoneme of English and therefore Iml - Ini is an exclusive opposition. This point will be taken up again and explained in detail in 6.2 in connection with the concept of neutralization of phonological oppositions. An exclusive opposition can be a simple opposition (e.g. /p/ - /b/ in English) or a multiple opposition (e.g. Iml - I n i - /g/ in English), and a non-exclusive opposition can be a simple opposition (e.g. Iml - Ini in a quick, casual or relaxed speech in English) or a multiple opposition (e.g. /p/ —/t/ —/k/ in English). For my detailed discussion of exclusive opposition and non-exclusive opposition, see Akamatsu (1988, pp. 52-63). For my short presentation of exclusive opposition and non-exclusive opposition, see Akamatsu (1992b, pp. 53-55). The distinction between exclusive opposition and non-exclusive opposition should never be confused or equated with the distinction between bilateral opposition and multilateral opposition19 - a distinction I neither accept nor operate with (as can be seen in e.g. Akamatsu 1992b) - that Trubetzkoy (1939, p. 60 et passim) proposes and operates with. The two sets of two types of the distinction in question have nothing to do with each other either conceptually or terminologically. (For my discussion of bilateral opposition and multilateral opposition, see Akamatsu 1988, pp. 41-52). As we shall see later, the distinction between exclusive opposition and non exclusive opposition is importantly relevant to neutralization and neutralizable opposition.
30
5.5
Japanese Phonology: A Functional Approach
Correlative opposition Martinet (1955, p. 70) writes as follows about 'correlation': Series et ordres presupposent une unite plus grande les groupant en un tout, a savoir, la CORRELATION, qui comprend deux series paralleles et un certain nombre de phonemes accouples appartenant aux memes ordres. ('Series and orders presuppose a larger unit grouping them into a whole, i.e. a correlation, which comprises two parallel series and a certain number of pairs of phonemes belonging to the same orders [my translation].')
For the concepts of 'orders' and 'series', see Martinet (1955, 3.8ff et passim), Martinet (1960, 3.15), and Martinet (1962, pp. 77 and 81-82). Martinet's definition above of 'correlation' happens not to include an explicit reference to either 'mark', or 'relevant feature’ which 'mark' is said to be. Notice in this connection his following passage in Martinet (1960, 3.15): Le trait pertinent qui distingue les deux series [of a correlation] s’appelle la marque. Ici20 la marque est la « sonorite » ('The relevant feature which distinguishes the two series [of a correlation] is called the mark. Here the mark is "voice" [my translation].’) I personally do not operate with 'mark'.21 I have defined 'correlation' otherwise in Akamatsu (1992b, p. 57) as follows. A correlative opposition is a phonological opposition whose terms are distinguished from each other through two bi-oppositional relevant feature (e.g. "oral" - "nasal", "voiced" - "voiceless", "aspirated" - "unaspirated") and which has at least one other phonological opposition whose terms too are distinguished from each other through the same two bi-oppositional relevant features. The two terms of a correlative opposition are often referred to as a 'correlative pair'. For example, in English, /p/ - /b/, l\J —/d/, /k/ —/g/, /tf/ - //tf/ and /cfc;/ systematically fail to occur in word-initial context before /r/, and so on. This phenomenon may be referred to as 'systematic non-occurrence' (of a phoneme or phonemes) and I shall return to this subject, albeit very briefly, later in the present section. The other possibility, which is the one that concerns us in this section, is that a phonological opposition is neutralized in some contexts (contexts o f neutralization) but is valid, i.e. is not neutralized, in other contexts (contexts o f relevance). For the purpose of explicating the concept of neutralization of a phonological opposition and that of neutralizable opposition, I shall present below three of the instances of neutralization which occur in English, a language that I assume the majority of readers of this book are well acquainted with. We shall first consider the phonological opposition /p/ ("voiceless bilabial non nasal") - fbl ("voiced bilabial non-nasal"). This phonological opposition is valid in, for example, word-initial context (as in pit /pit/ - bit /bit/) or word-final context (as in cap /kaep/ - cab /kseb/), both of these being contexts of relevance. However, after Is/ this phonological opposition is neutralized (as in crisp /kris p-b/ or spot /s p-b o t/) since in
36
Japanese Phonology: A Functional Approach
this context, the context of neutralization, there is no possibility of distinguishing the signifiers of different words by virtue of /p/ - /b/; in other words, there can be no distinction between /p/ and /b/ possible after Is/. We therefore say that /p/ - /b/ is neutralized after /s/, and that Ipl - /b/ in English is a neutralizable opposition. The distinctive unit, indicated as /p-b/ (in /kris p-b/ or /s p-b ot/), is an archiphoneme - the archiphoneme has already been briefly touched on in 3.3.2 but will be explained in detail in the next section, 6.3 - which is associated with the above-mentioned neutralization and which consists in the sum of the common base of /p/ and /b/, i.e. "bilabial non-nasal". The distinction between the relevant features "voiceless" (of /p/) and "voiced" (of /b/) is nullified (or dissolved) in the neutralization of /p/ - /b/. Note that the distinction between the relevant features itself is not neutralized, but is nullified (or dissolved).27 We shall next consider the phonological opposition /m/ ("bilabial nasal") - Ini ("apical nasal") - /rj/ ("dorsal nasal") in English. This phonological opposition is valid in some contexts (e.g. word-final context as in /sim/ Sim - /sin/ sin - /sir)/ sing), contexts of relevance, but is neutralized in other contexts (e.g. before certain consonant phonemes as in lint Hi m-n-g t/, sample /sse m-n-g pi/ or pinch /pi m-n-g tf/), since in such contexts there is no possibility of distinguishing the signifiers of different words by virtue of /ml -In i - /g/; in other words, there can be no distinction possible between Iml, Ini and /g/ in these contexts, contexts of neutralization. Therefore, Iml - Ini - Iql in English is a neutralizable opposition, and the archiphoneme /m-n-g/ which consists in the common base of /ml, Ini and /g/, i.e. "nasal", is the distinctive unit which occurs in the context of neutralization, i.e. 'before Itl' in this example. We lastly consider the phonological opposition /ml ("bilabial nasal") - Ini ("apical nasal") in English. (This is an instance of neutralization which I referred to briefly in 5.4 in connection with the distinction between 'exclusive opposition' and 'non-exclusive opposition'.) We observe that Iml - Ini is valid in some contexts (e.g. in word-initial context as in /met/ met - /net/ net or in word-final context as in /sim/ Sim - /sin/ sin). In a speech style characterizable as quick, casual or relaxed, this phonological opposition is often neutralized before If/ or /v/, in which case e.g. camphor and anvil are both pronounced with [tg ] which is a voiced labiodental nasal, instead of camphor being pronounced with [m] which is a voiced bilabial nasal and anvil with [n] which is a voiced apical-alveolar nasal. This means that there is no possibility of distinguishing the signifiers of different words by virtue of /ml - Ini before /fI or /v/; in other words, there can be no distinction possible between /ml and Ini before HI or /v/ in this speech style. The phonological opposition Iml - Ini is a neutralizable opposition in such a case, and the archiphoneme lm-nl is the distinctive unit which occurs 'before III or /v/', contexts of neutralization, and whose phonological content is "non-dorsal nasal", for /m-n/ and /g/ are opposed to each other before Ifl or /v/'. (cf. e.g. sunflower, sun visor, harmful and warm vest having lm-nl, and e.g. long fan and long van having /g/). The signifiers of camphor and anvil are thus /'kas m-n fa/ and /'ae m-n vil/, respectively. The neutralization of Iml - Ini before lil or /v/ was discussed in Akamatsu (1973) and Aicamatsu (1992b, pp. 88-89), and why the phonological content of the archiphoneme Im-nJ is not "nasal" but "non-dorsal nasal" is explained in Akamatsu (1992b, pp. 89-91). A neutralizable opposition is necessarily an exclusive opposition,28 which is equivalent to saying that the common base of the phonological contents of the phonemes of a neutralizable opposition in a given language is exclusive to these phonemes and is not found in any other phoneme of the same language. This is because the phonological content of a given archiphoneme is necessarily different from that of any of the phonemes or that (those) of any other archiphoneme(s) susceptible of occurring in the context of neutralization where the archiphoneme in question occurs, simply because a given archiphoneme is opposed to any other phoneme(s) or archiphoneme(s) that occur(s) in the same context. It stands to reason that if this were not the case, an archiphoneme would be confused with another archiphoneme or with
Chapter 6: Neutralization, neutralizable opposition, archiphoneme
37
any phoneme susceptible of occurring in the context of neutralization. That a neutralizable opposition is obligatorily an exclusive opposition is also essential so that neutralization is distinguished from systematic non-occurrence of a phoneme(s) (or a toneme(s), for that matter), It is of fundamental importance to understand that neither or none of the phonemes of a neutralizable opposition occurs in the context of neutralization and that the archiphoneme occurs instead of the phonemes of the neutralizable opposition. (This point will be re-emphasized in 6.3.) For example, in English, in the context of neutralization for /p/ - Pol, e.g. 'after Is/', for example, neither /p/ nor Pol occurs, but it is the archiphoneme /p-b/ that occurs. Likewise, in the context of neutralization for Iml Ini - /g/, i.e. before certain consonant phonemes, in English, it is the archiphoneme /m-ng/ definable as "nasal" that occurs in that context, and not any of Iml, Ini and /g/. If a neutralizable opposition is not valid in some contexts, i.e. in the contexts of neutralization, it is not due to the systematic non-occurrence of either or any of the phonemes of the neutralizable opposition. In the case of the neutralization of Iml - Ini before HI or /v/, it is the archiphoneme /m-n/ that occurs - neither Iml nor Ini - in the context of neutralization, in w'hich context /g/ occurs which is not involved in the neutralization of Iml - Ini, and as a result /m-n/ and /g/ are opposed to each other. It is of equally fundamental importance to understand that neutralization of a phonological opposition is not caused by, nor is a phenomenon resulting from, the non occurrence, in a given context, of the phonemes of a phonological opposition except one while the other phoneme or phonemes occurs or occur in that context. For example, the non-validity in word-initial context of ImJ - InJ - /g/ (e.g. /mas t/ mat - Inst 1/ gnat - */gaet/) in English has nothing to do with neutralization but is a consequence of the systematic non-occurrence of /g/ in word-initial context. 6.3
The archiphoneme
Some preliminary remarks about the archiphoneme were made in 3.3.2 in the course of my description of various types of distinctive unit in Chapter 3. My further remarks about the archiphoneme to be given in this chapter which deals with neutralization, neutralizable opposition as well as the archiphoneme, are complementary to them. My remarks in 3.3.2 and in the present chapter are to be considered as equally important. I have already quoted in 3.3.2 some definitions of the archiphoneme previously given in Akamatsu (1988, p. 199) and Akamatsu (1992b, p. 47). Here is another definition of the archiphoneme: the archiphoneme is the common base of the (two or more) phonemes of a neutralizable opposition and is the distinctive unit that occurs in the context of neutralization. Just as in the case of the phoneme, the sum of the relevant features in terms of which the archiphoneme is definable is referred to as its phonological content. I have already demonstrated in the preceding section by way of the examples of a few neutralizable oppositions in English how the archiphoneme is definable as the common base of the phonemes of a neutralizable opposition. Furthermore, just like the phoneme, the archiphoneme is manifested by phonetic segments, which are known as realizations, manifestations, implementations, actualizations or variants of the archiphoneme. For example, the archiphoneme /p-b/ is variously realized by [p=] (voiceless unaspirated lenis bilabial plosive, the bilabial closure being released after the hold stage) as in spit, [V]/[Pn] (like the preceding but without aspiration, with or without preglottalization, and with non-release of the bilabial closure) or [^p]/[p] (voiceless bilabial plosive, with little or no aspiration, preglottalized or not, and the bilabial closure being released) as in wasp. The archiphoneme /m-n-g/ is variously realized by [n] before III or Id/ as in rent and sand, by [m] before /p/ or Pol as in camp and amber, [g] before !\d or Igl as in anchor
38
Japanese Phonology: A Functional Approach
and anger, and so on. As in the case of realizations of a phoneme, those realizations of an archiphoneme which are physically appreciably different from each other associated with different contexts in which the archiphoneme occurs, as the above examples show, are referred to as combinatory variants (of the archiphoneme). It remains true, however, that, as in the case of realizations of a phoneme, there occur as many realizations of an archiphoneme as as the occurrences of the archiphoneme in speech, and it is therefore both impossible and meaningless to say that such-and-such an archiphoneme has such-and-such number of realizations. In the context of relevance where a neutralizable opposition is valid, there is neither need nor justification to postulate the archiphoneme, since in the context of relevance the common base of the phonemes of the neutralizable opposition does not function in any way as a distinctive unit. It goes without saying that in the context of neutralization, it is the archiphoneme to the exclusion of the phonemes of the neutralizable opposition that occurs. As previously emphasized (3.1), the archiphoneme fulfils, as does the phoneme, the distinctive function, and is therefore one of the types of distinctive unit of the second articulation. There does not seem to be, to my knowledge, a convenient term whereby to refer at the same time to both the phoneme and the archiphoneme, which are the minimum (distinctive) units of the second articulation. The full complement of the distinctive units of the second articulation of a given language consists of both the phonemes and the archiphonemes of the language (that is, if the phonological system has archiphonemes as well), so that any list of the distinctive units of a language relevant to the secondary articulation should list both the phonemes and the archiphonemes (if the archiphonemes are in the phonological system of the language), not just the phonemes.29 An archiphoneme is opposable to any other archiphoneme(s) and to any phoneme(s) which is(are) susceptible of occurring in the context of neutralization where the archiphoneme in question occurs. In fact, all these archiphonemes and the phonemes are opposed to each other in that context. Thus, for example, in English, in the context 'after Is/' (cf. spear) which is a context of neutralization for /p/ - /b/, there is opposition between all the distinctive units that are susceptible of occurring in this context; namely the archiphonemes /p-b/ (cf. spit Is p-b it/), /t-dJ (cf. still /s t-d il/) and /kg/ (cf. skill /s k-g il/), the phonemes /m/ (cf. Smith /smi0/), /n/ (cf. snap /snap/)), /]/ (cf. slack /slaek/), If I (cf. sphene /sfi:n/), /v/ (cf. svelt /svelt/) and Id (cf. Srinagar) /sri'riAga/. The above-mentioned opposition, a multiple opposition in this case, is of course possible because the phonological contents of all the archiphonemes and the phonemes susceptible of occurring in the context 'after Is/' are different from each other. The phonological content of the archiphoneme /p-b/ in English is "bilabial non-nasal", which is different from the phonological content of any other archiphoneme or any phoneme susceptible of occurring in the context 'after Is/’. The phonological content of the archiphoneme /t-d/ is "apical non-nasal" (cf. still /s t-d il/), that of the archiphoneme /k-g/ "dorsal non-nasal" (cf. skill Is k-g il/), that of the phoneme /ml "bilabial nasal" (cf. Smith /smiG/, that of the phoneme In/ "apical nasal" (cf. snap /snaep/), that of the phoneme f\J "lateral" (cf. slack /sleek/), that of the phoneme HI "voiceless labiodental” (cf. sphene /sfi:n/), that of the phoneme /v/ "voiced labiodental" (cf. svelt /svelt/), and that of the phoneme III "spirant" (cf. Srinagar /sn'nAga/). There is, even among functionalists, no hard-and-fast single way to symbolize an archiphoneme in phonological notation. Many functionalists prefer to employ a single special symbol (generally a capital letter, but not always) for an archiphoneme. Thus, some employ a capital letter /P/ rather than /p-b/ or /p/b/ for the archiphoneme associated with the neutralization of Ipl - Ibl in English. Some employ /N/ rather than /m-n-g/ or/m/n/g/ for the archiphoneme associated with the neutralization of /ml - InJ -
Chapter 6: Neutralization, neutralizable opposition, archiphoneme
39
/r)/ in English. There will, however, be a problem if a language has more archiphonemes than there are suitable capital letters to represent all of them. For example, a problem will arise if, having allotted the capital letter, N, hence /N/ rather than /m-n-g/, to notate the archiphoneme associated with the neutralization of /m/ - /n/ - /g/, one wishes to employ also a capital letter in order to notate the archiphoneme associated with the neutralization of ImJ - Ini before /f/ or l\l in a certain speech style (quick, relaxed, casual) in English, i.e. the archiphoneme that I have notated Im-nl further above. My own preference, when presenting archiphonemes individually, as in a list of the distinctive units of a language, is to notate archiphonemes by means of the symbols standing for the terms of neutralizable oppositions and linked by a hyphen(s), thus, e.g. /p-b/, Im-nl, /m-n-g/. Using a sequence of letters which stand for the terms of a neutralizable opposition, be it a simple opposition or a multiple opposition, cumbersome as it may be, is useful in clearly indicating straightaway and unambiguously the identity of a neutralizable opposition. In presenting archiphonemes in running phonological notation, however, in which such symbols for archiphonemes occur in combination with symbols for other distinctive units, as in /bae m-n-g k-g s-z/ banks, I provide spaces on the sides of the symbols for archiphonemes in order to make for desired legibility, as in /bae m-n-g k-g s-z/, and this practice on my part will be maintained throughout this book. Provision of space will be easily appreciated when one compares, for example Is p-b it1 with /sp-bit/, or /'ae m-n-g ka/ with /'ae m-n-gka/. Where an archiphoneme is associated with the neutralization of a multiple opposition whose terms are considerable in number, the use of a capital letter is convenient, especially in running phonological notation, and I myself am not averse to notating such an archiphoneme by means of a suitable capital letter in running phonological notation for the sake of readability, as will be seen later in this book, i.e. IQI (= /p-p’-b-b’ -t-d-k-k’g-g’--g-h-q-sz-ts-cq-z-r-r’/), INI (= /m-m’-n-n’-g-g’/ ) and ICI (= /p-p’ -b-b’-t-d-k-k’-g-g’-$-g-h-q-j-s-zts-cc-z-r-r’-m-m’-n-n’-g-g’-iq/), as in /haQkoo/, /haNko/ and /haCnoo/ in my phonological analysis of Japanese. In such cases, the above-mentioned use of spaces will be unnecessary, as in /laeNp/, /sPit/ and /sKip/ in which INI is equivalent to /m-n-g/, and IP/ to /p-b/ and /K/ to /k-g/. For a detailed general discussion by me on the question of symbolizing archiphonemes, see Akamatsu (1988, pp. 314-331). Symbolization of archiphonemes continues to interest functionalists, without reaching unanimity. One important reason for disfavouring the use of a capital letter to symbolize an archiphoneme is given by Martinet (1989, pp. 115-117), who says that symbolizing an archiphoneme by a capital letter (e.g. /T/ as in /ra:T/ for Rat ’council' or for Rad 'wheel' in German) may wrongly suggest that there is an alternation between /t/ (cf. Rate 'councils') and H I or between /d/ (cf. Rader ’wheels') and /T/, i.e. Id - H I or /d/ - HI, which there is not. Martinet therefore prefers the notation like /ra:t/d/ rather than /ra:T/ which will ward off any suggestion that there are alternations like /t/ ~ H I and /d/ - HI. The same view was seen previously expressed in Martinet (1976, p. 101) and Martinet (1978, pp. 8-9).30 It seems then that the question of how to symbolize an archiphoneme may not be purely a practical one but is pregnant with theoretical implication.
Chapter 7: The commutation test
41
Chapter 7 The commutation test 7.1
The commutation test, its place in functional phonology
We have seen in 5.1 that a phonological opposition logically precedes distinctive units like phonemes, archiphonemes, tonemes, architonemes and relevant features, and I have quoted in this connection a relevant passage from Trubetzkoy (1939, p. 60). (Trubetzkoy refers to phonemes, but not to archiphonemes, tonemes, architonemes and relevant features.) I am in agreement with Trubetzkoy here. Distinctive units are only conceivable and identifiable as terms of phonological oppositions. Trubetzkoy (1936b, p. 17) at an earlier date writes as follows. Der Anfang jeder phonologischen Beschreibung besteht in der Aufdeckung der in der betreffenden Sprache bestehenden bedeutungsdifferenzierenden Schallgegensatze. ('The beginning of every phonological description consists in discovering distinctive oppositions that exist in the language concerned [my translation].') Notice particularly that Trubetzkoy mentions phonological oppositions, not the phonemes, in the above quoted passage. The logical precedence of a phonological opposition over its terms having been accepted, the discovery of individual phonological oppositions and their terms are theoretically simultaneous. We confirm whether or not a given phonic difference, i.e. a difference between phonic elements, is phonologically distinctive or relevant. If the phonic difference proves to be distinctive, we have a phonological opposition and also the terms (distinctive units) of the phonological opposition. If, on the other hand, the phonic difference proves to be nondistinctive, i.e. phonologically irrelevant, we do not have a phonological opposition, and the phonic difference remains such. It is by recourse to the analytical device called the commutation test that we discover the phonological oppositions and the distinctive units which constitute the terms of the phonological oppositions of a given language. Actually, although phonological oppositions logically precedes the terms of the phonological oppositions, operationally the establishment of the phonemes precedes the identification of the phonological oppositions whose terms the phonemes are. It is also at the same time by recourse to the commutation test that we identify the relevant features in terms of which the distinctive units themselves are necessarily identified. Martinet (1947, p. 44 = 1965, p. 67) writes. ...il nous faudra pousser la commutation assez loin pour pouvoir degager non plus seulement les phonemes, mais les traits pertinents eux-memes. ('...we shall have to push the commutation test far enough to be able to elicit not only the phonemes, but the relevant features themselves [my translation].') Other essential tasks are also carried out during the course of the commutation test: (1) identifying the distinctive units of various types, by identifying at the same time the relevant features in terms of which the distinctive units are definable; (2) establishing the types of phonological opposition whose terms these distinctive units are; (3) discovering and identifying all the facts of neutralization such as the neutralizable opposition concerned and the archiphoneme associated therewith and defining the archiphoneme in terms of relevant features as well as the contexts of neutralization (distinct from those of relevance); and (4) identifying cases of systematic non occurrence of phonemes in some contexts.
42
Japanese Phonology: A Functional Approach
The commutation test is fundamentally founded on the two criteria of opposition, i.e. phonic difference (note, not phonetic similarity) and semantic difference.31 In other words, if a phonic difference is confirmed to link to a semantic difference, the phonic difference is regarded as constituting a phonological opposition. As I put it in Akamatsu (1992b, p. 60): Commutation is said to occur if the replacement of one phonic element by another or other phonic elements at a given point in a given set of sequences of phonic elements [...] which are identical but for the replaceable phonic elements results in a change in the identities of linguistic forms. For all its significant importance and necessity in functional phonology, the commutation test is generally not sufficiently explained. Detailed accounts of how to actually perform the commutation test step by step remain relatively scarce, so far, in functionalists' writings. It is important to mention Martinet (1947, pp. 41-46 = 1965, pp. 63-69) in particular and also Martinet (1960, 3.10-3.15) for both theoretical and practical aspects of the commutation test presented therein. I myself, in Akamatsu (1992b, pp. 60-80), have given an extended account of the principles of the commutation test and how to perform the commutation test on some material drawn from French and invite the readers to take a look at it, with regard to both theoretical and practical aspects of the commutation test. Incidentally, the readers will find a useful guide regarding the writings by functionalists on various aspects of the commutation test under the rubric 'Commutation' on page 347 in La Linguistique 21 (1981) or, in an updated version, under the same rubric on page 51 in La Linguistique (1996). 7.2
What is a commutative series?
A commutative series consists of a number (which may vary from one commutative series to another) of such word-size linguistic forms which are distinguished from each other through either minimal phonic differences or near-minimal phonic differences, the rest of the words being phonetically the same. What is meant by 'minimal phonic difference' and 'near-minimal phonic difference' by implication will be explained later. These words, which are constituents of a commutative series, if they are differentiated from each other through minimal phonic differences, will be referred to as 'minimal multiplets', but as 'near-minimal multiplets' if they are differentiated from each other through near-minimal phonic differences. The term 'minimal multiplet' is my coinage first employed in Akamatsu (1992b, p. 52 et passim), so that minimal multiplets are 'a set of more than two minimally differentiated terms' (Akamatsu 1992b, p. 52). The term 'nearminimal multiplet', also my coinage, which I employ in the present book for the first time, is equivalent to the term 'near-minimal item' or 'quasi-minimal item' which was already used in Akamatsu (1992b, p. 75 et passim). The term 'word-size linguistic forms', which I have used above, obviously lacks theoretical precision as used in linguistics, but is practically convenient. I shall use the term and concept of 'word' in my expositions in the rest of this book. A word is a significant unit, not necessarily a minimal significant unit in a language, which consists in a union of a signifier and a signified. A word should not be simplistically equated with a moneme, however. A word may correspond to a moneme, but also in many cases to a combination of monemes, either sequential or concomitant. A commutative series can be illustrated by, for example, a somewhat hypothetical series consisting of heed, hid, head, had, hard, hod, hoard. Hood, hood, Hudd, *?[had], hade, Hoad, hide, *howd, *hoyd, *hierd, -haired, hoard, *hoord ,32 (I will explain below why I say 'a somewhat hypothetical series'.) Different commutative series should be associated with different phonetic contexts so that the commutative series given
Chapter 7: The commutation test
43
above is associated with the phonetic context [h - d]. The dash in [h - d] indicates the point at which each of the vowel segments occurs whose phonological status the commutation test is intended to establish. In giving an example of commutative series above, I expediently presented in their orthographic form (thus, heed, hid, etc.) - as I continue to do in the rest of the book the words which constitute the (either minimal or near-minimal) multiplets of the commutative series. But of course the commutation test operates on the phonetic form of the multiplets (thus, [hi:d], [hid], etc.). Strictly speaking, the multiplets of the commutative series should be presented in phonetic notation, not in orthographic form. Let us not forget that we are here interested in any repercussion that their vocal expression may have on the corresponding semantic content. The commutative series, heed, hid, head, had, hard, hod, hoard, Hood, hood, Hudd, herd, *?[had], hade, Hoad, hide, *howd, *hoyd, *hierd, -haired, hoard, *hoord, given above, consists of minimal multiplets, as all the words are minimally differentiated from each other, i.e. only through the difference in the word-medial vowel segments of the respective words, the rest being identical. The commutative series given above is actually somewhat hypothetical in that those words which are marked with asterisks are non-existent in current English vocabulary but phonetically permissible in English (with the exception of *?[had] which is supposed to occur accented). (The orthographic form given to the non-existent words is of course concocted and my use of ? simply means that no appropriate orthographic form exists in English.) The rest of the words are of attested in English. It is merely for the sake of illustrating a commutative series which consists of minimal multiplets that I have resorted to the somewhat hypothetical commutative series above. A commutative series can also be illustrated by, for example, the series consisting of pit, bit, tit, dit, kit, git, mitt, knit, fit, vit, thick, this, sit, zit, shit, gite, chit, jet, writ, hit, yet and wit, which is associated with the phonetic context [- it]. Some of the words presented above are minimal multiplets (namely pit, bit, tit, dit, kit, git, mitt, knit, fit, vit, sit, zit, shit, chit, writ, hit, yet, wit), as each differs minimally from any of the others, i.e. only through the difference in the word-initial consonant segments, the rest being the same, while the remainder of the words (namely thick, this, jet) are near-minimal multiplets. Words that might be pronounced [6it], [Sit] and [d3 i t], which would constitute minimal multiplets in this commutative series, are unattested in present-day English, though phonetically permissible. Of the three near-minimal multiplets, thick differs from any one of the minimal multiplets by not merely having [0] initially but also having [k] finally instead of [t]; this by having not merely [8] initially but also [s] finally instead of [t]; and finally jet by having not merely [d3 ] initially but also [e] medially instead of [i]. It will be seen that thick, this and jet are near-minimal multiplets as they differ from each other and from the other multiplets (i.e. the minimal multiplets) nearminimally, i.e. through more than just the difference in the word-initial consonant segments. In the above-cited commutative series which consists of a mixture of minimal multiplets and near-minimal multiplets, it is the phonological status of the word-initial consonant segments that the commutation test is intended to establish. 7.3
Does a commutative series consist of minimal multiplets or nearminimal multiplets?
Commutative series consisting entirely of minimal multiplets are an exception rather than the rule, as such commutative series, though to all intents and purposes ideal, are difficult to come by in practice in languages. One example of a commutative series consisting entirely of minimal multiplets is a commutative series which is available in Japanese and whose minimal multiplets are [? i"1?] 'mind', [7 e"1?] 'painting', [7 a"1?] 'Good God!', [7 o'1?] 'tail', and [7 ui "*7] 'cormorant', associated with the phonetic context
44
Japanese Phonology: A Functional Approach
[#?-?#]. ([?] is a glottal plosive, the symbol # represents a pause, and the symbol "* signifies that any phonetic segment(s) that follow this symbol are pronounced on a lower pitch than the phonetic segments before this symbol.) In presenting a commutative series, it is customary to attach the glosses for the multiplets, as I have done above for the multiplets drawn from Japanese. I have deliberately not done so for those drawn from English as I assume that the readers will know the meanings of all of them. This is to confirm that the phonic difference, be it minimal or near-minimal, between the multiplets definitely corresponds to differences of semantic content. One will generally find, in establishing commutative series, that the predominant majority of commutative series that one can obtain are a mixture of minimal and nearminimal multiplets. The possibility that a commutative series consists in its entirety of minimal multiplets is in practice rare. This will have been testified by the fact that such a commutative series consisting of heed, hid, head, had, hard, hod, hoard, Hood, hood, Hudd, herd, *?[had], hade, Hoad, hide, *howd, *hoyd, *hierd, -haired, hoard and *hoord, given above, is in reality a somewhat hypothetical commutative series if it is to be viewed as consisting of minimal multiplets. We have seen that the commutative series consisting of pit, bit, tit, dit, kit, git, mitt, knit, fit, vit, thick, this, sit, zit, shit, gite, chit, jet, writ, hit, yet and wit is actually a mixture of minimal multiplets and near-minimal multiplets. Indeed, more often than not, the analyst will find himself performing the commutation test with commutative series which consist of near-minimal multiplets as well as minimal multiplets. 7.4
Are minimal multiplets and near-minimal multiplets equally valid in performing the commutation test?
