THE ACCENTUAL HISTORY OF

THE JAPANESE AND RYUKYUAN LANGUAGES

- A RECONSTRUCTION -
THE LANGUAGES OF ASIA SERIES
Series Editor : Alexander Vovin
University of Hawai’i at Mānoa

EDITORIAL BOARD
Wolfgang Behr
University of Bochun
Uwe Blæsing
University of Leiden
Stefan Georg
University of Bonn
Toshiki Osada
Research Institute for
Humanity and Nature (Kyoto)
The Accentual History
of the Japanese and
Ryukyuan Languages

A Reconstruction

MORIYO SHIMABUKURO
UNIVERSITY OF THE RYUKYUS
LANGUAGES OF ASIA SERIES
Volume 2
THE ACCENTUAL HISTORY OF THE JAPANESE
AND RYUKYUAN LANGUAGES
A RECONSTRUCTION
by Moriyo Shimabukuro

First published in 2007 by

GLOBAL ORIENTAL LTD
PO Box 219
Folkestone
Kent CT20 2WP
UK

www.globaloriental.co.uk

© Moriyo Shimabukuro 2007

ISBN 978-1-901903-63-8

All rights reserved. No part of this publication

may be reproduced or transmitted in any form or by any
electronic, mechanical or other means, now known
or hereafter invented, including photocopying and
recording, or in any information storage or retrieval
system, without prior permission in writing from
the publishers.

British Library Cataloguing in Publication Data

A CIP catalogue entry for this book is available
from the British Library

Printed and bound in England by TJI Digital, Padstow, Cornwall

To Hitomi and Morito
 CONTENTS

Preface xiii
List of Abbreviations and Notations xv

1 INTRODUCTION 1
1.1 BRIEF INTRODUCTION ON JAPONIC LANGUAGE
HISTORY 1
1.1.1 Proto-Japonic and Proto-Ryukyuan 2
1.1.2 Relation to Social and Political History of Japanese
Archipelago 3
1.1.3 Relation to Other Languages in the Region 4
1.2 CONTROVERSIAL ISSUES IN THE
RECONSTRUCTION OF PROTO-JAPONIC 5
1.2.1 Kindaichi's (1975) or Ramsey's (1979)
Hypotheses 5
1.2.2 Pitch and Vowel Length 7
1.2.3 Devoiced Vowels 8
1.2.4 How Pitch Accent Changes in Japanese 9
1.2.5 Whether Neighboring Languages such as
Korean and Ainu Have Kept Supra-segmental
and Segmental Features of Earlier Stages of the
Japanese Language 9
1.2.6 Whitman's Law 9
1.2.7 Register and Initial Voiced Obstruents 10
1.3 PITCH ACCENT IN RELATON TO MORAS,
SYLLABLES, PITCH, STRESS, AND TONES 11
viii CONTENTS

2 PREVIOUS SCHOLARSHIP: A CRITICAL REVIEW 18

2.1 HATTORI 18
2.2 KINDAICHI 34
2.3 TOKUGAWA 44
2.4 OKUDA 50
2.5 MCCAWLEY 53
2.6 RAMSEY 62
2.7 THORPE 69
2.8 MARTIN 74
2.9 MATSUMORI 78

3 HISTORICAL LINGUISTIC METHODS 88

3.1 KOKUGOGAKU IN JAPAN 88
3.2 WESTERN LINGUISTICS 89
3.3 METHODOLOGY 90
3.3.1 Methods for Reconstruction of Proto-Japonic and
Proto-Ryukyuan Accent 90
3.3.2 Establishment of Accent Classes 91
3.3.3 Filling the Classes with Phonological
Characteristics 92
3.3.4 Explanation of the Development of Modern
Accent with Phonetically Reasonable Sound
Changes 94
3.3.5 Phonological Framework 95

4 ACCENT CHANGE 98
4.1 INTRODUCTION 98
4.2 ACCENT CHANGE IN JAPONIC LANGUAGES 98
4.2.1 Accentual Change in Historical Written Records 99
4.2.2 Accent in Careful Speech and Connected
Speech 108
4.2.3 Accent Shift in Relation to Vowel Devoicing 111
4.2.4 Accent Change in Relation to Vowel Length 112
4.2.5 Accent Shift in Relation to Vowel Quality 113
4.2.6 Accent Conditioned by Syllable Structure 113
4.2.7 Accent in Two Different Generations 114
4.2.8 Overview 114
 CONTENTS ix

4.3 ACCENT CHANGE IN OTHER LANGUAGES 116

4.3.1 Korean 117
4.3.2 African Languages 120
4.3.2.1 Phonetic Tone Rules 120
4.3.2.2 Morphophonemic Tone Rules 125
4.3.2.3 Summary 126
4.3.3 Austronesian Languages 128
4.3.3.1 Bahasa Indonesia 128
4.3.3.2 Mukah 128
4.3.3.3 Sangiric Languages 129
4.3.3.4 Summary 129
4.3.4 Chinese 130
4.3.5 Summary 132

5 RECONSTRUCTION OF PROTO-RYUKYUAN
ACCENT 133
5.1 INTRODUCTION 133
5.2 AMAMI RUKYUAN 134
5.2.1 Description of Amami Ryukyuan Accent 134
5.2.1.1 Naze Dialect 135
5.2.1.2 Kamishiro Dialect 140
5.2.1.3 Kametsu Dialect 145
5.2.2 Correspondences and Reconstruction of
Proto-Amami Accent 149
5.2.3 Development of Accent from Proto-Amami to
Modern Amami Dialects 156
5.2.3.1 Development of Naze Accent 157
5.2.3.2 Development of Kamishiro Accent 158
5.2.3.3 Development of Kametsu Accent 161
5.3 OKINAWA RYUKYUAN 163
5.3.1 Description of Okinawa Ryukyuan 163
5.3.1.1 Nakijin Dialect 163
5.3.1.2 Shuri Dialect 169
5.3.1.3 Aguni Dialect 173
5.3.2 Correspondences and Reconstruction of
Proto-Okinawa Accent 176
5.3.3 Accent and Vowel Length 183
x CONTENTS

5.3.4 Development of Modern Okinawan Accent

from Proto-Okinawa Accent 188
5.3.4.1 Development of Nakijin Accent 189
5.3.4.2 Development of Shuri Accent 191
5.3.4.3 Development of Aguni Accent 194
5.4 MIYAKO RYUKYUAN 196
5.4.1 Description of Miyako Ryukyuan Accent 196
5.4.1.1 Ikema Dialect 197
5.4.1.2 Ôura Dialect 203
5.4.1.3 Uechi Dialect 208
5.4.2 Correspondences and Reconstruction of
Proto-Miyako Accent 212
5.4.3 Development of Modern Miyako from
Proto-Miyako 219
5.4.3.1 Development of Ikema Accent 219
5.4.3.2 Development of Ôura Accent 221
5.4.3.3 Development of Uechi Accent 223
5.5 YAEYAMA RYUKYUAN 226
5.5.1 Description of Yaeyama Ryukyuan Accent 226
5.5.1.1 Ishigaki Dialect 226
5.5.1.2 Sonai Dialect 229
5.5.1.3 Kuroshima Dialect 234
5.5.2 Correspondences and Reconstruction of
Proto-Yaeyama Accent 238
5.5.3 Development of Modern Yaeyama Dialects
from Proto-Yaeyama Accent 245
5.5.3.1 Development of Ishigaki Accent 246
5.5.3.2 Development of Sonai Accent 248
5.5.3.3 Development of Kuroshima Accent 250
5.6 YONAGUNI RYUKYUAN 253
Description of Yonaguni Ryukyuan Accent 253
5.7 CORRESPONDENCES AND RECONSTRUCTION
OF PROTO-RYUKYUAN ACCENT 257
 CONTENTS xi

5.8 DEVELOPMENT OF PROTO-AMAMI,

PROTO-OKINAWA, PROTO-MIYAKO,
PROTO-YAEYAMA, AND YONAGUNI
ACCENT FROM PROTO-RYUKYUAN 270
5.8.1 Development of Proto-Amami Accent 271
5.8.2 Development of Proto-Okinawa Accent 273
5.8.3 Development of Proto-Miyako Accent 276
5.8.4 Development of Proto-Yaeyama Accent 278
5.8.5 Development of Yonaguni Accent 281
5.9 CONCLUSION 283

6 RECONSTRUCTION OF PROTO-MAINLAND
JAPANESE ACCENT 286
6.1 INTRODUCTION 286
6.2 MAINLAND JAPANESE 287
6.2.1 Description of Tokyo Japanese Accent 288
6.2.2 Description of Kyoto Japanese Accent 291
6.2.3 Description of Myôgishô Accent 295
6.2.4 Description of Kagoshima Japanese Accent 298
6.3 CORRESPONDENCES AND RECONSTRUCTION
OF PROTO-MAINLAND JAPANESE ACCENT 302
6.4 DEVELOPMENT OF THE ACCENT SYSTEM
OF MAINLAND JAPANESE DIALECTS FROM
PROTO-MAINLAND JAPANESE 313
6.4.1 Development of Myôgishô Accent from
Proto-Mainland Japanese 314
6.4.2 Development of Modern Kyoto Accent from
Proto-Mainland Japanese 317
6.4.3 Development of Modern Tokyo Accent from
Proto-Mainland Japanese 320
6.4.4 Development of Modern Kagoshima Accent
from Proto-Mainland Japanese 324
6.5 CONCLUSION 327
xii CONTENTS

7 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 329

7.1 INTRODUCTION 329
7.2 CORRESPONDENCES AND RECONSTRUCTION 330
7.3 DEVELOPMENT OF PROTO-RYUKYUAN AND
PROTO-MAINLAND JAPANESE ACCENT FROM
PROTO-JAPONIC ACCENT 340
7.3.1 Development of Proto-Ryukyuan Accent
from Proto-Japonic Accent 341
7.3.2 Development of Proto-Mainland Japanese
Accent from Proto-Japonic Accent 345
7.4 CONCLUSION 348

8 CONCLUSION 350

REFERENCES 353
APPENDIX: JAPONIC COGNATES AND
RECONSTRUCTED FORMS 363
 PREFACE

H
ow does accent change? What were Proto-Japanese, Proto-
Ryukyuan, and Proto-Japonic accents like? This book
searches for answers to these questions.
Applying the comparative method, I have reconstructed the
accentual history of the Japonic languages. The reconstruction is
based on modern dialects of Japanese and Ryukyuan, and also on
historical materials.
The investigation of ‘natural accent changes’ has allowed me to
formalize rules for accent change. Using these rules, I account for
the developments of the accent systems of descendant dialects or
languages. I have also looked into the typology of accent, and
explained how typologically different accent systems develop.
The present book is based, by and large, on my doctoral thesis,
presented at the University of Hawai’i at Manoa. Additional new
material and the revision of other parts of the text have resulted in
the volume’s present form.
When I started going over the manuscript a few years ago, it was
expected that the book would be published within a year or even
within several months. In the event, it has been long in waiting for
its completion.
I am indebted to numerous people who assisted me in various
ways. Without their support, this book could not have been
published.
First, I would like to express my gratitude to the editor,
Alexander Vovin, who gave me an opportunity to publish this
xiv PREFACE

volume within his new series on the Languages of Asia. I am also

very grateful to his invaluable comments and suggestions.
In the writing of my doctoral thesis, I benefited greatly from
discussions with Leon A. Serafim. Without his insightful comments
and suggestions, the thesis could not have reached its completion.
I would also like to thank John Bentley for his careful reading of
the whole manuscript and suggestions for improvement. Needless
to say, any errors made are mine only.
I also wish to thank Paul Nourbury at Global Oriental for his
patience with my delayed submission of the manuscript.
Finally, my warmest thanks go to my dearest wife Hitomi. I
could not have completed this book without her support and
encouragement.
LIST OF ABBREVIATIONS AND NOTATIONS

assml Assimilation
C-smpl Contour Pitch Simplification
D-chng Domain Change
D-smpl Double Accent Simplification
F falling pitch
F-dltn Final Accent Deletion
F-loss Final Accent Loss
H high pitch
I-low Initial Pitch Lowering
I-gain Initial Accent Gain
L low pitch
L-sprd Low Tone Spreading
MK middle Korean
OJ old Japanese
PA Proto-Amami
PK Proto-Korean
PJ Proto-Japonic
PM Proto-Miyako
PMJ Proto-Mainland Japanese
PO Proto-Okinawa
PR Proto-Ryukyuan
PY Proto-Yaeyama
R rising pitch
shift Accent Shift
V-shrt Vowel Shortening
V-lngth Vowel Lengthening
 phonemic pitch fall between syllables or moras
 phonemic pitch fall within a syllable or mora
 phonemic pitch rise between syllables or moras
 phonemic pitch rise within a syllable or mora

x x is high register
_x x is low register
O syllable
o mora
 1

INTRODUCTION

T his introductory chapter gives the reader a brief background on

the Japanese and Ryukyuan languages, including their
linguistic relation to each other as well as to other languages in the
region. Cultural and historical settings are certainly not ignored.
Another important purpose of this chapter is to introduce
controversial issues that need to be kept in mind when one attempts
to reconstruct the Proto-Japonic language. The term ‘Japonic
language(s)’ was coined by Leon A. Serafim and has been widely
accepted among Japanologists; it is meant to indicate the Japanese
language spoken on the main islands of Japan (i.e., Hokkaido,
Honshu, Shikoku, and Kyushu) and the Ryukyuan language on the
archipelago consisting of the three main island groups (i.e., Amami,
Okinawa, and Sakishima) located between Kyushu and Taiwan.
Lastly, this chapter also touches upon theoretical issues in
suprasegmental phonology, and discusses a system of pitch accent
in relation to moras, syllables, pitch, stress, and tones.

1.1 BRIEF INTRODUCTION ON JAPONIC LANGUAGE

HISTORY
This section of the chapter, consisting of three main parts, gives a
brief overview of the history of the Japonic languages. The first
part, introducing different views from the literature, deals with the
question of what Proto-Japanese and Proto-Ryukyuan mean. The
second part discusses the social and political history of the Japonic
languages. The focus of the last part is on the Japonic languages in
relation to other languages in the region.
2 ACCENTUAL HISTORY

1.1.1 Proto-Japonic and Proto-Ryukyuan

According to Hattori (1959), Proto-Japonic was a language spoken
by a people who settled in northern Kyushu (southwestern Japan)
from the Korean peninsula around 300 B.C., and Ryukyuan
branched off from Proto-Japonic before 300 A.D., i.e., no more
than 600 years later. Thorpe (1983) also maintains that Ryukyuan
and Japanese had diverged significantly earlier than the eighth
century A.D. Both of the claims are based on the hypothesis that
Proto-Mainland Japanese and Proto-Ryukyuan split from Proto-
Japonic:

(1) Proto-Japonic

Proto-Ryukyuan Proto-Mainland Japanese

However, another view claims that Proto-Southwestern Japanese

branched off from Proto-Japonic and then Proto-Ryukyuan and
Proto-Southwestern Kyushu developed from Proto-Southwestern
Japanese. Many scholars prefer this view, especially those who
know that there are similarities between Ryukyuan and
Southwestern-Kyushu Japanese. Yukio Uemura (1972) is one of
the proponents of this approach.

(2) Proto-Japonic

Proto-Southwestern-Japanese Proto-else

Proto-Ryukyuan Proto-Southwestern-Kyushu

There is another view proposed by Miller (1971), who argues

that Middle Korean, Old Japanese, and Ryukyuan are sister
languages which split from Proto-Korean-Japanese.

(3) Proto-Korean-Japanese

Middle Korean Old Japanese Ryukyuan

 INTRODUCTION 3

Miller’s proposal is too unlikely to merit careful consideration,

because while the genetic relationship between Japanese and
Ryukyuan is obvious, the one between Korean and Japonic is still
controversial. Therefore, his proposal will not be considered in the
context of this discussion. However, since there is no doubt that
Ryukyuan is a daughter dialect of Proto-Japonic, the Japanese-
Ryukyuan connection will be carefully investigated.

1.1.2 Relation to Social and Political History of Japanese

Archipelago
Hattori (1959) claims that after settling in Northern Kyushu the
Yayoi people later moved to Central Japan, where the old capital
Kyoto is now situated, and also to the Ryukyus. It is known from
archaeological evidence that there were people (bearers of the
Jômon culture) in the Japanese archipelago before the Yayoi people
came. It is believed that the Ainu people were there before the
Yayoi, as well, and they are commonly thought to be descendants
of a northern Jômon people. However, it is not known from when
or where the Jômon people came.
Ryukyuan is a group of dialects spoken by 194,383 inhabitants
of the Ryukyu archipelago (OKCS 1998, KOS 1999).1 The
archipelago consists of three major island groups (Amami,
Okinawa, and Sakisima) comprising approximately 140 islands,
only 47 of which are populated (OPOCVB 1996). It is 500 miles
long and is located between Kyushu to the northeast and Taiwan to
the southwest. What eventually became the Ryukyu civilization
formed into an independent kingdom taking advantage of two
neighboring countries, China and Japan. However, in 1609, the
kingdom fell into the hands of the Shimazu clan of Kagoshima,

1
The accurate number of native Ryukyuan speakers is not known. The number
(194,383) is the total number of 65 year-old or older inhabitants in the Ryukyus,
assuming that native inhabitants who are 65 years old or older are able to speak
Ryukyuan: 163,230 in Okinawa prefecture (OKCS 1998) and 31,153 in Amami
(KOS 1999). Note that the number 194,383 still does not reflect the actual number
of native Ryukyuan speakers because quite a number of non-Ryukyu-born people
have migrated to the Ryukyus.
4 ACCENTUAL HISTORY

Kyushu, though it retained much of its internal autonomy. When

the Japanese government was centralized and the clans were
abolished in the second half of the nineteenth century, the
inhabitants of the Ryukyu islands were educated in Standard
Japanese in schools, and discouraged from using their own
language. A group of islands in the southern part of the Ryukyuan
chain became a prefecture of Japan in 1879, known as Okinawa.
WWII brought an end to this status, but it was regained after the
return of the islands to Japan from the post-war American
occupation, in 1972. The Amami islands in the northern part of the
island chain, part of Kagoshima Prefecture, had been returned to
Japan, much earlier, in 1953.

1.1.3 Relation to Other Languages in the Region

There have been a number of proposals on the genetic affiliation of
the language of mainland Japan to other languages. Among them,
the genetic relationship with Ryukyuan is the only one on which all
linguists agree. According to those who specialize in Japanese
linguistics, Ryukyuan is considered a dialect of Japanese. Other
than the proposal on the genetic relationship between Ryukyuan
and Japanese, the most widely debated and persuasive are those
which relate Japanese to the Altaic family, including Korean.
Scholars such as Martin (1966), Ramstedt (1968), and Whitman
(1985) have attempted to demonstrate a genetic relationship
between Japanese and Korean.
Another often-debated hypothesis is the Southern theory.
Proponents of the Southern theory claim that Japanese is a mixed
language related to both Altaic and Austronesian. According to
Murayama (1985), a proponent of the mixed-language hypothesis,
Japanese verb roots are of either Altaic or Austronesian origin,
while inflectional endings are Altaic.
A major problem faced by scholars investigating the genetic
relationship of Japanese to other languages, other than Ryukyuan,
has to do with the difficulty in establishing cognate sets. For
example, Martin (1966) compares 320 possible cognates between
Japanese and Korean and reconstructs their proto-forms, yet he
 INTRODUCTION 5

limits himself, strictly speaking, to “twenty items that show the

proper correspondences to be cognates and about which we have
little doubt” (Martin 1966: 196-7).

1.2 CONTROVERSIAL ISSUES IN THE

RECONSTRUCTION OF PROTO-JAPONIC
1.2.1 Kindaichi’s (1975) and Ramsey’s (1979) Hypotheses
In order to reconstruct the accentual history of Japanese and
Ryukyuan, the controversial issue of whether Tokyo-type or Kyoto-
type accent is more conservative cannot be avoided (see 1.3. for my
definition of accent - locus and register). Choosing one view or the
other is crucial to a reconstruction of the accentual history because
choosing one over the other will lead us to a completely different
reconstruction of the history. In other words, what matters here is
how to account for the geographical distribution of accentual types.
In Japanese and Ryukyuan, there are four main types of accent:
Tokyo-type, Kyoto-type, Kagoshima-type, and accentless-type (no
distinctive accent). These are distributed over the islands of Japan
as shown in (4). In the center of Japan, the dialects with Kyoto-type
accent are spoken and are surrounded by the dialects with Tokyo-
type accent. Next to the area where the Tokyo-type dialects are
spoken, the dialects with no distinctive accent are distributed.
Kagoshima-type accent is seen in the southwestern area of Kyushu,
and extended to the islands of the Ryukyus, which are in (4)
magnified for convenience.
6 ACCENTUAL HISTORY

(4) Geographical distribution of accentual types

 INTRODUCTION 7

If the Tokyo-type accent is more conservative, as maintained by

Ramsey, the distribution of Tokyo-type and Kyoto-type dialects
can be explained by the wave theory. That is to say, accentual
changes started in the area where Kyoto-type dialects are spoken,
and spread outward.

(5) Ramsey’s hypothesis

Tokyo-type Kyoto-type Tokyo-type

old new old

On the other hand, if the Kyoto-type accent is more conservative,

as supported by Kindaichi, it has to be assumed that the same
accent changes occurred in two different areas surrounding the
Kyoto-accent area:

(6) Kindaichi’s hypothesis

Tokyo-type Kyoto-type Tokyo-type

new old new

There are some advantages and disadvantages in both

hypotheses. However, overall, Ramsey’s has more problems that
are difficult to explain. For instance, Ramsey proposes leftward
accent shift for Japanese, but in universally-agreed-upon cases in
Japanese and Ryukyuan, accent shifts rightward; leftward accent
shift is not found in these languages (see chapter 4). Therefore, I
will follow Kindaichi’s hypothesis. However, this does not mean
that I agree with all of Kindaichi’s proposals.

1.2.2 Pitch and Vowel Length

There is a correlation between initial-syllable vowel length and
initial low pitch in Japanese and Ryukyuan. Martin (1987: 247)
suggests that there may have existed a correlation between initial-
syllable vowel length and initial low pitch (called low register) in
Japanese. He points out that some Ryukyuan disyllabic nouns
8 ACCENTUAL HISTORY

belonging to historical accent class 2.3 (LL), 2.4 (LH), and 2.5 (LF)
contain an initial long vowel, and that their initial-low pitch in
modern Japanese is a reflex of original Proto-Japonic initial-
syllable vowel length. However, not all nouns in these historical
accent classes show vowel length, that is to say, most of them lost
Proto-Japonic vowel length, but kept initial low register, e.g., duru
LL ‘mud’ in Shuri and du:ru LLL ‘mud’ in Ôhama (Ishigaki
Ryukyuan). This necessitates intensive study of Ryukyuan
suprasegmental features.

1.2.3 Devoiced Vowels

Development of the accent system of the Japanese and Ryukyuan
languages can be considered in relation to devoiced vowels. Nouns
with an initial devoiced vowel belonging to Proto-Japonic accent
type A (initial-high register) are type B (initial-low register) in
Ryukyuan dialects such as Aden, Nakijin, Hateruma, Shiraho,
Miyako, and Yonaguni. Shimabukuro (1998a) has claimed that
accent shift from A to B in Ryukyuan was caused by a phonetically
devoiced vowel in the first syllable existing in Proto-Ryukyuan,
and that the shift did not occur in Ryukyuan dialects such as Shuri
and Hentona, which lost the process of devoicing.
Finding correlations between suprasegmentals and segments or
between suprasegmentals and other suprasegmentals is important
for reconstructing a language because archaic forms are often kept
in either a segment or a suprasegmental after a certain correlation
between the segment(s) and suprasegmental(s) is lost, and a history
of the language can be reconstructed by recovering this correlation.
In the case of Japanese and Ryukyuan, there appear to be three
kinds of correlations between suprasegmentals and segments or
other suprasegmentals: one between accent shift and devoiced
vowel (see previous paragraph), one between voicing in initial
consonant and initial pitch height (see 6 in this section and also
footnote 6), and one between low register and vowel length (see 2
in this section). An examination of these correlations will help in
reconstructing the accentual system of Proto-Japonic and Proto-
Ryukyuan.
 INTRODUCTION 9

1.2.4 How Pitch Accent Changes in Japanese

This study will also look into other pitch-accent languages such as
Korean and Ainu, and tone languages in Africa and Southeast Asia,
in order to elucidate natural tendencies in pitch change. It is
important to investigate such tendencies because these help in
pitch-accent-system reconstruction.

1.2.5 Whether Neighboring Languages such as Korean and

Ainu Have Kept Suprasegmental and Segmental Features of
Earlier Stages of the Japanese Language
Since these languages have a history of contact with Japanese, they
may retain, in loanwords, features Japanese has lost. For example,
Vovin (1993b: 130) reconstructs PJ *tu:tu[-Ci] LLH ‘hammer’,
with initial-syllable vowel length, on the basis of the fact that the
Ainu borrowing *tu:ti LLH from Japanese has initial vowel length.
He suggests that the Ainu word is an early loan from Eastern Old
Japanese. If neighboring languages have kept linguistic features
that Japanese lost, then they would certainly contribute to a
reconstruction of the Japanese language.

1.2.6 Whitman's Law -- a problem occurring in analyses

relating to Whitman’s /-r-/ deletion rule and initial-syllable
long low-pitched vowels
What has come to be known as Whitman’s Law (Whitman 1985:
21-4, 190-8) states that /-r-/ is deleted when a preceding vowel is
short. However, there are a number of initially low-pitched nouns
with the syllable structure (C)Vr- in Old Japanese. If Whitman’s
hypothesis is correct, it is possible that those nouns originally had a
long vowel before the /-r-/ and that somehow the long vowel has
been shortened, but the word has kept the word-initial low register.
Therefore, by this logic, the word duru (LL) ‘mud’ in Shuri, for
example, was *do:ro (LLL) in Proto-Japonic. Whitman’s Law
plays a crucial role in the reconstruction of the accentual systems of
the Japanese and Ryukyuan languages, because it involves vowel
length, which is claimed to have a correlation with initial low
10 ACCENTUAL HISTORY

register. Therefore, the relation between Whitman’s Law and vowel

length needs to be further investigated.

1.2.7 Register and Initial Voiced Obstruents -- whether the

origin of accentual register in Japanese has to do with a voicing
distinction in word-initial obstruents
Through comparison with Proto-Altaic, Vovin (1997) proposes the
hypothesis that a register system in Japanese developed in
conjunction with the loss of contrast between initial voiceless and
voiced obstruents in Proto-Japonic. That is to say, words with a
Proto-Japonic initial voiced obstruent came to be initial low-
register words and ones with a Proto-Japonic initial voiceless
obstruent became initial high-register words as the distinction of
voicing in initial obstruents was lost.

(7) PJ low register < Proto-Altaic initial voiced obstruents

(*b, *d, *j, *g)
PJ high register < Proto-Altaic initial voiceless obstruents
(*ph, *p, *th, *t, *ch, *c, *kh, *k)

However, there are a number of exceptions, the following are

some examples:

(8) Proto-Japonic Proto-Altaic

(a) *da H ‘arrow’ *da ‘arrow’
(b) *ta L ‘rice-field’ *thala ‘field’

According to Vovin (1997), exceptions like (8a) might have lost an

initial laryngeal, that is to say, *Lda H (with L indicating a
laryngeal) became *da H. Vovin also says that exceptions such as
that in (8b) might have lost vowel length -- which Martin (1987:
247) claims existed in relation to low register in the Japanese and
Ryukyuan languages -- and have kept the low pitch. Vovin’s
hypothesis requires further research.
 INTRODUCTION 11

1.3 PITCH ACCENT IN RELATION TO MORAS,

SYLLABLES, PITCH, STRESS, AND TONES
In pitch-accent systems, phonemic pitch change can occur
anywhere within a string, either within or between syllables/moras.
For example, in Kyoto the word for ‘window’ is mado LF, in which
pitch falls within the second syllable (= mora). In the word hasi HL
‘bridge,’ pitch falls between the first and the second syllable (=
mora).
According to Haraguchi (1977: 340-3), both syllable and mora
can be the accent-bearing unit and tone-bearing unit (TBU)
depending on the language. Namely, what bears accent and/or bears
tone is determined by each language. The tone-bearing unit is not
necessarily also an accent-bearing unit. For example, in Tokyo,
tone-bearing units are moras, but accent-bearing units are syllables,
while in Kyoto, both tone-bearing units and accent-bearing units
are moras. In other words, in Tokyo there is no contrast between
CVV˘CV (HHL) and CV˘VCV (HLL) or between CVN˘CV and
CV˘NCV; pitch always falls after the first mora in a syllable that
contains two moras, if the syllable is accented. However, in Kyoto,
any mora can be accented. Notice also that in both of these dialects
each mora is linked to a tone. (Syllable boundaries are indicated by
a dot.)

(9) (a) Tokyo dialect

ni.hon ‘Japan’ ryoo.ri ‘cooking’

| | | | | |
L HL HLL

(b) Kyoto dialect

on.na ‘woman’ on.do ‘temperature’

|/ | | \ |
H L H L
12 ACCENTUAL HISTORY

However, phonological duration is measured in terms of the

mora. For example, haiku, Japanese poetry, consists of three lines
of 5, 7, and 5 moras.
Inaba (1996: 135) says that rhythm and pitch are independent of
each other in Japanese because “rhythm involves timing with the
alternation of strong and weak while pitch involves frequency with
patterns like LHHHL, LHH, and LHL. The point of contact
between rhythm and pitch is the location where the last foot-head
meets the last high pitch.” In the following example ‘<ex>’
indicates extrametricality (‘TJ’ is Transitional Juncture, which is
employed to meet the binary foot requirement,2 and ‘/tó/’ is an
accented syllable):

(10) Rhythm: <ex> (+ TJ) (+ -) (+ - -)

ya ma ho to tó gi su ‘cuckoo’
Pitch: L H H H H L L

With regard to differences between pitch accent, tone, and stress

languages, general characteristics of these three systems will be
briefly discussed below.
First, a pitch-accent language is one whose lexicon specifies
either the syllable on which an accent occurs or only whether a
word is one type, not the other, when there are only two types of
accent. For instance, in Tokyo Japanese, which is generally
considered to be a prototypical pitch-accent language (McCawley
1978), there are two types of nouns: accented and unaccented. In
this book, the term ‘accent’ denotes a distinctive fall or rise in pitch,
indicated by ‘˘’ or ‘È’ respectively. The following examples are
from Tokyo Japanese and Sonai (Iriomote) Ryukyuan.

2
A foot, or rhythmic unit, is required to have two moras in Japanese. Thus, ma
alone cannot be a foot. Therefore, ‘TJ’ is filled in with ma to consist of a foot.
 INTRODUCTION 13

(11) Tokyo Sonai

accented: ‘flower’ hana˘ LH3 paÈna LH
unaccented: ‘nose’ hana LH pana LL

As far as nouns in Tokyo and Sonai are concerned, without

knowledge of the existence and location of accent, the pitch shape
of the nouns is unpredictable. Thus, both the existence and location
of the accent need to be indicated in the lexicon. However, the
pitch shape of verbs and adjectives in Tokyo is predictable if we
know that a given verb or adjective is accented. There are, for
example, two pitch patterns for disyllabic verbs in Tokyo: HL for
accented and LH for unaccented. The verb meaning ‘tear (paper or
cloth)’ is accented, so it is sa˘ku, and the verb meaning ‘bloom’ is
saku LH because it is unaccented. Therefore, for verbs and
adjectives the location of accent does not need to be specified, but
the information on whether or not they are accented must be
indicated in the lexicon.
At the phonetic level, in pitch-accent languages not all words
have accent, while in stress-accent languages all lexical words have
stress, and in tone languages most lexical words have tone(s) (see
discussions below for more details on stress-accent and tone
languages). According to Onseigaku dai-jiten [A dictionary of
phonetics] (1976), there are Japanese dialects lacking a distinctive
pitch accent, such as spoken in Miyagi, the southern part of
Yamagata, Fukui, Hachijô-jima, etc. In those dialects, a given word
can be pronounced in many different ways because pitch is not
distinctive. For example, atama ‘head’ can be pronounced with
LHL, HHL, HLL, LLL, or LHH (cf. LHH in Tokyo), without
changing the meaning of the word. It is widely believed that accent
systems have become less complex in the history of Japonic
languages (Kindaichi 1960). Therefore, the modern Japonic non-

3
Pitch falls after the word. For example, when followed by the particle ga (i.e.,
hana-ga), the pitch is LH-L. Regarding the initial low pitch, in Tokyo Japanese its
word-initial pitch is always low unless accented (see chapter 6).
14 ACCENTUAL HISTORY

pitch-accent languages must have lost their pitch distinction in the

course of their development.
A system of pitch accent is potentially made up of two
typologically different sub-systems. One is locus accent, and the
other is initial-register accent. For example, Tokyo dialect has locus
accent, where pitch fall is distinctive, e.g., atama LHH(L) ‘head’ vs.
kokoro LHL(L) ‘heart’ (Hirayama 1960). Shuri Ryukyuan has
distinctive initial-register accent, where there are only two accent
classes: initial high (“A”) and initial low (“B”), e.g., hana A ‘nose’
vs. hana B ‘flower’ (Kokuritsu Kokugo Kenkyûjo 1963).
Furthermore, both locus accent and initial-register accent can
coexist in one pitch-accent system. Hyôgo dialect, for example,
distinguishes not only kaki HH ‘persimmon’ and hasi HL ‘bridge’
by means of locus accent, but also azuki HHL ‘red bean’ and
kabuto LHL ‘helmet’ by initial-register accent (McCawley 1968).
In tone languages, most words require an underlying tone or
tones, and each syllable must be specified for a tone at the
phonemic level. For example, in Mende, a language of Sierra
Leone, the words nikili ‘groundnut’, nyaha ‘woman’, and mba
‘companion’ have underlying tones or a tone melody LHL (Leben
1978: 186). At the phonetic level, nikili, nyaha, and mba are LHL,
LF, and RF respectively.

(12) (a) /nikili/ ni ki li LHL

‡ | | |
LHL L H L

(b) /nyaha/ nya ha LF

‡ | |\
LHL L HL

(c) /mba/ mba RF

‡ /|\
LHL LH L
 INTRODUCTION 15

McCawley (1968) points out that unlike Japanese, in Chinese the

information about a syllable does not give any information on the
other syllables in the words, except the fact that “third tone cannot
be followed by another third tone.” In Japanese, that one syllable
has ‘accent’ generally implies that the other syllables in the words
are not accented.
Note that Haraguchi (1977) analyses Japanese as a tone language.
In his analysis, in addition to underlying tones, accented words
need to specify the location of accent in the lexicon so that the
words can generate the ‘right’ pitch pattern. Although tone
languages generally specify only underlying tone(s) in their lexicon,
there are some tone languages (which have been called ‘pitch-
accent’ languages) which require a lexical ‘prelinked’ tone. That is
to say, the location of a certain tone link to a segment is specified
in the lexicon. The following example shows that H in HL has to
link to an accented syllable,4 indicated by /ó/, and a rule lowers the
initial syllable (Haraguchi 1977: 9-17). Compare (13) with the
examples in (12).

(13) kokoro ‘heart’ in Tokyo Japanese

/kokóro/ kokóro kokóro kokóro

‡ | | ‡ \ | | ‡ | | |
HL HL H L L HL

Regarding pitch-accent languages with no phonemic accent, the

Ôgami-island dialect of Miyako Ryukyuan, for example, has only
one phonetic pitch shape for all words, e.g. pa: LL ‘leaf’, pana LL
‘flower’, mitum LLL ‘woman’ (Hirayama et al. 1967: 31-32). In
this language, words do not need to specify underlying tone at all
because the pitch shape of the words is completely predictable.

4
‘Accented syllable’ in Haraguchi (1977) simply means the syllable to which the
H tone has to link.
16 ACCENTUAL HISTORY

On Haraguchi’s autosegmental analysis of Japanese pitch accent,

Vance questions the appropriateness of an autosegmental analysis
of the accent. He says:

[O]ne of the characteristics of African languages that

motivates the autosegmental approach is the crowding of n [(n
is any positive integer)] tones onto a word with fewer than n
syllables. In Standard Japanese, however, whenever the H and
L of the melody are crowded onto a single mora, Haraguchi’s
tone simplification rule [50] eliminates the contour tone. In
other words, Japanese does not work the way we expect a
language with relatively autonomous tones to work. (Vance
1987: 105)

Furthermore, more importantly, Haraguchi’s analysis obscures the

crucial distinction between locus and initial-register accent in the
Japonic languages.
According to Hyman (1975), a stress language is a language
whose lexicon specifies at most the syllable on which an accent
occurs. However, in some stress languages such as Turkish (stress
on the final syllable), Polish (stress on the penultimate), Hungarian
(stress on the initial syllable), the stressed syllable is predictable.
Therefore, in the lexicon of those languages, there is no
specification of stressed syllable. In addition, there is only one
primary accent per word, but there can be secondary accent.
However, there is also secondary accent in some pitch-accent
languages. For instance, according to Kindaichi (1975: 223), in the
Goka dialect of Iki-island, sakana ‘fish’ is pronounced MLH(L)
(pitch falls after the first mora and after the third mora).5 It also
seems to be true that all lexical words in stress languages have at
least one stressed syllable, but not all words in pitch-accent
languages need to have accent, e.g. sakana ‘fish’ (i.e., with no
accent) in Tokyo Japanese.

5
‘M’in MLH(L) stands for a mid-high pitch.
 INTRODUCTION 17

In a stress language, the presence or absence of stress can cause

segmental changes. For example, vowel reduction when a vowel is
not stressed (e.g., English).

(14) Vowel reduction in English (connected speech)

university [junIvŒ®sItI] ‡ [junvŒ®sItI] ‡ [junvŒ®sRI]

In a tone language, voicing in consonants causes changes in pitch

height6 (e.g., Mandarin Chinese). In a pitch-accent language, both
accent and segments affect the pitch shape of a word. For example,
Tokyo Japanese has a process of vowel devoicing,7 but a vowel in
accented syllables usually does not undergo the process (Han 1962:
81). However, according to McCawley (1977: 266), voiceless
vowels can shift an accent to the right.

(15) /ti8˘kaku/ ‡ ti8ka˘ku ‘near’

/hu8˘kaku/ ‡ hu8ka˘ku ‘deep’

Yet, in contrast, Vance (1987: 50) gives the example aki8˘kaze

‘autumn wind’, where the devoiced syllable ki carries accent.
According to Cheng (1973), stress and tone are not mutually
exclusive because stress exists in at least some tone languages. For
example, in Mandarin Chinese, weakening in unstressed syllables
occurs, e.g., /lí pa/ ‡ [lí b´] ‘fence’ where the initial syllable is
phonetically stressed.

6
Vovin (1997) claims that the contrast between initial voiceless and voiced
obstruents, which is Altaic in origin, existed in Proto-Japonic, and word-initial
voiced obstruents historically lowered the initial pitch of words. If this is true,
voicing in consonants would also affect pitch height in pitch-accent languages.
7
In Tokyo Japanese, high vowels (i and u) are devoiced when followed by a
voiceless consonant and before a morpheme boundary, if the word is tonic and the
preceding syllable has a voiced vowel (Martin 1987: 76).
 2

PREVIOUS SCHOLARSHIP:
A CRITICAL REVIEW

T his chapter will critically review previous major studies on the

accentual history of the Japonic languages, such as: Hattori
(1933a, 1937, and 1951), Kindaichi (1975 [1954]), Tokugawa
(1978 [1962]), Okuda (1975), McCawley (1977), Ramsey (1979),
Thorpe (1983), Martin (1987), and Matsumori (1993 and 1998). All
of the studies involve a reconstruction of history and pre-history of
accent systems in the Japanese language and/or the Ryukyuan
language.

2.1 HATTORI
As far as the accentual history of the Japonic languages is
concerned, Hattori’s series of articles in Hôgen (1931-33a) are the
first comparative studies on accent in Japanese dialects. In these
articles, Hattori points out regular correspondences in accent
between several dialects of Japanese, yet does not reconstruct a
proto-accent of the language. His first attempt to reconstruct a
system of the proto-Japanese accent appeared in Genshi nihongo no
ni-onsetsu meishi no akusento [The accent of disyllabic nouns in
proto-Japanese] in 1937. In this article, Hattori reconstructs an
accent system of Proto-Mainland Japanese only for disyllabic
nouns. Hattori demonstrates how accent classes can be
reconstructed by using the comparative method.
In Hattori’s system, there are five accentual distinctions as
shown in (1). His system is, in a way, very similar to those of
Myôgishô, Kyoto-type, i.e., 2.1 HH(H), 2.2 HL(L), 2.3 LL(L), 2.4
LH(H), and 2.5 LH(L). That is, there are five accentual distinctions,
 CRITICAL REVIEW 19

and classes 2.1-2 are high-initial, and classes 2.3-5 are low-initial.
Note that Hattori’s class 2.4 is identical to that of Myôgishô. The
difference is that there are rising and falling pitches in Hattori’s
reconstruction.

(1) Proto-Mainland Japanese accent for disyllabic nouns (Hattori

1937: 47-48)

2.11 *HH(F)2
2.2 *HF(L)
2.3 *LL(R)
2.4 *LH(H)
2.5 *LH(F)

Comparing Tokyo-type accent with Kyoto-type accent, Hattori

pointed out that the location of accent in Tokyo-type dialects is one
syllable right of the location of accent in Kyoto-type dialects. See
the following example.

(2) Kôchi (Kyoto-type) Tokyo

o˘tiru3 oti˘ru ‘fall, drop’

Based on this, Hattori (1933b) claimed that Tokyo-type accent

derived from Kyoto-type accent by shifting accent rightward.4 That

1
Numbers indicate corresponding accent classes of the Ruijumyôgishô (also called
Myôgishô), a Chinese-Japanese dictionary written in the eleventh century. This is
one of the oldest existing written records with accent marks. More specifically, the
first number indicates the number of syllables, and the last number shows an
accent class. Therefore, ‘2.1’ means Myôgishô disyllabic accent class 1. Similarly,
1.2 and 3.2 respectively indicate Myôgishô monosyllabic accent class 2 and
Myôgishô trisyllabic accent class 2.
2
A pitch in parentheses is for a following particle.
3
Following Shibatani’s phonemic analysis (1990), the phoneme /t/ is used, which
becomes palatalized to [c&] before /i/.
4
‘Tokyo-type’ and ‘Kyoto-type’ refer to modern Tokyo-type and modern Kyoto-
type respectively throughout this book. The Japanese dialect used in Myôgishô is
traditionally considered to be an earlier dialect of Kyoto Japanese.
20 ACCENTUAL HISTORY

is, Hattori’s hypothesis is completely the opposite of Ramsey’s (see

1.2 for details).
However, Hattori was never satisfied with his hypothesis as in
(3a) that Kyoto-type is more conservative than Tokyo-type accent
because it contradicts the fact that, as shown in (3b), newer lexical
items are found in Kyoto-type dialects and more archaic words are
in Tokyo-type dialects. (See Yanagita 1930 [1927] for details.)
However, it is worth mentioning that under Ramsey’s hypothesis
(see 1.2), the distribution of accent types and that of lexical items
match.

(3) (a) Geographical distribution of accent types

Tokyo-type Kyoto-type Tokyo-type

new old new

(b) Geographical distribution of lexical items

Tokyo-type Kyoto-type Tokyo-type

old new old

To avoid this contradiction, Hattori (1937) proposes a new

hypothesis that both Tokyo-type and Kyoto-type dialects developed
from a proto language different from either of them.

(4) Proto accent

Tokyo-type Kyoto-type

Hattori also revises the reconstruction of the accent system for

disyllabic nouns as below. Compare it with the earlier one in (1).
There are only three classes in (5) because Hattori claims that in
order to reconstruct proto classes for 2.4 and 2.5 classes he needs
the data from the Ryukyuan language. For the first three classes,
the only difference between (1) and (5) is in 2.3; *LL(R) in (1) and
*LH(L) in (5).
 CRITICAL REVIEW 21

(5) Proto-Mainland Japanese accent for disyllabic nouns (Hattori

1937: 51)

2.1 *HH(F)
2.2 *HF(L)
2.3 *LH(L)

Note that *LH(L) is the same as that of modern Tokyo-type

dialects. Hattori argues that a crucial advantage of the new
hypothesis over the old one is that the new one is in accordance
with the distribution of lexical items. That is to say, Tokyo-type
accent is more similar to Proto-Japonic accent than Kyoto-type
accent is. Therefore, (3a) should also be revised as follows:

(6) Distribution of accent types

Tokyo-type Kyoto-type Tokyo-type

old new old

Although Hattori contends that the system of Tokyo-type accent

is more conservative than that of Kyoto-type, it is not clear on what
he bases his claim. The following are the accent systems of Tokyo
and Kyoto dialects with diachronic changes, taken from Hattori
(1937: 52). Throughout this book, diachronic changes are indicated
by ‘>’, e.g. *X > Y; X changes to Y and ‘=’ indicates no change,
e.g. *Z = Z; Z remains the same.
A comparison of 2.1 and 2.2 classes in both (7a) and (7b) below
demonstrates that there are more changes in Tokyo than in Kyoto
(four changes vs. two). However, for 2.3, there is no change in
Tokyo, but there is a series of three changes from a proto form to a
modern form in Kyoto. If there is a way to determine which system
is more conservative, it should be presented. Furthermore, without
reconstructing a whole accent system, it would be quite difficult to
argue about the archaism of one system in comparison with another.
22 ACCENTUAL HISTORY

(7) (a) Tokyo dialect

*2.1 HH(F) > HH(H) > LH(H)
*2.2 HF(L) > HH(L) > LH(L)
*2.3 LH(L) = LH(L)

(b) Kyoto dialect

*2.1 HH(F) > HH(H)
*2.2 HF(L) > HL(L)
*2.3 LH(L) > LL(H) > HL(H) > HL(L)

In Ruijumyôgishô [A collection of words by classes], the pitch of

each syllable is marked in accordance with a Chinese system of
marking tones for syllables. In this system, dots are placed next to
the character; a dot placed at the lower left corner (•®) represents
the ‘even tone (E)’, a dot at the upper left corner (•®) represents
the ‘rising tone (R)’, a dot at the upper right (®•) indicates the
‘departing tone’, and a dot at the lower right (®•) denotes the
‘entering tone.’5 Of the four tones, only the first two are regularly
used to show phonological distinctions.

(8) Myôgishô accent classes and their pitch shapes

Disyllabic nouns Trisyllabic nouns
Accent in Traditional Accent in Traditional
Myôgishô interpretation Myôgishô interpretation
RR HH RRR HHH
RE HL RRE HHL
EE LL REE HLL
ER LH ERR LHH
ERE LHL
EER LLH
EEE LLL

5
Ramsey (1979: 161) says: “The [‘]entering tone[‘] represented in Chinese not a
distinctive pitch contour, but rather meant that the syllable ended in a -p, -t, or -k
stop; in Japanese language texts the use of this [‘]tone[‘] was rare and
inconsistent.”
 CRITICAL REVIEW 23

It is widely accepted in the literature that R and E are high pitch

and low pitch respectively. For example, a disyllabic word marked
with RR is HH pitch, a word with RE is HL, a word with EE is LL,
and so on. According to Hattori, there are four pitch shapes for
disyllabic words and seven for trisyllabic words in Myôgishô. In (8),
disyllabic examples are extracted from Kindaichi (1937) and
trisyllabic examples are from Ôhara (1942).
In his article (1951), pointing out that in Kyoto Myôgishô EEE
and Myôgishô RRE merged into HHL, Hattori suggests that
Myôgishô EEE be ‘mid falling pitch’ (chûchûge-gata), namely,
MML.

(9) Myôgishô Kyoto

EEE > HHL
RRE >

Therefore: Myôgishô Kyoto

MML > HHL
HHL >

Not: Myôgishô Kyoto

LLL > HHL
HHL >

His reasoning is that falling pitch easily merges with another falling
pitch, but merging of level pitch and falling pitch is unnatural.
Moreover, Hattori also suggests MLH for Myôgishô EER and ML
for Myôgishô EE by referring to Kindaichi’s claim that Myôgishô
EE and Myôgishô RE merged into HL in the Kamakura period
(1192-1333 A.D.).

(10) Myôgishô Kyoto

EE > HL
RE >
24 ACCENTUAL HISTORY

Therefore: Myôgishô Kyoto

ML > HL
HL >

Not: Myôgishô Kyoto

LL > HL
HL >

Therefore, Hattori’s view on Myôgishô classes and their pitch

shapes is as follows:

(11)
Disyllabic nouns Trisyllabic nouns
Accent in Hattori’s Accent in Hattori’s
Myôgishô interpretation Myôgishô interpretation
RR HH RRR HHH
RE HL RRE HHL
EE ML REE HLL
ER LH ERR LHH
ERE LHL
EER MLH
EEE MML

In his article Genshi nihongo no akusento [The accent of Proto-

Japonic language] (1951), Hattori expands his reconstruction
covering monosyllabic, disyllabic, and trisyllabic words. This
article includes two sets of reconstructions of the Proto-Japonic
accent system. Hattori reconstructs a Proto-Japonic accent system
and later in the article he revises it. The revised version is more
complex than the earlier one through adding alternative pitch
shapes.
 CRITICAL REVIEW 25

(12) Hattori’s reconstruction (1951)

First Reconstruction
Monosyllabic Accent
*H(H) (1.1) > Tokyo-type L(H)
(e.g., ti ‘blood’)
*H(L) (1.2) > H(H) > Tokyo-type L(H)
(e.g., na ‘name’)
*L(H) (1.3) > F(H) > Tokyo-type H(L)
(e.g., ki ‘tree’)
*RF.(L)6 (1.3) > Tokyo-type H(L)
(e.g., ha ‘tooth’)

Disyllabic Accent
*HH(H) (2.1) > Myôgishô HH(H) > Kyoto HH(H)
(e.g., usi > Tokyo LH(H)
‘cow’) > Morioka LL(L)
> Takamatsu HH(H)
> Kagoshima LH(L)

*HF(L) (2.2) > Myôgishô HL(L) > Kyoto HL(L)

(e.g., isi > Tokyo LH(L)
‘stone’) > Morioka LL(L)
> Takamatsu HL(L)
> HH(L) > Kagoshima LH(L)

*LH(L) (2.3) > Myôgishô ML(H) > Kyoto HL(L)

(e.g., asi > Tokyo LH(L)
‘foot, leg’) > Morioka LH(L)
> Takamatsu HH(H)
> Kagoshima LL(H)

6
The pitch ‘RF.’ occurs within a syllable and a dot following ‘RF’ indicates a
syllable boundary.
26 ACCENTUAL HISTORY

*LH(H) (2.4) > Myôgishô LH(H) > Kyoto LL(H)

(e.g., iki > ML(H) > Tokyo HL(L)
‘breath’) > Morioka HL(L)
> Takamatsu LL(H)
> Kagoshima LL(H)

*HL(L) (2.5) > Myôgishô LH(L) > Kyoto LH(L)

(e.g., aki > Tokyo HL(L)
‘autumn’) > Morioka HL(L)
> Takamatsu LF(L)
> LH(L) > Kagoshima LL(H)

Trisyllabic Accent
*HHH(H) (3.1) > Myôgishô HHH(H) > Kyoto HHH(H)
(e.g., kuruma > Tokyo-type
‘cart’) LHH(H)7

*HHF(L) (3.2) > Myôgishô HHL(L) > Kyoto HLL(L)

(e.g., tobira > LHF(L) > Tokyo-type
‘door’) LHH(H)

*HFL(L) (3.3) > Myôgishô HFL(L) > Kyoto HLL(L)

(e.g., tikara > LHL(L) > Tokyo-type
‘strength’) LHL(L)8

*LLH(L) (3.4) > Myôgishô MML(H) > HHL(L) > Kyoto

(e.g., kagami HLL(L)
‘mirror’) > LLH(L) > Tokyo-type
LHH(L)

7
For trisyllabic examples, Hattori (1951) gives the pitch patterns of Tokyo-type
(Chûgoku) dialects, not of Tokyo dialect.
8
In Tokyo, the pitch for the word ‘strength’ is LHH(L) (Kindaichi 1966, Nihon
Hôsô Bunka Kenkyûjo 1971). Cf. tika˘ra LHL in Hiroshima (Hirayama 1960).
 CRITICAL REVIEW 27

*LLH(H) (3.5) > Myôgishô MLH(H) > Kyoto HLL(L)

(e.g., abura > Tokyo-type
‘oil’) LHH(H)

*LHL(L) (3.5) > Myôgishô MLH(L) > Kyoto HLL(L)

(e.g., inoti ‘life’) > Tokyo-type
LHL(L)9

*LHH(H) (3.6) > Myôgishô LHH(H) > Kyoto LLL(H)

(e.g., nezumi > Tokyo-type
‘mouse’) LHH(H)

*LHF(L) (3.7) > Myôgishô LHL(L) > Kyoto LHL(L)

(e.g., kusuri > Tokyo-type
‘drug’) LHH(H)

*HLL(L) (3.7) > Myôgishô LHL(L) > Kyoto LHL(L)

(e.g., kabuto > Tokyo-type
‘helmet’) HLL(L)

Revised Reconstruction
Monosyllabic Accent
*H(H) (1.1) > Myôgishô H(H)
> Tokyo-type L(H)

*H(L) (1.2) > Myôgishô F(L)

> H(H) > Tokyo-type L(H)

*FR.(H) > R(H) > Myôgishô L(H)

~ R(H) (1.3) > F(L) > Tokyo-type H(L)

*FR.(L) > Myôgishô H(L)

~ H(L) (1.3) > F(L) > Tokyo-type H(L)

9
In Tokyo, the pitch for the word ‘life’ is HLL(L) (Kindaichi 1966, Nihon Hôsô
Bunka Kenkyûjo 1971).
28 ACCENTUAL HISTORY

Disyllabic Accent
*HH(H) (2.1) > Myôgishô HH(H)
> Tokyo-type LH(H)

*HF(L) (2.2) > Myôgishô HL(L)

> HL(L) > Tokyo-type LH(L)

*LF(L) > Myôgishô ML(H)

~ L.RF.(L) > HF(L) > Tokyo-type LH(L)
(2.3)

*FH(H) > Myôgishô LH(H)

~ LH(H) (2.4) > FL(L) > Tokyo-type HL(L)

*FH(L) > Myôgishô LH(L)

~ LH(L) (2.5) > FL(L) > Tokyo-type HL(L)

Trisyllabic Accent
*HHH(H) (3.1) > Myôgishô HHH(H)
> Tokyo-type LHH(H)

*HHF(L) (3.2) > Myôgishô HHL(L)

> HHF(L) > Tokyo-type
LHH(L)
*HFL(L) (3.3) > Myôgishô HLL(L)
> HFL(L) > Tokyo-type
LHL(L)
*LLF(L) > Myôgishô MML(H)
~ LL.RF.(L) > HHF(L) > Tokyo-type
(3.4) LHH(L)
*LLH(H) (3.5) > Myôgishô MLH(H)
> HHH(H) > Tokyo-type
LHH(H)
 CRITICAL REVIEW 29

*LFL(L) > Myôgishô MLH(L)

~ L.RF.L(L) > HFL(L) > Tokyo-type
(3.5) LHL(L)

*FHH(H) > Myôgishô LHH(H)

~ LHH(H) > HHH(H) > Tokyo-type
(3.6) LHH(H)

*FHH(L) > Myôgishô LHL(L)

~ LHH(L) > HHH(H) > Tokyo-type
(3.7) LHH(H)

*FHL(L) > Myôgishô LHL(L)

~ LHL(L) > FLL(L) > Tokyo-type
(3.7) HLL(L)

Adopting the comparative method, Hattori demonstrates how he

reconstructs accent classes for monosyllabic nouns based on the
data from Myôgishô, modern Tokyo, and Kyoto dialects. However,
as shown in (12) above, Hattori reconstructs five accent classes for
disyllabic nouns and nine for trisyllabic nouns without providing us
with any data of accentual correspondences.

(13)
2.1 2.2 2.3 2.4 2.5
Kyoto HH(H) HL(L) HL(L) LL(H) LH(L)
Tokyo LH(H) LH(L) LH(L) HL(L) HL(L)
Morioka LL(L) LL(L) LH(L) HL(L) HL(L)
Takamatsu HH(H) HL(L) HH(H) LL(H) LF(L)
Kagoshima LH(L) LH(L) LL(H) LL(H) LL(H)
Myôgishô RR(R) RE(R) EE(R) ER(R) ER(E)
PJ *HH(F) *HF(L) *LH(L) *LH(H) *HL(L)

When it comes to the reconstruction of phonetic pitch shapes for

the reconstructed accent classes, Hattori’s methodology is even
more puzzling. In (1951), Hattori does not indicate how he assigns
30 ACCENTUAL HISTORY

the phonetic values to each accent class, but shows how the pitch
shapes change from Proto-Japonic to the modern dialects of Tokyo,
Kyoto, Morioka, Takamatsu, and Kagoshima (see (13) above).
Although Hattori demonstrates the development of the pitch shapes
for accentual classes, it is not certain how he arrives at the pitch
shapes for the Proto-Japonic accent classes on the basis of the
correspondences below. The following correspondences are
extracted from (12) and arranged as they are given.10 Notice that
Hattori’s reconstruction of the phonetic values for the accent
classes is not systematic. In other words, for example, it is not clear
why Hattori reconstructs PJ *HF(L) for 2.2 based on the
correspondences, Kyoto HL(L) : Tokyo LH(L) : Morioka LL(L) :
Takamatsu HL(L) : Kagoshima LH(L) : Myôgishô RE(R).

(14) Hattori’s claim on the development of accent from Proto-

Japonic 2.2 to modern dialects

PJ *HF(L) > LH(L) Tokyo, Kagoshima

> RE(R) > HL(L) Kyoto
> HL(L) Takamatsu
> LL(L) Morioka

With regard to the development of accent in modern dialects

from Proto-Japonic, what Hattori does is present a series of
accentual changes without explanation. For example, Hattori claims
that Proto-Japonic 2.2 *HF(L) developed into LH(L) in Tokyo and
Kagoshima, HL(L) in Kyoto through Myôgishô RE(R), HL(L) in
Takamatsu, and LL(L) in Morioka, but no explanation is given how
the changes occurred. It is unclear what principles of accentual

10
In fact, it is not known what dialects are involved in Hattori’s reconstruction.
Since he shows the development of accent in Tokyo, Kyoto, Morioka, Takamatsu,
and Kagoshima, these dialects are assumed to have been used for the
reconstruction of Proto-Japonic. Furthermore, because Hattori does not list any
accentual correspondences, and also as I assume all the dialects mentioned in (12)
are used for the reconstruction, a set of correspondences based on the data in (12)
will be listed here.
 CRITICAL REVIEW 31

change are involved in Hattori’s reconstruction. The development

should be explainable in accordance with natural pitch change.
Without a detailed scholarly explanation, no hypothesis can be
acceptable.
According to Hattori (1979a), there are two subclasses of nouns
in each of the accent classes 2.3, 2.4, and 2.5. That is to say, some
nouns in accent classes 2.3, 2.4, and 2.5 have vowel length in the
initial syllable, and the initial-syllable vowel length of those nouns
in some dialects of Ryukyuan corresponds to a short vowel with
accent in other dialects. Let us call that subclass ‘(a)’, and the rest
subclass ‘(b)’, namely, 2.3a, 2.3b, 2.4a, 2.4b, 2.5a, and 2.5b. See
some examples below -- examples for 2.3a are taken from Hattori
(1979b: 100) and examples for 2.4a and 2.5a are from Hattori
(1979a: 103-104). Vowel length in the first syllable in Shuri
corresponds to a short vowel with accent in Nakijin.

(15) Correspondences between vowel length and accent

2.3a 2.4a 2.5a

‘jar’ ‘breath’ ‘shadow’
Shuri ka:mi B11 /i:ci B ka:gi B
Nakijin ha˘mi /i˘ci ha˘gi

Compare (15) with the examples in (16) for 2.3, 4, 5b which

constitute Hattori’s other class. Examples for 2.3b are from Hattori
(1979b: 101) and examples for 2.4b and 2.5b are from Hattori
(1979a: 105).

(16) 2.3b 2.4b 2.5b

‘hole’ ‘board’ ‘sweat’
Shuri /ana B /ica B /asi B
Nakijin /anaÈa hicaÈa hasiÈi

11
‘B’ following Shuri words stands for initial low pitch.
32 ACCENTUAL HISTORY

Hattori (1979a: 106) says that there are two possible hypotheses
regarding reconstruction of accent for the correspondences shown
in (15). They are as follows:

(17) I. The initial-syllable vowel length in 2.3a, 2.4a, and 2.5a

existed in Proto-Japonic, and the length gave rise to
accent when the long vowel became short.

II. The initial-syllable vowel length in 2.3a, 2.4a, and 2.5a

developed from a short vowel in the accented syllable.

Hattori chooses Hypothesis I. That is, vowel length in Shuri is a

remnant of Proto-Japonic vowel length and its corresponding short
vowel with accent in Nakijin is secondary. However, Hattori’s
argument is not persuasive. His explanation is as follows: if one
hypothesizes that initial-syllable accent is original in Proto-Japonic,
he or she would have to reconstruct two subclasses for each of the
2.3, 2.4, and 2.5. On the other hand, if one proposes initial vowel
length in Proto-Japonic, he or she would not have to reconstruct
two subclasses for each of the 2.3, 2.4, and 2.5. However, there is
no clear explanation given how it can be done and why it is so.
Furthermore, Hattori fails to mention that there is another possible
hypothesis in addition to two hypotheses mentioned in (17). That is
to say:

(18) Hypothesis III

Initial-syllable vowel length and initial-syllable accent
existed independently in Proto-Japonic.

From the point of view of the comparative method, one has to

reconstruct two subclasses for each of the 2.3, 2.4, and 2.5 accent
classes, unless he or she can account for some correspondences as
secondary, because there are two distinctive correspondences in
each of the 2.3, 2.4, and 2.5.
 CRITICAL REVIEW 33

Other evidence that Hattori provides to support his hypothesis

deals with the accent systems of the Ainu language. In different
dialects of Ainu, there are two distinct types of accent systems;
pitch accent and non-pitch accent. In a system of non-pitch accent,
there is vowel length distinction, but there is no such distinction in
the pitch accent system. The initial-syllable high pitch in pitch-
accent Ainu (i.e., Hokkaido) dialects corresponds to vowel length
in the corresponding syllable of words in non-pitch accent Ainu
(i.e., Sakhalin) dialects. For example (Hattori 1967: 219):

(19) ‘red’ ‘breathe’ ‘yesterday’

Saru (Hokkaido) hure HL hese HL numan HL
Raichiska (Sakhalin) hu:re he:se nu:man

Referring to these correspondences of accent and vowel length

between Hokkaido and Sakhalin Ainu, Hattori (1967: 220) says:

(20) ... to iu taiô wo miru to, karafuto hôgen no hô ga korera

no ten de wa hobo kokei wo tamotte iru mono to
kangaerareru. [Looking at such correspondences, it would
appear that Sakhalin Ainu seems to have retained an
archaic form.]

And then he reconstructs the initial-syllable vowel length in Proto-

Ainu for words like those in (19).12 It is not clear why Hattori
thinks that the vowel length in Sakhalin is more conservative than
the accent in Hokkaido. Hattori’s methodology is questionable. If
there are principles of universal accent change on which Hattori
relies, and/or if there are languages, which have developed accent
from vowel length, Hattori should have mentioned them. Without a
clear methodology and reasoning, Hattori’s argument is far from
convincing. Therefore, this argument for Proto-Ainu vowel length
does not support his argument for Proto-Japonic vowel length.

12
Arguing that the correspondence mentioned in (19) is not regular, Vovin
(1993b) reconstructs both vowel length and pitch accent for Proto-Ainu.
34 ACCENTUAL HISTORY

The last crucial drawback in Hattori’s reconstruction of the

accentual history of Proto-Japonic is that he does not take accent
typologies into account. As shown below, in general three types of
pitch accent are found in the Japonic languages. (See 5.2, 5.3, 5.4.
5.5, and 6.2)

(21) Pitch-Accent Typologies

(a) Register and locus accent e.g., Kyoto-type dialects
(b) Locus accent e.g., Tokyo-type dialects
(c) Register accent e.g., Kagoshima dialect,
some Ryukyuan dialects

Hattori’s reconstruction ignores typology of Proto-Japonic accent

and does not account for how typologies shift from one type to
another in pitch-accent languages. In fact, none of the studies
reviewed in this chapter discusses the accentual history of the
Japonic languages from the point of view of typological shift.
Problems related to pitch-accent typology have to be raised and
investigated.

2.2 KINDAICHI
Another seminal study on the accentual history of the Japanese
language is Kindaichi (1975 [1954]). In his article, Kindaichi
claims that Tokyo-type accent developed from Kyoto-type accent.13
Kindaichi adds that although not all Japonic dialects developed
from Kyoto-type Japanese, all the Japonic dialects evolved from a
dialect having the accent system of Myôgishô. That is to say, the
Myôgishô accent is identical to Proto-Japonic accent.

13
This claim is based on Hattori’s (1937) abandoned older hypothesis (see
Kindaichi 1975: 60-61).
 CRITICAL REVIEW 35

(22) Kindaichi’s hypothesis on history of Japanese accent

Myôgishô accent (= Proto-Japonic accent)

Kyoto-type accent other dialect accent systems

Kyoto-type accent Tokyo-type accent

(23) provides correspondences in pitch types for disyllabic

words between modern Tokyo-type and modern Kyoto-type
dialects (Kindaichi 1975: 56). Solid lines denote the
correspondences between words, which can be at the end of a
sentence, and dotted lines the correspondences between words,
which cannot stand at the end of a sentence.

(23) Modern Kyoto-type Modern Tokyo-type

HH (H) (a) LH (H)

(b)
HL (L) (c) LH (L)

LH (H) (d) HL (L)

(e)
LH (L)

Regarding correspondences (a), (b), and (c), Kindaichi argues

that it is difficult to assume that Kyoto-type accent developed from
Tokyo-type accent because there is no reason why Tokyo-type
LH(H) splits into HH(H) and HL(L) in Kyoto-type dialects. On the
other hand, it would be reasonable to hypothesize that Kyoto-type
HL(L) splits into LH(H) and LH(L) in Tokyo-type dialects because
there is a distributional difference between LH(H) and LH(L).
Furthermore, Kindaichi adds that in Japanese there is a rule of pitch
change from HL(L) to LH(L): shift of an accent rightward .
36 ACCENTUAL HISTORY

(24) Distributional difference between LH(H) and LH(L) in Tokyo

LH(H) -- cannot occur at the end of a sentence, e.g., ok-u

‘put, place (something)’, (‘rentai (Attributive)’
form).

LH(L) -- can occur at the end of a sentence, e.g., ok-e ‘put,

place (something)’, (‘meirei (Imperative)’ form).

However, Kindaichi’s argument is faulty because no such accent

split as described above exists. He would have noticed that Kyoto-
type HL(L) does not correspond to both LH(H) and LH(L) in
Tokyo-type dialects if he examined the data carefully. Kyoto-type
HL(L) corresponds only to LH(L). For example, according to the
data given by Kindaichi (1975: 51), the ‘rentai’ form of the verb for
‘to put, place (something)’ is HH(H) in Wakayama (Kyoto-type)
dialect and LH(H) in Kôfu (Tokyo-type) dialect. The ‘meirei’ form
of the verb is HL(L) in Wakayama and LH(L) in Kôfu.

(25) Wakayama Kôfu

ok-u (‘rentai’) HH(H) : LH(H)
ok-e (‘meirei’) HL(L) : LH(L)

That is to say, the split did not occur, but the Wakayama HH(H)
and HL(L) simply became LH(H) and LH(L) respectively. Since
Japonic languages generally place the verb is at the end of a
sentence, only verbs are dealt with here.
For correspondences (23d) and (23e), as Kindaichi notes,
although the change from LH(H) to HL(L) is a completely opposite
direction to the change stated in the rule mentioned above, he
claims that in Kyoto LH(H) and LH(L) changed to HL(L). He
explains that, according to Kamikô Kan’ichi’s report (see Kindaichi
(1975: 57)), younger people tend to pronounce LL(L) for words
with LH(H) pitch pronounced by older people in Osaka. Therefore,
he argues, LH(H) is in the process of becoming LL(L). Kindaichi
also mentions that in the area covering part of Yamagata Prefecture
 CRITICAL REVIEW 37

and of northern Miyagi Prefecture, words pronounced in careful

speech with LL(L) are HL(L) in fast speech. Thus, it seems that a
change from LL(L) to HL(L) is in process. From these phenomena
of pitch change, Kindaichi claims that pitch changes as follows:
LH(H) > LL(L) > HL(L). Based on this claim, Kindaichi argues
that Kyoto LH(H) and LH(L) might have merged to LL(L), then
became HL(L) in Tokyo.
Kindaichi brings our attention to a theoretical issue, as well. As
in (26), listing theoretically possible hypotheses on Proto-Japonic
accent, Kindaichi argues that (26c) is the least possible one because
Proto-Tokyo-Kyoto is not real; there is no evidence for the
existence of Proto-Tokyo-Kyoto and a hypothesis based on ‘non-
existing evidence’ is not preferable.

(26) (a) Tokyo-type accent developed from Kyoto-type accent.

(b) Kyoto-type accent developed from Tokyo-type accent.
(c) Both Tokyo-type and Kyoto-type developed from
Proto-Tokyo-Kyoto.