Given an incorporation, to varying degrees, of near-minimal multiplets in establishing commutative series and a significant measure of difficulty to obtain commutative series consisting entirely of minimal multiplets, it is a worthwhile question to ask whether near-minimal multiplets enjoy the same degree of validity for the commutation test to be performed. The answer to this question is that minimal multiplets and near-minimal multiplets of a commutative series are, for the purpose of the commutation test, of comparable and equal validity and are on an equal footing. There is no sense in which minimal multiplets can be said to be superior to near-minimal multiplets in this regard. (But see 7.5 below.) The role of near-minimal multiplets should never be underestimated. The near-minimal multiplets, thick, this and jet, in the commutative series associated with the phonetic context [- it], are as valid as any of the minimal multiplets (pit, bit, tit, dit, etc.). The acceptance of thick, this and jet results in modifying the very phonetic context [- it], with which the commutative series in question is supposed to be associated, to [- ik] (thick), [- is] (this) and [- et] (jet). However, this poses no problem at all to the validity of the near-minimal multiplets, as will be explained in the next section. 7.5
What is the criterion whereby to accept or reject near-minimal multiplets?
For near-minimal multiplets to enjoy the same degree of validity as minimal multiplets as constituents of a commutative series, a certain criterion must be met whereby the acceptance of near-minimal multiplets is unquestionably justified. Short of satisfying this criterion, prospective near-minimal multiplets are to be dismissed. The criterion in question is entirely of phonetic nature. We need to make sure that the occurrence of the phonetic segment whose phonological status the commutation
Chapter 7: The commutation test
45
test is intended to establish is independent of any phonetic influence that might be thought to be exerted by that/those phonetic segment(s) whose presence prevents a near-minimal multiplet from being a would-be minimal multiplet. For example, we need to confirm before accepting thick as a near-minimal multiplet in the commutative series associated with [- it] that the presence of [k] in word-final context in thick instead of [t] is not thought to obligatorily make [9] occur rather than some other consonant segment in word-initial context. Similarly, before accepting this as a near-minimal multiplet in the same commutative series, we must make sure that the occurrence of [s] in word-final context here is not thought to necessarily make [5] occur rather than some other consonant segment in word-initial context. Finally, likewise, before accepting jet as a near-minimal multiplet in the same commutative series, we should confirm that the occurrence of [e] instead of [i] (of [- it]) is not thought to necessarily make [d3 ] occur rather than some other consonant segment in word-initial context. The all-clear can be given in all three cases here, and so thick, this and jet can be unproblematically accepted as near-minimal multiplets, with equal validity with the minimal multiplets, in the commutative series concerned. I have earlier given a somewhat hypothetical commutative series, namely heed, hid, head, had, hard, hod, hoard, Hood, hood, Hudd, herd, *?, hade, Hoad, hide, *howd, *hoyd, *hierd, -haired, hoard and *hoord. This commutative series need in fact not be thrown overboard if we replace *howd, *hoyd, *hierd and *hoord by near-minimal multiplets, which might be, for instance, howdy, hoyden, beard and gourd, respectively. The afore-mentioned criterion whereby to accept or reject near-minimal multiplets is met: the addition of [d] (howdy) is not thought to necessarily bring about the vowel segment [au] (*howd); the addition of [(a )n] (hoyden) is not thought to obligatorily bring about the vowel segment [oi] (*hoyd)\ the change of [h] (*hierd) to [b] (beard) is not likely to necessarily bring about the vowel segment [ia ]; and the change of [h] (*hoord) to [g] (gourd) is unlikely to bring about the vowel segment [ua ]. The somewhat hypothetical commutative series whose multiplets are supposed to be all minimal multiplets can thus be unproblematically transformed into a commutative series consisting of a mixture of minimal multiplets and near-minimal multiplets. Undoubtedly the analyst's decision as to whether to accept or reject a given attested word in the language under phonological analysis as a near-minimal multiplet in a given commutative series requires a good prior knowledge of general phonetics and furthermore, in particular, the normal phonetic phenomena of the language which is being phonologically analyzed. Without such knowledge the analyst would be unnecessarily wary of admitting near-minimal multiplets or, alternatively, guilty of inappropriately admitting some attested words as near-minimal multiplets. Finally I should add that my concept of 'near-minimal multiplet' (and my acceptance of a near-minimal multiplet thus conceived) seems to be broader than Martinet's. We read as follows in Martinet (1956, 3.16). [...] Dans ce cas, on peut se permettre de rapprocher des mots d'aspect phonique assez different, mais qui presentent dans des contextes absolument identiques les deux phonemes qu'il s'agit de comparer [italics added], En frangais, par exemple, le rapprochement animal-bemol suffirait a etablir l'independance phonologique de /a/ et de h i puisque, places dans un contexte identique [italics added] entre /ml et /l/ en syllabe finale, ils ne se confondent pas. Les rapprochements de ce type ne represented, bien entendu, qu'un pisaller [italics added]. ('[...] In this case, we can allow ourselves to bring together words which are phonetically rather different but which present in absolutely identical contexts the two phonemes to be compared with each other. In French, for example, bringing together animal-bemol would be sufficient to establish the phonological independence of lal and h i since, placed in an identical context between ImJ and III in syllable-final context, /a/ and h i are not confused. The bringing together of this type represents, of course, only a last resort
46
Japanese Phonology: A Functional Approach
[my translation].') 7.6
Why multiplets, not minimal pairs, for the commutation test?
It will have been noticed that I do not have recourse to the concept and term of 'minimal pair’ which by definition consists of two and no more than two multiplets, whether they be minimal multiplets or near-minimal multiplets. There has been excessive importance attached to 'minimal pair' in the phonological literature in general (cf. e.g. Walter 1977, pp. 24-27, i.e. the section entitled 'LE TEST DE LA COMMUTATION ET LES PAIRES MINIMALES'). I have warned against excessive emphasis placed on minimal pairs or near-minimal pairs in performing the commutation test in the past (see e.g. Akamatsu 1992b, p. 52 et passim-, Akamatsu 1995). My warning, in the context in which it was sounded, related, partly but importantly, to my warning against excessive importance widely given to what I call 'simple opposition' as distinct from what I call 'multiple opposition’ (see e.g. Akamatsu 1992b, p. 52; Akamatsu 1995). Actually, my warning against excessive importance attached to minimal pairs is germane to the commutation test, too. It is true that strict recourse to minimal pairs would, unjustifiably, disallow recourse to near-minimal multiplets, but this is a minor reason. There are more serious reasons against objecting to adhering exclusively to minimal pairs in performing the commutation test. I will explain them below. Recourse to minimal pairs does not lead to correct identification of either the relevant features or the distinctive units which are sums of relevant features. The only exception to this would apply to a language in which there are two and no more consonant phonemes and/or two and no more vowel phonemes and consequently a commutative series consisting of three or more items is impossible. Whether or not such a language actually exists is irrelevant here. In functional phonology there is no question of being able to identify, for example, the distinctive units known as /p/ and /b/ in English as "voiceless bilabial non-nasal" and "voiced bilabial non-nasal", respectively (these definitions of /p/ and Ibl can only have been arrived at during the course of the commutation test performed on a number of commutative series associated with mutually different phonetic contexts) by recourse to a minimal pair like [phaet] pat vs. [baet] bat. (The small subscript symbol . as in [b] signifies 'devoicing', and the small raised symbol h in [phae] aspiration.) Even recourse to two or more minimal pairs like [pha»t] pat vs. [baet] bat, [mop] mop vs. [mob ] mob, ['jipan] Ripon vs. ['jiban] ribbon, and so on, still fails to lead to a correct identification of /p/ and Ibl. The most that one could elicit are the relevant features "voiceless" for /p/ and "voiced" for Ibl. The relevant feature "non-nasal" will not be elicited without opposing e.g. [phaet] pat and [baet] bat to [maet] mat, [mop] mop and [mob] mob to [mom] mom, [’jipan] Ripon and [’jiba n] ribbon to ['leman] lemon, and so on. Such will then be recourse to 'minimal quadruplets', no longer 'minimal pair', anyway. As for eliciting the relevant feature "bilabial", [phaet] pat should be opposed to [thaet] tat and [khaet] cat and [tjhaet] chat, since "bilabial" is opposed to "apical", "dorsal" and "hush". In other words, the relevant feature "bilabial" cannot be elicited by recourse to a minimal pair, [phaet] pat vs. [thaet] tat, or [phaet] pat vs.[ khaet] cat or [phaet] pat vs. [tfhset] chat. What has just been said above about eliciting the relevant feature "bilabial" equally applies to e.g. [bat] bat, which should be opposed to both [daet] DAT and [gaet] gat and [d3 et] jet (cited in the absence of an English word which would be pronounced [djas t] *jat), which would again mean operating with a minimal quadruplet and not a minimal pair. Apart from the fact that recourse to a minimal pair or minimal pairs is inadequate for the purpose of correctly identifying the phonological content of the phonemes of a language (English in this case), it ignores the functional principle that all the phonemes of a language are opposed to each other, so that, in the case of
Chapter 7: The commutation test
47
English, /p/, for example, is opposed to not only lb/, /m/, /1/ and fk/ - which are the direct neighbours in the English consonant phoneme system - but also to all the other consonant phonemes of English. The functionalist cannot forget the principle that a phoneme of a language is what the other phonemes are not. It is significant that recourse to a minimal pair or minimal pairs is at best tantamount to operating necessarily with a part of a. single commutative series. Recourse to multiplets does not automatically guarantee operating with a number of commutative series, either. This takes us to the next section. 7.7
The necessity of operating with a number of commutative series
I have mentioned in 7.2 that the functionalist performs the commutation test on a number of commutative series which are associated with mutually different phonetic contexts and have emphasized, in 7.3 and 7.6 in particular, that commutative series consist of multiplets (not a minimal pair(s)), be they minimal or near-minimal multiplets. It is neither sufficient nor defensible to operate with just one commutative series associated with one particular phonetic context and consisting of just one series of multiplets. During the course of the commutative test, we should compare with each other the results of the phonological analysis obtained on the basis of the different commutative series before arriving at the final conclusion about the relevant features and (consequently) the distinctive units themselves, and neutralization of phonological oppositions and/or systematic non-occurrence of a phoneme(s). It goes without saying that working with a single commutative series consisting of multiplets is definitely bound to miss out instances of neutralization of phonological oppositions and/or systematic non-occurrence of a phoneme(s). Depending on the different commutative series obtained, even working with a few commutative series may or may not lead the analyst to detect instances of neutralization of phonological oppositions and/or systematic non-occurrence of a phoneme(s). It is for this reason that 'incomplete commutative series' (which I shall call 'non-maximally differentiative commutative series' in the next section) are just as necessary and illuminating as 'complete commutative series' (called 'maximally differentiative commutative series' also in the next section) and both types should equally be taken into consideration. It will have been guessed that my definitional concept of the commutation test is in spirit basically in agreement with the theory of the micro-phoneme and the macro phoneme advanced in the mid-1930s by Twaddell (1935). It is true that what Twaddell meant by the term 'phoneme' has nothing to do with what functionalists understand by it. Twaddell did not mention either neutralization or the archiphoneme in his article in question, though in his private correspondence with me in 1981, Twaddell approved of both neutralization and the archiphoneme, and his theory of the micro-phoneme and the macro-phoneme is implicitly based on the commutation test and compatible with operating not with minimal pairs but with what I call multiplets. See Akamatsu (1984) for my reappraisal of Twaddell's theory of the micro-phoneme and the macro-phoneme from the point of view of the commutation test. 7.8
'Maximally differentiative commutative series' and 'non-maximally differentiative commutative series'
When a number of commutative series are being set up for phonological analysis of a given language, the analyst will notice that on the one hand, some commutative series exhibit 'maximum differentiation' in that such commutative series comprise the maximum number of those different phonetic segments, by virtue of which all the words cited therein are crucially distinguished from each other and whose phonological status is
48
Japanese Phonology: A Functional Approach
under analysis, in the respective phonetic contexts with which the individual commutative series are associated. There are consequently no lacunae in such commutative series. On the other hand, other commutative series exhibit ’non-maximal (i.e. 'less than 'maximal') differentiation' due to the presence of a lacuna or lacunae. I shall conveniently refer to the two types of commutative series in question as 'maximally differentiative commutative series’ and 'non-maximally differentiative commutative series', respectively. These are designations of my own coinage which I am using here for the first time. The distinction between the above-mentioned two types of commutative series and their above-mentioned designations are not ascribable to any other functionalists, as far as I know. The phonetic context with which a maximally differentiative commutative series is associated may be termed 'context of maximum differentiation', and that with which a non-maximally differentiative commutative series is associated may be termed 'context of non-maximum differentiation'. The distinction between the two types of phonetic contexts just mentioned and the two relevant designations are again ascribable to me, not to any other functionalist. A maximally differentiative commutative series may, as the case may be, consist of nothing but minimal multiplets, or of both minimal multiplets and near-minimal multiplets, or (at least potentially) of nothing but near-minimal multiplets. The distinction between minimal multiplets and near-minimal multiplets on the one hand and the distinction between maximally differentiative commutative series and nonmaximally differentiative commutative series are independent of each other. A large examples of both maximally differentiative commutative series and non-maximally differentiative commutative series will be encountered in what follows in the present chapter and also in the next two chapters, i.e. Chapters 8 and 9 in particular, but also here and there in Chapters 10 and 11. I should add lastly that the point that will have arisen prior to establishing the difference between maximally differentiative commutative series and non-maximally differentiative commutative series is whether or not it is necessary and sufficient to operate with the minimally necessary number of minimal pairs. We read as follows in Martinet (1956, 3.15): Theoriquement [italics added], il faudrait pour chaque phoneme faire tous les rapprochements necessaires pour montrer qu'il est distinct de tous les autres phonemes de la langue. En fran5 ais, par exemple, il faudrait prouver que [s] est distinct de chacune des consonnes, et meme de chacune des voyelles. En pratique [italics added], il suffit de montrer par la commutation que [s] est distinct des phonemes dont les realisations sont les plus nettement apparentees, c'est-ii-dire en franfais /s/ et /z/ [...]. (’Theoretically, we should have, for each phoneme, to do all the necessary comparisons in order to show that the phoneme is distinct from all the other phonemes of the language in question. In French, for example, we would have to prove that [s] is distinct from each of the consonants, and even from each of the vowels. In practice, it is sufficient to show through commutation that [s] is distinct from the phonemes whose realizations are clearly the closest to [5], that is to say, in French, Is/ and /z/[...] [my translation].')I I am in agreement with Martinet on these points. The theoretical aspects of the commutation test are being demonstrated throughout the present chapter, but also in Chapters 8, 9, 10 and 11 in my presentation of commutative series for my phonological analysis of Japanese in this book. One will no doubt detect there my precautionary reaction to the widespread overemphasis on recourse to minimal pairs. The practical aspects of the commutation test will be illustrated, also in Chapters 8 and 9 in particular, when I establish the relevant feature(s) of the individual vowel phonemes or consonant phonemes, and thereby at the same time establishing these phonemes themselves.
Chapter 7: The commutation test
7.9
49
Implications of non-maximally differentiative commutative series
It should be warned that generally speaking, the analyst will only infrequently obtain a number of maximally differentiative commutative series. It is indeed the case that he can as a general rule obtain non-maximally differentiative commutative series as well. Non-maximally differentiative commutative series are important, as they reflect instances of neutralization of phonological oppositions and/or systematic non occurrence of a phoneme(s), and for this reason they should be accorded just as much importance as maximally differentiative commutative series during the course of the commutation test. Here are a few examples. (1) In establishing commutative series on the basis of data from English, the analyst will meet with non-maximally differentiative commutative series associated with wordinitial context in that there is consistently the lacuna of [g] in that context. The phonological implication of the lacuna of [g] in word-initial context is that it is a reflection of the systematic non-occurrence of /g/ in word-initial context in English; this results from the fact that there is no English word whose signifier begins with /g/. (2) Also, in establishing commutative series on the basis of data from R.P. (= Received Pronunciation) British English, the analyst will meet with non-maximally differentiative commutative series associated with word-final context in that there are consistently the lacunae of [h], [j], [j] and [w] in that context. The phonological implication of the lacunae of [h], [j ], [j] and [w] in word-final context is that it is a reflection of the systematic non-occurrence of /h, hi, /j/ and /w/ in that context. (3) Furthermore, in establishing commutative series associated with the phonetic context 'after word-initial [s] or word-medial [s]' on the basis of data drawn from English, the analyst will meet with non-maximally differentiative commutative series in that there are consistently lacunae of [b]/[b], [d]/[d], [g]/[g], [5]/[§], [z]/[?], [3]/[S] and [d3 ]/[d$], while [p]/[ph], [t]/[th], [k]/[kh], [s] and [tf]/[tfh] are attested in that context. (It will be recalled that the diacritic mark , signifies devoicing (note, that e.g. [b] is not equivalent to [p]) and the diacritic mark h signifiers aspiration.) The fact that certain other consonant segments (i.e. [m], [n], [f], [v], [8], [1], [j],33 [j] and [w]) also occur in the context 'after word-initial or word-medial [s]', while certain other consonant segments (i.e. [g], [h], [s], [z]) do not occur in that context, should be noted as relevant. The phonological implication of the lacunae of the above-mentioned consonant segments is that these are a reflection of the neutralization of /p/ - lb/, It/ - Id/, Ikl - /g/ and /tf/ - /ct$/ in that context34 and the systematic non-occurrence of /g/ and /hi, /s/, /z/, /J/ and III in that context, too. The above-mentioned cases of neutralization of phonological oppositions and those of systematic non-occurrence of a phoneme(s) can only be detected and identified during the course o f the commutation test through careful phonological analysis, not sheerly on the basis of the non-occurrence of the above-mentioned consonant segments. That equal importance should be attached to maximally differentiative commutative series and non-maximally differentiative commutative series during the commutation test cannot be overemphasized. If maximally differentiative commutative series contribute to establishing all the phonemes of a given language, non-maximally differentiative commutative series serve to reveal instances of neutralization of phonological oppositions and/or systematic non-occurrence of a phoneme(s). The two types of commutative series are necessary and need to be compared with each other during the commutation test so that the results of the commutation test may be maximized. It is indeed well to remember that, depending on the language to be phonologically analyzed, there may be no possibility of obtaining any maximally differentiative commutative series (for at least part of the phonological system) for the simple reason
50
Japanese Phonology: A Functional Approach
that the language phonologically functions in such a way as to reflect itself in that way in commutative series available.35 In this case, it is necessary to operate with whatever non-maximally differentiative commutative series are obtainable. This of course does not mean that the commutation test is doomed to failure. 7.10
No multiplet should contain a potential pause
In accepting as multiplets of a given commutative series words of a language under phonological analysis which would constitute minimal or near-minimal multiplets, I wish to mention, in addition to the requirement I have already mentioned in 7.5, three more important requirements that should be observed. In this and the next two sections, I will explain them. The readers are invited to consult a succinct relevant exposition in Martinet (1960, 3.5, 3.6 and 3.7). All the words which are cited as multiplets of a commutative series should contain no potential pauses in them. For this reason, a word like lamppost (generally known as a compound word in English grammar) which contains a potential pause between lamp and post is unacceptable as a multiplet. For the same reason a word like linked (known as an inflected word) contains a potential pause between link and ed and is also unacceptable as a multiplet. On the other hand, a word like spoke is acceptable as a multiplet, as there is no potential pause in this word. It follows that e.g. pact is acceptable, but not packed, as a multiplet. The readers will have noticed that I employ the term ’word' above and hereafter in connection with the commutation test. I deliberately use this term as popularly understood. The use of the term 'moneme' would create complexity which, for practical reasons, I prefer to avoid so far as the task of establishing commutative series is concerned. If a word like pack is mono-monematic, words like lamppost, linked and packed are bi-monematic and consist of two monemes successively conjoined, while a word like spoke is also bi-monematic but consist of two monemes which are fused (the term 'amalgam' is given to such a fused form). All this would take us to what functionalists call 'synthematics' ('synthematique' (F)), which I prefer not to go into in this book. The use of the term 'word' seems quite appropriate and well-advised in performing the commutation test in Japanese too, as will be seen in Part II. But to return to the subject matter of this section, I quote the following passages in connection with the point that multiplets should not be contain potential pauses. II convient done de pratiquer l’analyse a partir de segments de l’enonce non susceptibles d'etre interrompus par une pause [be it actual or potential]... ('It is therefore appropriate to perform the [phonological] analysis on the basis of segments of utterances which are not susceptible of being interrupted by a pause [be it actual or potential]... [my translation].') (Martinet 1960, 3.5) II se trouve que, dans certaines langues, le comportement phonologique particulier que l'on rencontre generalement devant une pause virtuelle, se retrouve, de fa§on plus ou moins marquee, a l’interieur de ce qu'on appelle les mots, a la frontiere de deux monemes... ('It so happens that in certain languages, the phonological behaviour which we encounter before a potential pause also occurs, in a more or less obvious manner, inside of what are called words, at the boundary between two monemes... [my translation].') (Martinet 1960, 3.6) On utilisera done, pour l'analyse en phonemes, des segments de l'enonce dont on est stir qu'ils ne renferment pas de pauses virtuelles ('We should therefore use, for an analysis into phonemes, segments of utterances which we are certain do not contain potential pauses [my translation].') (Martinet 1960, 3.7)
Chapter 7: The commutation test
51
Martinet gives a few examples drawn from a few languages. I shall choose the wellknown examples from English, some of which are also cited by Martinet, to illustrate the point. The triplet night-rate, nitrate and Nye-trait36 have an identical sequence of phonemes, and the same accentual pattern. Yet, phonetically, the three words are different from each other in a few respects, in that [ai] is longest in Nye of Nye-trait (i.e. [an]), less so in nitrate (i.e. [an]) and still less so in night-rate (i.e. [ai]); the onset of the second syllable is situated before [U] in nitrate and Nye-trait but between [t] and [j ] in night-rate', [t] is fully aspirated in Nye-trait, less so in nitrate and very little in nightrate', and [j ] is devoiced in Nye-trait and nitrate but voiced in night-rate. Admission of such triplets as night-rate, nitrate and Nye-trait, two of which (i.e. night-rate and Nyetrait) containing potential pauses, unnecessarily complicates, indeed obstructs, phonological analysis, as the analyst would find it difficult to conclude in what specific phonological way the three words are crucially distinguished from each other. Is he to postulate three different vowel phonemes associated with three different lengths of [ai], i.e. [a:i], [a*i] and [ai]?; three different consonant phonemes associated with three different degrees of aspiration of [t] i.e. [th], [th] and [t]?; and two different consonant phonemes associated with [ j ] (voiced) and [ j ] (devoiced)? And which single phonological factor should we consider as crucially responsible for differentiating the three words from each other? An impossible question to answer. Actually the few phonetic variations mentioned above result ultimately from the fact that potential pauses intervene in night-rate (i.e. between night and rate) and Nye-trait (i.e. between Nye and trait) but not in nitrate. Other appropriate examples - adduced by Bloch et al. (1942, p. 47) and others - are minus and slyness, in which [ai] in minus (which contains no potential pause) is shorter than that in slyness (which contains a potential pause between sly and ness). Also syntax and tin-tax - adduced also by Bloch et al. {op. cit., loc. cit.) - in which [n] is longer in tin-tax (with a potential pause between tin and tax) than it is in syntax (which contains no potential pause). Hall (1964, p. 112) adduces sin-tax instead of tin-tax. The advisability to avoid words which contain potential pauses in establishing commutative series applies to these pairs of examples, too. Incidentally, some readers who are familiar with post-Bloomfieldian linguistics will have seen that what is referred to as 'potential pause' corresponds to 'internal open juncture' (or informally called 'plus juncture1) in the terminologies employed by postBloomfieldian linguists. However, the difference is not merely terminological. An internal open juncture is a phoneme (/+/) to post-Bloomfieldian linguists, while a potential pause is never a phoneme to functionalists. The direct opposition between nitrate and night-rate, as /naytreyt/ vs. /nayt+reyt/, as Trager et al. (1957, p. 38) advocate (the phonological notation is theirs), i.e. as 'minimal pairs' which are said to be distinguished through the phonological opposition between the absence and presence of /+/, is disallowed in functional phonology. 7.11
All multiplets of a given commutative series should share an identical suprasegmental context
Another important requirement is that all the linguistic forms cited in a given commutative series as multiplets should have 'in principle' an identical suprasegmental (or prosodic) context, not merely an identical or semi-identical segmental context except for the phonetic segments whose phonological status is to be phonologically identified. As Martinet (1960, 3.7) himself puts it,I II faudra.. ,ne rapprocher que des segments dont on est sur qu'ils component les memes traits prosodiques. ('We should...compare only those segments which we are sure bear the same prosodic [i.e. suprasegmental] feature [my translation].')
52
Japanese Phonology: A Functional Approach
We shall take a look at, for exemple, the commutative series consisting of dapper, rubber, utter, udder, succour, saga, summer, Anna, hangar, suffer, cover, other, tussore, Tozer, usher, azure, duchy, leger, colour, borough, Moho, hiya and kiwi. The segmental elements whose phonological status we wish to establish are those which occur, at least ideally, in the phonetic context [ a - a], i.e. at the point shown by the dash. This is so at least in principle. That is to say, ['a - a] applies to all the minimal multiplets in this commutative series (i.e. utter, udder, other, usher) and some of the near-minimal multiples (i.e. rubber, succour, summer, suffer, cover, tussore, colour, borough). The remaining near-minimal multiplets (dapper, Anna, hangar, Tozer, azure, duchy, leger, Moho, hiya, kiwi) are associated with such phonetic contexts which are, segmenially, variably and unproblematically different from ['a - a], but suprasegm entally (prosodically) identical with [ a - a ] , and are therefore unquestionably acceptable multiplets of this commutative series. If I said 'in principle' at the beginning of this section, this is because in specific cases, it may turn out in the course of the commutation test on data drawn from a language that the occurrence of the segmental elements whose phonological status the analyst intends to establish through the commutation test turns out, as the commutation test progresses, not to be influenced one way or another by different suprasegmental contexts. We shall see later that such is the case with the occurrences of segmental elements, be they vocalic or consonantal, in Japanese even in different accentual patterns and that this means that different suprasegmental contexts have no repercussion on the identification of the vowel phonemes or consonant phonemes and on their occurrences in Japanese. Nevertheless, I will see to it that, in principle and as far as possible, the above-mentioned requirement about the sameness in suprasegmental contexts is observed in my phonological analysis of Japanese to be presented in Part I. We shall find that after all, this requirement is little necessary in the case of Japanese. Be that as it may, it is highly desirable as a matter of general principle to observe as far as possible the requirement of 'identical suprasegmental context' for the phonetic context associated with commutative series. Depending on the language under analysis, flouting the requirement of 'identical suprasegmental context' may well result in erroneous phonological analysis. In a phonological analysis of Russian, for example, disregard of the said requirement would lead to establishing a single system of vowel phonemes of Russian rather than two sub-systems: one sub-system consisting of five vowel phonemes /i e a o u/ in accented syllables (or in strong positions, as Slavicists call them) and the other sub-system consisting of three vowel phonemes A a u/ in unaccented syllables (or in weak positions, as Slavicists call them). The establishment of the two sub-systems is correctly recognized by Martinet (1968a, p. 7) and is concurred with by me (cf. Akamatsu 1988, p. 355). In the case of a tone language, the requirement that the suprasegmental context should be identical for all the multiplets of a given commutative series concerns the tone (a phonetic, not a phonological, entity) and is mandatory. In an attempt to elicit and identify the phonemes of e.g. Mandarin Chinese, it is obligatory that a given commutative series associated with the phonetic context [-a] (i.e. followed by the vowel sound [a] bearing one of the four tonemes this Chinese dialect) should consist of words all of which are characterized by [ '], so that a commutative series consists of [pha] ’(to) fear1, [pa] '(to) stop', [tha] '(to) strike', [ta] 'big', [m&] '(to) curse', [na] '(to) stutter', etc. but does not include as well some words that have different tones, e.g. [p^a] '(to) climb up' (instead of [p^a] '(to) fear'), and [ma ] 'hemp' (instead of [ma] '(to) curse’), and so on, in one and the same commutative series.
Chapter 7: The commutation test
7.12
53
No multiplets should contain phonic manifestation associated with the expressive function
This is the last requirement I wish to mention. I have already explained what is meant by the expressive function in 2.5 and given a couple of examples of how this function is phonetically manifested in English. I said there that the word enormous in the phrase an enormous tree may well be pronounced with [-n:::-] and/or [-o:::-] by a speaker who is much impressed by the enormity of the tree and expresses his relevant state of mind in this particular phonetic manner. In establishing a commutative series in a phonological analysis of English, however, the word enorm ous as pronounced in the afore-mentioned way should not be incorporated as a multiplet, i.e. as pronounced in such a way that the phonic substance by which the word is manifested is associated with the expressive function in addition to the distinctive function (or any other function(s) involved). If the word enormous is to be included as a multiplet in a commutative series, it should be with the phonetic form which is devoid of association with the expressive function. The pronunciation of enormous as [i’no:mas], i.e. a neutral manner of pronouncing the word, is the one that is acceptable if the word is to be included in a commutative series. Indeed [i'no:mas] corresponds to a pronunciations of this word as entered in dictionaries (for Received Pronunciation of British English). The requirement which has been explained above is necessary to prevent the analyst from including as separate multiplets, say [I'noimas] and [Tnonmas], and establishing two separate vowel phonemes /a:/ and /o::::/ by claiming that the commutation between [o:] and [o::::] results in separate semantic contents. Similarly, the same requirement will prevent the analyst from establishing /p/ and /p:::/ in French, who might otherwise include as separate multiplets [sposibl] and [£p:::o sibl], the first being the neutral pronunciation of impossible and the second a pronunciation of the same word as in the utterance Cet enfant est imPOSsible! 'This child is imPOSsible!' said with considerable irritation (cf. Martinet 1960, 3.7). The elicitation of the phonemes in the course of the commutation test is crucially and exclusively concerned with the distinctive function fulfilled by the phonemes, and it is the neutral phonetic form of the multiplets of commutative series that leads to the identification of the phonemes. Consideration of phonic form induced by the expressive function has no place in the commutation test. The requirement which has been explained above has an important implication in that commutation between phonic elements by means of multiplets, if successful, should result in destruction of the identity of a multiplet and its total replacement by the identity of another multiplet, or, to rephrase it in popular parlance, in a change of one word by another. Such is not the case with the commutation between [i'no:mas] and [i'no::mas], or between [§po sibl] and [ep:::osibl], since no change of one word by another occurs in either case, and the commutation is therefore unsuccessful, which means that one does not elicit two separate phonemes in either case. 7.13
A few examples of commutative series
As will have been seen above, in order to avoid expository complexity, my remarks made on the commutation test has concerned themselves with what may be called 'segmental phonology'. I leave untouched what may be called 'suprasegmental phonology'. Based on the principles sketched above in this chapter, the analyst will establish a number of commutative series associated with different phonetic contexts for the purpose of performing the commutation test. Suppose the analyst wishes to elicit and define the consonant phonemes of English, he may establish commutative series such as
Japanese Phonology: A Functional Approach
54
the following and perform the commutation test on them. Commutative series 1 [Pi - 1 pip fib pit rid pick Pig Pym pin ping d ff sieve pith with piss fizz pish beige pitch ridge pill — — — —
Commutative series 2 [-.t] p it bit tit dit kit git mitt knit —
fit vit thick this sit zit shit gite chit je t lit writ hit yet wit
Commutative series 3 ['a - a ] Qd)apper (r)ubber (of whist) utter udder (,s)uccour (s)aga (s)ummer Anna (.h)angar (s)uffer (c)over Arthur other (t)ussore (T)ozer usher azure (d)uchy {l)edger (c)olour (b)orough (M)oho (h)iya (k)iwi
I attach a few explanatory remarks about the examples of the commutative series given above. (1) As already explained further above, all the multiplets of the commutative series, though expediently presented in their orthographic form, should be understood in terms of their phonetic forms, so that knit should be understood in terms of [nit] and writ in terms of [jit]. (2) In order to show to the readers the more easily what are those segmental elements whose phonological status the analyst intends to establish, I have deliberately presented in boldface the letters in the orthographic form of the multiplets that roughly correspond to the segmental elements in question. Thus, for example, thick means that [9] that orthographically corresponds to th is the segmental element in this multiplet that the analyst is concerned with for the purpose of the commutation test. (3) A good number of near-minimal multiplets are cited in the commutative series. They can and should be presented in their full (orthographic and phonetic) forms. However, just to make it easier for the readers to see the phonetic context with which each commutative series is associated, I have deliberately enclosed within parentheses those letters the exclusion of which will, albeit artificially, present the near-minimal multiplets closer in appearance to the corresponding minimal multiplets. For example, (c)olour is cited in Commutative series 3 for want of an existing English word which might be pronounced ['a la]. However, a case like (d)uchy (and there are others) in which the lettery in this case corresponds to [i], not [a], makes it clear that such use of parentheses does not necessarily mean that the segmental element [i] (instead of [a ]) which makes the cited example a near-minimal multiplet is obligatorily placed within
Chapter 7: The commutation test
55
parentheses. The artificial device I have resorted to should therefore be taken as no more than an imperfect expedient. (4) Each commutative series is associated with a given phonetic context which differs from the phonetic contexts associated with the other commutative series. Commutative series 1 is associated with [pi -], Commutative series 2 with [- it], and Commutative series 3 with ['a - a]. The dashes in such phonetic contexts indicate where the segmental elements whose phonological status the analyst is interested to establish occur. Each such phonetic context is intact in minimal multiplets (cf. pip, pit, pick, etc. in Commutative series 1) but is partially modified (cf. fib, rid, tiff, etc. also in Commutative series 1) or even entirely modified (cf. with and beige also in Commutative series 1) in near-minimal multiplets. Likewise in connection with Commutative series 2 and 3. Such modifications of the phonetic context with which each commutative series is associated are perfectly valid provided that their acceptability is justified on phonetic grounds - as it is in all three commutative series above - and all near-minimal multiplets themselves are for this reason as acceptable as any minimal multiplets (cf. 7.5). (5) We see that each of the three commutative series presented above is a mixture of minimal multiplets and near-minimal multiplets. To take Commutative series 2 as an example, the minimal multiplets are pit, bit, tit, dit, kit, git, mitt, knit, fit, vit, sit, zit, shit, gite, chit, lit, writ, hit and wit, while the near-minimal multiplets are thick, this, jet and yet. The ratio of minimal multiplets to near-minimal multiples within a commutative series varies from one commutative series to another, as comparison between Commutative series 2 and the two other commutative series easily reveals. The relative difficulty to obtain minimal multiplets (with the obvious consequence of incorporating a relatively large number of near-minimal multiplets) depends entirely on the specific phonetic context the analyst has chosen as being associated with the commutative series. This, however, is no reason why a commutative series like Commutative series 3 above which consists of a large number of near-minimal multiplets and a tiny number of minimal multiplets should be fought shy of or be considered less desirable for the purpose of the commutation test. What is important is that Commutative series 3 happens to be a maximally differentiative commutative series which is maximally conducive to eliciting all the consonant phonemes of English. It is of course desirable if the analyst can obtain yet another commutative series which is also a maximally differentiative commutative series, but he may or may not be lucky to. Commutative series 1 which we have already seen as consisting of a large number of minimal multiplets and an insignificant number of near-minimal multiplets happens to be a nonmaximally differentiative commutative series in that there are lacunae. A commutative series like Commutative series 1 is necessary as, on the one hand, it is conducive to eliciting a major portion of the English consonant phonemes and on the other hand, it reveals the instances of systematic non-occurrence of some English consonant phonemes (i.e. /r/, /h/, /j/ and /w/). The only lacuna observed in Commutative series 2 reveals an instance of systematic non-occurrence of a phoneme (i.e. /g/). (6 ) It will have been seen that the multiplets in the commutative series are arranged in a particular chosen parallel serial order in all three commutative series, with regard to the various segmental elements by virtue of which the multiplets are minimally or nearminimally differentiated from each other. In all three commutative series presented above, the sequential order (from top to bottom) is [p], [b], [t], [d], [k], [g], [m], [n], [rj], [f], [v], [0], [8 ], [s], [z], [f], [3 ], [tj], [d3 ], [1], [j], [h], [j] and [w] in this order. Any lacunae, which are the only factor that disturbs the completeness of the parallel have been indicated by long dashes meaning vacancies.I I omit details of how the analyst will then perform the commutation test on the basis of Commutative series 1, 2 and 3 with a view to eliciting and defining the English consonant phonemes in terms of relevant features (we shall assume that the analyst will
56
Japanese Phonology: A Functional Approach
have achieved this first task), as this would take me out of the confines of the present book on a phonology of Japanese and is better reserved in a separate book devoted to the phonological analysis of English from a functional point of view. My presentation of a phonological analysis of Japanese from a functional point of view to be presented in Part II will amply show how to perform the commutation test and the results obtained therefrom. Interested readers may also wish to take a look at Akamatsu (1992b, pp. 6580) in which I have shown how to perform the commutation test on material drawn from French and English. Apart from eliciting and defining the consonant phonemes of English in terms of relevant features on the basis of Commutative series 1, 2 and 3, the analyst will have found that Id, /h/, 1)1 and /w/ are non-occurrent in the context /pi - / (cf. Commutative series 1) and /r]/ is non-occurrent in the context /- it/ (cf. Commutative series 2). Further analysis with more commutative series will have shown to the analyst that in fact, Id, /h/, 1)1 and /w/ are systematically non-occurrent in word-final context in general, and that /g/ is systematically non-occurrent in word-initial context in English. But now, I need to present below one more commutative series which is associated with the phonetic context [(#)s - (#)] and which we shall call Commutative series 4. The symbol # signifies a pause, so that [(#)s - (#)] is to be interpreted as meaning that [s] optionally follows a pause and the segmental element under phonological analysis is optionally followed by a pause. The parentheses signifies 'optionality'. Commutative series 4 contains a number of lacunae which are indicated, not by dashes this time, but by blanks. The reason for my specifically presenting Commutative series 4 is that the lacunae are linked to neutralization of phonological oppositions in some cases, and to systematic non-occurrence of phonemes (cf. supra Commutative series 1 and 2) in some other cases. In presenting Commutative series 4, I also need to reproduce by its side Commutative series 3 (with a tiny modification in its format), so that the necessary comparison between the two commutative series is possible.
/p/ /p-b/ Ibl It! /t-d/ /d1 IkJ /k-g/ ____! s l _
Commutative series 3 ['a - a] upper
Commutative series 4 [(#)s - (#)] spit, Caspar, rasp
rubber utter stick, caste, mast udder succour skit, Oscar, husk saga
/ml Ini /n/ Ifl N!
summer Anna hangar suffer cover
/el ie-5i 161
Arthur
smock, asthma, buxom snip, listener, mason sphere, satisfy svelte aesthete, sixth
other
57
Chapter 7: The commutation test
Is/ ItJ IP IP
tussore Tozer usher azure
/If/ Af-# W
duchy ledger
hi /li/ V /w/
borough Moho hiya kiwi
question Srinagar Oxhide suet swan
It is assumed that the analyst has elicited and defined all the consonant phonemes of English on the basis of Commutative series 1, 2 and 3. This is why I have put the consonant phonemes on the leftmost column (/p/, /b/, /t/, Idl, etc.) The readers will see that the opposition between /p/ and Ibl is germane to Commutative series 3 but not to Commutative series 4 and that the archiphonemes /p-b/, /t-d/, /k-g/, /9-S/ and /tf-d$/, also placed in the leftmost column, are germane to Commutative series 4 but not to Commutative series 3, as my placement of the phonemes and the archiphonemes shows. What the comparison between Commutative series 3 and Commutative series 4 reveals are the following two separate phonological facts. First, for example, /p/ - Ibl (cf. upper vs. rubber) is valid in the context /'a - a/ (this is the phonological notation corresponding to [‘a - a]), i.e. in the context of relevance for /p/ - Ibl, but is neutralized (cf. spit, Caspar, rasp) in the context /(# )s - (#)/. Note specifically that in the phonological notation /(#)s - (#)/, #, by which I mean a pause, is not a distinctive unit; (#) means that a pause is optional.37 Likewise /t/ - Id/, Ik/ - Igl, /©/ -161 and / f /-/e] n n ’e] *foe] |'uier e",baa] 'Weber'53 *[Nel Commutative series 10 r-m r i [p’uir (tutto)] 'whizzingly'
Commutative series 9 [-o r] [p’or (ko'1 N)] 'jerkily (in bowing)' [por (kkar’i)] ’(surfacing) suddenly' [b’or (odoo)] 'equality'
[prq r (tsra "lN)j '((to) be cut) 'suddenly and neatly' [b’uir (ujkeN)] 'fallacy'
[bor (o)] 'rod' *[t’o] [tor (bakup] 'gambling' *[d’o] [dor (bta)] 'ditch' [k’or (obai)] 'auction' [kor (oeN)] 'park' [g’or ($mN)] 'fish-meal' [gor (ztu1 ui)] 'fifty'
[btnr (ai)] 'percentage' *ft’uil *[tin] [d’uir (e",tto)l56 'duet' *[dtn] [k’uir (ui fi>i)] 'suddenly' [ktn r (tsiu"')] 'shoe' [g’uir (ratoo)] 'cowpox' [gm r (cqi)] 'gmmbling'
[$ar ( V $a",are)] 'Fanfare (G.)' [gar (kra'1)] ' 10 0 '
[$or (ruimar’iN)] 'Formalin (G.)' (?or (obaN)] 'reputation'
[cJ»xnr (jTxi’’)] 'winter'
[jar (baN)] 'savageness' [qar (qiN)] 'photograph' [izar (ma)] 'hindrance' f?ar (ma)] 'hindrance' rsar (to"lo)l 'sugar' fzar (qci)l 'magazine' [dzar (qqi)] 'magazine' [tsar (ar’ Pzm mui)] 'Czarism' [cqar (ujaN)] 'rice bowl' [r,ar (kniztD)] 'sketch map' [rar (kuj)] 'comfort'
[jor (boo)] 'prevention' [cor (kcu)] 'employment' [jZOr (btiiN)] 'preface' [zor (buiN)] 'preface' [sor (me)] 'dyeing' [zor (oi)] 'bequest' [dzor (oi)] 'bequest' *[tso]
Commutative series 8 [-a r] *[p’a] [par (ase"lnto)] 'percent' [b’ar (kuj"lreN)] 'white lo tus' [bar (qo)] 'place' *[t’a] [tar (dare)] 'festering' *[d’a] [dar (be"Yuj)] '(to) chatter' [k'ar (kui)] 'guest' [kar (ba)] 'birch' [g’ar (kin)] 'inverse' [gar (kni)] 'tablet'
[cqor (oba)] 'counter' [r’or (o$uiui)] 'cool breeze' [ror (odoo)] 'labour'
[?rur (tuma t>i"lstq to)] 'humanist' [jtDr (iuai)] 'friendship' [cujr (cpoo)] 'insistence' [puir (N)] 'order' [?inr (N)] 'order' [srnr (iee)] 'swimming' [zoir (bo",N)] 'trousers' [dzra r (bo”’N)] 'trousers' [tsm r (bak(i) ^ )] 'spittle' [cqtur (uiqa)] 'parking' [r’uir (uiboktxj:)] 'driftwood' [nnr (i?i)] 'resemblance'
Chapter 9: Identifying the consonant phonemes of Japanese
n i ’al54 [lar (kui)l 'comfort' [har (na)] 'nose' [m’ar (kui)] 'pulse'
*ri’oi 54 rior (odoo)] 'labour' [hor (ei)] 'star'
[mar (cci)] 'town' *fn’al 55 [nar (ta)] 'hatchet' [i>ar (a"lgo)] 'miaow!'
fmor (cci)])] 'rice cake' *[n’ol5S fnor (okoo)] 'farming' [t>or (odoo)] 'urethra'
*[n’a] *[na] [ujar (ci)] 'eagle'
*[n’o] *[no] [rqo r (otaacui "’ll! to)] 'water chute' *[Nol
*[Na]
[m’or (obaN)] 'alum'
79
*ri’uil54 [luir (i?i)] 'resemblance' *[hui] [m’iur (uizikkuiho',oriii)] 'music hall' [muir (sai)l 'celibacy' *[n’ui] 55 [nui r (ime)] 'seam' [T>uir (ui)iN] 'hospitaliza tion' *[n’ui] *[gm] *[upar ] *[Nui]
Out of a number of commutative series that can be established for the purpose of eliciting and identifying the consonant phonemes of Japanese, I have chosen to present just ten above, which divide into two types (Commutative series 1 to 5 on the one hand and Commutative series 6 to 10 on the other) which are associated with two different accentual patterns. The following remarks are in order in connection with certain aspects of my presentation above of the multiplets in Commutative series 1 to 10 as a whole. (1) In order for the readers to see easily the directly relevant parts of the multiplets for the purpose of the commutation test, I have deliberately put, in presenting the multiplets, pairs of parentheses which enclose those parts of the multiplets which are outside the directly relevant parts (e.g. [p’i",(iktn)] 'summit' in Commutative series 1). (2 ) I already explained in 8 .2 the meaning of the symbols r and n (used either singly or in combination) in my presentation of multiplets of commutative series. The use of the symbol ^ (as in [kT"1] 'tree' cited in Commutative series 1) or r (as in [k’ir ] 'yellow (colour)' cited in Commutative series 6 ) is necessary to indicate that for example, in [k’i"lkara] 'from the tree', [k’i] is pronounced on a pitch higher than is [kara], while in [k’ir kara] 'from the yellow (colour)', [k’i] is pronounced on a pitch lower than is [kara]. It is of course inappropriate to cite [k’ir ] 'yellow (colour)' in Commutative series 1 and [k’i'1] 'tree' in Commutative series 6 , as the phonetic contexts associated with Commutative series 1 and 6 , respectively, would not be respected. (3) A general look at the examples in Commutative series 1 to 10 will show that I have tried to choose - with a large measure of success - such multiplets in which we have [- i"1], [- e"1], [- a"1], [- o’1] and [- in "'] on the one hand (for Commutative series 1 to 5) and [- ir ], [- er ], [- ar ], [- or ] and [- uir ] on the other (for Commutative series 6 to 10 ), for the sake of the uniformity in the accentual context for the commutative series presented. Actually, in Japanese, differences in the accentual context have no repercussion on the occurrence or non-occurrence of the consonant segments in [- i], [- e], [- a], [- o] and [- ui ] whose phonological status I am to establish on the basis of Commutative series 1 to 10. As already said earlier, any analyst of Japanese phonology will sooner or later discover during the course of the commutation test that the occurrence or non-occurrence of phonetic segments, be they vocalic or consonantal, is in no way either determined or influenced by different accentual patterns. One case in point has emerged in what has been presented as [d’ui "*(etto)] 'duet' in Commutative series 5 but as [d’rar (e",tto)] 'duet' in Commutative series 10. Remarks about this are
Japanese Phonology: A Functional Approach
80
made in notes 51 and 56. The fact that this happens to be an English loanword is irrelevant. A large number of Japanese words have alternative accentual patterns, e.g. [jzuir Vjo"lokaN]/[jzujr VjookaN] 'cruiser', [dar inoo]/ [da inoo] 'cerebrum', [her ndeVpo]/[her nde Vqo"1] 'transformer substation', [kamr pui",ui]/[kamr puiuj] 'cold wind'. What is significant is that the phonetic segments are identical in spite of different accentual patterns in all such Japanese words. (4) [t’] and [d’] should not be confused with [tj] and [dj], respectively, which occur in English words like tube and dew in the pronunciation of those British/American speakers who pronounce [tju:] and [dju:] (and not [tu:] and[du:]). [t’] and [d’] are single consonant segment (i.e. [t] with concomitant palatalization, and [d] with concomitant palatalization), while [tj] and [dj] are sequences of two phonetic segments, [t] + [j], and [d] + [j]. As is seen in Commutative series 1 to 5, [t’] and [d’] in Japanese occur only before [i] and [uj ] (hence *[t’e], *[t’a], *[t’ o], *[d’e], *[d’a], *[d’ o] as noted in Commutative series 2 to 4, and 7 to 9), and only in the pronunciation of loanwords. Exceptionally, asterisks are placed against [t’ru] and [d’ui] (or more exactly [t’uir ] and [d’uir ]) in Commutative series 10. This is because no loanwords in Japanese seem to be pronounced with [t’rur ] or [d’u ir ]- On the other hand, [t’ui"1] and [d’ui"1] are listed without asterisks in Commutative series 5, as they also do occur in loanwords in Japanese. In addition to [t’i‘lmpaT>i] 'timpani', [d’i'1 (sixiko)] 'disco' (both cited in Commutative series 1), [t’ui'luiba] ’tuba' and [d’ui "'etto] 'duet' (both cited in Commutative series 5), [t’ir iqatsui] 'T shirt' and [d’ir re"lttanto)] ’dilettante' (both cited in Commutative series 6 ), and [d’uir e"'tto] 'duet' (cited in Commutative series 10), here are some more relevant examples randomly chosen: [t’i r isuipui"luiN] 'tea-spoon', [p’i r it’iie",e] 'P.T.A.', [fin",uidaa] 'Tudor', [t’ui "Visur] 'Tunis', [f ui r uitoN] 'Teuton', [qir id,i"li] 'C.D.', [d’ir id’iit’i",i] 'D .D .T.', [ d ’ui ",ma] 'Dum as', [pin r rod’ui",uisaa] 'producer', [d’uir sseruido",rui(])xu] 'Diisseldorf. Note that even in standard Japanese pronunciation, not all such loanwords are pronounced by all individuals in the way indicated above and variant pronunciations with other consonant segments than [t’] and [d’ ] are currently heard; for example, timpani [cpi",mpai>i] (instead of [t’i'lmpar>i]) 'timpani', [cqui",uiba] (instead of [t’ui",mba]) 'tuba', [cqui",T>istu] (instead of [t’ui",T>isur]) 'Tunis', [cqui",uidaa] (instead of [t’ui "’uidaa]) 'Tudor', [cqui r uitoN] (instead of [t’ui r uitoN]) 'Teuton', [pui r rojzin",uisaa] (instead of [pui r rod’ui',uisaa]) 'producer', and [jzuir sseruido",rui$ur] (instead of [d’ui r ssenndo"Yui$ut]) 'Diisseldorf. These variant pronunciations are more in keeping with the more traditional Japanese phonetic practice. However, for example, tea-spoon, T-shirt, P.T.A., C.D., D.D.T, disco and duet are always pronounced in the way I have indicated. I have provided some relevant information about the occurrence of [t’] and [d’ ] before [i] and [in] in Japanese in Akamatsu (1997a, pp. 80-82). I wish to note finally the interesting case of [de^zitarm], the general rendition in Japanese of digital (in connection with audiovisual methods and products), rather than [jzinzitarui] which might be expected prima facie, apart from [d’inzitarui]. The general adoption of [de "'zitarui] can be considered as resulting ultimately from a phonetic phenomenon called dissimilation whereby a contiguous succession of an identical (or semi-identical) phonetic segment is avoided and consequently replaced by a different phonetic segment. In the present example, [jzi] is replaced by [de] in order that the sequence [j^i^zi] may be avoided. Once [d] has replaced [jz], [i] is substituted by [e] which is the nearest, front, vowel segment in Japanese that can occur after [d], ([d] cannot be followed by [i] in Japanese.) It goes without saying that if [d’] is selected, [i] is retained. The rendition [d’ Pzitarui] would be /V /
/N /
Chapter 9: Identifying the consonant phonemes of Japanese
81
somewhat closer to the pronunciation [’didjitfa )1] in English, the source language in this case. (5) In postpausal context, [dz] is known to be general, though [z] occurs in the speech of some Japanese. In intervocalic context, it is [z] rather than [dz] that is more common. For this reason I have listed both [dz] and [z], (6 ) [j] is included as occurrent in Commutative series 3 to 5, and 8 to 10, but as nonoccurrent in Commutative series 1 and 2, and 6 and 7. [uj] is included as occurrent in Commutative series 3 and 8 only, and is included as non-occurrent in Commutative series 1, 2, 4, 5, 6 , 7, 9 and 10. The phonological status of [j] in Japanese and [uj] in Japanese requires some discussion which will be provided in 9.8.8 and 9.8.9, respectively. (7) Asterisked items, namely *[pi], *[bi], *[ti], *[di], *[ki], *[gi], *[ji], *[si], *[zi], *[dzi], *[tm], *[tsi], *[ri], *[li], *[hi], *[mi], *[n’i], *[ni], *[gi], *[Ni] (see note 45), *[uji] (but see (8 ) below) *[p’e], *[b’ e], *[t’e], *[d’e], *[k’e], *[g’e], *[?e], *[je], *[r’e], *[l’e], *[m’e], *[n’e], *[n»e], *[g’e], *[Ne], *[t’a], *[d’a], *[n’ a], *[Na], *[t’o], *[d’o], *[tso], *[n’o], *[ujo] (but see (8 ) below), *[No], *[tm], *[dm], *[hm], *[n*m], *[Nui] and*[ujm] are non-occurrent sequences of phonetic segments in Japanese. It will have been noticed that no pitch pattern is indicated for any one of the asterisked items (neither *[pir ] nor *[pi"1], for example). A pitch pattern is irrelevant to a non-occurrent form. (8 ) The prevalent trend among Japanese speakers appears to be to use [in ir ] rather than [uj ir ], and [in r o] rather than [uj or ] in the rendition of loanwords, as in [mr iikme"lndo] 'weekend' and [ui r ootaaqm’,into] 'water chute'. The less prevalent trend appears to be to use [ujir imkme'1 ndo] and [ujor otaacm",mto], if not [ujir iinkmendo] or [ujor otaaqmmto] for these example words. This does not mean that [141 "’] or [ujo"1] does not occur in the rendition of loanwords; cf. [uji",tto] 'wit' cited in Commutative series 1 and [ujo^toka] 'vodka' cited in Commutative series 4. I should add in connection with [ujo"'toka] 'vodka' (< Russ. B oaxa) that the normal rendition is [m r o"'kka] and [ujo",toka] appears to be rare. I should further add that anyhow [ujo"1] (note, not [ujor ]) rarely, if ever, occurs in loanwords, still less in non-loandwords, in standard Japanese. With regard to [uje] for which I have cited [uje^ebui] in Commutative series 2 and [ujer e’1baa] in Commutative series 7 , 1 should say that the prevalent trend is to use [uie] rather than [uje], hence more commonly e.g. [m r e"lebm] 'wave' than [uje "’ebui}, [mr e",$aasui]/[mpe$a",asm:] 'wafers' than [uje",ij)aasm:]. Some loanwords are rendered with only [me], without the alternative with [uje], as in [m r e",etaa] 'waiter' (but not [ujer etaa] or [uj e"'etaa]), [m r eetor’i"’$rnt’iggm] 'weight lifting (but not [ujepetor, i",$u}t’iggra] or [uj e^etor’i$ujt’iggm]), [m r e"’etoresiu] 'waitress' (but not [ujer etoresm] or [uje^etoresiu]) and a number of other example loanwords.Whether [uje] may in future join the rank of [uji] and [ujo], depends on whether a knowledge of the pronunciation of loanwords in the donor languages (nowadays principally English) on the part of Japanese speakers will increasingly induce them to render the loanwords with [uji] (rather than [mi]) and [ujo] (rather than [mo]), and even [uje] (rather than [me]). This is an aspect of dynamic synchrony in Japanese phonology. Commutative series 1 to 10 which I have presented above represent, then, the prevalent usage concerning [uj] before [i], [e] and [o] at the present time, and may be subject to minor potential modification if there occurs relevant change in Japanese pronunciation. As for the implication about the prevalent and less prevalent trends in the use of [uj] before [i], [e] and [o] at present, see Mekai Nihongo Akusefito Jiten (1981, p. 22). What has been said above is based on what this accent dictionary records. As for [ujm], the asterisk I have placed against it (cf. Commutative series 5 and 10) is valid without reservation, [uj a] is altogether unproblematic, for it is attested (cf.
82
Japanese Phonology: A Functional Approach
Commutative series 3 where [uja"1] is found with a relevant multiplet cited, and Commutative series 8 where [ujar ] is found with a relevant multiplet cited. The occurrence, albeit less prevalent, of [me], [mi] and [tq o] in Japanese has repercussions on parts of Japanese phonology, as I will show later, towards the end of 9.8.9, 9.2
Remarks on Commutative series 1 to 5
(1) The phonetic contexts with which Commutative series 1 to 5 are associated are such that the sequence of a given consonant segment and a given vowel segment (e.g. [pi7], [pe7], [pa"1], [po7], [pm7]) is pronounced on a pitch higher than the phonetic segment or the sequence of phonetic segments, if any, that follow it, so that for example, [pe7 r’ii] is so pronounced that [pe] is on a pitch higher than that on which [r’ ii] is pronounced. The pitch pattern just mentioned does not apply, however, to a small number of multiplets (e.g. [(ha)r g’i7 (pir’i)] 'gnash' in Commutative series 1) which have been cited in the absence of other, suitable, multiplets showing the pitch pattern in question. (2 ) I have not specified, in indicating the phonetic context with which each of Commutative series 1 to 5 is associated, whether the sequence of a given consonant segment and a given vowel segment (e.g. [pi"1]) is necessarily in postpausal context, though it is true that for the sake of expositional simplicity, I have in principle cited in Commutative series 1 to 5 a good number of those cases in which the sequence of the phonetic segments in question occurs in postpausal context. (It goes without saying that the asterisked sequences of the phonetic segments are irrelevant to what has just been said.) However, this should not be taken to mean that the sequence of phonetic segments concerned does not occur in certain other contexts as well. For example, although I have cited [pi"1], this does not exclude e.g. [or pi7 dor’i] 'mandarin duck', a multiplet I could have cited as well. Given the phonetic context [- i7] for Commutative series 1 , such a multiplet as [or piate] 'guesswork' would not, strictly speaking, be a candidate multiplet for the reason that [ate], the sequence of phonetic segments that follows [pi], is pronounced on the same pitch as [pi]. However, in view of the absence of any repercussion of different accentual patterns on the occurrence or non-occurrence of the phonetic segments of Japanese words, mentioned in (3) in 9.2, a good deal of versatile interpretation of the phonetic context of each of Commutative series 1 to 5 is allowable, so much so that even the case of [or pi7 dor’ i] would be perfectly acceptable. On the other hand, I prefer to maintain a reasonable degree of uniformity in the matter of the accentual pattern associated with a commutative series, and it seems advisable in the meantime for the analyst to allow in e.g. [pi"1], [pi7 ro] 'white' and [or pi7 dor’i] for Commutative series 1 but to refrain from incorporating e.g. [or pi7 dor’ i] until later in the commutation test when he will have found for sure the absence of repercussion of different accentual patterns on the occurrence or non-occurrence of given phonetic segments in Japanese. What has been said above does not apply to Commutative series 6 to 10, as I will explain in 9.3. (3) As an example of the occurrence of [ga7] in Commutative series 3, I have cited not only [(ka)r ga7 (ipa)] 'assailant' but also [ga7] a word which some may feel rather too exceptional to be cited as an multiplet here and which occurs in postpausal context and is an adversative particle used at the beginning of an utterance which is obligatorily preceded by another utterance. This adversative particle, which roughly means 'However', corresponds in its function to the English adversative conjunction when used at the beginning of an utterance which is preceded by another utterance which ends with a pause.