Kindaichi also mentions that if with either (26a) or (26b), one could
not explain how the difference between the two dialects above
came into existence, then he or she should consider (26c). However,
since the development can be explained with (26a), (26b) and (26c)
should be disregarded. Moreover, Kindaichi notes that the reason
that linguists suggest hypothetical Proto-Indo-European is that they
cannot explain that all Indo-European languages developed from
Sanskrit. Apparently Kindaichi thinks that a language in the oldest
existing written record should be the origin of its descendent
languages.
There is no reason why (26c) should be excluded from possible
hypotheses just because a proto language is hypothetical. When we
reconstruct a proto language by the comparative method, a proto-
language is normally a hypothetical language. There are a number
of cases where corroborative evidence has come to light in support
of reconstructed forms, e.g., the evidence of Hittite for the Proto-
Indo-European laryngeals (Fox 1995: 173-181). The reason why
38 ACCENTUAL HISTORY

the hypothesis that the Indo-European languages developed from

Proto-Indo-European is necessary is not because we cannot explain
that all Indo-European languages developed from Sanskrit, but
because a reconstructed language need not be identical to any
attested language.
Like Hattori’s earlier hypothesis shown in (3a), Kindaichi’s
hypothesis conflicts with Yanagita’s findings in the distribution of
lexical items shown in (3b). To support his hypothesis, Kindaichi
incorrectly claims that the wave theory does not account for the
distribution of all aspects of language, and that vocabulary is one
aspect to which the theory applies, but not to phonology. From the
Rhenish Fan, it is already known that sound changes can also
spread (see Anttila 1989: 290-2).
If Kindaichi’s hypothesis were correct, it would have to be
assumed that exactly the same innovation independently occurred
in two different areas surrounding the Kyoto-type accent area.
Although Kindaichi claims that the accentual changes from Kyoto-
type to Tokyo-type can be explained by regular changes in accent,
he cannot account for why the Kyoto-type accent has remained the
same for centuries. Furthermore, Kindaichi should also explain
why Kagoshima-type dialects did not undergo the same changes as
Tokyo-type did and why the Tokyo-type accent is not seen in any
other place; if it is a result of natural accent change from the same
ancestral language, we should find the same type of accent system
randomly distributed throughout Japan.
In addition, a problem in the distribution of accentual systems
within Kyoto-type itself also needs to be raised here. Most Kyoto-
type dialects share the number of accentual classes and the pattern
of accent merger, namely, four accentual classes -- 2.2 and 2.3
merged, but phonetic pitch shape for each class varies depending
on the dialect. For example (Hyôgo examples are taken from
McCawley 1977):
 CRITICAL REVIEW 39

(27)
2.1 2.2, 3 2.4 2.5
Kyoto HH(H) HL(L) LH(H) LH(L)
Hyôgo HH(H) HL(L) LL(H) LH(L)

However, on the island of Ibuki, there is a dialect which has five

accent classes, like Myôgishô. Except for 2.1 and 2.5 accent classes,
the pitch shapes of the classes are very different from those of
Myôgishô. Compare Ibuki-island examples with the Myôgishô
accent system. In the Ibuki-island dialect, younger speakers
distinguish the same number of accent classes as older speakers,
but pitch shapes are not identical.

(28) Ibuki-island, Kagawa Prefecture (Wada 1966a)

2.1 2.2 2.3 2.4 2.5

Younger HH(H) HH(L) HL(L) LH, LL(H) LH(L)
Older HH(H) HL(L) HL, HH(L) LH, LL(H) LH(L)
Myô RR(R) RE(R) EE(R) ER(R) ER(E)

The distribution of Ibuki-island and Kyoto dialects seems to

suggest that a new change, namely, the merging of 2.2 and 2.3
accent classes occurred in Kyoto and the change did not affect the
Ibuki-island dialect, which is far from the central -- Kyoto -- area.
However, Kôchi, even farther away from the central area, has an
accent system identical to that of Kyoto dialect (Hattori 1933b).
That is to say, when we look at the geographical location of Kyoto,
Ibuki-island, and Kôchi, the wave theory does not seem to account
for the distribution of these accent systems.

(29) The distribution of Kyoto-type accent for disyllabic nouns

Myôgishô
⁄
Kyoto Ibuki-island Kôchi
4 classes 5 classes 4 classes
40 ACCENTUAL HISTORY

However, in the history of Japan, Kôchi came under the control of

the central court in the eighth century A.D. (Yamamoto 1970). That
is, there was constant contact between the center and Kôchi. On the
other hand, Ibuki-island has been isolated from the center because
of its remote location. Therefore, the distribution of 4-class type
accent and 5-class type accent among Kyoto, Ibuki-island, and
Kôchi can be explained.
Kindaichi agrees with the traditional view that the values of
‘even tone’ and ‘rising tone’ in Myôgishô are ‘low pitch’ and ‘high
pitch’ respectively. Under this view, the dialect described in
Myôgishô is a Kyoto dialect because of similarities between the
Myôgishô dialect and the modern Kyoto dialect. (30) illustrates
similarities between Kôchi dialect (Kyoto-type) and the Myôgishô
dialect in accent.

(30) Kindaichi’s view on pitch shapes for disyllabic nouns in

Myôgi-shô in comparison with Kôchi and Tokyo dialects

Myôgishô Kindaichi Kôchi Tokyo

2.1 RR(R) HH(H) HH(H) LH(H)
2.2 RE(R) HL(H) HL(L) LH(L)
2.3 EE(R) LL(H)
2.4 ER(R) LH(H) LH(H) HL(L)
2.5 ER(E) LH(L) LH(L)

Kindaichi explains how the Myôgishô accent system has become

that in Kôchi as in (31). Note that Myôgishô 2.1, 2.4, and 2.5 did
not change, and the only changes are seen in 2.2 and 2.3; HL(H)
changed to HL(L), and LL(H) merged with HL(L). If Kindaichi is
correct, the Kyoto-type accent did not change much for about nine
centuries, but the accent of all other dialects changed drastically by
shifting the accent rightward.
 CRITICAL REVIEW 41

(31) Accent changes from Myôgishô to modern Kôchi

Myôgishô Kôchi dialect

2.1 HH(H) = HH(H)

2.2 HL(H)
> HL(L) > HL(L)
2.3 LL(H)

2.4 LH(H) = LH(H)

2.5 LH(L) = LH(L)

As illustrated in (32), Kindaichi (1984: 19) shows how the

Kagoshima accent developed from the Myôgishô accent. Notice
here that the Tokyo accent evolved from the Kyoto accent, a
descendant of the Myôgishô dialect, but the Kagoshima accent
evolved directly from the Myôgishô dialect, although a series of
changes is involved in the course of its development.

(32) Kyoto Tokyo

HH(H) > LH(H) = LH(H)
--------- --------- ---------
HL(L) > LH(L) = LH(L)
HL(L) LH(L) LH(L)
-------- -------- --------
LL(H) > LL(L) > HL(L)
-------- --------
‰ LH(L) > LL(H) > HL(L)
Myôgishô
2.1 HH(H)
2.2 HL(H)
2.3 LL(H)
2.4 LH(H)
2.5 LH(L)
Ê
42 ACCENTUAL HISTORY

Ôita
HH(H) = HH(H) > LH(H) = LH(H) >
HH(H) HH(H) LH(H) LH(H)
--------- -------- -------- --------
LL(H) > HL(L) > LH(L) = LH(L) =
--------- -------- -------- --------
LH(H) = LH(H) > LL(L) > HL(L) =
LH(H) LH(H) LL(L) HL(L)

Hanji14
LL(L) = LL(L) > HL(L) > HL ~ LH(L)
LL(L) LL(L) HL(L) HL ~ LH(L)
-------- ------- -------- ---------------
LH(L) = LH(L) = LH(L) > LH ~ LL(H)
-------- LH(L) LH ~ LL(H)
HL(L) > LH(L) LH(L) LH ~ LL(H)
HL(L) LH(L)

Kindaichi (1975: 146) proposes a Proto-Ryukuan accent system

for disyllabic nouns as follows:

(33) Kindaichi’s reconstruction of Proto-Ryukuan accent

Accent
class Pitch shape Examples
2.1, 2 LH(H) kaze ‘wind’, hasi ‘bridge’
2.3 LH(L) hana ‘flower’
2.4, 5 HL(L) iki ‘breath’, kage ‘shadow’

In (33), ‘Accent classes’ refer to Myôgishô accent classes. That is,

in Proto-Ryukuan, Proto-Japonic 2.1 and 2.2 had merged, as had
Proto-Japonic 2.4 and 2.5. Notice that this system is identical to

14
Hanji Ryukyuan is spoken in Hanji, Kunigami-son, Okinawa Island [Okinawa
hontô].
 CRITICAL REVIEW 43

that of Ôita. Therefore, Kindaichi claims that the Ryukyuan

language evolved from Ôita-type Japanese.

(34) Kindaichi’s hypothesis (1984: 19)15

Ôita-type = Proto-Ryukuan
LH(H), LH(L)
HL(L)

Motobu-type Hanji-type
LH(H), LL(H) LL(L), LH(L)
HL(L) HL(L)

Kagoshima Shuri Naze

LH(L) HL(L) LL(H)
LL(H) LL(L) HL(L)

In addition, according to Kindaichi, Proto-Ryukuan accent split

into Motobu-type and Hanji-type first. And then the accent of
Hanji-type Ryukyuan evolved into other Kyushu (e.g., Kagoshima)
dialects and Ryukyuan dialects (e.g., Shuri Ryukyuan, Amami
Naze Ryukyuan).
Concerning the controversial issue of Proto-Japonic vowel
length, Kindaichi (1975: 138) believes that the initial-syllable
vowel length found in modern Ryukyuan did not exist in Proto-
Ryukuan. That is, it is a secondary development after the breakup
of Proto-Ryukuan. For his claim, Kindaichi argues that since the

15
This is a simplified version of Kindaichi’s chart in Kindaichi (1984: 19) to show
that the accent of Kagoshima (Kyushu), Shuri (Okinawa), and Naze (Amami)
developed from Hanji-type accent, not Motobu-type accent. In Kindaichi’s
original chart, there is a series of changes in accent from Hanji-type accent to
those modern dialects of the Japonic languages. In his analysis, Kindaichi lists
LH(H), LL(H), and HL(L) accent distinctions for Motobu dialect. However,
Hirayama et al. (1966) present data showing that Sakimotobu dialect has LH(H),
LR ~ LL(H), and HL(L). Kindaichi does not say specifically what dialect of
Motobu he is using.
44 ACCENTUAL HISTORY

Proto-Ryukuan accent system is identical to that of Ôita, and as

there is no initial-syllable vowel length in 2.3, 4, 5a accent classes,
in Ôita then some Ryukyuan dialects must have developed the
vowel length. Kindaichi (1975: 145) adds, the initial-syllable vowel
length in 2.3 nouns in Ryukyuan is phonologically conditioned --
the second syllable of those nouns contains the vowel /i/. However,
as with Hattori (1979a: 103), Kindaichi’s claim is faulty, because
there are words containing both /a/ in the second syllable and a
long vowel in the first syllable, e.g., na:ka ‘middle, inside’, and
also because there are words without the initial-syllable vowel
length even though they have /i/ in the second syllable, e.g., muzi
‘wheat’.
Kindaichi has not provided a clear explanation how vowel length
came into existence in Ryukyuan. A crucial shortcoming of his
hypothesis is that he ignores the existence of the correspondences,
in the initial syllable of 2.3, 4, 5a nouns, between, on the one hand,
vowel length and, on the other, a short vowel with accent (see
examples in (15) above). As stated by Hattori, these
correspondences are not due to chance, but are systematic.

2.3 TOKUGAWA
In his article, Tokugawa (1978 [1962]) reconstructs Proto-Japonic
accent for disyllabic nouns based on accent classes found in
modern dialects and also based on the geographical distribution of
those accent classes. However, Tokugawa does not reconstruct
phonetic values for each accent class.
Tokugawa’s reconstruction of Proto-Japonic accent has been
carried out with the following presuppositions (I) - (IV).
(I) Proto-Japonic had five accent classes for disyllabic nouns.
This is because there is no Japonic language with more than five
accent classes.
(II) Merger of accent classes is what happened in the
development of accent in modern dialects. In other words, no
accent class split into two. Moreover, once two accent classes
merged, they never come back together to the original accent
classes. For example, in Kyoto Japanese there are four accent
 CRITICAL REVIEW 45

classes: HH(H), HL(L), LL(H), and LH(L). The second class,

HL(L), is a result of the merger of the Proto-Japonic 2.2 and 2.3
classes. Later in the history of Kyoto Japanese, this class does not
split back to two original accent classes identical to Proto-Japonic
2.2 and 2.3, nor can it.

(35) This is what occurred:

Proto-Japonic 2.2
Kyoto 2.2, 3
Proto-Japonic 2.3

But, the following will never occur.

2.2
Kyoto 2.2, 3
2.3

In fact, though, as Tokugawa admits in the revisions to his

reconstruction, an accent class does split into two different classes.
Discussions on this matter will appear below.
(III) There were five stages in the history of the accentual
change from Proto-Japonic to accentless dialects in which all
accent classes merged. As far as the number of accent classes is
concerned, there were 5 at the first stage, 4 at the second, 3 at the
third, 2 at the fourth, and 1 at the fifth.

(36) Proto-Japonic Accentless

Stage: I II III IV V
Number of accent classes: 5 4 3 2 1

Development from Proto-Japonic accent occurred by means of

merger of accent classes. If a language goes through all the stages,
it would lose all the accent distinctions: all the accent classes merge.
A change occurs stage by stage; i.e., a language at the second stage
does not skip the third stage and become a language at the fourth
46 ACCENTUAL HISTORY

stage. However, Tokugawa fails to take typological shift into

consideration, e.g., a shift from locus accent to register accent in
Kagoshima.
(IV) (a) When languages with the same accent system are found
geographically next to each other, all of these languages have
evolved from an accentually identical earlier language. (b) When
accent system A and accent system B are located geographically
next to each other and the development of accent system B from
accent system A can be explained by accent merger, the following
must be true.

(37) accent system A (larger number of accent classes)

accent system B (smaller number of accent classes)

(c) Furthermore, if both accent systems above can be explained as

developments from an identical earlier accent system, the following
must be true.

(38) accent system AB

accent system A accent system B

Concerning Tokugawa’s presupposition (IV), (IVa) is not

necessarily true because languages influence one another by
contact. According to Thomason and Kaufman (1988), any
linguistic feature can be borrowed from one language to another by
contact. It is possible that due to contact between those languages
in question, they resulted in having the same accent system. For
(IVb), this is similar to Kindaichi’s method adopted in his
reconstruction of Proto-Japonic accent (see discussions on
Kindaichi’s reconstruction of Proto-Japonic above and also
Kindaichi 1975 [1954]). This method is problematic because a new
accent system can evolve from another by splitting accent classes.
That is to say, (39) can be true. Compare (39) with (37) and notice
 CRITICAL REVIEW 47

that they represent completely opposite reconstructions of the

history of accent systems A and B.

(39) accent system A (larger number of accent classes)

accent system B (smaller number of accent classes)

Tokugawa points out that there are logically 52 possible accent

types; 1 at the first stage (Proto-Japonic), 10 at the second stage, 25
at the third stage, 15 at the fourth, and 1 at the fifth stage.

(40) Logically possible accent types in terms of accent-class merger

(Tokugawa 1978 [1962]: 546-548)

Stage I: 1/2/3/4/5

Stage II: 12/3/4/5 13/2/4/5 14/2/3/5 15/2/3/4

1/25/3/4 1/2/3/45 1/23/4/5 1/2/34/5
1/24/3/5 1/2/35/4

Stage III: 1/2/345 123/4/5 15/3/24 1/3/245

12/34/5 13/2/45 1/235/4 13/24/5
14/2/35 1/234/5 14/23/5 15/2/34
145/2/3 12/35/4 1/23/45 135/2/4
13/25/4 1/24/35 134/2/5 15/23/4
1/25/34 125/3/4 12/3/45 124/3/5
14/3/25

Stage IV: 1/2345 12/345 135/24 2/1345

13/245 134/25 1245/3 14/235
125/34 1235/4 15/234 124/35
1234/5 145/23 123/45

Stage V: 12345
48 ACCENTUAL HISTORY

Although Tokugawa assumes that there have been five stages of

changes in accent, the number of possible accent changes is not
3750 (1 x 10 x 25 x 15 x 1 = 3750), but 180 (10 x 64 x 3 = 180).
The reason is because, for instance, accent type 1/23/4/5,16 where
2.2 and 2.3 merged, does not change to accent type 12/3/45, where
2.1 and 2.2 merged. As presupposition (II) states, the change from
1/23/4/5 to 12/3/4/5 is not possible because in the former 2.2 and
2.3 merged, but in the latter 2.1 and 2.2 merged. Therefore,
Tokugawa says that even though there are 180 possible changes,
the actual number of the changes occurring was much less.
Tokugawa (1978 [1962]: 562-3) first discusses his earlier
proposal that Proto-Japonic split into four different accent systems:
Eastern, Central, Northeastern Shikoku, and Western. Note that the
accent systems of Western Japanese and Eastern Japanese are
identical and unattested.

(41) Tokugawa’s reconstruction17

Proto-Japonic *1/2/3/4/5

Western Central North Eastern Shikoku Eastern

*1/2/3/45 1/23/4/5 13/2/4/5 *1/2/3/45

Hiroshima, Kyoto, Takamatsu, Tokyo,

etc. etc. etc. etc.

As shown in (42), Tokugawa (1978 [1962]: 565) revises his

earlier hypothesis on Proto-Japonic accent by suggesting that the
distinction of 2.4 and 2.5 is a later development in the central area

16
‘/’ indicates a boundary between accent classes. An accent system like 1/23/4/5
indicates that Proto-Japonic 2.2 and 2.3 classes merged into one, thus this
language has four accent classes.
17
The accent system for Central Japanese is 1/23/45 in Tokugawa (1978 [1962]:
562-3), but it is probably a misprint because the accent system of Kyoto Japanese
is 1/23/4/5. Thus, assuming that Tokugawa meant 1/23/4/5 for Central Japanese,
an adjustment in the chart have been made accordingly.
 CRITICAL REVIEW 49

of Japan, because the distinction of 2.4 and 2.5 is found only in

central Japanese and also because cases of accent split are found in
Japanese. In Tokugawa’s revised reconstruction, Proto-Japonic did
not distinguish 2.4 from 2.5, and Proto-Japonic accent split into
Western, Old Central, and Eastern accents. Northeastern Shikoku
and central accent developed from old central accent. This
hypothesis is similar to Thorpe’s (see discussions on Thorpe’s
reconstruction of Proto-Japonic below).

(42) Proto-Japonic *1/2/3/45

Western Old Central Eastern

*1/2/3/45 *1/2/3/4/5 *1/2/3/45

North Eastern Shikoku Central

13/2/4/5 1/23/4/5

For his explanation on the fourth class of Proto-Japonic accent

developing classes 2.4 and 2.5, Tokugawa (1978 [1962]: 565) says
that an accent class does split and that there are some examples for
it, but there is no case where accent class 2.5 was involved.18 If
there is no evidence for the development of class 2.5 in any dialects
of Japanese, how can Tokugawa claim that class 2.5 is an
innovation? It is false to argue that the split of 2.5 from 2.4
occurred if the only evidence that can be brought to bear is that a

18
“Rui no naibu de (aruiwa ittan tôgô shita ikutsu ka no rui ga, moto no kubetsu de
nai betsu no) kubetsu wo shôzuru koto wa ariuru shi, gen ni aru. ‘Rui no kubetsu’
ga tôgô no hôkô ni nomi susumu to iu koto wa, ‘kata’ no atarashii hassei zôka ga
nakatta to iu koto wo imi suru mono de wa nai. Shikashi 5-rui no kubetsu igai no
kubetsu wa, dai 1-ni, bunri ni kanren ga atta to omowareru jijitsu ga genzai
mitomerareru, ....” [After two or more accent classes have merged into one, it can
split to distinctive accent classes that are not the original ones. This sort of change
has been seen. Although there is a tendency toward merger of accent classes, it
does not mean that there are no cases where an accent class has not been created.
However, it is considered that an accent class (or classes) other than class 2.5 has
undergone a change by splitting of the class…]
50 ACCENTUAL HISTORY

similar split is found in other accent classes. If there is no direct

evidence for the development of 2.5, he must reconstruct class 2.5,
giving five accent classes for disyllabic nouns in Proto-Japonic.
Tokugawa’s reconstruction is an interesting one because of its
methods, excluding all possible external causes to language change,
but taking geographical distributions of accent types into account.
There is no doubt that geographical distributions of linguistic
changes and external influence on language change are correlated
with each other, because dialect geography also must deal with the
spread of change which is induced by external influence, namely,
language contact. Possibilities of language-contact-induced change
cannot be ignored when a history of a language is reconstructed.
Therefore, Tokugawa’s reconstruction is unrealistic.
Tokugawa’s attempt at reconstructing Proto-Japonic accent is
incomplete in the sense that he only reconstructs Proto-Japonic
accent for disyllabic nouns and also that he does not reconstruct the
phonetic values for the reconstructed accent classes.

2.4 OKUDA
On the basis of data from Myôgishô, Kunio Okuda (1975: 56)
proposes a Proto-Japonic accent system as in (43); eight accent
classes for disyllabic nouns and eleven for trisyllabic nouns.

(43) Okuda’s Proto-Japonic accent

Disyllabic Nouns
Pitch Myôgishô example
HH(H) hana-ga ‘nose-Nom(inative).’
H˘L(L) ka˘mi-ga ‘paper-Nom.’
LLÈ(H) hanaÈ-ga ‘flower-Nom.’
LÈH(H) kaÈsa-ga ‘umbrella-Nom.’
LÈH˘(L) maÈdo˘-ga ‘window-Nom.’
ÈRH(H) Ègoma-ga ‘sesame-Nom.’ (Kamei et al.
1964: 129)
ÈR˘L(L) Èha˘gi-ga ‘leg-Nom.’ (Hayata 1973: 146)
HH˘(L) Not attested
 CRITICAL REVIEW 51

Trisyllabic Nouns
Pitch Myôgishô example
HHH(H) iwasi-ga ‘sardine(s)-Nom.’
HH˘L(L) azu˘ki-ga ‘read bean(s)-Nom.’
H˘LL(L) ci˘kara-ga ‘strength-Nom.’
LLLÈ (H) atamaÈ-ga ‘head-Nom.’
LLÈH(H) kokoÈro-ga ‘heart-Nom.’
LÈHH(H) uÈsagi-ga ‘rabbit-Nom.’
LÈH˘L(L) kaÈbu˘to-ga ‘helmet-Nom.’
ÈRHH(H) Èsiwoni-ga ‘aster-Nom.’ (Hayata 1973: 146)
ÈR˘LL(L) Èe˘yami-ga ‘epidemic-Nom.’ (Hayata 1973:
146)
LLÈH˘(L) honoÈho˘-ga ‘flame-Nom.’ (Kamei et al.
1964: 129)
HHH˘(L) Not attested

Okuda (1975) suggests that pitch rise was phonemic in

Myôgishô and that such a rise in pitch became non-distinctive in
Middle Japanese because of a historical change involving mergers
in accentual patterns.
Looking into the Myôgishô data, Okuda claims that there are
constraints on the combination of pitch rise ( È) and pitch fall (˘ ) in
Myôgishô.

(44) The constraints are: ( È) may not be preceded by (˘ )

within a word, and ( È) and (˘ ), if both present, may not
be separated by more than one [syllable]. (Okuda 1975:
58)

Based on these constraints, Okuda points out that there are six
logically possible pitch patterns for disyllabic nouns (i.e., HH(H),
H˘L(L), LLÈ(H), LÈH(H), LÈH˘(L), and HH˘(L)) and eight for
trisyllabic nouns (i.e., HHH(H), HH˘L(L), H˘LL(L), LLLÈ(H),
LLÈH(H), LÈHH(H), LÈH˘L(L), and HHH˘(L)). Okuda suggests
that the Proto-Japonic accent system should include all the pitch
52 ACCENTUAL HISTORY

patterns. Most of the pitches are attested in Myôgishô except

HH˘(L) and HHH˘(L).
Okuda suggests another five accent classes for Proto-Japonic.
They are ÈRH(H), LLÈH˘(L), ÈR˘L(L), ÈR˘LL(L), and ÈRHH(H).
(‘R’ stands for rising pitch.). Regarding ÈRH(H) and LLÈH˘(L),
Okuda refers to Kamei et al. (1964) where the word goma ‘sesame’
is indicated with ÈRH(H) and honoho ‘flame is with LLH(L). For
ÈR˘L(L), ÈR˘LL(L), and ÈRHH(H), he mentions Hayata (1973) in
which hagi ‘leg’, eyami ‘epidemic’, and siwoni ‘aster’ are
described as ÈR˘L(L), ÈR˘LL(L), and ÈRHH(H), respectively.
Okuda argues, although those classes cannot be reconstructed based
on the constraints in (44), they are attested in Myôgishô. Thus, they
should be reconstructed in Proto-Japonic.
As can be seen in (45) below, Hayata (1973) points out that
polysyllabic words with the initial R in Myôgishô are problematic.

(45) Polysyllabic words with the “rising” mark are

problematic and might be accounted for in various
ways. Some outstanding illustrative examples are
discussed here. siwoni [RHH] ‘aster’ and goma [RH]
‘sesame’ are Chinese loan words. The rather unusual
form Ngoma is found as well as the more common
form goma [RH] in the Ruiju[-]myôgi[-]shô, where N
represents a grapheme interpreted as a syllabic nasal.
'eyami [RLL] ‘epidemic’ is probably a compound
word, since 'e was used also in isolation meaning
‘epidemic’, and yami is a nominalized form of a verb
yam- ‘to be sick’. (Hayata 1973: 147)

According to Okuda’s reconstruction of Proto-Japonic accent,

the number of accentual patterns is eight for disyllabic nouns and
eleven for trisyllabic nouns.

(46) Disyllabic: HH(H), HL(L), LL(H), LH(H), LH(L),

HH(L), RH(H), RL(L)
 CRITICAL REVIEW 53

Trisyllabic: HHH(H), HHL(L), HLL(L), LLL(H),

LLH(H), LHH(H), LHL(L), LLH(L),
HHH(L), RHH(H), RLL(L)

However, the constraints stated in (44) also enable reconstruction

of /˘OO/ [LL(L)] and /˘OOO/ [LLL(L)], although, like the classes
HH(L) and HHH(L), the LL(L) and LLL(L) classes do not exist in
Myôgishô. Okuda did not state why he did not reconstruct /˘OO/
and /˘OOO/.
The most crucial point of Okuda’s argument has to do with his
methodology. It is not certain why Okuda thinks that his method of
reconstruction can reveal the accentual history and prehistory of
Japonic languages. In other words, Okuda does not give any
reasons why logically possible accent classes based on the
constraints in Myôgishô are the same as the accent classes of Proto-
Japonic.
Another drawback of Okuda’s reconstruction is that it is based
exclusively on the Myôgishô data. Needless to say, in principle, the
more data we have, the better the form of proto-language we can
reconstruct. This is because the dialect used in Myôgishô has
developed over the centuries from Proto-Japonic and it’s highly
unlikely for any dialect to keep all accentual features that Proto-
Japonic had. Some other dialects might have kept some features
that the Myôgishô dialect has lost in the course of its development.
Therefore, a reconstruction based on one source of data is not
credible.

2.5 MCCAWLEY
In his article Accent in Japanese, James D. McCawley (1977) has
reconstructed the Proto-Japonic accent system for two-mora and
three-mora nouns. His methods of reconstruction involve two steps,
which are different in nature. The first step utilizes the comparative
method -- a reconstruction of a proto-form by comparing accent
types of modern dialects and Myôgishô, and the second step is
adoption of some accentual classes which are not reconstructed by
54 ACCENTUAL HISTORY

him, but which were originally suggested by Hayata (1973) and

Hattori (1951).

(47)
(a) Accent correspondences for two-mora nouns
Myôgishô Hyôgo Tokyo Akita/ Ôita W. Kyushu
2.1 HH(H) HH(H) LH(H) LH(H) F(alling)
2.2 HL(L) HL(L) LH(L)
2.3 LL(H) LH(L) L(evel)
2.4 LH(H) LH(H) HL(L) HL(L)
2.5 LH(L) LH(L)

(b) Accent correspondences for three-mora nouns

Myôgishô Hyôgo Tokyo Akita Ôita W.K
3.1 HHH(H) HHH(H) LHH(H) LHH(H) LHH(H) F
3.2 HHL(L) LHL(L)
3.3 HLL(L) HLL(L) LHL(L) LHL(L) LHH(L)
3.4 LLL(H) LHL(L) LHH(L) LHH(L) L
3.5 LLH(L) HLL(L) LHL(L) LHL(L) LHL(L)
3.6 LHH(H) LHH(H) LHH(H) HLL(L)
3.7 LHL(L) LHL(L) HLL(L)/
LHH(H)

With regard to the first step of McCawley’s reconstruction of

Proto-Japonic accent, it is simply based on accentual
correspondences between Myôgishô and five modern dialects:
Hyôgo (Kyoto-type), Tokyo, Akita (northeastern Honshu), Ôita
(northeastern Kyushu), and Western Kyushu dialects. Hyôgo
dialect is Kyoto-type, and both Akita and Ôita dialects are Tokyo-
type, but not identical to that of Tokyo. In Western Kyushu, there
are only two accent classes: falling and low. Note that these classes
of Western Kyushu systematically correspond to the initial pitch
height of the accent classes in Myôgishô.
On the basis of the correspondences shown above, McCawley
reconstructs Proto-Japonic accent as in (48). Except for the
 CRITICAL REVIEW 55

reconstructed 3.7b class, McCawley’s reconstruction of Proto-

Japonic accent is identical to the accent system of Myôgishô.

(48) McCawley’s reconstructed Proto-Japonic accent

Disyllables Trisyllables
2.1 *ÈOO HH(H) 3.1 *ÈOOO HHH(H)
2.2 *ÈO˘O HL(L) 3.2 *ÈOO˘O HHL(L)
2.3 *OOÈ LL(H) 3.3 *ÈO˘OO HLL(L)
2.4 *OÈO LH(H) 3.4 *OOOÈ LLL(H)
2.5 *OÈO˘ LH(L) 3.5 *OOÈO˘ LLH(L)
3.6 *OÈOO LHH(H)
3.7a *OÈOO˘ LHH(L)
3.7b *OÈO˘O LHL(L)

McCawley’s reconstruction at this point is not finished yet. He

will add one more accent class for two-mora nouns and two more
classes for three-mora nouns to his reconstruction. These additional
accent classes are HH(L), HHH(L), and LLH(H). The first two are
adopted from Hayata (1973) (see discussion above) and the third
one is from Hattori (1951). Therefore, the final version of
McCawley’s reconstruction is as shown in (49).
McCawley claims that, if this reconstruction is correct, the
following historical developments have occurred as in (50).
However, (50) does not explain how the changes took place.
Regarding the change from Proto-Japonic 2.2b */ÈO˘O/ to /OO/ in
Akita and Ôita, for example, in (50b) he says that accent was lost in
high-initial words in Akita and Ôita. This does not explain the
development in the dialects. It needs to be explained how the
change took place step by step. It would be more straightforward if
we propose that the development of PJ */ÈO˘O/ to /OO/ in Akita
and Ôita involves a series of accent changes, */ÈO˘O/ > /O˘O/ >
/OO˘/ > /OO/.
56 ACCENTUAL HISTORY

(49) McCawley’s reconstruction of Proto-Japonic accent19

2.1 *ÈOO HH(H) 3.1 *ÈOOO HHH(H)

2.2a *ÈOO˘ HH(L) 3.2a *ÈOOO˘ HHH(L)
2.2b *ÈO˘O HL(L) 3.2b *ÈOO˘O HHL(L)
2.3 *OOÈ LL(H) 3.3 *ÈO˘OO HLL(L)
2.4 *OÈO LH(H) 3.4 *OOOÈ LLL(H)
2.5 *OÈO˘ LH(L) 3.5a *OOÈO˘ LLH(L)
3.5b *OOÈO LLH(H)
3.6 *OÈOO LHH(H)
3.7a *OÈOO˘ LHH(L)
3.7b *OÈO˘O LHL(L)

Furthermore, (50) should include that a rising pitch ‘È’ is not

phonemic in Tokyo-type dialects; thus Proto-Japonic rising pitch
becomes non-distinctive in Tokyo-type dialects. For example, a
rising pitch becomes non-distinctive in the development of Proto-
Japonic 2.2b */ÈO˘O/ to /OO˘/ in Tokyo. Thus, the development
goes */ÈO˘O/ > /O˘O/ > /OO˘/, not */ÈO˘O/ > /OÈO˘/, because the
initial pitch height is predictable.

19
McCawley gives us two different representations of his Proto-Japonic accent.
He first considers only a falling pitch accent, marked with ‘˘’ (McCawley uses a
tick (‘) to indicate a falling pitch, but here ‘˘’ is used for convenience). And then
later he adopts Okuda’s analysis where a rising pitch (‘È’) is also distinctive. For
example, McCawley first gives PJ */˘O˘O/ for 2.3 and later by adopting Okuda’s
analysis he gives PJ */OOÈ/. Both Proto-Japonic forms indicate LL(H)
phonetically. However, there is a difference when the change from the Proto-
Japonic form to the Tokyo form is taken into consideration. According to (50), PJ
*/˘O˘O/ becomes /O˘O/ and then an accent shift one mora to the right /OO˘/ in
Tokyo, Akita, and Ôita. But there is no explanation for the development of PJ
*/OOÈ/ to Tokyo /OO˘/.
 CRITICAL REVIEW 57

(50) Historical developments in accent (McCawley 1977: 286)

Dialect Developments
a W. Kyushu Syllables past the first become L
(thus, only the pitch on the first
syllable remains distinctive).
b E. Kyushu (Ôita) and Accent is lost in high- initial
N. Honshu (Akita) words.
c All Japanese dialects, OOÈ LL(H) > ÈO˘O HL(L)20
except a couple of
localities where the
first three classes
remain mutually
distinct: Togi (Noto
peninsula) and Ibuki-
island (Kagawa
prefecture).
d E. Honshu (including Accent shifts one mora to the
Tokyo and Akita), W. right.
Honshu, E. Kyushu
(Ôita)

See McCawley’s reconstructed Proto-Japonic accent and its

development to those of the modern dialects summarized in (51).
An accent change or a series of accent changes is added in
parentheses where it is needed to explain the developments. The
developments for 2.2a, 3.2a, 3.5b, and 3.7a are missing because
they cannot be explained by any of the accent changes listed in (50),
and they will be examined below. The following abbreviations are
used for each dialect: Ky = Kyoto, To = Tokyo, Ak = Akita, Ôi =
Ôita, and WK = Western Kyushu.

20
McCawley mentions the change for only two-mora nouns in (50c), but based on
the data, it can be assumed a similar change (i.e., /OOOÈ/ > /OÈO˘O/) has also
occurred in three-mora nouns.
58 ACCENTUAL HISTORY

(51) McCawley’s Proto-Japonic accent and its developments

PJ Myôgishô Modern dialects

2.1 *ÈOO > ÈOO > ÈOO [Ky, To, Ak, Ôi]
> F [WK]

2.2a *ÈOO˘

2.2b *ÈO˘O > ÈO˘O > ÈO˘O [Ky]

> (O˘O >) OO˘ [To]
> (O˘O > OO˘ >) OO
[Ak, Ôi]
> F [WK]

2.3 *OOÈ > OOÈ > ÈO˘O [Ky]

> (ÈO˘O > O˘O >) OO˘
[To, Ak, Ôi]
> L [WK]

2.4 *OÈO > OÈO > OÈO [Ky]

> (OO >) O˘O [To, Ak,
Ôi]
> L [WK]

2.5 *OÈO˘ > OÈO˘ > OÈO˘ [Ky]

> (OO >) O˘O [To, Ak,
Ôi]
> L [WK]

3.1 *ÈOOO > ÈOOO > ÈOOO [Ky]

> OOO [To, Ak, Ôi]
> F [WK]

3.2a *ÈOOO˘
 CRITICAL REVIEW 59

3.2b *ÈOO˘O > ÈOO˘O > ÈOO˘O [Ky]

> (OOO˘ >) OOO [To,
Ak, Ôi]
> F [WK]

3.3 *ÈO˘OO > ÈO˘OO > ÈO˘OO [Ky]

> O˘OO / OO˘O [To]
> OO˘O [Ak]
> (OO˘O >) OOO˘ [Ôi]
> F [WK]

3.4 *OOOÈ > OOOÈ > (ÈO˘OO >) ÈOO˘O

[Ky]
> (OOO > O˘OO >
OO˘O >) OOO˘ [To,
Ak, Ôi]
> L [WK]

3.5a *OOÈO˘ > OOÈO˘ > (OOOÈ >) ÈO˘OO

[Ky]
> (OOO >) O˘OO [To]
> (OOO > O˘OO >)
OO˘O [Ak, Ôi]
> L [WK]

3.5b *OOÈO

3.6 *OÈOO > OÈOO > OÈOO [Ky]

> OOO [To]
> (OOO > O˘OO >)
OO˘O [Ak]
> (OOO >) O˘OO [Ôi]
> L [WK]
60 ACCENTUAL HISTORY

3.7a *OÈOO˘

3.7b *OÈO˘O > OÈO˘O > OÈO˘O [Ky]

> (OOO˘ > OOO >)
O˘OO [To, Ôi]
> (OOO˘ > OOO > O˘OO
>) OO˘O [Ak]
> L[WK]

When it comes to the reconstructed accent classes 2.2a, 3.2a,

3.5b, and 3.7a, the developments of these classes to the modern
dialects cannot be explained by the changes in (50). As shown in
(52), it seems, the simplest way to explain the developments is to
assume that accents have shifted one mora to the left. McCawley
agrees with Hayata that accent shifts one mora to the left for 2.2a
and 3.2a. This leftward accent shift can apply to 3.7a, as well. A
disadvantage of proposing a leftward accentual shift is that it is
very rare in Japonic languages.

(52) Proto-Japonic 2.2a, 3.2a, 3.5b, and 3.7a and their reflexes in
modern Kyoto

2.2a *ÈOO˘ > Kyoto ÈO˘O

3.2a *ÈOOO˘ > Kyoto ÈOO˘O
3.5b *OOÈO > Kyoto ÈO˘OO
3.7a *OÈOO˘ > Kyoto OÈO˘O

The form for Proto-Japonic 3.5b, */OOÈO/, is originally

Hattori’s reconstruction. But Hattori gives no clear explanation
how he has reconstructed the form and on what principles of accent
change his reconstruction is based. The development of PJ
*/OOÈO/ to Kyoto /ÈO˘OO/ and Tokyo /OO˘O/ can be explained
by both a rightward shift and a leftward shift, and both of them are
equally complex.
 CRITICAL REVIEW 61

(53) (a) Rightward shift

PJ *OOÈO > OOOÈ > ÈO˘OO (Kyoto)
PJ * OOÈO > OOOÈ > ÈO˘OO > OO˘O (Tokyo)

(b) Leftward shift

PJ *OOÈO > OOÈO˘ > OÈO˘O > ÈO˘OO (Kyoto)
PJ * OOÈO > OOO˘ > OO˘O (Tokyo)

In addition, McCawley also suggests a leftward accent shift for

the Tokyo 3.3 accent /O˘OO/ from PJ */ÈO˘OO/; that is, */ÈO˘OO/
> /OO˘O/ > /O˘OO/.

(54) In most ‘Tokyo-type dialects’ (e.g. Numazu and

Hiroshima, which otherwise agree closely with Tokyo
in accent), these words have OO˘O. In Tokyo, OO˘O
is comparatively rare in 3-syllable nouns: aside from
compound and derived nouns, only koko˘ro has that
accentuation. Tokyo thus appears to have undergone a
shift OO˘O > O˘OO. (McCawley 1977: 302)

What McCawley claims is that accent has shifted one mora to the
right and later shifted one mora to the left. If this is what actually
has occurred, a reconstruction of accentual history seems to be
impossible because accent shifts both directions. Rather, by taking
into consideration that a common direction of accent shift is
rightward, it is more reasonable to suggest an alternative way, that
Proto-Japonic 3.3 has remained the same in most of the 3.3 nouns
and in some 3.3 nouns accent shifted one mora to the right.
It is reasonable to question why the Proto-Japonic accent system
has been retained in Kyoto-type dialects -- almost the same for
centuries -- except 2.2a, 3.2a, 3. 5b, and 3.7a, but accent systems of
other dialects such as Tokyo-type and Western Kyushu-type have
undergone drastic changes. It is unrealistic to assume that the
Kyoto-type has not changed for centuries and others have
undergone a series of changes.
62 ACCENTUAL HISTORY

2.6 RAMSEY
S. Robert Ramsey (1979) challenges Kindaichi’s (1975) hypothesis,
which is one of the most widely accepted hypotheses on the
accentual history of the Japonic languages. With internal and
comparative evidence, Ramsey argues that the accent of Tokyo-
type dialects is more conservative than that of Kyoto-type dialects.
Ramsey compares Myôgishô with three types of dialects in modern
Japanese: Tokyo-type, Kyoto-type, and Kagoshima-type, and
presents structural evidence for his claim.
Ramsey claims that the actual pitch values of ‘even tone’ and
‘rising tone’ are not reflected directly in Myôgishô. There is no
simple way of knowing what the labels ‘even tone’ and ‘rising
tone’ mean. If we assign ‘low’ to ‘even tones’ and ‘high’ to ‘rising
tones’ as in Kindaichi’s hypothesis, the accent system of Myôgishô
looks like those of the modern Kyoto-type dialects. Ramsey says
they superficially resemble it in disyllabic nouns, but they do not in
longer words. However, in his historical studies of Chinese
phonology, Baxter (1992: 319-324) shows with evidence that the
pitch value of ‘rising tone’ is high.

(55)
Myôgishô Tokyo Kôchi Kagoshima
1.1 R(R) mi L(H) mi H(H) mi F ‘body’
1.2 R(R) hi L(H) hi H(L) hi F ‘sun’
1.3 E(R) ki H(L) ki L(H) ki L ‘tree’
2.1 RR(R) kane LH(H) kane HH(H) kane F ‘metal’
2.2 RE(R) isi LH(L) isi HL(L) isi F ‘stone’
2.3 EE(R) inu LH(L) inu HL(L) inu L ‘dog’
2.4 ER(R) umi HL(L) umi LH(H) umi L ‘sea’
2.5 ER(R) aki HL(L) aki LH(L) aki L ‘autumn’

Comparing the accent system of Myôgishô with those of Tokyo-

type and Kyoto-type dialects, Ramsey points out that accent in
Tokyo-type dialects corresponds to pitch change from E to R in
Myôgishô. There is no such correspondence between modern Kyoto
dialect and Myôgishô. For example, as shown in (55), pitch falls
 CRITICAL REVIEW 63

after the second syllable in the Tokyo noun for ‘dog’ and in
Myôgishô there is pitch transition from E to R after the second
syllable, whereas in Kyoto pitch falls after the first syllable. It
should be noted that the initial unaccented syllable is phonetically
low in Tokyo.
Ramsey maintains that, if the pitch value ‘high’ is assigned to E
and ‘low’ is assigned to R, we would be able to explain why there
are such correspondences in accent between Myôgishô and Tokyo
dialect. That is, E is distinctively higher pitch than R and the accent
in Myôgishô has been preserved by Tokyo-type dialects.
Although it has been claimed that Myôgishô had class 2.5, in
which nouns are marked with the so-called ‘east-dot’, Ramsey
doubts the existence of the class. The reasons Ramsey gives are “...
it is not clear what the ‘east-dot’ could have represented
phonologically because the pitch of a following particle was
apparently unaffected by it. (The essence of the 2.5 distinction in
the Kyoto-type dialects is that it lowers the pitch of a following
particle.)” and “... since only a tiny percentage of the 2.5 nouns
were ever marked with the dot, it seems unwise to generalize about
the entire class on this basis (Ramsey 1979: 164).”
However, one cannot deny the possibility that there was a 2.4 -
2.5 accentual distinction in Myôgishô because, as Martin (1987)
points out, there is evidence for class 2.5, in which nouns behave
differently from 2.4 nouns when followed by the particle /no/
(Genitive), even though evidence is limited (for the examples, see
Martin 1987: 173). Martin contends, “... the original class 2.5 may
have been larger than is indicated by the evidence of later dialects
and that through the years 2.4 has been acquiring members of
2.5 ...” Martin’s hypothesis cannot be ignored because it agrees
with philological evidence that the Kyoto dialect has developed its
accent system by merging accent classes, but not splitting them
(see discussions on the philological evidence in Kamei, Ôfuji, and
Yamada, eds. 1964: 122-138). Furthermore, the merger of accent
classes seems to be a natural way to explain how other Japonic
languages have developed their accent systems. I will support
Martin’s hypothesis in this book. That is, class 2.5 is ER(E).
64 ACCENTUAL HISTORY

Therefore, the correspondence shown for class 2.5 in (55) must be

revised as follows.

(56)
Myôgishô Tokyo Kôchi Kagoshima
2.5 ER(E) aki HL(L) aki LH(L) aki L ‘autumn’

According to Ramsey’s proposal, if R is low pitch and E is high

pitch, the Myôgishô 2.5 ER(E) would be HL(H). That is, the
particle following 2.5 nouns seems to have its own accent, which
contradicts Ramsey’s hypothesis -- particles did not and do not
have their own accentuation.
There is another problem with Ramsey’s hypothesis, in which he
proposes leftward accent shift. If R is ‘low’ and E is ‘high’, the
Myôgishô 2.2 RE(R), for example, is LH(L) and it has remained the
same in Tokyo, but it changed to HL(L) in Kyoto-type dialects.

(57)
Myôgishô Tokyo Kôchi
2.2 RE(R) = LH(L) isi LH(L) > isi HL ‘stone’

Supposing Ramsey is correct, it would have to be assumed, for

example, that 2.2 words such as isi ‘stone’ have undergone the
change shown in (57), that is, that accent shifted one mora or
syllable to the left.
Ramsey argues for leftward accent shift by mentioning a
phenomenon of accentual change when followed by the genitive
particle no. In Myôgishô, the particle no continues the pitch of the
preceding syllable, even though the syllable has an accent. Nouns
preceding the particle no become accentless. This phenomenon is
observed in modern dialects as well. For example, in Tokyo, when
word-final-accented nouns, such as LH(L) and LHH(L) are
followed by the particle no, the nouns lose their accent and become
accentless.
 CRITICAL REVIEW 65

(58) Tokyo dialect

ike LH(L) ‘pond’ ike no mawari LH H HHH

otoko LHH(L) ‘male’ otoko no gakusei LHH H HHHH

In Wakayama (Kyoto-type), unlike in Tokyo, the particle no does

not continue the pitch of the preceding syllable, but the particle
cancels an accent on a penultimate syllable, e.g., HL(L) > HH(H).
See the examples in (59). Ramsey explains that this accent change
indicates that the words ike and otoko were once HH(L) and
HHH(L), and an accent has shifted from the word final to the
penultimate syllable.

(59) Wakayama dialect

ike HL(L) ‘pond’ ike no mawari HH H HHL
otoko HHL(L) ‘male’ otoko no gakusei HHH H HHHH

However, an accent normally shifts rightward in Japonic languages.

Therefore, it is more reasonable to account for the accentual change
of the no-phrase by the ‘natural’ change of accent in Japonic
languages. (See 4.2.2)
Pointing out that the Kagoshima falling accent and level accent
correspond to word-initial R and E respectively in Myôgishô,
Ramsey claims that Kagoshima dialect has retained the initial pitch
height of Myôgishô. The Kagoshima accents F and L do correspond
to R and E respectively in Myôgishô, but the initial pitch heights of
the F and L do not match the values of R and E; as (60) illustrates,
their correspondences are completely opposite. Ramsey has to
account for why these correspondences are opposite in pitch height
and also how Kagoshima accent has developed that from Myôgishô.
66 ACCENTUAL HISTORY

(60) Ramsey’s hypothesis

Myôgishô Kagoshima
R low : F high initial
E high : L low initial

By contrast, if R is high and E is low, the pitch height of R and E

matches the initial height of F and L accent. Therefore, the widely-
accepted analysis is more reasonable for the claim that the initial
pitch height (register) of Myôgishô has been retained in Kagoshima
dialect.

(61) Widely-accepted analysis

Myôgishô Kagoshima
R high : F high initial
E low : L low initial

Ramsey says that the archaic accents of nouns are often

preserved in compounds. He demonstrates, for example, sake ‘rice
wine’ (2.1) is LH(H) in Tokyo and HH(H) in Kyoto. However,
when the noun is in the final position of compounds, it becomes -
zake LL in both Tokyo and Kyoto. Furthermore, the pitch LL is
identical to the accent of the word sake in Myôgishô, i.e. RR(R).
(62) presents more examples.

(62)
Gloss Tokyo Kyoto Myôgishô
‘belly’ hara(ga) LH(L) hara HL 2.3 EE(R)
~ -bara HH ~ -bara HH

‘soup’ siru HL siru LH 2.4 ER(R)

~ -ziru HL ~ -ziru HL

The evidence above justifies Ramsey’s hypothesis. However, it is

not known whether these three examples are the only ones that
Ramsey has found, whether the phenomenon is commonly seen in
 CRITICAL REVIEW 67

other compounds, or if this phenomenon is also found in

compounds involving trisyllabic nouns. I will examine this issue in
a separate study in the future.
Ramsey points out a problem that results from Kindaichi’s
analysis of the Myôgishô ‘even tones’ and ‘rising tones’. The
accent correspondences of words like kawa HL ‘river’ (2.2) and
yama HL ‘mountain’ (2.3) in modern Kyoto are respectively RE(R)
and EE(R) in Myôgishô. According to Kindaichi, both RE(R) and
EE(R), namely, HL(H) and LL(H), changed to HL. In his analysis,
a following particle appears to have independent accentuation. That
is, in Myôgishô particles had their own accentuation, but it is lost in
modern Kyoto.

(63)
Myôgishô Kindaichi’s Modern Kyoto Modern Tokyo
2.2 RE(R) HL(H) HL(L) LH(L)
2.3 EE(R) LL(H) HL(L) LH(L)

On the contrary, Ramsey says, if the ‘even tones’ were high pitched
instead of low, and the ‘rising tones’ were low pitched instead of
high, these particles would have low pitch. There is no need to
assume the change of the particle accentuation in Kyoto -- particles
do not have their own accentual system.

(64)
Myôgishô Ramsey’s Modern Kyoto Modern Tokyo
2.2 RE(R) LH(L) HL(L) LH(L)
2.3 EE(R) HH(L) HL(L) LH(L)

With regard to the distribution of Tokyo-type dialects and

Kyoto-type dialects, Ramsey’s hypothesis gives a simpler
explanation on how these dialects have come to be distributed as
they are now than Kindaichi’s. Agreeing with the wave theory, an
innovation must have occurred in the Kyoto area and then spread
out from there into the more conservative Tokyo-type areas, in
contrast with Kindaichi’s hypothesis that an innovation occurred in
68 ACCENTUAL HISTORY

two separate areas of Tokyo-type dialects. Ramsey’s hypothesis

explains in a simple manner why Tokyo-type dialects are spoken in
the northeastern part and the southwestern part of Japan. In the case
of an area of Tokyo-type dialect surrounded by Kyoto-type dialects
in the Kii peninsula, Ramsey explains that the isolated rural
villages are not influenced by the innovation.
Without touching upon Kyushu dialects, Ramsey’s hypothesis
seems to be correct. However, the reality of the distribution of
accent types is much more complex. For instance, the addition of
further-west accent types shows a possible lack of bilateral
symmetry, an important portion of Ramsey’s dialect-geographical
argument. Going from the center toward Kyushu, the following
distribution of accent types will be seen (a number in parentheses
indicates the number of accent classes for disyllabic nouns).

(65) Distribution of accent types

Kyoto-type Tokyo-type Miyazaki-type Kagoshima-type

(5) (3) (0) (2)

However, the same distribution is not found from the center toward
the northeastern part of Japan -- i.e., different accent types seem as
if they are scattered over the area, so it does not appear to be a
‘wave’-like distribution.
As Ramsey says, if an accentual innovation started in the central
area of Japan and spread outward, we would find more archaic
accent systems as we go outward. Under Ramsey’s hypothesis, the
farther we go from the center, the more conservative dialects we
find. That is, Tokyo-type dialects are more conservative than
Kyoto-type ones, Miyazaki-type dialects are more conservative
than Tokyo-type, and so on. If this is so, then Ramsey has to
account for why Kagoshima-type dialects, which have only a two-
way distinction, are more conservative than Miyazaki-type in
which there is no accentual distinction.
Ramsey has brought our attention to a new interpretation of
accent marks in Myôgishô texts and has revealed an intriguing
 CRITICAL REVIEW 69

hypothesis on the accentual history of the Japonic languages.

However, as discussed above, Ramsey’s hypothesis is generally
unconvincing.

2.7 THORPE
In Ryukyuan language history, Thorpe (1983) reconstructs Proto-
Japonic accent and Proto-Ryukuan accent, and explains how Proto-
Japonic evolved to Proto-Ryukuan and also how modern Ryukyuan
dialects developed from Proto-Ryukuan.
On the basis of data from Myôgishô and modern dialects, Thorpe
reconstructs the Proto-Japonic accent system for monosyllabic,
disyllabic, and trisyllabic nouns. In his reconstruction, pitch rise
([+rise]) and pitch fall ([+fall]) are phonemic and they are indicated
by ‘È’ and ‘˘’ respectively.

(66) Thorpe’s Proto-Japonic accent system for nouns (Thorpe

1983: 129)
Monosyllabic Disyllabic Trisyllabic
1.1 ÈO H(H) 2.1 ÈOO HH(H) 3.1 ÈOOO HHH(H)
1.2 O˘ H(L) 2.2 O˘O HL(L) 3.2 ÈOO˘O HHL(L)
1.3 O L(L) 2.3 OO LL(L) 3.3 O˘OO HLL(L)
2.4 OÈO LH(H) 3.4 OOO LLL(L)
2.5 OO˘ LH(L) 3.5 OOÈO LLH(H)
3.6 OÈOO LHH(H)
3.7 OO˘O LHL(L)

(67) presents Thorpe’s reconstruction of Proto-Ryukuan (numbers

indicate the correspondence of Thorpe’s reconstructed Proto-
Japonic accent classes to the most widely used classes). In
Thorpe’s dissertation, there is no discussion on his methodology of
reconstruction. What he does is present reconstructed forms
without a lucid explanation of how he reconstructs them, and
explains how the proto-forms have evolved to the corresponding
forms of the reflexes.
70 ACCENTUAL HISTORY

(67) Proto-Ryukuan accent system (1983: 129-30)

Monosyllabic Disyllabic Trisyllabic

1.1, 1.2 2.1, 2.2 3.1, 3.2, 3.3
ÈO H(H) ÈOO HH(H) ÈOOO HHH(H)
---------------- ---------------- --------------------
1.3 2.3 3.4, 3.5
OÈ L(H) OOÈ LL(H) OOOÈ LLL(H)
---------------- ---------------- ---------------------
2.4, 2.5 3.6, 3.7
OO˘ LH(L) OO˘O LHL(L)

In order to account for pitch accent phenomena including

accentual changes, Thorpe adopts a theory of non-transformational
generative phonology. In his analysis of the Proto-Japonic accent
system, there are two phonemic tonal melodies: [+rise] and [+fall].
Words have underlying tonal melody or melodies and the location
of the melody is specified in an underlying representation. For
example, the representation of words with /OÈO/ (LH) pitch shape
is [+rise, 2nd syllable] and of words with /OO˘/ LH(L) is [+fall,
2nd syllable]. However, in Proto-Ryukuan, the melodies [+rise]
and [+fall] merged into [+accent] if their location features are [1st
syllable] or [2nd syllable]. Beyond the second syllable all accents
are lost. For example, the Proto-Japonic /OÈO/ [+rise, 2nd syllable]
merges with Proto-Japonic /OO˘/ [+fall, 2nd syllable] and becomes
/OO˘/ [+accent, 2nd syllable]. Words with /ÈOO/ pitch shape are
[+accent, 1st syllable]. There are also words with no underlying
accent. They are /ÊO/ 21 ~ /OÈ(O)/, /OÈO/ ~ /OOÈ(O)/, /OOÈO/ ~
/OOOÈ(O)/. A final high pitch is automatic if there is no accent.
Notice that in Thorpe’s analysis, the term ‘accent’ does not
parallel the widely accepted usage of the term in Japonic linguistics.
That is, according to Thorpe, the reconstructed Proto-Ryukuan
accent class /ÈOOO/ [HHH(H)] has an accent, i.e., [+accent, 1st

21
‘Ê’ indicates a rising pitch. That is, pitch rises within a syllable.
 CRITICAL REVIEW 71

syllable], because of the initial rising accent. However, nouns

belonging to this class are high-level pitch -- no pitch fall.
According to Thorpe (1983: 119-127), there are four types of
changes in accent. The first type is a change occurring before a
pause (indicated by ‘#’). For instance, /O˘/# becomes /ÈO/#.
Another example is that in words without underlying melody their
final syllable becomes high pitch, e.g., /OO/ > /OÈO/ ~ /OOÈ(O)/.
Second, changes occur after a pause. For example, an initial high
pitch become low after pause, e.g., HHL > LHL. Third, the melody
[+fall] shifts to the right, and [+rise] shifts to the left. Thorpe
explains that this type of change occurs typically in Ryukyuan
dialects, where moras like Q, N, devoiced ï8, and the second
segment of long vowel or sequence of vowels cannot bear tonal
melody features. Finally, there is a type of changes called ‘echo’
fall. Echo fall is a predictable secondary pitch fall, which occurs
two syllables behind a primary fall or any final high pitch mora in
many dialects of Japanese and Ryukyuan.

(68) Echo fall

OOÈO > O˘O ÈO
OOO˘O > O˘OO˘O
OOOÈO > OO˘O ÈO

The first type agrees with ‘natural’ change in accent (see

Chapter 4). For instance, H(L) becomes H(H) by going through
/O˘/ [H(L)] (> /O/ [H(H)]) > /ÈO/ [H(H)], namely, rightward accent
shift. As Kindaichi (1974, 1975) discusses, the second type is also
commonly seen in the Japonic languages. For example, in Tokyo, it
occurred after losing ‘word initial register system’ -- the initial
mora is always low if it is not accented. The third type of change
contradicts what other studies have revealed. For example,
McCawley (1977: 266) states, an accent shifts to the right when it
is preceded by a voiceless syllable, e.g., /tasu˘ke-ta/ > tasu8ke˘-ta
‘rescue-Past’ (see also 4.1.3). Regarding the fourth type, Thorpe
mentions that echo fall is found in medieval Japanese (Wenck
72 ACCENTUAL HISTORY

1959: 408), Shimane dialect (Okuda 1975: 48), dialects in Iwate

and Kôchi prefectures (McCawley 1977),22 and some Ryukyuan
dialects. However, words with two peaks are rarely found in the
Japonic languages. The reason is, according to Kindaichi (1967),
that words with two peaks are not stable: there is a tendency for
change.
In addition, Thorpe (1983: 138) claims that in Aden (Kikai
island) [+fall] shifted one mora to the left (‘Fall Retraction’) in the
course of the development of its accent system from Proto-
Ryukuan. This is contradictory to the first and third types of
changes mentioned above.
Arguing against a hypothesis that Proto-Ryukuan developed
from an accent system like Ôita dialect (Hirayama and Nakamoto
1964: 69-102), Thorpe gives four reasons why he disagrees with
the hypothesis. First, Proto-Ryukuan evolved out of a language,
which had to resemble Proto-Japonic more than any of its
descendants in the prehistoric era. The following diagram shows
Thorpe’s view of the Japonic language relationship.

(69) Thorpe’s view of Japonic language relationship (1983: 236)23

Proto-Japonic

Kyushu-type Kyoto-type

Proto- unrecorded Old Tokyo-

Ryukyuan Kyushu dialects Japanese type

Thorpe (1983: 237) explains: “[T]he great historical value of the

22
According to Thorpe (1983), Iwate and Kôchi examples are on page 17 of
McCawley (1977). However, the McCawley (1977) starts from page 261 and ends
on page 302. Therefore, his examples cannot be located.
23
This tree diagram takes geographical information into account. This is the
reason that two lines cross: the line between Kyoto-type and Old Japanese and the
one between Kyushu-type and Tokyo-type.
 CRITICAL REVIEW 73

Ryukyuan languages resides in their preservation of [Proto-

Japonic] language elements inherited through Kyushu that have
been lost everywhere else.” Since Thorpe does not give reasons
why Proto-Ryukuan evolved from a language which had to be more
like Proto-Japonic than any of its descendants, this is not justified.
Second, Thorpe says that the accent system for three-mora nouns
in Ôita dialect is incompatible with that of Proto-Ryukuan. As far
as the pattern of accent merger is concerned, an accent system such
as that in Proto-Ryukuan cannot develop from that of Ôita because
in Ôita 3.3 merged with 3.4, but in Proto-Ryukuan 3.3 merged with
3.1 and 3.2.

(70) Ôita three-mora accent and Proto-Ryukuan three-mora accent

Myô Ôita Proto-Ryukuan
3.1 LLL(L) HLL(L)
3.2
3.3 HHH(H)
3.4 LLL(L)
3.5 HHL(L)
3.6 HLL(L) HHL(L)
3.7

Third, according to Thorpe, with apparently historically attested

and phonetically motivated accent changes, it is not possible to
explain naturally how modern dialects of Ryukyuan have
developed from a system like Ôita. However, it is not true that we
have to come up with ad hoc changes to explain the development of
Proto-Ryukuan from Ôita. For example, the last class (3.6, 7) of the
Proto-Ryukuan trisyllabic nouns is easily explained by ‘natural’
accent change, rightward accent shift (HLL > HHL). As mentioned
earlier, Thorpe’s hypothesis itself has problems with regard to
‘natural’ change.
Fourth, says Thorpe, it is not possible that East Tokunoshima
accent preserves the Proto-Ryukuan accent system, because, if so,
the development of other dialects of Ryukyuan has to be explained
74 ACCENTUAL HISTORY

by a number of independent unprecedented changes. Thorpe adds

that his reconstruction minimizes the number of changes, which
can account for the development of the Ryukyuan dialects.
Thorpe’s argument on ‘simplicity’ of a hypothesis is not justified
because no evidence is given.
In summary, Thorpe has accomplished a great deal in his
dissertation, making use of an enormous amount of data. However,
as pointed out, Thorpe’s hypothesis has problems in its explanation
of accent change. His reconstruction methodology is not clear. A
reconstruction of an accentual history of a language requires clear
explanations of methodology and has to be done based on attested
evidence or phonetically motivated changes.

2.8 MARTIN
Incorporating previous studies on accentual history, Martin (1987)
has set forth his version of the accentual history of the Japonic
languages. He reconstructs Proto-Japonic accent on the basis of 800
monosyllabic, disyllabic, and trisyllabic nouns in six modern
dialects of the Japonic languages: Tokyo, Kyoto, Kagoshima,
Shodon (Amami), Shuri (Okinawa), and Yonaguni. According to
Martin, adopting the comparative method, one has to reconstruct at
least three proto accent classes for monosyllabic nouns; at least five
for disyllabic nouns; and at least seven for trisyllabic nouns.
However, in his reconstruction, “despite various doubts about
them” (Martin 1987: 162), Martin includes some extra classes (2.2a,
3.2a, 3.5a, and 3.7a) that some linguists have proposed.
This reconstruction is carried out by means of the comparative
method based on a set of correspondences among the dialects listed
above, except the classes 1.3a, 2.2a (Hayata 1973), 3.2a (Hayata
1973), 3.5a (Hattori 1951), and 3.7a (McCawley 1977). Martin
explains, the fourth class for monosyllabic nouns, namely 1.3a,
cannot be reconstructed on the basis of the modern dialects, but
both Myôgishô and structural evidence suggests that earlier kinds
of Japanese must have had the class 1.3a. Martin (1987: 180-181)
observes that in Myôgishô 1.3 nouns, some are marked with even
tone, i.e., low pitch, (e.g., me ‘eye’, te ‘hand’, ki ‘tree’), and some
 CRITICAL REVIEW 75

are marked with going tone, i.e., rising pitch, (e.g., me ‘female’, mo
‘garment’, su ‘nest’).
On structural consideration, Martin (1987:243) points out that
the pitch pattern of a noun will be identical to that of the groups of
nouns of the appropriate number of syllables when the noun is
followed by a particle.

(71) 1.1 H(H) ‡ 2.1 HH(H)

1.2 H(L) ‡ 2.2 HL(L)
1.3a L(L) ‡ 2.3 LL(L)
1.3b L(H) ‡ 2.4 LH(H)

Furthermore, Martin (1987: 251) says, “Something over 60

percent of the nouns of two and three syllables are atonic. If 1.3a is
considered as an atonic low class, the same percentage prevails for
the monosyllabic nouns as well. (This provides an independent
argument in favor of making that distinction.)”
According to Martin, except for 2.2a, there is no philological
evidence for the distinction of 3.2a, 3.5a, and 3.7a. Moreover,
Martin implicitly rejects the distinction between 3.5a and 3.5b by
giving both classes LLH(H) pitch.

(72) Martin’s Proto-Japonic accent (1987: 273)

1.1 H(H) ti ‘blood’

1.2 H(L) ha ‘leaf’
(1.3a L(L))
1.3b L(H) te ‘hand’

2.1 HH(H) kaze ‘wind’

2.2a HH(L) > HL(L) hito ‘person’
2.2b HL(L) isi ‘stone’
2.3 LL(L) yama ‘mountain’
2.4 LH(H) matu ‘pine’
2.5 LH(L) saru ‘monkey’
76 ACCENTUAL HISTORY

3.1 HHH(H) katati ‘shape’

3.2a HHH(L) > HHL(L) tokage ‘lizard’
3.2b HHH(L) azuki ‘red bean’
3.3 HLL(L) hatati ‘twenty’
3.4 LLL(L) kagami ‘mirror’
3.5a LLH(H) sudare ‘reed screen’
3.5b LLH(H) (? < LLH(L)) namida ‘tears’
3.6 LHH(H) usagi ‘rabbit’
3.7a LHH(L) > LHL(L) kusuri ‘drug’
3.7b LHL(L) kabuto ‘helmet’

Martin claims that Proto-Japonic accent has two kinds of

typologically different features, initial register and locus accent,
and also that the Proto-Japonic register system is retained in
Kagoshima-type dialects and in some Ryukyuan dialects (such as
Shodon and Shuri).
When it comes to the developments of the modern dialects from
Proto-Japonic, Martin says, Kagoshima-type dialects lost locus
accent, but retained initial-register accent. Tokyo-type dialects lost
the register distinctions, but retained the locus accent. With regard
to the ‘low atonic’ classes, Martin suggests the following changes
(73).

(73)
Kyoto PJ Tokyo
2.3 HL < *LL(L) > LH(L)
3.4 HLL(L) < *LLL(L) > LHH(L)

That is to say, Proto-Japonic low-register atonic became high-

register prototonic in Kyoto-type dialects and oxytonic in Tokyo-
type dialects. At the same time, or earlier, Proto-Japonic 1.3a L(L)
merged with 1.3b L(H), resulting in three monosyllabic accent
classes.
Like most of the other major studies on Japonic accent, Martin
does not investigate ‘natural’ accent change in relation to the
development of phonetic pitch shape. Martin states: “There remain
 CRITICAL REVIEW 77

many problems in plotting the phonetic developments found in the

modern dialects and in accounting for the various mergers of accent
types ([section] 6). We will put these problems aside for future
work” (1987: 247).
Martin discusses certain accent changes caused by segmental
changes and analogy. First, in some dialects, the second mora of a
heavy syllable cannot be accented, e.g., ka˘i ‘shell’ in Tokyo vs.
kai˘ in Kyoto. According to Martin, this is due to the
monosyllabification of separate syllables in Tokyo dialect into a
heavy syllable in which the second mora cannot have a locus.
Second, devoiced vowels shifted the accent to an adjacent
‘stronger’ syllable.24 Third, by analogy, homophonic words shifted
their accent class to another. These changes, however, are not
enough to account for the accentual history of the Japonic
languages -- the nature of accent change has to be investigated in
order to account for the development of the accent system of the
Japonic languages.
Concerning the relation between vowel length and accent,
Martin suggests that a word initial low register is related to vowel
length. That is, words with initial-low register had vowel length
word initially. This hypothesis is based on the fact that there are a
number of long vowels in Shuri and some other Ryukyuan dialects
in 2.3, 2.4, and 2.5 nouns. However, Martin also raises the question
“why the majority of low-register nouns did not maintain the vowel
length and why so few verb stems seem to manifest the reflex”
(1987: 247). There is still no answer to this question. Further
investigation is needed.
As mentioned earlier, Hattori also noticed the phenomenon of
Ryukyuan vowel length, and contended that this is a remnant of
Proto-Japonic vowel length. It has been kept in some Ryukyuan
dialects, but in others it simply become short with high pitch.

24
Martin (1987: 253) calls syllables containing a full vowel ‘stronger syllables’ as
opposed to syllables with a devoiced vowel. Moreover, he does not mention which
direction the accent shifts.
78 ACCENTUAL HISTORY

(74) CVV-LL- (e.g., Shuri)

Proto-Japonic
*CVV-LL-
CV-H- (e.g., Nakijin)

In addition, Martin points out that the first syllable of some 2.2
and 2.3 nouns contains a long vowel, but the syllable is high in
pitch: HHL. Martin suggests: “possibly it is a shape-restricted
retention of the earlier vowel length we assume for Type 2.3
extended to the less-populated Type 2.2 with the merger of the two
[classes], even though the pattern of the merged type does not
begin with a low pitch” (1987: 248).
Martin does not touch upon the controversial issue of the
geographical distribution of accent types in relation to the
development of the modern dialects. As mentioned earlier,
choosing one or the other hypothesis would lead us to a completely
different history of Japonic accent.

2.9 MATSUMORI
Another study on the history of Japonic accent is Matsumori (1993).
She reconstructs Proto-Japonic based on a phenomenon of
‘downstep’ found in the dialects of Ibuki-island (Kagawa
Prefecture), Ômi (Kagawa Prefecture), Tsubata (Ishikawa
Prefecture), and Shiramine (Ishikawa Prefecture). Downstep is
considered a phenomenon of pitch change from high to mid-high
when the high pitch is preceded by another high pitch, e.g., `` ``
(HH) > `` æ (HM), cf. `` __ (HL). According to Hyman and
Schuh (1974), downstep historically results from losing a L tone
between two H tones.

(75)
OOO (HLH) > OOO (HHH) > OOO (HHM)
| | | | | |/ |
HLH H(L)M H M
 CRITICAL REVIEW 79

The patterns of downstep found in these dialects of Japanese are

HM for two-mora, HHM for three-mora, and HHMM for four-
mora words. See Matsumori’s analysis of some Ibuki-island
examples in (76).