Chapter 9: Identifying the consonant phonemes of Japanese
83
(4) Notice the non-occurrence of [z] in intervocalic context in the phonetic context [- e"1] with which Commutative series 2 is associated. On the other hand, Commutative series 1, 3, 4 and 5 are not affected where the occurrence of [z] in intervocalic context applies and relevant multiplets have been cited. In standard Japanese, the occurrence of [z] before [e] in intervocalic context is unattested, so that e.g. Magellan, as a loanword, is pronounced with [z] ([mar ze"lraN]), not [z]. (5) The same applies to the occurrence of [z] in intervocalic context in the phonetic context [- i"1] with which Commutative series 1 is associated. Commutative series 2, 3, 4 and 5 are not affected and the relevant multiplets have been cited. 9.3
Remarks on Commutative series 6 to 10
(1) The phonetic contexts with which Commutative series 6 to 10 are associated is such that the sequence of a given consonant segment and a given vowel segment (e.g. [b’ir ], [ber ], [bar ], [bor ], [buir ]) is pronounced on a pitch lower than the segment or the sequence of segments, if any, that follows, so that for example, [bar po] is so pronounced that [ba] is on a pitch lower than that on which [eo] is pronounced. (2 ) I have not specifically mentioned, in indicating the phonetic context with which each of Commutative series 6 to 10 is associated, whether the sequence of a given consonant segment and a respective vowel segment (e.g. [bar ]) is necessarily in postpausal context. I should add here that each such sequence can occur in other contexts than postpausal context. (3) It is recalled that I cited [(ha)r g’i",(cir’i)] 'gnash' as an instance of the occurrence of [g’i-1] in Commutative series 1, [(a)r ge"l(do)] 'push-up door' as an instance of the occurrence of [ge"1] in Commutative series 2, [(ka)r ga"l(iqa)] 'assailant' as an instance of the occurrence of [ga"1] in Commutative series 3, [(o)r go"'(soka)] 'austere' as an instance of the occurrence of [go"1] in Commutative series 4, and [(e)r gtu'1 (i)] 'acrid' as an instance of the occurrence of [gui"1] in Commutative series 5. It will have been seen that at corresponding points in Commutative series 6 to 10, no multiplets which would exhibit instances of the occurrence of [g’ ir ], [ger ], [gar ], [gor ] and [giur ] are cited. This is because no Japanese word begins (in postpausal context) with [g’ ] or [g], except for [ga] which is - let it be recalled - an adversative particle beginning an utterance which is obligatorily preceded by another utterance ending with a pause (rather corresponding to However in English when used at the beginning of an utterance preceded by another utterance ending with a pause), [ga*1] in question is a shortened form of [da^ga], It is easy to see that [ga"1] qualifies as a multiplet in Commutative series 3 but not in Commutative series 8 . (4) Although not specifically mentioned in the presentation of [- ir ] and [- ui p] as the phonetic contexts associated with Commutative series 6 and 10, respectively, it is to be understood that [i] or [in ], which is voiced, occurs when the consonant segment which precedes either is voiced (e.g. [b’ir koo)] 'nostril', [b’ujr uikeN] 'fallacy') but that [j] or [ui], which is voiceless, occurs when flanked by voiceless consonant segments (e.g. [?j(r kakui)] 'comparison', [kui r (tsui"’)] 'shoe'), but not otherwise (e.g. [cqir (ziN)] 'acquaintance', [p’i r (ano)] 'piano', [^'■ (jin'1)] 'winter', [smr (iee)] 'swimming'). Note that the occurrence of [i] or [j], or [ui] or [to], in similar segmental environment is irrelevant to Commutative series 1 or 5 as the phonetic contexts associated with them, i.e. [- in] or [—hi"’], make it the rule - on account of the accentual pattern involved - that [i] or [ui] occurs. (5) [j] is included as occurrent in Commutative series 8 to 10 but as non-occurrent in Commutative series 6 and 7, and [uj] is included as occurrent in Commutative series 8 only, and included as non-occurrent in Commutative series 6 , 7, 9 and 10.
84
9.4
Japanese Phonology: A Functional Approach
Defining the Japanese consonant phonemes in terms of relevant features
Each of Commutative series 1 to 10 presented further above lists, by way of the multiplets, 37 consonant segments whose phonological status I am to establish. These 37 consonant segments are arranged in an identical order in each commutative series, starting with [p’] and ending with [N]. The asterisked items correspond to lacunae, as they are non-occurrent. The identities of the consonant segments that might be found in the non-occurrent items are not the same in the individual commutative series; for example, [p] in Commutative series 1 but not in Commutative series 2, and [p’] in Commutative series 2 but not in Commutative series 1. None of the ten commutative series presented further above is a maximally differentiative commutative series, as is evident from the presence of a number of lacunae therein. We shall discover at a later stage of the commutation test that these lacunae are phonetic manifestations of instances of neutralization of phonological oppositions or systematic non-occurrence of individual phonemes. In spite of lacunae, the multiplets of the ten commutative series enable the analyst to elicit and define the consonant phonemes in Japanese. As previously suggested, the analyst will also find out that different accentual patterns have no repercussion on the occurrence or non-occurrence of individual consonant phonemes of Japanese. It is suggested that the readers should look first at Commutative series 3, 4, 5, 8 , 9 and 1 0 , i.e. with regard to the phonetic contexts [- a-1], [- o'1], [-in 11], [- ar ], [- or ] and [-u ir ]. These phonetic contexts can now be phonologically assessed in terms of / - a/, l- o l and /-in /. These commutative series contain very few lacunae, and moreover, lend themselves easily to eliciting and defining the consonant phonemes, in the first place in the contexts 'before /a/, 'before /o/' and /before /to/' in particular. Of course, with regard to the phonetic contexts [- P ], [-ui ’’], [- ip], and [- u ir ], phonologically assessed as the contexts 'before /iP and /before /to/', we need to look at Commutative series 1, 2, 6 and 7. In fact, we need to go forward and backward between the above-mentioned two groups of commutative series in order to elicit and define all the consonant phonemes of Japanese during the course of the commutation test. On the basis of the data presented in Commutative series 1 to 10, we can elicit and identify the following 29 consonant phonemes in terms of relevant features. As was done in 8.4 in the case of the vowel phonemes of Japanese, I will first enumerate below the relevant feature(s) of each consonant phoneme followed immediately, within parentheses, by the identity/ties of the other consonant phoneme(s) to which the consonant phoneme in question is directly opposed. Each instance of such a phonological opposition between the consonant phonemes represents 'direct opposition' and the terms of a direct opposition are direct neighbours in the Japanese consonant phoneme system. For instance, /p/ - /p’/, /p’/ - /k7, /p7 - lb' I and /p7 - /m7 represents each a direction opposition whose terms are direct neighbours in the system of the Japanese consonant phonemes. Table 1 The Japanese consonant phonemes and their relevant features /p/: "voiceless" (vs. Ibl), "non-palatalized" (vs. /p’/), "labial" (vs. /t/, IkJ), "non-nasal" (vs. /ml). /p7: "voiceless” (vs. Ib'l), "palatalized" (vs. /p/), "labial" (vs. /t/, Ik'/), "non-nasal" (vs. Im'/). Ibl: "voiced" (vs. /p/), "non-palatalized" (vs. Ib'l), "labial" (vs. /d/, /g/), "non-nasal" (vs. /m/). Ib 'l: "voiced" (vs. /p7), "palatalized" (vs. Ibl), "labial" (vs. /d/, /g’ /), "non-nasal" (vs. Im'f).
Chapter 9: Identifying the consonant phonemes of Japanese
85
/t/: "voiceless" (vs. Id/), "apical" (vs. /p/, /p’/, Pd, /k’/), "non-nasal" (vs. Inl, /n’/). Id! ■ "voiced" (vs. Id), "apical" (vs. /b/, /b’/, /g/, Ig'l), "non-nasal" (vs. Ini, In'f). IkJ: "voiceless" (vs. Ig'l) "non-palatalized" (vs. Ik'/), "dorsal" (vs. /p/, /t/), "non-nasal" (vs. /g/). Ik'/ ■ "voiceless" (vs. Ig'l), "palatalized" (vs. Ikl), "dorsal" (vs. /p’/), "non-nasal" (vs. /g’/). /g/: "voiced" (vs. Ikl), "non-palatalized" (vs. Ig'P), "dorsal" (vs. Pol, Id!), "non-nasal" (vs. /g/). Ig'l: "voiced" (vs. Ik'l), "palatalized" (vs. Igf), "dorsal" (vs. lb'I), "non-nasal" (vs. /g’/). /$/: "labial" (vs. /?/), "fricative" (vs. /p/, /p’/, /b/, /b’/, /ml, Im'l). /?/: "palatal" (vs. /$/), "fricative" (vs. /cq/, /j/). /j/: "palatal" (vs. Iu\l), "spirant" (vs. /?/, /cq/). /hi: "glottal" (vs. all the other consonant phonemes). /q/: "voiceless" (vs. /z/), "hush" (vs. /s/). /?/: "voiced" (vs. /q/), "hush" (vs. Izl). Is!: "voiceless" (vs. Izl), "hiss" (vs. /q/). It! : "voiced" (vs. Is/), "hiss" (vs. /?/). Its/: "apical" (vs. /cq/), "affricate" (vs. /t/, /d/, /n/, In'f). /cq/: "palatal" (vs. Its!), "affricate" (vs. /q/, /j/). Id: "non-palatalized" (vs. Ic'l), "liquid" (vs. all the other consonant phonemes except Ir'l). Ic'l: "palatalized" (vs. Id), "liquid" (vs. all the other consonant phonemes except Icf). /ml: "non-palatalized" (vs. Im'l), "labial" (vs. Ini, r)/), "nasal" (vs. /p/, Po!). Im'l: "palatalized" (vs. /ml), "labial" (vs. In'I, /rj’/), "nasal" (vs. /p’/, Po' P). Inl: "non-palatalized" (vs. In'/), "apical" (vs. /ml, /rj/), "nasal" (vs. /t/, Id/). In'/: "palatalized" (vs. Inl), "apical" (vs. Im'l, /g’/) "nasal" (vs. l(l, IdP). /g/: "non-palatalized" (vs. /rf/), "dorsal" (vs. Iml, Inf), "nasal" (vs. Ikl, Igf). /g’/ : "palatalized" (vs. /g/), "dorsal" (vs. /m’/, In’P), "nasal" (vs. Ik'l, Ig'l). /ilj/ : "dorsal" (vs. /j/), "spirant" (vs. Pd, Pa'I, Igl, lg' I, /g/, /g’/). /a/: (see 9.8.10 for my discussion of this phoneme). Here below is a convenient summary of the phonological contents of all the consonant phonemes of Japanese. The summary derives of course from the list presented just above. I shall provide in addition the information about the contexts in which each consonant phoneme is valid. The reason why I do so is that we are also interested not only in the context(s) in which each consonant phoneme is valid but also in that or those in which it is not valid for the reason that either the phoneme does not occur or the phonological opposition one of whose terms is the consonant phoneme in question is neutralized, as we shall see later.
^TableTl
The Japanese consonant phonemes with their phonological contents and their contexts of validity /p/: /p’/ : Pol: lb'I: Id: Id!: Pal: Ik'l: Ig/: Ig'l: /$/:
"voiceless non-palatalized labial non-nasal". Valid before /a/, lol and Iml. "voiceless palatalized labial non-nasal". Valid before /a/, lol and Iml. "voiced non-palatalized labial non-nasal". Valid before /a/, lol and Iml. "voiced palatalized labial non-nasal". Valid before /a/, lol and /ml. "voiceless apical non-nasal". Valid before I'd, lei, Id, lol and /ml. "voiced apical non-nasal". Valid before Pd, Id, Id and lol and Iml. "voiceless non-palatalized dorsal non-nasal". Valid before Id, lol and Iml. "voiceless palatalized dorsal non-nasal". Valid before Id, lol and Iml. "voiced non-palatalized dorsal non-nasal". Valid before Id, lol and Iml. "voiced palatalized dorsal non-nasal". Valid before Id, lol and Iml. "labial fricative". Valid before Id, Id, Id, lol and Iml.
86
/?/: /j/: /h/: /q/: /?/: /s/: /z/: /ts/: /cq/: Id : /r’/ : An/: An’/ : Id : /n’/ : A)/: A)’/ : /uj/: Id :
Japanese Phonology: A Functional Approach
"palatal fricative". Valid before /i/, /a/, /o/ and /tu/. "palatal spirant". Valid before /a/, /o/ and /in/. "glottal" .57 Valid before /e/, /a/ and /o/. "voiceless hush". Valid before /i/, /e/, /a/, /o/ and /ui/. "voiced hush". Valid before lil, /e/, /a/, /o/ and /in/. "voiceless hiss". Valid before Id, /a/, /o/ and /in/. "voiced hiss". Valid before Id, Id, lol and /ml. "apical affricate". Valid before Id, Id, lol and /in/. "palatal affricate". Valid before /i/, Id, Id lol and /tu/. "non-palatalized liquid". Valid before Id, lol and /in/. "palatalized liquid". Valid before Id, lol and /ui/. "non-palatalized labial nasal". Valid before Id, lol and /in/. "palatalized labial nasal". Valid before Id, lol and /in/. "non-palatalized apical nasal". Valid before Id, lol and /in/. "palatalized apical nasal". Valid before Id, lol and /in/. "non-palatalized dorsal nasal". Valid before Id, lol and /in/. "palatalized dorsal nasal". Valid before Id, lol and /in/. "dorsal spirant". Valid before lil, Id, Id and lol. (see 9.8.10 for my discussion of this phoneme).
Two consonant phonemes, i.e. /j/ 'palatal spirant" and /uj/ "dorsal spirant" will be discussed further below, in 9.8.8 and 9.8.9, respectively, as the phonological analysis of [j] and [iqa] calls for some explanation. The last phoneme (/a/) in Tables 1 and 2 above will be discussed in 9.8.10. 9.5
The structure of the Japanese consonant phoneme system
The 29 consonant phonemes of Japanese which we have elicited and defined in terms of relevant features are diagrammatically presented in Appendix 1 in terms of orders and series, which makes clear the nature of the system of the 29 Japanese consonant phonemes. The consonant phoneme system presented in Appendix 1 is the one applicable to what I call 'the minority speech'. At this point, I request the readers to consult Appendix 1 in order to follow my explanation I am to give below in the rest of the present section. The presentation of the 29 consonant phonemes of Japanese are to be seen in terms of 'orders' (i.e. "labial" order; "apical" order, "palatal" order, "hiss" order, "hush" order, "dorsal" order and "liquid" order) and 'series' (i.e. "non-nasal" series', "nasal" series, "voiceless" series, "voiced" series, "non-palatalized" series, "palatalized" series, "fricative" series, "affricate" series and "spirant" series). Notice that the relevant feature "glottal" does not belong to any series or order. Fourteen out of the 29 consonant phonemes take part in correlations. The interlocking of correlations can be clearly seen in the list of the Japanese consonant phonemes presented above. Thus, /p/ and IkJ (in the "voiceless" series) and /b/ and Igl (in the "voiced" series) form a correlation, and so do Ip'I and Ik.'I (in the "voiceless" series) and /b’/ and /g’ / (in the "voiced" series). Furthermore, Ipl, fkl, Ibl and /g/ (in the "nonpalatalized" series) and Ip'I, Ik' /, /b’/ and Ig 'l (in the "palatalized" series) form a correlation. Elsewhere, /ml, I d and /g/ (in the "non-palatalized" series) and /m’/, In'I and /g’/ (in the "palatalized" series) form a correlation. Moreover, Ipl, IkJ, Ibl, Igl Ip'I, Ik'I, lb' I and Ig'l (in die "non-nasal" series) and /ml, Id, /g/, Im'l, In'I and /g’/ (in the "nasal" series) form a correlation. Of the above-mentioned Japanese consonant phonemes, Ipl, Ip'I, Ibl, lb'I, IkJ, lk'l,lg/, lg'l,lm l,lm 'l, /g/and/g’/ form among themselves a bundle of correlations which can be schematically represented as follows.
Chapter 9: Identifying the consonant phonemes of Japanese
/p /-/p 7 | |
87
— An/-/m7
Ib l- lb 'l
I I /k /-/k ’/ I I /g /-/g ’/
-
/0/ - / 0V
Four other consonant phonemes of Japanese form a correlation. They are /s/ and /p/ (in the "voiceless" series) and /z/ and /?/ (in the "voiced" series). This correlation can be schematically represented as follows. /s/ I IzJ
— —
/p/ I Izl
IV and Id! (in the "non-nasal" series) and Ini and In7 (in the "nasal" series) do not form a correlation. If /t7 and Id' / had been elicited as well as Itl, Idl, In'I and Ini, they would form a bundle of correlations which could be represented as follows. /t/
—
It'/
Idl
—
Id'/
Ini
—
In'/
1
1
However, as I do not establish either /t7 or /d’ / in Japanese, such a bundle of correlations as the above is extraneous to Japanese. The other Japanese consonant phonemes cannot be said to form a correlation(s) or a bundle(s) of correlations. These are /t/, IdJ, Ini, In'I, Id, Idl, l$l, /?/, Its/, /cp/, 1)1, ln\l and Pnl. There are different degrees of non-correlatedness among these consonant phonemes. At one end of the spectrum there are Id and Id / which are bound with each other by both being "liquid" and are differentiated from each other through the opposition between "palatalized" and "non-palatalized", and /t/, Idl, Ini, In'I which are bound with each other by all being "apical", and Ixl and Idl bound with each other by both being "non-nasal" and Ini, In'I by both being "nasal". It! and Idl are differentiated from each other through the opposition between "voiceless" and "voiced", while Ini and In'/ are differentiated from each other through the opposition between "non-palatalized" and "palatalized". Similarly, /?/, /cp/ and /]/ are bound with each other by being "palatal" while being differentiated from each other through the opposition between "fricative", "affricate” and "spirant”. At the other end of the spectrum of non-correlatedness there is Ihl which consists of a single relevant feature "glottal". There lie, in between, /$/, Its/ and /uj/, each of which consists of two relevant features. One of the two relevant features of /$/, Itsl and /uj/ ("labial" in the case of /$/; "apical" in the case of /ts/; and "dorsal" in the case of In)/) is the relevant feature that is also possessed by some other consonant phonemes which do or do not partake of the correlations mentioned above, /j/ consists of two relevant features, one of which ("palatal") is also possessed by some other consonant phonemes and the other ("spirant") by one other consonant phoneme, neither phoneme partaking of a correlation.
88
9.6
Japanese Phonology: A Functional Approach
Realizations of the Japanese consonant phonemes
Here follows a table in which, all 29 consonant phonemes are again listed, but this time with information about the realizations of the respective consonant phonemes, followed by (to repeat) the specification of the contexts of validity, i.e. the contexts in which the respective consonant phonemes are valid. The phoneme notatd as Id, which has yet to be defined in terms of relevant features (see 9.8.10) is added to the 29 consonant phonemes in the following list. Table 3 The Japanese consonant phonemes with their realizations and their contexts of validity /p/: /p’/ : lb/: /b’/ : /t/: IdJ: IkJ: Ik’/ : /g/: /g’/ : /$/: Iql: /j/: Id : Id : /?/: Isl: Izl: /ts/: le d : Id: /r’/ : Iml: /m’/ : Id : In'/: Id : A)’/ : h i\l: Id :
[p] (voiceless non-palatalized bilabial plosive; before /a/, /o/ and /in/). [p’] (voiceless palatalized bilabial plosive; before /a/, /o/ and /ml), [b] (voiced non-palatalized bilabial plosive; before /a/, /o/ and An/). [b’] (voiced palatalized bilabial plosive; before /a/, /o/ and An/). [t] (voiceless non-palatalized apico-dentialveolar plosive; before Id, Id and /o/); [t’] (voiceless palatalized apico-dentialveolar plosive; before /i/ and /in/). [d] (voiced non-palatalized apico-dentialveolar plosive; before Id, Id and /o/); fd’] (voiced palatalized apico-dentialveolar plosive; before III and An/). [k] (voiceless non-palatalized dorso-velar plosive; before Id, lol and /in I). [k’] (voiceless palatalized dorso-velar plosive; before Id, lol and An /). [g] (voiced non-palatalized dorso-velar plosive; before Id lol and /in/). [g’] (voiced palatalized dorso-velar plosive; before Id, lol and /in/). [$] (voiceless bilabial fricative; before III, Id, Id, lol and An/). [] (voiced laminodorsoalveolopalatal nasal) as a realization of /n’/ before /a/, /o/ and An / (as already mentioned in Akamatsu 1997a, pp. 121-124) rather than, as the readers might well expect, [n’] (voiced palatalized apico-dentialveolar nasal). This is because, as I said in Akamatsu (1997a, p. 122), '...palatalization which is achieved by the raising of the front of the tongue tends to push the tip of the tongue backward and downward when the articulation of the consonant for which palatalization is concomitantly achieved involves the tongue', whereupon I appended a relevant note (note 342 on p. 320 in op. cit.) in which a passage from Catford (1988, p. 108) writing about Russian, was quoted as being indirectly relevant to [t>] in addition to, and in support of my own above explanation. In the passage of Catford, we read, among other things that 'The upward thrust of the anterodorsum of the tongue tends to retract and depresses the apex', which is what interests us in particular as it applies to Japanese as well as Russian. In other words, in the articulation of [r>], the tip of the tongue does not touch the dentialveolar region in front of the oral cavity (but hangs down), while in that of [n’] it does. A full description of [i>] can be found in Akamatsu (1997a, pp. 121-124) and the readers are further invited to look also at notes 340 to 347 in Akamatsu (op. cit, pp. 320-321) in connection with [t>]. It is to be added that [r>] (voiced laminodorsoalveolopalatal nasal) should not be confused with [p] (voiced dorso-palatal nasal) which occurs in French (as in montagne 'mountain') or Italian (as in gnomo 'gnome') or Portuguese (as in tomanho 'so big'), as Biedrzycki (1974, p. 56), independently of me, warns while referring to [t>] in Polish. The contact between the anterodorsum and the hard palate is over a substantially larger area for [p] than for [t>]. (4) [1], as a realization of Id, occurs preceded by [n] (e.g. [ka^nlo] 'honeydew'), while [P], as a realization of Id I, occurs preceded by [n’] (e.g. [bennT i] 'convenience'). Similarly, [r], as a realization of Id, occurs when not preceded by [n] (e.g. [a^ra^i]
90
Japanese Phonology: A Functional Approach
'storm'), and [r’], as a realization of /r’/, when not preceded by [n’J ([kmr r ’i",do] 'sliding door'). (5) [t] occurs as a realization of /t/ when followed by Id, Id and lol (e.g. [te"1] 'hand', [ta"*] 'paddy field', [to"1] 'whetstone'), while [t’] occurs as a realization of It/ when followed by HI (e.g. [t’i",mpaT>i] 'timpani', [t’ir icatsm] 'T shirt') and /m l (e.g. [t’m",raba] 'tuba'). As can be seen from the multiplets cited in Commutative series 1 and 5, and also those cited in Commutative series 6 (but, note, not those cited in Commutative series 10 ) and cited again just above, [t’J (voiceless palatalized apico-dentialveolar plosive) is a realization of /t/ before /i/ or /m/ in the pronunciation of loanwords only. This will have been seen to be the case from a number of relevant example words adduced in 9.1. As already shown in Commutative series 1, 5 and 6 (but not in 10), [ti] is non-occurrent in Japanese (though [t’i] can occur in loanwords), and [tm ] is non-occurrent in non loanwords in Japanese (though occasionally occurrent in loanwords). Loanwords with [t] in word-final context or word-medial context in the donor languages (e.g. English) are pronounced with [to] as in the case of [ba^nto] (< b u n t) or [ma'Vtoresm] (< mattress) or [tsm] as in [kar tsui retsm] (< cutlet), etc. Likewise, [d] occurs as a realization of /d/ when followed by Id, Id and lol (e.g. [denN] 'legend', [da^ba] 'jade', [do"1] 'degree'), while, as can be seen from the multiplets cited in Commutative series 1 and 5, and also in Commutative series 6 and 10, [d’] (voiced palatalized apico-dentialveolar plosive) occurs as a realization of /d/ when followed by HI (e.g. [d’i",(stqko)] 'disco', d’ir (re“,ttanto)] 'dilettante') and Ival (e.g. [d’m‘,etto]/[d’rar e‘ltto] 'duet'), that is, again in exclusive connection with loanwords, [di] is non-occurrent in Japanese (though [d’i] can occur in loanwords), while [dm] is non-occurrent in both non-loanwords and loanwords in Japanese. Loanwords with [d] in word-medial context or word-final context in the donor languages (e.g. English) are pronounced with [do] as in the case of [ra^ndor’ a] (< laundry), [bar do‘lm’intoN] ()] 'owner' [r>ui’,(xnzi)] 'baby' [(ka)r g’in(tD)] 'snail'
Commutative series 6 [-ir ] [p’i r (ano)] 'piano' *[pi] [b’ir (koo)l 'nostril' nbii
*[p’e] [per (r’ikaN)] 'pelican' *[b’e]
[k’ir ] 'yellow (colour)' *[ki] [g’ir (ree)] 'etiquette' *[gi]
*[k’e] [ker ntoo] 'boxing' *[g’e] [ger (r’i)] 'diarrhoea'
[m’ir (se"1)] ’shop’ *[mi] *[n’i] *[ni]
*(m’e] [mer (bae11)] 'bud' *fn’er l [ner (zmm’i)l 'rat' *[t>e] Tn’e] *[ge]
fT>ir ( u i a ) ] ’g a r d e n '
*[n’i]
Commutative series 7 C-er ]
[ber (Vpoo)] 'compensation'
Chapter 10: Establishing /g/, /g’/, /g /, /g’/ , k/ and /k’/, /in the minority speech) and /X/, /X’/, /Y/ and /Y7 (in the majority speech)
111
Commutative series 9 r-o ri [p’or (ko"lN)] 'jerkily (in bowing)' [por (kkar’i)] '(surfacing) suddenly' [b’or (odoo)] 'equality'
Commutative series 10 r-ixir i [p’rar (ratto)] 'whizzingly'
[bor (o)] 'rod'
[bra r (ai)] 'percentage'
[k’ar (kra)] 'guest' fkar (ba)] 'birch' [g’ar (kixi)] 'inverse' [gar (kiu)] 'tablet'
[k’or (obai)] 'auction' [kor (oeN)] 'park' [g’or ($tuN)] 'fish-meal' [gor (zra",m)] '50'
[k’rar (ra R>i)] 'suddenly' [krar tsra] 'shoe' [g’rar (ra)too] 'cowpox' [gra r (cqi)l 'grumbhng'
[m’ar (kui)] 'pulse'
[m’or (obaN)] 'alum'
fmar (cpi)] 'town' *[n’a] [nar (ta)] 'hatchet' [r>ar (a"lgo)] 'miaow!'
[mor (cci)] 'rice cake' *[n’o] [nor (okoo)] 'farming' [r>or (odoo)] 'urethra'
*[n’a] *[(ial
*[n’o] *[fio]
[m’rar (razikkcuho"'orra)] 'music hall' [mrar (sai)] 'celibacy' *[n’ra] [nra r (ime"1)] 'seam' [i>rar (ra)iN] 'hospitalization' *[g’ra] *[np>]
Commutative series r- ar l
8
*[p’a] [par (ase"'nto)] 'percent’ [b’ar (kra",reN)] 'white lo tus' [bar (co)] 'place'
10.5.2
[pra r (tsui "lN)] '((to) be cut) suddenly and neatly' [b’ixir (uikeN)] 'fallacy'
Defining /g/, /g’/, /rj/ and /g’/ in terms of relevant features
Such data as are given in Commutative series 3 and 8 (associated with [- a]), 4 and 9 (associated with [- o]) and 5 and 10 (associated with [- ra]), partially reproduced in 10.5.1, are conducive to, but less than adequate to, establishing /g/, /g’/, /rj/ and /g’/. This is because in word-initial context, whilst [g] and [g’j are occurrent before all of [a], [o] and [ra], [g] is non-occurrent before [o] and [in] and [gT] is non-occurrent before [a], [o] and [m ]. This makes it impossible to judge with certainty whether the difference between [g] and [g] and that between [g’] and [g’ ] are phonologically relevant. We therefore also need such data in word-medial context as show with certainty that the difference between [g] and [g] and that between [g’ ] and [g’] are phonologically relevant, like [j?ra",rago] '15' vs. [j?m",rago] 'the home front', and [se^ggo] '1,005' vs. [se^ggo] 'postwar', the members of each pair being minimally or near-minimally differentiated from each other, i.e. by virtue of the difference between [g] and [g]. Further examples will be given in 10.5.5 of such pairs of minimally or near-minimally differentiated words that will indicate the phonological relevance of the difference between [g] and [rj] and that between [g’ ] and [g’j. It is not only desirable but necessary to look at both word-initial context and word-medial context in examining
Japanese Phonology: A Functional Approach
112
the phonological relevance or otherwise of the difference between [g] and [g] and that between [g’] and [g*]. On the basis of evidence obtained in both word-initial context and word-medial context, we can establish and define the following in terms of relevant features: /g/: /g’/ : /g/: /g’/ :
"voiced non-palatalized dorsal non-nasal". "voiced palatalized dorsal non-nasal". "non-palatalized dorsal nasal". "palatalized dorsal nasal".
This is what I have already done in 9.4. These consonant phonemes are valid before /a/, /o/ and /ui/. In order to identify all the relevant features of /g/, /g’/, /g/ and /g’/, it is of course necessary to take into account not just the difference between [g], [g’], [g] and [g’], but the difference between [k], [k’], [n], [ tv], [m] and [m’] as well as [g], [g*], [g] and [O’ ] . 76 wdh regard to their occurrence before [a], [o] or [m]. In addition, we have established the following in 9.4. Ik/ : /k’/ : /m/: Im 'l: Ini: /n7 :
"voiceless non-palatalized dorsal non-nasal". "voiceless palatalized dorsal non-nasal". "non-palatalized labial nasal". "palatalized labial nasal". "non-palatalized apical nasal". "palatalized apical nasal".