(76) Downstep in Ibuki-island

i nu > i nu > inu ‘dog’

|\ | | | | |
HLH H(L)H HM

atama > atama > atama ‘head’

| | | |/ | |/ |
HL H H(L) H H M

As demonstrated above, words have underlying HLH tones no

matter how many moras the words consist of. The L tone gets
disconnected and the H tone preceded by another H becomes M.25
Note that the processes from an underlying form to a surface form
shown above are not observed -- only the surface form is observed.
Agreeing with Kindaichi (1967), Matsumori says, the reason why
the pitch shape HLH undergoes a change is that it is not stable due
to the presence of two peaks. According to her, in addition, there
are two other changes that HLH undergoes. They are as follows:

(77) (a) Single-peaking

OOO (HLH) > OOO (HLL)
| | |
H H H

25
Apparently the middle pitch, indicated by M, is phonemic because in Ibuki-
island there are three contrastive pitch patterns involving the M (Uwano 1985).
See the following examples.
kuruma HHH ‘car’
musume HHL ‘daughter’
atama HHM ‘head’
80 ACCENTUAL HISTORY

(b) Plateauing
OOO (HLH) > OOO (HHH)
| | |/ |
H H H H

Matsumori mentions that the HLH pitch shape is found in

Makurazaki (Kagoshima Prefecture) and some dialects in Oki
Island and that according to Kindaichi (1967), a pitch pattern with
two peaks is not stable; thus, it has great tendency to change.
However, there is no evidence for the development of the HLL,
HHH, and HHM from the HLH. By contrast, pitch changes such as
LL(L) > HL(L) are attested in Ryukyuan (see Hirayama and
Nakamoto 1964 and Hirayama, Ôshima, and Nakamoto 1967).
Thus, proposing the change LLL > HLL is more reasonable than
HLH > HLL, unless there is evidence for the latter. Since a
rightward accent shift is common in Japonic languages, the change
HLL > HHL is more natural, and also this analysis is much simpler
than Matsumori’s.
Disagreeing with a widely accepted claim that the Proto-Japonic
accent is similar to that of Myôgishô, Matsumori contends that the
Proto-Japonic accent system and even the accent system of Proto-
Mainland Japanese were significantly different from that of
Myôgishô. The reason for this claim is that the Myôgishô accent
does not account for the development of downstep in the modern
dialects mentioned above. Furthermore, she argues that in order to
account for the downstep, one has to reconstruct a Proto-Mainland
Japanese accent system which is more archaic than the Myôgishô
accent system.
According to Matsumori, the development of the downstep
phenomenon from Proto-Mainland Japanese and also from Proto-
Japonic can be explained by her suggested changes in (76), (77),
and the common changes discussed in Kindaichi (1973).
 CRITICAL REVIEW 81

(78) Matsumori’s proposal on the history of Japonic accent

Proto-Japonic

Ryukyuan Proto-Mainland Japanese

Myôgishô Ômi, Ibuki-jima,

(no downstep) Tsubata, Shiramine
(with downstep)

Matsumori adopts Kindaichi’s (1973) rules of accent change in

order to explain the historical development of the Japonic accent.
The following are Kindaichi’s three rules of accent change.

(79)
(a) Pitch height of a syllable (or mora) tends to assimilate
to that of its preceding one.

e.g., LHH > LLH > LLL

HLL > HHL > HHH

(b) The initial high pitch tends to become low pitch.

e.g., HHH> LHH

(c) The first L tends to become H when L’s are in sequence.

e.g., LLL > HLL

82 ACCENTUAL HISTORY

Based on changes listed in (76), (77), and (79), Matsumori

reconstructs Proto-Mainland Japanese as follows:

(80) Matsumori’s Proto-Mainland Japanese accent (1993: 52)

1.1 *F(H) 2.1 *FH, HL(H) 3.1 *HLH

1.2 *L(L) 2.2 *LL 3.2 *HLL
1.3 *H(H) 2.3 *LH 3.3 *LLL
2.4 *HH 3.4 *LLH
2.5 *HL 3.5 *LHH
3.6 *HHH
3.7 *HHL

Regarding Matsumori’s reconstruction, there are several

problems with her methodology. First of all, her method allows us
to reconstruct a phonetic shape for phonemic accent classes, but not
the phonemic accent classes themselves. Matsumori does not
reconstruct accent classes, nor does she discuss how she reaches the
conclusion that there are three monosyllabic, five disyllabic, and
seven trisyllabic accent classes in Proto-Mainland Japanese. She
presupposes that Proto-Mainland Japanese had three accent classes
for monosyllabic nouns, five for disyllabic nouns, and seven for
trisyllabic nouns. With her method it is not possible to reconstruct
phonemic accent classes of Proto-Mainland Japanese. In addition,
even if Matsumori’s method (or any other method) could allow her
to reconstruct the way she does above, it should be questioned why
the number of the classes have remained the same for about
fourteen centuries (the fourth century B.C. to the eleventh century
A.D.)26, but have become much fewer in number in modern dialects
of the Japonic languages over about nine centuries (the eleventh
century A.D. to present), in some modern dialects there being no
distinction in accent at all.

26
As far as the number of accent classes is concerned, Matsumori’s reconstruction
of Proto-Japonic is the same as that of her Proto-Mainland Japanese.
 CRITICAL REVIEW 83

Matsumori (1993: 49) purports to show how Ômi accent

developed from Proto-Mainland Japanese by undergoing the
following changes (81).

(81) (a) First mora becomes H when beginning with L.

(b) First mora becomes L when the first and second moras
are H.
(c) Downstep occurs.

However, the changes do not give us the right results. For instance,
according to Matsumori, as shown in (82), the Ômi 3.5 accent
HHM evolved from PMJ *LHH by undergoing two changes: (81a)
and (81c); (81b) is not applicable.

(82)
PMJ (81a) (81c) Modern Ômi
3.5 *LHH > HLH > HHM

In fact, as revealed below, by (81a), PMJ *LHH should have

become HHH, not HLH. And then HHH should have become LHH
by (81b). Therefore, Matsumori’s explanation fails.

(83) Expected
PMJ (81a) (81b) Modern Ômi
3.5 *LHH > HHH > LHH

Matsumori also shows how the modern Ibuki-island trisyllabic

accent evolved from Proto-Mainland Japanese by going through a
series of changes. These changes are listed in (84), and (85), and
illustrate the development of the Ibuki-island accent system.
Matsumori explains, (Ia) applies to 3.1, (Ib) to 3.6 and 3.7, (Ic) to
3.3 and 3.4, (Id) to 3.2, and (II) to HLH pitch of 3.4 and 3.5.
Notice that at the first stage none of the changes listed in (I) applies
to 3.5. Thus, it is not certain how PMJ *LHH becomes HLH.
84 ACCENTUAL HISTORY

(84) Changes proposed by Matsumori (1993: 51)

(I) (a) HLH > HHH

(b) first mora becomes L when the first and second
moras are H.
(c) first mora becomes H when beginning with L
(d) accent shift to the right

(II) Downstep

(85) The development of the Ibuki trisyllabic accent system

proposed by Matsumori (1993: 51)

PMJ (I) (II) Ibuki-island

3.1 *HLH > HHH = HHH
3.2 *HLL > HHL = HHL
3.3 *LLL > HLL = HLL
3.4 *LLH > HLH > HHM
3.5 *LHH > HLH > HHM
3.6 *HHH > LHH = LHH
3.7 *HHL > LHL = LHL

As is apparent from her demonstration above, at the first stage

the changes do not all occur at once; they occur one after another
within the stage. Therefore, as shown in (87), it can be interpreted
that there are five accent changes that Ibuki-island dialect has gone
through from Proto-Mainland Japanese.

(86) (a) HLH > HHH

(b) First mora becomes L when the first and second moras
are H.
(c) First mora becomes H when beginning with L.
(d) Accent shift to the right
(e) Downstep
 CRITICAL REVIEW 85

When Matsumori’s Proto-Mainland Japanese accent forms undergo

all the changes, only the forms for 3.1 and 3.2 are the same as those
of Ibuki-island. The rest of the outcomes do not match. (87)
demonstrates how all the classes of Proto-Mainland Japanese
undergo changes stage by stage. This demonstration reveals that
Matsumori’s hypothesis fails to account for the development of the
accent classes except 3.1 and 3.2 classes. The column under
‘Result’ shows my working out of the outcome of the changes. ‘?’
indicates uncertainty of applicability of a change or a form due to
lack of information.

(87)
PMJ (a) (b) (c) (d) (e) Result Ibuki
3.1 *HLH > HHH > LHH > HHH = HHH = HHH? HHH
3.2 *HLL > HHL = HHL HHL
3.3 *LLL > HLL > HHL = HHL HLL
3.4 *LLH > HLH > HHL? = HHL? HHM
3.5 *LHH > HHH > ? = HHH? HHM
3.6 *HHH > LHH > HHH > ? = HHH? LHH
3.7 *HHL > LHL > HHL > HHH(L) > HHM LHL

It is apparent that Matsumori’s hypothesis is problematic. Not only

does the analysis for the development of Ibuki-island have
problems, but also other analyses such as for Ômi accent and
Myôgishô accent exhibit similar problems.
Matsumori makes use of an autosegmental analysis to account
for the downstep phenomenon. She explains, Proto-Mainland
Japanese 2.1 *FH has a tonal melody HLH and *FH is HL(H)
when followed by a particle. Its analysis is shown in (88). Although
Matsumori does not mention any processes/rules which are
involved in this analysis, it is clear that tones link to a tone bearing
unit from right to left because H and L tones are linked to the tone
bearing unit on the left when there are two tone bearing units.
86 ACCENTUAL HISTORY

(88) (a) when in isolation (b) when followed by a particle

FH HL(H)
O O OO (O)
/| | | | |
HL H HL H

Matsumori’s analysis contradicts Haraguchi’s (1977) analysis, in

which tones link to a tone bearing unit from left to right. Under
Haraguchi’s analysis, they would be analysed as in (89). That is,
Proto-Mainland Japanese 2.1 is *HR, instead of *FH. Since
Matsumori does not discuss her analysis in detail from the
viewpoint of autosegmental phonology, it is not certain why her
analysis does not agree with Haraguchi’s and also whether two
different analyses are even possible.

(89) (a) when in isolation (b) when followed by a particle

HR HL(H)
OO OO (O)
| |\ | | |
H LH HL H

In her article (1998), Matsumori reconstructs the Proto-Ryukuan

accent system for monosyllabic and disyllabic nouns in (90). This
reconstruction is based on Hattori’s (1979a and b) discovery of the
two subclasses within 2.3, 4, and 5 (see discussions above).
Matsumori claims that these subclasses may not have existed in
Proto-Japonic because they are found only in Ryukyuan.

(90) Matsumori’s Proto-Ryukuan accent system

Monosyllabic accent
Class I Myôgishô 1.1 and 1.2 nouns
Class II Myôgishô 1.3 nouns
 CRITICAL REVIEW 87

Disyllabic accent
Class I Myôgishô 2.1 and 2.2 nouns
Class II Myôgishô 2.3a, 2.4a, and 2.5a
Class III Myôgishô 2.3b, 2.4b, and 2.5b

Regarding the controversial issue of the history of the initial-

syllable vowel length in the Japonic languages, Matsumori (1998)
argues that the vowel length may not have existed in Proto-Japonic,
but that it came into existence in Proto-Ryukyuan because the
correspondences between the initial-syllable vowel length and an
accented short vowel are found only in the Ryukyuan language, but
not in the Japanese language. Note that on the basis of the same
correspondences, Hattori (1979a and b) claims that vowel length
existed in Proto-Japonic.
Matsumori’s reasoning for excluding vowel length from the
reconstruction of Proto-Japonic is faulty. The reason is that, in
principle, a linguistic feature for a proto language has to be
reconstructed unless its existence can be explained. As discussed
earlier, we cannot account for the reason why there is such a
correspondence between the vowel length and accent, and also why
some Ryukyuan dialects have the vowel length in particular accent
classes.
 3

HISTORICAL LINGUISTIC
METHODS
3.1 KOKUGOGAKU IN JAPAN
n Japan there is a discipline or study of the Japanese language
I called Kokugogaku, literally meaning ‘the study of the national
language’ (i.e., koku ‘nation’, go ‘language’, and gaku ‘study’).
While Kokugogaku scholars have done a great number of studies
on historical aspects of not only the Japanese language but also the
Ryukyuan language, the systematic approach of the comparative
method has rarely been taken advantage of. With regard to the
studies of the Ryukyuan language, for example, Thorpe (1983: 4)
states as follows:

(1) ….. Ryukyuan descriptive materials [have]

been augmented with a whole series of almost uniformly
excellent studies by Japanese and Ryukyuan
linguists; this literature includes not only research on
particular languages and dialects but also several ambitious
comparative studies. Yet, surprisingly, a systematic effort to
solve the major problems of Ryukyuan language
history has hardly begun.

After more than two decades, this statement is still relevant not
only to the studies of the Ryukyuan language but also to
kokugogaku (studies of the national language) as a whole, with the
exceptional work of Shirô Hattori (see below).
As far as historical studies of the accentual system of the
Japanese and Ryukyuan languages are concerned, most scholars in
 HISTORICAL LINGUISTIC METHODS 89

kokugogaku confine their studies to the inspection of written

records and speculation on phonetic values of distinctive accentual
classes recognized in the written records. With this methodology,
the kokugogaku scholars have come to believe that the ancestral
language of all Japonic languages, as far as its accentuation is
concerned, is identical to the language described in Myôgishô (e.g.,
see Kindaichi 1975 [1954]), even though this hypothesis has crucial
problems (see 2.2).
Based on written records, Kindaichi (1947, 1954, 1960, 1974,
and 1975) discusses ‘laws of accent change’ to account for the
linguistic situation found today in Japan. However, none of his
‘laws’ are supported by comparative method evidence, and
furthermore, his alleged laws are based exclusively on the dialects
of Japanese available in written texts.
Quite contrary, in his series of articles, Kokugo sho-hôgen no
akusento gaikan [An overview of the accent of the national-
language dialects], Hattori (1931a) has demonstrated that when the
comparative method is properly applied, a reconstruction of earlier
accentual forms of the Japonic languages can be pursued, not
completely dependent on written records. Hattori has been the only
linguist who understood and properly used the methodology of the
comparative method in Japan. Oddly enough, although the
kokugogaku scholars’ central focus is on revealing what earlier
stages of the Japanese language had been like, no kokugogaku
linguists have taken advantage of the comparative method.

3.2 WESTERN LINGUISTICS

In Western linguistics, linguists have generally employed the
comparative method to reconstruct the accentual history of the
Japonic languages. The comparative method is a method of
reconstruction whereby one can hypothesize a ‘protophoneme’
(Anttila 1989: 278) on the basis of sets of correspondences coming
from different languages (Anttila 1989: 229). For example, based
on regular correspondences of pitch shapes in modern dialects of
the Japonic languages, McCawley (1977), Ramsey (1979), Thorpe
90 ACCENTUAL HISTORY

(1983)?,1 and Martin (1987) reconstruct proto accent systems (see

2.5, 2.6, 2.7, and 2.8). They also utilize old written texts in their
reconstruction. In addition, Ramsey’s work involves the inspection
of the written texts, and gives a new interpretation of the data in the
texts, which is completely different from a generally accepted
interpretation of the texts.
None of the work done by western linguists addresses or
discusses principles of ‘natural’ accentual change. Without the
principles of accentual change, it seems impossible to account for
the accurate development of accent system of modern languages
from that of Proto-Japonic. Reconstruction can be done because
sound change is regular. In other words, the phonological system of
a language does not change randomly, but rather in a very
structured way.

3.3 METHODOLOGY
3.3.1 Methods for Reconstruction of Proto-Japonic and Proto-
Ryukyuan Accent
The reconstruction of proto-forms of the Japanese and Ryukyuan
languages will in principle be done on the basis of the comparative
method. More specifically, the reconstruction will involve the
comparison of modern languages/dialects2 as well as old existing
texts such as Ruijumyôgishô [A collection of words by class] and
Bumôki [A dictionary of Shingon-shû].3
Crucial points in reconstructing the accent system of Proto-
Japonic are (i) establishment of accent classes, (ii) filling the

1
It is not certain that Thorpe uses the comparative method in his reconstruction.
Thus, here I list Thorpe with ‘?’.
2
In this book, all data of modern languages/dialects are extracted from published
materials that have extensive lists of words. Ryukyuan data are in general from the
materials that Hirayama, Ichirô Ôshima, and Masachie Nakamoto published in the
1960s. Tokyo and Kyoto data are taken from the accentual dictionaries that
Hirayama and Haruhiko Kindaichi compiled. And Kogoshima data are mostly
from Hirayama (1951).
3
Bumôki is a dictionary of words used in the Shingon sect of Japanese Buddhism.
It describes not only the meaning of the words, but also their pronunciation. This
dictionary was written in 1687.
 HISTORICAL LINGUISTIC METHODS 91

classes with phonological characteristics, and (iii) explanation of

the development of modern accent with phonetically reasonable
accent changes. These points are discussed in detail below.

3.3.2 Establishment of Accent Classes

In order to establish accent classes, accentuations of two or more
languages will be compared. Let me take a reconstruction of Proto-
Japonic disyllabic nouns as an example (taken from chapter 7 of
this book). For the establishment of Proto-Japonic disyllabic accent
classes, the Proto-Ryukyuan classes with those of Proto-Mainland
Japanese are compared. Proto-Ryukyuan disyllabic nouns have
three and Proto-Mainland Japanese disyllables have five accent
classes. As shown in (2), there are eight correspondence sets. This
leads me to a reconstruction of eight accent classes for Proto-
Japonic disyllables.

(2) Accentual correspondence between Proto-Ryukyuan and Proto-

Mainland Japanese disyllables
(a)
PR PMJ
*`OO˘ HH(L) *`OO HH(H)
‘nose’ (2.1) *`phana˘ *`pana

(b)
PR PMJ
*`OO˘ HH(L) *O˘O HL(L)
‘stone’ (2.2) *`/isi˘ *i˘si

(c)
PR PMJ
*_OO˘ LH(L) *_OO LL(H)
‘flower’ (2.3) *_phana˘ *_pana
92 ACCENTUAL HISTORY

(d)
PR PMJ
*ooOˆ LHF ~ LHH(L) *_OO LL(L)
‘bone’ (2.3) *phu:nïˆ *_poney

(e)
PR PMJ
*_OO˘ LH(L) *OÈO LH(H)
‘shoulder’ (2.4) *_khata˘ *kaÈta

(f)
PR PMJ
*ooOˆ LHF ~ LHH(L) *OÈO LH(H)
‘sea’ (2.4) */u:miˆ *uÈmyi

(g)
PR PMJ
*_OO˘ LH(L) *OO˘ LH(L)
‘rain’ (2.5) *_/amï˘ *amey˘

(h)
PR PMJ
*ooOˆ LHF ~ LHH(L) *OO˘ LH(L)
‘shadow’ (2.5) *kha:giˆ *kagey˘

3.3.3 Filling the Classes with Phonological Characteristics

After establishing accent classes, they will be filled with
phonological characteristics such as locus, register, and vowel
length. Phonetic pitch shape will be also assigned to each accent
class, determined by the phonemic information; e.g., the phonetic
pitch of /O˘O/ should be HL. These procedures result in the
following Proto-Japonic accent shown in (3), where Proto-Japonic
forms are added to the correspondences in (2).
 HISTORICAL LINGUISTIC METHODS 93

(3) Reconstruction of Proto-Japonic disyllabic accent classes

(a)
PR PMJ PJ
*`OO˘ *`OO *`OO
HH(L) HH(H) HH(H)
‘nose’ (2.1) *`phana˘ *`pana *`pana

(b)
PR PMJ PJ
*`OO˘ *O˘O *O˘O
HH(L) HL(L) HL(L)
‘stone’ (2.2) *`/isi˘ *i˘si *e˘si

(c)
PR PMJ PJ
*_OO˘ *_OO *_OO
LH(L) LL(H) LL(L)
‘flower’ (2.3) *_phana˘ *_pana *_pana

(d)
PR PMJ PJ
*ooOˆ *_OO *_ooO
LHF ~ LHH(L) LL(L) LLL(L)
‘bone’ (2.3) *phu:nïˆ *_poney *_po:nay

(e)
PR PMJ PJ
*_OO˘ LH(L) *OÈO LH(H) *OÈO LH(H)
‘shoulder’ (2.4) *_khata˘ *kaÈta *kaÈta

(f)
PR PMJ PJ
*ooOˆ *OÈO *ooÈO
LHF ~ LHH(L) LH(H) LLH(H)
‘sea’ (2.4) */u:miˆ *uÈmyi *u:Èmyi
94 ACCENTUAL HISTORY

(g)
PR PMJ PJ
*_OO˘ *OO˘ *OO˘
LH(L) LH(L) LH(L)
‘rain’ (2.5) *_/amï˘ *amey˘ *amay˘

(h)
PR PMJ PJ
*ooOˆ *OO˘ *ooO˘
LHF ~ LHH(L) LH(L) LHH(L)
‘shadow’ (2.5) *kha:giˆ *kagey˘ *ka:gay˘

3.3.4 Explanation of the Development of Modern Accent with

Phonetically Reasonable Sound Changes
After reconstructing proto accent, what follows is an explanation of
the development of modern languages from the proto-form. The
explanation has to be reasonable with natural accentual changes.
For example, as explained in chapter 4, accent shifts rightward in
the Japonic languages -- leftward changes rarely occur. Therefore,
the change from /O˘O/ to /OO˘/ is natural, but the other direction is
unnatural. It is also natural for similar accent types to merge. The
development of Proto-Ryukyuan disyllabic accent from Proto-
Japonic shown in (4) demonstrates this. After Accent Shift, accent
classes in (4a) and (4b) merged -- both are high-level in isolation.
Three accent classes in (4c, e, and g) also merged -- all the
accentuations are initial low (either LH or LL in isolation). And
another merger involves initial-low nouns with a long vowel in the
initial syllable.4
The rules of sound change occurring in a language generally
affect all words that meet conditions for application of the rules. In
other words, in general there are no rules that apply, for example, to
disyllabic nouns, but not to trisyllabic nouns, where both types of

4
Facts only relevant to the discussion are pointed out here. For the full explanation
on the development shown below, see chapter 7.
 HISTORICAL LINGUISTIC METHODS 95

nouns have the same phonological or phonetic structure -- the only

difference is in the number of syllables or moras. If I can account
for the development of a modern accent system as a whole from its
proto-form by the same set of rules, I do not have to propose a
separate set of rules for monosyllables, disyllables, and trisyllables,
unless there is a reason to do so.

(4) Development of Proto-Ryukyuan disyllabic accent from Proto-

Japonic
accent abductive
PJ shift change assml PR
(a) *`OO = `OO
> `OO˘ = *`OO˘
(b) *O˘O > OO˘

(c) *_OO = _OO

(e) *OÈO = OÈO > _OO˘ = *_OO˘
(g) *OO˘ ª _OO˘

(d) *_ooO = _ooO

(f) *ooÈO = ooÈO > _ooO˘ > *ooOˆ
(h) *ooO˘ ª _ooO˘

Furthermore, if there is more than one possible explanation or

hypothesis for the development of modern accentuation, my
selection is primarily based on simplicity and naturalness of the
development.

3.3.5 Phonological Framework

In Japanese kokugogaku tradition (see 3.1. above), reconstruction is
in general based on a traditional Japanese phonological analysis5;
no overtly stated theory of phonology is involved. In Western
linguistics, some of the previous reconstructions are done in the
framework of generative phonology, such as Thorpe (1983),

5
The traditional analysis of Kokugogaku is purely descriptive. Systematic analyses
of phonological processes are usually not involved.
96 ACCENTUAL HISTORY

Hayata (1973), and Okuda (1975). Others have pursued their work
fundamentally by taking advantage of the comparative method.
There are quite a number of synchronic theoretical analyses of
Japanese and Ryukyuan pitch-accent systems in the framework of
segmental phonology, autosegmental phonology, and metrical
phonology. I will introduce one analysis from each framework
below. They are in order McCawley’s (1977) segmental analysis,
Haraguchi’s (1977) autosegmental analysis, and Inaba’s (1996)
metrical analysis.
One of the well-known studies in the segmental approach is
McCawley (1977). McCawley analyses Japanese pitch accent using
the notion of ordered phonological rules. For example, McCawley
(1977: 261-262) gives the following three ordered rules to predict
the pitch of a phrase in Tokyo Japanese.

(5) (a) Make every mora H.

(b) Make every mora after the first mora of an accented
syllable L.
(c) (i) Make the first mora L if the second mora is H.
(ii) Make the first mora L if the second mora is H and not
in the same syllable.

In the autosegmental approach, segments and tones are presented

at different tiers or levels. A word has a tone melody at the tonal
tier, and it links to segments by rules. For example, Haraguchi
(1977) claims that Tokyo Japanese has a HL tone melody and it
links to segments as shown in (6).

(6) inoti ‘life’ in Tokyo Japanese (Haraguchi 1977: 7-18)

/ínoti/ ínoti ínoti

‡ | | ‡ | |/
HL HL HL

In the metrical approach, the systems of pitch accent are

analysed utilizing foot structures and rules. For instance, Inaba
 HISTORICAL LINGUISTIC METHODS 97

(1996: 138) analyses the word makudonarudo ‘McDonald’s’ in

Tokyo Japanese as shown in (7). In Inaba’s analysis, feet6 are built
directly on moras, which are primitive timing units in Japanese, and
two different levels of foot parsing are recognized: initial binary
foot parsing and surface ternary foot parsing. Inaba’s metrical
approach correctly predicts the pitch shape of the word.

(7) makudonárudo in Tokyo Japanese

- (+ -) (+ -) -
m m m m m m
Initial Foot Parsing: m a k u d o n a r u d o

+
Ø (<-> + -) (+ -) -
m m m m m m
Surface Foot Parsing: m a k u d o n a r u d o

All the theories mentioned above describe pitch accent

phenomena in a proper way and predict the pitch patterns of words
and phrases. My reconstruction will not be entirely in a particular
framework of phonology, but I will make use of recent findings in
phonological theory whenever they help me to explain certain
linguistic phenomena in the history of Japanese and Ryukyuan
languages.
Using the methods discussed above, I will reconstruct the proto-
accent systems (i.e., Proto-Ryukyuan, Proto-Mainland Japanese,
Proto-Japonic) and also explain how the proto systems have
evolved to the systems of the modern dialects.

6
Feet are metrical constituents, the minimal bracketed units. (Hayes 1995: 40)
 4

ACCENT CHANGE
4.1 INTRODUCTION

T he purpose of this chapter is to determine whether there are

natural accent changes in languages. The chapter consists of
two main parts: one is concerned with the Japonic languages
(Japanese and Ryukyuan) and the other focusing on languages
other than Japonic. Each part will discuss characteristics of accent
changes that are seen in languages.
The Japonic section is important for my reconstruction of earlier
forms of Japanese and Ryukyuan accent, because my
reconstruction is based on natural accent changes discussed in this
section. The second part will present some insight on accent change,
comparing Japonic accent changes with those of other languages.

4.2 ACCENT CHANGE IN JAPONIC LANGUAGES

This section discusses diachronic accent changes that took place in
the history of Japonic languages as well as synchronic processes.
Regarding diachronic changes, there have been a number of studies
on historical written materials including Ruijumyôgishô, Bumôki,
Wajitaikanshô,1 and so on. Accent changes seen in those materials
will be discussed.
As for synchronic processes, I will examine processes seen in
careless speech in comparison with careful speech, different accent
patterns between two different generations, and accent shifts in

1
Ruijumyôgishô, also called Myôgishô, is a Chinese-Japanese dictionary written in
the eleventh century. This is one of the oldest existing written records with accent
marks. Bumôki, literally ‘a book of repairing (something) forgotten’, is a
dictionary of words used in Shingonshû Buddhism. It describes not only the
meaning of the words, but also their pronunciation. This dictionary was written in
1687. Wajitaikanshô was written around the seventeen-nineteenth centuries. It is a
book on kana or Japanese syllabary.
 ACCENT CHANGE 99

relation to devoiced vowels, vowel length, and vowel quality.

4.2.1 Accentual Change in Historical Written Records

Kindaichi (1947) tries to determine rules of accent change on the
basis of both comparisons of historical materials with modern
Kyoto-type dialects and comparisons of careful speech with
careless speech.2 This is the first study that tries to formulate rules
of accent change in Japanese, based on the assumption that the
Kyoto dialect is a descendant of the Myôgishô dialect. In fact,
modern Kyoto dialect is not a direct descendant of that of the
Myôgishô. This is obvious from correspondences between
Myôgishô accent and Kyoto accent (see Chapter 6). I believe that
there were probably various Kyoto-type dialects in the eleventh
century, just as there are now. The Kyoto dialect has probably
developed from one of the Kyoto-type dialects similar to the dialect
of Myôgishô existing at that time. Another dialect that Kindaichi
mentions in his study is Toyama dialect. As far as the accent
correspondences based on Hirayama’s word list (1935: 28-51) are
concerned, the Toyama dialect is not a descendant of the Myôgishô
dialect, either. It is not known whether other Kyoto-type dialects
are direct descendants from the Myôgishô. A claim of historical
change based on uncertain ascendant-descendant relationships is
hardly sound. Therefore, when discussing historical accent change,
I do not compare modern dialects of Japanese with dialects in the
historical materials.
In order to see the patterns of diachronic accent changes,
Kindaichi (1960: 45-6) compares the accentual patterns in various
historical materials written in different eras. The language
described in those various materials is considered to be the dialect
spoken in the old capital, Kyoto. The data shown in (1) include
accentuation of monosyllabic, disyllabic, and trisyllabic nouns both
in isolation and when followed by the topic marker -wa (Top.)
As these data show, by the sixteenth century monosyllabic nouns
are two-mora in isolation, but they are one mora when followed by

2
Kindaichi’s comparisons of careful-careless speech will be discussed in 4.2.2.
100 ACCENTUAL HISTORY

a particle, e.g., mi ‘fruit’ > mi: ~ mi-wa ‘fruit (Nom.)’.

From the philological evidence in (1), it is obvious that the
number of accent distinctions became fewer as time went by. This
is a result of merging one accent class with another. A five-way
distinction of Myôgishô monosyllabic accent became three-way by
the time of Wajitaikanshô. As for disyllables, five distinctions
became four by losing one accent class, and the eight distinctions
of trisyllabic accent in Myôgishô became a five-way distinction by
the nineteenth century.
As presented in (1), Kindaichi believes that the language of
Myôgishô developed into modern Kyoto Japanese. However, his
chart in (1) has a discrepancy. Myôgishô LLH(H) (1o) and
Myôgishô LLF(H) (1p) should have merged during the fourteenth-
sixteenth centuries because both were HLL(L) at the time. There is
no reason why these two classes should not have merged. This is
another piece of evidence that modern Kyoto is not a direct
descendant of the language used in Myôgishô.

(1) Accentual patterns in various historical materials3

I. Monosyllables
Myôgi- Shiza Bumôki Wajitaikan Modern
shô kôshiki -shô4 Kyoto
12c. 12-14c. 14-16c. 17-19c.
(a) H(H)
> H(H) = HH ~ = HH ~ = HH ~
(b) R(H) H(H) H(H) H(H)

(c) F(H) > F(L)

> HL ~ = HL ~ = HL ~
(d) R(L) > H(L) H(L) H(L) H(L)

3
Examples are from Myôgishô.
4
Kindaichi uses three or four materials for each era (see Kindaichi 1960: 30). For
convenience, only one of them is listed for each era.
 ACCENT CHANGE 101

(e) L(H) = L(H) = LH ~ = LH ~ = LH ~

L(H) L(H) L(H)

Examples: (a’) mi ‘fruit’, mi-wa ‘fruit (Top.)’

(b’) su ‘nest’, su-wa ‘nest (Top.)’
(c’) fi ‘day’, fi-wa ‘day (Top.)’
(d’) fa ‘tooth’, fa-wa ‘tooth (Top.)’
(e’) ki ‘tree’, ki-wa ‘tree (Top.)’

II. Disyllables
Myôgi- Shiza Bumôki Wajitaikan Modern
shô kôshiki -shô5 Kyoto
12c. 12-14c. 14-16c. 17-19c.
(f) HH(H) = HH(H) = HH(H) = HH(H) = HH(H)

(g) HL(H) > HL(L)

> HL(L) = HL(L) = HL(L)
(h) LL(H) = LL(H)

(i) LH(H) = LH(H) = LH(H) > LH ~ = LH ~

LL(H) LL(H)

(j) LF(H)6 > LF ~ = LF ~ = LF ~ = LF ~

LH(L) LH(L) LH(L) LH(L)

Examples: (f’) kaze ‘wind’, kaze-wa ‘wind (Top.)’

(g’) kawa ‘river’, kawa-wa ‘river (Top.)’
(h’) yama ‘mountain’, yama-wa ‘mountain (Top.)’
(i’) sora ‘sky’, sora-wa ‘sky (Top.)’
(j’) ame ‘rain’, ame-wa ‘rain (Top.)’

5
Kindaichi uses three or four materials for each era (see Kindaichi 1960: 30). For
convenience, only one of them is listed for each era.
6
Note that there are four manuscripts of Myôgishô, and not all of them show a
falling pitch.
102 ACCENTUAL HISTORY

III. Trisyllables
Myôgi- Shiza Bumôki Wajitaikan Modern
shô kôshiki -shô7 Kyoto
12c. 12-14c. 14-16c. 17-19c.
(k) HHH(H) = HHH(H) = HHH(H) = HHH(H) = HHH(H)

(l) HHL(H) > HHL(L)

> HHL(L) = HHL(L)
(m) HLL(H)
> HLL(L)
(n) LLL(H)
> HLL(L)
(o) LLH(H) = LLH(H) > HLL(L) = HLL(L)

(p) LLF(H) > LLH(L) > HLL(L) > HLL(L) > HHL(L)

(q) LHH(H) = LHH(H) = LHH(H) > LLH ~

LLL(H)

(r) LHL(H) > LHL(L) = LHL(L) = LHL(L)

Examples: (k’) katati ‘shape’, katati-wa 'shape (Top.)'

(l’) azuki ‘red beans’, azuki-wa 'red beans (Top.)'
(m’) omote ‘surface’, omote-wa 'surface (Top.)'
(n’) tikara ‘strength’, tikara-wa 'strength (Top.)'
(o’) inoti ‘life’, inoti-wa 'life (Top.)'
(p’) fitoe ‘one layer’, fitoe-wa 'one layer (Top.)'
(q’) usagi ‘rabbit’, usagi-wa 'rabbit (Top.)'
(r’) kabuto ‘helmet’, kabuto-wa 'helmet (Top.)'

Looking into these data, it is clear that some accentual patterns

7
Kindaichi uses three or four materials for each era (see Kindaichi 1960: 30). For
convenience, only one of them is listed for each era.
 ACCENT CHANGE 103

underwent changes while others have remained the same. Accent

classes in (1a), (1f), and (1k), namely high-level atonic, kept their
accentuation. High-level pitch seems to be very stable. On the other
hand, other nouns underwent some change(s). In (2), all changes
seen in the historical materials are listed. There are seven types of
changes.

(2) Accentual changes seen in historical materials

(a) Simplification of contour pitch: R > H, F > H, F > L

(b) Lowering pitch of particles: F(H) > F(L), R(H) > H(L),
HL(H) > HL(L), LF(H) > LH(L), HHL(H) > HHL(L),
HLL(H) > HLL(L), LHL(H) > LHL(L), LLF(H) >
LLH(L)
(c) Change of final-high to initial-high: LL(H) > HL(L),
LLH(H) > HLL(L), LLL(H) > HLL(H) > HHL(L)
(d) Change in domain: LH(H) > LH ~ LL(H), LLH(H) > LLH ~
LLL(H)
(e) LF(H) > LF ~ LH(L)
(f) Low Tone Spreading: LHH(H) > LLH(H)
(g) LLH(L) > HLL(L)

First, contour pitches8 seem to be unstable. Rising pitch became

high (i.e., R > H) by the fourteenth century; e.g., su-wa R(H) ‘nest
(Top.)’ became H(H). Most falling pitch underwent leveling by the
sixteenth century except for LF ~ LH(L) in (1j); nouns with this
accent retain falling pitch on the final syllable in isolation, but lose
it when followed by a particle, e.g., ame LF ‘rain’, ame-wa LH-L.
According to Kindaichi (1960: 34), in addition to the accentuation
listed in (1), in Myôgishô there are also words with HF, FL, RH,
and RL pitches, although they are few in number, e.g., mizo HF
‘gutter’, nizi FL ‘rainbow’, kisa RH ‘elephant’, hagi RL ‘shank’. In

8
Contour pitches such as rising pitch and falling pitch can be analysed as a
sequence of two level pitches. That is, R-pitch is a sequence of L-pitch and H-
pitch. F-pitch is a sequence of H-pitch and L-pitch. See also Chapter 3.
104 ACCENTUAL HISTORY

addition to the changes shown above, Kamei et al. (1964: 128-129)

discovered more accentual patterns with contour pitch that
underwent changes. They are listed with some examples in (3).

(3) Additional accentual changes found in historical materials

by Kamei et al.

Myôgishô Shizakôshiki
12c. 12-14c.
(a) HF > HL
(b) LHF > LHL
(c) HHF > HHL
(d) LFL > LHL

Examples: (a’) ha-ga ‘leaf (Nom.), ka-mo ‘mosquito (Add.)’

(b’) kasa-mo ‘umbrella (Add.)’
(c’) take-mo ‘bamboo (Add.)’
(d’) ame-ga ‘rain (Nom.)’

Having closely examined the accentual changes involving

contour pitches, I formalize the changes in (4). The rule in (4i)
must precede the one in (4ii). That is, F-pitch becomes L when
preceded by H-pitch. Other than that circumstance, contour pitches
generally become high. As the autosegmental analysis in (4i)
illustrates, a HL-noun with two TBUs has an underlying HL
melody. The H-tone is linked to both TBUs and the L-tone is
associated with only the final TBU. This rule disconnects the H-
tone from the final TBU. As a result, one TBU receives one tone;
there is no double-linked TBU.
In (4ii), there are two examples. First, a monosyllabic noun with
F-accent has a HL melody. Both tones are associated with one TBU.
The rule in (4ii) eliminates the linkage between L-tone and a TBU;
thus, the H-tone remains connected. Another example is a
monosyllabic noun with R-accent. It has a LH melody and both
tones are linked with one TBU. The rule in (4ii) also simplifies this
contour pitch by delinking the L-tone from the TBU. Since the
 ACCENT CHANGE 105

rules simplify contour pitches, I call this phenomenon Contour

Pitch Simplification throughout this book.
(4) Contour Pitch Simplification

(i) F > L / H __

e.g., H F HL
O O OO
| /| > | |
H L H L

(ii) F, R > H

e.g., F H
O O
| > |
HL H

e.g., R H
O O
| > |
LH H

The second type of accent change is a lowering of the pitch of a

particle (2b). As far as the data in (1) are concerned, this change
was completed by the fourteenth century. Kindaichi (1960: 36-42)
claims that Myôgishô nouns had already started lowering the pitch
of particles and there were already alternative forms such as HL(H)
~ HL(L), LF(H) ~ LF(L), HHL(H) ~ HHL(L), HLL(H) ~ HLL(L),
LHL(H) ~ LHL(L), and LLF(H) ~ LLF(L). Kindaichi explains that
a ‘bond’ between nouns and particles became stronger and the
particles lost their original accent.9 As a result, particles take the

9
Kindaichi (1960: 36) mentions that this explanation on lowering pitch of
particles originated with Kamei Takashi, but gives no reference.
106 ACCENTUAL HISTORY

pitch of the preceding syllable. That is, particles are L when

preceded by low pitch; particles are H when preceded by high
pitch.
The third type of accent change is one of final-high nouns to
initial-high. This is seen in disyllabic and trisyllabic nouns. The
change from LLL(H) to HHL(L) is included here as well, because
this basically involves the same process followed by Accent Shift.
The change from LLL(H) to HHL(L) can be explained by Initial
Accent Gain and Accent Shift, namely LLL(H) > HLL(L) >
HHL(L). The Initial Accent Gain rule gives an accent on the initial
syllable, and the rule of Accent Shift shifts the accent onto the
following syllable. Furthermore, the change from final-high to
initial-high probably involves another stage between LLL(H) and
HLL(L), namely HLL(H).
In (2d), final-high nouns were LH and LLH in isolation, and they
were LH(H) and LLH(H) respectively when followed by a particle;
i.e., the final syllable of words in this pattern was high. However,
after the seventeenth century, those nouns expanded the domain of
accent to a ‘phonological word’ (i.e., a noun and a particle). Thus,
only the final syllable was high both in isolation and in a phrase (=
a ‘phonological word’). In Kagoshima dialect the domain of accent
is a phrase, as well (see Chapter 6).
According to the change in (2e), Myôgishô nouns with LF(H)
had developed to have alternative pitch shapes: LF in isolation and
LH(L) when followed by a particle. This change is not due to a
single event, but it involves the changes in (2b) and (2d). As
Kindaichi claims, in Myôgishô, a particle had its own accent, which
was high pitch. When the particle lost its accent, LF(H) became
LF(L). And then nouns with this accent expanded the domain of
accent to a phrase. As a result, they became LF in isolation, but
they are LH(L) in a phrase. An autosegmental analysis shown in (5)
gives us a clear picture of this phenomenon. (5a) illustrates that a
high tone is associated to the final syllable and a low is also
connected with the same syllable because there is no other TBU
available for the L-tone. In (5b), when a particle is added, the L-
tone of the HL melody is relinked and linked to the particle.
 ACCENT CHANGE 107

(5) LF(H) > LF ~ LH(L)

(a) OO > OO
/| /|
HL HL

(b) OO-O > OO-O

/| | |
HL HL

Another type of accent change seen in written documents is

rightward spreading of L-tone as in (2f). For example, in Myôgishô
the word usagi LHH ‘rabbit’ became LLH by the nineteenth
century. An autosegmental analysis in (6) illustrates this
phenomenon. In Myôgishô, a L-tone was associated with the first
syllable of the word and a H-tone was with both the second and the
final syllable. By the nineteenth century, the L-tone became
connected with the second syllable as well, and the association line
between the second syllable and the H-tone was delinked. As a
result, the H is connected only with the final syllable.

(6) Low Tone Spreading

12th c. 17-19th c.
usagi usagi ‘rabbit’
| | > |/ |
L H L H

Finally, the change in (2g) is very similar to the change in (2c).

Both changes result in initial-high accent. The difference is whether
or not the final syllable is accented.
In written materials, some words are recorded with different
accentuation in different materials in the same era. This presumably
shows that those words had alternative accent forms when they
were recorded. According to Sakurai (1978: 168), in Jusei Rongi
the words for kokoro ‘heart’, inoti ‘life’, kotahu ‘to answer’, etc.
108 ACCENTUAL HISTORY

are recorded with three different accentuations: LLH, HLH, and

HLL. Those words must have been pronounced in three different
ways at the time of Jusei Rongi. He also says that the change, LLH
> HLL, was in process at that time, and the pitch shape of HLH
was the middle stage from LLH to HLL. In addition, although it is
said that nouns with two peaks such as HLH are rare, HLH accent
is seen not only in historical materials but also in modern dialects
of Japanese. In the Narada dialect of Yamanashi, the word for
‘heart’ is kokoro HLH(L); cf. asagao HLHL(L) ‘morning glory’
(Uwano 1977: 293).

4.2.2 Accent in Careful Speech and Connected Speech

It has been pointed out that there is a difference in accent between
careful speech and careless speech in Japanese and also that
alternative accentuation is due to change a language is undergoing.
(See Kindaichi 1947 and Akinaga 1966). This section will
investigate what types of accentual changes are involved in this
process.
(7) lists changes with respect to careful and connected speech
that are found in the literature. The first type of change is shifting
an accent onto the following syllable. For example, as a result of
this change, nouns with HL(L) accent, for example, become HH(L).
I call this kind of accent shift Rightward Accent Shift. As shown in
(7a), there are some examples from Shizuoka dialect. In this dialect,
ame-ga ‘rain (Nom.)’ is HL-L when uttered carefully, but it is HH-
L when uttered carelessly. In Chiba, LHL-nouns (/OO˘O/) become
LLH (/OOO˘/) in careless speech. Notice that Shizuoka nouns keep
high pitch on the syllable from which the accent shifted (i.e., HL(L)
> HH(L)), but in Chiba the syllable previously accented becomes
low after the accent shifts onto the following syllable (i.e., LHL >
LLH). This is due to a phonetic rule that the dialects have.
 ACCENT CHANGE 109

(7) Accent change in connected speech

a Rightward Accent Shift Examples
HL(L) > HH(L), ame-ga ‘rain (Nom.)’, tori-ga
HLL > HHL ‘bird (Nom.)’ in Shizuoka.
(Kindaichi1947: 43)
LHL > LHH seen in Shizuoka, but no
examples are given.
(Kindaichi 1947: 39)
LH(L) > LL(H), LHL > hana-ga ‘flower(Nom.)’,
LLH yama-ga ‘mountain (Nom.)’
in Chiba. (Kindaichi 1947:
34-5 and 40)
b Atonic to Tonic Examples
i LL > HL seen in Yamagata and
northern Miyagi Prefectures,
but no examples are given.
(Kindaichi 1975: 58)
HHH > HLL seen in Mie and Kôchi, but no
examples are given.
(Kindaichi 1947: 28-9)
HHH > HHL katati ‘shape’, sakura
‘cherry’, etc. in Fukui.
(Kindaichi 1947: 27)
ii LL(L) > kaze ‘wind’ and tori ‘bird’ in
HH(L)/LH(L) Nîgata. (Kindaichi 1975: 249)
c Low Tone Spreading Examples
LH(H) > LL(H), sora-ga ‘sky (Nom.)’, suzume
LHH > LLH ‘sparrow’ in Kochi.
(Kindaichi 1947: 17)
otoko ‘male’, tikara ‘strength’
in Hiroshima. (Kindaichi
1974: 55)
d LLH > HLH seen in Saitama, but no
examples are given.
(Kindaichi 1947: 35)
110 ACCENTUAL HISTORY

Concerning the second type of change, there are two types, as

illustrated in (7b). The first one is a change from atonic to initial
accent. As (7bi) shows, according to Kindaichi (1975: 58), in the
Shônai dialect of Yamagata Prefecture and in northern Miyagi
Prefecture, LL-words become HL in careless speech.10 In Mie and
Kôchi dialects high-level atonic becomes initial-high in careless
speech (Kindaichi 1947). These changes are called Initial Accent
Gain (see 4.1.1). Moreover, in the dialect of Fukui certain nouns
are HHH when uttered with care, but they are HHL in connected
speech. Examples are given in (7bi).
The other kind of change categorized under Change from Atonic
to Tonic is in (7bii). Kindaichi (1975: 249) says that in the
Murakami dialect of Niigata Prefecture, LL(L)-nouns become
either HH(L) or LH(L) in careless speech, e.g., kaze ‘wind’.
In addition to this, in the dialect of Niigada, HL(L)-nouns also
shift their accent onto the second syllable, resulting in HH(L). This
change is called Rightward Accent Shift (see (7a) above). In
Murakami there are three accentual distinctions in disyllables in
careful pronunciation: HL(L), LH(L), and LL(L). Note that LH(L)
in (8c) does not undergo a change.

(8) Careful and careless speech in Murakami dialect of Niigata

careful careless examples

(a) HL(L) ‡ HH(L) haru ‘spring’, hasi ‘chopsticks’
(b) LL(L) ‡ HH(L)/LH(L) kaze ‘wind’, tori ‘bird’
(c) LH(L) = LH(L) ame ‘rain’, sora ‘sky’, etc.

When nouns are uttered carelessly, there are only two distinctions
because LL(L)-words merge with either LH(L)- or HH(L)-words.
(9) describes this phenomenon.

10
Unfortunately, Kindaichi (1975: 58) does not present any examples or the
sources where these examples can be found.
 ACCENT CHANGE 111

(9) Accent merger in Murakami

(a) LL(L) merges with HL(L) (b) LL(L) merges with LH(L)
careful careless careful careless
HL(L) HL(L) ‡ HH(L)
‡ HH(L)
LL(L)
LL(L)
‡ LH(L)
LH(L) = LH(L) LH(L)

Another change seen in careless speech is Low Tone Spreading

(7c), which is also found in historical materials. According to
Kindaichi (1947), in the Kôchi dialect, nouns with LH(H) or LHH
become LL(H) or LLH when carelessly uttered. Examples are
shown in (7c). Kindaichi (1947) and Akinaga (1966) also report
that in Hiroshima, words such as otoko ‘male’, tikara ‘strength’,
sakura ‘cherry’, katati ‘shape’, etc. are LHH in careful speech, but
LLH in careless speech.
The change in (7d) causes two peaks in words. This is seen in the
dialect of Saitama. In this dialect, nouns with LLH pitch become
HLH in connected speech (Kindaichi 1947: 35). Unfortunately,
Kindaichi does not give any examples. In addition, Hattori (1933)
claims that a series of changes LLH > HLH > HLL is natural, but
he gives no reasons for his claim. However, judging from his article,
his claim is probably based on his observation of Japanese dialects,
although he does not say in what dialects this phenomenon is
taking or has taken place. The change from LLH to HLH is
explained by Initial Accent Gain. The change from HLH to HLL is
called Double Accent Simplification in this book.

4.2.3 Accent Shift in Relation to Vowel Devoicing

In many modern Japanese dialects, high vowels between voiceless
consonants become devoiced (e.g., voiceless). These devoiced
vowels are indicated with a circle underneath when necessary, e.g.,
tu8˘ku ‘arrive’. When an accented syllable contains a devoiced
vowel, the accent often shifts onto the following syllable. For
112 ACCENTUAL HISTORY

example, in Tokyo, the word asita ‘tomorrow’ belongs to LHH-

accent class, but it is pronounced with LLH pitch because of the
devoiced syllable, si8 in asita (Kindaichi 1951). Vance (1987: 86)
also discusses this phenomenon of accent shift. Pointing out the
following examples (10), he explains that accent shifts rightward.
Initial-accent disyllabic nouns often shift the accent onto the final
syllable. For trisyllabic nouns with accent on the second syllable,
the accent sometimes remains in the same position when the final
syllable contains a devoiced vowel.

(10) Rightward Accent Shift in Tokyo

/tu8˘ku/ ~ /tu8ku˘/ ‘arrive’

/fu8˘ku/ ~ /fu8ku˘/ ‘blow’
/kaku8˘su/ ~ /kaku8su˘/ ~ /kaku˘su8/ ‘hide’

In addition, McCawley (1978: 266) says: “There is evidently a

rule shifting accent to the right when it falls on a voiceless syllable.
However, this rule has exceptions, both systematic and
idiosyncratic. Accent is never shifted off of an adjective stem onto
an ending nor off of a noun onto a particle.” And he gives the
following two examples: yorosi8˘ katta ‘was okay’ (ending -katta)
and mati8˘ kara ‘from the town’ (particle kara). In both examples,
the accent on devoiced syllables stays on the syllable even though
it contains a devoiced vowel.
According to Hirayama et al. (1967), in Ryukyuan, accent shifts
in relation to the devoicing of vowels are also found. For example,
in Ishigaki dialect of Yaeyama Ryukyuan, the word for ‘person’ is
`pï8tu LH(L); other high-register disyllables are HL(L) (see also
5.5.1.1). In Ishigaki, when HL(L)-nouns contain a devoiced vowel
in the initial syllable, the accent shifts onto the second syllable.

4.2.4 Accent Change in Relation to Vowel Length

On the basis of historical written records, Sakurai (1978) proposes
that lengthening of vowels caused accent change. In Amakusa-bon
 ACCENT CHANGE 113

Isoho Monogatari,11 the word for ‘head’ is ko:be with a long [o].
Sakurai explains that this is historically a compound consisting of
kabu ‘head’ and be ‘direction’ or kami ‘hair’ and he ‘edge’, and
that the long vowel [o:] in ko:be must have been derived from the
change kau-fe > ko:be. In Myôgishô, kaube is marked with LHH. In
Shiza Kôshiki (12-14th centuries), kaube is not only with LHH but
also the alternative accentuation LLH. Sakurai suggests that the
alternative pronunciation is due to vowel lengthening mentioned
above.

4.2.5 Accent Shift in Relation to Vowel Quality

Japanese accent is also sensitive to vowel quality. In some cases,
the shifting of accent onto the following syllable depends on vowel
quality within the syllable. Wada (1958) reports that in Takamatsu,
some 2.2-nouns12 are HL(L) and others are LF(L); both are
phonemically the same. HL-words have either /i/ or /u/ in the final
syllable and the LF-words have /a/, /e/, or /o/ in the final syllable.
That is, if a word contains a high vowel (i.e., either /i/ or /u/), the
accent is on the initial syllable. On the other hand, if the word has
non-high vowels (i.e., /a/, /e/, or /o/), it is realized as LF(L).
Examples are isi-ga HL-L ‘stone (Nom.)’ and uta-ga LF-L ‘song
(Nom.)’.

4.2.6 Accent conditioned by Syllable Structure

Kindaichi (1974: 177) observes that in Tokyo, nouns with LHL(L)
pitch such as kaki˘ne ‘fence’ in general retain their accent on the
second syllable even when uttered in a phrase. However, some of
LHL(L)-nouns become LHH(H) when followed by the possessive
marker no. The final syllable of those words consists of two moras.
Examples are sato: ‘sugar’, nioi ‘smell’, and nihon ‘Japan’.13

11
This is an Early Modern Japanese text written in the Latin alphabet in late
sixteenth century.
12
‘2.2’ is one of the five accent classes for Myôgishô disyllabic nouns.
13
The word for ‘smell’ is pronounced with three syllables in careful speech, but it
is pronounced with two syllables in careless speech. In the latter case, the final
syllable consists of two moras.
114 ACCENTUAL HISTORY

A similar phenomenon to this is also seen in many Ryukyuan

dialects (see Hirayama et al. 1966). For example, in the Kamishiro
dialect of Ryukyuan, some final-accent nouns are HL(L) or HLL(L)
pitch, e.g., hui˘ HL(L) ‘voice’, wui˘ HL(L) ‘bucket’, muQkwa˘
HLL(L) ‘bridegroom’; cf. usi˘ LF ~ LH(L) ‘mortar’. This occurs
only for words in the word structure of CVi or CVQCV.

4.2.7 Accent in Two Different Generations

According to Hirayama and Nakamoto (1964: 87), older and
younger speakers of the Hateruma dialect of Ryukyuan pronounce
some of the same words differently. For example, speakers who are
over 50 pronounce the word pana ‘nose’ with LL(L) accent, and on
the other hand, speakers who are in their 20s pronounce the same
word with HL(L). A similar phenomenon is seen in the dialect of
the Kudaka Ryukyuan. Hirayama et al. (1966: 189) report that
younger speakers pronounced low-level-atonic nouns with initial
accent. For example, older speakers pronounce pana ‘flower’ and
pasira ‘pillar’ as LL(L) and LLL(L) respectively, but for younger
ones, they are HL(L) and HLL(L).
In addition, it should be noted that Umegaki (1957) discusses
accentual change taking place in younger speakers of Osaka.
Examining junior high school students’ speech in comparison with
that of older speakers in Osaka, Umegaki reports that words with
HHL-accent in older speakers’ speech appear as either HLL, LHL,
HHH, or LLH in younger speakers’. However, it is difficult to
determine whether these changes are natural because words are
irregularly similar in accentuation with those of its neighboring
areas such as Kyoto, Hyogo, Nara, etc. For example, for younger
speakers in Osaka the word for ‘forest’ (3.4) is hayasi HHH. The
word is also HHH in Kyoto, HHL in Hyogo and Kochi, and LHL in
Toyama. Since there is high possibility of language contact, his
work will not be considered in my discussion of natural sound
change.

4.2.8 Overview
In this chapter, I have so far investigated both diachronic changes
 ACCENT CHANGE 115

and synchronic processes of accentuation in Japonic languages. Six

changes have been discussed, and they are summarized in (11).

(11) Accentual changes found in the Japonic languages

(i) Leveling of contour tones (Leveling)
(ii) Lowering of pitch of particles (Lowering)
(iii) Change from final-high to initial-accent
(Initial Accent Gain)
(iv) Change of domain
(v) Low Tone Spreading
(vi) Rightward accent shift (Accent Shift)

Changes in (11i), (11ii), and (11iv) are seen only in historical

materials. According to the texts, Leveling (11i) was completed by
the sixteenth century (with one exception) and Lowering (ii) by the
fourteenth century. The change of domain took place between the
seventeenth and nineteenth centuries. There is also a change that is
found in synchronic material, i.e., Accent Shift (11vi). Initial
Accent Gain (11iii) and Low Tone Spreading (11v) are found in
both diachronic and synchronic materials.
From a phonological and phonetic point of view, among the six
changes in (11), Initial Accent Gain, Accent Shift, and Low Tone
Spreading apparently occur under certain conditions. See (12). ‘+’
indicates a change occurs in relation to the condition under which it
is marked.
On the basis of (12), it should be pointed out that (i) Initial
Accent Gain takes place when speakers carelessly speak; (ii)
Accent Shift occurs in careless speech, and it also takes place when
an accented syllable contains a devoiced vowel; and (iii) Just like
the other two changes, Low Tone Spreading also takes place in
careless speech. This change is also seen in relation to vowel
quality and vowel length.
116 ACCENTUAL HISTORY

(12) Accentual changes in relation to their conditions

careless vowel vowel vowel
speech devoicing quality length
Initial Accent Gain +
Accent Shift + +
L-tone Spreading + + +

With a careful look at all the changes discussed above, patterns

of the changes are as follows: atonic (either high-level, low-level,
or final-high atonic) > initial-accent (i.e., HLL) > second-syllable
accent (either LHL or HHL) > final-accent (either LLH(L) or
LHH(L)) > initial- and final-high (i.e., HLH) > atonic. Changes in
(11i and ii) are only seen earlier in history. The change of domain
in (11iv) is seen throughout history. Based on the patterns of those
changes, proto accent systems of the Japonic languages will be
reconstructed in this book.

4.3 ACCENT CHANGE IN OTHER LANGUAGES

In this section I will investigate what kind of accentual changes
(both diachronic and synchronic) are seen in other languages. It is
difficult to compare Japonic pitch accent with other suprasegmental
systems such as stress accent (e.g., Bahasa Indonesia), tonal system
(Mandarin Chinese), and the like. What this section focuses on are
changes in pitch height in both tone and pitch accent languages,
and also changes in the location of stress in stress languages.
Needless to say, I exclude changes conditioned on grammatical
basis or changes caused by language contact. For example, Polish
words generally keep stress accent on the penultimate syllable,
even when suffixes are added: adding the adjectival suffix -owi to
the word nauczy@ciel ‘teacher’ (Nom. singular), causes the stress to
fall on the vowel /o/ in -owi, namely nauczyciel-ówi (Kenstowicz
1994: 552). It seems that the accent on the penultimate syllable of
nauczy@ciel shifted onto the initial vowel of the suffix -owi.
However, this ‘shift’ took place in reference to a word boundary;
i.e., it is morphologically conditioned. More importantly, this is a
 ACCENT CHANGE 117

matter of placement of stress in relation to the boundary, not a

phenomenon of accentual shift.
Although I have looked into a number of languages in different
language families such as Niger-Kordofanian, Austronesian, Uralic,
Turkic, Dravidian, Sino-Tibetan, etc., the changes in
suprasegmentals in these languages are not thoroughly discussed in
the literature. Among those I have investigated, some languages are
selected from different language families to show suprasegmental
changes seen in the languages. The languages presented in this
section are Korean, Ainu, African languages (e.g., Gwari, Kikuyu,
Fe/Fe/, etc.), Austronesian languages (e.g., Bahasa Indonesian,
Mukah, etc.), and Chinese (e.g., Mandarin and Cantonese).

4.3.1 Korean
There have been a number of studies on Korean accent (e.g.,
Ramsey 1978, Kim 1988) even though relatively few dialects of
Korean have distinctive accent (e.g., Hamkyeng dialect spoken in
the northeastern part of North Korea and Kyengsang dialect spoken
in southeastern part of South Korea). I will investigate what kind of
accentual changes are involved in the history of the Korean
language.
Ramsey (1978) contends that the accent system of the
Hamkyeng dialect is more conservative than that of the Kyengsang
dialect. This is mainly based on the fact that Hamkyeng accent and
Middle Korean (MK)14 accent are similar. Ramsey claims that the
development of Kyengsang accent can be explained by shifting an
accent of Hamkyeng one syllable to the left. Middle Korean texts
are marked with the ‘side dots’ to indicate accent of words. The
‘side dots’ which mark level tone and departing tone are indicated
by respectively L and D. A comparison of Middle Korean, South
Hamkyeng, and Kyengsang accents in (13) demonstrates Ramsey’s
claim. The accentuation of the South Hamkyeng dialect can be
determined by location of the departing tone in the corresponding

14
“[T]he language of Korean texts written in the fifteenth and sixteenth centuries.”
(Ramsey 1978: 2)
118 ACCENTUAL HISTORY

Middle Korean lexical item.

(13) Comparison of Middle Korean, South Hamkyeng, and

Kyengsang accent (Ramsey 1978: 82)
MK South Hamkyeng Kyengsang
side accent examples accent examples gloss
dots
L o kkoc-í o kkóch-i flower
D ó káp-i `o `kaps-i price
kkoc káp-i kkoch `kaps-i price of
flowers
LL oo palam-í oo plam-i wind
LD oó atúl-i óo átul-i son
DL, óo móki-ka `oo `mokwu-ka mosquito
DD
kaal móki-ka kasíl mokwu- autumn
ka mosquito
LLL ooo saytali-ká ooo saytalí-ka ladder
LLD ooó kamakwí-ka oóo kkamákwu-ka raven
LDL, oóo kamúlchi-ka óoo kámuchi-ka mullet
LDD
DLL, óoo thókkaypi-ka `ooo `thokkaypi-ka sprit
DLD,
DDL,
DDD

According to Jeong (1985), in Kyengsang dialect there are four

accent patterns for trisyllables: HLL, HHL, LHL, and LLL.15
Kyengsang nouns with HHL accent can be pronounced HLL as
long as the initial high pitch is pronounced clearly. LLL- and LHL-
nouns are also interchangeable. It seems that the initial pitch plays
a crucial role in distinguishing one accent from another.
Jeong also says that in Middle Korean the final pitch of adverbial
and prenominal forms varies between low and high, without
changing meaning of the words. He suggests the change of high to

15
It should be pointed out that Jeong’s analysis of Kyengsang accent is not
identical to Ramsey’s.
 ACCENT CHANGE 119

low at the end of words, namely …H# > …L#.16 Agreeing with Lee
Sung-nyong’s proposal in 1964, Jeong claims that in Middle
Korean, a change in accent starts from the end of a word. The
reason for this, he continues, is that the final pitch does not function
to distinguish one word from another; whether the final pitch is low
or high is not important in Middle Korean.
Another observation on Middle Korean, is in Lee Sang-ak (1985:
16). He points out that around the end of the fifteenth century, high-
level nouns were also pronounced HLH or HHL. He says that the
variations of accent existed until the end of the sixteenth century.
In the matter of Proto-Korean accent, there are two hypotheses.
One is that Proto-Korean did not have distinctive accent (Ramsey
1978). If this is correct, Middle Korean developed its accent system
and lost it by the end of sixteenth century.17 On the other hand,
Whitman (1994) argues that Proto-Korean had distinctive accent, at
least for nouns. It is not known what sorts of changes Proto-Korean
underwent to develop or to lose its accent because neither of the
hypotheses reveals how the system came to exist and to lose its
function. I leave this as it is for the time being with the hope to see
further studies in the future.

(14) Accentual changes found in Korean

(a) Leftward accent shift (e.g., HHL > HLL)

(b) Final high lowering, i.e., …H# > …L# (e.g., HHH > HHL)
(c) LHL > LLL
(e) HHH > HLH

In sum, on the basis of the discussions above, considering

variation of accent a change in progress, there are four types of
accentual changes found in Korean. (14) gives a summary of the
discussion.

16
The symbol ‘#’ indicates a word boundary.
17
Middle Korean is considered to be a precursor of Seoul dialect, which has no
distinctive accent.
120 ACCENTUAL HISTORY

As Jeong proposes, in Korean, accentual change starts from the

end of a word. This is completely opposite to the case of the
Japonic languages, where accent shifts rightward and lowering of
pitch goes from the left to the right.

4.3.2 African Languages

In this section, I will discuss tonal changes that take place in
African languages. For most of the African languages their tonal
systems consist of two or three tones, namely H-, L-, and Mid-
tones. This system is much simpler than those of Chinese or Thai,
where there are four or more tones.
Following Hyman’s discussions on tonal change (1975), my
discussion consists of two main parts, the first is on Phonetic Tone
Rules and the second is on Morphophonemic Tone Rules. I exclude
grammatically-conditioned rules from my discussion of
Morphophonemic Tone Rules.
Hyman and Schuh (1974) and Hyman (1975) discuss natural
tone rules. According to Hyman (1975), there are two types of
natural tone rules. One is a phonetic tone rule, and the other is a
morphophonemic rule. Phonetic tone rules will be discussed first,
and then morphophonemic rules.

4.3.2.1 Phonetic Tone Rules

First, Hyman and Schuh (1974) gives assimilation, simplification,
and rightward tonal shift for their natural phonetic rules. Under
assimilation, there are two subtypes, namely Vertical Assimilation
and Horizontal Assimilation. Vertical Assimilation involves a
change in tone height; either a low or high tone to a mid tone.
Horizontal Assimilation causes contour tones; either R (i.e., LH) or
F (i.e., HL).
In (15a), Hyman (1975) makes a clear distinction between
natural and unnatural vertical assimilation. Based on that, L-tones
become M-tones only when immediately followed by a H-tone.
And H-tones become M-tones only when immediately preceded by
a L-tone. For example, in Gwari /gyìwyé da#/ LHM ‘possessor of
 ACCENT CHANGE 121

money’ becomes gyìwye#da# LMM (Hyman 1975: 221-5).18

(15) Assimilation
(a) Vertical Assimilation
Natural Unnatural
L-H > M-H H-L > H-M
L-H > L-M H-L > M-L

(b) Horizontal Assimilation

Natural Unnatural
L-H > L-LH L-H > LH-H
H-L > H-HL H-L > HL-L

(c) Gwari examples

/òkpá/ > òkpa& ‘length’ ([a&] is rising tone.)
/súkNù/ > súkû˜ ‘bone’ ([û] is HL tone, [˜] indicates
a nasalized sound.)

Regarding Horizontal Assimilation, a H-tone preceded by a L-

tone becomes a R-tone, and a L-tone becomes a F-tone when
preceded by a H-tone. Since this is progressive assimilation, L- or
H-tones are not affected by a following tone. See the Gwari
examples in (15c) (Hyman 1975: 222).
Hyman also mentions that some languages undergo Complete
Horizontal Assimilation. It means that H- and L- become
assimilated completely to their preceding tones. That is, L-H-H
becomes L-L-H by undergoing L-LH-H (16). H-L-L becomes H-
HL-L and then finally H-H-L. This phenomenon is seen in Kikuyu,
e.g., /gòr/ + /írE¤/ > gòrìrE¤ ‘bought (immediate past)’

18
The symbols á, à, and a# indicate high-tone, low-tone, and mid-tone
respectively.
122 ACCENTUAL HISTORY

(16) Complete Assimilation

L-H-H > L-LH-H > L-L-H

H-L-L > H-HL-L > H-H-L

Another phonetic change that Hyman discusses is Simplification

-- one type of assimilation. Similar to Leveling of Contour Pitch
discussed earlier (see 4.1.1.), it simplifies contour tones to level
tones. For example, in Fe/Fe/ a rising tone becomes mid tone
when followed by a L-tone. Compare the following examples.

(17) Simplification in Fe/Fe/ (Hyman and Schuh 1974: 91)

[tu&u] ‘drum’
[tu#u lè] ‘this drum (near speaker)’
[tu&u la¤] ‘that drum (near hearer)’
[tu#u lì] ‘that drum (far from speaker and hearer)’

Moreover, Leben (1971: 203) reports that in Hausa, all LH rising

tones are simplified to H-level tone.
Regarding the direction of tonal shift, Hyman and Schuh (1974)
argue that tonal shift to the right is natural because it is motivated
by nothing more than neighboring tones. They consider tonal shift
to the left to be unnatural because there is an external explanation
for it. An example for natural change is as follows:

(18) Rightward tonal shift in Ewe

/lç$ç@ tìè/ ‡ lç$ç@ tíè ‘look for the pot’

In Ewe, H-tone of the word lç$ç@ ‘look for (it)’ shifts onto the initial
syllable of the word tìè and the initial syllable becomes high (i.e.,
tìè > tíè). L-tone of ‘look for (it)’ spreads to its second vowel (i.e.,
lç$ç > lç$ç$).
 ACCENT CHANGE 123

As for shifting of tones to the left, see the following Ga/anda

examples. In Ga/anda, the vowel of the perfective suffix -wú is
apocopated after a vowel. Thus, a phrase consisting of the verb và
‘shoot’ and the perfective suffix -wú becomes va&w.

(19) Leftward tone shift in Ga/anda (Hyman and Schuh 1974: 104)

/và-wú/ ‡ va&w ‘he shot (it)’

/ná dlèe-wú/ ‡ ná dle&w ‘I dipped (it) out’

Hyman and Schuh (1974) contend that after the loss of the vowel u,
the H-tone on the vowel shifts onto the syllable to its left because
there is no TBU on its right. They add: “Such [‘]extenuating
circumstances[’] probably account for most cases of shifting to the
left (Hyman and Schuh 1974: 104).” Moreover, in Wolaitta there
are examples for both rightward and leftward shift, but leftward
shift is grammatically conditioned (Amha 1996).
On the direction of tonal shift, Schuh (1978) also claims that the
most common tone spreading rules are rightward and they can be
schematically shown in (20) below (a tone in ‘X’ varies from
language to language).

(20) The most common tone spreading rules

H-tone Spreading: H LX > HHX

L-tone Spreading: LH X > LLX

See (21) below. Examples for H-tone Spreading are from Bade and
Bolanci (Schuh 1978: 226), and examples for L-tone Spreading are
from Duwai and Ngizim (Schuh 1978: 222).
124 ACCENTUAL HISTORY

(21) Examples for High and Low Tone Spreading

H LL > HH L
Bade: /n´@n kàtáw/ ‡ n´@n kâtáw ‘I returned’
Bolanci: /ku&m sàawùrà/ ‡ ku&m sáawùrà ‘ear of falcon’

LH H > LLH
Duwai: /k´$vús ∫ái/ ‡ k´$vùs ∫ái ‘it’s not a warthog’
Ngizim: /gùb´@s bái/ ‡ gùb´s bái ‘it’s not a warthog’

Schuh’s claim is supported by Maddieson (1976), who also

maintains that tones spread rightward.