The above consonant phonemes too are valid before /a/, /o/ and /m I. The attribution of "non-palatalized" to /g/ and that of "palatalized" to /g’ / results from the direct opposability between /g/ and /g’/. The attribution of "nasal" to both /g/ and /g’ / results from the direct opposability between /g/ on the one hand and /k/ and /g/ on the other, and from the direct opposability between /g’/ on the one hand and /k7 and Ig'l on the other. The attribution of "dorsal" to both /g/ and /g’/ results from the direct opposability between /g/ on the one hand and /m/ and Ini on the other, and from the direct opposability between /g’ / on the one hand and /m’/ and In'I on the other. The data appearing in Commutative series 1 and 6 (associated with [- i]) and those appearing in Commutative series 2 and 7 (associated with [- e]) are not conducive to establishing either Igl and Ig'l or /g / and /g’/. Before HI or /e/, there is no direct opposability not only between Igl and Ig'l but also between /g/ and /g7. This is because, as will be explained in the course of 11.3, Igl - Ig'l and /g/ - /g7 are neutralized before /i/ or Id. 10.5.3
The non-occurrence of [g] and [g’ ] in word-initial context in the minority speech and the majority speech
It is noticeable to whoever examines Commutative series 1 to 10 that [g] and [g’] do not occur in word-initial context (cf. Commutative series 8 , 9 and 10), though they occur in word-medial context (cf. Commutative series 3, 4 and 5). We are of course for the moment deliberately interested in the contexts [- a], [- o] and [- ru ] to the exclusion of [- i] and [—e]. I need to explain why [g] and [g’ ] are in principle non-occurrent in word-initial context except for [ga], i.e. [ga "*] (cf. Commutative series 3) but not [gar ] (cf. Commutative series 8 )). The occurrence of [ga"1] in word-initial context cum in postpausal context is highly exceptional in the sense that this occurrence is limited to one particular Japanese utterance-connective adversative particle meaning 'However'
Chapter 10: Establishing /g/, Ig 'l, /g/ , /g’/ , k/ and Ik 'I, /in the minority speech) and IXJ, fX.'I, IY I and IY '1 (in the majority speech)
113
and is pronounced [ga"1], a word already cited in (3) in 9.2. This single exception of [ga-1] apart, neither [g] nor [g’] occurs in word-initial context (whether or not coinciding with prepausal context) before any of the five vowel segments. In Japanese, it is the general rule that [g] or [g’J, but not [g] or [g’j (except of course for [ga"1] mentioned just above), occurs in word-initial context. What has been said up to this point about the non-occurrence of [g] or [g’ ] in wordinitial context actually concerns the minority speech. As for the majority speech, [g] or [g’j instead occurs in word-initial context. Even the above-mentioned utteranceconnective adversative particle is generally pronounced [ga], or occasionally either [ga] or [ga] in a random fashion. 10.5.4
Complexity of the occurrence of [g] or [g], or [g’ ] or [g’ ], in wordmedial context in the minority speech and its phonological implications
There are, in the minority speech, instances of the occurrence of [g] or [g’] in wordmedial context (e.g. [kor otooga"'kkoo] 'senior high school', [cir g’g’o'1 osoo] 'three written styles of Chinese characters') as well as those of the occurrence of [g] or [g’ ] (e.g. [ker ga] 'injury', [c or oga"'kkoo] 'primary school', [kar g’i'1] 'key', [sar g’g’oo] 'industry'). As can be surmised from such examples, the occurrence of [g] or [g], or that of [g’] or [g’]> in word-medial context is somewhat complex; I have given a brief relevant account in Akamatsu (1997a, p. 129), but the readers will find pertinent accounts in any good manual of Japanese phonetics. Incidentally, I have provided detailed phonetic accounts of [g], [g’]> [g] and [g’] in current standard Japanese in Akamatsu (1997a, pp. 124-133). It is a fact that if we also choose word-medial context for the commutation test, we shall encounter a substantial number of occurrences of [g], [g’], [g] and [g’] before [a], [o] or [ui]. However, our task of establishing the phonological oppositions /g/ - Ig'l and /g/ -/g ’/, and thereby the identification of /g/, Ig'l, /g/ and /g’/ themselves on the basis of these occurrences, is attended by complexity. An example of a pair of words like [or ogama] 'big toad' vs. [or ogama] 'scythe/cauldron' is of course ideal evidence on which to work, at least partially, towards establishing Igl and /g/ but another example of a pair of words like [kor otoog"lkkoo] 'senior high school' and [g] of [cor oga"1 kkoo] 'primary school' may initially puzzle the analyst, as the difference between [g] and [g] appears at first sight to be phonologically irrelevant-, both [gakkoo] and [gakkoo] relate to one and the same Japanese word meaning 'school'. It is by having confirmed the phonological relevance of the difference between [g] and [g] in word-initial context and also by having phonologically analyzed a case like [or ogama] 'big toad' vs. [or ogama] 'big scythe/cauldron' (in which the difference between [g] and [g ] is phonologically relevant) pertaining to word-medial context that the analyst can arrive at the conclusion that the difference between [g] and [g] in [kor otooga“,kkoo] and [cor oga"'kkoo] is phonologically relevant and [g] and [g] are realizations of Igl and /g/, respectively. What has just been said applies, mutatis mutandis, also to the difference between [g’] and [g’ ]. The occurrence of [g] or [g’] in word-medial context is typically very frequent in word-medial intervocalic context and in post-[g] cum word-medial context in the minority speech, and it is therefore rather the occurrence of [g] in wordmedial context that attracts our attention in the minority speech. The analyst will sooner or later discover that (i) the occurrence, in the minority speech, of [g] or [g’] in word-medial context (cf. [or ogama] (< [or o] 'big' + [gar ma] 'toad'), [kor otooga',kkoo] 'senior high school' (< [kor otoo] 'higher grade' + [gar kkoo] 'school'), [gar kkoog’o"'osee] 'school administration' (< [gar kkoo] 'school' + [g’o r osee] 'administration') is typically, if not necessarily, attested in composite words and that (ii) the occurrence of [g] or [g’] in word-medial context is in fact comparable with the
114
Japanese Phonology: A Functional Approach
occurrence of [g] or [g’] in word-initial context. For example, the occurrence of [g] in [kor otooga",kkoo] 'senior high school' (< [kor otoo] 'higher grade' + [gar kkoo] 'school') is comparable to the occurrence of [g] in [gar kkoo] when it is pronounced in isolation, though this will not account for the occurrence of [g] (instead of [g]) in [qor oga‘,kkoo] 'primary school' (< [co^o] 'small' + [gar kkoo] 'school'). Words like [or ogama], [kor otooga'lkkoo] are composite words (hence the occurrence of [g] in many cases), but [por oga"lkkoo], [or ogama], etc. are also composite words (in which [g] occurs). The apparent contradictory occurrence of [g] and [g ] in composite words may be explained in the following way. First, it is significant in the minority speech that in medial context within a simple word (that is, not a composite word), it is always [g] or [g’], never [g] or [g’], that occurs (e.g. [kar go] 'basket', [su^gra] 'immediately', [nor ga",sui] '(to) let go1, [kar g’i**] 'key', [p>ir g’ir’i] 'grip', [nong’i] 'beard (of barley)'), and instances of the occurrence of [g] medially in composite words can be seen as an extension of the regular occurrence of [g] in simple words. Second, it is notable that in the minority speech, it is [g] or [g’], not [g] or [g’j, that regularly occurs in word-initial context (the sole exception being the utterance-connective adversative particle [ga"1]), and instances of occasional occurrence of [g] or [g’] medially in composite words can be seen as the extension of the regular occurrence of [g’] or [g] in word-initial context. Incidentally, some scholars are of the opinion that when [g] or [g’] occurs at the beginning of the non-initial constituent in a composite word as in [qor oga"'kkoo], this is because the separate identity of the non-initial constituent is less in the speaker's mind than in e.g. [kor otooga"'kkoo]. If this is so, the occurrence of [g] in [qor oga"'kkoo] suggests that such a composite word is used as if it were more like a simple word. I have earlier mentioned, in 7.10, that it is necessary (in the case of some languages) or desirable (in the case of other languages) to perform the commutation test in which the multiplets should be linguistic forms that do not involve potential pauses, but added that this precaution can be largely relaxed or even does not apply to the commutation test where Japanese is concerned, given the lexical makeup of Japanese. Indeed, in the case of the phonological analysis of Japanese, both simple words (which do not involve potential pauses) and composite words (which may or may not)77 can and need be taken into consideration in choosing multiplets. This fact will have been seen not only just above but will have been seen further above in this book. It may be recalled that as I said in 7.10, a huge percentage of Japanese words are composite words. 10.5.5
Pairs of Japanese words conducive to establishing /g/ and /g/ (and 7g’ / and/g’/')
It should first be clarified that the reason why I have put 7g7 and /g’ /' within parentheses in the title of this section is that I have not been able to find examples of relevant pairs of Japanese words. This does not seem, prima facie, to detract from the potentiality of establishing /g’/ and/g’/ which differ from Igl and /g/ by having the relevant feature "palatalized" instead of "non-palatalized”. Much has been written in Japanese phonetics literature about the justification of establishing Igl and /g/ (this is presumably held to apply to /g’ / and /g’/ as well) in standard Japanese pronunciation. The traditional practice has been to adduce a number of pairs of minimally or near-minimally differentiated words (nearly always composite words) which are distinguished from each other basically by virtue of the difference between [g] and [g], and are hence said to be distinguished through Igl - /g/. When /g/ and Igl are identified in this way by non-functionalists, it is not necessary to bring in /k/, Ik'I, /ml, Im'l, Ini and In'I, as well as lg' I and /g’/, since a phoneme is to them not conceived as a sum of relevant features. At any rate, a limited (I emphasize, limited)
Chapter 10: Establishing /g/, /g’/, /g/ , /tj’/ , k/ and /k’/, /in the minority speech) and 1X1, IX 'I, A7 and IY ’1 (in the majority speech)
115
number of pairs of minimally or near-minimally differentiated words like the following are often cited by scholars for the purpose of establishing /g/ and /g/. It must be said that some of these words are either rare or even artificial or even unknown to average Japanese speakers. [or ogama] 'big toad' vs. [or ogama] 'scythe/cauldron'78 [kor gaN] 'young wild goose' vs. [kor gaN] 'lakeshore' [bogr goro] 'ground ball' vs. [bogr goro] 'round about the Bon Festival' [ar igo] 'well-matched go players' vs. [a^igo] 'protection' [itar ga'Irasxu] 'plate-glass' vs. [ir taga",rasm:] 'make (someone) feel pain' [k’i r tag’ie(i] 'Engineer Kita (surname)' vs. [k’j r tag’ic(i)] 'northern shore' [dor ktuga] 'poisonous moth' vs. [dor kiu',ga]' poisonous fang' [k’ir guim’i] 'oleaster' vs. [k’ir gtnmT'1] 'wooden framework'79 [jzui",ujgo] '15' vs. [j2 tij',,ujgo] 'the home front'79 [da^igo] '5th' vs. [da^igo] 'a kind of thick and sweet liquid made from milk'80 [se^ggo] '1,005' vs. [se^ggo] 'postwar'81 [ga"1] 'elegance' vs. [ga-1] 'However' [ser e?igo"lro] 'demagogue' vs. [or ginngpro] 'about noontime' [gar raga'1ra] 'toy'82 vs. [ir egara",j 'the standing of a family' [sor nogo'1] 'that (Japanese) game of go' vs. [sopnogo] 'thereafter' [ar kaigar astu ] 'red plate-glass' vs. [kar igara] 'shell' [kor ojunugakko] 'this kind of school' vs. [tju: r iuga‘,kkoo] 'junior high school'83 (It may be of interest to observe that of all the example words cited above, only [ga"1], [ga-1] and [gar raga",ra] are simple words, all the others being composite words.) Of the above cited examples, [k’j r tag’iqi] 'Engineer Kita (surname)' and [k’i r tag’ip(i)] 'northern shore' are irrelevant to my phonological analysis in establishing /g/,/g’/,/g/ and /g’/, since, as already said towards the beginning of this section, /g/ - /g’/ and /g/ - /g’/ are neutralized before /i/ or Id, so that none of Igl, Ig'l, /g/ and /g’/ occurs in these contexts. It is expected that [k], [k’J, [n], [t>], [m] and [m’] too are potentially occurrent at the points where [g] or [g] (or for that matter, [g’J or [g’]) occurs before [a], [o] or [ra] in the above cited words and other phonetically possible examples. This will be conducive to establishing Igl, /g/, Ig'l and /g’/, which will be directly opposable to fkl, Ik’/, /ml, Im'l, Id and /n’/. 10.5.6
Consequence of the establishment of Igl, Ig'l, /g/ and /g’/ on phonological notation
The establishment of Igl, Ig'l, /g / and /g’/ in the consonant phoneme system ascribable to the minority speech, as I have done, has repercussion on phonological notation in that [g], [g’j, [g] and [g’ J we have seen in the phonetic notation of the example words can now be analyzed to correspond to the respective consonant phonemes of which they are realizations. Recall that we are still concerned with the context 'before /a/, lol or An /' to the exclusion of 'before /i/ or Id'. Here are some representative examples of the phonological notation in question, taken from among the examples I have adduced during the course of this chapter so far. I shall only indicate Igl, Ig'l, /g/ and /g’/, omitting the rest of the phonemes or archiphonemes that should otherwise be included in the phonological notation of the example words. Also, I shall omit the indication of the pitch patterns (a manifestation of accentual patterns).
116
Japanese Phonology: A Functional Approach
The phonetic notation is placed on the left and its corresponding phonological notation on the right. [or ogama] [or ogama] [kor gaN] [kor gaN] [qor oga'1 kkoo] [kor otooga"'kkoo] [gar kkoog’o ’’osee] [kar go] [rja-1] [se^ggo] [se^ggo] [t>or goga",qima] [g’ar ktur’i] [gar kin] [kar g’aktu]
/■■■g•••/ /...g../ l-g -l /...g..y /...g..y l-g -l lg...g'...l /...g ../ /g..y /...g.../(i.e. [g] of [gg]) /.. .g.../ (i.e. the second [g] of [gg]) /...g...g..y /g’..y ig -i /... g’..y
It should not be misunderstood that the phonological notation above which centres round Igl, Ig'l, /rj/ and /r)7 results from a simplistic transfer of [g], [g’], [g] and [g’]> respectively, to Igl, Ig'l, /g/ and /g’/; this would be a non-functionalist's operation based on the criterion of phonetic similarity. The implications of the phonological notation above are, functionally, that where Ig'l occurs, /g’ / does not, and vice versa, and where Igl occurs, /g/ does not, and vice versa. 10.6
Identifying 1X1, IX'I, IYI and IY ’ / in the majority speech
It is only in the consonant phoneme system of Japanese practised by the dwindling minority speakers (including myself) of standard Japanese of our day that the establishment of /g/ and /g’/ is possible and valid. For the increasing majority speakers of standard Japanese today, /g/ and /g’ / cannot be said to form part of their consonant phoneme system. It is therefore necessary to pursue our phonological analysis of Japanese by looking at that certain relevant part of the consonant phoneme system ascribable to the majority speech which differs from the corresponding section of that of the minority speech. This I intend to do below, in 10.6.1 to 10.6.5. 10.6.1
The pattern of occurrence of [g], [g’j, [rj ] and [g’ ] in the majority speech
So far as the majority speakers are concerned, partial reproduction (in 10.5.1) of Commutative series 1 to 10 (originally presented in whole in 9.1) is actually inapplicable in the sense that all the instances of [g] should be replaced by [g/[g] and all the instances of [g’j should be replaced by [g’]/[g’ ]- My notation [g]/[g] means '[g] or [g] in free variation’ and [g’]/[g’] means '[g’j or [g’j in free variation' though, so far as my personal observation goes, with a predominance of the occurrence of [g] or [g’j even in the above-mentioned free variation.84 Thus, for example, [sar g’o",oi] 'working clothes', [(kar ga"lica] 'assailant', [g a"*] 'However', [qo r oga"lkkoo] 'primary school', [o^ggaktu] 'music', as observed in the minority speech, are pronounced [sar g’o",oi]/[sar g’o",oi], [kar g'liqa)]/[kar ga",ica], [ga"l]/[ga‘1], [qor oga'1 kkoo]/[por oga',kkoo] and fo^ggaktu]/
Chapter 10: Establishing /g/, /g’/, /rj/, /g’/ , k/ and /k’/, /in the minority speech) and IXJ, /X’/, /Y/ and /Y7 (in the majority speech)
117
[o^ggakin] in the majority speech. As for a minimal pair like [se^ggo] '1,005' vs. [se^ggo] 'postwar' (valid for the minority speech), the two members of the pair are pronounced either identically, i.e. [se’1 ggo]/[se"lggo] '1,005', or [se"'ggo]/[se'1 ggo] 'postwar', in the majority speech. On the other hand, any instances of [g] or [g’j in word-initial context, or those in word-medial context in some if not all composite words, observed in the minority speech, remain unaffected in the majority speech too, so that, for example, [gar ijmuj] 'overseas trip' and [kor otooga"'kkoo] 'senior high school' (in the minority speech) are also pronounced in this way in the majority speech. What emerges, from a phonological point of view, from the occurrence of [g]/[g] and [g’]/[g’] in the majority speech is that none of /g/, Ig'l, /g/,/g’/, IkJ and /k’/ that we have seen forming part of the consonant phoneme system attributable to the minority speech have any place in the consonant phoneme system attributable to the majority speech. Notice that the question has nothing to do with the absence of /g/ and /g’ / and the presence of Igl and /g’/ in the majority speech. To amplify this statement somewhat, in the minority speech of Japanese, /g/, /g’/, /g/ and /g’/, respectively, are distinguished from each other, coupled with the fact that /k/ and /k’/ also exist in their consonant phoneme system. In the case of the majority speakers of Japanese, on the other hand, no such distinction exists as between Igl, Ig'l, /g/ and /g’/, respectively, with the attendant fact that neither Ikl nor Ik'l exist in their consonant phoneme system, either. All this will be demonstrated in 1 0 .6 .2 onward. It will be recalled that a relatively small section of Japanese speakers may consistently pronounce [g] or [g’j rather than [g]/[g] and [g’]/[g’ ]. Phonologically, such speakers' consonant system falls in step with the consonant phoneme system of the majority speech and, anyway, these Japanese speakers definitely belong to the majority speakers, evidently not to the minority speakers who are 'consistent [g]-speakers'. 10.6.2
Setting up commutative series conducive to establishing 1X1, IX ’ I, IYI and n r I
Given the different pattern of occurrence of [g], [g’j, [g] and [g’j in the majority speech indicated in the preceding section, it is necessary to modify partially, as follows, Commutative series 1 to 10 (presented in 9.1 and, subsequently, partially reproduced in 10.5.1) that we have operated with in connection with the minority speech. We need a few additional commutative series such as the following. Commutative series V to 1 0 ' are meant to include additional alternatives (applicable to the majority speech) concerning [g], [g’j, [g] and [g’j to Commutative series 1 to 10 earlier presented (applicable to the minority speech). Commutative series 11 to 20, which are presented for the first time below, apply to the majority speech only. Commutative series 6 to 10 are associated with the phonetic contexts [Vg-V], and Commutative series 16 to 20 with the phonetic context [Vg’-V]. [V] is a cover notation for any one of [i], [e], [a], [o] and [in]. Of the two [V]'s in [Vg-V] or [Vg’-V], the second [V] is indicated in the respective phonetic context associated with each of Commutative series 11 to 20, while the first [V] stands for [i], [e], [a], [o] or [i d ] in each of these commutative series. The asterisked phonetic contexts and asterisked examples are of course non-occurrent ones, as neither [gi] nor [g’ e] is permitted in Japanese. Despite their non-occurrence, they are nevertheless listed below for the sake of completeness. The dashes in [- a], [Vg - e], [Vg’ - i], etc. indicate the places where [k], [k*], [g], [g’ ], [g] or [g’], as the case may be, occurs.
118
Japanese Phonology: A Functional Approach
C o m m u ta tiv e series 1'
C o m m u ta tiv e series
[k’i -*] ’tree’ *[ki] J g P ] 'craft' *[gi]
[(h a)r ti’i 'l(cif’i)] 'g n ash ' [(h a)r g ’i", (cir’i)] 'g n ash '
Commutative series 3' r-al
2'
[ - e"1]
[-H
* [k ’e] flee-1] 'd iv in a tio n sign' * [g ’e] [ge"1] 'lo w g rad e'
*[g’e] [(a)r ge", (d o )] 'p u sh -u p d o o r' [(a)r tie"l(d o )] 'p u sh -u p d o o r'
Commutative series 4'
Commutative series 5'
r-oi
[-ID]
fk’a‘1 (kuibin)l ’leg’ [ka",(ba)] ’birch1 [g’a"l(koir’i)l 'paradox' [ga"1] 'elegance'
[k’o‘1(o)] 'today' [ko~*(i)] 'carp' [g’o(omui)] 'business' [go-1! '5'
rk’iu",(iD)l '9' [kta n] 'pain' [gui "*] 'foolishness' [g’tn’Vra)] 'cow'
[(ka)r g,a(kxu)] 'reversible reaction' [(ka)r g’a(kur)] 'reversible reaction' [ga**] 'However'/ [(ka)r tia‘1 (iqa)] 'assailant' [ga-1] 'However'/ [(ka)r ga"l(ica)] 'assailant'/
[(sa)r g’o-l(oi)] 'working clothes'
[(ka)r r]’ui(m)] 'snail'
[(sa)r g’o"'(oi)] 'working clothes'
[(ka)r g’iu(iu)] 'snail'
[(o)r tjo"l(soka)] 'austere'
[(e)r t)tu",(i)] 'acrid'
[(o)r go"l(soka)] 'austere'
[(e)r gra'l(i)] 'acrid'
Chapter 10: Establishing Igl, /g’/, /rj/, /r)’/ , k/ and /k’/, /in the minority speech) and /X/, /X’/, /Y/ and /Y7 (in the majority speech)
Commutative series 7; [- er l
Commutative series 6 ' t-ir]
[k’ir ] 'yellow (colour)' *[ki] fg’ir (ree)l 'etiquette' *[gi]
*[k’e] |'ker ntool ’boxing' *[g’e] rger (r’i)l 'diarrhoea'
[m’ir (se"1)] 'shop' *[mi] *rn’il *[ni] ft>ir (ma)l 'garden'
*[m’e] [mer (bae'1)] 'bud' *[n’ej fner (zujm’i)l 'rat' *[i>e] *rn’el •foe]
*[ni1
Commutative series 8 ' r-a r l
Commutative series 9' r-o ri
Commutative series 10' r-rarl
[k’ar (ktn)] 'guest1 [kar (ba)] 'birch' [g’ar (km)] 'inverse' [gar (ktn)] 'tablet'
[k’or (kaN)j 'very big man' [kor (oeN)] 'park' [g’or ($uiN)] 'fish-meal' [gor (?tn‘1 in)] '50'
[k’tDr (tnka)] ’holiday’ [ktur tsru] ’shoe1 [gui r (cci)] ’grumbhng' [g’tnr (tn)tool 'cowpox'
[m’ar (ktn)l 'pulse' fmar (cci)l 'town' *rn’al [nar (ta)] 'hatchet' [t>ar (a‘1 r)o)] 'miaow!'
[m’or (o?oo)] 'Venus' [mor (cci)l 'rice cake' *[n’o] [nor (okoo)l 'farming' [t>or (got]a",cima)] 'Nyogoga Shima'ss [t>or (goga\ima)] 'Nyogoga Shima'ss *tn’o] *[go]
[m’ui",(uiztu)] 'Muse' [imnr (sai)l 'celibacy' *[n’tn] [mu r (ime"1)] 'seam' [t>ror (tnr)aN)] 'spiritual awakening '86 [t>ujr (iugaN)] 'spiritual awakening '86
[T>ar (a"’go)] 'miaow!' *[g’a] *toa1
119
*[gm]
120
Japanese Phonology: A Functional Approach
Chapter 10: Establishing /g/, /g’/, /r)/, /r)’/ , k/ and /k’/, /in the minority speech) and 1X1, /X’/, /Y/ and /Y’/ (in the majority speech)
? 87
TaVg’a] 'pilgrimage' [a"Vg’a] 'pilgrimage' [og’g’a] ? 87 [og’g’a]
? 87
[uig’g’a]
? 87
frog’g’a]
? 87
10.6.3
1
[eg’g’a]
Commutative series 19 o'
Commutative series 18 rvn’ - a] [i"lg’g,a] 'cruel indecency' [i’V g ’a] 'cruel indecency' [eg’g’a] ? 87
[ir g’g’o"l(o)] ’signet' [ir g’g’o'1 (o)] 'signet' [(k)er g’g’o(o)] 'side business' [(k)er g’g’o(o)] 'side business' r(s)ar g’q’o(o)] 'industry' f(s)ar g’g’o(o)] 'industry' [(h)o-1 g’g’o(o)] 'main occcupation' [(h)o'Vg’o(o)] 'main occupation 1 [(b)rar g’g,o(o)] 'division of labour' [(b)uir g’g’o(o)] 'division of labour'
121
Commutative series 20 [Vn’ -m ] rir g’g’uj] 'pleading at a trial' [ir q’g’uj] 'pleading at a trial' [(ke)r g’g’uj(ui)] 'Altair' [(ke)r g’g’ui(ui)] 'Altair' [ag’g’mj ? 87 [ag’g’rn] ? 87 [og’g’ui] ? 87 [og’g’ui]
? 87
[mg’g’ui]
? 87
[ui’gg’m]
? 87
Defining 1X1, /X’/, IYI and /Y’ / in terms of relevant features
It should first be recalled that in establishing /g/, /g’/, /g/ and /g’/ in the minority speech, I mentioned at the end of 10.5.2 that the phonetic contexts [- i] and [- e] (hence, Commutative series 1, 2, 6 and 7) are, as we shall see in the course of 11.3, contexts of neutralization for Igl - Ig'l and /g/ - /g’/ and are therefore irrelevant and unhelpful to establishing /g/, Ig'l, /g/ and /g’/. Furthermore, as again we shall see also in the course of 11.3, the phonetic contexts [- i] and [- e] are contexts of neutralization for the four consonant phonemes which we shall establish in the consonant phoneme system ascribable to the majority speech and which we shall notate as IXJ, IX'I, IYI and fY 'l below. The readers are requested to have another look at Commutative series V to 10' presented in 10.6.2. The contents of Commutative series 1', 2', 6 ' and 7' are to be left out of account again. Only the contents of Commutative series 3', 4 ', 5', 8 ', 9' and 10 ' form the object of our examination in this section. The phonological irrelevance of the difference between [g] and [g’j, and between [g] and [g’J, before [a], [o] or [ra], in word-medial context and word-initial context, is easily confirmed from examples like [(ka)r g’a(ktu)]/[(ka)r g’a(ktu)] 'reversible reaction’ (cf. Commutative series 3'), [(ka)r ga",(ipa)]/[(ka)r gan(ipa)] 'assailant1 (cf. Commutative series 3'), [(sa)r g’o",(oi)]/[(sa)r g’o",(oi)] ’working clothes' (cf. Commutative series 4'), [(o)r go",(soka)]/[(o)r go'1 (soka)] 'austere' (cf. Commutative series 4'), [(ka)r g’ra(nj)]/ [(ka)r g’m(ra)] 'snail' (cf. Commutative series 5'), [(e)r gui",(i)]/[(e)r gtu",(i)] 'acrid' (cf. Commutative series 5'), [T>ar (a"lgo)]/[T>ar (a"lgo)] ’miaow!' (cf. Commutative series 8 '), [T>or (goga",cima)]/[T>or (goga‘,gima) ] 88 'Nyogoga Shima' (cf. Commutative series 9') and [ga",]/[ga'1] 'However' (cf. Commutative 3'). Notice that there is only one example of the phonological irrelevance of the difference between [ga"1] and [ga"1] in wordinitial context, i.e. [ga",]/[ga’1] ’However'. Apart from the above-mentioned examples, numerous examples can actually be adduced of words which have [g] or [g’ ] in wordmedia/ context. To give just a few more such examples, [cor ogak"lkoo]/[gor ogak",koo]
122
Japanese Phonology: A Functional Approach
'primary school', [ri,uir uig,unn]/[ri.txir txig’iuui] 'milch cow', [har g’g’akui]/ [har g’g’aktu] 'rebellion', and [ha‘1ge]/[ha"lge] 'baldness'. The list could be easily lengthened. Add to this a pair of words like [se^ggo]/ se^ggo] in the majority speech for both the word meaning '1,005' and the word meaning 'postwar', as this also manifests the phonological irrelevance of the difference between [g] and [g]. The phonological irrelevance of the difference between [g] and [g’j, and between [g] and [g* ], before [i] and [e] - this occurs in word-medial context only - can also be seen in the contents of Commutative series 1', 2', 6' and 7', and consequently examples like [(ha)r g’i “l(pir’i)]/[(ha)r g’i",(pir’i)] 'gnash' (C om m utative series 1 '), [(a)r ge'l(do)]/[(a)r ge',(do)] 'push-up door' (cf. Commutative series 2') are left out of account just at this juncture for the reason that this would take us to one of the instances of neutralization which I shall treat in detail in 11.10. On the basis of the evidence found in Commutative series 3', 4 ', 5', 8', 9', 10' and Commutative series 6 to 15 all of which have been presented in 10.6.2, and also that found in Commutative series 3 to 5 and 8 to 10 earlier presented in 9.1, it is possible to establish four consonant phonemes which I conveniently notate as IXJ, /X’/, IYI and /Y’/ and which form part of the consonant phoneme system ascribable to the majority speech. The establishment of IXJ, IX'I, IYI and IY'1 is a necessity consequent, in the majority speech, on the phonological irrelevance of the difference between [g] and [g’j, and that between [g] and [g’ ], unlike in the minority speech. Here are IXJ, IX'I, IYI and IY'1 defined in terms of relevant features. IXJ: IX 'I: IYI: IY 'I:
"voiced" (vs. IYI), "non-palatalized" (vs. IX'/), "dorsal" (vs. e.g. Ibl). "voiced" (vs. IY'I), "palatalized" (vs. IXJ), "dorsal" (vs. e.g. lb'I). "voiceless" (vs. IXJ), "non-palatalized" (vs. IY'I), dorsal" (vs. e.g. Ipf). "voiceless" (vs. IX'f), "palatalized" (vs. IYI), "dorsal" (vs. e.g. /p’/).
Here now is a convenient summary of the phonological contents of IXJ, IX'/, IYI and IY'I deriving from the above. IXI: IX'/: IYI: IY 'I:
"voiced non-palatalized dorsal". Valid before /a/, Jol and /ml. "voiced palatalized dorsal". Valid before laJ, Jo/ and /in/. "voiceless non-palatalized dorsal". Valid before /a/, loJ and /ml. "voiceless palatalized dorsal". Valid before /a/, Jo/ and /ml.