(22) Maddieson’s Tone Spreading Rules in Bantu languages

High Tone Spreading: HL{L or #} > HH{L or #}

Low Tone Spreading: LH{H or #} > LL{H or #}

Contrary to Hyman and Schuh’s rightward shift hypothesis, there

are also counterexamples reported in the literature. One of them is
given by Odden (1985). According to him, Kimatuumbi has a
Retraction Rule, which shifts the H-tone from the final syllable
onto the preceding one that contains a long vowel. In (23), ‘VV’
indicates a long vowel in a syllable, and ‘.’ in ‘VV.’ means a
syllable boundary.

(23) Retraction Rule (Odden 1985: 359-60)

H H
| > /
VV.V VV.V

H H
| /
e.g., mboopo > mboopo
 ACCENT CHANGE 125

Furthermore, disagreeing with Hyman and Schuh, Schadeberg

(1977) discusses leftward movement. He states, “In the case of
non-tonal phonological phenomena, rules of various sorts are
known to operate from right to left as well as from left to right,
synchronically as well as diachronically. From a general linguistic
point of view perseveration (progressive assimilation) and
anticipation (regressive assimilation) both seem to be perfectly
normal, natural process.” However, he adds, “It is true that for
some phenomena, one direction may be more normal than the
other; e.g., t > c / i is more natural as an anticipation rule / __ i than
as a perseveration rule /i __ (Schadeberg 1977: 201-2).”

4.3.2.2 Morphophonemic Tone Rules

This section will introduce Hyman’s Morphononemic Tone Rules:
Dissimilation, Copying, Polarization, and Floating Tones.
First, Hyman (1975: 224) states, “Just as dissimilation most
frequently is bound to certain morphemes or constructions, the
same is true of tonal dissimilations.” For example, in Hausa a
sequence of two L-tones at the end of a word results in L-H when a
vowel in the final syllable is long.

(24) Dissimilation in Hausa

Underlying form
/káràntà:/ ‘to read’ ‡ káràntá: HLH

Second, “Copying refers to the process by which a syllable (most

frequently a grammatical morpheme such as a pronoun) is
considered to have no underlying tone of its own, but rather
receives its tone from a neighboring syllable” (Hyman 1975: 224).
For instance, in Kru the relative clause marker a is high after a H-
tone verb and low after a L-tone verb. That is, the tone of the
preceding verb is copied on the following marker.
Polarization is a rule that assigns a tone opposite to that of the
neighboring syllable to a morpheme – this is a completely opposite
126 ACCENTUAL HISTORY

process to Copying. For example, in Hausa the direct object

pronoun s&i: takes a H-tone when the tone of the preceding verb is
low, and a L-tone when preceded by a H-tone. (Hyman 1975:
224-5)

(25) Polarization in Hausa19

/mún ká:mà s&i:/ ‘We seized it’ ‡ mún ká:màs&í: H HL H
/mún sàyé: s&i:/ ‘We bought it’ ‡ mún sàyé: s&ì: H LH L

For a process involving Floating tones, Hyman (1975: 225) gives

examples from Central Igbo and Aboh Igbo as follows:

(26) Floating tones

Central Igbo /àgbà ´ èNwè/ ‘jaw of monkey’ ‡ àgbá èNwè
Aboh Igbo /ègbà ´ èNwè/ ‘jaw of monkey’ ‡ ègbà éNwè

The phrase for ‘jaw of monkey’ consists of two morphonemes, i.e.,

àgbá èNwè in Central Igbo and ègbà éNwè in Aboh Igbo. Focusing
on tones, although both have the same underlying form, their
surface forms are not identical. This is because in Central Igbo a
floating H-tone between two morphemes is assigned to the
preceding vowel (i.e., /à/), and in Aboh Igbo the H-tone is assigned
to the following vowel (i.e., /è/). In both cases, the original tones of
the vowels are replaced with the H-tone.

4.3.2.3 Summary
According to Hyman (1975), there are phonetic tone rules and
morphophonemic tone rules. The former rules involve Assimilation
(Vertical and Horizontal), Simplification, and Rightward Tonal

19
In Hausa, Polarization precedes Dissimilation. Furthermore, a domain of
Dissimilation is a word, while that of Polarization is a phrase. The following
example demonstrates this: /mún káràntà: s&i/ ‡ mún káràntà: s&í: ‡ múnkáràntá:
s&í: ‘we read it’ (Hyman 1975: 224).
 ACCENT CHANGE 127

Shift, and the latter Dissimilation, Copying, Polarization, and

Floating Tones. Our discussions so far are summarized in (27) and
(28).

(27) Phonetic tone rules

(i) Assimilation
Vertical – This can be either rightward or leftward. The
right-ward shift is similar to L-tone Spreading
seen in Japonic languages.
Horizontal – This can also be either rightward or leftward.
(ii) Simplification – This is similar to Leveling of Contour Pitch
seen in Japonic.
(iii) Rightward Tonal Shift – There are counter examples against
this claim.

(28) Morphophonemic Tone Rules

(i) Dissimilation – Two tones dissimilate to each other.
(ii) Copying – A grammatical morpheme receives a tone from
a stem; this is similar to a characteristic of
Japanese particles. This takes place in a
rightward direction.
(iii) Polarization – This is very similar to Dissimilation.
(iv) Floating Tones – Floating tones can shift either rightward
or leftward.

In African languages, changes involving tones vary. Although

Hyman and Schuh claim that rightward tonal changes are natural
and that leftward changes are unnatural, there are some examples of
leftward changes. However, it is likely that one language tends to
have one direction of change, not both directions. For example, in
Central Igbo floating tones attach to the preceding syllable; on the
other hand, in Aboh Igbo floating tones attach to the following
syllable. As mentioned earlier, some of the changes in African
languages are very similar to ones found in the Japonic languages.
128 ACCENTUAL HISTORY

4.3.3 Austronesian Languages

This section looks into the nature of accent shift in Austronesian
languages. My focus especially addresses the following two
questions: under what condition does accent shift and also in what
direction does it shift? In Austronesian languages, accent is ‘stress’.
Bahasa Indonesia, Mukah, and Ratahan will be described to show
characteristics of the languages.

4.3.3.1 Bahasa Indonesia

Bahasa Indonesia is a stress language, where a stress falls on a
penultimate syllable (Halim 1975). In the language, if the vowel of
a penultimate syllable is a schwa followed by a single consonant,
stress shifts to the final syllable of disyllables. If a schwa is
followed by a consonant cluster, stress does not shift. The examples
in (29) demonstrate this -- in the words in (29a and b), accent is on
their penultimate syllable, while in (29c) the accent is on the final
syllable. From this, scholars conclude that in Bahasa Indonesia,
accent shifts to the right.

(29) Accent shift in Bahasa Indonesia (Halim 1975)

(a) ádu ‘fight’
(b) ´@mpat ‘4’
(c) k´nái ‘know’

4.3.3.2 Mukah
According to Blust (1988), Mukah, spoken in Borneo, is a stress-
accent language. In general the penultimate syllable of a word
receives accent in Mukah -- monosyllabic words are also accented.
In Mukah, accent shifts to the final syllable under the following
three conditions: (i) if the penultimate syllable contains a schwa,
(ii) if the last two syllables consist of a consonant and the vowel /a/
in the penultimate syllable, and a high vowel in the last syllable
(the high vowel is lengthened), and (iii) if the nucleus of the
penultimate syllable is followed by a glottal stop (in the last case,
stress optionally shifts to the final syllable). It should be noted that
 ACCENT CHANGE 129

a schwa in the final syllable also can receive accent.

(30) Accent in Mukah (Blust 1988: 178-9)

/t´l´w/ t´l´@w ‘three’
/t´ba/ t´bá ‘well’
/pai/ paí: ‘stingray’
/da/´m/ dá/´n ~ da/´@n ‘leaf’
/ña/´m/ ñá/´m ~ ña/´@m ‘water’

The alternative forms in the last two examples above demonstrate

that an accent shifts rightward in Mukah.

4.3.3.3 Sangiric Languages

In his study of the Sangiric languages, James N. Sneddon (1984)
accounts for how Proto-Sangiric (PS) stress accent evolved in
Ratahan. In Proto-Sangiric, accent is on a penultimate high vowel
and in Ratahan accent is on the final vowel. That is to say, the
accent of Proto-Sangiric shifted from a penultimate vowel to an
immediately following vowel.

(31) Development of Ratahan accent from Proto-Sangiric

PS Ratahan
*Ríud > iúr ‘to pull’
*tían > tián ‘belly’
*súan > suán ‘to plant’
*túid > tuír ‘stump’ (Sneddon 1984: 53)

4.3.3.4 Summary
This discussion makes it clear that the similar types of accentual
shift is seen in Austronesian languages. In Bahasa Indonesian,
accent shifts rightward. For Mukah words, accent also shifts
rightward. In the development of Ratahan accent from Proto-
Sangiric, accent shifted rightward, as well.
130 ACCENTUAL HISTORY

(32) Accent changes in Austronesian languages

(a) Bahasa – rightward accent shift
(b) Mukah – rightward accent shift
(c) Proto-Sangiric to Ratahan – rightward accent shift

4.3.4 Chinese
With regard to Chinese, it is known that tones in Chinese change
depending on their environments. In this section, I am looking into
characteristics of tone sandhi.
First, according to Chao (1965: 33), Mandarin Chinese tones can
be phonetically described in numbers as shown below. Chao uses
five levels of pitch to describe tones. The first tone (i.e., 55) is
high-level, the second tone (i.e., 35) rises from level 3 and ends in
high pitch, the third tone (i.e., 214) falls from level 2 to 1 and rises
to level 4, and the fourth tone (i.e., 51) falls from high to low.
Referring to Chao’s notation, tone sandhi will be described.

(33) Mandarin Chinese tones

1st tone : 55 high-level
2nd tone : 35 rising
3rd tone : 214 falling-rising
4th tone : 51 falling

In Mandarin tone sandhi, 35 rising tone becomes high-level (i.e.,

55) when preceded by high-tone (either 55 or 35) (Chao and Yang
1962 and Cheng 1973: 44). In the following formula, ‘t’ stands for
any tone except neutral tone. The following examples are from
Chao and Yang (1962).
 ACCENT CHANGE 131

(34) Mandarin tone sandhi

35 ‡ 55 / {5} __ t

dong 55 + nan 35 + feng 55 ‡ dong nan feng 55 55 55

‘east’ ‘south’ ‘wind’ ‘southeast wind’

san 55 + nian 35 + ji 51 ‡ san nian ji 55 55 51

‘three’ ‘year’ ‘class’ ‘third year class’

In Cantonese, 53 falling tone becomes 55 (high-level) when

followed by either 55 or 53 (Hashimoto 1972 and Cheng 1973).
The pitch of level 3 is raised to 5. The following examples are
extracted from Hashimoto (1972: 112).20 Hashimoto gives
another tone sandhi rule in Cantonese. See (35) below. In this case,
falling tone 21 becomes level tone 22 when followed by either 21
or 22. Both rules basically have the same characteristic – the lower
part of the tones is raised to the same height as the following pitch.

(35) Cantonese tone sandhi

53 ‡ 55 / __ {5}
21 ‡ 22 / __ {2}

i‡IN 53 + kç:i‡ 53 ‡ i‡IN kç:i‡ 55 53 ‘should, must’

såm 53 + kåp 5 ‡ såm kåp 55 5 ‘anxious’
mA: 21 + i‡åu& 21 ‡ mA: i‡åu& 22 21 ‘sesame oil’

Regarding the direction of change, tone spreading from the left

to the right can account for Mandarin tone sandhi. However, in
Cantonese, tone spreads leftward. That is, in Chinese, tones can
spread either rightward or leftward depending on the dialect.

20
I follow Hashimoto’s notation for Cantonese phonemes. See Hashimoto (1972).
132 ACCENTUAL HISTORY

4.3.5 Summary
This chapter has discussed natural accent change not only in the
Japonic but also in other languages such as Korean, African
languages, Austronesian languages, and Chinese. The findings are
summarized as follows:

(36) Summary of accent changes

(a) Japonic accent changes: rightward direction
(b) Korean: starting from the end (… H# > … L#)
(c) African: either direction, depending on the language
(d) Austronesian: rightward direction
(e) Chinese: either direction

In the Japonic languages, regular sound change occurs from the

left to the right. However, in Korean the accent change starts from
the end of words. In African languages, tones shift either to the
right or to the left, depending on the language. In Austronesian
languages, stress shifts to the right. Finally, Chinese tones involved
in tone sandhi spread in either direction, depending on the dialects.
To conclude, as far as our discussions are concerned, strictly
speaking, changes involved in the development of accent differ
from language to language. On the other hand, from the viewpoint
of direction in change, a language tends to shift its accent in one
direction. For example, in Japanese, its accent shifts rightward, and
Korean shifts accent leftward.
Regarding reconstruction of proto accent forms of earlier Japonic
languages, with the findings of the patterns of the accent changes
and the comparative method, the accentual history of the Japanese
and Ryukyuan languages will be reconstructed in the following
chapters.
 5

RECONSTRUCTION OF
PROTO-RYUKYUAN ACCENT
5.1 INTRODUCTION
hroughout this work, my reconstruction of accent proto-forms
T has dealt only with nouns. Neither verbs nor adjectives are
included, because the accent patterns of Japonic nouns are
much more complex than those of verbs or adjectives and also
because the noun accent types reflect all types of verb and adjective
accent.
By excluding compounds and loanwords, I have carefully
selected Japonic cognates. However, some cultural words, such as
the one for ‘chopsticks’, are included because the accentuation of
those words parallels other words in the same class. Furthermore,
the phonology of these cultural words that I have included does not
seem to match that of neighboring languages, from which Japonic
languages might have borrowed, e.g., Chinese or Austronesian
languages. Therefore, I assume that these words might have existed
at the time of Proto-Japonic.
In some cases, it is difficult to see patterns in correspondences
between Ryukyuan nouns and their cognates in Myôgishô because
of missing cognates from the correspondence sets. In those cases,
to capture the patterns, I have added more cognates, which are not
on my original word list (see (1) below).
This chapter gives accentual descriptions of the Ryukyuan
dialects, and attempts to reconstruct Proto-Amami, Proto-Okinawa,
Proto-Miyako, and Proto-Yaeyama accent. In this chapter I treat
Yonaguni dialect as one independent dialect group because of its
134 ACCENTUAL HISTORY

peculiar characteristics (see Hirayama et al. (1964) for details).

However, Proto-Yonaguni is not reconstructed because only Sonai
data are available; no data on other Yonaguni dialects are available
in the literature. Therefore, the section of YONAGUNI
RYUKYUAN is structured differently.

(1) Japonic cognates used for my reconstruction

Myô List of cognates
1.1 hair, blood
1.2 leaf, name
1.3 three, eye, tooth
2.1 nose, cow, bird, loins, wind, beard
2.2 stone, paper, bridge
2.3 flower, mountain, cloud, bone, dog
2.4 boat, sea, shoulder, breath, board, mortar,
chopsticks, needle
2.5 sweat, rain, bridegroom, voice, shadow, bucket
3.1 mark, smoke, forehead, yawn
3.2 centipede, two
3.3 strength, gold
3.4 bag, fan, treasure, mirror
3.5 pillar, fat, life, tears, pillow, heart, cousin
3.6 eel, earthworm
3.7 medicine, whale, field

5.2 AMAMI RYUKUAN

5.2.1 Description of Amami Ryukyuan Accent
Amami Ryukyuan is made up of dialects spoken in the Amami
islands, including Amami Ô-shima, Tokunoshima, Kikai Island,
Okinoerabu Island, and Yoron Island. I make use of three Amami
dialects of Ryukyuan for my reconstruction: Naze dialect of
Amami Ô-shima, Kamishiro dialect of Okinoerabu-Island, and
Kametsu dialect of Tokunoshima. These dialects have been
carefully chosen in order to cover the characteristics of Amami
accent.
The main sources for the Amami dialects are as follows:
Hirayama et al. (1966 and 1967) for the Naze dialect, Hirayama et
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 135

al. (1966) and Kyûgakkai Rengô Amami Ô-shima Kyôdôchôsa

Iinkai, ed. (1959) for the Kamishiro dialect, and Hirayama et al.
(1966) and Hattori (1979a and 1979b) for the Kametsu dialect.

5.2.1.1 Naze Dialect

There are three nominal accent classes in the Naze dialect, nouns
with atonic, penultimate accent, and final accent. No register
distinction is involved. Atonic nouns have high pitch on the final
mora of a ‘prosodic word’, e.g., hana LH ‘nose’, or hana-ga LL-H
‘nose (Nom.)’. For monomoraic atonic nouns, they are high in
pitch, but L(H) when uttered with a particle, e.g., ci H ‘blood’, ci-
ga L-H ‘blood (Nom.).
Penultimate-accent nouns have high pitch on the penultimate
mora and the rest of each word is low in pitch, e.g., ha˘ri HL
‘needle’, ha˘ri-ga HL-L ‘needle (Nom.)’. Notice that the pitch
shape is not affected by the addition of a particle. That is, the
domain of the accent seems to be a lexical word.
For final-accent nouns, they keep a high pitched mora word-
finally, and the rest of the moras on the words are low in pitch, e.g.,
/unagi˘ LLH ‘eel’, /unagi˘-ga LLH-L ‘eel (Nom.)’. The high pitch
stays on the final mora of a lexical word. The domain of locus
accent is a lexical word, while that of atonic accent is a prosodic
word.
With regard to TBU, the Naze TBU is the mora, because the
pitch changes between moras within the same syllable, e.g., kagan
LLH ‘mirror’. If the syllable were the TBU, the pitch would change
at the syllable boundary.
The details of the accent system of Naze monosyllables,
disyllables, and trisyllables will be discussed below.
Monosyllables: Unlike many other Ryukyuan dialects, Naze
monosyllables have no vowel length. The data show that the
monosyllables are basically CV-structure, except for the word /in
‘dog’.
Naze monosyllabic nouns do not have distinctive accent; all the
nouns are atonic. They are generally high-level in isolation, but
136 ACCENTUAL HISTORY

low-pitch in a phrase with a particle, i.e., L(H) -- low on the stem

and high on the particle. Furthermore, when a noun contains two
moras, namely /in ‘dog’, the noun is LH. However, it is low-level
when followed by an enclitic, e.g., /in LH, /in-nu LL-H.

(2) Naze monosyllabic accent

Phonemic Phonetic Myô List of nouns
O H ~ L(H) 1.1 khï ‘hair’, ci ‘blood’
1.2 ha ‘leaf’, na ‘name’
1.3 khï ‘tree’, mï ‘eye’, ha ‘tooth’
LH ~ LL(H) 2.3 /in ‘dog’

In the data, except for /in ‘dog’, all Naze synchronically

monosyllabic nouns are also historically monosyllabic. With regard
to the word /in ‘dog’, it is historically disyllabic, but has become
monosyllabic after losing the final vowel /u/. Based on many
Japonic cognates (e.g., inu in Tokyo and Kyoto), the earlier Japonic
form for ‘dog’ should be something like *inu with the vowel /u/ at
the end.
Disyllables: In (3) the synchronic analysis of Naze disyllabic
nouns shows that there are three accentual types: atonic,
penultimate accent, and final accent. In disyllabic atonic nouns,
there are two types. One type consists of two moras, and the other
contains three moras. The former type is historically disyllabic, but
the latter is historically trisyllabic. Nouns belonging to the latter
type, three-mora disyllabic, have either a long vowel or a syllable-
final moraic nasal.
The two-mora atonic nouns correspond to Myôgishô accent
classes 2.1-5, not restricted to certain types of Myôgishô accent
class or classes. The three-mora atonic corresponds to both 3.2 and
3.4 in Myôgishô; however there is no systematic correspondence.
In disyllabic penultimate-accent nouns there are also two-mora
nouns and a three-mora noun. The three-mora noun, i.e., /o˘:gi ~
/u˘gi ‘fan’, is historically trisyllabic, cf. afuki LLL ‘fan’ in
Myôgishô (Mochizuki 1974: 31).
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 137

(3) Naze disyllabic accent

Phonemic Phonetic Myô List of nouns
OO LH ~ 2.1 hana ‘nose’, /usi ‘cow’, thuri
LL(H) ‘bird’, khusi ‘loins’, higi
‘beard’, hazye ‘wind’
2.2 /isi ‘stone’, khabi ‘paper’, hasi
‘bridge’
2.3 hana ‘flower’, yama
‘mountain’, kumu ‘cloud’
2.4 /ita ‘board’, mugi ‘barley’,
khata ‘shoulder’
2.5 /asi ‘sweat’, /amï ‘rain’
LLH ~ 3.2 ta:ci ‘two’
LLL(H) 3.4 kagan ‘mirror’
O˘O HL(L) 2.3 hu˘nï ‘bone’
2.4 hu˘nï ‘boat’, /u˘mi ‘sea’, ha˘ri
‘needle’, /u˘si ‘mortar’, ha˘si
‘chopsticks’, /i˘ki ‘breath’
2.5 kha˘gë ‘shadow’, khu˘i ‘voice’,
mo˘ho ‘bridegroom’, u˘hï
‘bucket’
HLL(L) 3.4 /o:˘gi ~ /u˘gi ‘fan’
ooO˘ LLH(L) 3.6 më:za˘ ‘earthworm’

The word khu˘i ‘voice’ (2.5) has a CVV structure. I treat it as

disyllabic because there is no need to treat it as monosyllabic and
also because historically it is attested as disyllabic, i.e., kowe
LH(L)1 ‘voice’ in Myôgishô (Mochizuki 1974: 226). Moreover, in
Naze, historical monosyllabic nouns generally have CV structure.2

1
It is LH in Mochizuki (1974: 226). That is, it is not clear whether or not the word
for ‘voice’ has accent on the final syllable. However, in Martin (1987) the word is
LH(L). Following Martin, I treat the word as final-accent.
2
In Naze, historical disyllables also appear as CV, e.g., nu ‘wheat’, /o ‘millet’,
/ma ‘horse’, /yu ‘fish’, hi ‘breast’, /o ‘foam’ (Hirayama et al. 1967: 194-387). In
addition, they are also in CV: and CVn derived from historical disyllables, e.g.,
138 ACCENTUAL HISTORY

Naze two-mora penultimate-accent nouns correspond to

Myôgishô accent classes 2.3-5. The word for ‘fan’, corresponding
to Myôgishô accent class 3.4, has two alternative forms; one with
three moras (i.e., /o:˘gi) and the other with two (i.e., /u˘gi). From
the data, it is obvious that the two-mora form is regular and the
three-mora form is irregular. Probably the three-mora form is a
borrowing from a Japanese dialect because of the vowel /o:/, on the
basis of the fact that the Japanese mid vowels (i.e., /o/ and /e/) in
general correspond to high vowels (i.e., /u/ and /i/) in Naze (e.g.,
ko˘e ‘voice’ in Tokyo, but ku˘i ‘voice’ in Naze).
The third accentual type is represented by nouns with final
accent. There is only one noun for the class, and the word is
originally trisyllabic, cf. mimizu LHH ‘earthworm’ in Myôgishô
class 3.6 (Mochizuki 1974: 521). It should be noted that më:za˘
‘earthworm’ behaves like trisyllabic nouns (see trisyllabic accent).
Trisyllables: There are three patterns of accentuation in Naze
trisyllabic nouns: atonic, penultimate accent, and final accent.
Atonic nouns are final-high in a phrase; i.e., LLH or LLL(H).
The data in (4) show that words in this accent class correspond to
Myôgishô classes 3.1 and 3.3-5. However, an investigation of a
larger data set reveals that Naze atonic trisyllables correspond to
Myôgishô classes 3.1-7; sakura ‘cherry’ (3.2), /azïki ‘red
beans’(3.2), hadasi ‘barefoot’ (3.6), /usiru ‘behind’ (3.7), kaigo
‘silkworm’ (3.7), mabuta ‘eyelid’ (3.7).3
Regarding the penultimate-accent nouns, they are normally
LHL(L), but HLL(L) only when the second syllable contains the
high-front vowel (i.e., /i/). Instead of creating another accent class,
Hirayama, et al. (1966: 124) treat the words hasi˘ra and kuzi˘ra as
penultimate accent because their pitch shape is predictable. The
data below show that Naze nouns with penultimate-accent

mun ‘thing’, na˘: ‘inside’, de: ‘bamboo’, min ‘ear’ (Hirayama et al. 1967: 196-
387).
3
The cognates for ‘cherry’, ‘red beans’, ‘eyelid’, ‘barefoot’, ‘behind’, and
‘silkworm’ are taken from Hirayama, ed. Gendai Nihongo hôgen dai-jiten and the
one for ‘left’ is from Hirayama, et al. (1966).
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 139

correspond to Myôgishô cognates in accent classes 3.2, 3.4-5, and

3.7. Looking into more cognates, I have found that there is a noun
corresponding to Myôgishô class 3.6 as well: /usa˘gi ‘rabbit’ (3.6)
(Hirayama, et al. 1966: 124). There is also another 3.5-noun with
penultimate-syllable accent: /inu˘cï ‘life’ (Hirayama, ed. 1982:
374). Therefore, Naze trisyllables in this class correspond to
Myôgishô 3.2 and 3.4-7 nouns. In addition, because most of Naze
nouns in this class correspond to low-register in Myôgishô, except
for muka˘zi ‘centipede’ (3.2), the accentuation of ‘centipede’ might
be irregular.

(4) Naze trisyllabic accent

Phonemic Phonetic Myô List of nouns
OOO LLH ~ 3.1 kibusi ‘smoke’, /akubi ‘yawn’
LLL(H) 3.3 kugani ‘gold’, cikyara ‘strength’
3.4 takara ‘treasure’
3.5 /abura ‘fat’, inuci ‘life’
OO˘O LHL(L) 3.2 muka˘zi ‘centipede’
3.4 huku˘ro ‘bag’
3.7 kusu˘ri ‘medicine’
HLL(L) 3.5 hasi˘ra ~ ha˘rya ~ ha˘ra ‘pillar’
3.7 kuzi˘ra ‘whale’
OOO˘ LLH(L) 3.6 /unagi˘ ‘eel’

The word for ‘pillar’ (3.5) has three alternative forms, i.e.,
ha˘sira ~ ha˘rya ~ ha˘ra. Since Hirayama et al. list those forms
without any discussion, it is not known under what circumstances
the alternative forms are uttered -- e.g., casual speech vs. careful
speech -- and also whether some of their informants or all of the
informants alternate the forms.
The pitch shapes of Naze final accent nouns are LLH(L). In (4),
there is only one noun in this class: /unagi˘ ‘eel’; it corresponds
only to Myôgishô accent class 3.6.
140 ACCENTUAL HISTORY

5.2.1.2 Kamishiro Dialect

Considering the system as a whole, the accent system of the
Kamishiro dialect is one of the most complex among the Ryukyuan
dialects because it utilizes two typologically different distinctions
(i.e., locus accent and register).
In Kamishiro, there are three patterns of accentuation: high-
register atonic, low-register atonic, and tonic. High-register nouns
are phonologically high-level, but their initial syllable is
phonetically in low pitch. This is due to a phonetic rule which
lowers the pitch of an unaccented initial syllable.
Low-register atonic nouns start in low and end in high pitch.
When uttered without a particle, the final mora of the prosodic
word rises in pitch. However, it is high when the prosodic word
ends in a particle, e.g., _hi: LR ‘tree’, _hi:-ga LL-H ‘tree (Nom.)’.
This indicates that the TBU of the Kamishiro accent is the mora. It
also indicates that the prosodic word is the domain of accent.
The third type of Kamishiro accentual pattern is tonic; it has an
accent on the second syllable. The accentuation of this type is only
seen in disyllables and trisyllables. Tonic disyllables are LF in
isolation, but they are LH(L) when followed by a particle. For
trisyllables, they are LHL(L).
Monosyllables: There are two distinctive accent classes for
monosyllabic nouns and both of them are atonic -- there is no pitch
fall, but nouns are distinguished by register: high-register and low-
register. The high-register nouns are in high pitch throughout a
phrase. The low-register nouns are LR in their citation forms, but
they become LL(H) in a phrase with a particle.4
By contrast with Naze monosyllabic nouns, all Kamishiro
monosyllabic nouns are two moras long with vowel length.

4
According to Tokugawa (1958), Kamishiro monosyllables are in either CV: or
CV. The former type is HH(H) pitch and the latter is R pitch (i.e., pitch rises
within a syllable). Nouns with rising pitch are a little shorter than two-mora nouns,
but a little longer than monomoraic nouns. He distinguishes them using CV> in
contrast with CV: and CV in length. Furthermore, in Tokugawa’s analysis, two-
mora nouns become monomoraic when followed by a particle. Thus, for example,
hi: HH ‘hair’ is hi-nu H-H, and hi> R ‘tree’ is hi-nu L-H.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 141

The Kamishiro monosyllabic data show regular correspondences

between the Kamishiro nouns and certain Myôgishô accent classes.
More specifically, Kamishiro high-register monosyllabic nouns
correspond to Myôgishô accent classes 1.1-2, and low-register
nouns to Myôgishô class 1.3.

(5) Kamishiro monosyllabic accent5

Phonemic Phonetic Myô List of nouns
`oo HH(H) 1.1 `hi: ‘hair’, `ci: ‘blood’
1.2 `ha: ‘leaf’, `na: ‘name’
_oo LR6 ~ 1.3 _hi: ‘tree’, _mi: ‘eye’, _ha:
LL(H) ‘tooth’

Disyllables: Kamishiro disyllabic nouns have three accent

classes. Two of them are atonic: high-register and low-register, and
the other one is final accent.
Kamishiro has a phonetic rule for pitch shape. This rule lowers
the pitch of the initial unaccented syllable if the syllable consists of
one mora, e.g., `hana LH(H) ‘nose’ vs. `ta:ci HHH(H) ‘two’.
Therefore, most of the high-register atonic nouns are phonetically
LH(H); the word `ta:ci ‘two’ is HHH(H). The word for ‘two’ is
historically trisyllabic. As a result of losing one syllable, it became
disyllabic; cf. futatu LHH(L) ‘two’ in Tokyo (Nihon Hôsô Bunka
Kenkyûjo, ed. 1971), futatu HHL/HLL ‘two’ in Myôgishô
(Mochizuki 1974: 475). On the other hand, all Kamishiro low-
register disyllabic nouns are LR in isolation, but LL(H) when
pronounced with a particle.

5
The word for ‘blood’ is extracted from Kyûgakkai Rengô Amami Ô-shima
Kyôdôchôsa Iinkai, ed. (1957: 411).
6
This pitch shape is exactly the same as that of disyllabic low-atonic nouns, where
the pitch rises within the second syllable. However, from Hirayama et al. (1966),
for CV: nouns, it is not clear whether the pitch rises at the end of the words or the
pitch gradually rises within a word from the beginning of the words.
142 ACCENTUAL HISTORY

(6) Kamishiro disyllabic accent7

Phonemic Phonetic Myô List of nouns
`OO LH(H) 2.1 `hana ‘nose’, `/usi ‘cow’,
`thui‘bird’, `hazi ‘wind’
2.2 `hasi ‘bridge’, `/isi ‘stone’
2.4 `/uni ‘sea’, `/iki ‘breath’
HHH(H) 3.2 `ta:ci ‘two’
_OO LR ~ 2.3 _hana ‘flower’, _yama
LL(H) ‘mountain’
2.4 [_kata] ‘shoulder’, [_/ita]
‘board’, [_ha:ri] ‘needle’
2.5 _/ami ‘rain’
OO˘ LF ~ 2.4 hi:ni˘ ‘boat’, /usi˘ ‘mortar’, hasi˘
LH(L)/ ‘chopsticks’
LLH(L) 2.5 haga˘ ‘shadow’
HL(L)/ 2.5 hui˘ ‘voice’, wui˘ ‘bucket’,
HHL(L) muQkwa˘ ‘bridegroom’
LLH(L) 3.4 /o:gi˘ ‘fan’

Synchronically all historically-disyllabic atonic nouns begin

with low pitch, yet the difference between high-register and low-
register nouns is that in the former the pitch rises after the first
syllable, and in the latter the pitch rises at the end of a phrase.
As stated in 5.2.1, nouns with CVV structure such as `thui ‘bird’
are treated as disyllabic in this book. The reasons are as follows: (i)
there is no reason why they have to be treated as monosyllabic, (ii)
they are historically disyllabic, and (iii) Ryukyuan monosyllabic
nouns are normally CV: structure.
Low-register atonic disyllabic nouns are phonetically LR in their
citation forms, but LL(H) when followed by a particle. Most of the

7
Words for ‘bird’, ‘boat’, ‘sea’, ‘breath’, and ‘rain’ are taken from Kyûgakkai
Rengô Amami Ô-shima Kyôdôchôsa Iinkai, ed. (1957: 410-422). Words in square
brackets are written in kanji (Chinese characters) in the original sources (see
Hirayama et al. 1966: 159), but the discussion makes clear the vowel-length
distinctions.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 143

nouns are two-mora, but the word for ‘needle’ (2.4) has three moras,
_ha:ri. The pitch shape of this word is not certain because
Hirayama does not discuss the pitch shape of three-mora nouns in
this class. However, based on Kamishiro (three-mora) trisyllabic
low-register nouns, the word for ‘needle’ is probably LLR ~
LLL(H).
There are three allophonic subtypes of tonic-accent nouns. The
first type is phonetically LF in isolation, but LH(L) when followed
by a particle. Like other ordinary disyllables, this type of noun has
(C)V(C)V structure, with one exception; hi:ni˘ ‘boat’. This word
has a pitch shape of LLF ~ LLH(L).
The second type is HL(L) in the surface representation. The
pitch shape of the words with CVV or /Q/ in the initial syllable is
predictable. Words with CVV structure are HL(L), and words with
CVQCV are HHL(L). Regarding the word for ‘bridegroom’ (2.5),
according to Serafim (1994: 1),8 there are two possible explanations
for the development of muQkwa˘. First, the form might have
resulted from an earlier form *mo:kura because in Ryukyuan the
morpheme Qkwa is derived from *kura or *ku:ra, cf. *kura > Qkwa
HL(L) ‘child’ in Shuri. In addition, the word for ‘bridegroom’
might have had an initial-syllable long vowel in an earlier form. A
number of Ryukyuan dialects show the initial-syllable vowel length
in their cognates. For example, in Shuri and Onna dialects of
Okinawa Ryukyuan, their cognate forms show length. Vowel
length will be discussed in 5.3.3. The other explanation is that
muQkwa˘ has been contaminated by the word makura ‘pillow’
which resulted in maQkwa or the like in Ryukyuan. For example,
the word maQkwa in Shuri is historically trisyllabic, cf. makura
LLH ‘pillow’ in Myôgishô (Mochizuki 1974: 497). In its history,
the word lost a moraic segment, and in order to compensate for the
loss of the segment, the gemination of /k/ took place.

8
More specifically, it is according to both Serafim (1994) and discussions with
Leon A. Serafim during my writing of this book. For convenience, I cite only
Serafim (1994).
144 ACCENTUAL HISTORY

The third type of tonic-accent disyllabic noun in Kamishiro is

phonetically LLH(L), and for this type there is only one example,
e.g., /o:gi˘ ‘fan’. This is historically a trisyllabic noun. It is listed
as afuki LLL ‘fan’ in Myôgishô (Mochizuki 1974: 31).
The accentuation of Kamishiro words for `/iki ‘breath’ (2.4) and
`/uni ‘sea’ (2.4) must be irregular because 2.4 nouns are usually
either low-register or final-accent in Kamishiro. Notice that
Kamishiro high-register nouns correspond to Myôgishô class 2.1 or
2.2 and ‘breath’ and ‘sea’ are the only high-register nouns
belonging to Myôgishô accent class 2.4.
Trisyllables: The Kamishiro trisyllabic nouns have a three-way
distinction in their accent system. The distinctions are high-register
atonic, low-register atonic, and tonic. They are phonetically
realized as follows: LHH(H), LLR ~ LLL(H), and LHL(L)
respectively. Notice that all Kamishiro trisyllabic nouns begin with
low pitch because of a phonetic rule that lowers the pitch of an
unaccented initial syllable. While the pitch rises after the first
syllable in high-register atonic nouns, in low-register atonic nouns
the pitch rises at the end of the phrase.

(7) Kamishiro trisyllabic nouns

Phonemic Phonetic Myô List of nouns
`OOO LHH(H) 3.1 `hibusi ‘smoke’, `/akubi
‘yawn’, `sirusi ‘mark’
_OOO LLR ~ 3.4 _hagani ‘mirror’, _takara
LLL(H) ‘treasure’
OO˘O LHL(L) 3.4 huku˘ru ‘bag’
3.7 guzi˘ya ‘whale’

As far as the data in (7) are concerned, high-register trisyllabic

nouns correspond to Myôgishô accent class 3.1, and the low-
register trisyllables to Myôgishô class 3.4. tonic-accent nouns are
either class 3.4 or 3.7 in Myôgishô. The data in (7) lack nouns
corresponding to Myôgishô accent classes 3.2, 3.3, 3.5, and 3.6.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 145

5.2.1.3 Kametsu Dialect9

In the Kametsu dialect, there are three types of nominal accent
classes: atonic accent, penultimate accent, and final accent. Atonic
nouns are phonetically low-pitched on the initial syllable;
monosyllables are high-level.
For penultimate-accent nouns, the penultimate syllable is high in
pitch and the rest are low-pitched, e.g., muka˘re LHL ‘centipede’,
muka˘re-ga LHL-L ‘centipede (Nom.)’. When the penultimate
syllable consists of two moras, the pitch falls between the moras,
e.g., gun˘za HLL ‘whale’. In addition, there are no penultimate-
accent monosyllables.
In final-accent nouns, the pitch falls within a final mora, and the
rest of moras in words are low in pitch, when uttered without a
particle. However, they are final-mora-high when followed by a
particle, e.g., yama LF ‘mountain’, yama-ga LH-L ‘mountain
(Nom.)’. This is very similar to Kyoto (see 6.2.2).
Monosyllables: Hirayama et al. (1966: 149) state that Kametsu
monosyllabic nouns are generally two moras long with a long
vowel, but they are occasionally pronounced as monomoraic words.
For example, ha:˘ ‘tooth’ is occasionally pronounced as ha˘ F, ha˘-
nu H-L.
In Kametsu, there are atonic and final-accent monosyllabic
nouns. Atonic nouns are phonetically high-level, but the phonetic
pitch shape of the tonic nouns is rather complex, with two types.
One is LF in isolation, but HH(L) when used in a phrase. The other
type is HL(L) both in isolation and in a phrase.
On the basis of the data below, the pitch shape of the HL-nouns
is predictable. In other words, the pitch falls after the initial mora
when the second mora is a nasal, i.e., /in˘ HL(L) ‘dog’ and /un˘
HL(L) ‘sea’. Furthermore, these nouns with a moraic nasal are
historically disyllabic; cf. inu LL ‘dog’, umi LH ‘sea’ in Myôgishô
(Mochizuki 1974: 67 and 94).

9
Most of the nouns listed here are taken from Hirayama et al. (1966: 149-50),
unless other sources are indicated.
146 ACCENTUAL HISTORY

(8) Kametsu monosyllabic accent10

Phonemic Phonetic Myô List of nouns
oo HH(H) 1.1 kï: ‘hair’, ci: ‘blood’
1.2 ha: ‘leaf’, na: LH ‘name’
oo˘ LF ~ HH(L) 1.3 kï:˘ ‘tree’, mï:˘ ‘eye’, ha:˘ ‘tooth’
HL(L) 2.3 /in˘ ‘dog’
2.4 /un˘ ‘sea’

The data show that there are regular correspondences between

Kametsu monosyllabic accent and Myôgishô accent; Kametsu
monosyllabic atonic nouns correspond to Myôgishô classes 1.1-2,
and tonic nouns to Myôgishô class 1.3.
Disyllables: For disyllabic nouns, there are three types of
accentuation in Kametsu. They are atonic, penultimate accent, and
final-falling accent, and are phonetically realized as LH(H), HL(L),
and LF ~ LH(L) respectively.

(9) Kametsu disyllabic accent11

Phonemic Phonetic Myô List of nouns
OO LH(H) 2.1 hana ‘nose’, /usi12 ‘cow’, thuI
‘bird’, khusI ‘loins’, sïgi ‘beard’,
khadI ‘wind’
2.2 /IsI ‘stone’, kabi ‘paper’, hasi13
‘bridge’
3.2 ta:ci14 ‘two’

10
The words for ‘name’, ‘eye’, ‘dog’, and ‘sea’ are extracted from Hirayma (1986:
532, 712, 81, and 109 respecitvely). The pitch of the word na: ‘name’ is described
as LH. Cf. the word for ‘eye’ is mI˘ ‘eye’ in Kyûgakkai Rengô Amami Ô-shima
Kyôdôchôsa Iinkai, ed. (1959: 437).
11
Kametsu disyllabic nouns such as thuI ‘bird’, khusI ‘loins’, and khadI ‘wind’ are
extracted from Hattori (1979b: 101-104), and /I˘khI ‘breath’, /amˆ˘ ‘rain’, and
/asI˘ ‘sweat’ are from Hattori (1979a: 102-5).
12
/usï ‘cow’ in Hirayama et al. (1966: 449).
13
hasï ‘bridge’ in Hirayama et al. (1966: 449).
14
Cited from Kyûgakkai Rengô Amami Ô-shima Kyôdôchôsa Iinkai, ed. (1959:
437).
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 147

O˘O HL(L) 2.3 hu˘nI bone’

2.4 hu˘nï ‘boat’, ha˘i ‘needle’, /u˘sï
‘mortar’, /I˘khI ‘breath’
2.5 ku˘i ‘voice’, ka˘gï ‘shadow’
3.4 hu˘Qku ‘bag’, o˘gi ‘fan’
3.5 ha˘ra ‘pillar’
3.6 mï˘da ‘earthworm’
3.7 gun˘za ‘whale’
OO˘ LF ~ 2.3 hana˘ ‘flower’, yama˘
LH(L) ‘mountain’, kumo˘ ‘cloud’
2.4 khata˘ ‘shoulder’
2.5 /amˆ˘ ‘rain’, /asI˘ ‘sweat’,15
/ukï˘16 ‘bucket’
3.5 /amba˘ ‘fat’

Most of the Kametsu disyllabic atonic nouns correspond to

Myôgishô classes 2.1-2, except for ta:ci ‘two’ (3.2). Kametsu
disyllabic tonic nouns generally correspond to Myôgishô 2.3-5
accentuation, with the exception of historically-trisyllabic nouns.
Moreover, some of Myôgishô 2.3-5 nouns correspond to Kametsu
penultimate accent and others to final accent.
Trisyllables: Kametsu trisyllabic nouns have a three-way accent
distinction: atonic, penultimate accent, and final accent, and they
are phonetically LHH(H), LHL(L), and LLF ~ LLH(L) respectively.
The structure of the trisyllabic nouns is (C)VCVCV.
Based on my original word list, I have analyzed Kametsu
trisyllabic accent as in (10a), which lacks nouns corresponding to
Myôgishô classes 3.2-3, 3.5, and 3.7. From this, it is not clear what
type of accentuation those cognates have in Kametsu. In order to
know what type of accentuation those missing cognates have and
also to know whether or not there are even more accent distinctions
for Kametsu trisyllables, I have expanded my word list by adding
words from Hirayama (1986: 83-658), except for the ones for

15
The words for ‘rain’ and ‘sweat’ are extracted from Hattori (1979a).
16
This is from Hirayama (1986: 128).
148 ACCENTUAL HISTORY

‘rabbit’ and ‘centipede’, which are extracted from Hirayama (1966:

150) and Uemura Kouji (1959: 11) respetively. The result of this
investigation is shown in (10b).

(10) Kametsu trisyllabic accent

(a) Analysis of Kametsu trisyllables limited to my original word list
Phonemic Phonetic Myô List of nouns
OOO LHH(H) 3.1 sirusi ‘mark’, khïbusi ‘smoke’,
/akubi ‘yawn’
OO˘O LHL(L) 3.6 /una˘gi ‘eel’
OOO˘ LLF ~ 3.4 kagami˘ ‘mirror’
LLH(L)

(b) Analysis of Kametsu trisyllables with an expanded word list

Phonemic Phonetic Myô List of nouns
OOO LHH(H) 3.1 sirusi ‘mark’, khïbusi ‘smoke’,
/akubi ‘yawn’
3.2 sakura ‘cherry’
3.3 cïkyara ‘strength’
3.6 hadaka ‘naked’
OO˘O LHL(L) 3.2 muka˘re ‘centipede’
3.4 kata˘na ‘sword’
3.5 huta˘ru ‘firefly’, sïga˘ta
‘figure’
3.6 /una˘gi ‘eel’, hada˘sï
‘barefoot’, sïza˘i ‘left’, nizï˘mi
‘mouse’, /usa˘gi ‘rabbit’
3.7 kusu˘i ‘drug’, cuba˘ki
‘camellia’
OOO˘ LLF ~ 3.4 kagami˘ ‘mirror’, hasami˘
LLH(L) ‘scissors’, sïkari˘ ‘light’
3.5 /inucï˘17 ‘life’

17
There is another form for ‘life’ in the same source; /inoci˘ (Hirayama 1986:
832).
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 149

The data in (10b) reveal that Kametsu trisyllables with atonic

accent correspond to Myôgishô 3.1-3 and 3.6, ones with
penultimate accent to Myôgishô 3.2 and 3.4-7, and ones with final
accent to Myôgishô 3.4-5. This is rather systematic except for two
nouns with possible irregular accentuation: hadaka ‘naked’ (3.6)
and muka˘re ‘centipede’ (3.2). If they turn out to be irregular, there
is an accent system that Kametsu atonic and penultimate accent
nouns respectively correspond to Myôgishô 3.1-3 (high-register)
and Myôgishô 3.4-7 (low-register). Because many Kametsu
historically-trisyllabic nouns have become disyllabic, I am not able
to find other 3.2- and 3.6-cognates in order to determine whether or
not the words in question are regular.

5.2.2 Correspondences and Reconstruction of Proto-Amami

Accent
Presenting correspondences among the three Amami dialects, in
this section, I am going to show my reconstructed accentual forms
for monosyllables, disyllables, and trisyllables.
Monosyllables: Based on the following two sets of
correspondences, in principle I am able to reconstruct two
accentual classes for Proto-Amami (PA) monosyllabic nouns. One
of the proto-accentual patterns is high-level atonic, and the other is
final accent. Proto-Amami monosyllables are two moras long. This
means that in the Naze dialect they evolved into monomoraic,
while the other two dialects have retained the length.
Except for the word */in˘ ‘dog’, reconstructed monosyllables
listed below are historically monosyllabic. The word for ‘dog’ is
historically disyllabic; however, it is monosyllabic in modern
Amami dialects. Thus, in principle, we have to reconstruct it as a
monosyllable for Proto-Amami. At the time of Proto-Amami, it had
already become monosyllabic.
150 ACCENTUAL HISTORY

(11) Proto-Amami accent for monosyllabic nouns

(a)
Naze Kamishiro Kametsu PA
O `oo Oo *oo
H ~ L(H) HH(H) HH(H) HH(H)
‘hair’ (1.1) khï `hi khï: *khï:
‘blood’ (1.1) ci `ci: ci: *ci:
‘leaf’ (1.2) ha `ha: ha: *ha:
‘name’ (1.2) na -- na: *na:

(b)
Naze Kamishiro Kametsu PA
O _oo oo˘ *oo˘
H ~ L(H)18 LR ~ LL(H) LF ~ HH(L)19 LH(L)
h
‘tree’ (1.3) kï _hi: khï:˘ *khï:˘
‘eye’ (1.3) mï _mi: mi:˘ *mï:˘
‘tooth’ (1.3) ha _ha: ha:˘ *ha:˘
‘dog’ (2.3) /in -- /in˘ */in˘

All Naze dialect forms have lost Proto-Amami vowel length.

The reason for proposing vowel length in Proto-Amami
monosyllables is mainly based on a phenomenon of long-vowel
shortening in Japonic languages. For example, Proto-Okinawa
*mu:˘ku ‘bridegroom’ became mu˘hu in Nakijin, cf. _mu:ku in
Shuri. Furthermore, this vowel length in the initial syllable is
deeply related to the accent system of Ryukyuan languages (see
5.3.3).
Disyllables: As shown below, there are three sets of accentual
correspondences. For convenience, I summarize the
correspondences of accent classes and reconstructed Proto-Amami
accent classes in (12); nouns that I have actually reconstructed are
listed in (13).

18
For ‘dog’, its pitch shape is LH ~ LL(H).
19
The pitch shape of the word for ‘dog’ is HL(L).
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 151

(12) Summary of accentual correspondences and reconstruction of

Proto-Amami accent for disyllables
Naze Kamishiro Kametsu PA
a OO `OO OO *OO
LH ~ LL(H) LH(H)20 LH(H) HH(H)
b OO _OO OO˘ *OO˘
LH ~ LL(H) LR ~ LL(H) LF ~ LH(L) LH(L)
c O˘O OO˘ O˘O *O˘O
HL(L) LF ~ LH(L)21 HL(L) HL(L)

On the basis of those, I propose three accent classes for Proto-

Amami disyllabic nouns: high-level atonic, oxytonic, and
penultimate accent. That is, the Proto-Amami disyllables
distinguish one from another by locus. Notice that the accent
system of Kametsu dialect is phonemically identical to that of
Proto-Amami.

(13) Proto-Amami accent for disyllabic nouns

(a)
Naze Kamishiro Kametsu PA
OO `OO OO *OO
LH ~ LL(H) LH(H) LH(H) HH(H)
‘nose’ (2.1) hana `hana hana *hana
‘cow’ (2.1) /usi `usi /usi */usi
‘bird’ (2.1) thuri `thui thuI *thuri
‘loins’ (2.1) khusi -- khusI *khusi
‘beard’ (2.1) higi -- sïgi *higi
‘wind’ (2.1) hazye `hazi khadI *khazï
‘stone’ (2.2) /isi `/isi /IsI */isi
‘paper’ (2.2) khabi -- khabi *khabi
‘bridge’ (2.2) hasi `hasi hasi *hasi
‘two’ (3.2) ta:ci `ta:ci ta:cI *ta:cï

20
The word for ‘two’ is HHH(H).
21
The pitch shape of ‘voice’, ‘bucket’, and ‘bridegroom’ are HL(L), and ‘fan’ has
LH(L) pitch.
152 ACCENTUAL HISTORY

(b)
Naze Kamishiro Kametsu PA
OO _OO OO˘ *OO˘
LH ~ LR ~ LL(H) LF ~ LH(L)
LL(H) LH(L)
‘flower’ (2.3) hana _hana hana˘ *hana˘
‘mountain’(2.3) yama _yama yama˘ *yama˘
‘cloud’ (2.3) kumu -- kumo˘ *khumu˘
‘shoulder’ (2.4) khata [_kata] khata ˘ *khata ˘
‘board’ (2.4) /ita [_/ita] -- */ita˘
‘rain’ (2.5) /amï _/ami /amï˘ */amï˘
‘sweat’ (2.5) /asi -- /asI˘ */asï˘
‘tears’ (3.5) nada -- -- *nada˘

(c)
Naze Kamishiro Kametsu PA
O˘O OO˘ O˘O *O˘O
HL(L) LF ~ LH(L) HL(L) HL(L)
‘bone’ (2.3) hu˘nï -- hu˘nI *hu˘nï
‘mortar’ (2.4) /u˘si usi˘ /u˘sï */u˘si
‘chopsticks’(2.4) ha˘si hasi˘ -- *ha˘si
‘boat’ (2.4) hu˘nï hi:ni˘ hu˘nï *hu:˘nï
‘needle’ (2.4) ha˘ri [ha:ri˘] ha˘i *ha:˘ri
‘breath’ (2.4) /i˘ki `/iki /I˘ki */i˘ki
‘sea’ (2.4) /u˘mi `/uni /u˘n */u˘mi
‘shadow’ (2.5) kha˘gë haga˘ ka˘gï *kha˘gï
‘voice’ (2.5) khu˘i hui˘ ku˘i *khu˘i
‘bridegroom’(2.5) mo˘ho muQkwa˘ (muQkwa)22 *mu˘kwa
‘bucket’ (2.5) u˘hï wui˘ ukï˘ *wu˘khï
‘fan’ (3.4) /o˘:gi o:gi˘ /o˘gi */o:˘gi
~/u˘gi

22
The accentuation of this word is not available (Hirayama 1966: 307).
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 153

In (13c), the Kamishiro words `/iki ‘breath’ and `/uni ‘sea’

have irregular accent. For explanations, see the discussion in
5.2.1.2 above. In the same correspondence set, the Kametsu
cognate for ‘bucket’ has irregular accentuation, as well. Kametsu
cognates in this correspondence set have penultimate accent.
With regard to nouns corresponding to Myôgishô accent class
2.3, almost all of 2.3-cognates are in (13b); one in (13c).23
Trisyllables: For Proto-Amami trisyllabic nouns, I have
reconstructed four accent classes on the basis of the following four
sets of correspondences, as shown in (14): */OOO/, */OOO˘/,
*/OOOˆ/, and */OO˘O/.

(14) Summary of accentual correspondences and reconstruction of

Proto-Amami accent for trisyllables
Naze Kamishiro Kametsu PA
a OOO `OOO OOO *OOO
LLH ~ LHH(H) LHH(H) HHH(H)
LLL(H)
b OOO _OOO OOO˘ *OOO˘
LLH ~ LLR ~ LLF ~ LLH(L)
LLL(H) LLL(H) LLH(L)
c OO˘O OO˘O OO˘O *OOOˆ
LHL(L) LHL(L) LHL(L) LHF ~
LHH(L)
d OOO˘ OO˘O *OO˘O
---
LLH(L) LHL(L) LHL(L)

Like Proto-Amami monosyllabic and disyllabic accent, Proto-

Amami trisyllabic accent has a locus system. Among the Amami
Ryukyuan dialects, the Kamishiro dialect is the only one that
utilizes both register and locus. Moreover, in Proto-Amami

23
The investigation that I conducted on the accentuation of twelve 2.3-cognates
shows a similar pattern: eleven 2.3-cognates belong to (13b), and one cognate for
‘flea’ (i.e., no˘mi in both Naze and Kametsu) has the same pattern as in (13c).
Cognates that I have investigated are ‘net’, ‘mud’, ‘grave’, ‘pigeon’, ‘color’,
‘tortoise’, ‘god’, ‘rice’, ‘weed’, ‘parent’, ‘moon’, and ‘flea’.
154 ACCENTUAL HISTORY

trisyllables, final accent (*/OOO˘/) is distinguished from final-

falling accent */OOOˆ/). The following correspondence sets
contain cognates not included in my original word list (see p134). I
have added these to see more general patterns in correspondences
in terms of Myôgishô accent classes. For example, without
additional data, I would not be able to discover that Naze and
Kametsu 3.2-cognates also show a regular pattern for the
correspondence in (15a); the 3.2-cognate for ‘centipede’ seems to
have irregular accentuation (see below). Additional data are shown
in boldface; Kametsu cognates are from Hirayama (1986), and
Naze cognates are extracted from Hirayama (1992).
Nouns in (15a) regularly correspond to Myôgishô 3.1-3 accent
classes, in which nouns start in high pitch. Cognates in (15b) are
3.4-5 nouns and they are all low-initial in Myôgishô. In (15c), there
are nouns corresponding to 3.2 (high-register) and 3.4-7 (low-
register). Many of nouns are 3.4-7 nouns, and 3.2-noun is only one,
i.e., ‘centipede’. The 3.2-noun must have irregular accentuation. In
(15d), nouns correspond only to Myôgishô accent class 3.6. For the
correspondence set in (15d), I have reconstructed */OO˘O/ soley on
the basis of Naze and Kametsu data; Kamishiro data are not
available.

(15) Proto-Amami accent for trisyllabic nouns

(a)
Naze Kamishiro Kametsu PA
OOO `OOO OOO *OOO
LLH ~ LHH(H) LHH(H) HHH(H)
LLL(H)
‘mark’ (3.1) sirusi `sirusi sirusi *sirusi
‘smoke’(3.1) khibusi `hibusi khïbusi *khïbusi
‘yawn’ (3.1) /akubi `akubi akubi */akubi
‘cherry’(3.2) sakura -- sakura *sakura
‘red beans’ (3.2) /azïki -- -- *azïki
‘strength’(3.3) cikyara -- cïkyara *cikyara
‘gold’ (3.3) khugani -- -- *khugani
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 155

(b)
Naze Kamishiro Kametsu PA
OOO _OOO OOO˘ *OOO˘
LLH ~ LLR ~ LLF ~ LLH(L)
LLL(H) LLL(H) LLH(L)
‘mirror’(3.4) khagan _hagani khagami˘ *khagami˘
‘treasure’(3.4) thakara _takara -- *thakara˘
‘light’ (3.4) hikyari -- sïkyaI˘ *hïkari˘
‘calendar’ (3.4) kuyumi -- -- *kuyumi˘
‘fat’ (3.5) /abura -- /amba˘ */abura˘
‘life’ (3.5) /inuci -- /Inucïv */inuci˘
‘pillow’ (3.5) makura -- maQkwa˘ *makura˘
‘cousin’ (3.5) /itoko -- -- */itoko˘

(c)
Naze Kamishiro Kametsu PA
OO˘O OO˘O OO˘O *OOOˆ
LHL(L) LHL(L) LHL(L) LHF ~
LHH(L)
‘centipede’ (3.2) muka˘zi -- muka˘re *mukaziˆ
‘bag’ (3.4) huku˘ro huku˘ru hu˘Qku *hukuruˆ
‘sword’(3.4) kata˘na -- kata˘na *katanaˆ
‘pillar’(3.5) hasi˘ra24 -- ha˘ra *hasiraˆ
‘figure’ (3.5) sigata -- siga˘ta *sïgataˆ
‘rabbit’ (3.6) usa˘gi -- usa˘gi *usagiˆ
‘medicine’ (3.7) khusu˘ri -- kusu˘I *khusuriˆ
‘whale’ (3.7) kuzi˘ra guzi˘ya gu˘nza *kuziraˆ

(d)
Naze Kamishiro Kametsu PA
OOO˘ -- OO˘O *OO˘O
LLH(L) LHL(L) LHL(L)
‘eel’ (3.6) unagi˘ -- /una˘gi */una˘gi
‘earthworm’(3.6) më:za˘ -- mï˘da *më(C)ë˘za

24
This word has alternative forms: ha˘rya and ha˘ra (Hirayama et al. 1966: 125)
156 ACCENTUAL HISTORY

There are obvious disyllabic words on the list of

correspondences for trisyllables. Among them, the cognates listed
for ‘earthworm’ (15d) are both disyllabic. In Naze its cognate is
më:za˘ and in Kametsu mï˘da. The Naze form indicates that the
word used to be trisyllabic; the vowel /ë:/ must have come from a
sequence of two vowels, i.e., CV(C)V.25 Therefore, I have
reconstructed *më(C)ëza˘ ‘earthworm’. However, on the basis of
the data, the value of the consonant in the second syllable is not
known.
The vowel /o/ in the word huku˘ro ‘bag’ is irregular because a
vowel corresponding to the Japanese vowel /o/ is generally /u/ in
Naze. Therefore, I have reconstructed /u/ in the final syllable of the
Proto-Amami noun for ‘bag’.
Regarding the Kamishiro and Kametsu cognates for ‘whale’
(15c), their forms have a /g/ at the word-initial position, while the
Naze form has a /k/. In Ryukyuan, nouns related to animals tend to
begin with a voiced consonant, e.g., gani ‘crab’ (Serafim 1993: 1).
Ryukyuan dialects seem to have developed this characteristic; it is
not seen in the Japanese language.

5.2.3 Development of Accent from Proto-Amami to Modern

Amami Dialects
I have reconstructed the Proto-Amami accent system as shown
below.

(16) Proto-Amami accent

Monosyllables Disyllables Trisyllables
*oo [HH(H)] *OO [HH(H)] *OOO [HHH(H)]
*oo˘ [LH(L)] *OO˘ [LH(L)] *OOO˘ [LLH(L)]
*O˘O [HL(L)] *OOOˆ [LHF ~ LHH(L)]
*OO˘O [LHL(L)]

25
In the discussion of the vowel /ë/ in më:za˘ ‘earthworm’ with Leon A. Serafim,
he pointed out that the vowel /ë/ is derived from a sequence of two vowels.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 157

In this section I will explain how this system has evolved to the
accent systems of the three Amami Ryukyuan, i.e., Naze,
Kamishiro, and Kametsu. Typological accent change is not
involved; i.e., a change in register or locus is not seen in the
developments of Amami Ryukyuan.

5.2.3.1 Development of Naze Accent

Although the development of Naze accent is not complex, it has
turned out to be drastically different from Proto-Amami accent. For
instance, by undergoing Final Accent Loss and Vowel Shortening,
Naze monosyllabic nouns lost characteristics of Proto-Amami
accent; Proto-Amami accentual distinctions were lost and bimoraic
nouns became monomoraic. In the following, I explain the
development of Naze monosyllabic, disyllabic, and trisyllabic
accent.
Monosyllables: In the Naze dialect, with Final Accent Loss
Proto-Amami monosyllabic final-accent nouns lost their accent and
merged with PA */oo/ as LH(H). And then later the length of nouns
became short by Vowel Shortening (V-shrt) -- two mora nouns
became monomoraic.

(17) Development of Naze monosyllabic accent

PA F-loss V-shrt
(a) *oo
> oo [LH] > O [H ~ L(H)]
(b) *oo˘

Disyllables: The development from Proto-Amami accent to

modern Naze accent involves Final Accent Loss only. Because of
this sound change, PA */OO/ and */OO˘/ merged as /OO/ [LH(H)].
Proto-Amami initial-accent nouns have remained the same.
158 ACCENTUAL HISTORY

(18) Development of Naze disyllabic accent

PA F-loss
(a) *OO
> OO [LH ~ LL(H)]
(b) *OO˘

(c) *O˘O = O˘O

Trisyllables: The development of Naze trisyllabic accent from

Proto-Amami is simple. Three sound changes contributed to its
development. They are Final Accent Loss, Accent Shift, and
Contour Pitch Simplification (C-smpl).
By means of the first change, PA */OOO˘/ (19b) lost its accent.
This change caused the merger into an atonic accent from two
sources, namely PA */OOO/ (19a) and */OOO/ (< PA */OOO˘/)
(19b). The second change shifted the accent of PA */OO˘O/ (19d)
onto the final syllable. Finally, the Contour Pitch Simplification
rule converted PA */OOOˆ/ (19c) to /OO˘O/ (see 4.1.1).

(19) Development of Naze trisyllabic accent

PA F-loss shift C-smpl

(a) *OOO
> OOO = OOO = OOO
(b) *OOO˘ [LLH(H)]
[LLH(L)]

(c) *OOOˆ = OOOˆ = OOOˆ > OO˘O

[LHL(L)]

(d) *OO˘O = OO˘O > OOO˘ = OOO˘

[LLH(L)]

5.2.3.2 Development of Kamishiro Accent

Among the Amami Ryukyuan dialects introduced in this book, the
development of Kamishiro accent is the most complex. However,
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 159

Kamishiro has as many accentual distinctions as Proto-Amami has.

The following shows how Kamishiro nouns have developed their
accent system.
Monosyllables: Compared with its disyllables and trisyllables,
the development of Kamishiro monosyllabic accent is rather simple.
It involves only one sound change, namely Final Accent Loss, and
it is followed by a phonetic change (L-sprd). Final Accent Loss
turned PA */oo˘/ [LH(L)] into /_oo/ [LH(H)], whose pitch shape
eventually became LR ~ LL(H). This is because the L-pitch of the
initial syllable influenced the pitch of the following syllable, i.e.,
LH(H) > LR ~ LL(H).

(20) Development of Kamishiro monosyllabic accent

PA F-loss L-sprd
(a) *oo ª `oo = `oo
(b) *oo˘ > _oo > _oo
[LH(L)] [LH(H)] [LR ~ LL(H)]

Disyllables: In the development of Kamishiro disyllabic accent

from Proto-Amami, there are two sound changes and three phonetic
changes. The first change is Final Accent Loss, which converted
PA */OO˘/ [LH(L)] (21b) to /_OO/ [LH(H)]. Along with this
change, PA */OO/ (21a) became /`OO/. As shown below, register
distinctions appreared at this point.
The other sound change is Accent Shift, which took place at the
second stage. It shifted the accent of PA */O˘O/ (21c) onto the
second syllable, resulting in /OO˘/ [HH(L)].
The last three changes are phonetic. Accent Shift is followed by
Low Pitch Spreading, which turned the pitch shape of /_OO/
[LH(H)] into LR ~ LL(H). Next, assimilation made /OO˘/ [HH(L)]
(21c) into /OO˘/ [HF ~ HH(L)]. And then at the final stage, it
became LF ~ LH(L). /`OO/ also underwent Initial Pitch Lowering
(I-low).
160 ACCENTUAL HISTORY

(21) Development of Kamishiro disyllabic accent

PA F-loss shift L-sprd assml I-low

(a) *OO ª `OO = `OO = `OO = `OO > `OO
[HH(H)] [LH(H)]

(b)*OO˘ > _OO = _OO > _OO = _OO = _OO

[LH(H)] [LR ~
LL(H)]

(c) *O˘O = O˘O > OO˘ = OO˘ > OO˘ > OO˘
[HH(L)] [HF ~ [LF ~
HH(L)] LH(L)]

Trisyllables: Going through Final Accent Loss, Accent Shift,

Contour Pitch Simplification, Low Pitch Spreading, and Initial
Pitch Lowering, Kamishiro trisyllabic nouns have developed their
present-day accent.
First, Final Accent Loss deleted the accent on PA */OOO˘/
[LLH(L)] (22b), which resulted in /_OOO/ [LLH(H)]. This change
also caused a secondary change: PA */OOO/ (22a) > /`OOO/ in
contrast with the /_OOO/. Undergoing the second change, i.e.,
Accent Shift, PA */OO˘O/ (22d) merged with /OOOˆ/ (22c). At the
third stage, the rule of Contour Pitch Simplification changed PA
*/OOOˆ/ (22c) to /OO˘O/. On the final two changes (Low Pitch
Spreading and Initial Pitch Lowering), it should be pointed out that
they caused changes to the phonetic pitch shape of some nouns.
Low Pitch Spreading turned the pitch of /_OOO/ [LLH(H)] into
LLR ~ LLL(H), and at the final stage, the initial pitch of /`OOO/
became low.
With regard to typology of accent, it is important to note that in
the Kamishiro line, a locus system of Proto-Amami accent became
a locus-register system. When Proto-Amami final accent was lost,
register distinctions appeared and were kept through the
development of the Kamishiro accent system.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 161

(22) Development of Kamishiro trisyllabic accent

PA F-loss shift C-smpl L-sprd I-low

(a) *OOO ª `OOO = `OOO = `OOO = `OOO > `OOO
[HHH(H)] [LHH(H)]

(b) *OOO˘ > _OOO = _OOO = _OOO > _OOO = _OOO

[LLH(H)] [LLR ~
LLL(H)]

(c) *OOOˆ = OOOˆ

> OOOˆ > OO˘O = OO˘O = OO˘O
(d) *OO˘O = OO˘O

It should be also noted that assimilation had no effect on

monosyllables and trisyllables. This is because conditions for the
change to take place were not met. There are two such conditions.
First, the assimilation in question occurs towards the end of the
development, between Low Pitch Spreading and Initial Pitch
Lowering. Second, this change requires at least final two syllables
in high pitch. For monosyllables, since Final Accent Loss made all
final-accent monosyllables atonic before assimilation, there were
no words to which assimilation could apply.

5.2.3.3 Development of Kametsu Accent

The development of Kametsu accent is rather simple. Kametsu
monosyllables underwent one change, disyllables two changes, and
trisyllables three changes. Unlike Kamishiro, Kametsu accent did
not involve the development of register distinctions; a locus system
like that of Proto-Amami has been retained.
Monosyllables: The development of Kametsu monosyllabic
accent from Proto-Amami involves only assimilation. That is, PA
*/oo˘/ became /oo˘/ [LF ~ HH(L)]. PA */oo/ has remained the
same.
162 ACCENTUAL HISTORY

(23) Development of Kametsu monosyllabic accent

PA assml
(a) *oo [HH(H)] = oo [HH(H)]
(b) *oo˘ [LH(L)] > oo˘ [LF ~ HH(L)]

Disyllables: As shown in (24), Assimilation and Initial Pitch

Lowering contributed to the development of Kametsu disyllabic
accent. First, by means of assimilation, PA */OO˘/ [LH(L)] (24b)
has evolved to /OO˘/ [LF ~ LH(L)]. And then, the next change
lowered the initial pitch of PA */OO/ (24a), making it LH(H).
Because of this phonetic rule, Kametsu nouns are generally low on
the initial syllable except initial-accent nouns (i.e., /O˘O/). This
phenomenon is also seen in Tokyo.

(24) Development of Kametsu disyllabic accent

PA assml I-low
(a) *OO = OO > OO [LH(H)]
(b) *OO˘ > OO˘ = OO˘ [LF ~ LH(L)]
(c) *O˘O = O˘O = O˘O [HL(L)]

Trisyllables: In the development of Kametsu trisyllabic nouns

from Proto-Amami, there are three changes: Contour Pitch
Simplification, assimilation, and Initial Pitch Lowering.

(25) Development of Kametsu trisyllabic accent

PA C-smpl assml I-low

(a) *OOO = OOO = OOO > OOO
[HHH(H)] [LHH(H)]

(b) *OOO˘ = OOO˘ > OOO˘ = OOO˘

[LLH(L)] [LLF ~ LLH(L)]
(c) *OOOˆ
> OO˘O = OO˘O = OO˘O
(d) *OO˘O
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 163

The first change converted PA */OOOˆ/ (25c) into /OO˘O/. As a

result, it merged with PA */OO˘O/ (25d). The second change
turned PA */OOO˘/ [LLH(L)] (25b) to /OOO˘/ [LLF ~ LLH(L)].
And the final rule made the pitches of PA */OOO/ [HHH(H)] (25a)
to LHH(H) -- the rule lowered the height of the pitch of the first
syllable.

5.3 OKINAWA RYUKUAN

5.3.1 Description of Okinawa Ryukyuan
My reconstruction of Proto-Okinawa Rukyuan involves three
varieties of Okinawa Ryukyuan: Nakijin, Shuri, and Aguni dialects.
The Nakijin and Shuri dialects are spoken on the island of Okinawa,
and Aguni is spoken on the island of Aguni, which is located about
60 km offshore northwest from Naha. These dialects are carefully
chosen in order to reflect characteristics of Okinawa Ryukyuan
accent.

5.3.1.1 Nakijin Dialect

The Nakijin dialect is one of the northern Okinawa dialects. All
Nakijin data used in this research are extracted from Nakasone
(1983).
There are three types of accent classes in Nakijin: high-register
atonic, low-register atonic, and penultimate accent classes. Nakijin
has a phonetic rule that lowers an initial-syllable pitch if it is not
accented. Therefore, high-register atonic nouns have low pitch on
the initial syllable, e.g., `husi: LHH ‘cow’. This does not apply to
monosyllabic nouns, i.e., `ci: HH ‘blood’.
Low-register atonic nouns start, needless to say, low in pitch and
end high in pitch. In Nakijin, a following enclitic takes the same
pitch of its preceding mora. Thus, an enclitic of atonic nouns is
always high in pitch.
The TBU of Nakijin accent is the mora. This is because the final
mora of low-register atonic disyllables takes a H tone, e.g., _yama:
LLH ‘mountain’. If its TBU were the syllable, the expected word
would be LHH. For trisyllabic nouns with two moras in the final
syllable, the final syllable is high-pitched, e.g., khaga:mi: LLLHH
164 ACCENTUAL HISTORY

‘mirror’. A final long syllable can be high-pitched in longer words.

That is, if there are at least two moras before a final long syllable,
the long syllable can be high in pitch, e.g., _phasira: LLHH ‘pillar’.
Monosyllables: Nakijin has two accent types for monosyllabic
nouns: high-register and low-register. Monosyllabic nouns are
distinguished only by register; there are no locus distinctions.
Concerning the phonetic pitch shape of Nakijin monosyllables, the
high-register monosyllables are HH(H), and the low-register,
LH(H).

(26) Nakijin monosyllabic accent

Phonemic Phonetic Myô List of nouns
`oo HH(H) 1.1 `khi: ‘hair’, `ci: ‘blood’
1.2 `pha: ‘leaf’, `na: ‘name’
_oo LH(H) 1.3 _khi: ‘tree’, _mi: ‘eye’, _pha:
‘tooth’

In a comparison of Nakijin monosyllabic accent with Myôgishô

accent, there are regular correspondences. All Nakijin high-register
monosyllabic nouns are 1.1-2 class-nouns in Myôgishô, and low-
register nouns belong to Myôgishô class 1.3.
Disyllables: There are three distinctive accent classes for
Nakijin disyllabic nouns. Two of them are atonic: high-register and
low-register, and the other is penultimate-accent.
With the exception of the word `thui HH(H) ‘bird’, nouns in the
high-register class are phonetically LHH(H); the pitch begins low
and rises after the first mora. The reason why the word ‘bird’ has
high-level pitch is due to its internal structure; i.e., CVi. This
structure often influences its pitch shape (see also Ikema dialect in
5.4.1.1). Low-register disyllabic nouns are phonetically LLH(H);
the first two moras are low pitch and the pitch rises at the end of
words. The phonetic pitch shape of penultimate-accent nouns is
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 165

HL(L). The Nakijin dialect lowers the pitch of the initial syllable if
it is not accented and if it is not in CVV structure.26

(27) Nakijin disyllabic accent

Phonemic Phonetic Myô List of nouns
`Ooo LHH(H) 2.1 `phana: ‘nose’, `husi: ‘cow’,
`husi: ‘loins’, `pizi: ‘beard’,
`hazi: ‘wind’
2.2 `/isi: ~ `hisi: ‘stone’, `habi:
‘paper’, `phasi: ‘bridge’
3.1 `phice: ‘forehead’
HH(H) 2.1 `thui ‘bird’
_Ooo, LLH(H) 2.3 _phana: ‘flower’, _yama:
_ooO ‘mountain’, _kumu: ‘cloud’
2.4 _hata: ‘shoulder’, _hica: ‘board’
2.5 _/ami: ‘rain’, _hasi: 27 ‘sweat’
3.4 _/o:zi ‘fan’
3.5 _/anda: LLHH ‘fat’, _nada:
‘tears’, _maQka: LLHH ‘pillow’
3.7 _gunza ‘whale’
O˘O, HL(L), 2.3 phu˘ni(:) ‘bone’
oo˘O HHL(L)
2.4 phu˘ni(:) ‘boat’, /u˘mi(:) ‘sea’,
pha˘i ‘needle’, /u˘si(:) ‘mortar’,
/i˘ci(:) ‘breath’
2.5 hu˘i ‘voice’, ha˘gi(:) ‘shadow’,
mu˘hu(:) ‘bridegroom’, hu˘khi(:)
‘bucket’
3.2 ta:˘ci HHL‘two’

26
This phenomenon is also seen in Tokyo (McCawley 1968: 133) and other
Ryukyuan dialects discussed in this book.
27
According to Nakasone (1973: 634), the change from //a/ to [ha] occurs when
followed by a voiceless obstruent. Therefore, the word hasi: should be /asi:
phonemically. Furthermore, this devoicing rule applies to //u/ as well. Therefore,
the noun hu˘khi(:) ‘bucket’ should be phonemically /u˘khi(:).
166 ACCENTUAL HISTORY

Many Nakijin disyllables end with a long vowel. Atonic nouns

always have a long vowel in the final syllable, e.g., _hata:
‘shoulder’, _hata:-nu ‘shoulder (Nom.)’, but penultimate accent
nouns lose the length when pronounced in isolation, e.g., /u˘mi
‘sea’, /u˘mi:-nu ‘sea (Nom.)’. To distinguish these two types of
vowel length, the latter type of words are indicated with ‘(:)’ at the
end of the words, e.g., /u˘mi(:) ‘sea’.
The data above include not only historical disyllables but also
historical trisyllables that have merged with disyllabic nouns. Most
of the historically-disyllabic nouns have a long vowel at the end of
the words, except for ¯thui ‘bird’ and pha˘i ‘needle’. The word final
vowel length is predictable. In other words, Nakijin historical
disyllabic nouns have a long vowel word-finally and the ones with
CVV structure have a word-final short vowel, e.g., ‘needle’ and
‘voice’.
In the data, there are seven former trisyllables in the class of
low-register disyllabic nouns: _/o:zi ‘fan’ (3.4), _/anda: ‘fat’ (3.5),
_nada: ‘tears’ (3.5), _maQka: ‘pillow’ (3.5), _gunza ‘whale’ (3.7),
and ta:˘ci ‘two’ (3.2). And the high-register disyllabic class has
only one historical trisyllabic, namely `phice: ‘forehead’. Except
for the words meaning‘whale’, tears’, and ‘forehead’, those
historically-trisyllabic nouns are structurally different from other
disyllabic nouns in the same class. For example, the words meaning
‘fat’ and ‘pillow’ consist of four moras. The word ‘fan’ is
composed of three moras, but, unlike other disyllables, the initial
syllable is heavy, as is that of the word ta:˘ci ‘two’.
According to Nakasone (1983: 634), in Nakijin, word-initial //i/
becomes voiceless /hi/ when it is unaccented and followed by a
voiceless obstruent except /s/. However, he adds, the word for
‘stone’ has alternative forms, i.e., `/isi: ~ `hisi:, devoicing of the
initial syllable is rare.
When it comes to correspondences between Nakijin disyllabic
accent and Myôgishô accent, the correspondences are regular.
Nakijin high-register atonic nouns correspond to Myôgishô classes
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 167

2.1-2, Nakijin low-register atonic nouns to Myôgishô classes 2.3-5,

and Nakijin penultimate-accent to Myôgishô 2.3-5.
Trisyllables: Nakijin trisyllables have a two-way accent system:
atonic and penultimate accent. For atonic nouns, the pitch begins
low and rises at the end of the word. In penultimate-accent nouns,
the pitch also begins low, but it rises on the second syllable and
falls after the penultimate syllable.

(28) Nakijin trisyllabic accent28

Phonemic Phonetic Myô List of nouns
_OOO LLH(H) 3.2 _muka:zi ‘centipede’
3.4 _phuku:ru ‘bag’, _thaka:ra
‘treasure’, _khaga:mi: 29
‘mirror’
3.5 _phasira: ‘pillar’, /inuci: ‘life’
3.6 _/una:zi ‘eel’, _miminza
‘earthworm’
3.7 _khusui ‘medicine’
OO˘O LHL(L) 3.1 siru:˘si ‘mark’, khibu:˘si
‘smoke’, haku:˘bi‘yawn’
3.3 cika:˘ra ‘strength’

Concerning the word structure of trisyllabic nouns, there are four

types. The first type has a long vowel in the second syllable (i.e.,
CVCV:CV). The second type has a long vowel in the final syllable
(i.e., CVCVCV:). The third type has a long vowel in both second
and third syllables (i.e., CVCV:CV:). The fourth has no vowel
length at all (i.e., CVCVCV). According to the data, most of the

28
In this analysis, /O/ in /OOO/ indicates a syllable, and phonetic pitch is given to
each syllable, not to each mora. Thus, both khaga:mi: ‘mirror’ and khusui
‘medicine’ are LLH(H).
29
According to Hattori (1979: 101 [#5]), the Nakijin word for ‘mirror’ is a
loanword from Shuri. However, the accentuation of the word is regular; it has the
same accentuation as other nouns in the same historical accent class, namely 3.4
(e.g., ‘bag’, ‘treasure’). Therefore, I include the word ‘mirror’ in my
reconstruction.
168 ACCENTUAL HISTORY

trisyllables are the first type. For the second type, there are two
examples: phasira: ‘pillar’ and /inuci: ‘life’, and interestingly both
correspond to Myôg-shô class 3.5. The third and fourth types of
word structure have only one example: khaga:mi: ‘mirror’ for the
third type, and khusui ‘medicine’ for the fourth type.
Comparing the accentuation of Nakijin trisyllabic nouns with
their cognates in Myôgishô, we know that Nakijin atonic
trisyllables correspond to Myôgishô classes 3.2 and 3.4-7, and
Nakijin penultimate-accent trisyllables to Myôgishô classes 3.1 and
3.3. Finding another 3.2-noun saku:˘ra ‘cherry’ (not in my original
word list), now we know that Nakijin penultimate-accent nouns
regularly correspond to Myôgishô high-register classes 3.1-3. This
correspondence of the noun _muka:zi ‘centipede’ (3.2) to the same
class as 3.4-7 nouns does not seem to be regular because in
Myôgishô, 3.1-3 nouns are high-register (or high-initial) and 3.4-7
nouns are low-register (or low-initial). Myôgishô register
distinctions are usually reflected in correspondences between
Myôgishô accent classes and modern Ryukyuan accent classes. For
example, in Shuri, trisyllabic high-register nouns correspond to
Myôgishô high-register accent classes 3.1-3, and trisyllabic low-
register nouns to Myôgishô low-register classes 3.4-7.30

(29) Accentual correspondences between Shuri trisyllabic nouns

and their cognates in Myôgishô

Shuri Myôgishô
high-register high-register classes 3.1-3
low-register low-register classes 3.4-7

In the Nakijin dialect, the word muka:zi ‘centipede’ which earlier

belonged to the penultimate-accent class, might have merged with

30
In my Shuri data, no synchronically-trisyllabic tonic nouns corresponding to
Myôgishô class 3.2 are listed because all those nouns have become disyllabic by
losing a syllabic segment. However, the word `sakura ‘cherry’ (3.2) in Shuri
proves the Shuri trisyllabic tonic to Myôgishô 3.2 correspondence.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 169

atonic nouns. Therefore, as I hypothesize, the correspondences used

to be as in (30a), but they became the ones of (30b).

(30) Accentual correspondences between Nakijin trisyllabic nouns

and their cognates in Myôgishô

(a) Nakijin Myôgishô

penultimate high-register classes 3.1-3
atonic low-register classes 3.4-7

(b) Nakijin Myôgishô

penultimate high-register classes 3.1-3
atonic low-register classes 3.2 and 3.4-7

To sum up, as shown below, there are two accentual distinctions

in Nakijin for monosyllables, three for disyllables, and two for
trisyllables. Monosyllabic nouns are distinguished by register only.
For disyllabic nouns, both register and locus play a crucial role to
distinguish one word from another. Trisyllables are of two types:
atonic and tonic (penultimate-accent). That is, register does not
play a role. I analyzed LHL(L)-nouns as penultimate, not high-
register because nouns distinguished by register are atonic in both
monosyllables and disyllables. Analyzing LHL(L) as high-register
creates incoherence throughout the system. Therefore, Nakijin
lacks high-register trisyllabic nouns.

(31) Summary of Nakijin accent system

H-register `oo HH(H) `OO LHH(H)
L-register _oo LH(H) _OO LLH(H) _OOO LLH(H)
Penultimate O˘O HL(L) OO˘O LHL(L)

5.3.1.2 Shuri Dialect

The Shuri dialect of southern Okinawan is known as a descendant
of a dialect used by the Kings of the Ryukyus and the intellectuals,
170 ACCENTUAL HISTORY

and later was used as ‘a common language’ throughout the

Ryukyus. All Shuri data are from OGJ (1963).31
The accentuation of Shuri monosyllabic, disyllabic, and
trisyllabic nouns is a two-way system; nouns are either high-
register or low-register. Phonetically low-register nouns are low-
level throughout the words, and for high-register nouns the pitch
falls after the first mora for two-mora nouns, but for three-mora
nouns the pitch falls after the second mora.
Monosyllables: Shuri monosyllabic nouns have a two-way
accentual distinction; nouns are either high- or low-register. In the
former type of nouns, the pitch falls after the initial mora. The pitch
shape of the latter type is low-level. It is the register that
distinguishes one type from another in Shuri, because locus is not
distinctive.

(32) Shuri monosyllabic accent

Phonemic Phonetic Myô List of nouns
`oo HL(L) 1.1 `ki: ‘hair’, `ci: ‘blood’
1.2 `hwa: ‘leaf’, `na: ‘name’
_oo LL(L) 1.3 _ki: ‘tree’, _mi: ‘eye’, _ha:
‘tooth’
2.3 _/in ‘dog’
2.5 _kwi: ‘voice’

The nouns _/in ‘dog’ and _kwi: ‘voice’ are synchronically

monosyllabic, but they are diachronically disyllabic. Their
corresponding forms in Myôgishô are inu LL and kowe LH(L)
respectively (Mochizuki 1974: 226).
As shown above, the correspondences between Shuri accent
classes and Myôgishô accent classes for monosyllabic nouns are
regular. High-register nouns correspond to Myôgishô classes 1.1-2,
and low-register nouns to Myôgishô class 1.3.

31
There are two varieties of the Shuri dialect - one spoken by the nobility and the
other by common people. OGJ includes these two.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 171

Disyllables: The accentuation of Shuri disyllabic nouns is very

similar to that of its monosyllabic nouns; both have high-low
register distinctions.

(33) Shuri disyllabic accent

Phonemic Phonetic Myô List of nouns
`OO HL(L) 2.1 `hana ‘nose’, `/usi ‘cow’, `tui
‘bird’, `kusi ‘loins’, `hwizi
‘beard’, `kazi ‘wind’
2.2 `/isi ‘stone’, `kabi ‘paper’, `hasi
‘bridge’
3.1 `hwice: ‘forehead’
3.2 `ta:ci ‘two’
_OO LL(L) 2.3 _hana ‘flower’, _yama
‘mountain’, _kumu ‘cloud’, _huni
‘bone’
2.4 _huni ‘boat’, _/umi ‘sea’, _ha:i
‘needle’, _/u:si ‘mortar’,
_ha(:)si ‘chopsticks’, _kata
‘shoulder’, _/i:ci ‘breath’, _/ica
‘board’
2.5 _ka:gi ‘shadow’, _mu:ku
‘bridegroom’, _/ami ‘rain’, _/asi
‘sweat’, _u:ki ‘bucket’
3.2 _nkazi ‘centipede’
3.4 _kagan ‘mirror’, _/o:zi ‘fan’
3.5 _ha:ya ‘pillar’, _/anda ‘fat’,
_nuci ‘life’, _nada ‘tears’,
_maQkwa ‘pillow’
3.6 _/nnazi ‘eel’32

32
Just as a syllable-final nasal consonant in the structure of CVn (see Poser 1990
and Vance 1987), I consider a nasal consonant in the structure such as (/)nCV- or
(/)mCV- in the word initial position as moraic, not syllabic. The reason for this
analysis is that there is no evidence to support the claim that the initial nasal in
question is syllabic.
172 ACCENTUAL HISTORY

In low-register nouns the pitch shape of the words is low-level.

In two-mora high-register nouns the pitch falls after the first
syllable (e.g., `hana HL ‘nose’), and in three-mora high-register
nouns the pitch falls after the second mora (e.g., `hwice: HHL
‘forehead’).
As shown below, the list of disyllabic nouns includes formerly-
trisyllabic nouns, which have become disyllabic by losing a
syllabic segment (i.e., namida > nada ‘tears’ and hitai > hwice:
‘forehead’).
Notice also that some of the originally-disyllabic low-register
nouns are three-mora, with a long vowel in the initial syllable, e.g.,
_ka:gi ‘shadow’. No originally-disyllabic high-register nouns show
this phenomenon. Furthermore, those nouns with the length are in
either class 2.4 or 2.5 in Myôgishô.
There are regular correspondences between Shuri historically-
disyllabic nouns and Myôgishô disyllabic nouns. The Shuri low-
register nouns correspond to Myôgishô accent classes 2.3-5, and the
Shuri high-register nouns to Myôgishô classes 2.1-2.
Trisyllables: Shuri trisyllabic nouns also have two accent
classes: high-register and low-register. High-register nouns are in
HHL(L) pitch, and low-register nouns have the shape of LLL(L).

(34) Shuri trisyllabic accent

Phonemic Phonetic Myô List of nouns
`OOO HHL(L) 3.1 `sirusi ‘mark’, `kibusi ‘smoke’,
`/akubi ‘yawn’
3.3 `cikara ‘strength’, `kugani ‘gold’
_OOO LLL(L) 3.4 _hukuru ‘bag’, _takara ‘treasure’
3.6 _mimizi ‘earthworm’
3.7 _kusui ‘medicine’, _guzira
‘whale’, _hataki ‘field’

As far as my data are concerned, Shuri trisyllabic high-register

nouns correspond to Myôgishô classes 3.1 and 3.3, and trisyllabic
low-register nouns to Myôgishô classes 3.4 and 3.6-7.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 173

In the data, all trisyllabic nouns corresponding to Myôgishô

classes 3.2 and 3.5 merged with disyllabic nouns when they lost a
syllabic segment. However, from other Shuri nouns such as
`sakura ‘cherry blossom’ (3.2), _/icuku ‘cousin’ (3.5), _cinuku
‘mushroom’ (3.5), and _waraba: ‘child’ (3.5), it is possible to
discover to what accent class 3.2- and 3.5-nouns belong in Shuri.
That is, nouns corresponding to Myôgishô classes 3.1-3 are high-
register, and nouns corresponding to Myôgishô classes 3.4-7 are
low-register in Shuri.

(35) Correspondences between Shuri trisyllabic accent and

Myôgishô accent

Shuri accent : Myôgishô accent

high register : 3.1-3
low register : 3.4-7

5.3.1.3 Aguni Dialect

As a dialect of Aguni island, I will use the Nishi dialect spoken on
the western side of the Island of Aguni. All data of the language are
extracted from Hirayama et al. (1966: 196-8). In the data below,
bold-faced words (i.e., kata ‘shoulder’, /ame ‘rain’, and /ase ‘sweat’)
are written in either Japanese syllabary or Chinese characters in the
original source material.
The Aguni accent is typologically a locus system -- words are
distinguished by location of accent. In Aguni, there are three
accentual classes: prototonic, a class accented on the second-
mora,33 and atonic.
Prototonic nouns have a high pitch on the initial mora and other
moras following it are low in pitch, e.g., si˘rusi HLL ‘mark’. For
nouns accented on the second-mora, their pitch shape depends on

33
‘Second-mora-accent’ means that an accent is located on the second mora from
the beginning of words. Throughout this book, indicating the location of a specific
mora or syllable using the words such as ‘first’ or ‘initial’, ‘second’, and ‘third’, I
count moras or syllable from the left to the right.
174 ACCENTUAL HISTORY

the number of syllables they consist of. Bimoraic disyllables are

simply LH(L), e.g., umi˘ LH ‘sea’. Three-mora disyllables are
LHL(L), e.g., ha:˘i LHL ‘needle’. Three-mora trisyllables are
HHL(L) with a following enclitic, but low-level in isolation, e.g,
gara˘si LLL ‘crow’, gara˘si-ga HHL-L ‘crow (Nom.)’. Aguni
atonic nouns are simply low-level pitch throughout the words.
Monosyllables: There are two accentual classes for
monosyllabic nouns: initial accent and low atonic. I do not treat this
accent system as a register system even though there are two accent
classes; one is initial-high and the other is initial-low. The reason
why I analyze Aguni monosyllabic accent as a system without
register is that its disyllables and trisyllables function entirely on
the basis of locus (see below).
In the initial-accent nouns, the pitch falls between the initial
mora and the final mora. All monosyllabic nouns are composed of
two moras, with a long vowel.

(36) Aguni monosyllabic accent

Phonemic Phonetic Myô List of nouns
o˘o HL(L) 1.1 ki˘: ‘hair’
1.2 ha˘: ‘leaf’
oo LL(L) 1.3 ki: ‘tree', ha: ‘tooth’

A comparison of Aguni monosyllabic nouns and their cognates

in Myôgishô shows that Aguni initial-accent nouns correspond to
1.1-2 class nouns in Myôgishô, and Aguni atonic nouns to
Myôgishô 1.3 nouns.
Disyllables: The Aguni accent system for disyllabic nouns is
complex. There are three types of accentuation: atonic, initial
accent, and second-mora accent. Atonic nouns are phonetically
low-level pitch. Initial-accent nouns are HL or HLL pitch; pitch
falls between the initial mora and the second mora from the
beginning of the words. For second-mora accent nouns, pitch
basically falls immediately after the second mora from the
beginning of the words. There are two types of second-mora-accent
nouns; one contains two moras and the other three moras with a
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 175

long vowel. For the former type, the pitch falls at the word final
position, i.e., /OO˘/, and for the latter, the pitch falls between the
second and the final moras, e.g., hice˘: ‘forehead’. Furthermore, of
second-mora-accent, ones with three moras become low-level
atonic in isolation, e.g., ha:i-nu LHL-L ‘needle (Nom.)’, ha:i LLL
‘needle’.
With regard to accentual correspondences between Aguni
disyllables and their cognates in Myôgishô, low-atonic nouns
correspond to Myôgishô accent classes 2.3-5. Initial-mora-accent
disyllables correspond to Myôgishô 2.1-2 nouns. Second-mora-
accent cognates belong to accent classes 2.4-5 in Myôgishô. It
should be noted that this description of accentual correspondences
includes original disyllables only; thus synchronically-disyllabic
nouns corresponding to Myôgishô trisyllables are excluded, e.g.,
nkazi ‘centipede’ (3.2), etc.

(37) Aguni disyllabic accent

Phonemic Phonetic Myô List of nouns
OO LL(L) 2.3 hana ‘flower’, yama ‘mountain’
2.4 kata ‘shoulder’
2.5 /ame ‘rain’,/ase ‘sweat’
3.2 nkazi ‘centipede’
O˘O HL(L) 2.1 ha˘na ‘nose’
2.2 /i˘si ‘stone’, ha˘bi ‘paper’
HHL(L) 3.1 kin˘si ‘smoke’
OO˘, LH(L) 2.4 huni˘ ‘boat’, /umi˘ ‘sea’
oo˘O, 2.5 muku˘ ‘bridegroom’
Oo˘o LHL(L) 2.4 ha:˘i ‘needle’, /u:˘si ‘mortar’,
ha:˘si ‘chopsticks’
3.1 hice˘: ‘forehead’
3.5 ha:˘ya ‘pillar’

In my synchronic analysis of Aguni disyllabic accent, there are

nouns corresponding to historical trisyllabic accent, e.g., kin˘si
‘smoke’ (3.1), hice˘: ‘forehead’ (3.1), ha:˘ya ‘pillar’ (3.5), and -
176 ACCENTUAL HISTORY

nkazi ‘centipede’ (3.2). They are historically trisyllabic, but in the

course of their development, they became disyllabic.
Trisyllables: For Aguni trisyllabic nouns, there are three accent
classes: low atonic, initial accent, and second-mora accent. Because
many Aguni trisyllabic cognates on my word list are not available
and also since many trisyllables became disyllabic, I have added
two more words in (37) (i.e., kuyumi ‘calendar’ (3.4) and usa˘gi
‘rabbit’ (3.6)), which are not on the list. As shown below, there are
only one or two examples for each class.
Regarding the pitch shape of words, atonic nouns are LLL(L).
The initial-accent noun is HLL(L). Second-mora-accent nouns are
phonetically HHL(L) when followed by a particle, but they are
low-level in isolation, e.g., /usa˘gi LLL ‘rabbit’, /usa˘gi-nu HHL-
L ‘rabbit (Nom.)’.

(38) Aguni trisyllabic accent

Phonemic Phonetic Myô List of nouns
OOO LLL(L) 3.4 kuyumi ‘calendar’
O˘OO HLL(L) 3.1 si˘rusi ‘mark’
OO˘O LLL ~ 3.6 /usa˘gi ‘rabbit’
HHL(L)

Based on the data, it is difficult to know a general pattern of the

correspondences between Aguni trisyllabic nouns and their
cognates in Myôgishô because each class has only one or two
examples. However, based on the larger data set, Hirayama et al.
(1966: 197) state that second-mora-accent nouns generally
correspond to Myôgishô classes 3.4-7, and initial-accent nouns to
Myôgishô 3.1-3. Although Hirayama et al. claim regular
correspondences, they do not present concrete examples to support
their claim.

5.3.2 Correspondences and Reconstruction of Proto-Okinawa

Accent
In this section, I show sets of correspondences for monosyllabic,
disyllabic, and trisyllabic accent. On the basis of the
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 177

correspondence sets, I also set forth my reconstructed Proto-

Okinawa (PO) accent system.
Monosyllables: Based on the following two sets of
correspondences, I reconstruct tonic accent (i.e., */o˘o/) and atonic
accent (i.e., */oo/) classes for Proto-Okinawa monosyllabic nouns.
Specifically, for the correspondence in (39b), low-atonic accent can
be reconstructed because all corresponding accent classes show
low-atonic accentuation.

(39) Reconstructed Proto-Okinawa accent for monosyllabic nouns

(a)
Nakijin Shuri Aguni PO
`oo `oo o˘o *o˘o
HH(H) HL(L) HL(L) HL(L)
‘hair’ (1.1) `khi: `ki: ki˘: *khi˘:
‘blood’ (1.1) `ci: `ci: -- *ci˘:
‘leaf’ (1.2) `pha: `hwa: ha˘: *pha˘:
‘name’ (1.2) `na: `na: -- *na˘:

(b)
Nakijin Shuri Aguni PO
_oo _oo Oo *oo
LH(H) LL(L) LL(L) LL(L)
‘tree’ (1.3) _khi: _ki: ki: *khi:
‘eye’ (1.3) _mi: _mi: -- *mi:
‘tooth’ (1.3) _pha: _ha: ha: *pha:
‘dog’ (2.3) (/innukwa:) _/in -- */in

On the other hand, the correspondence in (39a) shows that

phonetically, the Nakijin form is high-level atonic, and Shuri and
Aguni are initial-high. This difference has to be reflected in my
reconstruction. In other words, my reconstruction has to explain the
development of these different pitch shapes.
Notice that I reconstruct the historically-disyllabic noun */in
‘dog’ as monosyllabic in Proto-Okinawa exclusively on the basis of
the Shuri word, because Nakijin and Aguni do not have a cognate
form. As far as the data are concerned, at the time of Proto-
178 ACCENTUAL HISTORY

Okinawa the word for ‘dog’ might have already become

monosyllabic.
Disyllables: My Okinawa Ryukyuan data show that there are
three sets of correspondences for disyllabic nouns.

(40) Summary of accentual correspondences and reconstruction

of Proto-Okinawa accent for disyllables
Nakijin Shuri Aguni PO
a `Ooo LHH(H) `OO HL(L) O˘O HL(L) O˘O HL(L)
b _Ooo, _ooO _OO LL(L) OO LL(L) *OO LL(L)
LLH(H)
c O˘O HL(L) OO˘ LH(L) *OO˘ HH(L)

Correspondence sets in (41) lack a Nakijin cognate for

‘chopsticks’ and a number of Aguni cognates. In Nakijin, the word
for ‘chopsticks’ is a compound noun, thus it has been excluded
from the data. For the correspondences where Aguni cognates are
missing, I reconstructed Proto-Okinawa accent forms simply based
on Nakijin and Shuri forms.

(41) Reconstructed Proto-Okinawa accent for disyllabic nouns

(a)
Nakijin Shuri Aguni PO
`Ooo LHH(H) `OO HL(L) O˘O HL(L) *O˘O HL(L)
‘nose’ (2.1) `phana: `hana ha˘na *pha˘na
‘cow’ (2.1) `husi: `/usi -- */u˘si
‘bird’ (2.1) `thui `tui -- *thu˘i
‘loins’ (2.1) `husi: `kusi -- *khu˘si
‘beard’ (2.1) `pizi: `hwizi -- *phi˘zi
‘wind’ (2.1) `hazi: `kazi -- *kha˘zi
‘stone’ (2.2) `/isi: ~ `hisi: `/isi /i˘si */i˘si ~ *hi˘si
‘paper’ (2.2) `habi: `kabi ha˘bi *kha˘bi
‘bridge’ (2.2) `phasi: `hasi -- *pha˘si
‘forehead’(3.1) `phice: `hwice: hice˘: *phi˘ce:34

34
The forms for ‘forehead’ in Shuri, Aguni, and Proto-Okinawa are phonetically
HHL(L).
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 179

(b)
Nakijin Shuri Aguni PO
O˘O HL(L) _OO LL(L) OO˘ LH(L), *OO˘ HH(L)
LHL(L)
‘bone’ (2.3) phu˘ni(:) _huni -- *phuni˘
‘mortar’ (2.4) /u˘si(:) _/u:si /u:˘si */u:si˘
‘chopsticks’ (2.4) -- _ha(:)si ha:˘si *pha:si˘
‘needle’ (2.4) pha˘i _ha:i ha:˘i *pha:i˘
‘boat’ (2.4) phu˘ni(:) _huni huni˘ *phuni˘
‘sea’ (2.4) /u˘mi(:) _/umi /umi˘ */umi˘
‘breath’ (2.4) /i˘ci(:) _/i:ci -- */i:ci˘
‘shadow’ (2.5) ha˘gi(:) _ka:gi -- *kha:gi˘
‘bucket’ (2.5) hu˘khi(:) _u:ki -- *u:khi˘
‘bridegroom’(2.5) mu˘hu(:) _mu:ku muku˘35 *mu:ku˘
‘voice’ (2.5) hu˘i _kwi -- *kui˘

(c)
Nakijin Shur Aguni PO
_Ooo, _ooO _OO LL(L) OO LL(L) *OO LL(L)
LLH(H)
‘flower’ (2.3) _phana: _hana hana *phana
‘mountain’ (2.3) _yama: _yama yama *yama
‘cloud’ (2.3) _kumu: _kumu -- *kumu
‘shoulder’ (2.4) _hata: _kata kata *khata
‘board’ (2.4) _hica: _/ica -- */ica
‘rain’ (2.5) _/ami: _/ami /ame */ami
‘sweat’ (2.5) _hasi: _/asi /ase */asi
‘fan’ (3.4) _/o:zi _/o:zi -- */o:zi
‘tears’ (3.5) _nada: _nada -- *nada
‘fat’ (3.5) _/anda: _/anda -- */anda
‘pillow’ (3.5) _maQka: _maQkwa -- *maQkwa

With regard to the long vowel in the initial syllable of some of

the Shuri and Aguni cognates in (41b), see my discussions in 5.3.3,

35
Some speakers pronounce mu˘ku ~ muku˘-nu (Hirayama et al. 1966: 196).
Alternative forms are usually due to a transitional stage from one form to another.
In this case, the form mu˘ku is probably shifting its accent onto the final syllable,
i.e., muku˘.
180 ACCENTUAL HISTORY

where I revise the Proto-Okinawa accent form in (41b) in relation

to vowel length. For the time being, I leave the reconstruction in
(41b) as it is.
I do not reconstruct a Proto-Okinawa form for ‘two’ (3.2) even
though, as shown in (42), there are cognates in Nakijin and Shuri.
The word for ‘two’ is originally trisyllabic, but has become
disyllabic in Nakijin and Shuri by losing a syllabic segment. The
reason why this correspondence is excluded is that the
correspondence does not match any of the historically-disyllabic
accentual patterns, although other former trisyllables fit in one of
the disyllabic accentual correspondence patterns, e.g., see the
correspondences for ‘tears’ (3.5), ‘fat’ (3.5), ‘pillow’ (3.5), ‘fan’
(3.4), and ‘forehead’ (3.1) in (41). Since none of the original
disyllables show the same correspondence pattern as the one in (42),
the correspondence may possibly be irregular. The following
historically-trisyllabic noun ta:ci ‘two’ has an irregular
accentuation in either Nakijin or Shuri. However, without an Aguni
cognate, I am not able to determine which dialect has an irregular
form. Erring on the side of caution, I do not include the
correspondence for ‘two’ in my reconstruction.

(42)
Nakijin Shuri Aguni
O˘O HL(L) `OO HL(L) --
‘two’ (3.2) ta:˘ci `ta:ci --

As shown in (41), Proto-Okinawa forms for the historically-

trisyllabic nouns such as ‘fan’, ‘tears’, ‘fat, ‘pillow’, and ‘forehead’
are disyllabic: */o:zi ‘fan’, *nada ‘tears’, */anda ‘fat’, *maQka
‘pillow’, and *phi˘ce: ‘forehead’, because there is no evidence that
those words were trisyllabic in Proto-Okinawa.
Finally, regarding the reconstructed word-final long vowel in the
word ‘forehead’, it has to be reconstructed because it is not
predictable. The final vowel of the word has been lengthened in
order to compensate a syllabic segment that has been lost in the
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 181

course of its development in Nakijin, Shuri, and Aguni (see also

discussion in 5.3.1.1).
Trisyllables: For Okinawa Ryukyuan trisyllabic nouns, there are
three sets of correspondences. Based on this, for Proto-Okinawa
trisyllabic nouns three accent classes are reconstructed: prototonic,
mesotonic, and low-level atonic.

(43) Summary of accentual correspondences and reconstruction

of Proto-Okinawa accent for trisyllables

Nakijin Shuri Aguni PO

a OO˘O `OOO O˘OO *O˘OO
LHL(L) HHL(L) HLL(L) HLL(L)
b OO˘O *OO˘O
_OOO _OOO HHL(L) HHL(L)
c LLH(H) LLL(L) OOO *OOO
LLL(L) LLL(L)

It should be noted that the Aguni and Proto-Okinawa accent

patterns are phonemically identical. That is to say, Aguni has
retained the Proto-Okinawa accentuation, and Nakijin and Shuri
have undergone changes (see 5.3.4 for the development of their
accentual system).
Although the Aguni dialect lacks many cognates in the data, I
am still able to reconstruct Proto-Okinawa accent. This is based on
both the accentual patterns that Nakijin and Shuri dialects show and
regular correspondences between Nakijin and Shuri trisyllabic
accents in relation to Myôgishô accent classes (see above for the
regular correspondences).
Based on the Shuri form only, PO *khu˘gani ‘gold’ has been
reconstructed in (44a), because in the data Nakijin and Aguni lack a
cognate form for ‘gold’. Needless to say, I assume the Shuri form
has regular accentuation.
182 ACCENTUAL HISTORY

(44) Reconstructed Proto-Okinawa accent for trisyllabic nouns

(a)
Nakijin Shuri Aguni PO
OO˘O `OOO O˘OO *O˘OO
LHL(L) HHL(L) HLL(L) HLL(L)
‘mark’ (3.1) siru:˘si `sirusi si˘rusi *si˘rusi
‘smoke’(3.1) khibu:˘si `kibusi kin˘si *khi˘busi
‘yawn’(3.1) haku:˘bi `/akubi -- *ha˘kubi
‘cherry’ (3.2) saku:˘ra `sakura -- *sa˘kura
‘strength’ (3.3) cika:˘ra `cikara -- *ci˘kara
‘gold’ (3.3) -- `kugani -- *khu˘gani

(b)
Nakijin Shuri Aguni PO
_OOO _OOO OO˘O *OO˘O
LLH(H) LLL(L) HHL(L) HHL(L)
‘bag’ (3.4) _phuku:ru _hukuru -- *phuku˘ru
‘treasure’ (3.4) _thaka:ra _takara -- *thaka˘ra
‘mirror’ (3.4) _khaga:mi: _kagan -- *khaga˘mi
‘pillar’ (3.5) _phasira: _ha:ya ha:˘ya *phasi˘ra
‘life’ (3.5) _/inuci: _nuci -- */inu˘ci
‘heart’ (3.5) _khuku:ru _kukuru -- *khuku˘ru
‘cousin’ (3.5) _hici(:)ku: _/icuku -- */icuku
‘eel’ (3.6) _/una:zi _/nnazi -- */una˘zi
‘earthworm’(3.6) _miminza _mimizi -- *mimi˘zi36
‘whale’(3.7) _gunza _guzira -- *guzi˘ra37
‘medicine’ (3.7) _khusui _kusui -- *khusu˘i
‘field’ (3.7) -- _hataki -- *phata˘ki

36
I am assuming that a long vowel became short and a nasal segment came into
existence in Nakijin form _miminza ‘earthworm’. This is based on Martin's
hypothesis that the Japonic voiced consonants are prenazalized in their earlier
forms (see Martin 1987: 20).
37
In Ryukyuan, a sequence consisting of a consonant, the vowel /u/ or /i/, and /r/
(i.e., /Cur-/ or /Cir-/) often becomes /nd-/ or /nz-/, e.g., *abura ‘oil’ > anda in
Shuri, *gudira ‘whale’ > gunzya in Shodon. Thus, I reconstruct *guzira for
‘whale’.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 183

(c)
Nakijin Shuri Aguni PO
_OOO _OOO OOO *OOO
LLH(H) LLL(L) LLL(L) LLL(L)
‘centipede’(3.2) _muka:zi _nkazi nkazi *mukazi
‘calendar’ (3.4) _khuyumi: _kuyumi kuyumi *khuyumi

As for (44c), this accent class is reconstructed on the basis of

two correspondence sets: ‘centipede’ and ‘calendar’. The latter is
not in my original word list, but I included it here in order to show
that there is another accent class, which cannot be revealed by the
limited amount of data.
On vowel length in the second syllable and at the end of words
in Nakijin trisyllables, I hypothesize that the length is a secondary
development. That is, Proto-Okinawa trisyllabic nouns did not have
the length. In the Nakijin line the feature appeared, but not in Shuri
and Aguni.

5.3.3 Accent and Vowel Length

Hattori (1979a) points out that in Ryukyuan there are two
subclasses for disyllabic nouns corresponding to Myôgishô classes
2.3-5; namely, 2.3ab, 2.4ab, and 2.5ab (see also discussion on this
matter in chapter 2 of this book). For convenience, I use the
following examples shown earlier in (15) and (16) in chapter 2.

(45) Subclasses 2.3-5a

Nakijin Shuri
‘jar’ (2.3a) ha˘mi _ka:mi
‘breath’ (2.4a) /i˘ci _/i:ci
‘shadow’ (2.5a) ha˘gi _ka:gi
184 ACCENTUAL HISTORY

(46) Subclasses 2.3-5b

Nakijin Shuri
‘hole’ (2.3b) _/ana: _/ana
‘board’ (2.4b) _hica: _/ica
‘sweat’ (2.5b) _hasi: _/asi

In subclasses 2.3-5a, the initial accent in Nakijin regularly

corresponds to initial vowel length in Shuri. On the other hand, in
subclasses 2.3-5b, there is no such correspondence.

(47) Nakijin accent-Shuri vowel length correspondence

Nakijin Shuri
‘mortar’ (2.4a) /u˘si(:) _/u:si
‘chopsticks’ (2.4a) -- _ha(:)si
‘breath’ (2.4a) /i˘ci(:) _/i:ci
‘shadow’ (2.5a) ha˘gi(:) _ka:gi
‘bucket’ (2.5a) hu˘khi(:) _u:ki
‘needle’ (2.5a) pha˘i _ha:i
‘bridegroom’ (2.5a) mu˘hu(:) _mu:ku

(48) No accent-vowel length correspondence

Nakijin Shuri
‘flower’ (2.3b) _phana: _hana
‘mountain’ (2.3b) _yama: _yama
‘cloud’ (2.3b) _kumu: _kumu
‘shoulder’ (2.4b) _hata: _kata
‘board’ (2.4b) _hica: _/ica
‘rain’ (2.5b) _/ami: _/ami
‘sweat’ (2.5b) _hasi: _/asi
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 185

Hattori’s claim holds true in most of my data as well, except for

the words meaning ‘bone’ (2.3), ‘boat’ (2.4), and ‘sea’ (2.4), which
lack a long vowel in the corresponding initial syllable in Shuri.

(49) Irregular correspondence

Nakijin Shuri
‘bone’ (2.3a) phu˘ni(:) _huni
‘boat’ (2.4a) phu˘ni(:) _huni
‘sea’ (2.4a) /u˘mi(:) _/umi

In Shuri the initial-syllable vowel of ‘bone’ (2.3), ‘boat’ (2.4),

and ‘sea’ (2.4) must have been long earlier in its history, but it has
been irregularly shortened. Many Ryukyuan dialects underwent
shortening of the long vowel, but some Ryukyuan dialects have
kept the length that Shuri has lost. Examples are given below.

(50) Examples for initial-syllable vowel length lost in Shuri, but

kept in other Ryukyuan dialects

‘boat’ (2.3) hi:ni˘ in Kamishiro (Hirayama et al. 1966:

159); pu:ni in Onna (Hattori 1979a: 104);
hu:ni in Matsuda, Ginoza village
(Ôyama 1962: 45)

‘bone’ (2.4) hu:ni in Maja dialect, Nakazato village,

Kumejima (Ôyama 1962: 19)

‘sea’ (2.4) /u:mi in Maja dialect, Nakazato village,

Kumejima (Ôyama 1962: 19)

Furthermore, this initial-syllable vowel length in ‘boat’, ‘bone’,

and ‘sea’ is also seen in some of the Kyushu dialects, implying that
the length in historical accent classes 2.3-5a go back to an earlier
stage than Proto-Ryukyuan. I discuss this matter in chapter 7.
186 ACCENTUAL HISTORY

Based on my hypothesis that Shuri nouns for ‘bone’, ‘boat’, and

‘sea’ have irregularly shortened a long vowel in the initial syllable,
I reconstruct vowel length for historical classes 2.3-5a. Therefore, I
revise my reconstruction of Proto-Okinawa accent as shown below.
Changes are seen only in (51b). See below for explanations for the
changes. For convenience, only Nakijin and Shuri forms are listed
with Proto-Okinawa forms.

(51) Revised Proto-Okinawa disyllabic accent

(a)
Nakijin Shuri Previous PO Revised PO
`OO `OO *O˘O *O˘O
LHH(H) HL(L) HL(L) HL(L)
‘nose’ (2.1) `phana: `hana *pha˘na *pha˘na
‘cow’ (2.1) `husi: `/usi */u˘si */u˘si
‘bird’ (2.1) `thui `tui *thu˘i *thu˘i
‘loins’ (2.1) `husi: `kusi *khu˘si *khu˘si
‘beard’ (2.1) `pizi: `hwizi *phi˘zi *phi˘zi
‘wind’ (2.1) `hazi: `kazi *kha˘zi *kha˘zi
‘stone’ (2.2) `/isi: ~ `/isi */i˘si ~ */i˘si ~
¯hisi: *hi˘si *hi˘si
‘paper’ (2.2) `habi: `kabi *kha˘bi *kha˘bi
‘bridge’(2.2) `phasi: `hasi *pha˘si *pha˘si
‘forehead’ (3.1) `phice: `hwice: *phi˘ce: *phi˘ce:

(b)
Nakijin Shuri Previous PO Revised PO
O˘O _OO *OO˘ *oo˘O
HL(L) LL(L) HH(L) LHL(L)
‘bone’ (2.3) phu˘ni(:) _huni *phuni˘ *phu:ni˘
‘mortar’(2.4) /u˘si(:) _/u:si */u:si˘ */u:˘si
‘chopsticks’ (2.4) -- _ha(:)si *pha:si˘ *pha:˘si
‘boat’ (2.4) phu˘ni(:) _huni *phuni˘ *phu:˘ni
‘sea’ (2.4) /u˘mi(:) _/umi */umi˘ */u:mi˘
‘breath’(2.4) /i˘ci(:) _/i:ci */i:ci˘ */i:˘ci
‘shadow’ (2.5) ha˘gi(:) _ka:gi *kha:gi˘ *kha:gi˘
‘bucket’(2.5) hu˘khi(:) _u:ki *u:khi˘ *u:khi˘
‘needle’(2.5) pha˘i _ha:i *pha:i˘ *pha:i˘
‘bridegroom’ (2.5) mu˘hu(:) _mu:ku *mu:ku˘ *mu:ku˘
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 187

(c)
Najijin Shuri Previous PO Revised PO
_OO _OO *OO *OO
LLH(H) LL(L) LL(L) LL(L)
‘flower’(2.3) _phana: _hana *phana *phana
‘mountain’ (2.3) _yama: _yama *yama *yama
‘cloud’ (2.3) _kumu: _kumu *kumu *kumu
‘shoulder’(2.4) _hata: _kata *khata *khata
‘board’ (2.4) _hica: _/ica */ica */ica
‘rain’ (2.5) _/ami: _/ami */ami */ami
‘sweat’(2.5) _hasi: _/asi */asi */asi
‘fan’ (3.4) _/o:zi _/o:zi */o:zi */o:zi
‘tears’ (3.5) _nada: _nada *nada *nada
‘fat’ (3.5) _/anda: _/anda */anda */anda
‘pillow’ (3.5) _maQka: _maQkwa *maQkwa *maQkwa

Furthermore, there is another piece of independent evidence to

support the hypothesis that Shuri initial-syllable vowel length is a
remnant of an earlier form of Ryukyuan language. Comparing
Shuri disyllabic nouns with those of Nakijin reveals that there are
two sets of correspondences, as shown below. In (52), the Shuri
initial-syllable #/V:-38 corresponds to the syllable #/V- in Nakijin.
By contrast, in (53), the Shuri syllable #/V- corresponds to #hV- in
Nakijin. That is to say, when Shuri words contain a long vowel in
the initial syllable, the corresponding Nakijin syllable has an initial
glottal stop ///. On the other hand, when Shuri words have an
initial-syllable short vowel, the corresponding Nakijin syllable has
an initial /h/.

(52) Correspondence between Nakijin #/V- and Shuri #/V:-

Nakijin Shuri
‘mortar’ (2.4) /u˘si(:) _/u:si
‘breath’ (2.4) /i˘ci(:) _/i:ci

38
The symbol ‘#’ means a word boundary, and ‘-’ indicates that there is a segment
which follows.
188 ACCENTUAL HISTORY

(53) Correspondence between Nakijin #hV- and Shuri #/V-

Nakijin Shuri
‘board’ (2.4) _hica: _/ica
‘sweat’ (2.5) _hasi: _/asi
‘cow’ (2.1) `husi: `/usi
‘stone’ (2.2) `/isi: ~ `hisi: `/isi

What is significant is that Nakijin words corresponding to Shuri

words with an initial short vowel underwent devoicing (i.e., #/V- >
#hV-); by contrast, a Nakijin noun corresponding to a Shuri noun
with an initial long vowel has remained the same. That is, the
initial-syllable long vowel, which existed in earlier forms in
Nakijin, did not allow the syllable to undergo vowel devoicing in
‘mortar’ (*/u:˘si(:) > /u˘si(:)) and ‘breath’ (*/i:˘ci(:)) > /i˘ci(:)).
On the other hand, words in (52) underwent it, because their earlier
forms did not have a long vowel in the initial syllable (e.g., *_/asi:
> _hasi: ‘sweat’).
Notice that in (51b) /OO˘/ was reconstructed first, and then it
was revised to /oo˘O/. I consider that both have second-mora
accent because Proto-Okinawa puts an accent on the second mora.
When the second mora is part of a long syllable (i.e., -CV: or -CVn,
pitch falls immediately after the syllable, e.g., gara:˘si HHHL
‘crow’.

5.3.4 Development of Modern Okinawan Accent from Proto-

Okinawa Accent
In (54) I have summarized the Proto-Okinawa accent system
reconstructed in the previous section. There are two classes for
monosyllables, three for disyllables and trisyllables. I will explain
how this Proto-Okinawa accent system has evolved into Nakijin,
Shuri, and Aguni dialects below.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 189

(54) Proto-Okinawa accent

Monosyllables Disyllables Trisyllables
*o˘o HL(L) *O˘O HL(L) *O˘OO HLLL(L)
*oo LL(L) *oo˘O LHL(L) *OO˘O HHHL(L)
*OO LL(L) *OOO LLL(L)

5.3.4.1 Development of Nakijin Accent

Monosyllables: In the development of Nakijin monosyllabic
accent, Proto-Okinawa monosyllabic accent has undergone two
sound changes: Accent Shift and Final Accent Loss. First, Accent
Shift turned PO */o˘o/ into /oo˘/; the locus shifted from the initial
mora to the second mora. Second, Final Accent Loss deleted the
locus of /oo˘/ [HH(L)], resulting in /`oo/. Along with this change,
PO */oo/ [LL(L)] became /_oo/.
In addition, regarding phonetic pitch shape of words, Nakijin
dialect has a phonetic rule, which raises word-final pitch of low-
level words. This rule raised the pitch of the final mora, namely
LL(L) > LH(H).

(55) Development of Nakijin monosyllabic accent

PO shift F-loss
(a) *o˘o > oo˘ > `oo
[HL(L)] [HH(L)] [HH(H)]

(b) *oo = oo ª _oo

[LL(L)] [LH(H)]

When it comes to the ordering of changes, Accent Shift must

have preceded Final Accent Loss. The rule of final pitch rising
must have taken place towards the end of the development.
Disyllables: The development of Nakijin disyllabic accent is
complex. As shown in (56), there are five changes: Accent Shift,
Final Accent Loss, Vowel Shortening, Vowel Lengthening (V-
lngth), and Initial Pitch Lowering. They took place as follows.
190 ACCENTUAL HISTORY

First, Accent Shift, applicable only for high-initial nouns, shifted

the accent of PO */O˘O/ (56a) onto the final syllable, and then it
underwent Final Accent Loss, resulting in /`OO/. In addition to this,
PO */OO/ (56c) underwent a secondary change, i.e., */OO/ >
/_OO/. Notice that PO */oo˘O/ (56b) did not undergo Accent Shift,
because its initial pitch is low.

(56) Development of Nakijin disyllabic accent

PO shift F-loss V-shrt V-lngth I-low

(a) *O˘O > OO˘ > `OO = `OO > `Ooo > `Ooo
[HH(L)] [LHH(H)]

(b) *oo˘O = oo˘O = oo˘O > O˘O > O˘o(o) = O˘o(o)

[LHL(L)]

(c) *OO = OO ª _OO = _OO > _Ooo = _Ooo

[LL(L)] [LLH(H)]

As for Vowel Shortening and Vowel Lengthening, the former

must have taken place some time after Accent Shift occurred;
otherwise, PO */oo˘O/ would have been affected by the Accent
Shift rule. Concerning Vowel Lengthening, it is not ordered in
relation to other changes. However, for convenience, I place it as
shown in (56).
Undergoing Vowel Shortening, PO */oo˘O/ (56b) became /O˘O/.
Finally, Proto-Okinawa disyllabic nouns underwent Vowel
Lengthening. As mentioned earlier, Nakijin dialect raises the word-
final pitch. Therefore, /_Ooo/ has raised its pitch toward the end of
the word, i.e., /_Ooo/ [LLH(H)]. In addition, in Nakijin there is
another characteristic in pitch shape. In short, unaccented initial
syllable becomes low-pitch. Thus, /`Ooo/ (56a) has LHH(H) pitch.
This phonetic rule is also seen in Tokyo dialect (see chapter 6).
Trisyllables: In Nakijin trisyllabic nouns, there are two sound
changes (Accent Shift and Final Accent Loss) and one phonetic
rule (I-low). First, undergoing Accent Shift, PO */O˘OO/ (57a) and
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 191

*/OO˘O/ (57b) became /OO˘O/ [HHL(L)] and /OOO˘/ [LLH(L)]

respectively. At the second stage, /OOO˘/ was converted into
/OOO/ by means of Final Accent Loss. This caused a merger of the
/OOO/ and the already-existing atonic class, i.e., PO */OOO/ (57c).
Finally, an Initial Pitch Lowering rule lowered the initial pitch of
/OO˘O/ [HHL(L)], resulting in LHL(L).

(57) Development of Nakijin trisyllabic accent

PO shift F-loss I-low

(a) *O˘OO > OO˘O = OO˘O > OO˘O
[HLL(L)] [HHL(L)] [LHL(L)]

(b) *OO˘O > OOO˘

> OOO = OOO
(c) *OOO = OOO [LLL(L)] [LLH(H)]
[LLL(L)]

Regarding the ordering of the two sound changes contributing to

the development of Nakijin trisyllabic accent, Accent Shift has to
precede Final Accent Loss. The phonetic pitch rule could have
occurred before or after Final Accent Loss, but it must have taken
place after Accent Shift.

5.3.4.2 Development of Shuri Accent

Considering the accent system as a whole, Shuri has a complex
history in developing its accent system (although the development
of Nakijin disyllabic accent is the most complex one). In the
development of Shuri accent, there is a series of three changes, but
in trisyllables only two took place because one of the changes is
conditioned (i.e., Initial Accent Gain is applicable only to high-
initial).
From the point of view of accent typology, in the Shuri dialect,
the locus system of Proto-Okinawa accent evolved into a register
system. The shift from locus to register took place when final-
accent nouns lost their accent.
192 ACCENTUAL HISTORY

Monosyllables: The accent system of Shuri monosyllabic nouns

has resulted from undergoing three sound changes: Accent Shift,
Final Accent Loss, and Initial Accent Gain. Accent Shift converted
PO */o˘o/ (58a) to /oo˘/, and the next change (Final Accent Loss)
canceled the accent of the /oo˘/, resulting in /`oo/. Finally, Initial
Accent Gain, applicable only to high-initial, made the /`oo/
[HH(H)] phonetically initial-accent, i.e., /`oo/ [HH(H)] became
/`oo/ [HL(L)].

(58) Development of Shuri monosyllabic accent

PO shift F-loss I-gain

(a) *o˘o > oo˘ > `oo > `oo [HL(L)]
[HH(H)]

(b) *oo = oo ª _oo = _oo [LL(L)]

Shuri monosyllabic accent nouns are phonetically either HL(L)

or LL(L). Instead of analyzing them as initial-accent and low-
atonic accent, I analyze them as high-register and low-register
because of the predictability of their pitch patterns by initial pitch
height, and also because of coherence in my analysis (see 5.3.1.2).
Disyllables: In the Shuri line, Proto-Okinawa disyllables
underwent three sound changes: Accent Shift, Final Accent Loss,
and Initial Accent Gain. By these changes, the Proto-Okinawa
three-way accent system has become a two-way system.
First, undergoing Accent Shift, both PO */O˘O/ (59a) and
*/oo˘O/ (59b) became final-accent: /OO˘/ and /ooO˘/ respectively.
Second, the Final Accent Loss rule eliminated the accent of the
/OO˘/ and the /ooO˘/, and then the latter merged with PO */OO/
[LL(L)] (59c). At this point there are two accent classes. As
indicated below, the low-atonic class includes nouns with initial-
syllable long vowel and ones without it.
Third, Initial Accent Gain took place. This is a conditioned
change; that is, it is applicable only to high-initial nouns. Thus, the
/OO/ (< PO */O˘O/) became phonetically initial-high, but low-
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 193

atonic nouns did not undergo the change. It should be pointed out
that Initial Accent Gain converted /`OO/ into initial-accent at the
phonetic level – no effect at the phonemic level.

(59) Development of Shuri disyllabic accent

PO shift F-loss I-gain

(a) *O˘O > OO˘ > `OO > `OO
[HH(H)] [HL(L)]

(b) *oo˘O > ooO˘

[LH(L)]
> _ooO/_OO = _ooO/_OO
(c) *OO = OO [LLL(L)]
[LL(L)]

In the Shuri line, by losing locus distinctions, the accent system

has become simpler than that of Proto-Okinawa for disyllabic
nouns. Typologically speaking, Proto-Okinawa only had locus
distinctions, but the locus became nondistinctive when PO */oo˘O/
lost its accent, resulting in Shuri developing a register system.
In the Shuri dialect, most of the Proto-Okinawa initial vowel
length has been retained. Exceptions are _/umi (< PO */u:˘mi)
‘sea’, _huni (< PO *phu:˘ni) ‘boat’, and _huni (< PO *phu:˘ni)
‘bone’ (see also (51)). I assume PO */ooO/ (59b) sporadically
underwent Vowel Shortening in the Shuri dialect.
Trisyllables: Although its monosyllables and disyllables
underwent three sound changes, Shuri’s trisyllables underwent only
two: Accent Shift and Final Accent Loss. This is because the third
change (i.e., Initial Accent Gain) is not applicable. This rule is
applicable only to high-initial nouns. Therefore, it had no effect on
PO */OOO/ [LLL(L)].
In the development, Accent Shift shifted the accent of PO
*/O˘OO/ (60a) and PO */OO˘O/ (60b) onto the following syllable,
resulting in /OO˘O/ and /OOO˘/ respectively. PO */OOO/ (60c)
remained the same. At the next stage, Final Accent Loss eliminated
194 ACCENTUAL HISTORY

the accent of the /OOO˘/. As a result, it merged with PO */OOO/

(60c), becoming /_OOO/ in contrast with /`OOO/ (60a).

(60) Shuri trisyllabic accent

PO shift F-loss
(a) *O˘OO > OO˘O ª `OOO
[HLL(L)] [HHL(L)]

(b) *OO˘O > OOO˘

[HHL(L)]
> _OOO
(c) *OOO = OOO [LLL(L)]
[LLL(L)]

As shown in (60), I analyzed the Shuri /OO˘O/ and /OOO/ as

initial-high register and initial-low register respectively, because of
the reason mentioned above (see the description of Shuri
monosyllables and disyllables and see also 5.3.1.2).

5.3.4.3 Development of Aguni Accent

Among the Okinawa dialects in this book, Aguni accent is the most
similar to the Proto-Okinawa accent. Aguni monosyllabic and
trisyllabic accent did not undergo any change at all. That is, both
types of nouns are identical to those of Proto-Okinawa. As for
Aguni disyllabic accent, compared with other Okinawa dialects, it
underwent rather simple development.
Monosyllables: In Aguni monosyllabic nouns, there are two
accent classes: initial accent and low atonic, which are identical
both phonemically and phonetically to the corresponding accent
classes of Proto-Okinawa monosyllabic nouns. In other words,
Proto-Okinawa monosyllabic accent has not changed at all in the
Aguni dialect.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 195

(61) Development of Aguni monosyllabic accent

PO
(a) *o˘o [HL(L)] = o˘o [HL(L)]
(b) *oo [LL(L)] = oo [LL(L)]

Disyllables: In the development of Aguni disyllabic accent, PO

underwent Accent Shift and Vowel Shortening. First, Accent Shift
shifted the accent of PO */oo˘O/ (62b) onto the final syllable. Since
this rule is applicable only to low-initial nouns, it had no effect on
PO */O˘O/ (62a). At the following stage, by the vowel shortening
rule, /ooO˘/ became /OO˘/. In fact, not all Proto-Okinawa nouns in
this class underwent the change. The Proto-Okinawa nouns
*phu:˘ni ‘boat’, */u:˘mi ‘sea’, and *mu:˘ku ‘bridegroom’ became
huni˘, umi˘, and muku˘ respectively, but other Proto-Okinawa
nouns such as */u:˘si ‘mortar’, *pha:˘si ‘chopsticks’, and *pha:˘i
‘needle’ did not undergo the change. At present, I do not know why
some words did not undergo the change and others did.39

(62) Development of Aguni disyllabic accent

PO shift V-shrt
(a) *O˘O = O˘O = O˘O

(b) *oo˘O > ooO˘ = OO˘

[LHL(L)]

(c) *OO = OO = OO
[LL(L)]

39
A possible analysis on this is that disyllabic nouns with high vowels in both
syllables underwent the change. However, the word PO *_/u:˘si has high vowels
in both syllables. This may be considered as an exception. Another possible
explanation is that they are loans from the Shuri dialect. Without enough data, it is
difficult to draw a conclusion. I need to investigate this matter more in my future
studies.
196 ACCENTUAL HISTORY

Trisyllables: As shown below, the accent system of Proto-

Okinawa trisyllabic nouns and that of Aguni trisyllabic nouns are
identical.

(63) Development of Aguni trisyllabic accent

PO
(a) *O˘OO = O˘OO
[HLL(L)]

(b) *OO˘O = OO˘O

[HHL(L)]

(c) *OOO = OOO

In Aguni disyllables, Accent Shift took place, but it did not

occur in monosyllables and trisyllables. The reason is that the rule
is applicable only to low-initial nouns. In the disyllables and
trisyllables, all accented nouns are high-initial, and thus it did not
apply.
According to my reconstruction of Okinawa Ryukyuan accent,
the accent system of the Aguni dialect preserves most of the Proto-
Okinawa accentuation, and the Nakijin dialect preserves the
characteristics of the Proto-Okinawa accent least. Like the Aguni
dialect, the Shuri dialect has kept Proto-Okinawa monosyllabic
accent, but Proto-Okinawa disyllabic and trisyllabic nouns
underwent changes.

5.4 MIYAKO RYUKYUAN

5.4.1 Description of Miyako Ryukyuan Accent
In this section, I will compare the Ikema, Ôura, and Uechi dialects
of Miyako Ryukyuan and reconstruct Proto-Miyako (PM) accent.
To reflect the characteristics of Miyako accent, the three dialects
are carefully chosen from available data according to the
complexity of their accent systems. All Miyako data are taken from
Hirayama et al. (1967).
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 197

5.4.1.1 Ikema Dialect

The Ikema dialect is spoken on the island of Ikema, which is
located approximately 1.5 km to the northwest of Miyako-island.
The accent system of the Ikema dialect is a simple two-way
system, namely a system with tonic-atonic distinctions. Ikema tonic
monosyllables, disyllables, and trisyllables behave in the same way
when followed by a particle (pitch falls between the final mora and
a following particle). However, in isolation the monosyllables are
high-level, the disyllables are LH, and the trisyllables are LHL. In
Ikema, there is a phonetic rule which lowers the pitch of the initial
syllable if it is not accented, and also if it consists of one mora.
The other accentual class in Ikema is atonic. This is phonetically
low level in isolation, but high-level when followed by a particle.
Monosyllables: In Ikema, all monosyllabic nouns have two
moras; their word structure is either CV: or (C)Vn. There are two
types of accentuation; tonic and atonic. The former type of nouns
are phonetically LH(L) when followed by a particle, but they are
high-level pitch in isolation.

(64) Ikema monosyllabic accent

Phonemic Phonetic Myô List of nouns
oo˘ HH ~ 1.1 ki:˘ ‘hair’, hu:˘ ‘sail’, Qfa˘ ‘child’
LH(L)40 1.2 ha:˘ ‘leaf’, na:˘ ‘name’
1.3 ki:˘ ‘tree’, mi:˘ ‘eye’, ha:˘ ‘tooth’
2.3 in˘ ‘dog’, hmmu˘41 ‘cloud’
oo LL ~ 2.4 in ‘sea’
HH(H)

40
To describe the accent of Miyako dialects, Hirayama et al. (1967) use a
nominative particle nudu, which is comparable to the particle nu in Okinawa
Ryukyuan. Since pitch does not change within the particle, for convenience I use
(H) to indicate that the following particle is high-pitch and (L) to show that the
particle is low-pitch., even though the particle consists of two syllables.
41
/m/ preceded by /h/ becomes voiceless, e.g., /hmmu/ [m8:u] ‘cloud’, /hnna/ [n8:a]
‘rope’ (Hirayama et al. 1967: 105). Although PJ */ku/ and */pu/ normally
correspond to /fu/ in Miyako Ryukyuan, the word for ‘cloud’ is /hmmu/ in Ikema
< *fu8mu.
198 ACCENTUAL HISTORY

The form in˘ ‘dog’ (2.3) is derived from *inu; it lost its final
vowel and /n/ became moraic. The word for ‘sea’ is identical to the
form for ‘dog’, although their accentuation differs.
The word hmmu˘ ‘cloud’ (2.3) is another historically-disyllabic
noun. The loss of a vowel in the initial syllable led to the
compensatory lengthening of the following /m/, cf. kumo in
Japanese.
Disyllables: Ikema disyllabic nouns have two types of
accentuation: final accent and atonic. In general, final-accent nouns
are phonetically LH(L), e.g., hana˘ LH(L) ‘nose’ (2.1) for two-
mora nouns. If a word consists of three moras, it is either LHL ~
LHH(L) (e.g., nkadi˘ ‘centipede’) or HHL ~ HHH(L) (e.g., kyu:sï˘
‘smoke’) depending on its internal structure.
Each accentual class contains historical disyllables and
trisyllables. Historical disyllabic nouns are generally in (C)VCV
structure. As for historical trisyllables, there are two types. The first
type consists of two moras (e.g., nada˘ ‘tears’ (3.5)), and the
second is three moras long (e.g., maQfa˘ ‘pillow’ (3.5)).

(65) Ikema disyllabic accent

Phonemic Phonetic Myô List of nouns
OO˘42 LH(L) 2.1 hana˘ ‘nose’, usï˘ ‘cow’, tui˘
HL ‘bird’, kusï˘ ‘loins’, higi˘
‘beard’, kadi˘ ‘wind’
2.2 isï˘ ‘stone’, hasï˘ ‘bridge’
2.3 hana˘ ‘flower’, yama˘
‘mountain’
2.4 icya˘ ‘board’
2.5 ami˘ ‘rain’, asi˘ ‘sweat’
3.5 nada˘ ‘tears’

42
According to Hirayama et al. (1967), nouns with a locus on the final syllable are
either LH(L) or HL ~ LH(L). Depending on their internal structure, those nouns
are realized as LH(L) or HL ~ LH(L) at the phonetic level. Words with LH(L)
pitch have a (C)VCV structure, and ones with HL ~ LH(L) have a structure of CVi,
e.g., kui ‘this’.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 199

LHL ~ 2.2 kabi:˘ ‘paper’

LHH(L) 3.2 nkadi˘ ‘centipede’
3.4 kagan˘ ‘mirror’
HHL ~ 3.1 kyu:sï˘ ’smoke’
HHH(L) 3.5 aQva˘ ‘fat’, maQfa˘ ‘pillow’
3.7 fuQza˘ ‘whale’
OO LL ~ 2.3 puni ‘bone’
HH(H) 2.4 funi ‘boat’, pai ‘needle’, usï
‘mortar’, icï ‘breath’
2.5 kui ‘voice’, kagi ‘shadow’,
muku ‘bridegroom’, wu:ki
‘bucket’
3.5 hara ‘pillar’, nnucï ‘life’
3.7 Qfui ‘medicine’

A comparison of these original trisyllables and their cognate

forms in other Japonic languages shows that the word kagan˘ has
lost its final vowel and that the nasal in the final syllable became
moraic. In the course of development, the words for ‘fat’, ‘pillow’,
and ‘whale’ have lost a syllabic segment and eventually they
geminated the intermediate consonant to compensate for the lost
syllabic segment.

(66) Comparison between Ikema second-mora-accent nouns and

their corresponding Tokyo, Kyoto, and Myôgishô forms43

Ikema Tokyo Kyoto Myôgishô

‘mirror’ (3.4) kagan kagami kagami kagami
‘fat’ (3.5) aQva abura abura afura44
‘pillow’ (3.5) maQfa makura makura makura
‘whale’ (3.7) fuQza kuzira kuzira kuzira

43
Because here my focus is segmental, accent is not indicated.
44
Regarding /f/ in the Myôgishô cognate, since modern Japanese has a /b/ and
since all Ryukyuan dialects point to */b/, the voiceless consonant of the Myôgishô
form is to be taken as simply the tendency of older texts not to always write a
voiced consonant. It is therefore not to be trusted as phonologically literally true.
200 ACCENTUAL HISTORY

There are more nouns which are of trisyllabic origin in my data.

For instance, the words hara ‘pillar’ (3.5), nada˘ ‘tears’ (3.5), and
Qfui ‘medicine’ (3.7), cf. fasira LLH ‘pillar’, namida LLH ‘tears’,
and kusuri LHL ‘medicine’ in Myôgishô (Mochizuki 1974: 190).
The word kabi:˘ ‘paper’ (2.2) has a long vowel in the final
syllable, different from other historical disyllabic nouns. The vowel
length may be secondary. In Japonic languages the cognate form
with ‘paper’ normally has a CVCV structure, except in Nakijin
habi: ‘paper’.45 In Tokyo, Kyoto, and Myôgishô, their
corresponding forms are respectively kami˘ LH(L), ka˘mi HL(L),
and kami LL ‘paper’.
Excluding trisyllabic-origin words, according to the data, the
correspondence between Ikema and Myôgishô disyllabic nouns are
generally as in (67): Ikema atonic nouns to Myôgishô 2.3-5 nouns
and Ikema tonic nouns to 2.1-5 in Myôgishô. Notice that there is an
overlap; some nouns corresponding to Myôgishô 2.3-5 classes are
also /OO˘/ in Ikema.