Notice that IXJ and Ibl, IX'I and lb' I, IYI and /p/, and IY'I and Ip'I, are two by two all direct neighbours in the consonant phoneme system ascribable to the majority speech, in an analougous sense that Ibl and ImJ, and /p/ and /m/, are direct neighbours, and lb'I and Ira'I, and Ip' / and /m’/, are also direct neighbours in the same consonant phoneme system. Just as the distinction between "voiced" and "voiceless", which is relevant to /p/, Ibl, Ip'I and lb'I, is extraneous to Ival or /m’/, so the distinction between "nasal" and "non-nasal", which is relevant to /p/, Ibl, Ip'I and lb' I, is extraneous to IXJ, IX'I, IYI and IY'I. It will now be understood why I deliberately use symbols like X, X’, Y and Y’ to indicate the four consonant phonemes in the majority speech as IXJ, IX'I, IYI and IY'I whose phonological contents are, respectively, "voiced non-palatalized dorsal", "voiced palatalized dorsal", "voiceless non-palatalized dorsal" and "voiceless palatalized dorsal", respectively. It would of course be downright wrong to indicate Igl instead of IXJ, Ig'l instead of IX'I, Ikl instead of IYI and /k’/ instead of IY'I, as the phonological contents of Igl ("voiced non-palatalized dorsal non-nasal") and IXI ("voiced non-palatalized
Chapter 10: Establishing /g/, /g7, /rj/, /g 7 , k/ and /k’/, /in the minority speech) and IXJ, fX 'l, IYJ and /Y7 (in the majority speech)
123
dorsal"), those of /g7 ("voiced palatalized dorsal non-nasal") and /X’/ ("voiced palatalized dorsal"), those of IkJ ("voiceless non-palatalized dorsal non-nasal") and H I ("voiceless non-palatalized dorsal"), and those of /k’/ ("voiceless palatalized dorsal nonnasal") and /Y7 ("voiceless palatalized dorsal"), respectively, are not the same, /g/, /g’/, IkJ and /k7 are four of the phonemes found in the consonant phoneme system ascribable to the minority speech, but not in the consonant phoneme system ascribable to the majority speech. 10.6.4
Correlative structure of IXJ, IX '/, H I and H ' /
/X/./X7, /Y/and/Y’/, as I have defined in 10.6.3 in terms of relevant features, form a correlation which is shown below.89 "dorsal" "voiced" "voiceless" 10.6.5
"non-palatalized" IX/ A7
"palatalized" IX'I /Y7
Realizations of IXJ, IX 'I, H I and H ' /
Realizations of IXJ, IX'I, H I and/Y7 are as follows. (Recall that these phonemes are valid before /a/, lol or An / only.) I provide examples of the phonetic forms of a few words in which these realizations occur. (1) IXJ is realized: (i) by [g] (voiced non-palatalized dorso-velar plosive) and [g] (voiced nonpalatalized dorso-velar nasal) in free variation in word-medial intervocalic context (e.g. [ha"*ge]/[ha",ge]/ 'baldness') or at the beginning of the non-initial constituent(s) within a composite word (e.g. [por ogak",koo]/[por ogak",koo] 'primary school' (< [pono] 'small' + [gar kkoo] 'school')); (ii) also by [g] and [g] in free variation in post-[g] context (which is always also in word-medial context) (e.g. [har ggakui]/[har ggakui] 'half price' (< [ha^N] 'half + [ga^kui] 'amount')); (iii) by [g] in word-initial context (e.g. [gor ma] 'sesame') and at the beginning of the first constituent within a composite word (e.g. [ga'Mdokm] 'harm' (< [ga"*i] 'damage' + [dor kuj] 'poison')); and (iv) also by [g] at the beginning of the non-initial constituent(s) within a composite word (e.g. [ker ezaiga"liroN] 'introductory economics' (< [kenezai] 'economics' + [ga^ifoN] 'general outline')). (2) IX7 is realized: (i) by [g’j (voiced palatalized dorso-velar plosive) and [g’ ] (voiced palatalized dorso-velar nasal) in free variation in word-medial intervocalic context (e.g. [sa'1g’oo]/[sa",g’oo]/ 'work' or at the beginning of the non-initial constituent(s) within a composite word (e.g. [ha‘1g’oo]/[ha"lg’oo] 'ha-row' (< [ha"1] 'the Japanese kana character for [ha]' + [g’o"^] 'row')); (ii) also by [g’] and [g’] in free variation in post-[g’] context (which is always also in word-medial context) (e.g. [har g’g’akux]/[har g’g’akui] 'rebellion' (< [ha^N] 'anti-') +
124
Japanese Phonology: A Functional Approach
[gakux] 'opposite')); (iii) by [g’] in word-initial context (e.g. [g’ o^oko] ’congelation’) and at the beginning of the initial constituent within a composite word (e.g. [g’ar kuise",ndeN] ’counter-propaganda’ (< [g’ar kui] ’inverse’-!- [ser ndeN] ’propaganda’)); and (iv) also by [g’] at the beginning of the non-initial constituent(s) within a composite word (e.g. [gar kkoog’o",osee] ’school administration’ (< [gar kkoo] ’school’ + [g’or osee] ’administration’)). 3) /Y/ is reahzed; (i) by [k] (voiceless non-palatalized dorso-velar plosive) in word-initial context (e.g. [koToba"1] 'word'), in word-medial intervocalic context (e.g. [nor kor’ i"1] 'remainder'), at the beginning of the first constituent within a composite word (e.g. [kar teese"lekatsru] 'home life' (< [kar tee] 'home' + [ser ekatsxu] 'life')), and at the beginning of the non initial constituent(s) within a composite word (e.g. [te",krab’i] 'wrist' (< [te"1] 'hand' + [kixir b,i] ’neck’)); (ii) also by [k] in post-[g] context (e.g. [i^gko] 'parakeet'). (4) /Y7 is realized: (i) by [k’] (voiceless palatalized dorso-velar plosive) (e.g. [k’o r obai] 'auction', [cjr k’umi] 'womb', [k,ar kuiaq(i)] 'clientele' (< [k’ar kuj] 'customer' + [ar ci""] 'foot/leg'), [ger enook’o‘1ktu] 'entertainments department’(< [ger enoo] 'public entertainments' + [k’or kuj] 'department'); (ii) also by [k’J in post-[g’] context (e.g. [se"Vk’o] 'election'). Realizations of /Y/ and /Y 7 are straightforward in that in any of the various contexts specified, the former phoneme is always realized by [k’J and die latter by [k]. It will also have been seen that what is phonologically relevant is the difference between [g] and [g] on the one hand and [k] on the other, and also the difference between [g] and [k]. What phonologically matters here is the distinction between "voiced" and "voiceless", while "nasal" or "non-nasal" is extraneous. The same can be said of [g’], [g’] and [k’]. I earlier mentioned in note 72 that a relatively small number of Japanese speakers may consistently pronounce [g] and [g’] (e.g. [ka^gru] 'furniture', [ka^ggo] 'nursing' or IjzPg’oo] 'business') and said further, at the end of 10.6.1, that these Japanese may pronounce [g] and [g’] rather than [g]/[g] and [g’]/[g’], adding that phonologically these people's speech belongs to the majority speech, not to the minority speech. The consonant phoneme system ascribable to the speech of those who consistently pronounce [g] and [g’] (rather than [g]/[g]) and [g’]/[g’J has also fXJ, /X7, /Y/ and /Y7, which are realized by [g], [g’j, [k] and [k*], respectively. This of course differs from realizations of /X/, /X7, /Y/ and /Y7 listed above, but from a phonological point of view, despite such phonetic differences, it is of minimal interest and relevance whether individual majority speakers pronounce [g]/[g] and [g’]/[g’], or [g] and [g’j. 10.7
Consequence of the establishment of /X/, /X 7, /Y/ and /Y’ / on phonological notation
Just as I showed in 10.5.6 the consequence of the establishing /g/, /g7, /g/ and /g7 in the consonant phoneme system ascribable to the minority speech, I shall show below what the consequence is on phonological notation following the establishment of /X/,
Chapter 10: Establishing /g/, /g’/, /g /, /r)’/ , k/ and /kV, /in the minority speech) and IXJ, /X’/, A7 and /Y7 (in the majority speech)
125
IX'/, /Y/and/Y7 in the consonant phoneme system attributable to the majority speech. Some relevant example words, taken from among those already given in 10.5.5, will be cited again below to indicate their pronunciations in accordance with the majority speech. The glosses are omitted but can be recovered by the readers by referring back to 10.5.5. [or ogama]/[or ogama] or [or ogama]90 [or ogama]/[or ogama] or [or ogama]90 [kor gaN]/[kor gaN] or [kor gaN] [cor ogankkoo]/[cor oga",kkoo] or [por oga"'kkoo] [kor otooga'1kkoo]91 [gar kkoog’o nosee]92 [a^igoj/ta^igo] or [a'Mgo] [ga’1]/[ga’1] or [ga-1] [se^ggo]93 [se"lggo]/[se"1ggo] or [se^ggo]94 [T>or goga\im a]/[t>or goga’,pima] [g’ar kuir’i] [gar kro] [kar g’aktn:]/[kar g,akuj:] or [kar g’akxu]
/.. .X.../ /.. .X.../ /.. .X.../ I...X..J /...X.../ [X...X'..J /...X...I IX...I 1.. .X...I (i.e. [g] of [gg]) 1.. .X ...I(i.e. the second [g] of [gg]) 1.. .X...X...I r x \.j IX...I 1.. .X '..J
I add the following few example words in reference to the occurrence of IYI and /Y7. [kanba] (taken from 10.5.1) [k,a",kcubui] (taken from 10.5.1) [nor okoo] (taken from 10.5.1) [se'V k’o] (taken from 10.6.5)
IY..J IY '..J /.. .Y.../ I...Y' ..J
In the case of the minority speech, these last example words will of course be phonologically notated as lk...l and Ik' ...I, respectively. 10.8
1X1, !X'I, IYI and /Y’ / in the majority speech and Igl, Ig'l, /g/ and /g’ / in the minority speech
The readers will have seen that none of IXJ, IX'I, IYI and IY' I which have been established in connection with the majority speech corresponds to any of the consonant phonemes established for the minority speech. That Igl, Ig'l, /g/ and /g7 are ascribable to the minority speech can be readily confirmed by comparing the phonological contents of 1X1, IX'I, IYI and IY'1 with those of the consonant phonemes ascribable to the minority speech, as shown in 9.4. The "dorsal" order in the consonant phoneme system ascribable to the minority speech consists of IkJ, Ik'I, Igl, Ig'l, /g/ and /g7 as well as /uj/ and IU. while the "dorsal' order in the consonant phoneme system ascribable to the majority speech95 consists of IXJ, IX'I, IYI and IY 'I as well as /tg/ and Ikl. It is only in the "dorsal" order that the two different consonant phoneme sub-systems differ from each other. 10.9
The phonological implications of IXJ, IX'I, IYI and IY' / in the majority speech
We now turn our attention to how the use in the majority speech of IXJ, IX'I, IYI and
126
Japanese Phonology: A Functional Approach
/Y7 can be an alternative to the use of /g/, /g/,/g7,/g’/, Dd and /k7 in the minority speech. In word-initial context, the distinction between [g] and [k] or between [g’] or [k’] is binding in the majority speech just as much as in the minority speech. In other words, it is essential to all speakers of Japanese that in word-initial context, [g] and [k] should be distinguished from each other, and so should [g’j and [k’ ]. In fact there are many pairs of words which are minimally differentiated from each other through the difference between [g] and [k] or between [g’j and [k’ j. Here are some examples: [ga^kui] ’amount' vs. [ka^kui] 'nucleus'; [gor ocoo] 'tycoon' vs. [kor oqoo] 'negotiation'; [gui",zm] 'laggard' vs. [kui‘,zin] 'detritus'; [g’ar kui] 'inverse' vs. [k’ar kui] 'guest'; [g’o"1?^] 'fish manure' vs. [k’o"1?^] 'refusal'. What is phonologically crucial here is the distinction between voicedness (in [g] and [g’]) and voicelessness (in [k] and [k’ ]). In word-initial context, the difference between [g] and [g] is minimal in the majority speech in that the sole case of the utterance-connective adversative particle I have earlier mentioned on a few occasions, which is pronounced [ga-1] in the minority speech, is pronounced [ga",]/[ga'1] or even [ga-1] in the majority speech. If, in the majority speech, the vital distinction between voicedness and voicelessness in [g] and [k], or [g’j and [k’j, in word-initial context, is made to apply to the fullest extent to word-medial context as well, nothing would prima facie prevent the occurrence of [g] or [g’j to the exclusion o f [g] or [g’ ] because this would leave the necessary distinction between [g] and [k] or between [g’j and [k’j intact (e.g. [sa"lgaN] 'sandstone' vs. [sa’ kaN] 'plasterer'; [?ir g’oo] 'commencement of work' vs. [pjr k’oo] 'market conditions'; [ka"lggo]/[ka"lggo] 'nursing' vs. [ka^gko] 'ovation').96 This is what actually happens in the speech of a small portion of the majority speakers. However, for other, more numerous, majority speakers, what happens is that not only [g] or [g’j but also (for some reason or other) [g] or [g’j occur in such a way that [g] and [g], and [g’j and [g’ ], occur in free variation (e.g. [sa",gaN]/[sa"'gaN] 'sandstone' vs. [sa^kaN] 'plasterer'; (pi r g’oo]/[cir g’oo] 'commencement of work' vs. [cjr k’oo] 'market conditions').97 In this case, [g]/[g] are distinctively different from [k], and [g’]/[g’j from [k’j, which means of course that not only are [g] and [k], or [g’j and [k’j, distinctively different from each other as voiced vs. voiceless, but [g] and [k], or [g’j and [k’j, are distinctively different from each other as voiced vs. voiceless, not as nasal vs. non-nasal. This is why the relevant feature "voiceless" is possessed by IYI (realized by [k]) and /Y7 (realized by [k’]) and "voiced" is possessed by IXI (realized by [g]/[g]) and /X’ / (realized by [g’]/[g’ ]). All this is compatible with the fact that [g]/[g] is a realization of /X/ ("voiced non-palatalized dorsal"), [k] a realization of /Y/ ("voiceless non-palatalized dorsal"), [g’]/[g’] a realization of /X7 ("voiced palatalized dorsal") and [k’ ] a realization of /Y7 ("voiceless palatalized dorsal"). The difference between nasality and non-nasality is phonologically irrelevant when in combination with dorsality and the presence or absence of palatalization. This is why none of /X/, /X7, /Y/ and /Y7 has the relevant feature "nasal" or "non-nasal". This represents an economy of two articulations, nasality and non-nasality (this is why I see the present case as an instance of phonological catalysis; cf. 14.3.2), without detracting from the efficiency of verbal communication in Japanese. We confirm that the vital distinction between voicedness and voicelessness in [g] and [k], or [g’j and [k’j, in word-initial context is made to apply to the fullest extent to word-medial context as well. The presence of IXJ, fX 7, IYI and IY' I in the consonant phoneme system ascribable to the majority speech is compatible with the extension of the opposability between voicelessness and voicedness (when in combination with dorsality and the presence or absence of palatalization) obligatorily in force in wordinitial context to word-medial context as well. The efficiency of verbal communication in the majority speech is equal to that in the minority speech both in word-initial context and word-medial context. Even for the
Chapter 10: Establishing /g/, Ig ’ l, /g/ , /g’/ , k/ and /k’/, /in the minority speech) and IXJ, D C /,/Y / and /Y ’ l (in the majority speech)
127
insignificant pairs of words (listed in 10.5.5) for which the distinction between fg] and [g] is said to serve phonologically (cf. /g/ - /g/) in the minority speech, it is hardly likely that any serious inconvenience arises in the majority speech in which the distinction between [g] and [g] is eliminated.98 The somewhat irregular occurrence of either [g] (e.g. [kor otooga'1kkoo] 'senior high school') or [g] (e.g. [por oga"lkkoo] 'primary school') in the minority speech is replaced in the majority speech by [g]/[g] or [g] in a uniform fashion. Quite apart from phonetic and phonological facts and considerations, it is instructive to know that the Japanese (with the exception of phoneticians and phonologists), both the minority and majority speakers, do not see the question of /g/, Ig’l, /g/, /g’/, and IXJ, IX’I, IYI and fY ’l, in the light I have been describing. The average Japanese speaker sees or thinks of one and the same relevant kana character, e.g. z" and is unaware of different consonant segments (together with [o] which follows) that are pronounced by different Japanese speakers, i.e. [go], [go], [go]/[go], in different contexts. Such phonetic differences are masked by the use of the same kana character in Japanese writing system, as I have already indicated in note 78. What the average Japanese speaker is aware of and knows as being important is that Z* is distinct from Z which always corresponds to [ko]. The same can be said, mutatis mutandis, of ft", tf", ], [n], [g’j, [g], [f’] and [r], (Notice that [t> -] is germane to its occurrence before [i], [a], [o] and [tn], while [n -] is germane to its occurrence before [e], [a], [o] and [to].) I reproduce below that part of Commutative series 1, 2, 6 and 7 (earlier presented in 9.1) which pertain to [p’j, [p], [b’j, [b], [g’j, [g], [k’j, [k], [m’j, [m], [i>], [n], [g’j, [g], [r’] and [r] only (omitting the rest) occurring before [i] and before [e]. Commutative series 1 r - i -1! [p’i"'(ikxu)] 'summit' *[pi] [b’i"1! 'beauty' *[bi]
Commutative series 2 t-e " 1] *[p’e] [pe"'(ke)l 'no good' *[b’e] fbe^lddo)! 'bed'
Chapter 11: Neutralization of oppositions between the consonant phonemes of Japanese
[k’i"1] 'tree' *[ki] [g’i"1] 'craft' *[gi]
[r’i"1] 'advantage' *[ri] [(sar n’)l’i(Vea)] 'tricycle' *[li]
*[k’e] [ke"1] 'divination sign' *[g’e] [ge*1] 'low grade'
*[r’e] [re"1(baa)] 'liver (of animal)' *[l’e] [le^Cbaa)] 'liver (of animal)'
*[m’e]
[m’i"1] 'winnow' *[mi] *[n’i] *[ni]
[me"1] 'eye' *[n’e] [ne’1] 'root (of tree)'
[T>i~"]' 'two' [(ha)r ti’i",(cir’i)] 'gnash' *[giJ
*[t>e] *[f)’e] [(a)r ne",(do)] 'push-up door'
Commutative series 6 b irl rp’i r (ano)] 'piano' *[pi] rb’ir (koo)l 'nostril'
Commutative series 7 [-e rl *[p’e] rper (r’ikaN)l 'pelican' *[b’e]
*fbil
/N/
fber (Vpoo)l 'compensation'
[k’ir ] 'yellow (colour)' *[ki] [g’ir ree] 'etiquette' *[gi]
*[k’e] [ke(r ntoo)] 'boxing' *[g’e] [ger (r’i)] 'diarrhoea'
[r’ir (ppkur)] 'wrath'
*[r’e] [rer (ekai)] 'regular meeting' *[l’e] [ler (ekai)] 'regular meeting'
*[ri] R’ir l
131
132
Japanese Phonology: A Functional Approach
*[m’e]
[m’i^se"1)] 'shop' *[mi] *[n’i] *[ni] [T>ir (tqa)] 'garden' [(ka)r n’i',1 'key' *[Qi]
[mer (bae"1)] 'bud' *[n’el [ner (zium’i)] 'rat' *[i>e] *[g’e] [(a)r ge"Viu)] '(to) raise'
I emphasize that the occurrence of these consonant segments need not be limited to word-initial context. They occur in word-medial context as well. The present discussion about the neutralization of oppositions between consonant phonemes therefore apply to both word-initial and word-medial context. All the contexts being considered for the neutralization in question are before I'll, Id, /a/, /o/ and /m l. Note that when the neutralization of /p/ - /p’/, /b/ —/b’ /, /k/ —Ik'l, Igl - /g’/, /ml - /m7, Id - Id I, /g’/ - /rj/, and /r/ - /r’ / occurs in word-medial context, it occurs within both simple words and composite words. My conclusion, drawn in 8.7, that Id is systematically non-occurrent after [k’] and liJ is systematically non-occurrent after [k], would equally apply on the basis of the data which are to be found in the other commutative series, so that Id is systematically nonoccurrent after [p’], [b’], [g’], [m’], [t>], [rj’] and [r’], while HI is systematically nonoccurrent after [p], [b], [g], [m], [n], [g] and [r]. It is apparent that [k] and [k’] occurring before [a], [o] and [tu ] (identified as realizations of /a/, lol and /ml, respectively, in Chapter 8) are realizations of two different distinctive units, i.e. Ikl and Ik'l, respectively, definable, as I have already indicated, as "voiceless non-palatalized dorsal non-nasal" and "voiceless palatalized dorsal non nasal". However, before [i] where we find [k’i] but not [k’e], and before [e] where we find [ke] but not [ki], the distinction between Ik'l and Ikl is not valid. If the distinction were to be valid, we should find both [k’i] and [ki] and both [k’e] and [ke], which is not the case. The distinctive unit that occurs before [i] is then one that is extraneous to the distinction between non-palatalization (as in [k]) and palatalization (as in [k’]) but retains the sum of "voiceless", "dorsal" and "non-nasal", i.e. the common base of IkJ and Ik'l. As this common base is exclusive to Ikl and Ik' I in the Japanese consonant phoneme system, it is nothing but the archiphoneme Ik-k'l definable as "voiceless dorsal non-nasal" which is associated with the neutralization of Ikl - Ik'l with the cancellation of the distinction between "non-palatalized" (of Ikl) and "palatalized” (of Ik'l). [i] in [k’i] can be considered as a realization of HI, as [a], [o] and [in] also occur after [k’], if not [e]. We saw in 8.7 that /i/ - Id is not a neutralizable opposition. This is compatible with our earlier analysis that after [k’], Id is non-occurrent but laJ, lol and Iml are, and also with the fact that the phonetic quality of [i] in [k’i] is not inherently dictated by the occurrence of [k’] before it. That [i] in [k’ i] has an independent phonetic quality is demonstrated by the fact that [a], [o] and [m] also occur after [k’]. The archiphoneme Ik-k'l is realized by [k’] before /il, one of the two contexts of neutralization for IkJ - Ik'l. 'Before Id , 'before /o/1and 'before Iml' are contexts of relevance for Ikl - Ik'l. The same analysis applies, mutatis mutandis, to the distinctive unit that occurs before [e] in [ke] and is extraneous to the distinction between non-palatalization (as in [k]) and palatalization (as in [k’]) but retains the sum of "voiceless", "dorsal" and "non-
Chapter 11: Neutralization of oppositions between the consonant phonemes of Japanese
133
nasal", i.e. the common base of IkJ and /k’/ that is exclusive to these two phonemes in the Japanese consonant phoneme system. Before Id , there takes place the neutralization of IkJ - /k’/, the distinction between "non-palatalized" and "palatalized" being cancelled, and the archiphoneme /k-k’/ definable as "voiceless dorsal non-nasal" occurs before Id. The archiphoneme /k-k’/ is realized by [k] before /e/, which is the other of the two contexts of neutralization for /k/ - Ik'/. As already mentioned, 'before /a/', 'before /o/’ and 'before /ui/' are contexts of relevance for IkJ - /k’/. [e] in [ke] is a realization of Id, as [a], [o] and [ui] also occur after [k], if not [i], and we have seen in 8.7 that lil - Id is not a neutralizable opposition. The non-palatalization in [k] is not dictated by the occurrence before [e], since [a], [o] and [xu] also occur after [k], [k] and [k’] are combinatory variants of one and the same archiphoneme /k-k’/, that is to say, [k’] when this archiphoneme occurs before III and [k] when this archiphoneme occurs before Id. Likewise, the same phonological analysis applies, mutatis mutandis, to [pe], [p’i], [be], [b’i], [ge], [g’i], [me], [m’i], [ne], [T%i], [ge], [rj’i], [re] and [r’i]. Just as in the case of IkJ - Ik'I, the contexts of neutralization for all of /p/ - /p’ /, Ibl - lb'I, Igl - /g’/, /ml - /m’/, Id - In'/, /rj/ - /g’/ and I d -Ic'l are 'before /i/' and 'before Id', while the contexts of relevance are 'before /a/’, 'before /o/' and 'before hii /'. Each of these phonological oppositions is neutralized in the contexts of neutralization consequent on the cancellation of the opposition between "non-palatalized and "palatalized", while the remaining relevant feature(s) of both terms of each neutralizable opposition, i.e. the common base of both terms, i.e. the archiphoneme, stays relevant. Each of the archiphonemes associated with the neutralization of the abovementioned phonological oppositions, viz. /p-p’/ ("voiceless labial non-nasal"), /b-b’/ (voiced labial non-nasal"), /g-g’/ ("voiced dorsal non-nasal"), /m-m’/ ("labial nasal"), In-dl ("apical nasal"), /g-g’/ ("dorsal nasal") and /r - r ’/ ("liquid") as well as /k-k’ / ("voiceless dorsal non-nasal") is opposable, before HI or Id, not only to each other but also to each of those other distinctive units (all being phonemes) which are all susceptible of occurring before III or before Id, as well as certain other contexts, viz. /$/ ("labial fricative"), / q/ ("palatal fricative"), /q/ ("voiceless hush"), /?/ ("voiced hush") and led ("palatal affricate"). Whether or not the afore-mentioned seven archiphonemes are also opposable, before III or Id, to Is/ ("voiceless hiss"), /z/ ("voiced hiss"), Its/ ("apical affricate"), /j/ ("palatal spirant") and Id ("dorsal spirant") and Aq/ (also "dorsal spirant") will be examined later. Here below is a summary of all the instances of the neutralization of oppositions between consonant phonemes mentioned so far, with relevant information about the associated archiphonemes and their realizations. /p/ ("voiceless non-palatalized labial non-nasal") - /p’/ ("voiceless palatalized labial non-nasal") is a neutralizable opposition, valid before /a/, lol and /ml. It is neutralized before lil and Id. The archiphoneme /p-p’/ ("voiceless labial non-nasal") is realized by [p’] (voiceless palatalized bilabial plosive) before lil and by [p] (voiceless nonpalatalized bilabial plosive) before Id. Ibl ("voiced non-palatalized labial non-nasal") - fb'l ("voiced palatalized labial non nasal") is a neutralizable opposition, valid before Id, lol and Ixa I. It is neutralized before Id and Id. The archiphoneme /b-b’/ ("voiced labial non-nasal") is realized by [b’ ] (voiced palatalized bilabial plosive) before hi and by [b] (voiced non-palatalized bilabial plosive) before Id. IkJ ("voiceless non-palatalized dorsal non-nasal") - /k’/ ("voiceless palatalized dorsal non-nasal") is a neutralizable opposition, valid before Id, lol and ha I. It is neutralized before lil and Id. The archiphoneme /k-k’/ ("voiceless dorsal non-nasal") is realized by [k’] (voiceless palatalized dorso-velar plosive) before lil and by [k] (voiceless nonpalatalized dorso-velar plosive) before Id.
134
Japanese Phonology: A Functional Approach
/g/ ("voiced non-palatalized dorsal non-nasal") - /g’/ ("voiced palatalized dorsal non-nasal") is a neutralizable opposition, valid before lal, lol and /m l. It is neutralized before III and Id. The archiphoneme /g-g’/ ("voiced dorsal non-nasal") is realized by [g’J (voiced palatalized dorso-velar plosive) before III and by [g] (voiced non-palatalized dorso-velar plosive) before Id. Iml ("non-palatalized labial nasal") - /m’/ ("palatalized labial nasal"), is a neutralizable opposition, valid before /a/, lol and /ui /. It is neutralized before III and Id. The archiphoneme /m-m’/ ("labial nasal") is realized by [m’] (voiced palatalized bilabial nasal) before l'il and by [m] (voiced non-palatalized bilabial nasal) before Id. Ini ("non-palatalized apical nasal") - /n’/ ("palatalized apical nasal") is a neutralizable opposition, valid before /a/, lol and /in /. It is neutralized before I'll and Id. The archiphoneme /n-n’/ ("apical nasal") is realized by [t>] (voiced laminodorsoalveolopalatal nasal) before III and by [n] (voiced non-palatalized apico-dentialveolar nasal) before Id. Iql ("non-palatalized dorsal nasal") - /g ’/ ("palatalized dorsal nasal") is a neutralizable opposition, valid before /a/, lol and Iml. It is neutralized before III and Id. The archiphoneme /g-g’/ ("dorsal nasal") is realized by [if ] (voiced palatalized dorso-velar nasal) before III and by [g] (voiced non-palatalized dorso-velar nasal) before Id. /r/("non-palatalized liquid")-/r’/ ("palatalized liquid") is a neutralizable opposition, valid before /a/, lol and Iml. It is neutralized before I'll and Id. The archiphoneme /r-r’/ ("liquid") is realized by [r’]/[l’] (voiced palatalized apico-alveolar tap/lateral) before lil and by [r]/[l] (voiced non-palatalized apico-alveolar tap/lateral) before I d ." It goes without saying that in the context 'before /if, the archiphonemes /p-p’/, lbb’/, /k-k’/, /g-g’/, /m-m’/, /n-n’/, /g-g’/an d /r-r’/ are opposable to not only each other but also those consonant phonemes that also occur 'before /if, namely /t/, Id , /$/, /?/, Id, /ts/, /cp/ and /?/ (but not Ihl, Is/, /z/, /j/ and /rq/(/a/) which do not occur 'before lil’). Likewise, in the context 'before Id', the above-mentioned archiphonemes are opposable to not only each other but also those consonant phonemes that also occur 'before Id', namely lil, Idl, /$/, Ihl, Id, IsJ, Id, /ts/, /cp/, /?/, (but not /?/, 1)1, /xq/(/a/) which do not occur 'before Id ). It is the above-mentioned opposabilities 'before lil' and 'before Id' that serve to establish and confirm the phonological contents of the above-mentioned archiphonemes, i.e. /p-p’/, Ib-b'l, /k-k’/, /g-g’/, /m-m’/, /n-n’/, /g-g’/ and /r-r’/. 11.4
The neutralization of lil - /ts/?, /d/ - /z/?, /cp/ - It/?, /? / - /z/?, Id - Is/?, /? /-/$ /? , and /?/ - /h/?