(67) Correspondence between Ikema and Myôgishô disyllabic

accent
Ikema Myôgishô
/OO˘/ : 2.1-5
/OO/ : 2.3-5

With a detailed examination of this correspondence pattern, I

have found that among those Ikema disyllables, atonic nouns
systematically correspond to Proto-Okinawa cognates with initial-
syllable vowel length (see (68)), and tonic nouns to Proto-Okinawa
cognates without the vowel length (see (69)).

45
In Nakijin, many disyllabic nouns have a long vowel in the final syllable. This
vowel length may also be secondary.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 201

(68) Ikema PO Myôgishô

‘bone’ puni *phu:˘ni 2.3
‘boat’ funi *phu:˘ni 2.4
‘needle’ pai *pha:˘i 2.4
‘mortar’ usï */u:˘si 2.4
‘breath’ icï */i:˘ci 2.4
‘shadow’ kagi *kha:˘gi 2.5
‘bridegroom’ muku *mu:˘khu 2.5
‘bucket’ wu:ki *u:˘khi 2.5

(69) Ikema PO Myôgishô

‘flower’ hana˘ *phana 2.3
‘mountain’ yama˘ *yama 2.3
‘board’ icya˘ */ica 2.4
‘rain’ ami˘ */ami 2.5
‘sweat’ asi˘ */asi 2.5

This implies that in their earlier forms Ikema nouns in (68) might
have been distinguished from ones in (69) by initial-syllable vowel
length. I will discuss this matter below in relation to the vowel
length in other Ryukyuan dialects.
Trisyllables: For Ikema trisyllables there are two accent classes:
final accent and atonic. They are phonetically LHL ~ LHH(L) and
low-level respectively.

(70) Ikema trisyllabic accent

Phonemic Phonetic Myô List of nouns
OOO˘ LHL ~ 3.1 sïrusï˘ ‘mark’, futai˘
LHH(L) ‘forehead’, afuzï˘ ‘yawn’
3.2 futa:cï˘ ‘two’
3.3 kugani˘ ‘gold’
3.4 fukuru˘ ‘bag’, takara˘
‘treasure’
OOO LLL ~ 3.4 auzï ‘fan’
HHH(H) 3.6 unazï ‘eel’
202 ACCENTUAL HISTORY

The word futa:cï˘ ‘two’ (3.2) is trisyllabic, but contains four

moras. As far as my Ikema data are concerned, this is the only word
consisting of four moras. Although it is not known how it came
into existence, the vowel length in the second syllable might have
been secondary. Many of the Miyako and Yaeyama dialects have a
trisyllabic form with a long vowel in the second syllable for the
word meaning ‘two’, e.g., futa:cï ‘two’. However, its cognates in
many other Ryukyuan dialects, Japanese dialects, and Myôgishô do
not show the length. Furthermore, the Northern Ryukyuan word for
‘two’ usually has a disyllabic form with a long vowel in the initial
syllable, i.e., ta:ci; the vowel might have become long after
dropping the original initial syllable fu (cf. futatu˘ in Tokyo and
futa˘tu in Kyoto and futa˘tu HHL ‘two’ in Myôgishô).
In (71), Ryukyuan cognates for ‘one’ through ‘nine’ are listed in
comparison with Tokyo cognates. This suggests that Miyako and
Yaeyama dialects lengthen the penultimate vowel if a word meets
certain templates, namely either /CVCVCV/ or /CVCV/ where
vowels in the initial and final syllables are high vowels. Although
Ryukyuan dialects in general do not show the length for the word
‘five’, there are some exceptions. The Yaeyama dialects of Ôhama
and Sonai have the length: icï:cï in Ôhama and ici:ci in Sonai
(Hirayama et al. 1967: 227).

(71) Correspondences in numerals

Tokyo Shuri Ikema Ôura Ishigaki
‘one’ hito˘tu _ti:ci sïsïcï pit:˘cï pïti:zï
‘two’ futatu˘ `ta:ci futa:cï˘ futa:cï˘ uta:˘zï
‘three’ mittu˘ `mi:ci mi:cï˘ mi:cï˘ mi:˘cï
‘four’ yottu˘ `yu:ci yu:˘cï yu:cï˘ yu:˘cï
‘five’ itu˘tu _icici icïcï icï˘cï icïcï
‘six’ muttu˘ `mu:ci nncï˘ mmcï˘ nn˘cï
‘seven’ nana˘tu _nanaci nanacï nana˘cï nanacï
‘eight’ yattu˘ `ya:ci ya:cï˘ ya:cï˘ ya:˘cï
‘nine’ kokono˘tu _kukunuci Qkunucï˘ kuku˘nucï kukunucï

Turning now to the accentual correspondence between Ikema

and Myôgishô trisyllables, as far as my data are concerned, Ikema
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 203

atonic nouns correspond to Myôgishô classes 3.4 and 3.6, and tonic
nouns correspond to Myôgishô classes 3.1-4. Tonic and atonic
classes in Ikema overlap with class 3.4. My data lack examples for
trisyllabic nouns that correspond to Myôgishô class 3.5. According
to Hirayama et al. (1967), Myôgishô class 3.5 nouns are atonic in
Ikema, e.g., icyufu ‘cousin’ (3.5).

5.4.1.2 Ôura Dialect

The Ôura dialect is spoken in the northern part of Miyako island. It
is 4.5 km from the central city (Hirara-shi) of Miyako-island.
Like Ikema, Ôura accent has a locus system – distinctions are
penultimate accent and final accent. In penultimate-accent nouns,
the pitch falls after the second mora from the end of words. In
final-accent nouns, the pitch falls at the end of words. When
penultimate-accent nouns are followed by a particle, they become
identical to final-accent nouns (see below for details). Ôura also has
a phonetic rule that lowers the pitch of the initial syllable unless it
consists of two moras.
Hirayama et al. (1967) gives two different descriptions using
two different particles: mai, which is equivalent to mo ‘too, as well’
in Tokyo Japanese, and nudu, which is equivalent to ga
(nominative marker) in Tokyo Japanese. Since other Miyako
dialects are described with the particle nudu, I use his Ôura data
with the same particle in order to be consistent in my analysis.
Monosyllables: Ôura monosyllabic nouns distinguish
penultimate and final accent. The former is phonetically HL
without a particle (e.g., pa˘: HL ‘tooth’) and HH(L) with a particle
following (e.g., pa:˘-nudu HH-LL ‘tooth’).
In (72), nouns in bold-faced letters are not on my word list. They
are added from Hirayama et al. (1967) in order to fill in missing
spots in my data so that general patterns of Ôura monosyllabic
accent can be captured.
As the data show, there are three historical disyllables, i.e., pa:˘
‘nose’ (2.1), i˘m ‘sea’ (2.4), and in˘ ‘dog’ (2.3). The word pa:
‘nose’ is historically a disyllabic noun; cf., hana LH(H) in Tokyo
and fana HH ‘nose’ in Myôgishô. It presumably lost the syllable na
204 ACCENTUAL HISTORY

and lengthened the vowel /a/. For the cognates for ‘sea’ and ‘dog’,
see discussions above.
Excluding those historically-disyllabic nouns, the word pa˘:
‘tooth’ (1.3) is the only cognate that is penultimate-accent. Other
monosyllables are final-accent in Ôura. The accentuation of the
word for ‘tooth’ (1.3) must be irregular because as shown in (72),
nouns corresponding to Myôgishô accent class 1.3 are normally
final-accent.

(72) Ôura monosyllabic accent

Phonemic Phonetic Myô List of nouns
o˘o HL(L) 1.3 pa˘: ‘tooth’
2.4 i˘m ‘sea’
oo˘ HH(L) 1.1 pu:˘ ‘sail’, fa:˘ ‘child’
1.2 pa:˘ ‘leaf’, na:˘ ‘name’
1.3 ki:˘ ‘tree’, mi:˘ ‘eye’,46 ti:˘
‘hand’, Qsï˘ ‘nest’, pi:˘ ‘fart’,
yu:˘ ‘hot water’, pa:˘ ‘blade’
2.1 pa:˘ ‘nose’
2.3 in˘ ‘dog’

A syllable-final nasal consonant is moraic. That is to say, the

words such as in ‘dog’ and i˘m ‘sea’ are monosyllabic. They
became monosyllables when the final vowel was dropped, cf. inu˘
‘dog’ and u˘mi ‘sea’ in Tokyo and inu LL ‘dog’, umi LH ‘sea’ in
Myôgishô (Mochizuki 1974: 67 and 94).
Disyllables: There are also penultimate and final accent for Ôura
disyllabic nouns. Final-accent nouns are either LH(L), LHH(L), or
HH(L). They are HH(L) when their word structure is CVV (e.g.,
tuï˘ ‘bird’); otherwise, they are LH(L) for two-mora (e.g. usï˘
‘cow’), and LHH(L) for three-mora nouns (e.g., kiQfu˘ ‘smoke’).
In nouns with penultimate accent, the pitch falls after the

46
Hirayama et al. (1967: 361) also give another form for ‘eye’, namely mi˘: , but
this must be irregular or a typographical error.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 205

penultimate mora in nouns with two moras (e.g., pu˘ni ‘bone’, u:˘ki
‘bucket’).

(73) Ôura disyllabic accent

Phonemic Phonetic Myô List of nouns
O˘O HL ~ 2.3 pu˘ni ‘bone’
LH(L) 2.4 fu˘ni ‘boat’, pi˘ï ‘needle’, u˘sï
‘mortar’, i˘kï ‘breath’
2.5 ku˘i ‘voice’, ka˘gi ‘shadow’,
mu˘ku ‘bridegroom’
HHL(L) 2.5 u:˘ki ‘bucket’
3.4 o:˘gï ‘fan’
3.5 maQ˘fa ‘pillow’, nnu˘cï ‘life’
3.7 Qsu˘ï ‘medicine’
OO˘ LH(L) 2.1 usï˘ ‘cow’, kusï˘ ‘loins’, pïgi˘
‘beard’, kazi˘ ‘wind’
2.2 isï˘ ‘stone’, kabï˘ ‘paper’, pasï˘
‘bridge’
2.3 pana˘ ‘flower’, yama˘
‘mountain’, fumu˘ ‘cloud’
2.4 icya˘ ‘board’
2.5 ami˘ ‘rain’, asi˘ ‘sweat’
3.5 nada˘ ‘tears’, para˘ ‘pillar’
LHH(L) 3.1 kiQfu˘ ‘smoke’
3.2 nkazi˘ ‘centipede’
3.4 kagan˘ ‘mirror’
3.5 aQva˘ ‘fat’47
HH(L) 2.1 tuï˘ ‘bird’

The synchronic analysis of Ôura disyllabic nouns includes not

only original disyllabic nouns but also historical trisyllabic nouns.
The former type is normally composed of two moras, and the latter
consists of three moras. There are some exceptions to both types.

47
There is a discrepancy in Ôura data in Hirayama et al. (1976). On page 70, the
pitch of the word for ‘fat’ is LHH(L), but on page 272 it is HHH(L).
206 ACCENTUAL HISTORY

The noun u:˘ki ‘bucket’ (2.5) is historically-disyllabic, but is

composed of three moras. Furthermore, the words nada˘ ‘tears’
(3.5) and para˘ ‘pillar’ (3.5) are historically-trisyllabic, but they are
two moras long.
On the correspondences between Ôura historically-disyllabic
nouns and their Myôgishô cognates, Ôura final-accent nouns
correspond to Myôgishô 2.1-5 classes, and Ôura penultimate-accent
nouns to Myôgishô classes 2.3-5. There is an overlap. Some of
penultimate-accent and final-accent nouns correspond to Myôgishô
2.3-5 classes. Interestingly, the pattern of the correspondences is
parallel to that of corresponding Ikema cognates. See (74) and (75)
below.

(74) Ôura Ikema

‘bone’ (2.3) pu˘ni puni
‘boat’ (2.4) fu˘ni funi
‘needle’ (2.4) pa˘i pai
‘mortar’ (2.4) u˘sï usï
‘breath’ (2.4) i˘cï icï
‘shadow’ (2.5) ka˘gi kagi
‘bridegroom’ (2.5) mu˘ku muku
‘bucket’ (2.5) wu:˘ki wu:ki

(75) Ôura Ikema

‘flower’ (2.3) pana˘ hana˘
‘mountain’ (2.3) yama˘ yama˘
‘board’ (2.4) icya˘ icya˘
‘rain’ (2.5) ami˘ ami˘
‘sweat’ (2.5) asi˘ asi˘

As shown in (74) and (75), there are two groups of words

corresponding to Myôgishô 2.3-5 classes in both Ôura and Ikema,
and each group contains the same cognates in both Ôura and Ikema.
Furthermore, this phenomenon is also seen in Nakijin and Shuri
Okinawan (see 5.3.3). This distinction is reflected in my
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 207

reconstruction of Proto-Okinawa, Proto-Ryukyuan, and Proto-

Japonic accent.
Trisyllables: In the Ôura trisyllabic nouns, there are two types
of accentuation, i.e., penultimate accent and final accent. Final-
accent nouns are phonetically LHH(L). Penultimate-accent nouns
are phonetically LHL and they are LHH(L) when followed by a
particle. The pitch shape of penultimate nouns is identical to that of
final-accent nouns when used in a phrase. Penultimate nouns with
four moras are HHLL(L).
A word with the initial-syllable long vowel such as mi:ma˘zï
‘earthworm’ has four moras. This vowel length is found in Miyako
and Yaeyama dialects; it is not seen in Amami dialects, Okinawa
dialects, or Mainland Japanese dialects (see Hirayama et al. 1967:
71).

(76) Ôura trisyllabic accent

Phonemic Phonetic Myô List of nouns
OO˘O LHL ~ 3.6 una˘gï ‘eel’
LHH(L) 3.7 kuzi˘ra ‘whale’, pata˘gi
‘field’48
HHLL(L) 3.6 mi:ma˘zï ‘earthworm’
OOO˘ LHH(L) 3.1 sïrusï˘ ‘mark’, futai˘
‘forehead’, afukï˘ ‘yawn’
3.2 futa:cï˘ ‘two’
3.3 cïkara˘ ‘strength’, kugani˘
‘gold’
3.4 fukuru˘ ‘bag’, takara˘
‘treasure’

Just like in Ikema, the word futa:cï˘ ‘two’ may have secondary
vowel length. See the discussion in 5.4.1.1.

48
Hirayama et al. (1967) list two Ôura forms for ‘field’: pata˘gi (page 69) and
pa˘ri (page 309). pata˘gi is cognate with hatake, and pa˘ri is cognate with OJ pari
‘newly opened land for agriculture’.
208 ACCENTUAL HISTORY

Concerning accentual correspondences of Ôura trisyllabic nouns

to their Myôgishô cognates, they are systematic. Ôura final-accent
nouns regularly correspond to 3.1-4 class nouns in Myôgishô, and
the penultimate-accent nouns to Myôgishô 3.5-7 classes. My data
lack nouns corresponding to Myôgishô class 3.5, but the word icï˘fu
‘cousin’ (3.5) indicates that there is a correspondence between
Ôura penultimate-accent trisyllables and Myôgishô class 3.5.

5.4.1.3 Uechi Dialect

The Uechi dialect is spoken in the southwestern area of Miyako
island, approximately 6 km from the central Hirara city.
Typologically speaking, Uechi accent is a locus system. In Uechi
there are three accentual patterns: atonic, penultimate accent, and
final accent. Atonic nouns are low-level throughout the words.
Penultimate-accent nouns have an accent on the penultimate
syllable. For final-accent nouns, the pitch falls word-finally. A
phonetic rule lowers the pitch height of an unaccented initial
syllable. Interestingly, the domains for penultimate-accent and
final-accent nouns differ. Penultimate-accent nouns generally shift
the location of the accent depending on whether they are uttered
with or without an enclitic, e.g., fukuru LHL ‘bag’, fukuru-nudu
LHH-LL ‘bag (Nom.)’. On the other hand, final-accent nouns
normally keep an accent on the final syllable of words, e.g., pana˘
LH(L) ‘flower’.49
Monosyllables: In Uechi, there are two accent classes for
monosyllabic nouns: final accent and atonic. The tonic nouns,
excluding historical disyllables, correspond to their cognates in
Myôgishô classes 1.1 and 1.3, and its atonic nouns to Myôgishô
class 1.2. As far as my word list is concerned, there are no Uechi
nouns corresponding to Myôgishô class 1.1. In order to determine
what accent Uechi 1.1-nouns have and to what Myôgishô class they

49
Final-accent disyllables consisting of three moras are HHL(L). This is
completely predictable by their internal word structure (see the description of
Uechi disyllabic accent in this section).
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 209

correspond, other nouns have been added to the data in (77), which
are not on my original word list. Those words are boldfaced.
My list of monosyllabic nouns also includes historical
disyllables which have become monosyllabic in the course of their
development from earlier forms, e.g., pa: ‘nose’ (2.1), in˘ ‘dog’
(2.3), and im˘ ‘sea’ (2.4).

(77) Uechi monosyllabic accent

Phonemic Phonetic Myô List of nouns
oo LL(L) 1.2 pa: ‘leaf’, na: ‘name’
2.1 pa: ‘nose’
oo˘ HH(L) 1.1 pu:˘ ‘sail’, Qfa˘ ‘child’
1.3 ki:˘ ‘tree’, mi:˘ ‘eye', pa:˘
‘tooth’
2.3 in˘ ‘dog’
2.4 im˘ ‘sea’

The word pa: ‘nose’, originally disyllabic, became monosyllabic,

cf. hana LH ‘nose’ in Tokyo and fana HH ‘nose’ in Myôgishô
(Mochizuki 1974: 438). The word lost the final syllable na and
lengthened the initial-syllable vowel. For ‘dog’ and ‘sea’, see
discussions in 5.2.1.
Disyllables: For Uechi disyllabic nouns, there are also two
accentual classes: final accent and atonic. The phonetic pitch shape
of the tonic nouns with two moras is LH(L), but ones with three
moras are HHL. By contrast, atonic nouns are simply low-level.
Just like Ikema and Ôura nouns meaning ‘pillar’ (3.5) and ‘tears’
(3.5), in Uechi the corresponding nouns also became disyllabic (see
5.4.1.1 and 5.4.1.2). They are two moras long, different from other
historically-trisyllabic nouns that have become three-mora
disyllabic.
There are words that could be treated as if they were
monosyllabic, such as tuï ‘bird’ (2.1), kui˘ ‘voice’ (2.5), piï˘
‘needle’ (2.4). I consider these words as disyllabic. For ‘bird’ and
‘voice’, see discussions in 5.4.1.1 above. I also treat the word piï
210 ACCENTUAL HISTORY

‘needle’ as disyllabic. There are two reasons why I treat these as

disyllabic. First, the word contains a rearticulated vowel; i.e., the
word is not CV: or CVn, which is a typical Uechi monosyllabic
form. Second, the word is not only disyllabic in modern Japonic
languages, but also historically is disyllabic, e.g., ha˘ri ‘needle’ in
Tokyo and fari LH ‘needle’ in Myôgishô (Mochizuki 1974: 447).

(78) Uechi disyllabic accent

Phonemic Phonetic Myô List of nouns
OO LL(L) 2.1 usï ‘cow’, tuï ‘bird’, kusï ‘loins’,
pïgi ‘beard’, kazi ‘wind’
2.2 isï ‘stone’, kabï ‘paper’, pasï
‘bridge’
3.1 kiQfu ‘smoke’
3.2 mmkazi ‘centipede’
3.4 kagam ‘mirror’
OO˘ LH(L) 2.3 pana˘ ‘flower’, yama˘
‘mountain’, fumu˘ ‘cloud’,
puni˘ ‘bone’
2.4 funi˘ ‘boat’, piï˘ ‘needle’, usï˘
‘mortar’, ikï˘ ‘breath’, icya˘
‘board’
2.5 kui˘ ‘voice’, kagi˘ ‘shadow’,
muku˘ ‘bridegroom’, ami˘
‘rain’, asi˘50 ‘sweat’
3.5 para˘ ‘pillar’, nada˘ ‘tears’
HHL 2.5 u:ki˘ ‘bucket’
3.4 wa:gï˘ ‘fan’
3.5 aQva˘ ‘fat’, maQfa˘ ‘pillow’,
mnucï˘ ‘life’

50
According to Hirayama et al. (1967), it is as, which is irregular in comparison
with other disyllabic nouns in the language because it lacks a final vowel.
However, it may be asï8 where the final vowel is devoiced; thus, it sounds like
there is no vowel at the end.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 211

Words with a heavy syllable containing a geminate consonant

/Q/, moraic nasal, or vowel length are also treated as disyllabic, e.g.,
kiQfu ‘smoke’ (3.1), kagam ‘mirror’ (3.4), mmkazi ‘centipede’ (3.2),
wa:˘gï ‘fan’ (3.4), aQ˘va ‘fat’ (3.5), maQ˘fa ‘pillow’ (3.5), and
mnu˘cï ‘life’ (3.5). Moreover, all of those nouns are attested as
trisyllabic in other Japonic cognates including Myôgishô. It is likely
the case that the nouns have lost a segment or segments in the
course of the development of their accentuation, and some
resyllabified a nasal consonant as moraic and others lengthened a
vowel or consonant in order to compensate the loss of (a)
segment(s). This phenomenon is commonly seen in Ryukyuan
dialects.
With the exclusion of historically-trisyllabic nouns, a
comparison of Uechi trisyllabic nouns with their Myôgishô
cognates reveals that atonic disyllables regularly correspond to
Myôgishô classes 2.1-2, and tonic nouns to Myôgishô classes 2.3-5.
Trisyllables: In Uechi trisyllabic nouns, there are also two
accentuation types: tonic and atonic. The former type of nouns are
mostly LHL ~ LHH(L), except the noun mi:mizï˘ HHHL(L)
‘earthworm’ (3.6), where the initial syllable contains two moras.
Similar to other Miyako dialects, Uechi dialect has a phonetic rule
that lowers the initial pitch of nouns, if the syllable does not carry a
locus and if the syllable does not contain two moras. As for atonic
nouns, they are phonetically low-level pitch.

(79) Uechi trisyllabic accent

Phonemic Phonetic Myô List of nouns
OOO LLL(L) 3.1 sïrusï ‘mark’, futai ‘forehead’,
afugï ‘yawn’
3.2 futa:cï ‘two’
3.3 kugani ‘gold’
OOO˘ LHL ~ 3.4 fukuru˘ ‘bag’
LHH(L) 3.6 unagï˘ ‘eel’
3.7 fusuï˘ ‘drug’, kuzira˘ ‘whale’
HHHL 3.6 mi:mizï˘ ‘earthworm’
212 ACCENTUAL HISTORY

With the exception of the word mi:mizï˘ ‘earthworm’, Uechi

trisyllabic nouns are generally (C)VCV(C)V in structure. A
corresponding form for ‘earthworm’ in other Japonic dialects and
in Myôgishô is CVCVCV in structure. The initial-syllable vowel
length might be secondary.
Comparing Uechi trisyllabic nouns with corresponding
Myôgishô nouns, Uechi atonic nouns correspond to accent classes
3.1-3 in Myôgishô, and tonic nouns correspond to Myôgishô classes
3.4 and 3.6-7. Based on the data above, it is not possible to know
which class Uechi nouns correspond to Myôgishô class 3.5 belong
because the data simply do not have these words. However,
according to Hirayama et al. (1967: 258), the noun icufu˘ ‘cousin’
(3.5) is tonic. Hence, it can be stated that Uechi trisyllabic tonic
nouns correspond to Myôgishô classes 3.4-7.

5.4.2 Correspondences and Reconstruction of Proto-Miyako

Accent
Based on the descriptions of the Miyako dialects in the previous
section, this section will present all correspondences among the
dialects and reconstruct Proto-Miyako (PM) accent classes for
monosyllabic, disyllabic, and trisyllabic nouns. Proto-Miyako
nouns belonging to each class will be also reconstructed.
Monosyllables: There are two sets of accentual correspondences
among Ikema, Ôura, and Uechi monosyllabic nouns. On the basis
of the correspondences, I have reconstructed low-register final
accent and high-register final accent for Proto-Miyako
monosyllabic nouns. As far as accent typology is concerned, Proto-
Miyako accent is a locus-register system as a whole. However, for
Proto-Miyako monosyllables, only register is distinctive.
Correspondences between Miyako and Myôgishô accent are
regular. However, the correspondences in (80) show an interesting
phenomenon. Unlike Amami and Okinawa monosyllables, Miyako
nouns corresponding to Myôgishô accent 1.1 and 1.3 are in the
same correspondence set (80a), and Miyako nouns that correpond
to Myôgishô accent 1.2 are independently in (80b).
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 213

(80) Reconstructed Proto-Miyako accent for monosyllabic nouns

(a)
Ikema Ôura Uechi PM
oo˘ oo˘ oo˘ *_oo˘
LH(L) HH(L) HH(L) RH(L)
‘sail’ (1.1) hu:˘ pu:˘ pu:˘ *_pu:˘
‘child’(1.1) Qfa˘ fa:˘ Qfa˘ *_fa:˘
‘hair’ (1.1) ki:˘ -- -- *_ki:˘
‘tree’ (1.3) ki:˘ ki:˘ ki:˘ *_ki:˘
‘eye’ (1.3) mi:˘ mi:˘ mi:˘ *_mi:˘
‘tooth’ (1.3) ha:˘ pa˘: pa:˘ *_pa:˘
‘dog’ (2.3) in˘ in˘ in˘ *_in˘
‘sea’ (2.4) in i˘m im˘ *_in˘

(b)
Ikema Ôura Uechi PM
oo˘ oo˘ oo *`oo˘
LH(L) HH(L) LL(L) HH(L)
‘leaf’ (1.2) ha:˘ pa:˘ pa: *`pa:˘
‘name’ (1.2) na:˘ na:˘ na: *`na:˘

In the correspondence sets, there are some irregular accentual

patterns. The accentuation of Ôura nouns for ‘tooth’ and ‘sea’ in
(80a) is irregular; these two are initial-accent, whereas others are
final-accent in this correspondence set.
As discussed earlier, in my original word list there are no
cognates that correspond to Myôgishô 1.1 nouns. Therefore, other
1.1-nouns such as ‘sail’ and ‘child’ are added in order to
reconstruct Proto-Miyako nouns that correspond to Myôgishô
accent class 1.1. Moreover, PM *_ki˘: ‘hair’ is reconstructed
entirely based on the Ikema form, ki:˘ because Ôura and Uechi lack
equivalent cognates.
Miyako form for ‘dog’ (2.3), i.e., in˘ (Ikema), in (Ôura), and in˘
(Uechi) became monosyllabic after losing a syllabic segment (cf.
inu ’dog’ in Tokyo).
Disyllables: As far as the data are concerned, for Miyako
disyllabic nouns there are three sets of correspondences; therefore,
for Proto-Miyako disyllabic nouns three accent classes can be
214 ACCENTUAL HISTORY

reconstructed. The following is an overview of accentual

correspondences among Miyako disyllables along with my
reconstructed forms.

(81) Overview of disyllabic accent correspondences in Miyako

dialects
Ikema Ôura Uechi PM
a OO˘ OO˘ OO *`OO˘
LH(L) LH(L) LL(L) HH(L)
b OO˘ *_OO˘
LH(L) LH(L)
c OO O˘O *OO
LL(L) ~ HH(H) HL ~ LH(L) LL(L)

Based on the correspondences, the Proto-Miyako classes that I

have reconstructed are high-register tonic, low-register tonic, and
low atonic. Reconstructed nouns for each accent class are listed
under (82) below.
As in (82a), PM *`pana˘ ‘nose’ is reconstructed because the
Ikema form has retained the segment that Ôura and Uechi have lost.

(82) Reconstructed Proto-Miyako accent for disyllabic nouns

(a)
Ikema Ôura Uechi PM
OO˘ OO˘ OO *`OO˘
LH(L) LH(L) LL(L) HH(L)
‘nose’ (2.1) hana˘ pa:˘ pa: *`pana˘
‘cow’ (2.1) usï˘ usï˘ usï *`usï˘
‘bird’ (2.1) tui˘ tuï˘ tuï *`tuï˘
‘loins’ (2.1) kusï˘ kusï˘ kusï *`kusï˘
‘beard’ (2.1) higi˘ pïgi˘ pïgi *`pïgi˘
‘wind’ (2.1) kadi˘ kazi˘ kazi *`kadi˘
‘stone’ (2.2) isï˘ isï˘ isï *`isï˘
‘paper’ (2.2) kabi:˘ kabï˘ kabï *`kabï:˘
‘bridge’ (2.2) hasï˘ pasï˘ pasï *`pasï˘
‘centipede’ (3.2) nkadi˘ nkazi˘ mmkazi *`mmkazi˘
‘mirror’ (3.4) kagan˘ kagan˘ kagam *`kagam˘
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 215

(b)
Ikema Ôura Uechi PM
OO˘ OO˘ OO˘ *_OO˘
LH(L) LH(L) LH(L) LH(L)
‘flower’ (2.3) hana˘ pana˘ pana˘ *_pana˘
‘mountain’ (2.3) yama˘ yama˘ yama˘ *_yama˘
‘cloud’ (2.3) hmmu˘ fumu˘ fumu˘ *_fumu˘
‘board’ (2.4) icya˘ icya˘ icya˘ *_icya˘
‘rain’ (2.5) ami˘ ami˘ ami˘ *_ami˘
‘sweat’ (2.5) asi˘ asi˘ asi˘ *_asi˘
‘calendar’ (3.4) kuyun˘ kuyun˘ kuyum˘ *_kuyum˘
‘tears’ (3.5) nada˘ nada˘ nada˘ *_nada˘
‘fat’ (3.5) aQva˘ aQva˘ aQva˘ *_aQva˘
‘pillow’ (3.5) maQfa˘ maQfa˘ maQfa˘ *_maQfa˘

(c)
Ikema Ôura Uechi PM
OO O˘O OO˘ *OO
LL ~ HL ~ LH(L) LL(L)
HH(H) LH(L)
‘bone’(2.3) puni pu˘ni puni˘ *puni
‘boat’ (2.4) funi fu˘ni funi˘ *paï
‘needle’ (2.4) pai pi˘ï piï˘ *paï
‘mortar’ (2.4) usï u˘sï usï˘ *usï
‘breath’ (2.4) icï i˘kï ikï˘ *ikï
‘voice’ (2.5) kui ku˘i kui˘ *kui
‘shadow’ (2.5) kagi ka˘gi kagi˘ *kagi
‘bridegroom’ (2.5) muku mu˘ku muku˘ *muku
‘bucket’ (2.5) wu:ki u:˘ki u:ki˘ *wu:ki
‘pillar’ (3.5) hara para˘ para˘ *para
‘life’ (3.5) nnucï nnu˘cï mnucï˘ *nnucï

My reconstruction of disyllabic nouns includes not only

historically disyllabic nouns but also historically trisyllabic nouns
such as ‘pillow’, ‘pillar’, ‘life’, ‘tears’, ‘fat’, ‘mirror’, and
‘centipede’. On the basis of the data, these were already disyllabic
at the time of Proto-Miyako. Among these historically trisyllabic
nouns which have become disyllabic in Proto-Miyako, the word for
‘pillow’ shows an irregular correspondence; maQ˘fa in Ôura (82b).
216 ACCENTUAL HISTORY

Trisyllables: For Proto-Miyako trisyllabic nouns, three accent

classes are reconstructed on the basis of the correspondence sets
shown in (83).

(83) Summary of accentual correspondences and reconstruction of

Proto-Miyako accent for trisyllables
Ikema Ôura Uechi PM
a OO˘O OOO˘ OOO *`OOO˘
LHL ~ LHH(L) LHH(L) LLL(L) HHH(L)
b OOO˘ *_OOO˘
LHL ~ LHH(L)
c OOO OO˘O LHH(L) *OOO
LLL ~ HHH(H) LHL ~ LHH(L) LLL(L)

Reconstructed accent classes are high-register tonic, low-register

tonic, and a class with falling accent. Reconstructed trisyllabic
nouns for each class are listed in (84).
As for cognates for ‘fan’ in (84b), Ikema has presented the
trisyllabic structure, while Ôura and Uechi became disyllabic.
Notice also that the accentuation of Ikema and Ôura cognates is
aberrant; they differ from the accentuation of other 3.4-nouns in
their correspondence set.
In (84c), there is no Ikema cognate for ‘earthworm’. Hirayama et
al. (1967) list zïmizï for ‘earthworm’, however the initial consonant
/z/ corresponds to an /m/ in both Ôura and Uechi, which is not a
regular correspondence because an /m/ in both Ôura and Uechi
regularly corresponds to /m/ in Ikema. Therefore, the word zïmizï
probably is not cognate with the Japonic word mimizu ‘earthworm’.
It is possibly a contamination from zïï ‘earth, dirt’.
In Ikema and Ôura, the cognate for ‘medicine’ listed under (84a)
is disyllabic. Only Uechi has kept the original trisyllabic form. The
Ikema noun for ‘whale’ is also disyllabic and its accentuation is
irregular.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 217

(84) Reconstructed Proto-Miyako accent for trisyllabic nouns

(a)
Ikeam Ôura Uechi PM
OOO˘ OOO˘ OOO *`OOO˘
LHL ~ LHH(L) LLL(L) HHH(L)
LHH(L)
‘mark’ (3.1) sïrusï˘ sïrusï˘ sïrusï *`sïrusï˘
‘forehead’ (3.1) futai˘ futai˘ futai *`futai˘
‘smoke’ (3.1) kyu:sï˘ kiQfu˘ kiQfu *`kifusï˘
‘yawn’ (3.1) afuzï˘ afukï˘ afugï *`afugï˘
‘two’ (3.2) futa:cï˘ futa:cï˘ futa:cï *`futa:cï˘
‘gold’ (3.3) kugani˘ kugani˘ kugani *`kugani˘

(b)
Ikema Ôura Uechi PM
OOO˘ OOO˘ OOO˘ *_OOO˘
LHL ~ LHH(L) LHL ~ LHH(L)
LHH(L) LHH(L)
‘bag’ (3.4) fukuru˘ fukuru˘ fukuru˘ *_fukuru˘
‘treasure’ (3.4) takara˘ takara˘ -- *_takara˘
‘fan’ (3.4) auzï o:˘gï wa:gï˘ *_augï˘

(c)
Ikeam Ôura Uechi PM
OOO OO˘O OOO˘ *OOO
LLL ~ LHL ~ LHL ~ LLL(L)
HHH(H)] LHH(L) LHH(L)
‘cousin’ (3.5) icyufu icï˘fu icufu˘ *icufu
‘eel’ (3.6) unazï una˘gï unagï˘ *unagï
‘earthworm’ (3.6) -- mi:ma˘zï mi:mizï˘ *mi:mizï
‘medicine’ (3.7) Qfui Qsu˘ï fusuï˘ *fusuï
‘whale’ (3.7) fuQza˘ kuzi˘ra kuzira˘ *kuzira
‘field’ (3.7) -- pata˘gi -- *patagi

Regarding the correspondence for ‘field’ in (84c), Ikema and

Uechi dialects do not have a cognate word; the Ikema form for
‘field’ is hai and the Uechi form is pari˘, which are not cognate
with the Ôura form pata˘gi. Only Ôura has retained a Japonic
218 ACCENTUAL HISTORY

cognate form. Thus, the Proto-Miyako word for ‘field’ is entirely

based on the Ôura form.
It should be noted that PM *`kifusï˘ for ‘smoke’ is reconstructed
even though all corresponding cognates are three-mora disyllabic.
This trisyllabic form is reconstructable on the basis of the
correspondence because the Ikema form has kept the final syllable
(i.e., sï) that Ôura and Uechi lost.
The word meaning ‘treasure’ (3.4) is not available for the Uechi
dialect, but available for Ikema (takara˘) and Ôura (takara˘).
Without an Uechi cognate, it seems difficult to determine whether
this cognate belongs to a correspondence in (84b) or (84c), since I
do not know the accentuation of the Uechi cognate. However, on
the basis of accentual correspondence patterns in the data, it is
reasonable for us to assume that the Uechi cognate for ‘treasure’ be
high-register, unless it has irregular accent, because another Uechi
noun corresponding to Myôgishô accent 3.4 (e.g., `fukuru ‘bag’) is
high-register accent. Furthermore, Miyako nouns corresponding to
Myôgishô accent 3.4 must have developed from the same proto
accent form. the following correspondences are excluded because
some items are compounds, non-cognate with other Japonic words,
or unavailable in the data.

(85) Indeterminable correspondences

Ikema Ôura Uechi

‘hair’ ki:˘ pïgi pïgi
‘blood’ aka˘cï ha:˘cï akacï
‘chopsticks’ umasï m:ya˘:sï umi˘sï
‘shoulder’ kata˘musï: kata˘musï kata˘musï
‘strength’ taya cïkara taya˘

The form pïgi in Ôura and Uechi originally meant ‘beard’, but
semantic shift occurred and it became ‘hair’, cf. fige HH 'beard' in
Myôgishô. Words for ‘blood’, ‘chopsticks’, and ‘shoulder’ are
compounds in Ikema, Ôura, and Uechi. In the three dialects, the
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 219

form for ‘blood’ consists of an adjective aka ‘red’ and a noun cï

‘blood’. The form for ‘chopsticks’ is made up of an honorific
morpheme plus a noun for ‘chopsticks’. The word for ‘shoulder’ is
also composed of two morphemes, i.e., kata ‘shoulder’ and musï ‘?’.
Because these are compounds, I exclude them from my
reconstruction.
In Miyako, there are two different forms for ‘strength’: taya and
cïkara. They are not cognate. The form cïkara is cognate with other
Japonic words, but the form taya is not. The origin of taya is
unknown, as well as how it entered the language of Miyako..

5.4.3 Development of Modern Miyako from Proto-Miyako

As summarized below, not only is locality of accent distinctive but
also register is distinctive in Proto-Miyako accent. In this section, I
will account for the development of modern Miyako Ryukyuan
accent systems from the Proto-Miyako accent system.

(86) Proto-Miyako accent

Monosyllables Disyllables Trisyllables
*_oo˘ RH(L) *`OO˘ HH(L) *`OOO˘ HHH(L)
*`oo˘ HH(L) *_OO˘ LH(L) *_OOO˘ LHH(L)
*OO LL(L) *OOO LLL(L)

5.4.3.1 Development of Ikema Accent

In the development of Ikema accent, Abductive Change played a
crucial role. Due to the occurrence of this change, in the Ikema line
Proto-Miyako register lost its distinction. Eventually the Ikema
accent system makes use of a tonic-atonic distinction, except that
monosyllables do not have distinctive accent classes.
Monosyllables: In the Ikema line, Proto-Miyako monosyllabic
accent classes merged into one as /oo˘/ when word-initial pitch
became no longer distinctive to the speakers. Interestingly, the
pitch shape of the merged class reflects the pitch shapes of Proto-
Miyako accent classes, i.e., the pitch is HH in isolation, but it
becomes LH when uttered with an enclitic.
220 ACCENTUAL HISTORY

(87) Development of Ikema monosyllabic accent

abductive
PM change
(a) *_oo˘
> oo˘ [HH ~ LH(L)]
(b) *`oo˘

Regarding Abductive Change, it causes a merger of accent

classes that are phonetically similar. I consider the change natural
because it is phonetically driven.
Disyllables: The development of Ikema disyllabic accent is very
simple. It evolved from Proto-Miyako by undergoing only
Abductive Change. Like its monosyllables, PM */`OO˘/ (88a) and
*/_OO˘/ (88b) merged as /OO˘/ [LH(L)] when the register
distinction was lost. The pitch shape of /OO˘/ is LH(L) except for
nouns with two moras in the initial syllable, where the initial
syllable is high pitch (e.g., kyu:sï˘ HHH(L) ’smoke’). Regarding
PM */OO/ (88c), phonemically it has remained the same. However,
at the phonetic level, the pitch has become high-level when uttered
with a particle; it is low-level without a particle.

(88) Development of Ikema disyllabic accent

abductive
PM change
(a) *`OO˘
> OO˘ [LH(L)]
(b) *_OO˘

(c) *OO = OO [LL ~ HH(H)]

Trisyllables: The development of Ikema trisyllabic nouns

involves Abductive Change, assimilation, and Contour Pitch
Simplification. At the first stage, with the loss of the register
distinction, PM */`OOO˘/ (89a) and */_OOO˘/ (89b) became one,
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 221

namely /OOO˘/ [LHH(L)]. By means of the second change, the

pitch of the class /OOO˘/ [LHH(L)] became LHF ~ LHH(L), then
the application of Contour Pitch Simplification turned /OOO˘/
[LHF ~ LHH(L)] into /OO˘O/ [LHL ~ LHH(L)]. Notice that PM
*/OOO/ (89c) did not undergo any change.

(89) Development of Ikema trisyllabic accent

abductive
PM change assml C-smpl
(a) *`OOO˘
> OOO˘ > OOO˘ > OO˘O
(b) *_OOO˘ [LHH(L)] [LHF ~ LHH(L)] [LHL ~
LHH(L)]

(c) *OOO = OOO = OOO = OOO

Assimilation played a role in the development of Ikema

trisyllabic accent. However, it did not affect monosyllabic and
disyllabic accent, because the conditions for the rule were not met.
Assimilation takes place when there are at least two high pitches at
the end of a lexical word. Because of the pitch shape of /OO˘/
(88ab) (i.e., LH(L)), assimilation had no effect on it.

5.4.3.2 Development of Ôura Accent

Like Ikema, Ôura also lost the Proto-Miyako register distinction by
Abductive Change. In the development of Ôura accent, there are a
total of four changes. They are either phonetic or phonemic.
Monosyllables underwent one, disyllables underwent two, and
trisyllables went through three changes. I will describe how those
changes occurred below.
Monosyllables: Ôura monosyllabic accent resulted from Proto-
Miyako by undergoing Abductive Change. Since Proto-Miyako,
this is the only change that took place in the development of the
Ôura monosyllabic accent system. As a result of the change, the
two accent classes of Proto-Miyako merged as /oo˘/ [HH(L)].
222 ACCENTUAL HISTORY

(90) Development of Ôura monosyllabic accent

abductive
PM change
(a) *_oo˘ [RH(L)]
> oo˘ [HH(L)]
(b) *`oo˘ [HH(L)]

Disyllables: There are two crucial stages in the development of

Ôura disyllabic accent from Proto-Miyako accent. At the first stage,
PM */`OO˘/ (91a) and */_OO˘/ (91b) merged into one as /OO˘/ by
means of Abductive Change. Because of the change, in the Ôura
line the register distinction of Proto-Miyako was lost. At the second
stage, PM */OO/ (91c) gained an accent on the initial syllable.
I should point out that the phonetic pitch shape of /O˘O/ is HL ~
LH(L); the form keeps an accent on the penultimate syllable. Due
to this, the penultimate-accent nouns have the same accentuation as
final-accent nouns when uttered with enclitics.

(91) Development of Ôura disyllabic accent

abductive
PM change I-gain
(a) *`OO˘
> OO˘ = OO˘
(b) *_OO˘ [LH(L)]

(c) *OO = OO > O˘O

[HL ~ LH(L)]

Trisyllables: Like monosyllables and disyllables, Proto-Miyako

trisyllables also underwent Abductive Change, which put PM
*/`OOO˘/ (92a) and */_OOO˘/ (92b) into one class, i.e., /OOO˘/.
At this point, the Proto-Miyako register distinction was lost. The
next change (Initial Accent Gain) changed PM */OOO/ (92c) to
/O˘OO/. Finally, Accent Shift altered the accent of /O˘OO/ to the
following syllable.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 223

(92) Development of Ôura trisyllabic accent

abductive
PM change I-gain shift
(a) *`OOO˘
> OOO˘ = OOO˘ = OOO˘
(b) *_OOO˘ [LHH(L)]

(c) *OOO = OOO > O˘OO > OO˘O

With regard to Accent Shift, this caused a change in trisyllables,

but there was no effect on disyllables. This may be due to a high
vowel in the second syllable; i.e., the vowel prevented the accent
from shifting.

5.4.3.3 Development of Uechi Accent

In the history of Uechi Accent, four types of changes played roles.
Those changes applied to Uechi nouns when conditions for them
were met. Speaking of accent typology, like the other Miyako
Ryukyuan dialects, Uechi also lost a register system. I will explain
in detail below how the Uechi accent system evolved from Proto-
Miyako accent.
Monosyllables: In the development of Uechi monosyllabic
accent, there are three types of changes: Final Accent Loss,
Contour Pitch Simplification, and Pitch Leveling. The first sound
change, applicable only to high-initial words, canceled out the
accent of PM */`oo˘/, making it high-level atonic. And then, by
Pitch Leveling, it eventually became low-level. The other class (i.e.,
*/_oo˘/) did not undergo Final Accent Loss and Pitch Leveling.
However, Contour Pitch Simplification changed the initial pitch of
the /oo˘/ [RH(L)], resulting in HH(L) pitch (see (4ii) in 4.1.1).
224 ACCENTUAL HISTORY

(93) Development of Uechi monosyllabic accent

PM F-loss C-smpl Leveling

(a) *_oo˘ ≡ oo˘ > oo˘ = oo˘
[RH(L)] [RH(L)] [HH(L)]

(b) *`oo˘ > oo = oo > oo

[HH(L)] [HH(H)] [LL(L)]

Disyllables: Three accentual changes took place in the

development of Uechi disyllabic nouns: Final Accent Loss,
Abductive Change, and Pitch Leveling. First, undergoing Final
Accent Loss, PM */`OO˘/ (94a) became /`OO/. Along with this
change, the initial pitch of PM */_OO˘/ (94b) lost its
distinctiveness and, by contrast, the initial pitch of PM */OO/
[LL(L)] (94c) became distinctive. The second change is Abductive
Change. This change united the two low-initial classes (i.e., /OO˘/
(94b) and PM */OO/ (94c)) into /OO˘/. Finally, Pitch Leveling
changed the pitch of /OO/ [HH(H)] (< PM */`OO˘/) to low-level.

(94) Development of Uechi disyllabic accent

abductive
PM F-loss change Leveling
(a) *`OO˘ > `OO ≡ OO > OO
[HH(H)] [LL(L)]

(b) *_OO˘ ≡ OO˘

> OO˘ = OO˘
(c) *OO ª _OO

Trisyllables: Uechi trisyllabic accent resulted from the same

sequence of the changes as its disyllabic accent did. The first
change, Final Accent Loss, converted PM */`OOO˘/ (95a) to
/`OOO/. When the change took place, the initial pitch of PM
*/_OOO˘/ (95b) became no longer distinctive. However, the initial
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 225

pitch of PM */OOO/ [LLL(L)] (95c) was distinctive in contrast

with /`OOO/ [HHH(H)] (95a). At the second stage, /OOO˘/ (< PM
*/_OOO˘/) and /_OOO/ (95c) merged as /OOO˘/. Notice that
Abductive Change merged low-initial classes. The final change is a
phonetic change that made the high-level pitch of /OOO/ into low-
level.

(95) Development of Uechi trisyllabic accent

abductive
PM F-loss change Leveling
(a) *`OOO˘ > `OOO ≡ OOO > OOO
[HHH(H)] [LLL(L)]

(b) *_OOO˘ ≡ OOO˘

> OOO˘ = OOO˘
(c) *OOO ª _OOO

The changes took place in the order listed. However, Contour

Pitch Simplification affected monosyllables even though it did not
cause any change to disyllables and trisyllables. Moreover,
Abductive Change did not take place in monosyllables, but it did in
disyllables and trisyllables. When it comes to the ordering of these
two changes, Abductive Change must have taken place before
Contour Pitch Simplification because the former merges accent
classes depending on their initial pitch height. If Abductive Change
had taken place after Contour Pitch Simplification, Proto-Miyako
monosyllables would have lost their accent distinctions by merging
the classes (i.e., /oo˘/ [HH(L)] (93a) and /oo/ [HH(H)] (93b)).
Concerning typology of accent, in the Uechi line, the locus-
register system of Proto-Miyako became a locus system. This is a
result of a merger of accent distinctions caused by Abductive
Change.
226 ACCENTUAL HISTORY

5.5 YAEYAMA RYUKYUAN

5.5.1 Description of Yaeyama Ryukyuan Accent
Three dialects of Yaeyama Ryukyuan will be used in order to
reconstruct Proto-Yaeyama (PY): Ishigaki dialect (spoken in
Ishigaki City, Ishigaki Island), Sonai dialect (spoken in Sonai,
Iriomote Island), and Kuroshima dialect (spoken in Nakamoto,
Kuroshima). All data on the accent of those Yaeyama dialects used
in this book are taken from Hirayama et al. (1967).

5.5.1.1 Ishigaki Dialect

The accent system of Ishigaki nouns is simple -- nouns are either
high-register or low-register. In high-register nouns, the pitch falls
after the initial mora for two-mora nouns, and for three-mora nouns
the pitch falls after the second mora. In low-register nouns, the
pitch is low-level throughout the nouns.
Monosyllables: All monosyllables are two moras long; they
consist of either a long vowel or a sequence of a vowel and a nasal
consonant.

(96) Ishigaki monosyllabic accent

Phonemic Phonetic Myô List of nouns
`oo HL(L) 1.1 `ki: ‘hair’, `cï: ‘blood’
1.2 `pa: ‘leaf’, `na: ‘name’
_oo LL(L) 1.3 _ki: ‘tree’, _pa: ‘tooth’, _mi: ‘eye’
2.3 _in ‘dog’

The noun _in ‘dog’ is originally disyllabic, but it merged with

monosyllabic low-register nouns after dropping the final vowel /u/,
cf. inu˘ ‘dog’ in Tokyo and inu LL ‘dog’ in Myôgishô (Mochizuki
1974: 67).
Putting the original disyllabic nouns aside, it is clear that
Myôgishô class 1.1-2 nouns correspond to Ishigaki high-register
nouns, and also that Myôgishô 1.3 nouns are low-register in
Ishigaki.
Disyllables: Just like monosyllabic nouns, there are two
accentual types as well. High-register nouns are phonetically HL(L),
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 227

but the phonetic locus shifts one mora to the right when the first
mora is devoiced; they phonetically become LH(L), e.g., `pï•tu
‘person’.
Ishigaki disyllabic nouns are in general two moras long in the
(C)V(C)V structure. There are also three-mora disyllabic nouns
with a long vowel in the initial syllable, e.g., _a:mi ‘rain’, _u:ki
‘bucket’.

(97) Ishigaki disyllabic accent

Phonemic Phonetic Myô List of nouns
`OO HL(L) 2.1 `pana ‘nose’, `usï ‘cow’, `turï
‘bird’, `kusï ‘loins’, `pïni
‘beard’, `kazi ‘wind’
2.2 `isi ‘stone’, `kabï ‘paper’, `pasï
‘bridge’
_OO LL(L) 2.3 _pana ‘flower’, _yama
‘mountain’, _humu ‘cloud’, _puni
‘bone’
2.4 _huni ‘boat’, _parï ‘needle’, _usï
‘mortar’,_pasï ‘chopsticks’,
_kata ‘shoulder’, _ikï ‘breath’,
_ita ‘board’
2.5 _kui ‘voice’, _kai ‘shadow’,
_muku ‘bridegroom’, _a:mi
‘rain’, _asi ‘sweat’, _u:ki‘bucket’
3.4 _kangan ‘mirror’, _ongï ‘fan’
3.5 _para: ‘pillar’, _aba ‘fat’, _nucï
‘life’, _nada‘tears’, _mahwa
‘pillow’

As explained in 5.2.1, the noun _kui ‘voice’ is treated as

disyllabic. Similarly, I also treat the word _kai ‘shadow’ as
disyllabic. This word is historically disyllabic, i.e., kage LH
‘shadow’ in Myôgishô (Mochizuki 1974: 141).
Some Ishigaki disyllabic nouns are originally trisyllabic. They
are listed below in comparison with their Myôgishô cognates.
228 ACCENTUAL HISTORY

(98) Ishigaki Myôgishô

‘mirror’ (3.4) _kangan kagami
‘fan’ (3.4) _ongï afuki
‘pillar’ (3.5) _para: hasira
‘fat’ (3.5) _aba abura
‘life’ (3.5) _nucï inoti
‘tears’ (3.5) _nada namida
‘pillow’ (3.5) _mahwa makura

All Ishigaki words listed above have lost at least a syllabic

segment in the course of their development; as a result, they
became disyllabic. Specifically, the noun _kangan has lost the final
vowel and the final nasal became moraic, but the word also gained
another moraic nasal in the initial syllable. The word for ‘fan’ also
gained a nasal consonant. When words for ‘fat’, ‘life’, ‘tears’, and
‘pillow’ lost one syllable, they simply became two-mora disyllabic
nouns, but others such as ‘pillar’ compensated for the loss and
became three-mora.
The correspondences between Ishigaki disyllabic low-register
nouns and their cognates in Myôgishô are systematic. To put it
simply, the Ishigaki low-register nouns correspond to Myôgishô
low-initial nouns, and the Ishigaki high-register nouns are high-
initial nouns in Myôgishô.
Trisyllables: Ishigaki low-register nouns are phonetically low-
level, and high-register nouns are either HHL(L) or LHL(L). When
the initial syllable contains a devoiced vowel,51 the word is
LHL(L); otherwise HHL(L).
Some Ishigaki trisyllabic nouns show vowel length in either the
initial or second syllable, e.g., `huta:zï ‘two’, _u:nai ‘eel’, _mi:mizï
‘earthworm’. These correspond to futatu HHL ‘two’, munagi LHH
‘eel’, and mimizu LHH ‘earthworm’ in Myôgishô (Mochizuki 1974:
521). The reason why these words have vowel length in Ishigaki is
unknown.

51
In Ishigaki, high vowels become devoiced between two voiceless consonants.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 229

(99) Ishigaki trisyllabic accent

Phonemic Phonetic Myô List of nouns
`OOO HHL(L) 3.1 `sirusï ‘mark’, `kibusï ‘smoke’,
`akubï ‘yawn’
3.3 `kugani ‘gold’
LHL(L) 3.1 `hutai ‘forehead’
3.2 `huta:zï ‘two’ LHLL(L)
3.3 `cïkara ‘strength’
_OOO LLL(L) 3.4 _hukuru ‘bag’, _takara ‘treasure’
3.6 _u:nai ‘eel’, _mi:mizï
‘earthworm’
3.7 _husirï ‘medicine’, _kuzira
‘whale’, _patagi ‘field’

I will now turn to the correspondences of trisyllabic cognates

between Ishigaki and Myôgishô accentuation. Ishigaki low-register
nouns regularly correspond to Myôgishô classes 3.4-7, and Ishigaki
high-register nouns to Myôgishô classes 3.1-3.

5.5.1.2 Sonai Dialect

As far as available Yaeyama data are concerned, the accent system
of Sonai (Iriomote) dialect is the most complex one. There are three
accentual classes in the language: atonic, penultimate accent, and
final-rising accent. Monosyllabic nouns have a two-way distinction
in accent: tonic (penultimate-accent) and atonic. Disyllabic and
trisyllabic nouns distinguish three types of accentuation. To
distinguish one word from another, the accent system of the Sonai
dialect utilizes locus, not register. It is possible to analyze Sonai
monosyllabic accent as a register system, which leads to the
conclusion that only monosyllabic nouns have register, but that
disyllables and trisyllables only make use of locus. However,
taking the accent system as a whole into account, it is natural to
consider that accent functions in a uniform system. Therefore,
when we consider the accent system of the Sonai dialect as a whole,
it is reasonable to analyse monosyllabic accent as a locus system
than as a register system.
230 ACCENTUAL HISTORY

All Sonai words are taken from Hirayama et al. (1967). However
there is a discrepancy in the data. According to the description in
Hirayama et al. (1967: 45-9), the language has /OÈO/ [LH ~ LL(H)]
and /OOÈO/ [LLH ~ LLL(H)] classes, but in the word lists (1967:
60-75 and 193-504) the /OÈO/ and /OOÈO/ are described as /OO˘/
and /OOO˘/ respectively. I adopt the classes /OÈO/ and /OOÈO/
throughout this book because they are clearly described as such in
the description of the Sonai dialect.
Monosyllables: For Sonai Monosyllabic nouns, there are two
phonemic accent classes: penultimate-accent and low-atonic. They
are phonetically HL(L) and LL(L) respectively. Sonai
monosyllabic nouns are composed normally of two moras with a
long vowel.

(100) Sonai monosyllabic accent

Phonemic Phonetic Myô List of nouns
oo LL(L) 1.1 ki: ‘hair’, ci: ‘blood’
1.2 pa: ‘leaf’, na: ‘name’
o˘o HL(L) 1.3 ki˘: ‘tree’, pa˘: ‘tooth’

Accentual correspondences between Sonai monosyllables and

their cognates in Myôgishô are very systematic. Penultimate-accent
nouns regularly correspond to Myôgishô accent class 1.3, and low-
atonic nouns to Myôgishô 1.1-2.
Disyllables: For Sonai disyllabic nouns, there are three
accentual classes, namely low-level atonic (/OO/), a class with
penultimate-accent (/O˘O/), and a class with final-rising accent
(/OÈO/). As shown below, the disyllabic data in (101) include
nouns corresponding to Myôgishô trisyllabic accentuation. Those
nouns are originally trisyllabic, but became disyllabic when they
lost a syllabic segment in their development from earlier forms.
Not taking the historically-trisyllabic nouns into account, generally
atonic nouns correspond to Myôgishô accent classes 2.1-2. As for
tonic nouns, the data show that penultimate-accent nouns
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 231

correspond to Myôgishô accent classes 2.4-5 and final-rising accent

nouns to Myôgishô classes 2.3-5.
Words for ‘centipede’, ‘fan’, and ‘pillar’ are described as accent-
final in Hirayama et al. (1967: 67-70). This contradicts their
analysis in Hirayama et al. (1967: 45-9). As listed in (101), I write
them as final-rising because of the reason mentioned earlier.

(101) Sonai disyllabic accent

Phonemic Phonetic Myô List of nouns
OO LL(L) 2.1 pana ‘nose’, usi ‘cow’, kusi
‘loins’, pini ‘beard’, kazi ‘wind’
2.2 isi ‘stone’, kabi ‘paper’, pasi
‘bridge’
3.1 hute ‘forehead’
O˘O HL(L) 2.4 pa˘ri ~ pa˘:ri ‘needle’, u˘si
‘mortar’, pa˘si ‘chopsticks’, i˘ki
‘breath’
2.5 ku˘i ‘voice’
3.4 kaga˘n LHL ‘mirror’
OÈO LH ~ 2.3 paÈna ‘flower’, huÈmu ‘cloud’,
LL(H) iÈnu ‘dog’, puÈni ‘bone’
2.4 kaÈta ‘shoulder’, iÈta ‘board’,
fuÈni ‘boat’
2.5 aÈmi ‘rain’, aÈsi ‘sweat’,
muÈku ‘bridegroom’, uÈgi
‘bucket’
3.2 mo:Èza ‘centipede’
3.4 onÈgi ‘fan’
3.5 paÈra ‘pillar’, na:Èda ‘tears’,
mo:Èra ‘pillow’
3.6 o:Èni ‘eel’

As seen in Nakijin, Shuri, Ikema, and Ôura, Sonai also has

subclasses in disyllabic nouns corresponding to Myôgishô accent
classes 2.3-5, namely 2.3-5a and 2.3-5b distinctions that I discussed
232 ACCENTUAL HISTORY

earlier. Nouns belonging to subclass 2.3-5a are /O˘O/ in Sonai, and

ones belonging to subclass 2.3-5b are /OÈO/.
Turning now to originally-trisyllabic nouns, there are two types.
One type is two-mora disyllabic; these did not compensate for the
length of the lost segment. The other type is three-mora disyllabic,
resulting from compensatory lengthening. Examples for the former
type are hute ‘forehead’ and pa˘ra ‘pillar’, and for the latter type,
kaga˘n ‘mirror’, mo:za˘ ‘centipede’, on˘gi ‘fan’, na:Èda ‘tears’,
mo:Èra ‘pillow’, o:Èni ‘eel’. In Myôgishô, those examples are
attested as trisyllabic (see discussion on Miyako Ryukyuan in 5.4.
above).
As discussed earlier, words with CVV structure such as ku˘i
‘voice’ are treated as disyllabic, not monosyllabic (see discussion
in Miyako Ryukyuan). According to Hirayama et al. (1967),
characteristics of penultimate-accent nouns are that (i) they are
generally in (C)VV structure (e.g., ku˘i ‘voice’, fu˘i ‘pole’) and (ii)
they have a devoiced vowel in the final syllable (e.g., u˘si8
‘mortar’).
Trisyllables: According to Hirayama et al. (1967: 45-9), Sonai
trisyllabic nouns belong to a three-way accent system. More
specifically, nouns are either low-level atonic, penultimate-accent,
or final-rising accent. The atonic nouns are phonetically LLL(L)
and the penultimate-nouns are LHL(L). It should be noted that
among the penultimate-accent nouns, many 3.5-6 nouns have a
devoiced vowel in the final syllable. For the rising accent nouns,
their pitch shape is LLH in isolation, but they are LLL(H) when
followed by a particle -- the final syllable of a phrase is high.
In (102), hukuÈru ‘bag’, takaÈra ‘treasure’, and usaÈgi ‘rabbit’
are taken from Hirayama et al. (1967: 68-9) where final-rising
accent (i.e., /OOÈO/) is analysed as accent-final (i.e., /OOO˘/).
Because Hirayama et al. clearly describe Sonai as having final-
rising accent and also because those ‘final-falling accent’ nouns
regularly correspond to final-rising accent, I will assume that final-
falling accent nouns have been mistakenly listed.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 233

(102) Sonai trisyllabic accent

Phonemic Phonetic Myô List of nouns
OOO LLL(L) 3.1 sirusi ‘mark’, kibusi ‘smoke’,
aburi ‘yawn’,
3.2 huta:ci ‘two’
OO˘O LHL(L) 3.3 cika˘ra ‘strength’, kuga˘ni
‘gold’52
3.5 inu˘ci ‘life’
3.6 mimin˘ci ‘earthworm’
OOÈO LLH ~ 3.4 hukuÈru ‘bag’, takaÈra ‘treasure’
LLL(H) 3.5 abuÈra ‘fat’, kukuÈru ‘heart’
3.6 usaÈgi ‘rabbit’
3.7 husiÈri ‘medicine’, kuziÈra
‘whale’

The data show that Sonai atonic nouns correspond to Myôgishô

accent classes 3.1-2, the penultimate-accent to Myôgishô accent
classes 3.3 and 3.5-6, and the final-rising accent to Myôgishô
classes 3.4-7.
Although most Sonai trisyllabic nouns are three moras long,
there are two exceptional forms in the data. They are huta:ci ‘two’,
miminci ‘earthworm’. By contrast with earlier forms for those
words attested in Myôgishô (i.e., futatu HHL ‘two’ and mimizu
LHH ‘earthworm’ (Mochizuki 1974: 475 and 521)), the former has
a long vowel in the second syllable, and the latter has a moraic
nasal in the second syllable. See the discussion on the word huta:ci
‘two’ in 5.4.1.1. For the word for ‘earthworm’, see 5.3.2 for my
discussion on the origin of the nasal in relation to the Proto-
Okinawa form for ‘earthworm’. Moreover, note that the word for
‘earthworm’ shows irregular vowel length in other Ryukyuan
dialects.
As discussed above, there are three types of accentuation in the
Sonai dialect: atonic, penultimate, and final-rising accent. Among

52
This word is kugani LHL in isolation, but it is kugani ndu LLL-LH when
followed by a particle (Hirayama et al. 1967: 47).
234 ACCENTUAL HISTORY

the three accent types, in monosyllables, only two (i.e., atonic and
penultimate) are distinctive, while in disyllables and trisyllables
there are all the three types. This may be because rising accent is
sensitive to the syllable, not the mora. In other words, the locus of
rising accent is on the syllable, not on the mora.

(103) Summary of Sonai accent

Monosyllables Disyllables Trisyllables
Atonic oo OO OOO
LL(L) LL(L) LLL(L)
Penultimate o˘o O˘O OO˘O
HL(L) HL(L) LHL(L)
Final-rising OÈO OOÈO
LH ~ LL(H) LLH ~ LLL(H)

5.5.1.3 Kuroshima Dialect

The accent of Kuroshima nouns is relatively simple. Disyllabic and
trisyllabic nouns are contrastive in two ways, but monosyllabic
nouns do not have any phonemic accentual distinction. I have
analyzed the two-way distinctions of Kuroshima disyllables and
trisyllables as a tonic-atonic system.
Monosyllables: The Kuroshima accent system is simpler than
that of other Yaeyama dialects. Kuroshima monosyllabic nouns are
atonic; phonetically either LH(H) or HH(H). In the data below, the
latter phonetic shape is found only in the word in ‘dog’, which is
originally disyllabic.

(104) Kuroshima monosyllabic accent

Phonemic Phonetic Myô List of nouns
oo LH(H) 1.1 ki: ‘hair’, si: ‘blood’
1.2 pa: ‘leaf’, na: ‘name’
1.3 ki: ‘tree’, pa: ‘tooth’, mi: ‘eye’
HH(H) 2.3 in ‘dog’

In Kuroshima, syllable structure affects the phonetic pitch shape

of words. That is, if a noun is composed of the (C)Vn syllable
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 235

structure, the syllable will have high-level pitch instead of low on

the first mora and high on the second, e.g., in HH ‘dog’, cf. pana
LH ‘nose’. See also the trisyllabic noun han˘garï HHLL(L)
‘mirror’.
Disyllables: Kuroshima disyllabic nouns are either tonic or
atonic. Tonic nouns are phonetically HL(L), and atonic disyllables
are phonetically LH(H).

(105) Kuroshima disyllabic accent

Phonemic Phonetic Myô List of nouns
OO LH(H) 2.1 pana ‘nose’, usi ‘cow’,turï
‘bird’, kusi ‘loins’, pini ‘beard’,
kazi ‘wind’
2.2 isi ‘stone’, habi ‘paper’, pasi
‘bridge’
2.3 pana ‘flower’, yama ‘mountain’,
fumu ‘cloud’
2.4 hata ‘shoulder’, ica ‘board’
2.5 ami ‘rain’, asi ‘sweat’
3.5 aQva ‘fat’, nada ‘tears’
O˘O HL(L) 2.3 pu˘ni ‘bone’
2.4 fu˘ni ‘boat’, pa˘rï53 ‘needle’,
u˘si ‘mortar’, pa˘si ‘chopsticks’,
i˘ki ‘breath’
2.5 ku˘i ‘voice’, ha˘i ‘shadow’,
mu˘ku ‘bridegroom’, u:˘ki HHL
‘bucket’
3.2 nka˘za LHL ‘centipede’
3.4 on˘gi HHL ‘fan’
3.5 pa˘ra ‘pillar’, nu˘ci ‘life’,
maQ˘fa HLL ‘pillow’

53
According to Hirayama et al. (1967: 126), in Kuroshima, the vowel /ï/ is on its
way out. However, it appears after /s/ and /r/, and they are devoiced (e.g., parï8
‘needle’, turï8 ‘bird’, sï8kara ‘strength’).
236 ACCENTUAL HISTORY

When it comes to correspondences in accent between Kuroshima

original disyllabic nouns and Myôgishô cognates, the Kuroshima
atonic nouns correspond to all of the Myôgishô accent classes, i.e.,
classes 2.1-5, and tonic nouns to Myôgishô classes 2.3-5. Note that
some of the historical accent classes correspond to both tonic and
atonic nouns in Kuroshima.
Furthermore, as shown in (106) and (107), excluding
historically-trisyllabic nouns, there are two subclasses among
Kuroshima disyllabic nouns corresponding to the historical accent
classes 2.3-5. Nouns in (106) are tonic in Kuroshima and the ones
in (107) are atonic. These two groups exist in Ikema and Ôura
Miyako dialects as well (see (74) and (75) above). For an historical
perspective on this matter, see the historical explanations below.

(106) Kuroshima
‘bone’ (2.3) pu˘ni
‘boat’ (2.4) fu˘ni
‘needle’ (2.4) pa˘rï
‘mortar’ (2.4) u˘si
‘breath’ (2.4) i˘ki
‘shadow’ (2.5) ha˘i
‘bridegroom’ (2.5) mu˘ku
‘bucket’ (2.5) u:˘ki

(107) Kuroshima
‘flower’ (2.3) pana
‘mountain’ (2.3) yama
‘board’ (2.4) ica
‘rain’ (2.5) ami
‘sweat’ (2.5) asi

Most of the historical disyllabic nouns are two-mora in

Kuroshima. The noun u:˘ki ‘bucket’ is the only one that has a long
vowel in the initial syllable. This vowel length might have existed
in earlier forms in relation to accentuation (see 5.3.3.).
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 237

Some of the trisyllabic nouns have become disyllabic after

losing one syllable, and now have disyllabic accentuation, e.g.,
pa˘ra ‘pillar’ (3.5), nu˘ci ‘life’ (3.5), and nada ‘tears’ (3.5). There
are also historically-trisyllabic nouns that have disyllabic accent
classes, but are three moras long, e.g., nka˘za ‘centipede’ (3.2),
aQva ‘fat’ (3.5), maQ˘fa ‘pillow’ (3.5), and on˘gi ‘fan’ (3.4).
For the words with CVV structure (e.g., ku˘i ‘voice’, ha˘i
‘shadow’), see the discussion on Miyako Ryukyuan, where I argue
for the treatment of those words as disyllabic.
Trisyllables: Phonetically atonic trisyllabic nouns are LHH(H),
and tonic nouns are either LHL(L) or HHL(L); the nouns are
HHL(L) when the initial syllable is CVn-, otherwise LHL(L). In
other words, Kuroshima has a phonetic rule that lowers the initial
syllable of high-register nouns if the pitch does not fall
immediately after the syllable.

(108) Kuroshima trisyllabic accent

Phonemic Phonetic My List of nouns
ô
OOO LHH(H) 3.1 sirusi ‘mark’, kibusi ‘smoke’,
aubi ‘yawn’, futai ‘forehead’
3.2 futa:ci ‘two’
3.3 sïkara ‘strength’, kugani ‘gold’
3.4 takara ‘treasure’
OO˘O LHL(L) 3.4 huku˘rï ‘bag’, pata˘ki ‘field’,
icï:˘cï ‘five’
3.5 kuku˘ru ‘heart’, yara˘bi ‘child’
3.6 una˘gi ‘eel’, mi:ma˘zi
‘earthworm’
3.7 fusi˘rï ‘medicine’, kuzi˘ra
‘whale’, pata˘ki ‘field’
HHLL(L) 3.4 hanga˘rï ‘mirror’

The description of Kuroshima trisyllabic accent in (108)

includes nouns that are not in my original word list. In order to see
the whole picture of corresponding patterns between Kuroshima
238 ACCENTUAL HISTORY

and Myôgishô accentuation, more words have been added: pata˘ki

‘field’ (3.4), icï:˘cï ‘five’ (3.4), kuku˘ru ‘heart’ (3.5), and yara˘bi
‘child’ (3.5).
According to (108), Kuroshima atonic trisyllabic nouns
correspond to Myôgishô 3.1-4 nouns, and trisyllabic tonic nouns
correspond to Myôgishô classes 3.4-3.7. Some of the Kuroshima
nouns overlap different Myôgishô classes -- which implies the
occurrence of irregular sound change. Because Kuroshima nouns
corresponding to Myôgishô class 3.4 are normally tonic (huku˘rï
‘bag’, pata˘ki ‘field’, hanga˘rï ‘mirror’, icï:˘cï ‘five’), the word
‘treasure’ (3.4) must be irregular. Therefore, the correspondence
between Kuroshima and Myôgishô accent is systematic. That is,
Kuroshima atonic corresponds to Myôgishô classes 3.1-3, and
Kuroshima high-register corresponds to Myôgishô classes 3.4-7.
Among the trisyllabic nouns listed above, there are two with
vowel length, e.g., futa:ci ‘two’ and mi:ma˘zi ‘earthworm’. The
length might have been secondary because of its irregularity. That
is to say, in Kuroshima tonic trisyllables, the pitch falls after the
second mora. However, for the word mi:ma˘zi, the pitch falls after
the syllable ma, which is the third mora from the left, because the
pitch regularly falls after the second mora in Kuroshima trisyllables,
except for ‘earthworm’. This implies that the word might have
lengthened the initial-syllable vowel in its evolution, but the reason
for the lengthening is not known.
There is another noun that has four moras in the data: hanga˘rï
HHLL ‘mirror’, where there is a nasal in the initial syllable. This is
not rare in Ryukyuan; it is also seen in some other Ryukyuan
dialects (cf. kagami ‘mirror’ in Myôgishô, Tokyo, and Kyoto).
Kuroshima dialect is one of those Ryukyuan dialects, but it is not
clear how the moraic nasal came into existence.

5.5.2 Correspondences and Reconstruction of Proto-Yaeyama

Accent
This section contributes to a reconstruction of Proto-Yaeyama
accent. The reconstruction is straightforward; the number of
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 239

accentual distinctions reflects the number of correspondence sets.

Concerning typology of accent, the system of Proto-Yaeyama
accent utilizes locus and register. For more detail, see the following
discussions.
Monosyllables: There are two sets of accentual correspondences
regarding Yaeyama monosyllabic nouns. Notice that there is a
systematic correspondence between Myôgishô accent classes and
Proto-Yaeyama accent classes. This regularity of accent patterns
suggests that there should be two accent classes in Proto-Yaeyama.
Based on the correspondence sets in (109), I propose */`oo˘/ and
*/_oo˘/ for Proto-Yaeyama monosyllabic accent, phonetically
HH(L) and LH(L) respectively. Proto-Yaeyama monosyllabic
nouns are distinguished by register.

(109) Reconstructed Proto-Yaeyama accent for monosyllabic nouns

(a)
Ishigaki Sonai Kuroshima PY
`oo oo oo *`oo˘
HL(L) LL(L) LH(H) HH(L)
‘hair’ (1.1) `ki: ki: ki: *`ki:˘
‘blood’ (1.1) `cï: ci: si: *`cï:˘
‘leaf’ (1.2) `pa: pa: pa: *`pa:˘
‘name’ (1.2) `na: na: na: *`na:˘

(b)
Ishigaki Sonai Kuroshima PY
_oo o˘o oo *_oo˘
LL(L) HL(L) LH(H) LH(L)
‘tree’ (1.3) _ki: ki˘: ki: *_ki:˘
‘tooth’ (1.3) _pa: pa˘: pa: *_pa:˘
‘eye’ (1.3) _mi: -- mi: *_mi:˘

In (109b), a Sonai cognate for ‘eye’ is not available; Hirayama et

al. (1967) gives min˘ci for ‘eye’ in Sonai, but this is probably a
compound.
Disyllables: On the basis of the following correspondences, four
accent classes can be reconstructed for Proto-Yaeyama. The
240 ACCENTUAL HISTORY

reconstructed classes are */`OO˘/, */_OO˘/, and */OO/. See the

following summary of the correspondence sets and reconstruction.
For reconstructed nouns for each class, see (111).

(110) Summary of accentual correspondences and reconstruction of

Proto-Yaeyama accent for disyllables
Ishigaki Sonai Kuroshima PY
a `OO OO OO *`OO˘
HL(L) LL(L) LH(H) HH(L)
b _OO OÈO *_OO˘
LL(L) LH ~ LH(L)
LL(H)
c O˘O O˘O *OO
HL(L) HL(L) LL(L)

Proto-Yaeyama disyllabic accent makes use of register and locus

to distinguish one word from another. In the Ishigaki line, this has
become a register system. In the Sonai and Kuroshima lines, it
became a locus system.

(111) Reconstructed Proto-Yaeyama accent for disyllabic nouns

(a)
Ishigaki Sonai Kuroshima PY
`OO OO OO *`OO˘
HL(L) LL(L) LH(H) HH(L)
‘nose’ (2.1) `pana pana pana *`pana˘
‘cow’ (2.1) `usï usi usi *`usï˘
‘bird’ (2.1) `turï -- turï *`turï˘
‘loins’ (2.1) `kusï kusi kusi *`kusï˘
‘beard’ (2.1) `pïni pini pini *`pïni˘
‘wind’ (2.1) `kazi kazi kazi *`kazi˘
‘stone’ (2.2) `isi isi isi *`isi˘
‘paper’ (2.2) `kabï kabi kabi *`kabï˘
‘bridge’ (2.2) `pasï pasi pasi *`pasï˘
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 241

(b)
Ishigaki Sonai Kuroshima PY
_OO OÈO OO *_OO˘
LL(L) LH ~ LH(H) LH(L)
LL(H)
‘flower’ (2.3) _pana paÈna pana *_pana˘
‘mountain’ (2.3) _yama (yamana) yama *_yama˘
‘cloud’ (2.3) _humu huÈmu fumu *_fumu˘
‘dog’ (2.3) _in iÈnu in *_inu˘
‘board’ (2.4) _ita iÈta ica *_ita˘
‘shoulder’ (2.4) _kata kaÈta hata *_kata˘
‘rain’ (2.5) _a:mi aÈmi ami *_a:mi˘
‘sweat’ (2.5) _asi aÈsi asi *_asi˘
‘tears’ (3.5) _nada na:Èda nada *_na:da˘

(c)
Ishigaki Sonai Kuroshima PY
_OO O˘O O˘O *OO
LL(L) HL(L) HL(L) LL(L)
‘bone’ (2.3) _puni puÈni pu˘ni *puni
‘needle’ (2.4) _parï pa˘ri, pa˘rï *pa:rï
pa˘:ri
‘mortar’ (2.4) _usï u˘si u˘si *usï
‘chopsticks’ (2.4) _pasï pa˘si pa˘si *pasï
‘boat’ (2.4) _huni huÈni fu˘ni *funi
‘breath’ (2.4) _ikï i˘ki i˘ki *ikï
‘shadow’ (2.5) _kai ka˘i ha˘i *kai
‘voice’ (2.5) _kui ku˘i ku˘i *kui
‘bucket’ (2.5) _u:ki uÈgi u:˘ki *u:ki
‘bridegroom’ (2.5) _muku muÈku mu˘ku *muku
‘fan’ (3.4) _ongï onÈgi on˘gi *ongï
‘pillar’ (3.5) _para: paÈra pa˘ra *para:

In the matter of cognates in the correspondences in (111), the

words meaning ‘tears’ (3.5), ‘fan’ (3.4), and ‘pillar’ (3.5) are
historically trisyllabic nouns, but I reconstruct them as disyllabic,
i.e., PY *_na:da˘ ‘tears’ (111b), *ongï ‘fan’ (111c), and PY *para
‘pillar’ (111c), because none of the cognate forms are trisyllabic.
242 ACCENTUAL HISTORY

At the time of Proto-Yaeyama, these words were probably

disyllabic.
In the correspondences in (111c), Sonai cognates for ‘boat’,
‘bucket’, ‘bridegroom’, ‘fan’, and ‘pillar’ have irregular
accentuation; they are rising accent while most of cognates in the
correspondence set are initial-accent.
As discussed earlier, it appears that there are two subclasses for
2.3-5 words, and they are seen in Shuri, Nakijin, Ikema, Ôura, and
Kuroshima Ryukyuan. Cognates belonging to each subclass are
identical among these other Ryukyuan dialects (see 5.3.3, 5.4, and
5.5.1.4).
Trisyllables: In a comparison of the trisyllabic nouns of
Yaeyama Ryukyuan, we can establish four accentual classes. The
following is a summary of correspondences and reconstructed
accent classes for Proto-Yaeyama trisyllables.
According to the reconstruction, the Proto-Yaeyama trisyllabic
accent is a system utilizing both register and locus distinctions.
Register distinguishes high-initial final falling accent from low-
initial final falling accent.

(112) Summary of accentual correspondences and reconstruction

of Proto-Yaeyama accent for trisyllables
Ishigaki Sonai Kuroshima PY
a `OOO OOO OOO *OOO
HHL(L)/ LLL(L) LHH(H) HHH(H)
b LHL(L) OO˘O *`OOOˆ
LHL(L) HHF ~
HHH(L)
c _OOO OO˘O *OO˘O
LLL(L) LHL(L) LHL(L)
d OOÈO *_OOOˆ
LLH ~ LHF ~
LLL(H) LHH(L)

The following list of correspondences includes disyllabic nouns,

e.g., _mahwa ‘pillow’ in Ishigaki (113d), kagan ‘mirror’ in Sonai
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 243

(113c), and nu˘ci ‘life’ in Kuroshima (113c), which are historically

trisyllabic. (See 5.5.1 above for detailed discussions on those
historical trisyllables.) Because these can be reconstructed as
trisyllabic in Proto-Yaeyama in comparison with other cognate
forms in Yaeyama dialects, they are listed in (113).

(113) Reconstructed Proto-Yaeyama accent for trisyllabic nouns

(a)
Ishigaki Sonai Kuroshima PY
`OOO OOO OOO *OOO
HHL(L)/ LLL(L) LHH(H) HHH(H)
LHL(L)
‘mark’ (3.1) `sirusï sirusi sirusi *sirusï
‘forehead’ (3.1) `hutai hute futai *futai
‘yawn’ (3.1) `akubï aburi aubi *akubï
‘smoke’ (3.1) `kibusï kibusi kibusi *kibusï
‘two’ (3.2) `huta:zï huta:ci futa:ci *futa:cï

(b)
Ishigaki Sona Kuroshima PY
`OOO OO˘O OOO *`OOOˆ
HHL(L)/ LHL(L) LHH(H) HHF ~
LHL(L) HHH(L)
‘strength’ (3.3) `cïkara cika˘ra sïkara *`cïkaraˆ
‘gold’ (3.3) `kugani kuga˘ni kugani *`kuganiˆ

(c)
Ishigaki Sonai Kuroshima PY
_OOO OO˘O OO˘O *OO˘O
LLL(L) LHL(L) LHL(L) LHL(L)
‘mirror’ (3.4) _kangan kaga˘n hanga˘rï *kaga˘nï
‘life’ (3.5) _nucï inu˘ci nu˘ci *inu˘cï
‘earthworm’ (3.6) _mi:mizï mimin˘ci mi:ma˘zi *mi:mi˘zï
‘field’ (3.7) _patagi -- pata˘ki *pata˘ki
244 ACCENTUAL HISTORY

(d)
Ishigaki Sonai Kuroshima PY
_OOO OOÈO OO˘O *_OOOˆ
LLL(L) LLH ~ LHL(L) LHF ~
LLL(H) LHH(L)
‘centipede’ (3.2) _ngaza mo:Èza nka˘za *_mukazaˆ
‘bag’ (3.4) _hukuru hukuÈru huku˘rï *_hukuruˆ
‘treasure’ (3.4) _takara takaÈra takara *_takaraˆ
‘fat’ (3.5) _aba abuÈra aQva *_aburaˆ
‘pillow’ (3.5) _mahwa mo:Èra ma˘Qfa *_mafuraˆ
‘heart’ (3.5) _kukuru kukuÈru kuku˘ru *_kukuruˆ
‘cousin’ (3.5) _icihu icihu isiku LHH *_icikuˆ
‘eel’ (3.6) _u:nai o:Èni una˘gi *_u:nagiˆ
‘medicine’ (3.7) _husirï husiÈri fusi˘rï *_fusurïˆ
‘whale’ (3.7) _kuzira kuziÈra kuzi˘ra *_kuziraˆ

When it comes to irregular correspondences, the accentuation of

the word takara ‘treasure’ (113d) in Kuroshima must be irregular;
it must have been tonic. That is to say, it must be the same as other
nouns corresponding to Myôgishô class 3.4. A group of words
sharing the same historical accentuation generally undergoes the
same line of changes; thus those words should share the same
accentuation in their modern forms, unless there have been external
influence on the language, such as language contact. Between
Ishigaki trisyllabic nouns and their corresponding Myôgishô nouns,
there are systematic correspondences, e.g., Myôgishô 3.1-3 nouns
are high-register in Ishigaki, and Myôgishô 3.4-7 nouns are low-
register in Ishigaki. On the other hand, in Kuroshima, some
trisyllabic nouns corresponding to Myôgishô 3.4 are tonic and
others are atonic, e.g., huku˘rï ‘bag’ (3.4), takara ‘treasure’ (3.4),
and aQva ‘fat’ (3.4). The fact that Myôgishô 3.4 nouns correspond
to both tonic and atonic suggests the occurrence of irregular sound
change(s) in Kuroshima.
Regarding Yaeyama nouns corresponding to Myôgishô accent
class 3.2, there are two patterns, namely ‘two’ in (113a) and
‘centipede’ in (113d). The former has the same pattern as nouns
corresponding to Myôgishô class 3.1, and the latter to Myôgishô
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 245

classes 3.4-7. On the basis of merging patterns of accent classes in

Ryukyuan, it is natural to assume that a correspondence set for
‘two’ is regular and the one for ‘centipede’ (3.2) is irregular. In the
three Yaeyama dialects above, a cognate for ‘centipede’ has
somehow come to have the same accent as nouns corresponding to
Myôgishô accent classes 3.4-7.
I am not able to determine the accent type for the word for ‘sea’
because the Ishigaki and Kuroshima forms (i.e., tumo:rï and
tuma˘rï) are not cognate with the Japonic word for ‘sea’, and the
Sonai word Qsuna ‘sea’ also does not seem to be cognate; it is
possibly a compound form.

(114) Indeterminable correspondence

Ishigaki Sonai Kuroshima

‘sea’ tumo:rï Qsuna tuma˘rï

5.5.3 Development of Modern Yaeyama Dialects from Proto-

Yaeyama Accent
I have proposed a reconstruction of Proto-Yaeyama accent in the
previous section. In this section, I will explain how the modern
Yaeyama dialects (i.e., Ishigaki, Sonai, and Kuroshima) have
developed their accent systems from the proto accentual system.
(115) gives a summary of my reconstructed Proto-Yaeyama accent
system.

(115) Proto-Yaeyama accent system

Monosyllables Disyllables Trisyllables
*`oo˘ HH(L) *`OO˘ HH(L) *OOO HHH(H)
*_oo˘ LH(L) *_OO˘ LH(L) *`OOOˆ
HHF ~ HHH(L)
*OO LL(L) *OO˘O LHL(L)
*_OOOˆ
LHF ~ LHH(L)
246 ACCENTUAL HISTORY

5.5.3.1 Development of Ishigaki Accent

Although the developments of Ishigaki monosyllabic and disyllabic
accent are simple, involving only two accent changes, its
trisyllables underwent a complex course of development. From a
typological point of view, the word-initial register system existing
in Proto-Yaeyama has been kept intact in the Ishigaki dialect.
Monosyllables: As (116) illustrates, Final Accent Loss and
Initial Accent Gain rules contributed to the development of Ishigaki
monosyllabic accent.

(116) Development of Ishigaki monosyllabic accent

PY F-loss I-gain
(a) *`oo˘ > `oo > `oo
[HH(L)] [HL(L)]

(b) *_oo˘ > _oo = _ oo

[LH(L)] [LL(L)]

The application of the first rule made Proto-Yaeyama

monosyllables atonic (i.e., /`oo/ and /_oo/). By the following sound
change, /`oo/ [HH(H)] (< PY */`oo˘/) became /`oo/ [HL(L)]. This
rule does not apply to low-initial words.
Disyllables: The development of Ishigaki disyllabic accent
involves two sound changes, i.e., Final Accent Loss and Initial
Accent Gain. At the first stage, by the Final Accent Loss rule, PY
*/`OO˘/ (117a) and */_OO˘/ (117b) lost their final accent -- they
became /`OO/ and /_OO/ respectively. And the latter resulted in a
merger with PY */OO/ [LL(L)] (117c). In other words, the rule
converted a Proto-Yaeyama three-way system to a two-way
system: high-register atonic and low-register atonic.
At the next stage, the Initial Accent Gain rule, applicable only to
high-initial nouns, made high-register nouns HL(L) at the phonetic
level. Therefore, the Ishigaki dialect preserves Proto-Yaeyama
register, even though their phonetic pitch shapes are different.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 247

(117) Development of Ishigaki disyllabic accent

PY F-loss I-gain
(a) *`OO˘ > `OO > `OO
[HL(L)]

(b) *_OO˘
> _OO = _OO
(b) *OO

Trisyllables: As shown below, there are four stages in the

development of Ishigaki trisyllabic accent.

(118) Development of Ishigaki trisyllabic accent

abductive
PY shift change C-smpl F-loss
(a) *OOO = OOO
> OOOˆ > OO˘O ª `OOO
(b) *`OOOˆ = `OOOˆ [HHL(L)]

(c) *OO˘O > OOO˘

> OOO˘ = OOO˘ > _OOO
(d) *_OOOˆ = _OOOˆ

At the first stage, Accent Shift turned PY */OO˘O/ (118c) into

/OOO˘/. Speakers of the Ishigaki dialect began to lose the
distinction between PY */OOO/ [HHH(H)] (118a) and */`OOOˆ/
[HHF ~ HHH(L)] (118b), and also between /OOO˘/ [LHH(L)]
(118c) and */_OOOˆ/ [LHF ~ LHH(L)] because of similarity in
their pitch shapes. Notice that high-initial classes merged together
and low-initial classes merged as one. Furthermore, when this
merger took place, the Proto-Yaeyama register distinction was
completely lost. At the third stage, Contour Pitch Simplification
affected /OOOˆ/ [HHF ~ HHH(L)] and it became /OO˘O/
[HHL(L)]. Finally, /OOO˘/ (118cd) lost its accent by undergoing
248 ACCENTUAL HISTORY

Final Accent Loss, becoming /_OOO/ [LLL(L)]. This caused the

change from /OO˘O/ to /`OOO/ [HHL(L)].
Regarding ordering of the changes in the Ishigaki dialect, there
are five changes total. Basically the changes occur as listed in (118).
Since Initial Accent Gain takes place after Final Accent Loss as
shown in (116) and (117), it is ordered at the end in the
development of Ishigaki accent.

5.5.3.2 Development of Sonai Accent

The development of Sonai accent involves a total of five different
changes, including both phonemic and phonetic ones. Earlier in the
development, Proto-Yaeyama register was lost, and Sonai accent
became a locus system. Furthermore, unlike Ishigaki, a merger of
accent classes is rare; one merger is seen in trisyllables. The
following section will reveal how Sonai accent reached its current
state.
Monosyllables: The Sonai monosyllabic noun accent system
resulted from undergoing Final Accent Loss, Initial Accent Gain,
and Pitch Leveling. All Proto-Yaeyama monosyllabic nouns lost
their final accent by means of Final Accent Loss. All nouns became
atonic. At this point, distinctions are made by register, i.e., /`oo/
and /_oo/. When the next change (I-gain) took place, /_oo/ (119b)
was converted into /o˘o/. This is a conditioned rule, applying only
to low-initial words. Thus, /`oo/ (119a) did not undergo the change.
However, its initial pitch was no longer distinctive; the change
from /`oo/ to /oo/ occurred.

(119) Development of Sonai monosyllabic accent

PY F-loss I-gain Leveling

(a) *`oo˘ > `oo ≡ oo > oo [LL(L)]
[HH(L)]

(b) *_oo˘ > _oo > o˘o = o˘o

[LH(L)]
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 249

In addition to those sound changes, there is a phonetic change as

well. Notice that PY */`oo/ [HH(H)] (119a) became low-level in
the Sonai line. This is because Proto-Yaeyama high-level nouns
underwent lowering in pitch after register became nondistinctive.
Disyllables: In the development of Sonai accent, Proto-
Yaeyama disyllables underwent the same series of accent changes
as monosyllables. The first change deleted the final accent of PY
*/`OO˘/ (120a) and */_OO˘/ (120b). The former became /`OO/
and the latter /OÈO/. It also caused a change to PY */OO/, although
it is secondary. PY */OO/ became /_OO/ in contrast with /`OO/
(120a).
In this development, the register distinction was lost when the
Initial Accent Gain rule turned /_OO/ into /O˘O/; the initial pitch of
/`OO/ became nondistinctive. The rule is not applicable to high-
initial nouns, thus the high-initial atonic accent remained the same.
Finally, the application of Pitch Leveling lowered the pitch shape
of /OO/. That is, HH(H) became LL(L).

(120) Development of Sonai disyllabic accent

PY F-loss I-gain Leveling

(a) *`OO˘ > OO ≡ OO > OO
[HH(H)] [LL(L)]

(b) *_OO˘ > OÈO = OÈO = OÈO

(c) *OO ª _OO > O˘O = O˘O

Trisyllables: There were three changes related to Sonai

trisyllabic accent historically: Low Tone Spreading (L-sprd),
Contour Pitch Simplification, and Pitch Leveling.
First, the initial pitch of PY */_OOOˆ/ (121d) had an influence
at the phonetic level, and turned H-pitch to low. Therefore, LHF ~
LHH(L) became LLF ~ LLH(L). Second, Contour Pitch
Simplification changed PY */`OOOˆ/ (121b) and */_OOOˆ/ (121d)
to /OO˘O/ and /OOÈO/ respectively. As a result, the former merged
250 ACCENTUAL HISTORY

with PY */OO˘O/ (121c). Moreover, the register system of Proto-

Yaeyama was lost as well. At the final stage of the development,
PY */OOO/ (121a) lowered its pitch from HHH(H) to LLL(L).

(121) Development of Sonai trisyllabic accent

PY L-sprd C-smpl Leveling

(a) *OOO = OOO = OOO = OOO
[HHH(H)] [LLL(L)]

(b) *`OOOˆ = `OOOˆ

> OO˘O = OO˘O
(c) *OO˘O = OO˘O [LHL(L)]

(d) *_OOOˆ > _OOOˆ > OOÈO = OOÈO

[LHF ~ [LLF ~
LHH(L)] LLH(L)]

As mentioned above, there are five changes in the Sonai dialect:

Final Accent Loss, Initial Accent Gain, Pitch Leveling, Low Tone
Spreading, and Contour Pitch Simplification. The first three had an
effect on both monosyllables and disyllables, but the other two did
not. For trisyllables, the last three took place, but the first two did
not. Some changes did not occur because conditions for the
changes were not met. As a whole, considering the developments in
(119 - 121), the changes seen in the Sonai line must have been
ordered as follows: Low Tone Spreading > Final Accent Loss >
Initial Accent Gain > Contour Pitch Simplification > Leveling.
Note that Low Tone Spreading does not have any effect on
accented syllables; thus, for example, /OO˘/ [LH(L)] would not
become LL(L) or LL(H).

5.5.3.3 Development of Kuroshima Accent

As Kuroshima accent evolved, its monosyllables underwent two
changes, disyllables three changes, and trisyllables three changes.
The three changes that took place in the development of the
trisyllabic accent are very different from the sound changes
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 251

involved in the monosyllabic and disyllabic accents. The following

will show how those changes played roles in the development of
Kuroshima accent.
Monosyllables: In the development of Kuroshima monosyllabic
accent from Proto-Yaeyama, Final Accent Loss and Abductive
Change played a crucial role. By the first rule, PY */`oo˘/ and
*/_oo˘/ became /`oo/ and /_oo/ respectively. Next, Abductive
Change merged them into one class, namely /oo/ [LH(H)]. As a
result, the Proto-Yaeyama register distinction was lost.

(122) Development of Kuroshima monosyllabic accent

abductive
PY F-loss change
(a) *`oo˘ > `oo
[HH(L)]
> oo
[LH(H)]
(b) *_oo˘ > _oo
[LH(L)]

Disyllables: With the applications of Initial Accent Gain and

Final Accent Loss, Kuroshima disyllabic accent evolved from
Proto-Yaeyama.

(123) Development of Kuroshima disyllabic accent

abductive
PY I-gain F-loss change
(a) *`OO˘ = `OO˘ > `OO
> OO
[LH(H)]
(b) *_OO˘ = _OO˘ > _OO

(c) *OO > O˘O = O˘O = O˘O

252 ACCENTUAL HISTORY

When the first rule was applied, PY */OO/ (132c) became initial
accent. The second rule made an accent on the final syllable
nondistinctive. That is to say, PY */`OO˘/ and */_OO˘/ became
/`OO/ and /_OO/ respectively. Afterwards, they merged as /OO/
[LH(H)] at the final stage by means of Abductive Change. Due to
this change, the Proto-Yaeyama register system was lost.
Trisyllables: In the Kuroshima dialect, Proto-Yaeyama accent
experienced a rather drastic change by undergoing Abductive
Change, Contour Pitch Simplification, and Initial Pitch Lowering.
The four accent classes of Proto-Yaeyama trisyllables have been
reduced in half, and the Proto-Yaeyama register distinction has
been lost.
Undergoing Abductive Change, PY */OOO/ (124a) and
*/`OOOˆ/ (124b) merged as /OOO/. Because of similarity in pitch
shapes, speakers of the Kuroshima dialect lost the distinction
between them. At this point, the register was not distinctive any
more. The second change (namely, Contour Pitch Simplification)
converted PY */_OOOˆ/ (124d) to /OO˘O/, and it merged with PY
*/OO˘O/ [LHL(L)] (124c). Finally, /OOO/ [HHH(H)] (124a and b)
became LHH(H). Low pitch on unaccented initial syllables is
commonly seen in Yaeyama Ryukyuan.

(124) Development of Kuroshima trisyllabic accent

abductive
PY change C-smpl I-low
(a) *OOO
> OOO = OOO > OOO
(b) *`OOOˆ [LHH(H)]

(c) *OO˘O = OO˘O

> OO˘O = OO˘O
(d) *_OOOˆ ≡ OOOˆ

Looking into the changes that took place in the development of

Kuroshima accent as a whole, we know that Initial Accent Gain
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 253

follows Initial Pitch Lowering. If the ordering is reversed,

Kuroshima /OOO/ [LHH(H)] might have become initial-accent.

5.6 YONAGUNI RYUKYUAN

Description of Yonaguni Ryukyuan Accent
The Yonaguni dialect that will be analysed in this section is the
Sonai dialect of Yonaguni island. Yonaguni is geographically the
most-west island of the Ryukyus. The closest island to it is the
island of Iriomote (approximately 80 km), and is very close to
Taiwan (approximately 110 km).
As far as the Ryukyuan language is concerned, the accent system
of Yonaguni Ryukyuan is one of the most complex systems. In its
accent system, there are three accent classes and they are
distinguished by register and locus. One of the accent classes is a
high register class. The second class is a low register class; it is
low-level throughout. The last accent class is a class with a final
accent. Nouns with this accent start in low pitch and the pitch falls
within the final mora, e.g., huni˘ LF ‘boat’. However, the pitch falls
immediately after the final mora when uttered with a particle, e.g.,
huni˘ga LH-L. See the following description for more examples.
Monosyllables: Yonaguni monosyllabic nouns have two accent
classes: high-register atonic and low-register atonic. The high-
register nouns are phonetically high-level. All the low-register
nouns are low-level pitch. As for the structure of the monosyllables,
they are all comprised of two moras.

(125) Yonaguni monosyllabic nouns

Phonemic Phonetic Myô List of nouns
`oo HH(H) 1.1 `khi: ‘hair’, `ci: ‘blood’
1.2 `ha: ‘leaf’, `na: ‘name’
_oo LL(L) 1.3 _khi: ‘tree’, _ha: ‘tooth’, _mi:
‘eye’

When it comes to accentual correspondences between Yonaguni

monosyllables and their Myôgishô cognates, Yonaguni high-
254 ACCENTUAL HISTORY

register monosyllables correspond to Myôgi-shô classes 1.1-2, and

low-register nouns to Myôgishô class 1.3.
Disyllables: In Yonaguni disyllabic nouns, there are three
distinctive accent patterns: high-register atonic, low-register atonic,
and final-falling accent.

(135) Yonaguni disyllabic nouns

Phonemic Phonetic Myô List of nouns
`OO LH(H) 2.1 `uci ‘cow’, `khuci ‘loins’, `khadi
‘wind’, `Ngi ‘beard’
2.2 `haci ‘bridge’, `khabi ‘paper’
3.2 `ta:ci ‘two’
HH(H) 3.1 `tai ‘forehead’
_OO LL(L) 2.3 _hana ‘flower’, _dama
‘mountain’, _inu 'dog', _mmu
‘cloud’
2.4 _kata ‘shoulder’, _ita ‘board’
2.5 _ami ‘rain’, _asi ‘sweat’
3.4 _ku:ru ‘bag’, _kaNan ‘mirror’
3.5 _anda ‘fat’, _nuda ‘tears’
OOˆ LF ~ 2.3 huniˆ ‘bone’
LH(H) 2.4 uciˆ ‘mortar’, haciˆ ‘chopsticks’,
itiˆ ‘breath’, nniˆ ‘boat’
2.5 khaNiˆ ‘shadow’, muguˆ
‘bridegroom’, ugiˆ ‘bucket’
3.5 hiraˆ ‘pillar’, nutiˆ54 ‘life’
3.7 cu:riˆ ‘medicine’
HL ~ 2.4 haiˆ ‘needle’
HH(H) 2.5 khuiˆ ‘voice’

Yonaguni disyllables are distinguished by not only register but

also by locus. The high-register nouns are phonetically either
LH(H) or HH(H). They are generally LH(H), but HH(H) when they
have CVi word structure. The low-register nouns are low-level

54
This is taken from Hirayama et al. (1967).
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 255

throughout the words. In final-falling accent nouns, there are two

types of phonetic pitch shapes; one is LF and the other HL in their
citation form. The latter type of nouns is in the CVi structure, e.g.,
haiˆ ‘needle’, as opposed to (C)VCV or CCV structure of the
former type. Furthermore, the pitch shape of those final-falling
accent nouns changes when they are used in a phrase; LF-nouns
become LH(H), and HL-nouns become HH(H) when followed by a
particle.
Notice that nCV-words such as `Ngi ‘beard’ (2.1), _mmu ‘cloud’
(3.4), and nniˆ ‘boat’ (2.4) are analysed as disyllabic. This is
because they behave as disyllabic and they do not act like other
monosyllables do. If nniˆ ‘boat’ (2.4) were analyzed as
monosyllabic, another accent class would have to be established
only for this word. In addition, the nCV-words are disyllabic in
Japanese (e.g., hige ‘beard’, kumo ‘cloud’, and fune ‘boat’ in Tokyo
and Kyoto), and they are historically disyllabic, as well.
The list of disyllabic nouns in (126) includes not only original
disyllabic nouns but also historically trisyllabic nouns such as `tai
‘forehead’, `ta:ci ‘two’, _ku:ru ‘bag’ , _kaNan ‘mirror’, _anda ‘fat’,
_nuda ‘tears’, hiraˆ ‘pillar’, nutiˆ ‘life’, and cu:riˆ ‘medicine’.
These formerly-trisyllabic nouns are attested in Myôgishô as
displayed in (127) (see Mochizuki 1974)55. The formerly-trisyllabic
nouns lost a syllable in the course of their development in the
dialect.
A comparison of Yonaguni disyllabic nouns with their
corresponding Myôgishô cognates makes it clear that there is a
systematic accent correspondence in accent between them.
Yonaguni high-register atonic disyllables correspond to Myôgishô
high-initial accent classes, and Yonaguni low-register atonic and
final-falling accent disyllables to Myôgishô low-initial accent
classes.

55
Specific page numbers for the words are as follows: fitafi (p. 457), futatu (p.
475), fukuro (p. 471), kagami (p. 134), afura (p. 32), namida (p. 394), fasira (p.
434), inoti (p. 68), kusuri (p. 190), and mukade (p. 526).
256 ACCENTUAL HISTORY

(127) Historical trisyllabic nouns

Yonaguni Myôgishô
‘forehead’ `tai fitafi HHH/HH…
‘two’ `ta:ci futatu HHL
‘bag’ _ku:ru fukuro LLL
‘mirror’ _kaNan kagami LLL
‘fat’ _anda afura LLH/LHL
‘tears’ _nuda namida, namuda LLH
‘pillar’ hiraˆ fasira LLH
‘life’ nutiˆ inoti LLH
‘medicine’ cu:riˆ kusuri LHL
‘centipede’ Nkadiˆ mukade HHL

Trisyllables: The accent system of Yonaguni trisyllabic nouns is

nearly identical to that of its disyllabic nouns -- the only difference
between them is in the number of syllables. In the system of the
trisyllables, high-register nouns are phonetically LHH(H), low-
register nouns LLL(L), and final-falling accent nouns LHF ~
LHH(H). Note that Yonaguni has a rule that lowers the initial-
syllable pitch of words if the syllable is not accented.
The word Nkadiˆ ‘centipede’ is analysed as trisyllabic because it
acts as trisyllabic (see also the discussion on nCV-words above).
The accentuation of the word seems irregular. In general, words
belonging to historical high-register accent classes 3.1-3 and words
in historical low-register classes 3.4-7 (except 3.6 class; no data
available for the class) systematically develop their accent; they do
not generally ignore register distinction. However, the word for
‘centipede’ seems to be irregular in Ryukyuan dialects -- it is not
listed with nouns corresponding to historical accent classes 3.1 and
3.2. Its accent might have become irregular earlier than Proto-
Ryukyuan. For the time being, I leave this matter for future
research.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 257

(128) Yonaguni trisyllabic nouns

Phonemic Phonetic Myô List of nouns
`OOO LHH(H) 3.1 `ciruci ‘mark’, `khibunci
‘smoke’, `agui ‘yawn’
3.3 `sikara ‘strength’, `khuNani
‘gold’
3.7 `hatagi ‘field’56
_OOO LLL(L) 3.4 _tagara ‘treasure’, _kuyumi
‘calendar’, _icici ‘five’
3.5 _magura ‘pillow’
OOOˆ LHF ~ 3.2 Nkadiˆ ‘centipede’57
LHH(H) 3.7 khudiraˆ ‘whale’, taraiˆ ‘basin’

Regarding the accentual correspondences between Yonaguni

trisyllabic nouns and their Myôgishô cognates, Yonaguni high-
register trisyllabic nouns correspond to Myôgishô classes 3.1, 3.3,
and 3.7, Yonaguni low-register nouns to Myôgishô classes 3.4-5,
and Yonaguni tonic nouns to 3.7 accent class of Myôgishô. The
accentuation of the word `hatagi ‘field’ (3.7) must be irregular
because Yonaguni nouns corresponding to Myôgishô class 3.7 are
normally final-falling accent (e.g., khudiraˆ ‘whale’, taraiˆ ‘basin’).

5.7 CORRESPONDENCES AND RECONSTRUCTION OF

PROTO-RYUKYUAN ACCENT
My main goals are twofold for this section. The first is to present
accentual correspondences for monosyllables, disyllables, and
trisyllables among Proto-Amami, Proto-Okinawa, Proto-Miyako,
Proto-Yaeyama, and Yonaguni. The second goal is to reconstruct
the Proto-Ryukyuan accent system on the basis of the
correspondences. I will also reveal the individual Proto-Ryukyuan
nouns belonging to each accent class.

56
In contrast, Hirayama et al. (1967) list the word with different accentuation:
hatagiˆ ‘field’.
57
In Hirayama et al. (1967), the word for ‘centipede’ is `Nkadi.
258 ACCENTUAL HISTORY

As I explained, there is no Proto-Yonaguni. This is because a

reconstruction of it is not possible at present due to a lack of data
on other Yonaguni dialects; only Sonai data are available.
Therefore, I directly compare the Sonai data with Proto-Okinawa,
Proto-Amami, Proto-Miyako, and Proto-Yaeyama data in order to
reconstruct Proto-Ryukyuan accent.
Monosyllables: As shown in (129), there are three sets of
accentual correspondences for Ryukyuan monosyllabic nouns. This
enables us to hypothesize three Proto-Ryukyuan accentual
distinctions for Proto-Ryukyuan monosyllabic nouns. The
distinctions that I propose are high-level atonic (*/oo/ [HH(H)]),
high-register final accent (*/`oo˘/ [HH(L)]), and low-register final
accent (*/_oo˘/ [LH(L)]).
Compared with Myôgishô accentuation, which is thought to retain
archaic accentual distinctions, each correspondence set in (129)
regularly corresponds to Myôgishô accent classes 1.1, 1.2, and 1.3.
Furthermore, from the viewpoint of word-initial pitch height, most
of the cognates in the correspondence sets and Myôgishô cognates
match.

(129) Correspondences and reconstruction of Proto-Ryukyuan

accent for monosyllabic nouns
(a)
PA PO PM PY Yona58 PR
*oo *o˘o *_oo˘ *`oo˘ `oo *oo
HH(H) HL(L) LH(L) HH(L) HH(H) HH(H)
‘hair’ (1.1) *khï: *khi˘: *_ki:˘ *`ki:˘ `khi: *khï
‘blood’ (1.1) *ci: *ci˘: -- *`cï:˘ `ci: *ci:

(b)
PA PO PM PY Yona PR
*oo *o˘o *`oo˘ *`oo˘ `oo *`oo˘
HH(H) HL(L) HH(L) HH(L) HH(H) HH(L)
‘leaf’ (1.2) *ha: *pha˘: *`pa:˘ *`pa:˘ `ha: *`pha:˘
‘name’ (1.2) *na *na˘: *`na:˘ *`na:˘ `na: *`na:˘

58
‘Yona’ stands for Yonaguni.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 259

(c)
PA PO PM PY Yona PR
*oo˘ *oo *_oo˘ *_oo˘ _oo *_oo˘
LH(L) LL(L) LH(L) LH(L) LL(L) LH(L)
‘tree’ (1.3) *khï:˘ *khi: *_ki:˘ *_ki:˘ _khi: *_khï:˘
‘tooth’ (1.3) *ha:˘ *pha: *_pa:˘ *_pa:˘ _ha: *_pha:˘
‘eye’ (1.3) *mï:˘ *mi: *_mi:˘ *_mi:˘ _mi: *_mï:˘

Disyllables: A comparison of Proto-Amami, Proto-Okinawa,

Proto-Miyako, Proto-Yaeyama, and Yonaguni disyllabic accent
gives us three correspondences, which lead us to reconstruct three
accent classes for Proto-Ryukyuan disyllabic nouns. The
correspondence sets and the reconstructed accent classes are shown
in (130). Reconstructed individual nouns for each class are listed in
(131) below.

(130) Summary of accentual correspondences and reconstruction of

Proto-Ryukyuan accent for disyllables
PA PO PM PY Yona PR
a *OO *O˘O *`OO˘ *`OO˘ `OO *`OO˘
HH(H) HL(L) HH(L) HH(L) LH(H) HH(L)
b *OO˘ *OO *_OO˘ *_OO˘ _OO *_OO˘
LH(L) LL(L) LH(L) LH(L) LL(L) LH(L)
c *O˘O *oo˘O *OO *OO OOˆ *ooOˆ
HL(L) LHL(L) LL(L) LL(L) LF ~ LHF ~
LH(H) LHH(L)

On the basis of the correspondences, three accent classes are

reconstructed for Proto-Ryukyuan disyllabic nouns, i.e., */`OO˘/
[HH(L)], */_OO˘/ [LH(L)], and */ooOˆ/ [LHF ~ LHH(L)]. PR
*/`OO˘/ corresponds to Myôgishô classes 2.1-2, and both PR
*/_OO˘/ and */ooOˆ/ correspond to Myôgishô classes 2.3-5.
PR */`OO˘/ (130a) is identical to that of Proto-Miyako and
Proto-Yaeyama. PR */_OO˘/ (130b) shares the same accent form
with Proto-Amami, Proto-Miyako, and Proto-Yaeyama (although
the Proto-Amami form does not involve register distinctions). PR
260 ACCENTUAL HISTORY

*/ooOˆ/ (130) is similar to both the Proto-Okinawa and Yonaguni

forms.
As shown in (131), correspondences lack some cognates. A
number of missing cognates are seen especially in Yonaguni
Ryukyuan. This is either because my sources lack those cognates or
because they have been replaced by another word. Despite the lack
of those cognates, a reconstruction has been pursued on the basis of
other Ryukyuan cognates.

(131) Correspondences and reconstruction of Proto-Ryukyuan

accent for disyllabic nouns
(a)
PA PO PM PY Yona PR
*OO *O˘O *`OO˘ *`OO˘ `OO *`OO˘
HH(H) HL(L) HH(L) HH(L) LH(H) HH(L)
‘nose’ (2.1) *hana *pha˘na *`pana˘ *`pana˘ -- *`phana˘
‘cow’ (2.1) */usi */u˘si *`usï˘ *`usï˘ `uci *`/usi˘
‘bird’ (2.1) *thuri *thu˘i *`tuï˘ *`turï˘ -- *`thuri˘
‘loins’ (2.1) *khusï *ku˘si *`kusï˘ *`kusï˘ `khuci *`khusi˘
‘beard’ (2.1) *higi *pi˘zi *`pïgi˘ *`pïni˘ `Ngi *`pigi˘
‘wind’ (2.1) *khazï *ka˘zi *`kadi˘ *`kazi˘ `khadi *`khazï˘
‘stone’ (2.2) */isï */i˘si *`isï˘ *`isi˘ -- *`/isi˘
~*hi˘si
‘bridge’ (2.2) *hasi *pha˘si *`pasï˘ *`pasï˘ `haci *`phasi˘
‘paper’ (2.2) *khabi *ka˘bi *`kabi:˘ *`kabï˘ `khabi *`khabi:˘

(b)
PA PO PM PY Yona PR
*OO˘ *OO *_OO˘ *_OO˘ _OO *_OO˘
LH(L) LL(L) LH(L) LH(L) LL(L) LH(L)
‘flower’ (2.3) *hana˘ *phana *_pana˘ *_pana˘ _hana *_phana˘
‘mountain’ (2.3) *yama˘ *yama *_yama˘ *_yama˘ _dama *_yama˘
‘cloud’ (2.3) *khumu˘ *kumu *_fumu˘ *_fumu˘ _mmu *_kumu˘
‘dog’ (2.3) */in˘ */in *_in˘ *_inu˘ _inu *_/inu˘
‘shoulder’ (2.4) *khata˘ *khata -- *_kata˘ _kata *_khata˘
‘board’ (2.4) */ita˘ *hica *_icya˘ *_ita˘ _ita *_/ita˘
‘rain’ (2.5) */amï˘ *ami *_ami˘ *_a:mi˘ _ami *_/amï˘
‘sweat’ (2.5) */asï˘ *hasi *_asi˘ *_asi˘ _asi *_/asï˘
‘tears’ (3.5) *nada˘ *nada *_nada˘ *_na:da˘ _nuda *_na:da˘
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 261

(c)
PA PO PM PY Yng PR
*O˘O *oo˘O *OO *OO OOˆ *ooOˆ
HL(L) LHL(L) LL(L) LL(L) LF ~ LHF ~
LH(H) LHH(L)
‘bone’ (2.3) *hu˘nï *phu:˘ni *puni *puni huniˆ *phu:nïˆ
‘breath’ (2.4) */i˘ki */i:˘ci *ikï *ikï itiˆ */i:kiˆ
‘sea’ (2.4) */u˘mi */u:˘mi *i˘m -- -- */u:miˆ
‘boat’ (2.4) *hu:˘nï *phu:˘ni *funi *funi nniˆ *phu:nïˆ
‘mortar’ (2.4) */u˘si */u:˘si *usï *usï uciˆ */u:siˆ
‘chopsticks’ *ha˘si *pha:˘si -- *pasï haciˆ *pha:siˆ
(2.4)
‘needle’ (2.4) *ha:˘ri *pha:˘i *pai *parï haiˆ *pha:riˆ
‘shadow’ (2.5) *kha˘gï *kha:˘gi *kagi *kai khaNiˆ *kha:gïˆ
‘bridegroom’ *mu˘kwa *mu:˘ku *muku *muku muguˆ *mu:kuˆ
(2.5)
‘bucket’ (2.5) *wu˘khï */u:˘khi *wu:ki *u:ki ugiˆ *u:khïˆ
‘voice’ (2.5) *khu˘i -- *kui *kui khuiˆ *khu:iˆ

In (131), there are some irregular forms. First, there are some
monosyllables in the correspondence. The Proto-Amami, Proto-
Okinawa, and Proto-Miyako cognates for ‘dog’ are monosyllabic,
but in Proto-Yaeyama and Yonaguni the cognates are disyllabic.
Second, in (131b), the Proto-Yaeyama words *_a:mi˘ ‘rain’
(2.5) and *_na:da˘ ‘tears’ (3.5) contain a long vowel in the initial
syllable; none of words in this correspondence set has this feature.
The length in ‘tears’ can be explained by compensatory lengthening.
That is, the word used to be trisyllabic, but it lost a syllable in its
development. When the syllable was lost, the vowel in the initial
syllable became long. The vowel length of the word for ‘rain’ is
aberrant because no other words in this correspondence have this
feature. Thus, I consider the length as innovation.
For the correspondence in (131c), I have reconstructed vowel
length in the initial syllable. Proto-Okinawa cognates have a long
vowel in the initial syllable; on the other hand, nouns in Proto-
Amami, Proto-Miyako, Proto-Yaeyama, and Yonaguni generally
do not show initial-syllable vowel length. Exceptions to this are PA
*hu:˘nï ‘boat’ and *ha:ri ‘needle’, PM *wu:ki ‘bucket’, and PY
*u:ki ‘bucket’. They are remnants of Proto-Ryukyuan vowel length.
262 ACCENTUAL HISTORY

I thoroughly discuss the vowel length in relation to accentuation

below.
In this correspondence set, there are two irregular forms. First,
the accentuation of PM *i˘m ‘sea’ is irregular. This is initial accent,
but other Proto-Miyako words in the same correspondence set are
low-atonic. Moreover, the word is monosyllabic.
As discussed earlier, Hattori (1979a) points out that the
distinctions of the subclasses (i.e., 2.3-5a and 2.3-5b)
corresponding to traditional accent classes 2.3-5 need to be
reconstructed in Proto-Ryukyuan (see 5.3.3). As shown, I have
reconstructed two accent classes (PR */_OO˘/ [LH(L)] and */ooOˆ/
[LHF ~ LHH(L)]) which correspond to Myôgishô accent classes
2.3-5. More importantly, the reconstructed two classes correspond
to Hattori’s subclasses 2.3-5a and 2.3-5b.
According to the data, the subclasses in traditional accent classes
2.3-5 that Hattori (1979a) discovered are also seen in Kamishiro,
Kametsu, Ikema, Ôura, and Kuroshima. (132) gives examples of
Kametsu, Ôura, and Kuroshima cases along with Shuri and Nakijin.
With the exception of /ukï˘ ‘bucket’, Kametsu words in (132a) are
initial accent and ones in (132b) are final accent. Similar to the case
of Kametsu, Ôura nouns in (132a) are initial-accent and ones in
(132b) are final-accent. Kuroshima 2.3-5 nouns contrast in register;
high register in (132a) and low register in (132b). Nakijin 2.3-5
nouns in (132a) have initial accent and the ones in (132b) belong to
low-register. Unlike Nakijin, Shuri distinguishes the subgroups by
means of vowel length. That is, nouns in (132a) generally contain a
long vowel in the initial syllable except for _huni ‘bone’ and _huni
‘boat’, while ones in (132b) do not show the length at all.
The data in (132) suggests that the subclasses 2.3-5ab existed in
Proto-Ryukyuan. The distinctions have been retained in some
modern Ryukyuan dialects even though what distinguishes one
subclass from the other differs in each case.
With regard to the vowel length in Ryukyuan, Hattori (1959:
280-281, 327 and 1979b: 101-114) hypothesizes that vowel-length
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 263

distinctions in the initial syllable in Shuri nouns are inherited from

Proto-Ryukyuan,59 and furthermore, he assumes that the
dictinctions exixted in Proto-Japonic as well. In other words, Shuri
has retained Proto-Japonic vowel length, while Nakijin lost Proto-
Japoinc vowel length distinctions, but developed initial accent in
corresponding cognates. However, Hattori does not touch upon a
fundamental question why this occurred. In other words, Hattori’s
hypothesis is given without explaning why initial vowel length
must be original, while initial-syllable accent is secondary.

(132) Correspondence sets for 2.3-5 subclasses in Kametsu, Ôura,

Kuroshima, Shuri, and Nakijin
(a)
Kame Ôura Kuro Shuri Nakijin
‘bone’ (2.3) hu˘nI pu˘ni `puni _huni phu˘ni(:)
‘boat’ (2.4) hu˘nI fu˘ni `funi _huni phu˘ni(:)
‘needle’ (2.4) ha˘i pi˘ï `parï _ha:i pha˘i
‘mortar’ (2.4) /u˘sI u˘sï `usi _/u:si /u˘si(:)
‘breath’ (2.4) /I˘khI i˘kï `iki _/i:ci /i˘ci(:)
‘shadow’ (2.5) kha˘gˆ ka˘gi `hai _ka:gi ha˘gi(:)
‘bridegroom’ (2.5) -- mu˘ku `muku _mu:ku mu˘hu(:)
‘bucket’ (2.5) /ukï˘ u:˘ki `u:ki _u:ki hu˘khi(:)

(b)
Kame Ôura Kuro Shuri Nakijin
‘flower’ (2.3) hana˘ pana˘ _pana _hana _phana:
‘mountain’ (2.3) yama˘ yama˘ _yama _yama _yama:
‘board’ (2.4) -- icya˘ _ica _/ica _hica:
‘rain’ (2.5) /amˆ˘ ami˘ _ami _/ami _/ami:
‘sweat’ (2.5) /asI˘ asi˘ _asi _/asi _hasi:

Moreover, in relation to his vowel length hypothesis, Hattori

mentions that a similar phenomenon is also seen in Ainu, and

59
Hattori used the word ‘archaic form of Ryukyuan’, instead of ‘Proto-Ryukyuan’.
Since the existence of the vowel-length distinctions have to be hypothecized when
it comes to a reconstruction of Proto-Ryukyuan, I interpret ‘archaic form of
Ryukyuan’ to mean Proto-Ryukyuan.
264 ACCENTUAL HISTORY

contends that the vowel length is original.60 This reasoning of

Hattori’s is circuler, because it is based on his hypothesis on the
development of Ainu accent in relation to vowels, where he does
not explain why it is so.

(133) Hattori’s hypothesis on the initial-syllable vowel length

Proto-Ryukyuan
*CV:-
(e.g., *i:ci ‘breath’)61
 Ê
Shuri Nakijin
CV:- CV˘-
(e.g., i:ci ‘breath’) (e.g., /i˘ci ‘breath’)

As far as a reconstruction of Proto-Ryukyuan vowel length is

concerned, I certainly agree with Hattori. However, in contrast to
Hattori, I hypothesize a correlation between vowel length and
initial-syllable low register in earlier forms. There are four reasons
for this. First, a systematic correspondence exists between the
vowel length of some dialects (e.g., Shuri) and the initial accent of
other dialects (e.g., Nakijin). Second, a number of disyllabic nouns
belonging to traditional accent classes 2.3-5 have a long vowel in
the initial syllable and these nouns begin with a low pitch. More
specifically, as far as I know, it is Martin (1987: 247) who first

60
According to Hattori (1959 and 1979), initial-syllable vowel length in Sakhalin
Ainu corresponds to an accent in Hokkaidô Ainu. He claims that the vowel length
in Sakhalin is original and it has developed to the word initial accent in Hokkaidô.
The following examples are extracted from Hattori (1959: 85). See also 2.1 in
chapter 2 of this book.
‘thin’ ‘taste’ ‘hot’ ‘laugh’ ‘red’
Sakhalin a˘ne ke˘ra se˘sek mi˘na hu˘re
Hokkaidô aane keera seeseh miina huure
61
Based on Hattori’s claim, I have reconstructed a proto form for ‘breath’, because
Hattori does not provide us with reconstructed proto forms. Forms for Shuri and
Nakijin are taken from Hattori (1979).
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 265

suggested a correlation between initial-syllable vowel length and

initial low pitch by pointing out that Ryukyuan disyllabic nouns
with the initial vowel length belong to historical accent classes 2.3-
5. He also suggests that this vowel length be a reflex of original
vowel length in Japonic languages and that it might have been
retained in some Ryukyuan dialects. Martin has not developed an
actual reconstruction of the initial vowel length of Proto-Japonic
and Proto-Ryukyan. With regard to a reconstruction of Proto-
Japonic or Proto-Ryukyuan vowel length, following Martin,
Shimabukuro (1997) reconstructs Proto-Japonic initial-syllable
vowel length in relation to initial-low register. The third reason is
that a long vowel tends to have lower intrinsic pitch in comparison
with its short vowel (Shimabukuro 1998b). This characteristic
supports a correlation between a word-initial low pitch and a word-
initial vowel length. Finally, it is alos due to accountability of the
vowel length in relation with Whitman’s Law (see below for detail).
With regard to a correlation between initial-syllable vowel
length and low pitch, Shimabukuro (1997) has examined Okinawan
disyllabic nouns containing an initial long vowel and has found that
in general they are initially low in pitch. Examples are as follows.

(134)
Shuri Nakijin Onna Ô
‘tortoise’ _ka:mi: _ka:mi: -- _ka:mi:
‘ax’ _u:n _u:nu: -- _u:n
‘relations’ _na:ka na˘ha(:) _na:ka --

There are few exceptions, but they are explained as compounds,

borrowed words, or words with secondary vowel length (see also
Serafim 1994: 1). The following Shuri examples, taken from
Shimabukuro (1997), show some examples of secondary vowel
length.
With some exceptions, there are in general no 2.1 and 2.2 nouns
with initial-syllable vowel length in Shuri. The initial-syllable
vowel length found in 2.1 and 2.2 classes is monosyllabic, e.g., ta:
‘who, someone’. Nouns with original initial-syllable vowel length
266 ACCENTUAL HISTORY

are found only in classes 2.3, 2.4, and 2.5 in Shuri as well as in
Onna and Nakijin.
(135) Shuri examples for exceptions to a correlation between vowel
length and low pitch

`ku:ri ‘ice’ : This is a borrowed word. The

Ryukyus are located in a semi-
tropical area. No natural ‘ice’ is
found. The word for ‘ice’ in
Japanese is ko:ri HHH 3.1.

`ki:musi ‘caterpillar’ : This word is composed of the

words ki: ‘hair’ and musi ‘bug’.

`ka:ra ‘river’ : The initial-syllable vowel length is

due to a /w/ deletion between the
vowels in the syllable; kawara >
ka:ra.

As mentioned above, Whitman’s Law may account for a

phenomenon of Ryukyuan vowel length. Whitman’s /-r-/ deletion
rule (Whitman 1985: 21-23 and 1991) deletes /-r-/ when a
preceding vowel is short. However, there are a number of initially
low-pitched nouns with the syllable structure (C)Vr- in the Japonic
line. If Whitman’s hypothesis is correct, it is possible that those
nouns originally had a long vowel before the /-r-/ and somehow the
long vowel was shortened but word-initial low register remained.
Therefore, the word doro LL ‘mud’, for example, was *do:ro LLL
in Proto-Japonic. The following are Japonic nouns belonging to the
historical accent class 2.3, taken from Martin (1987) and compared
with Okinawan cognates.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 267

(136)
Tokyo Shuri Nakijin PJ
‘color’ iro _/iru _/iru: *i:ro
‘black’ kuro _kuru: _kuru: *ku:ro
‘paste’ nori _nui _nui *no:ri
‘mud’ doro _duru _duru: *do:ro/*nto:ro

A problem with this analysis is that cognates of this type in

Ryukyuan do not contain the expected initial long vowel. Further
research needs to be done regarding the lack of vowel length where
expected in the given subset of Okinawan examples. In addition, in
both Ôhama and Ishigaki, the word for ‘mud’ is du:ru. This is the
only evidence to support the form PJ *do:ro / *nto:ro.
Trisyllables: The reconstruction of Proto-Ryukyuan accent that
I have reconstructed is based on six sets of correspondences. In
(137), I have summarized the accentual correspondences and
reconstructed accent classes for Proto-Ryukyuan trisyllables. The
reconstructed classes are */`OOO/ (137a), */`OOO˘/ (137b),
*/_OOO˘/ (137c), */_OOO/ (137d), */OO˘O/ (137e), and */OOOˆ/
(137f). Reconstructed individual nouns for each class are listed in
(138).

(137) Summary of accentual correspondences and reconstruction of

Proto-Ryukyuan accent for trisyllables
PA PO PM PY Yona PR
a *OOO *O˘OO *`OOO˘ *OOO `OOO *`OOO
HHH(H) HLL(L) HHH(L) HHH(H) LHH(H) HHH(H)
b *`OOOˆ *`OOO˘
HHF ~ HHH(L)
c *OOO˘ *OO˘O *_OOO˘ HHH(L) _OOO *_OOO˘
LLH(L) HHL(L) LHH(L) LLL(L) LHH(L)
d *OOO *_OO˘ *OO *_OOO
LLL(L) LH(L) LL(L) LLH(H)
e *OO˘O *OO˘O *OOO *OO˘O OOOˆ *OO˘O
LHL(L) HHL(L) LLL(L) LHL(L) LHF ~ LHL(L)
f *OOOˆ *OO˘O *_OOOˆ LHH(L) *OOOˆ
LHF ~ HHL(L) LHF ~ LHF ~
LHH(L) LHH(L) LHH(L)
268 ACCENTUAL HISTORY

In the correspondences in (138), there are a number of

disyllables, which are historically trisyllabic, but became disyllabic
at some point in their development. Most of those words are found
in (138c), and there are six in (138a). There are also words with
irregular accent in (138). For instance, in (138a), the Proto-Amami
form for ‘centipede’ is *`muka˘zi, which is different from other
Proto-Amami nouns listed.
(138) Correspondence and reconstruction of Proto-Ryukyuan
accent for trisyllabic nouns
(a)
PA PO PM PY Yonaguni PR
*OOO *O˘OO *`OOO˘ *OOO `OOO *`OOO
HHH(H) HLL(L) HHH(L) HHH(H) LHH(H) HHH(H)
‘mark’ (3.1) *sirusi *si˘rusi *`sïrusï˘ *sirusï `ciruci *`sirusi
‘smoke’ (3.1) *khïbusi *khi˘busi *`kifusï˘ *kibusï `khibunci *`khïbusi
‘yawn’ (3.1) *akubi */a˘kubi *`afugï˘ *akubï `agui *`/akubi
‘forehead’(3.1) -- *phi˘ce: *`futai˘ *pitai `tai *`phitai
‘two’ (3.2) *ta:cï HHH -- *`futa:cï˘ *futa:cï `ta:ci *`futacï
‘centipede’(3.2) *`muka˘zi *mu˘kazi *`m:kazi˘ *_mukazaˆ Nkadiˆ *`mukadi
‘cherry’ (3.2) *sakura *sa˘kura -- -- -- *`sakura

(b)
PA PO PM PY Yonaguni PR
*`OOOˆ
*OOO *O˘OO *`OOO˘ `OOO *`OOO˘
HHF ~
HHH(H) HLL(L) HHH(L) LHH(H) HHH(L)
HHH(L)
‘strength’(3.3) *cikyara *ci˘kara -- *`cïkaraˆ `sikara *`cikara˘
‘gold’ (3.3) *khugani *khu˘gani *`kugani˘ *`kuganiˆ `khuNani *`khugani˘

(c)
PA PO PM PY Yonaguni PR
*`OOOˆ
*OOO˘ *OO˘O *_OOO˘ _OOO *_OOO˘
HHF ~
LLH(L) HHL(L) LHH(L) LLL(L) LRH(L)
HHH(L)
‘mirror’ (3.4) *khagami˘ *khaga˘mi: *`kagam˘ *_kaga˘nï _kaNan *_khagami˘
‘treasure’(3.4) *thakara˘ *thaka˘ra *_takara˘ *_takaraˆ _tagara *_thakara˘
‘fan’ (3.4) */o:˘gi */o:zi LLL *_augï˘ *o:gi -- *_augi˘
‘bag’ (3.4) *`huku˘ru *phuku˘ru *_fukuru˘ *_hukuruˆ _ku:ru *_pukhuru˘
‘pillow’ (3.5) *makura˘ *maQkwa *_maQfa˘ *_mafuraˆ _magura *_makura˘
‘life’ (3.5) */inuci˘ */inu˘ci *n:ucï *_inu˘cï nuti˘ *_/inuci˘
‘fat’ (3.5) */abura˘ */anda LLL *_aQva˘ *_aburaˆ _anda *_/abura˘
‘pillar’ (3.5) *`hasi˘ra *phasi˘ra *para *para: hira˘ *_phasira˘
‘cousin’ (3.5) */itoko˘ *hicu˘ku *icufu *_icïkuˆ itigu˘ *_/itoko˘
‘heart’ (3.5) -- *khuku˘ru -- *_kukuruˆ _kuguru *_khukhuru˘
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 269

(d)
PA PO PM PY Yonaguni PR
*OOO˘ *OOO *_OO˘ *OO _OOO *_OOO
LLH(L) LLL(L) LH(L) LL(L) LLL(L) LLH(H)
‘calendar’(3.4) *kuyumi˘ *khuyumi *_kuyum˘ *kuyun kuyumi *_khuyumi

(e)
PA PO PM PY Yonaguni PR
OOOˆ
*OO˘O *OO˘O *OOO *OO˘O *OO˘O
LHF ~
LHL(L) HHL(L) LLL(L) LHL(L) LHL(L)
LHH(L)
‘eel’ (3.6) *_/una˘gi */una˘zi *unagï *_unagiˆ -- */una˘gi
‘earthworm’
*_më(C)ë˘za *mimin˘zi *mi:mizï *_mi:mi˘zï -- *mi:min˘zï
(3.6)

(f)
PA PO PM PY Yonaguni PR
*OOOˆ *_OOOˆ OOOˆ *OOOˆ
*OO˘O *OOO
LHF ~ LHF ~ LHF ~ LHF ~
HHL(L) LLL(L)
LHH(L) LHH(L) LHH(L) LHH(L)
‘whale’ (3.7) *kuziraˆ *gunzi˘ra *kuzira *_kuziraˆ khudiraˆ *khudiraˆ
‘field’ (3.7) -- *hata˘ki *patagi *_fusurïˆ hatagiˆ *hatakiˆ
‘medicine’
*khusuriˆ *khusu˘i *fusuï *_mi:mi˘zï cu:riˆ *khusuriˆ
(3.7)

In addition, our data lack Proto-Amami cognates for ‘forehead’

and ‘field’, a Proto-Okinawa cognate for ‘two’, a Proto-Miyako
cognate for ‘strength’, or Yonaguni cognates for ‘fan’, ‘eel’, and
‘earthworm’. They are simply not found in the sources.
Looking into the Proto-Ryukyuan accent system as a whole,
there are five types of accentuation: high- and low-register atonic,
high- and low-register final accent, final-falling accent, and
penultimate accent. Among those, monosyllables and disyllables
make use of only three types. All accentual types are seen in
Trisyllables.
Typologically, Proto-Ryukyuan accent is a locus-register system.
As illustrated below, register distinguishes two types of final-
accent nouns throughout the system. However, in trisyllables,
register also distinguishes high-register from low-register atonic
nouns.
270 ACCENTUAL HISTORY

(139) List of Proto-Ryukyuan accent patterns according to their

types
Monosyllables Disyllables Trisyllables
high atonic *oo *`OOO
low atonic *_OOO
high-initial *`oo˘ *`OO˘ *`OOO˘
final accent
low-initial *_oo˘ *_OO˘ *_OOO˘
final accent
final falling *ooOˆ *OOOˆ
penultimate *OO˘O

Also characteristic of Proto-Ryukyuan nouns is that a number

have a long vowel. As in most of the Ryukyuan dialects,
monosyllables are bimoraic. For disyllables, nouns with final-
falling accent have a long vowel in the initial syllable, namely
*/ooOˆ/. Unlike its monosyllables and disyllables, Proto-Ryukyuan
trisyllables can have a long vowel (or two-mora-long syllable) in
the second syllable of words. Nouns with low-register atonic and
with penultimate accent do not have the length.

5.8 DEVELOPMENT OF PROTO-AMAMI, PROTO-

OKINAWA, PROTO-MIYAKO, PROTO-YAEYAMA, AND
YONAGUNI ACCENT FROM PROTO-RYUKYUAN
This section describes how the Proto-Amami, Proto-Okinawa,
Proto-Miyako, Proto-Yaeyama, and Yonaguni accent systems
evolved from Proto-Ryukyuan accent. As the summary of Proto-
Ryukyuan accent below illustrates, there are three distinctions for
monosyllables, three for disyllables, and six for trisyllables. From
the viewpoint of accent typology, Proto-Ryukyuan in general
utilizes locus and register. Some modern Ryukyuan dialects have
retained the locus distinctions, some have lost the register
distinctions, and others have still kept both the locus and register
features.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 271

(140) Proto-Ryukyuan accent

Monosyllables Disyllables
(a) *oo [HH(H)] (a) *`OO˘ [HH(L)]
(b) *`oo˘ [HH(L)] (b) *_OO˘ [LH(L)]
(c) *_oo˘ [LH(L)] (c) *ooOˆ [LHF ~ LHH(L)]

Trisyllables
(a) *`OOO [HHH(H)]
(b) *`OOO˘ [HHH(L)]
(c) *_OOO˘ [LHH(L)]
(d) *_OOO [LLH(H)]
(e) *OO˘O [LHL(L)]
(f) *OOOˆ [LHF ~ LHH(L)]

5.8.1 Development of Proto-Amami Accent

The development of Proto-Amami accent is quite simple -- one
change took place in monosyllables, two in disyllables, and three in
trisyllables.
From the viewpoint of the typology of accent, Proto-Amami lost
the Proto-Ryukyuan register distinctions and vowel length. The
loss of the register distinctions was due to mergers of accent classes.
The following discussion will present the development of Proto-
Amami accent in detail.
Monosyllables: In the development of Proto-Amami accent,
Proto-Ryukyuan monosyllables underwent Final Accent Loss,
applicable only to high-initial words. This sound change made a
three-way distinction of Proto-Ryukyuan accent a two-way system
in the Proto-Amami line by merging PR */oo/ (141a) and */`oo˘/
(141b) as /oo/. Because of this, the register distinction of Proto-
Ryukyuan was lost.
272 ACCENTUAL HISTORY

(141) Development of Proto-Amami monosyllabic accent

PR F-loss
(a) *oo
> oo
(b) *`oo˘

(c) *_oo˘ ≡ oo˘

Disyllables: By undergoing Final Accent Loss and Vowel

Shortening (V-shrt), Proto-Ryukyuan disyllabic accent evolved to
Proto-Amami accent. First, Final Accent Loss, applicable only to
words with high initial pitch, cancelled the accent of PR */`OO˘/
(142a). At the same time, the Proto-Ryukyuan register distinction
was lost. At the next stage, PR */ooOˆ/ (142c) became /O˘O/ by the
rule of Vowel Shortening. When it took place, final-falling accent
became (/ooOˆ/) became penultimate accent (/O˘O/). This is
probably because the final falling pitch is possible only for longer
words (at least three-moras long). Since /ooOˆ/ became bimoraic,
the final-falling pitch was lost.

(142) Development of Proto-Amami disyllabic accent

PR F-loss V-shrt
(a) *`OO˘ > OO = OO

(b) *_OO˘ ≡ OO˘ = OO˘

(c) *ooOˆ = ooOˆ > O˘O

Trisyllables: As my hypothesis on the development of Proto-

Amami trisyllabic accent in (143) suggests, Proto-Ryukyuan accent
underwent two changes, i.e., Final Accent Loss and Abductive
Change.
The first rule that had an effect on Proto-Amami trisyllabic
accent was Final Accent Loss. As mentioned earlier, it is applicable
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 273

to high-initial words, but not to low-initial words. Thus, PA

*/`OOO˘/ (143b) underwent the change, and merged with PR
*/`OOO/ (143a). Because of this change, the initial pitch of PR
*/_OOO˘/ (143c) became nondistinctive. At the next stage, /OOO˘/
(143c) and PR */_OOO/ (143d) merged as /OOO˘/[LLH(L)]. This
also made the initial pitch of /`OOO/ no longer distinctive. At the
time of the change, speakers of the Proto-Amami line must have
considered these two classes in (143c) and (143d) as one because
they are very similar at the phonetic level. With this change, the
Proto-Ryukyuan register system was completely lost.

(143) Development of Proto-Amami trisyllabic accent

abductive
PR F-loss change
(a) *`OOO
> `OOO ≡ OOO
(b) *`OOO˘

(c) *_OOO˘ ≡ OOO˘

[LRH(L)]
> OOO˘
(d) *_OOO = _OOO
[LLH(H)]

(e) *OO˘O = OO˘O = OO˘O

[LFL(L)]

(f) *OOOˆ = OOOˆ = OOOˆ

5.8.2 Development of Proto-Okinawa Accent

Compared with the Proto-Ryukyuan accent system, Proto-Okinawa
accent is much simpler. Proto-Ryukyuan has more distinctions than
Proto-Okinawa. The complex system of Proto-Ryukyuan evolved
into that of Proto-Okinawa by a series of five changes. From the
point of view of accentual typology, the locus-register system of
Proto-Ryukyuan became simplified to a locus system in the Proto-
274 ACCENTUAL HISTORY

Okinawa line. The development of Proto-Okinawa accent from

Proto-Ryukyuan will be explained below.
Monosyllables: There are two sound changes contributing to the
development of Proto-Okinawa monosyllabic accent. They are
Final Accent Loss and Initial Accent Gain. The former made the
three-way distinction of Proto-Ryukyuan monosyllabic accent a
two-way system by merging PR */oo/ (144a) and */`oo˘/ (144b) as
/`oo/. Simultaneously, PR */_oo˘/ (144c) also lost its final accent,
becoming low-atonic. And the other change, Initial Accent Gain,
gave an accent on the initial syllable of high-register nouns (/`oo/).
This rule is not applicable to low-initial words, and thus /_oo/ did
not undergo the change. However, its initial pitch became no longer
distinctive.