The reason why I have put an interrogative mark for each of the above phonological oppositions is that phonological analyses of Japanese by a number of scholars seem to suggest that these phonological oppositions can be interpreted as neutralizable oppositions in functional terms. Whether I agree with such a suggestion will be indicated below. We saw in Commutative series 5 presented in 9.1 that [to ] is unattested but that [tsui] is attested. It might be thought at first sight that It! occurs before Iml and is realized by [ts], hence [tsui]. But it should be recalled that we have established /ts/ definable as "apical affricate", which occurs before Id, lal, lol and /m l and is realized by [ts]. This might suggest that there occurs the neutralization of III - /ts/ before Iml, the associated archiphoneme /t-ts/ being realized by [ts].100 But what is the common base of /t/ and /ts/? The common base of "voiceless apical non-nasal" {lil) and "apical affricate" (/ts/) is "apical", but lil and /ts/ are not the only consonant phonemes of Japanese to possess "apical" {Id, Ini and /n’/ also possess "apical"), which makes lil - /ts/ a non-exclusive opposition, hence a non-neutralizable opposition. The conclusion is therefore that while /ts/ occurs before Iml (as well as before Id, lal and lol), lil occurs not
Chapter 11: Neutralization of oppositions between the consonant phonemes of Japanese
135
only before Id, Id and lol where /1/ is realized by [t] but also before /ui/ (and HI) where /t/ is realized by [t’]. It should be added straightaway that the occurrence of /t/ before /m l and III is limited to the rendition of loanwords (e.g. [t’i"'(mpaT>i)] 'timpani', [t’m', (mba)] 'tuba'). On other hand, It! is non-occurrent before Iml and I'll in non-loanwords. What has been said above about [ts] (the realization of /ts/ before Id, Id and Iml) should be compared with [dz], phonetically the voiced counterpart of [ts] which occurs before [e], [a], [o] and [in], but not before [i]. I shall discuss this point below in order to show that [ts] and [dz] before [uj ] is functionally not comparable with each other. We have seen that Idl ("voiced apical non-nasal") occurs before Id, Id and lol and is realized by [d], and before Iml and is realized by [d’]. [d’tu] is attested in Commutative series 5. It is reminded that [d’m] only occurs in a widespread pronunciation of loanwords from European languages but not otherwise. On the other hand, [dui ] is unattested. But we see that in Commutative series 5, [dzui] is attested. Are we to analyze [dzuj] as a realization of /d/ before Iml or is [dziu] linked to the neutralization of some phonological opposition? Given that we have [dzui] instead of [tsra], the point in hand might seem at first glance parallel to the point discussed above as to whether /t/ —/ts/ is neutralizable before Iml. But actually, as we shall see below, there is no parallelism between the two. Some preliminary remarks are in order. We first need to look at the phonological status of [dze], [dza], [dzo] as well as [dzui] (attested in Commutative series 2, 3, 4 and 5 presented in 9.1). [dzi] has been shown to be unattested (cf. Commutative series 1). In actual fact, [dz] (voiced laminoalveolar affricate) is one of the two variant realizations of /z/, the other being [z] (voiced lamino-alveolar fricative). These variant realizations can be expediently indicated as [z]/[dz] with the use of a virgule, or even better as [(d)z]. It is important to point out that the presence or absence of [d] in [dz] in all these cases is phonologically irrelevant, and [z] only is phonologically relevant. This is why Izl is definable as "voiced hiss". We have seen just above that [dz] is correctly to be understood as [(d)z] since [d] is phonologically irrelevant. The Japanese phoneme whose realizations are such that they possess all the phonetic characteristics of [z] can only be Izl. Some might wish to argue that [(d)z] might be linked to the neutralization of some consonant opposition and is a realization of the archiphoneme associated therewith.101 But such an argument can be refuted. The only potential candidate neutralizable opposition would be /d/ - Izl which would be considered neutralized before Iml. Izl would obviously not be involved, as the difference between [(j)z] and [(d)z] before [to] corresponds phonologically to /?/ - Izl before Iml. Recall that Idl has been defined as "voiced apical non-nasal" and Izl as "voiced hiss". The common base of /d/ and Izl is "voiced", which is possessed by a number of other consonant phonemes of Japanese as well. Therefore Idl - Izl is a non exclusive opposition, hence a non-neutralizable opposition. The conclusion is that [(d)z] is a realization of Izl before Iml (as well as Id, Id and lol) and that Idl occurs not only before Id, Id and lol where Idl is realized by [d] but also before Iml (and /i/) where Idl is realized by [d’]. It should be added straightaway that the occurrence of IdJ before Ad/ and IU is limited to the rendition of loanwords (e.g. [d’i"'(srako)] 'disco', [d’ra'1(etto)] 'duet'). On the other hand, IdJ is non-occurent before Iml and IU in non-loanwords. What has been said in the preceding paragraph confirms that [(d)z], which occurs before Id, Id, lol and /to/, is a realization of Izl. Despite the phonetic affinity between [dz] (or rather [(d)z]) and [ts] in Japanese, there is phonologically a fundamental difference between them in the sense that [d] of [(d)z] is phonologically irrelevant and [(d)z] is therefore a realization of Izl, while [t] of [ts] is an essential component of the realization of /ts/, a single phoneme. In other words, [t] of [ts] is phonologically relevant (this is why I do not notate [(t)s]) whereas [d] of [(d)z] is not. Both [dz] and [z] are realizations of Izl, whereas [ts] can only a realization of /ts/ and a realization of Is/ can only be [s].
136
Japanese Phonology: A Functional Approach
We need to look at another case which might at first sight suggest an instance of neutralization. We have seen that [cpi] is attested (in Commutative series 1) as well as [t’i]. This means that the difference between [cp] and [t’] is phonologically relevant and may be interpretable as corresponding to /cp/ - /t/. (The palatalization in [t’ ] may be regarded as being caused by I'll which follows.) Therefore there is no case for the neutralization of /cp/ - hJ before /i/, so far as I am concerned. However, for those few speakers of standard Japanese who regularly pronounce [cp] (and not [t’ ]) before [i] and phonetically do not distinguish, for example, the two loanwords meaning 'tip' (i.e. 'gratuity') and 'chip (< 'microchip', 'potato chip'), one might wish to suggest that this is a manifestation of the neutralization of /cp/ - hJ before /i/.102 However, the common base of /cp/ "palatal affricate" and /t/ "voiceless apical non-nasal" is indeterminable, and consequently /cp/ —/t/ is a non-exclusive, hence non-neutralizable, opposition. A case comparable to the phonological irrelevance of [d] of [dz] (i.e. [(d)z]) is the phonological irrelevance of [j] of [jz] (i.e. [(j)z]) which is attested before III, Id, Id, lol and An/. I have shown further above that [(j)z]) is a realization of /?/ definable as "voiced hush”. Notice that I have not attributed anything like the relevant feature "fricative" or "affricate" to /?/. In the case of /?/, the distinction between "fricative” and "affricate" is inapplicable, unlike in the case of /?/ "palatal fricative" vs. /cp/ "palatal affricate" and this is demonstrated in realizations of /?/, i.e. [(j)z]. It might be envisaged by some that there occurs the neutralization of /z/ - Izl before /i/ and the associated archiphoneme /z-z/ is realized by [(j)z]-103 However, the common base of Id and Id is "voiced", which is also found in some other consonant phonemes of Japanese. This renders /z/ - /z/ a non-exclusive opposition and consequently a nonneutralizable opposition. Commutative series 1 to 10 presented in 9.1 show that [p] occurs before [i], [e], [a], [o] and [ui] while [s] occurs before [e], [a], [o] and [in] but not before [i]. Does this suggest that there occurs the neutralization of /p/ - Is/ before /i/,104 or alternatively that Is/ is non-occurrent before /ill We have seen earlier that Id is definable as "voiceless hush" and Isl as "voiceless hiss", so that the common base of Id and Is/ is "voiceless", which is also found in some other consonant phonemes of Japanese. This renders Id Isl a non-exclusive opposition, and consequently a non-neutralizable opposition. My conclusion is that Isl is non-occurrent before I'll but occurs before the remaining four vowel phonemes. Some might suggest that Id - Is/ is neutralized before l\J through the cancellation of the opposition between "hush" and "hiss". There is even some phonetic motivation which can support this suggestion; HI which follows the presumed archiphoneme /p-s/ has a palatalizing effect and /p -s/ can be realized by [p] rather than [s]. However, the fact that the common base of Id and Isl is "voiceless", as already pointed out, is the decisive factor which renders Id - /s/ a non-exclusive opposition, hence a non-neutralizable opposition. We have seen earlier that /?/ ("palatal fricative") is valid before I'll, Id, lol and lm I, but is non-occurrent before Id. We have also seen that hd ("glottal") is valid before Id, Id and lol, but not before III and /in /. In accounting for the non-attestation of [?] before Id, there is no question of envisaging the neutralization of either /?/ - /h/ (i.e. "palatal fricative" vs. "glottal") or that of /?/ - /$ / (i.e. "palatal fricative" vs. "labial fricative") before Id. The common base of /?/ and f\d is indeterminable, so that Id - /h/ cannot be said to be an exclusive opposition, and is consequently a non-neutralizable opposition. My conclusion is therefore that /p/ is non-occurrent before I d and that /h/ is nonoccurrent before III and ha I. As for the possibility or otherwise of the neutralization of /? /-/$ / before Id, in accounting for the non-attestation of [?] before Id, we see that the common base of /?/ and /$/ is "fricative", a relevant feature that no other consonant phonemes of Japanese possess, and one might be tempted to conclude that / ? / - / $ / is
Chapter 11: Neutralization of oppositions between the consonant phonemes of Japanese
137
therefore an exclusive opposition. However, we have already seen in 9.4 that /$/ occurs before all five vowels phonemes of Japanese, so that the non-attestability of /?/ before Id should be interpreted as the systematic non-occurrence of /?/ before Id. My conclusion given in the preceding paragraph that /h/ is non-occurrent before /txi/ diverge importantly from the time-honoured traditional analysis (attributable mainly to non-functionalists) that /h/ in Japanese has three 'allophones', namely [h] before Id, Id and lol, [?] before HI, and [$] before /u i/.l0S Notice that I partially agree with the traditional analysis, that is, to the extent that [h] occurs before [e], [a], and [o], that [?] occurs before [i] (or [j]), and that [$] occurs before [m] (or [ui ]). (This partial agreement does no violence to the non-functionalists' concept of the relation between phonetic segments and phonemes.) It is from a functional point of view that I reject this traditional analysis. The consonant phoneme whose 'allophones' are [h], [?] and [$] will have to be defined in terms of such relevant features as derive from the distinctive common phonetic characteristics of [h], [?] and [$], i.e. those phonetic characteristics which are common to [h], [?] and [$] and by virtue of which the phoneme is distinguished from other phonemes of the Japanese consonant system. Such distinctive common phonetic characteristics of [h], [$] and [$] cannot be determined. Besides, my phonological analysis of Japanese has already established /?/ (valid before HI, Id, lol and fwf) and/$/ (valid before /i/, Id, Id, lol and An/) as well as /h/ (valid before Id, Id and lol). In other words, the occurrence of /?/ is not limited to 'before /i/’, that of /$/ not limited to 'before An/', while the occurrence of Dd before the same three vowel phonemes is as traditionally agreed on. Any possibility of attributing [h], [?] and [$] to one and the same consonant phoneme (regardless of how it should be defined) does not arise according to my phonological analysis. Non-functionalists typically identify a phoneme by the criterion of phonetic similarity and modes of distribution, not in terms of relevant features as functionalists do, and can therefore analyze [h], [p] and [$] as being 'allophones' (which are in complementary distribution) of one and the same phoneme of Japanese, i.e. Dd, though what the phonetic similarity remains in such a case is problematic. 11.5
Is li l - /j/ neutralized in some contexts?
In 9.4 and also in 9.8.8 I have defined /j/ as "palatal spirant". We now need to examine whether III - /j/ is neutralized in the two cases to be mentioned below. (i) First, how about those cases where [i] is preceded by a consonant segment as in [k’ir atstu] 'atmospheric pressure’, [pir o"1] 'salt' and [r’PraN] 'good fortune'? Is [i] in such cases to be considered a realization of III or a realization of the archiphoneme associated with the neutralization of, say, /j/ - /i/? Note in this connection that [j] does not occur after consonant segments so that forms like *[k’jatsur], *[cjo] and *[r’juiN], whatever the attendant accentual patterns, are unattested. We see that /j/ - III is a non-exclusive opposition, since the common base of /j/ "palatal spirant" and lil "close front" is indeterminable, and consequently it is impossible to confirm whether or not /j/ - HI is an exclusive opposition which a neutralizable opposition must necessarily be. The abovecited examples are therefore phonologically notated as /k-k’ ir atoi/, /gir o"V and /f-r’ i r uiN/, respectively, and we conclude that 1)1 is non-occurrent after consonant phonemes and archiphonemes (all archiphonemes in Japanese are necessarily consonant archiphonemes). (ii) Secondly, in view of the fact that [ji] or [je] is non-occurrent but [ii] and [ie] are occurrent in Japanese, is the first [i] of [ii] or [i] of [ie] a product of the neutralization of 1)1 - HI before III or /e/? We have seen above (in (i)) that 1)1 - lil cannot be determined to be either an exclusive opposition or not. The conclusion is then that [ii] and [ie] are to be
138
Japanese Phonology: A Functional Approach
Some might be tempted to reinterpret "close" as "spirant", or the other way about, and "front" as "palatal", or the other way about, and identify the common base of IH and /j/ as either "close front" or "palatal spirant", with no forced choice between the two. However, this would be untenable for two reasons. First, it would no longer be possible to have two different phonemes, i.e. /i/ and /j/, since the two phonemes would not have any relevant features by virtue of which they would be distinguished from each other. Secondly, the relevant features to which "close" is opposed to are different from those to which "spirant" is opposed, and the relevant features to which "front" is opposed to are different from those to which "palatal" is opposed, and consequently "close" and "spirant" are not equivalent, nor are "front" and "palatal". 11.6
Is An/ - /iq/ neutralized in some contexts?
In 9.4 and also in 9.8.9 I presented /u[/ as a consonant phoneme definable as "dorsal spirant". Moreover, in the course of 9.8.9, we confirmed that /ui/ - /uj/ is valid before /a/ in word-medial intervocalic context. This can be demonstrated in /mauj api/ 'sumo wrestler's loincloth' vs. /main api/ 'dancing legs', and /kaujaC/ 'I won't buy it!' vs. /kaiuaC/ 'a plan to buy', phonetically presented in 9.8.9 as [mar ujap(i)] vs. [mar raap(i)] and [kar uja",N] vs. [kar ui",aN]. (/C/ is an archiphoneme definable as "consonantal" which will be explained in 11.8.4.) On the other hand, also in the course of 9.8.9, we found that it is doubtful that /in/ - /uj/ is valid in various other contexts and that it is possible that either /in / - /uj / is neutralized in some of these contexts, or either /ui / or /uj/ is non-occurrent in the remaining contexts. We shall examine this question in the present section. The various contexts in question and the occurrence or otherwise of [ui ] and/or [uj] in those contexts, though already identified in 9.8.9, will be succinctly re-presented here below at the risk of some repetitiveness, accompanied by a minimum of relevant examples. The glosses of the example words are omitted but can be found in 9.8.9. (1) In word-initial context before [a], [uj] occurs but not [in] (cf. [uj ar pi] 'eagle'; *[in ar pi],*[ina "lp], and [1] and [n], that non-nasality is chosen by the speaker for [p’]/*[b’ ] while nasality is chosen for [m*], and likewise for the rest of the above-mentioned cases. This leads to the conclusion that non-nasality in e.g. [p’] serves as the relevant feature "non-nasal" and the nasality in [m’] as the relevant feature "nasal". The readers will have noticed that what I have said just above is identical with what I earlier said in identifying nasality in 11.8.2 as the relevant feature "nasal". My repetition is deliberate in that I wished and wish to emphasize that we identify "nasal" and "non-nasal" at the same time, and conversely, "non-nasal" and "nasal". This is because a given relevant feature is only conceivable and identifiable and definable in being opposed to another or other relevant features of the same language. This is the point I emphasized in setting forth what I called 'the functional principle of the relevant feature' (Akamatsu 1988, pp. 90-99; Akamatsu 1992b, pp. 37, 83, 94, 95, 162 and note 2 on p. 54). This is why it is essential to study Commutative series 1 to 18 (Group 1) and Commutative series 19 to 36 (Group 2) in conjunction with each other in identifying the two relevant features "nasal" and "non-nasal". (13) The readers will have noticed that certain cases of comparison between sequences of consonant segments might seem to be missing in (12). Such omissions are not accidental, as I will explain. Comparison between [d’d’ ] (no commutative series provided) and [n ’d ’] (Commutative series 7) was not mentioned, as no example word can be cited for [d’d’ ]. It would be legitimate to assume the difference between [t’] and [n’ ] to be the difference between non-nasality and nasality. Comparison between [ddz] (no commutative series provided) and [ndz] (Commutative series 14) was not mentioned, as [ddz] is non-occurrent. The difference between [d] and [n] is the difference between non-nasality and nasality. Comparison between [pp] (Commutative series 30) and [V p] (Commutative series 47, to be introduced in 11.8.5) was not mentioned - [ i> p ] does not occur in Japanese /\y but we shall consider this comparison in 11.8.5. The difference between [p] and [V] is the difference between non-nasality and nasality. Comparison between [ss] (Commutative series 32) and [Vs] (Commutative series 49, to be introduced in 11.8.5) was not mentioned - [Ss] does not occur in Japanese but we shall consider this comparison in 11.8.5. The difference between [s] and [V] is again the difference between non-nasality and nasality. Comparison between [hh] (Commutative series 36) and [Vh] (cf. Commutative series 51, to be introduced in 11.8.5) was not mentioned, but will be considered in /v 11.8.5. [Vh], which is phonetically permissible in Japanese, occurs, albeit very rarely, as in [paVha',i] 'Shanghai'. The difference between [h] and [V] is the difference between non-nasality and nasality. /n /
152
Japanese Phonology: A Functional Approach
Comparison between [1*1* ] (no commutative series provided) and [n’P] (Commutative series 17) was not mentioned, as [1*1’ ] is non-occurrent. On the other hand, I have mentioned comparison between [11] (cf. Commutative series 35) and [nl] (cf. Commutative series 18). I need to explain why it is justified to consider the difference between [1] and [n] to be the difference between non-nasality and nasality. Note that the soft palate is in its raised position for [1] and in its lowered position for [n]. It might seem odd at first sight that [1], not [d] or [t], is compared with [n] for the difference to be said to be the difference between non-nasality and nasality. Actually, no such sequence of phonetic segments as [dl] or [tl] occurs in Japanese; there is consequently no possibility of comparing [d] or [t] with [n] before [1], The phonetic context associated with Commutative series 35, i.e. [- 1], obligatorily allows [1] only to occur through homorganicity by regressive assimilation. The laterality of [1] is imposed by the phonetic context [- 1], The only other consonant segment susceptible of occurring in that phonetic context in Japanese is [n]. The difference between [1] and [n] is therefore the difference between non-nasality and nasality. This analysis is certainly justified. (14) Neither voicelessness nor voicedness can be said to be a phonetic feature common to [p’], [p], [b], [t*], [t], etc., since whether each of these non-nasal consonant segments is voiced or voiceless is determined by the voicedness or voicelessness of the phonetic segment (e.g. [p’j, [p], [b], [t*], [t], etc.) with which a given commutative series is associated. This signifies that the distinction between voicedness and voicelessness lies outside the choice of the speaker and is phonologically irrelevant. There is consequently no question of establishing the relevant features "voiced" and "voiceless" in these cases. (15) In Commutative series 19 to 36, the speaker has no choice over the different places of articulation in [p*], [p], [b], [t* ], [t], etc., either, for example, bilabiality in [p’], [p], [b] and [$], apico-dentialveolarity in [t*], [t], [d’] and [d], dorso-velarity in [k* ], [k], [g*] and [g], glottality in [h], and so on. The speaker has no choice over the different manners of articulation, either, for example, plosivity in [p*], [p], [b], [t* ], [t], [j], [c], etc., fricativeness in [$], [p], [s], [z] and [h], laterality in [I], and of course non-nasality in all these consonant segments. Furthermore, the presence or absence of palatalization, for some of the consonant segments (cf. [p’j, [t* ], [k’ ], etc.; [p], [t], [k], etc.), eludes the speaker’s choice. All these phonetic characteristics are automatically determined by the various consonant segments before which the non-nasal consonant segments in question occur. (16) The non-distinguishability between /p/, /p’/, /b/, /b7,/t/,/d/,/k/,/k’/,/g/,/g’/,/$/,/?/, /j/, Ihl, Id, Id, /s/, /z/, As/, led, Id, Idl and /u|/ in the contexts specified in Commutative series 19 to 36, with only non-nasality being the speaker's choice, points to the conclusion that there occurs in those contexts the neutralization of /p/ - /p’/ —/b/ —/b’ / —/t/ —/d/ — /k/ - Ik'l - Igl - /g’/ - / $ / - / ? / - /j/ - I h l - I d - / z / - / s / - / z / - / t s / - / c p / - / r / - / f 7 - / t q / , while this phonological opposition is valid before Id and lol. (17) The archiphoneme associated with the above-mentioned neutralization is definable as "non-nasal" and can be symbolized by /p-p’-b-b’ -t-d-k-k’-g-g’-$-p-h-p-j-sz-ts-cp-?-r-r’-u|/, or conveniently, by IQI. The symbol Q is traditionally employed by many who are concerned with Japanese phonology to designate a phonological entity which is not quite equivalent to that which I consider as the above-mentioned archiphoneme. (18) The archiphoneme IQI definable as "non-nasal" is directly opposable to the archiphoneme IN/ definable as "nasal" which we have earlier identified, and also to another archiphoneme definable as "consonantal" which I shall identify in 11.8.4 and symbolize by /C/.112 For the moment we shall be concerned with the opposition between IQI and INI. The opposition between these two archiphonemes can be seen in those contexts in which both are susceptible of occurring. Witness, for instance, the
Chapter 11: Neutralization of oppositions between the consonant phonemes of Japanese
153
following pairs randomly chosen: [he’1m’p’ina] 'out-of-the-way' (phonologically /he^N p-p’ ina/) vs. (inp’p’ i] 'helping hand' (phonologically /i^Q p-p’ i/); [gui "'ntai] 'army' (phonologically /gru ’’Ntai/) vs. [ka^ttai] 'leprosy' (phonologically /ka^Qtai/); [se’’i>ccoo] 'ship's captain' (phonologically /se^Ncpoo/) vs. [g’ Pccpo] 'left-handed' (phonologically /g-g’ i^cpo/). (19) The phonological content of an archiphoneme is traditionally defined as the common base of the phonological contents of the terms of a neutralizable opposition. (I have critically discussed this point in Akamatsu 1988, pp. 272-277.) Under this traditional view, it would be hard to see how IQI is defined as "non-nasal", which is the sole relevant feature of which IQI consists. Some of the terms of the neutralizable opposition /p/ - /p’/ - /b/ - /b’/ - /t/ - /d/ - /k/ - /k’/ - /g/ - Ig'l - /$/ - /?/ —/j/ —/h/ —/p/ / ? / - / s / - / z / - / t s / - / c p / - / r / - / r 7 - / q / possess the relevant feature "non-nasal" (i.e. /p/, /p7, /b/, Po'I, l\l, 161, /k/, Ik'I, /g/ and Ig'l) but not the others (i.e. /$/, /?/, /j/, Ihl, /p/, /?/, Id, /z/, /ts/,/cp/,/r’/,/r ’/a n d /q /). My own view, forcefully put forward in Akamatsu (1988), Akamatsu (1997a) and a few other works of mine is that it is during the course o f the commutation test that the phonological content of an archiphoneme is arrived at. This is equivalent to identifying the archiphoneme in the context(s) of neutralization where the archiphoneme occurs in opposition to the other distinctive units of the second articulation, in particular those which are directly opposed to the archiphoneme, in the context(s) of neutralization. In some cases, the phonological content of an archiphoneme thus arrived at happens to coincide with the phonological content of the archiphoneme traditionally arrived at by identifying the common base of the phonological contents of the terms of the neutralizable opposition. One such case concerns /N/ definable as "nasal" in Japanese, which is associated with the neutralization of /m/ - /m’/ - Ini - In'I - /g/ - /g’/, since /m/ is definable as "non-palatalized labial nasal", /m’/ as "palatalized labial nasal", Ini as "non-palatalized apical nasal", In' I "palatalized apical nasal", /g/ as "non-palatalized dorsal nasal" and /g’/ as "palatalized dorsal nasal". The common base is of course "nasal". Any suggestion that IQI is the archiphoneme associated with the neutralization of Ipl - /p’/ - /b/- lb' / - 111 - Id/ - IkJ- Ik'I - Igl - lg' I, with the common base of these consonant phonemes being "non-nasal", is rejected. The reason for my rejection of the above suggestion is that what is meant by the neutralization of Ipl - Ip'l - Ibl - lb' I - It! - Id! - IkJ - Ik'I - Igl - lg' I - /$/ - /?/ - /j/ - Ihl - /p/ - Id - Is/ - Izl - Its/ - /cp/ - Id - Ic'l /q/ is the impossibility of distinguishing between Ipl, Ip'l, Pol, Po'l, Id, Idl, IkJ, fk' I, lg/, Ig'l, /$/, Id, 1)1, Ibl, Id, Id, /s/, Izl, /ts/, /cp/, Id, Ic'l and /q / in each of the various contexts to be specified in (20) below, the contexts which figure, in phonetic terms, in Commutative series 19 to 36. (20) The contexts of neutralization for Ipl - Ip'l - Pol - Po' I - Id - Idl - Pd - /k’/ - Igl -Ig 'l - /$/ - /?/ -1)1 - P d - /p/ - Id - I d - I d - /ts/ - /cp/ - Id -Ic 'l - /q / are as follows: 'before the archiphoneme /p-p’/' (cf. [i"*p’p’ i}] 'helping hand', [ir ppeN)j 'once'), 'before Ipl (cf. [?}r ppa"Ym] '(to) puli'), 'before Ip'l' (cf. [har p’p’ akur] '800'), 'before Pol (cf. [ro"lbbui] 'lob (in tennis'), 'before/t/' (cf. [bar t’t’iggui] 'batting', [ka^tte] 'formerly'), 'before Idl' (cf. [be^ddo] 'bed'), 'before the archiphoneme lk-k'l' (cf. [sa^R’k’(i] 'a short while ago', [ha^kkee] 'the eight beauty spots by Lake Biwa'), 'before Pd' (cf. [sa^kka] 'novelist'), 'before Ik'l' (cf. [ka"lk’k’ o] 'angular distance'), 'before the archiphoneme /g-g’/ (cf. [mar g’g’ i'Vui] 'McGill'),
154
Japanese Phonology: A Functional Approach
'before/g/' (cf. [ba^ggra] 'bag'), 'before /$/' (cf. [sa^^oo] 'Sappho'), 'before /h/' (cf. [ba^hha] 'Bach'), 'before /?/' (cf. [sanqq ij 'booklet'), 'before li t (cf. [ro"'jj?i] 'lodge'), 'before Is/' (cf. [ka^sseN] 'battle'), 'before /ts/1(cf. [pM^ttsa] 'pizza'), 'before /c?/1(cf. [sanccq(i ] 'inference'), 'before kP (cf. [mar f’inel"1la] 'Marinella'. Contexts specifiable as 'before /b’/’, 'before /b-b’/', 'before /g’/1, 'before /?/', 'before /z/\ 'before/r-rV, 'before /r’/' and 'before /iq/' are not mentioned as contexts of neutralization fo r/p /-/p 7 -/b /-/b 7 -/t/-/d /-/k /-/k 7 -/g /-/g 7 - /$ /- /? /- /j/- /h /- /q /- /z /- /s /Izl - /ts/ - /cq/ - / r / - / r ’ / - /u j/, as Commutative series 19 to 36 do not point to the involvement of these particular contexts. (20) Any neutralizable opposition has by definition both a context(s) of relevance and a context(s) of neutralization. In the case of the neutralizable opposition /p/ - /p7 /b /-/b 7 -/t/-/d /-/k /-/k 7 -/g /-/g 7 -/$ /-/? /-/j/-/h /-/c /-/? /-/s /-/z /-/ts /-/c q / - / r / - / r ’/- /u j/, the contexts of relevance are 'before /a/' and '/before /o/', where this neutralizable opposition is valid, as can be seen from the information provided in 9.4.
11.8.4
Commutative series 37 to 43 (Group 3)
We next turn our attention to Commutative series 37 to 43 (Group 3)
Group 3 Commutative series 37 t-m ’] [(sa",)m’m’(i)] 'acidity' **[m’m’(e)] [(ra)r m’m’(akxu)] 'disorder' [(qi)r m’m’(oo)l 'docile' *[-m’m’(ru)l?113
Commutative series 38 1-ml **[mm(i)] [(ta"l)mm(ee)] 'short life' [(to"')mm(a)] 'simpleton' [(tu*l)mm(o)] 'mica' [(guj",)mm(ui)l 'military affairs'
Commutative series 39 T- t>1 ((k a ^ ^ ^ liN )! 'forbearance' **[tM>e] [(ha"l)TJbT>(a)] 'demoness' [(ke)r T>r>(oo)] 'urine test' [(he)r i>T>(tnui)] 'incorporation'
Commutative series 40 [-n1 **[nn(i)] [(ka)r nn(eN)] 'conception' [(ka'l)nn(aN)l 'hardships' [(ka'1)nn(oo)] 'mastery' [(ka"l)nn(uiq(i))] 'Shinto priest'
Chapter 11: Neutralization of oppositions between the consonant phonemes of Japanese
Commutative series 41 f-n ’] [(ci)r g’ g’(i"'kai)] 'deliberate council' * * [g ’
g’(e)]
[(ka^)g’ g’(a)] 'refinement' [(k’i",)g’ g’(o)] 'goldfish' [(ke)r g’ g’(uiui)] 'Altair'
155
Commutative series 42 [ - 0] **[on(i)] [(ka‘,)gg(e)] 'soliciting contributions for religious purposes' [(o‘1 )gg(aktu)] 'music' [(se'*)gg(o)] 'postwar' [(ga‘l)gg(m)] 'toy'
Commutative series 43 _________________________ r - # i
[(car k’k’i‘1 )N] 'debt'___________ [(se^N] '1,000'________ [(han)N] 'printing block'________ [(so^N] 'loss'_________________ [(tQ N ] 'destiny'______________ The following analytical observations can be made in connection with Commutative series 37 to 43. (1) The phonetic contexts associated with Commutative series 37 to 43 - the sole exception being that in Commutative series 43 - differ importantly from those associated with Commutative series 1 to 36 (Groups 1 and 2) in that the phonetic segments before which [m’], [m], [i>], [n], [g’ ] and [g] occur are nasal consonant segments. As a result we have in Commutative series 37 to 43 sequences of two identical nasal consonant segments of various places of articulation, i.e. [m’m’], [mm], [t> t>], [nn], [g’ g’] and [gg]. In the exceptional case, in Commutative series 43, the context in which [N] occurs is prepausal. (2) In Commutative series 43, the context in which [N] occurs is 'prepausal' (indicated by the symbol #). In this respect Commutative series 43 differs from the rest of the commutative series assembled as Group 3. Firstly, the prepausal context consists of no phonetic segment, be it a non-nasal consonant segment or a nasal consonant segment. Secondly, there consequently occurs no sequence of two identical nasal consonant segments. (3) It is evident that there occurs homorganicity involving regressive assimilation between the two nasal consonant segments in each of the sequences of [m’m’ ], [mm], [t>T>], [nn], [()’o’] and [rjg] and that consequently there is no choice available for the speaker to choose either the different places of articulation or the presence or absence of palatalization with regard to [m’]> [m], [t>>], [n], [o’] and [o]. (4) As to whether or not we can conceive of homorganicity between [N] and the pause that follows it, it may be argued that [N] exhibits regressive assimilation in relation to the pause since the posture of articulatory rest (when one is not speaking) corresponds to the dorso-uvular articulatory posture of [N]. The validity of this view rests on whether a dorso-uvular closure obtains during normal breathing.114 (5) Only nasal consonant segments, i.e. [m’], [m], [r*>], [n], [g’]> [g] and [N], occur, but no non-nasal consonant segments do. Nasality, which is common to [m’], [m], [i>], [n], [g’] and [g], occurs as a result of regressive assimilation due to the occurrence of [m’], [m], [r>], [n], [g’] and [g] before which [m’ ], [m], [i>], [n], [g’] and [g], respectively, occur. The nasality in question therefore eludes the speaker's choice. Contrast this with the nasality in [m’], [m], [n’ ], [n], [g], [g’ ] and [ t>>] which we have earlier seen in connection with Commutative series 1 to 18 (Group 1) and which we have seen as being the speaker's choice as the above-mentioned nasal consonant segments are
156
Japanese Phonology: A Functional Approach
opposable to [p’], [p], [b], [t’ ], [t], [d’], [d], [k*], [k], [g’] and [g] in Commutative series 19 to 28 (part of Group 2 consisting of Commutative series 19 to 36). One might be tempted at this stage to see the nasality in question being linked to the relevant feature "nasal" in terms of which the archiphoneme that we have symbolized by /N/ is defined. That this is not so will be explained in (7), (8 ), (9), (10), (11) and (12) below. (6 ) In Commutative series 43 in which [N] occurs before a pause, it is obvious that the speaker has no choice of distinguishing between different places of articulation or the presence or absence of palatalization any more than in Commutative series 37 to 42. It should be added in this connection that in prepausal context no consonant segment occurs in Japanese with the exception of [N]. As for whether nasality in [N] is the speaker's choice, let it be noted that [q] (voiceless dorso-uvular plosive) does not occur in prepausal context where [N] occurs and that this fact prevents us from confirming whether or not the nasality in [N] is the speaker's choice, ([q] does not occur in any context in Japanese anyway.) (7) In our attempt to identify the distinctive unit whose realizations are [m’], [m], [i>], [n], [o’], [rj] and [N] in the contexts associated with Commutative series 37 to 43, it is important to note the following facts. (i) In the contexts 'before [m’]', 'before [m]', 'before [t>]', 'before [n]', 'before [g’]' and 'before [g]’, [nT], [m], [i>], [n], [g’j and [g], respectively, are not commutable with any other consonant segments, which do not occur in these contexts anyway. Nothing like even, for example, [p’m’ j or [b’m’] occurs, which would permit commutation between [p’] and [b’] on the one hand and [m’ ] ([m’m’] does occur; see Commutative series 37) on the other. Indeed, no consonant segments occur at all in the abovementioned contexts. In the context 'before a pause', i.e. in prepausal context, [N] which is the only consonant segment that occurs in that context is not commutable with any other consonant segment, which does not occur anyway in that context. (ii) Japanese has, in word-medial intervocalic context or otherwise, no such sequences as [p’m’ ]/[b’m’], [pm]/[bm], [ct>]/[jt>], [tn]/[dn], [k’g’]/[g’g’], and [kg]/[gg], which would enable comparison between [m’m ’ ] and [p’m ’]/[b’ m ’], [mm] and [pm]/[bm], [i>r>] and [ci>]/[jn.], [nn] and [tn]/[dn], [g* o’] and [k’g’]/[g’g’], and [gg] and [kt)]/[gg]- The occurrence of [p’m’]/[b’ m’], [pm]/[bm], [ct>]/[p>], [tn]/[dn], [k’g’]/[g’g’], and [kg]/[gg] would serve to show the nasality in [m’], [m], [»>>], [n], [g’ ] and [g] as being the speaker's choice, for reason of the difference between [m’] and [p’ ]/[b’], [m] and [p]/[b], [r*>] and [c]/[j], [n] and [t]/[d], [g*] and [k’ ]/[g’], and [g] and [k]/[g], assuming that [m’m’] and [p’m’ ]/[b’m’], or [mm] and [pm]/[bm], etc., are not in free variation. (iii) On the other hand, [m’], [m], [i>], [n], [g’], [g] and [N] in [m’ m’], [mm], [t»>t>], [nn], [g’ g’], [gg] and [N], respectively, in the contexts associated with Commutative series 37 to 43 are commutable with any of the vowel segments, i.e. [i], [e], [a], [o] and [ui], (8 ) I present below five more commutative series which will lead us to identify the distinctive unit whose realizations are [m’], [m], [i>], [n], [g’], [g] and [N] in the contexts associated with Commutative series 37 to 43. These five commutative series are new and additional and will be serially numbered as Commutative series 37', 38', 39', 40', 4T, 42' and 43' (and not as Commutative series 44, 45, 46, 47, 48, 49 and 50) as they are just an expansion of Commutative series 37 to 43. What will be shown in Commutative series 37', 38', 39', 40', 4T, 42' and 43' is the commutability between [m’], [m], [t>], [n], [g*], [g] and [N], respectively, and each of the five vowel segments, i.e. [i], [e], [a], [o] and [m]. Commutative series 37' r-m ’l [(ka"')m’m’(i)] 'sweetness' [(ka"')im’(i)] 'agreeable sensation'
Commutative series 38' [-m l **[-mm(i)] **[- im(i)]
Chapter 11: Neutralization of oppositions between the consonant phonemes of Japanese
f(cpi)r em’(i)] 'a Japanese female forename' [(jut)r am’(i)] 'a bath' f(na)r om’(i)] 'a Japanese female forename' [($txi"l)TDm’(i)l 'flavour' (e)] **[- im’(e)] **[—em’(e)] **[- am’(e)] **[- om’(e)] **[- mm’(e)]
**[- em(i)] **[- am(i)] **[- om(i)] **[- tnm(i)] [(jzi)r mm(e)] 'person's life' [(ka)r im(e)] 'changing one's name' r(se",)em(e)l 'life' [(kja^mfe)] 'tortoise' [(co)or m(ee)] 'illumination' [(jui)r iBm(ee)] 'fame'
[(k’i)r m’m’(akur)] 'vein of gold'
[(j?i"l)mm(a)] 'nettle'
[(da)ir m’(akxij:)] 'locum tenens'
[(ta",)im(a)] 'hemp'
[(he)1"em’(akin)] 'normal temperature' *[- am’(a’)] [(do)r om’(akttr)] 'artery'
[(ke)r em(a)] 'knight (in chess)' [(jzi)r am(ar’i)] 'hypermeter' [(go"l)om(a)] 'conquest of the devil'
[(k’ui)r uim’(akuj:)] 'quick pulse'
[(ko)r nim(a)] 'pony'
[(ci)r m’m’(o)] 'mysteriousness' r(ka)r im’(oo)l 'changing one's name' f(ke)r em’(oo)l 'light and clever' *[- am’(o)] f(ko)r om’(oo)] 'dexterity' * [- UJ m’(o)]
[(k’i)r mm(otsxu)] 'tabooed thing' [(k’)ir m(o)] 'liver' [(ke)r em(o)] 'enlightenment' [(kJa^mCo)] 'wild duck' [(m>i)o*lm(o)] 'pregnancy' [(.uvaf tnm(oo)] 'animal hair'
*[- m’m’ (ui)] *[-im ’(ui)] *[- em’(tu)] r(b)ar m’(ui",tuda)] 'Bermuda' f- (k)or m’(uj",uiN)] 'commune' *[- nj m’(ui)]
[(k’i"l)mm(ui)] 'work' [(san)im(ui)] 'debt' f(ke",)em(ui)l 'police affairs' [(cca",)am(ui)] 'charm' [(so’1 )om(ui)] 'general affairs' [(mro)r tnm(Txi",ui)] 'muumuu'
(C o m m u ta tiv e se rie s 39' r - t>l
r(a", )T>>T>(i)l 'im phcitly' [(e)ir r> (ise)] 'lo n g -e s ta b lis h e d sh o p '
C o m m u ta tiv e series 4 0 ' [ - n] * * [ - n n (i)] * * [ - in(i)]
[(k e)e’1r>(ikm :)] 'c h ic k e n '
* * [ - en(i)]
[(k er r|)aT>(iN)] 'in ju re d perso n ' [(ko)r o t> (iN )] 'su c c e sso r' [(sui)r tDT>(iN)] 'se v e ra l p erso n s'
* * [ - an (i)] * * [ - on(i)] * * [ - r a n ( i) ]
* * [ - t>»r>(e)] * * [ - ii> (e)]
r(h a)r n n (e)l 'h alf-p rice' [(k a)r in(e)] 'p u rc h a se p rice'
157
158
Japanese Phonology: A Functional Approach
**[- ei>(e)] **[- at>(e)] **[- OT>(e)] **[- id T>(e)]
[(te)r en(eN)] 'retirement age' [(k)ar n(e)] 'money' [(h)or n(e)] 'bone' [(tsui",)Tnn(eN)] 'generally accepted idea'
[(ha",)i>T>(a)] 'demoness' *[- if>(a)] *[- er>(a)] *[- at>(a)] *[- ot>(a)] *[- to T>(a)]
[(ka"')nn(a)] 'canna' [(ka",)in(a)J 'arm' [(ta)r e’,n(artn])] 'exquisite' [(ke)r an(a)] 'pores of skin' [(so)r on(aN)l 'disaster' [(kui)r ran(aN)] 'impending calamity'
[(ci~1 )i>T>(o)] 'goddess' [(?)ir T>(oo)] 'urination' * [ - er t>(oo)] *[- a r t>(oo)l *r- Or T>(00)l [(ker ts)ni J>(oo)] 'haematuria'
[(qi)r nn(oo)] 'pericardium' [(m’)Pn(o)] 'straw raincoat' [(ge)r en(oo)] 'public entertainments' [(p’i)r an(o)] 'piano' [(te)r on(o)] 'hatchet' [(k’iu)r inn(oo)] 'rhinencephalon'
[(ha)r i>r>(imxi)] 'carrying (things) into'
[(ka^nnOnp i )] 'Shinto priest'
[(ka)r iT>(tmn)] 'intervention'
[(kan)in(mp i )] 'owner of pet animal'
*[- ei>(ui)] *[- ai>(tu)] [(ko)r QT>(txiui)] 'purchase' [($m)r uiT>(uitu)] 'enclosure'
[(ha)er n(rak’ i)] 'purity' [(k)ann(ixnn)] 'canoe' *[- on(xu)] *[-inn(ui)]
Commutative series 41' [-if] [(ha"1) o’ n’(i)l 'printing block' [(kan)ig’(i)] 'meeting' [(ge'*)eti,(i)] 'geisha girl' [(ke'*)aii’(i)] 'meeting' [(ko'l)ori’ (i)] 'protestation' [($uj",)ujg’(i)] 'public morals'
Commutative series 42' [ - 0] **[-30(1)] **[-io( 0 ] **[—eg(i)] **[-ag(i)] **[- og(i)] **[—m g(i)]
**[- if)’(e)] **[- eg’ (e)] **[- ag’(e)] **[- og’(e)]
[(sm)r gg(ek’(i)] 'dramatic sketch' [(ka)r ig(eN)] 'change of the era' [(se)r eg(eN)] 'restriction' [(ci)r ag(e)] 'completion' [(ko)r og(eN)] 'tableland'
**[-ing’(e)]
[(j?uifiDg(ek’(i))] 'rifle shooting'
[(ha)r g’ g’(aiu)] 'rebellion'
[(sa)r gg(ai)] 'second floor'
Chapter 11: Neutralization of oppositions between the consonant phonemes of Japanese
[(ka)r ig’(akiu)] 'pleasantry' *[- er)’(a)] [(k)ar g’(akut)] 'reversible reaction' *[- orf (a)] *[- in o’(a)]
159
[(sa)r ig(ai)] 'calamity' [(he)r eg(ai)] 'evil' [(pi)r ag(ar’i)] 'completion' [(ro)r og(aN)] 'presbyopia' [(sui)r ujg(akxn:)] 'mathematics'
[(ka)r g’ q’(oo] 'government enterprise' [(ka)r iri’(oo)] 'opening of a shop’ *[—erj’ (o)] [(k)a"V(oo)] 'business' [(ko^lonToo)! 'industry' [(7>ui",)inr)’(o)] 'emperor's withdrawal to the inner palace'
[(sa^qgto)] 'coral' [(ka)r ig(oo)] 'meeting' [(ke)r eq(o)] 'honorific expression' [ar g(o11)] 'jaw' [(ko)r oq(o)] 'colloquial speech' [(kni)r ing(oo)] 'dry moat'
r(ka)r g’ tf(ra|ru)] 'Altair' *[- ir)’ (in)] *[—eg’ (hi)] * [-ag ’(u!)] *[- og’(iu)] [(T)»uj)iDg’(u0 ] 'milch cow'
(pi~l)gg(Tn)] 'bedding' [(ka^iqtuiN)] 'navy' [er q(ruN)] 'military reinforcement' [(^a^gliora)! 'Shinto music and dancing' [(k)onq(m)] ’(to) row' [($)m"1 g(ui)] 'swellfish'
Commutative series 43' ________________ L N ]_______________ [(k a ‘l )N ] 'a p p e a ra n c e '______________________
[(ka'*)i] 'seashell'______________________ [(ma'l)e] 'front'________________________ [(ka'1)a] 'motor-car'____________________ [(ro'l)o] 'wax'________________________ [(ka)r tn] '(to) buy'____________________ A few remarks - (i), (ii) and (iii) below - are in order about Commutative series 37' to 43' presented just above, (i) The double asterisks indicate that the phonetic sequences in question are nonoccurrent in Japanese. For example, **[- mm(i)], **[- im(i)], **[- em(i)], **[- am(i)], **[- om(i)] and **[- in m(i)] in Commutative series 38' since [m] is non-occurrent before [i] in Japanese. (ii) The different pitch patterns (which are associated with different accentual patterns) have no repercussion whatever on the occurrence or non-occurrence of those phonetic segments which are indicated in boldface in the respective commutative series and which are the object of our consideration in this section. (iii) Accidental gaps in the occurrence of certain sequences of phonetic segments in lexical items of Japanese (just as happens in any individual language) have the consequence that a number of lacunae occur in parts of the commutative series presented above. For example, in Commutative series 37', I have not been able to find any existing citable item for [- am’(a)], for which I have therefore put a single asterisk. This lacuna, for instance, is potentially fillable without doing violence to the general combinabilities of phonetic segments in Japanese. The lacunae (with single asterisks) that result from the above-mentioned accidental gaps should not be confused with the
160
Japanese Phonology: A Functional Approach
lacunae (with double asterisks) which result from sequential phonetic impossibilities in Japanese. I continue below my observations being made in connection with Commutative series 37 to 43, but now by taking into consideration Commutative series 37' to 43' as well. (9) As can be judged from Commutative series 37' to 43', the distinctive unit whose realizations are [m’], [m], [t>], [n], [g’], [g] and [N] is commutable with any of the five vowel segments, i.e. [i], [e], [a], [o] and [to], but with no consonant segments. This distinctive unit is the archiphoneme associated with the neutralization of /p/ - /p’/ - /b/ - / b ’/ - / t / - / d / - / k / - / k ’/ - / g / - / g ’/ - / $ / - / ?/ - / j / - / h / - / p / - / ? / - / s / - / z / - / t s / - / c c / / r / - / r ’/ - / m / - /m’/ - / n / - I n 'I - /g/ - /g’/ - /uj/ in the contexts which I shall specify below in (13). This archiphoneme is definable as "consonantal" 1 1 5 and will be symbolized by /C/. The characterization of /C/ does not require the additional term 'nasal', though realizations of /C/ are without exception nasal. All realizations of /Cl are nasal and homorganic with the respective nasal consonant segments that follow them with regard to voicedness, place of articulation, and palatalization or non-palatalization. Neither the place of articulation nor the distinction between nasality and non-nasality is phonologically relevant, and only the characteristic of consonantality is phonologically relevant, hence "consonantal". But the nasality which is common to all the realizations of/C/, i.e. [m’], [m], [ t> ], [n], [g’ ] and [g], the various places of articulation of, and palatalization or non-palatalization in, [m’], [m], [i>], [n], [gT] and [g] are automatically determined by the presence of [m’], [m], [i>], [n], [g’] and [g] before which they occur (cf. regressive assimilation). As for [N], the determining factor for its nasality, and for its place of articulation, i.e. uvularity, is the physiological state at rest, in the relevant area of the vocal tract, associated with a pause. The factors directly involved in this physiological state are the raised back of the tongue, the uvula with which the tongue is in contact and the velic opening (i.e. the opening between the soft palate and the nasopharynx). As to why [N] occurs rather than some other consonant segment, the reason is simply that no consonant segment other than [N] can occur in prepausal context in Japanese. If a consonant segment is to occur, as it does, before a pause, the consonant is optimally [N]. As for [N] which is the only consonant segment that occurs in prepausal context, this too is a realization of the archiphoneme /C/ definable as "consonantal". [N] is commutable with vowel segments only, in fact all of [i], [e], [a], [o] and [to], as is seen in Commutative series 43'. (10) It goes without saying that my definition of the archiphoneme as "consonantal" has been arrived at during the course of the commutation test. It cannot have been arrived at through the traditional definition of the archiphoneme as being the sum of the relevant features common to the terms of the neutralizable opposition. (11) [m’], [m], [i>], [n], [g’j, [g] and [N] which we saw in Commutative series 37 to 43 are realizations, in fact, combinatory variants, of the archiphoneme /C/. (12) My phonological analysis of [m*], [m], [i>], [n], [g’], [g] and [N] as realizations of the archiphoneme /C/ definable as "consonantal", not as realizations of the archiphoneme /N/ definable as "nasal", constitutes a major departure from the traditionally accepted analyses, performed by both Japanese and non-Japanese scholars, who consider both /N/ and /Cl to be one and the same distinctive unit to which they generally refer as 'mora nasal' or 'moraic nasal' and symbolized by /N/. A few reasons can be suggested for the traditional analysis in question. Firstly, and most importantly, other scholars do not identify the distinctive units of Japanese on the basis of the commutation test which is a device for phonological analysis fundamentally based on the concept of opposition. Secondly, the very fact that the phonetic segments which constitute the object of the present phonological analysis, i.e. [m’J, [m], [■*>], [n], [O’], [g] and [N], are all characterized by nasal articulation, sways other scholars in the
Chapter 11: Neutralization of oppositions between the consonant phonemes of Japanese
161
direction of the conclusion that the distinctive unit in question is (phonologically as well as phonetically) characterized by nasality, hence "nasal". Thirdly, the fact that the archiphoneme /N/ definable as "nasal" is identified - correctly - in those cases which are comparable to what I have presented in terms of Commutative series 1 to 18, easily tempts scholars to see another instance of /N/ also in the contexts which I have presented as being associated with Commutative series 37 to 43. Fourthly, Japanese scholars at large, and also those non-Japanese scholars who are acquainted with Japanese writing system, are influenced in their phonological analyses by the occurrence of the relevant kana character - A, (hiragana) or V (katakana) - which is traditionally understood to correspond to /N/, even in the contexts where the archiphoneme /C/ - or in their minds the archiphoneme /N/ - occurs. The hold of the Japanese kana characters, in the present case L or V, cannot be underestimated. Note therefore that according to my own phonological analysis, both /N/ and /Cl correspond to A/ or V, which goes against the traditional phonological analyses of Japanese. (13) The contexts of neutralization for /p/ - /p’/ —/b/ —/b’ / —/t/ —/d/ —/k/ —/k’/ - /g/ —/g’/ —/$/ - /?/ —/j/ —/h/ —/q/ - /?/ - /s/ - Izl - /ts/ - /cq/ - Id - Id I - Iml - /m’/ - / n / - / n ’/ /g /-/g 7 -/tq /, the contexts where ICI occurs, as already adumbrated in Commutative series 37 to 43, are as follows: (i) ’before/m’A(cf. [far m’m’akin] 'disorder'); 'before /ml' (cf. [to"’mma] 'simpleton'); 'before In'/' (cf. [ha"1 i>T>a] 'demoness'); 'before InJ' (cf. [ka^nnaN] ’hardships'); 'before /gY (cf. [ka"'g’ g’a] 'refinement');116 and 'before Iqf (cf. [o^ggakui] 'music') ;116 (ii) 'before the archiphoneme /m-m’/' (cf. e.g. [sa^m’m’i] 'acidity', [ta‘1 mmee] 'short life'); 'before the archiphoneme /n-n’/' (cf. e.g. [ka^r^^iN] 'forbearance', [ka^nneN] 'conception'); and 'before the archiphoneme /g-g’/' (cf. e.g. [ko',g’ g’i] 'nuptials', [kangge] 'soliciting contributions for religious purposes') ;116 and (iii) 'before a pause' (cf. [qark’k’ PN] 'debt'). (14) The broad context in which the sequences consisting of two identical nasal consonant segments, [m’m ’ ]. [mm], [tm>], [nn], [g’ g’] and [gg ], occur can be characterized as 'post-vocalic word-medial pre-consonantal context'. Nevertheless, it is necessary to note the occasional occurrence of [mm] in word-initial prevocalic context as in [mr ma"1] instead of [ui r ma'1] 'horse', [m r me] instead of [ui r me] 'plum', [mr mo(rerin)] '(to) get buried', etc. and their derivatives. 117 The first [m] of [mm] in [mr ma"1] and the other related examples is a realization of the archiphoneme /C/.118 Traditional analyses of course postulate here 'mora nasal' or 'moraic nasal' and symbolize it by INI. The second [m] in [mpma], [mr me] and other examples are /ml or /m-m’/, as the case may be. 11.8.5
Commutative series 44 to 52 (Group 4)
Next we will examine Commutative series 44 to 52 which I have brought together as Group 4. Group 4 Commutative series 44 [-,], 'before the archiphoneme fX - X’/ (N.B. not /n-n’/)'; [n] or [n’ ], 'before/X/(N.B. not/t)/)'; [rjj, 'before /X’/ (N.B. not /rj’/)'; [g’l 'before the archiphoneme /m-mY; [m] or [m’l 'before the archiphoneme /n-n’/'; [n] or [i> 4 'before a pause'; [N]. This concludes the commutation test which I have carried out on the basis of the data from standard Japanese in Chapters 8 , 9, 10 and 11.
Chapter 12: The commutation test reviewed
179
Chapter 12 The commutation test reviewed 12.1
What has the commutation test achieved in Chapters 8, 9,10 and 11?
I have explained in some detail in Chapter 7 the importance of the commutation test in functional phonology and shown how to perform the commutation test. Recall that there were already occasional references to the commutation test in parts previous to Chapter 7. Throughout Chapters 8 , 9, 10 and 11,1 have performed the commutation test on the basis of data obtained from standard spoken Japanese. It is worth at this point reviewing what the commutation test has achieved in my phonological analysis of Japanese. (1 ) 1 have elicited, by recourse to the commutation test in Chapters 8 , 9, 10 and 11, the distinctive units of the second articulation, i.e. the phonemes and the archiphonemes of Japanese. All these distinctive units have been defined in terms of relevant features, i.e. their respective, mutually different, phonological contents. Recall that the identification of the phonological content of a distinctive unit is equivalent to defining the distinctive unit itself. (2) The Japanese phonemes elicited have been 'tentatively' classified as vowel phonemes the consonant phonemes. The distinction between vowel phonemes and consonant phonemes is basically allied to the distinction between syllabic distinctive units and nonsyllabic distinctive units. In my view, both these distinctions are little useful in a functional description of Japanese phonology, and the Japanese phonemes and archiphonemes will be reclassified later (18.20). Incidentally, in terms of the syllabic-nonsyllabic dichotomy, all the archiphonemes of Japanese will be regarded as consonant archiphonemes. Japanese has no vowel archiphonemes. (3) The phonological system constituted by the five vowel phonemes is valid in the whole of standard Japanese pronunciation, and was presented diagrammatically in terms of 'orders' (i.e. "close" order; "mid" order; and "open" order) and 'series' (i.e. "front" series; "back" series) in 8.5. The phonological system constituted by the consonant phonemes is not uniformly applicable to the whole of standard Japanese pronunciation, and two sub-systems are envisaged which are applicable to what I call 'minority speech' and 'majority speech', respectively. For the main differences between the two sub systems, see 12.2.3 further below. The two sub-systems of the consonant phonemes of Japanese can be seen presented in a tabular form in Appendixes 1 and 2. As for the set of the archiphonemes of Japanese, this does not form a system conceivable in terms of orders and series. (4) The contexts in which the phonemes elicited occur have been identified. (5) Realizations of each of the phonemes in the contexts in which it occurs have been described. ( 6 ) Various instances of neutralization of oppositions between consonant phonemes in Japanese have been discovered. Each neutralizable opposition has been identified. The contexts of relevance and the contexts of neutralization for each neutralizable opposition have been identified. (7) The archiphoneme which is associated with each neutralizable opposition and which by definition occurs in the contexts of neutralization has been defined in terms of its phonological content. Realizations of each archiphoneme in the contexts in which it occurs have been described. (8 ) Instances of systematic non-occurrence of some individual phonemes in certain contexts in Japanese, which are distinct from those of neutralization, have been discovered. The distinction between systematic non-occurrence of phonemes and
180
Japanese Phonology: A Functional Approach
neutralization of phonological oppositions is made possible by the criterion of 'exclusive relation', hence the concept of 'exclusive opposition' (5.4). A neutralizable opposition is obligatorily an exclusive opposition. (On the other hand, an exclusive opposition may be a neutralizable opposition or not, as the case may be.) (9) In connection with neutralization, I wish to further mention the following. The phenomenon of neutralization has been shown to demonstrate excellently the functionalist principle to the effect that 'there is no necessary correspondence between physical reality and linguistic function' (cf. 2.7). To cite just one example from Japanese phonology (in connection with the minority speech), [m] (voiced bilabial nasal consonant segment) is a realization of the phoneme /m/ "palatalized labial nasal" (as in /mui/ 'naught') or a realization of the archiphoneme /m-m’/ "labial nasal" (as in /m-m’ e"1/ 'eye') or a realization of the archiphoneme /N/ "nasal" (as in /ar Nma/ 'massagist') or a realization of the archiphoneme ICI "consonantal" (as in /sar Cma/ 'saury'), which means that [m] in Japanese should be regarded as linguistically different in each such case. In other words, the same physical reality ([m]) corresponds to different linguistic functions. It goes without saying that the above-mentioned archiphonemes, i.e. /m-m’/, INI and ICI are associated with three different instances of neutralization, while /m/ is a phoneme which, as such, is unassociated with neutralization. (10) Considerable space (the whole of Chapter 11) has been devoted to discussions of various instances of neutralization in Japanese which are discovered in the course of the commutation test. This is totally justified in functional phonology for the following reason. The concept of neutralization, amongst other functionalist concepts, derives directly from that of opposition. It is phonological oppositions that play the principal role in functional phonology, and the phonemes are only conceivable and identifiable in their capacity as the terms of phonological oppositions (Trubetzkoy 1939, p. 60).132 It is of extreme importance to confirm which of the phonological oppositions of a given language being analyzed are valid in all contexts (i.e. constant oppositions), or valid in some contexts but not valid in other contexts. In the latter case, the phonological oppositions may be neutralizable oppositions (valid in some contexts but not in other contexts) or non-neutralizable cum non-constant oppositions (valid in some contexts but not in other contexts) consequent on the fact that one or more of the terms of phonological oppositions is or are systematically non-occurrent in certain contexts, but not consequent on neutralization. It is very important to sort out individual cases of validity and non-validity of phonological oppositions in different contexts. (11) As will have been clearly shown above, the commutation test is not intended only to elicit the distinctive units of a given language, as often misunderstood, but to carry out other tasks as well, as indicated above. Also, it cannot be overemphasized that the whole operation of the commutation test is based on the concept of opposition.133
12.2
Inventories of all the distinctive units of the second articulation of Japanese and their phonological contents
In mentioning in the preceding section the various tasks the commutation test I carried out in chapters 8 , 9, 10 and 11 has achieved in connection with Japanese, I deliberately did not adduce concrete examples (except in (9)) in order to avoid overburdening the presentation. Indeed I did not need to reproduce the list of all the concrete results of the commutation test which had already been shown here and there in Chapters 8 , 9, 10 and 11 anyway. However, I shall make an exception at this point by reproducing at one go the list of all the distinctive units of the second articulation of Japanese (in both the minority speech and the majority speech), i.e. the phonemes and the archiphonemes of Japanese, together with the specification of their phonological contents. I make this exception, as I shall be concerned with the question of combinabilities between the distinctive units of Japanese in the next chapter, i.e.
Chapter 12: The commutation test reviewed
181
Chapter 13, and provision of the list of the distinctive units of Japanese below will be helpful. That we have already performed the commutation test with the resultant achievements does not mean that we are leaving phonematics behind now, for we shall continue to be in the domain of phonematics in Chapters 13 to 18.
12.2.1
The inventory of the distinctive units of the second articulation of Japanese (in the m in o rity speech)
The following are the distinctive units of the second articulation of Japanese ascribable to the minority speech. 1. The vowel phonemes: /i/, /e/, /a/, /o/, /ui/.
Table 4 The Japanese vowel phonemes (minority speech) hi: Id: laJ: lol: /ui/:
"close front" "mid front" "open" "mid back" "close back"
2. The consonant phonemes: /p/, /p’/, Ibl, lb’ /, /t/, /d/, Ikl, /k7, /g/, /g7, /$/, /