(144) Development of Proto-Okinawa monosyllabic accent

PR F-loss I-gain
(a) *oo
> `oo > o˘o
(b) *`oo˘

(c) *_oo˘ > _oo ≡ oo

Disyllables: From Proto-Ryukyuan, Proto-Okinawa disyllabic

accent evolved by undergoing Final Accent Loss, Contour Pitch
Simplification, and Initial Accent Gain.

(145) Development of Proto-Okinawa disyllabic accent

PR F-loss C-smpl I-gain

(a) *`OO˘ > `OO = `OO > O˘O

(b) *_OO˘ > _OO = _OO ≡ OO

(c) *ooOˆ = ooOˆ > oo˘O = oo˘O

 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 275

The first rule converted PR */`OO˘/ (145a) and */_OO˘/ (145b)

into /`OO/ and /_OO/ respectively. Undergoing the second change,
PR */ooOˆ/ became /oo˘O/. Finally, the third rule created an accent
on the initial syllable of /`OO/, resulting in /O˘O/. When this
happened, /_OO/ was indirectly affected by the rule. Its initial pitch
became nondistinctive; the Proto-Ryukyuan register system was
lost.
Trisyllables: As shown in (146), Proto-Okinawa trisyllabic
accent evolved from Proto-Ryukyuan by undergoing Abductive
Change, assimilation, Contour Pitch Simplification, and Initial
Accent Gain.

(146) Development of Proto-Okinawa trisyllabic accent

abductive
PR change assml C-smpl I-gain
(a) *`OOO
> `OOO = `OOO = `OOO > O˘OO
(b) *`OOO˘

(c) *_OOO˘ ≡ OOO˘

> OOOˆ
(f) *OOOˆ = OOOˆ
> OO˘O = OO˘O
(e) *OO˘O = OO˘O = OO˘O

(d) *_OOO = _OOO = _OOO = _OOO ≡ OOO

First, PR */`OOO/ (146a) and */`OOO˘/ (146b) merged as

/`OOO/. Along with the change, the initial pitch of PR */_OOO˘/
(146c) became nondistinctive. And the following rule turned
/OOO˘/ (146c) into /OOOˆ/. As a result, it merged with PR
*/OOOˆ/ (146f) merged as /OOOˆ/. When Contour Pitch
Simplification converted /OOOˆ/ (146cf) to /OO˘O/, it merged with
PR */OO˘O/ (146e). Finally, with the Initial Accent Gain rule,
/`OOO/ became /O˘OO/. The rule does not apply to low-initial
276 ACCENTUAL HISTORY

words; thus, /_OOO/ did not undergo the change. However, it lost
its word-initial register distinction.
Ordering of some of the changes is critical to account for the
development of Proto-Okinawa accent. First, Final Accent Loss
must precede Initial Accent Gain. Second, assimilation has to
precede Final Accent Loss. Third, Abductive Change must have
taken place before assimilation. Finally, Contour Pitch
Simplification must have occurred after assimilation. The changes
took place when conditions were met. Therefore, some types of
nouns underwent a change and others did not undergo the same
change.

5.8.3 Development of Proto-Miyako Accent

Unlike Proto-Amami and Proto-Okinawa, Proto-Miyako has kept
the register distinctions of Proto-Ryukyuan, although it underwent
a complex development as the other two dialects did. I will account
for the development of Proto-Miyako monosyllabic, disyllabic, and
trisyllabic accent in detail below.
Monosyllables: As (147) shows, the development of Proto-
Miyako monosyllables involves two changes: Initial Pitch
Lowering and Abductive Change. By undergoing Initial Pitch
Lowering, PR */oo/ [HH(H)] (147a) became LH(H). Because of its
pitch shape, it merged with PR */_oo˘/ (147c). The Initial Pitch
Lowering rule did not affect PR */`oo˘/ (147b) because its initial
pitch was distinctive at the time when the change was occurring.
Eventually, the three distinctive accent classes resulted in two.

(147) Development of Proto-Miyako monosyllabic accent

abductive
PR I-low change
(a) *oo [HH(H)] > oo [LH(H)]
> _oo˘
(c) *_oo˘ = _oo˘

(b) *`oo˘ = `oo˘ = `oo˘

 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 277

Disyllables: The development of Proto-Miyako disyllabic accent

is quite simple. As noticed, PR */ooOˆ/ (148c) is the only one that
underwent changes, while other Proto-Ryukyuan classes retained
their accent. First, PR */ooOˆ/ underwent Vowel Shortening. As a
result, it became /OOˆ/ [LF ~ LH(L)]. Following that change,
Contour Pitch Simplification turned /OOˆ/ to /OO/ [LH(H)] (see
4.1.1). Finally, its pitch became low-level.
With regard to accent typology, in the Proto-Miyako line, the
Proto-Ryukyuan register system has been kept.

(148) Development of Proto-Miyako disyllabic accent

PR V-shrt C-smpl Leveling

(a) *`OO˘ = `OO˘ = `OO˘ = `OO˘

(b) *_OO˘ = _OO˘ = _OO˘ = _OO˘

(c) *ooOˆ > OOˆ > OO > OO

[LHF ~ [LF ~ LH(L)] [LH(H)] [LL(L)]
LHH(L)]

Trisyllables: Unlike the development of its monosyllabic and

disyllabic accent, the development of Proto-Miyako trisyllabics is
complex. There are four changes: Abductive Change, again,
Abductive Change, Contour Pitch Simplification, and Leveling.
Note that Abductive Change took place twice, but they affected
different classes at different stages.
First, Abductive Change took place. It merged two classes with
falling pitch, i.e., PM */OO˘O/ (149e) and */OOOˆ/ (149f). At the
second stage, another Abductive Change occurred. This time it put
/`OOO/ (149a) and /`OOO˘/ (149b) into one class, namely
/`OOO˘/. The change also caused a merger of /_OOO˘/ (149c) and
/_OOO/ (149d). Contour Pitch Simplification brought a change to
/OOOˆ/ [LLF ~ LLH(L)], making it /OOO/ [LLH(H)]. And it
underwent the final change Leveling, resulting in /OOO/ [LLL(L)].
278 ACCENTUAL HISTORY

(149) Development of Proto-Miyako trisyllabic accent

abductive abductive
PR change change C-smpl Leveling
(a) *`OOO = `OOO
> `OOO˘ = `OOO˘ = `OOO˘
(b) *`OOO˘ = `OOO˘

(c) *_OOO˘ = _OOO˘

> _OOO˘ = _OOO˘ = _OOO˘
(d) *_OOO = _OOO
[LLH(H)]

(e) *OO˘O
> OOOˆ = OOOˆ > OOO > OOO
(f) *OOOˆ [LLF ~ [LLH(H)] [LLL(L)]
LLH(L)]

In Proto-Ryukyuan, there were six accent distinctions for

trisyllables, but in Proto-Miyako there were three. Half of the
distinctions were lost in the Proto-Miyako line. Although Proto-
Miyako lost some of Proto-Ryukyuan characteristics, Proto-
Miyako has still retained some of the Proto-Ryukyuan register
distinctions even though they are not identical to that of Proto-
Ryukyuan.

5.8.4 Development of Proto-Yaeyama Accent

Compared with Proto-Amami, Proto-Okinawa, and Proto-Miyako,
Proto-Yaeyama underwent one of the simplest developments as a
whole system. Although the Proto-Yaeyama trisyllabic accent
system is drastically different from that of Proto-Ryukyuan, Proto-
Yaeyama monosyllables and disyllables largely preserved the
Proto-Ryukyuan accentual distinctions. In addition, it should be
pointed out that not only is the Proto-Yaeyama disyllabic accentual
system identical to that of Proto-Miyako but also the developments
of Proto-Yaeyama and Proto-Miyako disyllabic accent are identical.
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 279

Monosyllables: There is only one change in the development of

Proto-Yaeyama monosyllabic accent, Abductive Change. This
caused a merger of PR */oo/ (150a) and */`oo˘/ (150b) as /`oo˘/.
PR */_oo˘/ (150c) stayed the same. Although the high-atonic
accent was lost, the Proto-Ryukyuan register distinction remained
the same.

(150) Development of Proto-Yaeyama monosyllabic accent

abductive
PR change
(a) *oo
> `oo˘
(b) *`oo˘

(c) *_oo˘ = _oo˘

Disyllables: As (151) shows, there are three stages that

contributed to the development of Proto-Yaeyama accent.
Interestingly, PR */ooOˆ/ (151c) is the only one that underwent the
changes, while PR */`OO˘/ (151a) and */_OO˘/ stayed unchanged.
During the first stage, PR */ooOˆ/ became /OOˆ/ [LF ~ LH(L)].
The following rule turned it to /OO/ [LH(H)]. Finally, it became
/OO/ [LL(L)].

(151) Development of Proto-Yaeyama disyllabic accent

PR V-shrt C-smpl Leveling

(a) *`OO˘ = `OO˘ = `OO˘ = `OO˘

(b) *_OO˘ = _OO˘ = _OO˘ = _OO˘

(c) *ooOˆ > OOˆ > OO > OO

[LHF ~ [LF ~ LH(L)] [LH(H)] [LL(L)]
LHH(L)]
280 ACCENTUAL HISTORY

Trisyllables: Proto-Ryukyuan trisyllabic accent has evolved to

Proto-Yaeyama accent by undergoing Initial Accent Gain, Accent
Shift, and assimilation. Any of these changes did not have any
effect on monosyllables and disyllables.

(152) Development of Proto-Yaeyama trisyllabic accent

PR I-gain shift assml

(a) *`OOO ≡ OOO = OOO = OOO

(b) *`OOO˘ = `OOO˘ = `OOO˘ > `OOOˆ

[HHF ~ HHH(L)]

(c) *_OOO˘ = _OOO˘ = _OOO˘

> _OOOˆ
(f) *OOOˆ = OOOˆ = OOOˆ [LHF ~ LHH(L)]

(d) *_OOO > O˘OO

> OO˘O = OO˘O
(e) *OO˘O = OO˘O

First, PR */_OOO/ (152d) underwent Initial Accent Gain,

resulting in /O˘OO/. When the following change shifted its accent
onto the second syllable, making it /OO˘O/, a merger of /OO˘O/
(152d) and /OO˘O/ (152e) occurred. Finally, under the effect of
assimilation, /`OOO˘/ [HHH(L)] (152b) and /_OOO˘/ [LHH(L)]
(152c) became /`OOOˆ/ [HHF ~ HHH(L)] and /_OOOˆ/ [LHF ~
LHH(L)] respectively. Furthermore, the latter then merged with
/OOOˆ/ [LHF ~ LHH(L)] (152f).
The ordering of some changes is crucial to the development of
Proto-Yaeyama. The changes seen in the development of Proto-
Yaeyama trisyllabic accent took place as shown in (152). Contour
Pitch Simplification must have been followed by assimilation. If
they were ordered in the other direction, /`OOOˆ/ (152b) and
/_OOOˆ/ (152cf) would have become /`OO˘O/ and /_OO˘O/
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 281

respectively. Concernig Abductive Change, the rule is ordered after

assimilation. In this way, trisyllabic high-register atonic and high-
register final accent would not be able to merge, as they did in
monosyllables.
Proto-Yaeyama has a tendency to avoid an accent on the third
syllable, cf. this phenomenon is also seen in the developments of
Myôgishô and of Kyoto dialects (see chapter 6). Assimilation was
applicabe only in trisyllables.

5.8.5 Development of Yonaguni Accent

The development of Yonaguni accent as a whole involves relatively
complex stages of changes. While monosyllables underwent only
one sound change, disyllables underwent three changes and
trisyllables underwent four changes, including phonetic changes.
From the point of view of accent typology, Proto-Ryukyuan
register was kept in the Yonaguni dialect, although the register
distinction is among atonic nouns after the loss of final accent.
Monosyllables: Yonaguni monosyllables developed their accent
system from Proto-Ryukyuan accent in a simple manner. When
Final Accent Loss canceled the final accent of both PR */`oo˘/
(153b) and */_oo˘/ (153c), the former merged with PR */oo/,
becoming /`oo/ in contrast with /_oo/ (< PR */_oo˘/).

(153) Development of Yonaguni monosyllabic accent

PR F-loss
(a) *oo
> `oo
(b) *`oo˘

(c) *_oo˘ > _oo

Disyllables: Comparing Yonaguni accent with Proto-Ryukyuan

accent below, differences are seen in classes in (154a) and (154b).
That is, Proto-Ryukyuan final-accent (*/`OO˘/ and */_OO˘/)
became atonic under the effect of Final Accent Loss. Notice that
282 ACCENTUAL HISTORY

the register distinction remains unchanged. Moreover, /`OO/ (< PR

*/`OO˘/ [HH(L)]) lowered the pitch of the initial syllable by Initial
Pitch Lowering.
As far as accentuation is concerned, both PR */ooOˆ/ and
Yonaguni /*OOˆ/ have final-falling accent; that is, they are
identical. In both systems, the length is not distinctive. In the
development of Yonaguni accent, it underwent Vowel Shortening.

(154) Development of Yonaguni disyllabic accent

PR V-shrt F-loss I-low

(a) *`OO˘ = `OO˘ > `OO > `OO
[HH(H)] [LH(H)]

(b) *_OO˘ = _OO˘ > _OO = _OO

(c) *ooOˆ > OOˆ = OOˆ = OOˆ

[LF ~
LH(H)]

Trisyllables: In the development of its trisyllabic accent,

Yonaguni underwent three changes. With these changes, six
distinctive accent classes of Proto-Ryukyuan became three in the
Yonaguni line. Despite this drastic change, Yonaguni has kept the
original Proto-Ryukyuan register distinction in atonic nouns.
The first change that took place was Final Accent Loss. It caused
a merger of /`OOO/ (155a) and /`OOO˘/ (155b) as /`OOO/, and
also another merger of /_OOO˘/ (155c) and /_OOO/ (155d) as
/_OOO/. The last two stages in the development are Accent Shift
and assimilation. Undergoing Accent Shift, /OO˘O/ (155e) became
final-accent (i.e., /OOO˘/). Finally, assimilation changed it to
/OOOˆ/ [LHF ~ LHH(H)]. As a result, it merged with PR */OOOˆ/.
It should be pointed out that there are no final-accent nouns in
Yonaguni, although there are final-falling nouns. That is, Proto-
Ryukyuan final-accent classes were eliminated in the development.
This is similar to Proto-Yaeyama trisyllabic accent. In Proto-
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 283

Yaeyama, there are final-accent monosyllables and disyllables, but

there are no final-accent trisyllables.

(155) Development of Yonaguni trisyllabic accent

PR F-loss shift assml

(a) *`OOO
> `OOO = `OOO = `OOO
(b) *`OOO˘

(c) *_OOO˘
> _OOO = _OOO = _OOO
(d) *_OOO
[LLH(H)]

(e) *OO˘O = OO˘O > OOO˘

> OOOˆ
(f) *OOOˆ = OOOˆ = OOOˆ [LHF ~ LHH(H)]

5.9 CONCLUSION
In this chapter, the accentuation of modern Ryukyuan dialects has
been discussed. Based on that, I have attempted to reconstruct the
Proto-Amami, Proto-Okinawa, Proto-Miyako, and Proto-Yaeyama
accent systems. Then, on the basis of these reconstructions, I have
proposed a reconstruction of Proto-Ryukyuan accent (see (140)
above). The Proto-Ryukyuan accent system utilizes locus and
register distinctions. Proto-Ryukyuan has falling accent, such as
/ooOˆ/. There are three distinctions for monosyllables and
disyllables, and six distinctions for trisyllables in the system.
Regarding word structure, monosyllables are bimoraic. As for
disyllables, nouns with falling-accent have a long vowel in the
initial syllable. Of the six accent classes of the trisyllables, none of
them has a long vowel.
This chaper has also shown how Proto-Amami, Proto-Okinawa,
Proto-Miyako, Proto-Yaeyama, and Yonaguni accent systems
evolved from Proto-Ryukyuan accent. As they evolved, some of
the Proto-Ryukyuan accent classes underwent changes; others
284 ACCENTUAL HISTORY

remained the same. For example, in the Proto-Yaeyama, PR

*/ooOˆ/ became /OO/, but the other two classes */`OO˘/ and
*/_OO˘/ stayed unchanged. Compared with the number of
accentual distinctions in Proto-Ryukyuan, the five descendants
have fewer distinctions than those of Proto-Ryukyuan, except for
disyllabic accent. Interestingly there is no merger in the
development of the descendant disyllabic accent.
With regard to accentual typology, similar to Proto-Ryukyuan
accent, the accent system of Proto-Miyako, Proto-Yaeyama, and
Yonaguni is basically a locus-register system. By contrast, the
accent system of Proto-Ryukyuan evolved to a locus system in
Proto-Amami and Proto-Okinawa. Furthermore, there are no
descendant proto-languages that evolved into a pure register system.
One of the characteristics seen in Proto-Ryukyuan is a ‘long
syllable’, either a long vowel or two-mora-long syllable. As far as
monosyllables are concerned, all the proto descendants kept their
monosyllables bimoraic. However, Proto-Okinawa is the only one
that retained Proto-Ryukyuan length in disyllables and trisyllables
as well.
As for final falling-accent, it is seen in Proto-Amami, Proto-
Miyako, Proto-Yaeyama, and Yonaguni. However, Proto-Okinawa
does not have the feature. The difference is reflected in the ordering
of Contour Pitch Simplification and assimilation. That is, due to the
application of Contour Pitch Simplification preceded by
assimilation, there is no final-falling accent. Moreover, in Proto-
Miyako and Proto-Yaeyama, the rule of assimilation did not apply
to shorter nouns such as monosyllables and disyllables; the rule
affected trisyllables. Therefore, final-falling accent is seen only in
trisyllables.
Finally, in the developments of Proto-Amami, Proto-Okinawa,
Proto-Miyako, Proto-Yaeyama, and Yonaguni, some of the changes
are conditioned. For example, in Proto-Amami, Final Accent Loss
is applicable only to high-initial nouns. Looking into those
conditioned changes, there is an interesting characteristic among
the descendant languages. As mentioned above, in Proto-Amami,
Final Accent Loss is conditioned to high-initial nouns. In Proto-
 RECONSTRUCTION OF PROTO-RYUKYUAN ACCENT 285

Okinawa, Initial Accent Gain is applicable only to high-initial

nouns. On the other hand, Initial Accent Gain is applicable only to
low-initial nouns in Proto-Yaeyama. In both Proto-Miyako and
Yonaguni, Accent Shift affects low-initial nouns. (Note that there
are no high-initial tonic nouns. Therefore, it is not known whether
the change is applicable to high-initial. However, it is certain that
the rule is at least applicable to low-initial nouns). That is, it seems
that, as far as application of rules is concerned, Proto-Amami and
Proto-Okinawa are conditioned to high-initial pitch, and that Proto-
Miyako, Proto-Yaeyama, and Yonaguni are sensitive to low-initial
pitch.
 6

RECONSTRUCTION OF
PROTO-MAINLAND
JAPANESE ACCENT
6.1 INTRODUCTION
his chapter discusses a reconstruction of Proto-Mainland
T Japanese accent and explains how Proto-Mainland Japanese
accent evolved into its descendants. ‘ Mainland Japanese’
refers to Japanese dialects spoken on the islands of Hokkaidô 1,
Honshû, Shikoku, and Kyûshû, excluding the islands of Amami
and the islands belonging to Okinawa Prefecture. Mainland
Japanese dialects can be classified into three types in terms of
accent. These three types are Eastern-type (or Tokyo-type)
accent, Western-type (or Kyoto-type) accent, and Southwestern-
type (or Kagoshima-type) accent. A reconstruction of Proto-
Mainland Japanese has been pursued using the Tokyo dialect for
Eastern-type Japanese, Kyoto dialect for Western-type Japanese,
and Kagoshima for Southwestern-type Japanese.
In section 6.2 accent systems of modern Tokyo, Kyoto, and
Kagoshima are synchronically analyzed. In section 6.3 all
correspondences among the dialects and reconstructions of an
accent class for each correspondence set are listed. The
following section, 6.4, demonstrates how the accent systems of
the mainland Japanese dialects have evolved from the
reconstructed Proto-Mainland Japanese accent.

1
Hokkaido was first settled by people from Tôhoku (northeast part of Honshû)
in the Kamakura era (14th century) and around the era of Edo (17th century)
people from different parts of Japan started to immigrate to the place.
(Hirayama ed. 1992: 54)
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 287

All data are primarily taken from Hirayama (1960). However,

accent patterns in NHK Hôsô Bunka Kenkyûjo, ed. (1971) are
also mentioned when they are different from those of
Hirayama’s.

6.2 MAINLAND JAPANESE

In this book I discuss three types of Japanese accent following
Hirayama, ed. (1992)2: ‘ Tokyo-type’ , ‘ Kyoto-type’ , and
‘ Kagoshima-type’ . This classification is mainly based on the
number of accent classes, the merging patterns of accent classes
corresponding to Myôgishô accent, and phonological or
phonetic rules (e.g., lowering high pitch on the unaccented
word-initial mora in Tokyo-type accent).
In the Tokyo-type dialects, traditional accent classes for
disyllables, for example, have merged as follows: 2.1/2.2-3/2.4-5.
That is, there are three accent classes for Tokyo disyllabic nouns.
Another characteristic, as mentioned earlier, in Tokyo-type
accent, is a rule that lowers high pitch on an unaccented word-
initial mora. I use mainly Tokyo dialect as an example for
Eastern-type accent, yet other Tokyo-type dialects such as
Hiroshima dialect are also mentioned when necessary.
A number of Kyoto-type dialects have four accent classes for
disyllabic nouns.3 Here I focus on Kyoto dialect as a
representative of Western-type accent. In the Kyoto dialect there
are four accent classes for disyllabic nouns, and they correspond
to traditional accent classes as follows: 2.1/2.2-3/2.4/2.5. The
second class resulted from a merger of two accent classes,
namely 2.2 and 2.3.
I also use the data from Kagoshima dialect, whose accent is
Southwestern-type. This dialect has only two accent classes.
Referring to Myôgishô accent classes, in Kagoshima, one class
has been derived from a merger of Myôgishô classes 2.1 and 2.2,
and the other class has resulted from a merger of Myôgishô
classes 2.3, 2.4, and 2.5. The Kagoshima accent system can be
analyzed as a register system: word-initial high and low register.

2
More precisely Hirayama ed. (1992: 27-29) gives five types of Japanese
accent. The other two (‘Miyakonojô-type’ and ‘Sendai-type’) are not relevant
because they do not have distinctive accent.
3
Ibuki-jima dialect has five distinctions for disyllabic nouns (Wada 1966a and
1966b).
288 ACCENTUAL HISTORY

6.2.1 Description of Tokyo Japanese Accent

Tokyo nouns are distinguished by locus only. Another
characteristic of Tokyo dialect is that it has a synchronic
phonetic rule that lowers the word-initial high pitch when the
mora is not accented.
Monosyllables: In Tokyo monosyllabic nouns are either tonic
or atonic. Tonic monosyllables are phonetically H(L); high on
the noun itself and low on a following particle. The phonetic
pitch shape of atonic monosyllabic nouns is L(H). Compared
with Myôgishô monosyllabic nouns, Tokyo tonic nouns
regularly correspond to Myôgishô accent class 1.3, and Tokyo
atonic nouns to Myôgishô accent classes 1.1-2. In addition,
Tokyo monosyllabic nouns are generally monomoraic.

(1) Tokyo monosyllabic accent

Phonemic Phonetic Myô List of nouns
O˘ H(L) 1.3 ki˘ ‘tree’, ha˘ ‘tooth’, me˘ ‘eye’
O L(H) 1.1 ke ‘hair’, ti ‘blood’
1.2 ha ‘leaf’, na ‘name’

Disyllables: There are three distinctive accent patterns in

Tokyo disyllabic nouns. They are atonic (/OO/), prototonic
(/O˘O/), and oxytonic (/OO˘/). Phonetically atonic nouns are
LH(H), prototonic nouns are HL(L), and oxytonic nouns are
LH(L).
My synchronic analysis of Tokyo disyllabic nouns reveals
that disyllables include not only historical disyllables but also a
historical trisyllabic noun; o:gi˘ ‘ fan’ , which belongs to
Myôgishô accent class 3.4. However, in Tokyo dialect it has
become disyllabic by losing a syllabic segment, which,
nevertheless resulted in trimoraic structure.
Regarding the correspondences between Tokyo disyllabic
nouns and their counterparts in Myôgishô, Tokyo atonic nouns
regularly correspond to Myôgishô class 2.1. Tokyo prototonic
disyllables seem to correspond to 2.3-5 in Myôgishô. However,
the accent of kumo ‘ cloud’ (2.3) is irregular because Myôgishô
2.3 nouns generally correspond to Tokyo oxytonic. Based on
the data in (2), except for kumo ‘ cloud’ , Tokyo initial accent in
general corresponds to Myôgishô 2.4-5. Tokyo oxytonic
disyllables correspond to Myôgishô accent classes 2.2-3. In other
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 289

words, accentual correspondences between Tokyo disyllables

and their corresponding Myôgishô nouns are systematic, see (3)
below:

(2) Tokyo disyllabic accent

Phonemic Phonetic Myô List of nouns
OO LH(H) 2.1 hana ‘nose’, usi ‘cow’, tori
‘bird’, kosi ‘loins’, kaze ‘wind’,
hige ‘beard’
O˘O 2.3 ku˘mo ‘cloud’
2.4 hu˘ne ‘boat’, u˘mi ‘sea’, ka˘ta
‘shoulder’, i˘ki ‘breath’, i˘ta
‘board’, u˘su ‘mortar’, ha˘si
‘chopsticks’, ha˘ri ‘needle’
2.5 a˘se ‘sweat’, a˘me ‘rain’,
mu˘ko ‘bridegroom’, ko˘e
‘voice’, ka˘ge ‘shadow’, o˘ke
‘bucket’
OO˘ LH(L) 2.2 isi˘ ‘stone’, kami˘ ‘paper’,
hasi˘ ‘bridge’
2.3 hana˘ ‘flower’, yama˘
‘mountain’, hone˘ ‘bone’, inu˘
‘dog’, doro˘ ‘mud’
3.4 o:gi˘ ‘fan’

(3) Correspondences between Tokyo and Myôgishô accent

Tokyo disyllables Myôgishô

Atonic 2.1
Oxytonic 2.2-3
Prototonic 2.4-5

Trisyllables: Tokyo trisyllabic nouns can be grouped into

four distinctive accent classes: atonic (/OOO/), prototonic
(/O˘OO/), mesotonic accent (/OO˘O/), and oxytonic (/OOO˘/).
The phonetic pitch shapes of these accent classes are respectively
LHH(H), HLL(L), LHL(L), and LHH(L).
290 ACCENTUAL HISTORY

My analysis of Tokyo trisyllabic accent in (4) includes not

only nouns from my original word list but also other words. I
have added more trisyllabic nouns in order to generalize the
accentual patterns. In (4), the additional words are in bold. With
the additional data, we are able to determine that hutatu ˘ ‘ two’
(3.2) is irregular because Tokyo nouns corresponding to
Myôgishô accent class 3.2 are normally atonic (i.e., /OOO/) in
Tokyo.
By adding more nouns corresponding to Myôgishô accent
class 3.5, it becomes clear that there is a mesotonic accent class
(i.e., /OO˘O/) in Tokyo. This class contains only nouns
belonging to class 3.5. Additional nouns corresponding to
Myôgishô accent class 3.7 also make it clear that nouns of 3.7
class are either atonic or prototonic in Tokyo.

(4) Tokyo trisyllabic accent

Phonemic Phonetic Myô List of nouns
OOO LHH(H) 3.1 sirusi ‘mark’, kemuri ‘smoke’,
hitai ‘forehead’, akubi ‘yawn’
3.2 mukade ‘centipede’, tubasa
‘wing’, tokage ‘lizard’,
turube ‘well-bucket’, ibara
‘thorn’
3.5 hasira ‘pillar’4, abura ‘fat’,
k e y a k i ‘zelkova’, tuzura
‘wicker’
3.6 unagi ‘eel’, mimizu
‘earthworm’, usagi ‘rabbit’,
suzume ‘sparrow’, nezumi
‘mouse’
3.7 kusuri ‘medicine’, kuzira
‘whale’, hatake ‘field’, usiro
‘behind’, tarai ‘washtub’

4
NHK Hôsô Bunka Kenkyûjo ed. (1971: 721), says that the word for ‘pillar’
can be either atonic hasira or oxytonic hasira˘. However, according to
Hirayama ed. (1992: 4059), the word can only be atonic: hasira.
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 291

O˘OO HLL(L) 3.3 ko˘gane ‘gold’5, ha˘t a t i ‘20

years old’, a˘w a b i ‘abalone’,
s a˘zae ‘top shell’
3.5 i˘noti ‘life’, na˘mida ‘tears’,
ma˘kura ‘pillow’, s u˘gata
‘figure’, a˘ware ‘pity’,
h o˘taru ‘firefly’
3.7 h i˘t o r i ‘one person’, h i˘t o t u
‘one’, ka˘b u t o ‘helmet’,
ka˘i k o ‘silkworm’, m a˘buta
‘eyelid’, o ˘tona ‘adult’,
s i˘zuka ‘quiet’, h o˘noka
‘faint’
OO˘O LHL(L) 3.5 i t o˘k o ‘cousin’, k o k o˘r o
‘heart’, nana˘t u ‘seven’,
i t u˘t u ‘five’
OOO˘ LHH(L) 3.2 hutatu˘ ‘two’
3.3 tikara˘ ‘strength’
3.4 hukuro˘ ‘bag’, takara˘
‘treasure’, kagami˘ ‘mirror’

Among nouns corresponding to Myôgishô accent class 3.3,

tikara ˘ ‘ strength’ does not share its accentuation with others.
The accentual pattern of ‘strength’ is probably irregular.

6.2.2 Description of Kyoto Japanese Accent

Kyoto accent is more complex than that of Tokyo or of
Kagoshima, because it makes use of two different accent
parameters: locus and word-initial register; thus it has more
accent classes than any other accent system in the Japonic
languages.
Monosyllables: In Kyoto monosyllabic nouns can be
analyzed as being either high-register, low-register, or tonic.
5
However, Hirayama (1960: 246) and NHK Hôsô Bunka Kenkyûjo ed. (1971:
302) lists not only k o˘gane but also kogane for ‘gold’.
292 ACCENTUAL HISTORY

Kyoto monosyllables are composed of two moras; there are no

monomoraic nouns in Kyoto.

(5) Kyoto monosyllabic accent

Phonemic Phonetic Myô List of nouns
`oo HH(H) 1.1 `ti: ‘blood’
_oo LH ~ 1.3 _ki: ‘tree’, _me: ‘eye’
LL(H)
o˘o HL(L) 1.1 ke˘: ‘hair’
1.2 ha˘: ‘leaf’, na˘: ‘name’
1.3 ha˘: ‘tooth’

Thus, the phonetic pitch shapes of those accent classes are

HH(H) for high-register, LH ~ LL(H) for low-register, and
HL(L) for tonic accent. Low-register nouns are LH in isolation,
but LL(H) when followed by a particle.
At first glance, the accentuation of Kyoto monosyllables does
not seem to systematically correspond to Myôgishô accent
classes when compared with Myôgishô accent classes. However, it
becomes clear that Kyoto monosyllabic accent regularly
corresponds to Myôgishô accent when we take into account
additional data on Kyoto monosyllables. In particular, the
accentuation of the word ke˘: ‘ hair’ (1.1) is irregular because
Kyoto nouns corresponding to Myôgishô accent class 1.1 are
usually high-register, e.g., `e: ‘ a handle’ , `ka: ‘ mosquito’ , `ko:
‘ child’ , `to: ‘ door’ , `ho: ‘ sail’ , `mi: ‘ fruit’ . The accent of the
word ha ˘: ‘ tooth’ (1.3) is also irregular because nouns
corresponding to Myôgishô accent 1.3 are normally low-register
(e.g., _o: ‘tail’, _te: ‘hand’, _hi: ‘fire’).

(6) General correspondences between Kyoto monosyllabic

accent and that of Myôgishô

Kyoto accent Myôgishô accent

High-register 1.1
tonic-accent 1.2
Low-register 1.3
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 293

To sum up, with some exceptions, Kyoto high-register nouns

generally correspond to Myôgishô accent class 1.1, Kyoto low-
register nouns to Myôgishô accent class 1.3, and Kyoto tonic-
accent nouns to Myôgishô accent class 1.2.
Disyllables: Kyoto disyllabic nouns have four distinct accent
classes: high-register, low-register, initial-accent, and final-accent.
High-register nouns are phonetically high-level in pitch. Low-
register nouns are LH in isolation, but LL(H) when used with a
particle. Initial-accent nouns are simply HL(L). In final-accent
nouns, the pitch begins low and falls on the second syllable (i.e.,
LF) when uttered by itself, but LH(L) when followed by a
particle.

(7) Kyoto disyllabic accent

Phonemic Phonetic Myô List of nouns
`OO HH(H) 2.1 hana ‘nose’, `usi ‘cow’, `tori
‘bird’, `kosi ‘loins’, `kaze
‘wind’, `hige ‘beard’
_OO LH ~ 2.4 _hune ‘boat’, _umi ‘sea’, _kata
LL(H) ‘shoulder’, _iki ‘breath’, _ita
‘board’, _usu ‘mortar’, _hasi
‘chopsticks’, _hari ‘needle’
O˘O HL(L) 2.2 i˘si ‘stone’, ka˘mi ‘paper’,
ha˘si ‘bridge’
2.3 ha˘na ‘flower’, ya˘ma ‘mountain’,
ku˘mo ‘cloud’, ho˘ne ‘bone’, i˘nu
‘dog’, do˘ro ‘mud’
3.4 o:˘gi ‘fan’
OO˘ LF ~ 2.5 ase˘ ‘sweat’, ame˘ ‘rain’, muko˘
LH(L) ‘bridegroom’, koe˘ ‘voice’, kage˘
‘shadow’, oke˘ ‘bucket’

As mentioned earlier, the synchronic data for disyllables

include a historical trisyllabic noun, namely o:˘gi ‘ fan’ . As its
counterpart afuki LLL in Myôgishô indicates, the earlier
Japanese form for ‘ fan’ must have been trisyllabic as well (see
Section 6.3).
294 ACCENTUAL HISTORY

The correspondences between Kyoto and Myôgishô disyllabic

accent classes are regular. That is, Kyoto high-register disyllables
regularly correspond to Myôgishô class 2.1, Kyoto low-register
nouns to Myôgishô class 2.4, Kyoto initial-accent words to both
Myôgishô classes 2.2 and 2.3, and Kyoto final-accent nouns to
Myôgishô class 2.5.
Trisyllables: In Kyoto there are four accentual types for
trisyllabic nouns. They are as follows: high-register-atonic, low-
register-atonic, initial-accent, and second-syllable-accent. High-
register-atonic nouns are phonetically high level, but low-
register-atonic nouns are final high; i.e., LLH in isolation and
LLL(H) when followed by a particle. As for tonic nouns, initial-
accent nouns have HLL(L) pitch shape. Second-syllable-accent
nouns are LHL(L).
A comparison of Kyoto trisyllabic nouns and their Myôgishô
cognates shows general systematic correspondences in accent
forms, except for an irregular correspondence of the word for
‘ gold’ (3.3). Kyoto high-register accent corresponds to
Myôgishô accent class 3.1-3, Kyoto low-register accent to
Myôgishô accent 3.6, Kyoto initial-accent to Myôgishô accent
3.2-5, and second-syllable accent to Myôgishô 3.2 and 3.7. The
Kyoto word `kogane ‘ gold’ is probably irregular because
nouns corresponding to Myôgishô class 3.3 are normally initial-
accent (/O˘OO/) in Kyoto, e.g., a ˘wabi ‘ abalone’ , ha ˘tati ‘ 20
years old’.

(8) Kyoto trisyllabic accent

Phonemic Phonetic Myô List of nouns
`OOO HHH(H) 3.1 sirusi ‘mark’, `kemuri ‘smoke’,
`hitai ‘forehead’, `akubi ‘yawn’
3.2 `aida ‘interval, `sakura
‘cherry’, `e n o k i ‘hackberry’,
`tobari ‘curtain’
3.3 `kogane ‘gold’
_OOO LLH ~ 3.6 _unagi ‘eel’, _mimizu
LLL(H) ‘earthworm’
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 295

O˘OO HLL(L) 3.2 a˘z u k i ‘red beans’, t o˘bira

‘door’, h i˘g a s i ‘east’, ne˘doko
‘alcove’
3.3 ti˘kara ‘strength’, ha˘t a t i ‘20
years old’, k o˘m u g i ‘wheat’,
a˘w a b i ‘abalone’
3.4 hu˘kuro ‘bag’, ta˘kara
‘treasure’, ka˘gami ‘mirror’
3.5 ma˘kura ‘pillow’, a˘bura ‘fat’,
i˘noti ‘life’, na˘mida ‘tears’,
ha˘sira ‘pillar’, ko˘koro‘heart’,
i˘toko ‘cousin’
OO˘O LHL(L) 3.2 muka˘de ‘centipede’, huta˘tu
‘two’, tuba˘s a ‘wing’, huta˘r i
‘two people’, turu˘b e ‘well-
bucket’, t o k a˘g e ‘lizard’,
i b a˘ra ‘thorn’
3.7 kusu˘ri ‘medicine’, h i t o˘r i
‘one person’, h i t o˘t u ‘one’,
kabu˘t o ‘helmet’, k u z i˘ra
‘whale’, hata˘ke ‘field’

Finally, in (8) there are additional nouns that are not on my

original word list. They are bold-faced. With the additional data,
it becomes clear that nouns corresponding to Myôgishô accent
class 3.2 correspond to /`OOO/, /O˘OO/, and /OO˘O/ in Kyoto.

6.2.3 Description of Myôgishô Accent

Compared with other Japonic languages, the accent system of
Myôgishô is much more complex than even present day Kyoto.
It has more accent distinctions than any other dialect of the
Japonic languages known to us. There are three distinctions for
monosyllables, five for disyllables, and seven for trisyllables.
These distinctions are based on combinations of word-initial
register and locus. As far as typology of accent system is
concerned, register and locus distinctions are very different from
each other. In Myôgishô, these two different systems coexist.
296 ACCENTUAL HISTORY

Monosyllables: There are three types of accentuation for

Myôgishô monosyllabic nouns: /`O/ [H(H)], /O˘/ [H(L)], and /_O/
[L(H)]. Following Martin (1987), in this book I define them, in
order, class 1.1, class 1.2, and class 1.3. In this accent system,
word-initial register and locus are distinctive.

(9) Myôgishô monosyllabic accent

Phonemic Phonetic Myô List of nouns
`O H(H) 1.1 `ke ‘hair’, `ti ‘blood’
O˘ H(L) 1.2 fa˘ ‘leaf’, na˘ ‘name’
_O L(H) 1.3 _ki ‘tree’, _me ‘eye’, Èfa6
‘tooth’

Notice that the word Èfa ‘ tooth’ has rising pitch, different
from other nouns in the same class. The word is phonetically R
in isolation, but L(H) when followed by a particle. Because this
type of words is rare in Myôgishô, they are usually treated as
nouns of class 1.3. Following a traditional analysis, I put the
word Èfa ‘ tooth’ in class 1.3 for the time being, but with the
rising pitch symbol ‘ È’ .
Disyllables: Myôgishô disyllabic accent is a five-way system.
The accent classes are /`OO/ (2.1), /O˘O/ (2.2), /_OO/ (2.3), /OÈO/
(2.4), and /OO˘/ (2.5). As its monosyllables do, Myôgishô
disyllabic nouns also utilize two different types of accent
systems: register (word-initial-high and word-initial-low register)
and locus accent.
There are two types of regarding locus accent: falling accent
(indicated by ‘ ˘’ ) and rising pitch (indicated by ‘ È’ ). Both types
of locus accent change pitch between syllables/moras.
Initial-high register words are high pitch throughout the
words. Initial-low register words are low-level pitch. Words with
initial-accent are phonetically HL(L). Words with final-syllable
rising accent are LH(H) pitch. Finally, words with final-accent
are LH(L).

6
The symbol ‘È’ indicates a rising pitch. For monosyllables, pitch rises
within a syllable. For disyllables and trisyllables, syllables before ‘È’ are low
and syllables following the symbol are high pitch; e.g., /ÈO/ [R ~ L(H)],
/OÈOO/ [LHH(H)], /OOÈO/ [LLH(H)] or [LLH ~ LLL(H)].
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 297

(10) Myôgishô disyllabic accent

Phonemic Phonetic Myô List of nouns
`OO HH(H) 2.1 fana ‘nose’, `usi ‘cow’, `tori
‘bird’, `kosi ‘loins’, `kaze
‘wind’, `fige ‘beard’
O˘O HL(L) 2.2 i˘si ‘stone’, ka˘mi ‘paper’,
fa˘si ‘bridge’
_OO LL(L) 2.3 _fana ‘flower’, _yama
‘mountain’, _kumo ‘cloud’,
_fone ‘bone’, _inu ‘dog’,
_doro ‘mud’
OÈO LH(H) 2.4 fuÈne ‘boat’, uÈmi ‘sea’, kaÈta
‘shoulder’, iÈki ‘breath’, iÈta
‘board’, uÈsu ‘mortar’, faÈsi
‘chopsticks’, faÈri ‘needle’
OO˘ LH(L)7 2.5 ase˘ ‘sweat’, ame˘ ‘rain’,
muko˘ ‘bridegroom’, kowe˘
‘voice’, kage˘ ‘shadow’, woke˘
‘bucket’

Trisyllables: In Myôgishô there are seven distinctive accent

types for trisyllabic nouns: /`OOO/, /`OO˘O/, /O˘OO/, /_OOO/,
/OOÈO/, /OÈOO/, and /_OO˘O/. Adopting Martin’ s terminology
(see 1987: 371), I define these Myôgishô accent classes as 3.1,
3.2, 3.3, 3.4, 3.5, 3.6, and 3.7 respectively.
Trisyllabic nouns belonging to Myôgishô accent class 3.1 are
high-register, phonetically high-level throughout the words.
Nouns with /`OO˘O/ accent belong to class 3.2, phonetically
HHL(L). Initial-accent trisyllables (3.3) have HLL(L) pitch.
Nouns in accent class 3.4 are low-register, phonetically low-level.

7
Kindaichi (1975: 52) explains that class 2.5 nouns is LH(L); no mention on
the pitch pattern of 2.5-nouns without a particle, but Kamei et al. eds. (1964:
116) define it as LF without a particle; it is not clear whether the pitch changes
when followed by a particle.
298 ACCENTUAL HISTORY

Nouns belonging to accent classes 3.5 and 3.6 make use of

‘ rising pitch’ (i.e., ‘ È’ ), instead of falling pitch (i.e., ‘ ˘’ ). 3.5-
and 3.6-nouns are /OOÈO/ [LLH(H)] and /OÈOO/ [LHH(H)]
respectively; that is, the location of the rising pitch is crucial in
distinguishing 3.5-nouns from 3.6-nouns. Finally, low-register
second-syllable-accent nouns (i.e., /_OO˘O/) are phonetically
LHL(L) and they are called 3.7 nouns. In addition, interestingly
enough, there are no final-accent trisyllables in the Myôgishô
dialect.8

(11) Myôgishô trisyllabic accent

Phonemic Phonetic Myô List of nouns
`OOO HHH(H) 3.1 `sirusi ‘mark’, `kemuri
‘smoke’, `fitafi ‘forehead’,
`akubi ‘yawn’
`OO˘O HHL(L) 3.2 `muka˘de ‘centipede’,
`futa˘tu ‘two’
O˘OO HLL(L) 3.3 ko˘gane ‘gold’, ti˘kara
‘strength’
_OOO LLL(L) 3.4 _fukuro ‘bag’, _takara
‘treasure’, _kagam ‘mirror’,
_afuki ‘fan’
OOÈO LLH(H) 3.5 makuÈra ‘pillow’, abuÈra
‘fat’, inoÈti ‘life’, namiÈda
‘tears’, fasiÈra ‘pillar’,
kokoÈro ‘heart’, itoÈko ‘cousin’
OÈOO LHH(H) 3.6 uÈnagi ‘eel’, miÈmizu
‘earthworm’
_OO˘O LHL(L) 3.7 _kusu˘ri ‘medicine’, _kuzi˘ra
‘whale’, _fata˘ke ‘field’

6.2.4 Description of Kagoshima Japanese Accent

Kagoshima accent for nouns is very simple. There are two types
of accent: high-register and low-register. In the former type of

8
Martin (1987: 273) suggests that some of LHL-nouns have been derived from
the earlier final-accent nouns; i.e., *LHH(L) > LHL , e.g., kusuri ‘drug’.
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 299

nouns, a phonetic locus is on a penultimate syllable in a prosodic

word,9 e.g., `hana HL ‘ nose’ , `hana ga LHL ‘ nose (Nom.)’ .
Notice that the domain of the accent is not a word itself, but it is
a prosodic word, namely a word and a following particle.
Because a phonetic rule lowers the pitch of unaccented syllables
in high-register nouns, they are LH(L) in disyllables and
LLH(L) in trisyllables when followed by a particle.
With regard to rising accent nouns, they begin in low pitch
and end in high pitch within the domain, e.g., _hana LH
‘ flower’ , _hana ga LLH ‘ flower (Nom.)’ . The final syllable of
a prosodic word is always high in pitch and the rest of the
syllables are low.
Typologically the accent system of Kagoshima is very
different from a locus system (e.g., Tokyo accent, where only
locus distinguishes one word from another).
All Kagoshima data are taken from Hirayama (1951),
Hirayama et al. (1967), and Hirayama ed. (1992).
Monosyllables: Kagoshima monosyllabic nouns are either
high-register or low-register. As (12) shows, modern Kagoshima
monosyllables include not only historical monosyllables but also
historical disyllables. The former type is monomoraic, but the
latter consists of two moras. High-register monomoraic nouns
are phonetically F in isolation, but H(L) when followed by a
particle. Two-mora high-register nouns are HL ~ HH(L).

(12) Kagoshima monosyllabic accent

Phonemic Phonetic Myô List of nouns
`O F ~ H(L) 1.1 `ke ‘hair’, `ti ‘blood’
1.2 `ha ‘leaf’, `na ‘name’
HL ~ 2.2 `kan10 ‘paper’
HH(L)
_O R ~ L(H) 1.3 _ki ‘tree’, _ha ‘tooth’, _me ‘eye’
HH ~ 2.3 _in ‘dog’
LL(H) 2.4 _un11 ‘sea’, _iQ ‘breath’

9
Following McCarthy and Prince (1995: 321), I use the term ‘prosodic word’
to indicate a phrase consisting of at least two moras or syllables.
10
This is kan HH(L) (Hirayama ed. 1992: 1291).
11
According to Hirayama ed., the noun for ‘sea’ is un HH(L) (1992: 655).
300 ACCENTUAL HISTORY

For low-register nouns, monomoraic words are R in isolation,

but they are L(H) when followed by a particle, e.g., _ki R ‘ tree’ ,
_ki ga L-H ‘ tree (Nom.)’ . Two-mora nouns with low-register
accent have high-level pitch when uttered alone, but otherwise
they are in low pitch with high pitch on the following particle,
e.g., _in HH ‘dog’, _in ga LL-H ‘dog (Nom.)’.
Regarding correspondences between Kagoshima and
Myôgishô accentual classes, as long as monosyllables are
concerned, Kagoshima high-register monosyllables regularly
correspond to Myôgishô accent classes 1.1-2, and its low-register
monosyllables to Myôgishô class 1.3.
Disyllables: In Kagoshima disyllabic nouns, just like in
monosyllables, there are two accentual distinctions: high-register
and low-register. High-register nouns have a phonetic pitch fall
immediately after the penultimate syllable within a prosodic
word. Low-register nouns always keep high pitch on the final
syllable in their domain; thus, they are pronounced LH without a
particle and LL(H) with a particle. In the data, some nouns are
historically trisyllabic, e.g., `hutaQ ‘ two’ . They consist of three
moras while original disyllables are two moras long.
Compared with their corresponding Myôgishô disyllabic
nouns, Kagoshima disyllables systematically correspond to
Myôgishô accent types. Kagoshima high-register nouns are
accent classes 2.1-2 in Myôgishô, and low-register disyllabic
nouns correspond to Myôgishô accent classes 2.3-5.
In the Kagoshima dialect the word-final syllables such as /si/,
/zi/, /su/, and /zu/ become voiceless. The word-final /Q/ is
phonetically implosive [t], e.g., `hutaQ ‘ two’ . Word-final
syllables corresponding to Tokyo syllables /ki/, /gi/, /ku/, /gu/, /ti/,
/di/, /tu/, /zu/, /bi/, and /bu/ are /Q/ in Kagoshima (Hirayama 1951).
There is one exception for this, i.e., _o:gi ‘fan’.
Kagoshima disyllabic low-register class includes the word
_o:gi ‘ fan’ , which was historically trisyllabic. Unlike other
Kagoshima disyllabic nouns, it has a long vowel in the initial
syllable. Monophthongization and compensatory lengthening
can account for the shape of the word, i.e., *apukyi > *auki >
o:gi.
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 301

(13) Kagoshima disyllabic accent

Phonemic Phonetic12 Myô List of nouns
`OO HL~ LH(L) 2.1 `hana ‘nose’, `usi ‘cow’, `toi
‘bird’, `kosi ‘loins’, `kaze
‘wind’, `hige ‘beard’
2.2 `isi ‘stone’, `hasi ‘bridge’
3.2 `hutaQ ‘two’
_OO LH ~ LL(H) 2.3 _hana ‘flower’, _yama
‘mountain’, _kumo ‘cloud’,
_hone ‘bone’, _doro ‘mud’
2.4 _hune ‘boat’, _kata ‘shoulder’,
_ita ‘board’, _usu ‘mortar’,
_hasi ‘chopsticks’, _hai
‘needle’
2.5 _ase ‘sweat’, _ame ‘rain’,
_muko ‘bridegroom’, _koe
‘voice’, _kage ‘shadow’,
_oke ‘bucket’
3.1 _akuQ13 ‘yawn’
3.4 _o:gi ‘fan’, _kagan ‘mirror’
3.5 _inoQ ‘life’, _nanda ‘tears’
3.6 _unaQ ‘eel’, _mimiQ
‘earthworm’

Trisyllables: Like its monosyllables and disyllables,

Kagoshima trisyllabic nouns have two accent patterns. They are
high-register and low-register. Within its domain of accent (i.e.,
prosodic word), high-register nouns have high pitch on a
penultimate-syllable. Thus, they are LHL in isolation and
LLH(L) with a particle. The other accent type is phonetically

12
Each letter (H or L) represents pitch height of one syllable. This is the same
for nouns with three moras such as `hutaQ ‘two’ -- high pitch on the first
syllable (/hu/) and low pitch on the second syllable (/taQ/).
13
The word for ‘yawn’ is akut˘ LHH(L) in Hirayama ed. (1992: 56).
302 ACCENTUAL HISTORY

high on the final syllable in the prosodic word; the other

syllables in the domain are in low pitch (e.g., LLH ~ LLL(H)).
When it comes to correspondence between Kagoshima
trisyllabic accent and Myôgishô accent, as far as the data in (14)
are concerned, the correspondences are in general systematic,
although the data lack nouns corresponding to Myôgishô accent
class 3.6. Kagoshima high-register nouns regularly correspond
to Myôgishô accent classes 3.1-3 and Kagoshima low-register
nouns generally to Myôgishô accent classes 3.4-5, and 3.7.

(14) Kagoshima trisyllabic accent

Phonemic Phonetic14 Myô List of nouns
`OOO LHL ~ 3.1 `sirusi ‘mark’, `kemui
LLH(L) ‘smoke’, `hitai ‘forehead’
3.2 `mukade ‘centipede’
3.3 `tikara ‘strength’, `kogane
‘gold’
_OOO LLH ~ 3.4 _hukuro ‘bag’, _takara
LLL(H) ‘treasure’
3.5 _makura ‘pillow’, _abura
‘fat’, _hasita ‘pillar’,
_kokoro ‘heart’, _itoko
‘cousin’
3.7 _kusui ‘medicine’, _kuzira
‘whale’, _hatake ‘field’

6.3 CORRESPONDENCES AND RECONSTRUCTION OF

PROTO-MAINLAND JAPANESE ACCENT
In this section, I will reconstruct a Proto-Mainland Japanese
accent system on the basis of accentual correspondences among
Tokyo, Kyoto, Myôgishô, and Kagoshima dialects. In
reconstructed Proto-Mainland Japanese, there are four accent
distinctions for monosyllables, five for disyllables, and twelve for
trisyllables. With regard to typology of accent system, in Proto-

14
Each letter (H or L) represents pitch height of one syllable.
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 303

Mainland Japanese two different types of accentuation (i.e.,

locus accent and register) coexist in Proto-Mainland Japanese.
Monosyllables: For monosyllabic nouns, there are four sets of
accentual correspondences among Tokyo, Kyoto, Myôgishô, and
Kagoshima. This leads us to a reconstruction of four Proto-
Mainland Japanese accent classes for monosyllables. The classes
that I have reconstructed are high-register atonic (* /`oo/), final
accent (* /oo˘/), low-register atonic (* /_oo/), and rising accent
(* /Èoo/). The accent system of Proto-Mainland Japanese
monosyllables is identical to that of Myôgishô, with variation of
phonetic pitch shape in some classes.

(15) Correspondences and reconstructed Proto-Mainland

Japanese accent for monosyllabic nouns
(a)
Tokyo Kyoto Myô Kago PMJ15
O `o o `O `O *`o o
L(H) HH(H) H(H) F ~ H(L) HH(H)
‘hair’ (1.1) ke ke˘: ke `ke *`ke:y
‘blood’ (1.1) ti `t i `t i `t i *`ti:y

(b)
Tokyo Kyoto Myô Kago PMJ
O o˘o O˘ `O *o˘o
L(H) HL(L) H(L) F ~ H(L) HL(L)
‘leaf’ (1.2)16 ha ha˘: fa˘ `ha *pa˘:
‘name’ (1.2) na na˘: na˘ `na *na˘:

15
My reconstruction also reflects characteristics of Old Japanese seen in the
writings of the late seventh century and the eighth century, although Old
Japanese cognates are not listed. For example, Old Japanese had two variants
for the vowels /i/, /e/, and /o/, and they are indicated /yi/, /iy/, /ye/, /ey/, /wo/,
and /´/ (Regarding the symbols, I mostly adopt Martin’s system (1987: 49-
52)), except for the distinction between /o/ and /o/ .
16
Numbers in parentheses indicate traditional accent classes.
304 ACCENTUAL HISTORY

(c)
Tokyo Kyoto Myô Kago PMJ
O˘ _o o _O _O * _o o
H(L) LL(H) L(H) R ~ L(H) LL(H)
‘tree’ (1.3) k i˘ _ki: _k i _k i * _ki:y
‘eye’ (1.3) me˘ _me: _me _me * _me:y

(d)
Tokyo Kyoto Myô Kago PMJ
O˘ o˘o ÈO _O *oÈo
H(L) HL(L) R~ L(H) R ~ L(H) LH(H)
‘tooth’ (1.3) ha ha˘: Èfa _ha *paÈ:

Regarding phonetic pitch shapes of the reconstructed accent

classes, the high-register and final accent are HH(H) and HH(L)
respectively. For low-register atonic its pitch shape is LL(L).
And for the rising accent, pitch goes from L to H, and it is
LH(H) when uttered with a particle.
My Proto-Mainland Japanese monosyllables are bimoraic.
That is to say, I consider that vowel length in Kyoto is a remnant
of Proto-Mainland Japanese vowel length because shortening of
segments is one of the characteristics of the Japonic languages,
e.g., monophthongization (see Martin 1987: 2 and 4). It
becomes obvious on the basis of the data cited above that
Kagoshima has a general tendency to shortening words, i.e.,
apocope. In Myôgishô, some monosyllabic nouns are recorded
as bimoraic, e.g., ka: 'mosquito', although there are not many of
them. This suggests that there were bimoraic monosyllables in
Myôgishô.
Disyllables: As shown in (16), there are five sets of
correspondences for mainland Japanese disyllabic accent. Based
on that, I have reconstructed five accentual classes for Proto-
Mainland Japanese disyllabic nouns. The reconstructed accent
classes are high-register (* /`OO/), initial accent (* /O˘O/), low-
register (* /_OO/), rising accent (* /OÈO/), and final accent
(* /OO˘/). These classes are identical to that of Myôgishô.
Reconstructed nouns for each accent class are listed in (17).
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 305

(16) Synopsis of reconstructed accent classes

Tokyo Kyoto Myô Kagoshima PMJ
a OO `OO `OO `OO *`OO
LH(H) HH(H) HH(H) HL ~ LH(L) HH(H)
b OO˘ O˘O O˘O *O˘O
LH(L) HL(L) HL(L) HL(L)
c _OO _OO *_OO
LL(L) LH ~ LL(H) LL(L)
d O˘O _OO OÈO *OÈO
HL(L) LH ~ LL(H) LH(H) LH(H)
e OO˘ OO˘ *OO˘
LF ~ LH(L) LH(L) LH(L)

Proto-Mainland Japanese initial-high register accent is

phonetically high level; there is no fall in pitch. The second class
for Proto-Mainland Japanese disyllables is initial-syllable accent;
high pitch on the initial syllable and low on both the second
syllable and a following particle. The third class, low-register
accent, is phonetically low level. The fourth Proto-Mainland
Japanese disyllabic accent is rising (* /OÈO/). Its pitch rises
between the first and second syllables, LH(H). Finally, the fifth
accentuation that I have reconstructed for Proto-Mainland
Japanese disyllables is final-accent, where pitch is low on the first
syllable and high on the second syllable, and then pitch falls
after the second syllable – it falls between the final syllable and a
following particle when the words are followed by a particle.
In the correspondences most forms are regular, except for the
accentuation of the Tokyo word ku ˘mo ‘ cloud’ (2.3) in (17c) --
In Tokyo, nouns corresponding to the historical accent class 2.3
are generally final-accent. Kindaichi (1971: 7 and 10) claims
that the word for ‘ cloud’ underwent an analogical change and
resulted in having the identical accentuation to ku ˘mo ‘ spider’
(2.5) because they are homophonic.
306 ACCENTUAL HISTORY

(17) Reconstructed Proto-Mainland Japanese accent for

disyllabic nouns
(a)
Tokyo Kyoto Myô Kago PMJ
OO `OO `OO `OO *`OO
LH(H) HH(H) HH(H) HL ~ LH(L) HH(H)
‘nose’ (2.1) hana `hana `fana `hana *`pana17
‘cow’ (2.1) usi `usi `usi `usi *`usi
‘bird’ (2.1) tori `tori `tori `toi *`t´ri
‘loins’ (2.1) kosi `kosi `kosi `kosi *`k´s i
‘wind’ (2.1) kaze `kaze `kaze `kaze *`kazey
‘beard’ (2.1) hige `hige `fige `hige *`pyigey

(b)
Tokyo Kyoto Myô Kago PMJ
OO˘ O˘O O˘O `OO *O˘O
LH(L) HL(L) HL(L) HL ~ LH(L) HL(L)
‘stone’ (2.2) isi˘ i˘s i i˘s i `isi *i˘s i
‘paper’ (2.2) kami˘ ka˘mi ka˘mi `kan *ka˘byi
‘bridge’ (2.2) hasi˘ ha˘s i fa˘s i `hasi *pa˘s i

(c)
Tokyo Kyoto Myô Kago PMJ
OO˘ O˘O _OO _OO *_OO
LL(L) LH ~ LL(H) LL(L)
LH(L) HL(L)
‘flower’ (2.3) hana˘ ha˘na _fana _hana * _pana
‘mountain’ (2.3) yama˘ ya˘ma _yama _yama * _yama
‘cloud’ (2.3) ku˘mo ku˘mo _kumo _kumo * _kumwo18
‘bone’ (2.3) hone˘ h o˘ne _fone _hone * _poney
‘dog’ (2.3) inu˘ i˘nu _inu _i n * _inu
‘mud’ (2.3) doro˘ do˘ro _doro _doro * _doro

17
The Modern Japanese phoneme /h/ corresponds to /p/ in Old Japanese.
18
A Western Old Japanese cognate for ‘cloud’ is kumwo (Martin 1987).
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 307

(d)
Tokyo Kyoto Myô Kago PMJ
O˘O _OO OÈO _OO *OÈO
LH ~ LH ~
HL(L) LH(H) LH(H)
LL(H) LL(H)
‘board’ (2.4) i˘ta _ita iÈta _ita *iÈta
‘sea’ (2.4) u˘mi _umi uÈmi _un *uÈmyi
‘shoulder’ (2.4) ka˘ta _kata kaÈta _kata *kaÈta
‘breath’ (2.4) i˘k i _iki iÈk i _iQ *iÈkyi
‘boat’ (2.4) hu˘ne _hune fuÈne _hune *puÈney
‘mortar’ (2.4) u˘su _usu uÈsu _usu *uÈsu
‘chopsticks’ (2.4) ha˘s i _hasi faÈs i _hasi *paÈsiy
‘needle’ (2.4) ha˘ri _hari faÈri _hai *paÈriy

(e)
Tokyo Kyoto Myô Kago PMJ
O˘O OO˘ OO˘ _OO *OO˘
LF ~ LH ~
HL(L) LH(L) LH(L)
LH(L) LL(H)
‘sweat’ (2.5) a˘se ase˘ ase˘ _ase *asey˘
‘rain’ (2.5) a˘me ame˘ ame˘ _ame *amey˘
‘bridegroom’(2.5) mu˘k o muko˘ muko˘ _muko *mwokwo˘19
‘voice’ (2.5) k o˘e koe˘ kowe˘ _koe *k´we˘
‘shadow’ (2.5) ka˘ge kage˘ kage˘ _kage *kagey˘
‘bucket’ (2.5) o˘ke oke˘ woke˘ _oke *wokey˘

Trisyllables: Among trisyllabic nouns, there are twelve sets of

correspondences among the mainland Japanese dialects. These
twelve sets of correspondences are listed below in (18a) through
(18m). On the basis of these correspondence sets, I have
reconstructed twelve Proto-Mainland Japanese accent classes.
They are * /`OOO/ (18a), * /OOOˆ/ (18b), * /OOO˘/ (18c),
* /O˘OÈO/ (18d), * /OÈOO˘/ (18e), * /OOÈO˘/ (18f), * /OOÈO/ (18g),

19
In Western Old Japanese the word for ‘bridegroom, son-in-law’ is mwokwo
(Martin 1987).
308 ACCENTUAL HISTORY

* /OOÊO/ (18h), * /_OOO/ (18i), * /OÈOO/ (18j), * /OÈOOˆ/ (18k),

and */OOÈOˆ/ (18m).

(18) Synopsis of reconstructed accent classes

Tokyo Kyoto Myô Kagoshima PMJ
a OOO OOO `OOO `OOO *`OOO
LHH(H) HHH(H) HHH(H) LHL ~ HHH(H)
LLH(L)
b OO˘O `OO˘O *`OOOˆ
LHL(L) HHL(L) HHF ~
HHH(L)
c `OOO *`OOO˘
HHH(H) HHH(L)
d O˘OO O˘OO O˘OO *O˘OÈO
HLL(L) HLL(L) HLL(L) HLH(H)
e OOO˘ _OOO _OOO *OÈOO
LHH(L) LLL(L) LLH ~ LHH(H)
LLL(H)
f `OOO *_OOO
HHH(H) LLL(L)
g O˘OO O˘OO OOÈO *OOÈO
HLL(L) HLL(L) LLH(H) LLH(H)
h OO˘O *OOÊO
LHL(L) LLR ~
LLL(H)
i OOO *OOÈO˘
LHH(H) LLH(L)
j _OOO OÈOO *OÈOO˘
LLH ~ LHH(H) LHH(L)
LLL(H)
k OO˘O _OO˘O *OÈOOˆ
LHL(L) LHL(L) LHF ~
LHH(L)
m O˘OO *OOÈOˆ
HLL(L) LLF ~
LLH(L)

Just as with its monosyllabic and disyllabic nouns, two

typologically different accent systems (namely, locus and
register) coexist in Proto-Mainland Japanese trisyllables to
distinguish one word from another. Furthermore, within the
locus system there are two different types, i.e., falling accent
(‘ ˘’ ) and rising accent (‘ È’ ). For falling accent, the pitch falls
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 309

immediately after an accented syllable or mora, while for rising

accent, the pitch rises immediately before an accented syllable or
mora in nouns. There are also accent types shown with ‘ ˆ’ and
‘ Ê’ . They indicate a falling or rising pitch within the syllable or
mora, e.g., /_OOOˆ/ [LHF ~ LHH(L)], /_OOÊO/ [LLR ~ LLL(H)].
In (18) I have summarized accentual correspondences and
also have listed the reconstructed accent classes. In (20),
reconstructed Proto-Mainland Japanese trisyllabic nouns for
each class are listed. Of those reconstructed nouns, there are also
cognates that are not on my original word list. They are added in
order to compensate for lack of data.
The data in (20) generally show regular correspondences
between each correspondence set and traditional accent classes.
However, not all cognates show regular accentuation. There are
some cognates with irregular accent. For example, in (20d), the
accent of the Tokyo form for ‘ strength’ (i.e., tikara ˘) is
irregular because Tokyo nouns corresponding to Myôgishô class
3.3 are generally HLL(L). Examples in (19) elucidate this.

(19) Examples for Tokyo trisyllabic nouns corresponding to

Myôgishô 3.3 accent class (Hirayama 1960)

‘ abalone’ awabi (3.3) HLL(L) (prototonic)

‘ wreath shell’ sazae (3.3) HLL(L) (prototonic)
’20 years old’ hatati (3.3) HLL(L) (prototonic)
‘ cape, promontory’ misaki (3.3) LHH(H) (atonic
~HLL(L) ~ prototonic)

Regarding the reconstructed Proto-Mainland Japanese form

for ‘ fan’ in (20e), Tokyo and Kagoshima cognates are
disyllabic while the Myôgishô form is trisyllabic. It is apparent
that the Tokyo and the Kagoshima forms underwent
monophthongization (see 6.2.2). Based on the Myôgishô form, a
trisyllabic form for ‘fan’ can be reconstructed.
In (20b), the Tokyo cognates for ‘ centipede’ (3.2) and ‘ two’
(3.2) have different accent patterns. Looking into accentuation
of other 3.2-nouns such as sakura ‘ cherry’ , tubasa ‘ wing’ ,
tokage ‘ lizard’ , aida ‘ interval’ , and many more, I have found
that they are atonic, namely/OOO/, which is identical to that of
‘ centipede’ . Therefore, the accent of the word hutatu ˘ ‘ two’
must be irregular. In addition, in Hiroshima dialect (which has
310 ACCENTUAL HISTORY

Tokyo-type accent), the accentuation of the word for ‘ two’ is

/OOO/ [LHH(H)]. This also confirms the irregularity of the
Tokyo accent for ‘two’.

(20) Reconstructed Proto-Mainland Japanese accent for

trisyllabic nouns
(a)
Tokyo Kyoto Myô Kagoshima PMJ
OOO OOO `OOO `OOO *`OOO
LHH(H) HHH(H) HHH(H) LHL ~ LLH(L) HHH(H)
‘mark’ (3.1) sirusi `sirusi `sirusi `sirusi *`syir´syi
‘smoke’(3.1) kemuri `kemuri `kemuri `kemuri *`keyburyi
‘forehead’ (3.1) hitai `hitai `fitai `hitai *`pyitapyi
‘yawn’ (3.1) akubi `akubi `akubi _akubi *`akubi

(b)
Tokyo Kyoto Myô Kagoshima PMJ
OOO OO˘O `OO˘O `OOO *`OOOˆ
LHH(H) LHL(L) HHL(L) LHL ~ LLH(L) HHF ~ HHH(L)
‘centipede’ (3.2) mukade muka˘de `muka˘de `mukade *mukadeyˆ
‘two’ (3.2) hutatu˘ huta˘tu `futa˘tu `hutatu *putatuˆ
‘wing’ (3.2) tubasa tuba˘sa `tuba˘sa `tubasa *tubasaˆ
‘lizard’ (3.2) tokage toka˘ge `toka˘ge `kokage *tokageˆ

(c)
Tokyo Kyoto Myô Kagoshima PMJ
OOO `OOO `OO˘O `OOO *`OOO˘
LHH(H) HHH(H) HHL(L) LHL ~ LLH(L) HHH(L)
‘interval’ (3.2) aida `aida `afi˘da `aida *apyida˘
‘cherry’ (3.2) sakura `sakura `saku˘ra `sakura *sakura˘
‘hackberry’ (3.2) enoki `enoki `eno˘k i `enoki *eyn´kiy˘
‘curtain’ (3.2) tobari `tobari `toba˘ri `tobari *tobari˘
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 311

(d)
Tokyo Kyoto Myô Kagoshima PMJ
O˘OO O˘OO O˘OO `OOO *O˘OÈO
HLL(L) HLL(L) HLL(L) LHL ~ LLH(L) HLH(H)
‘strength’ (3.3) tikara˘ t i˘kara t i˘kara `tikara *t i˘kaÈra
‘gold’ (3.3) k o˘gane k o˘gane k o˘gane `kogane *kwo˘gaÈne

(e)
Tokyo Kyoto Myô Kagoshima PMJ
OOO˘ O˘OO _OOO _OOO *OÈOO
LHH(L) HLL(L) LLL(L) LLH ~ LLL(H) LHH(H)
‘bag’ (3.4) hukuro˘ hu˘kuro _hukuro _hukuro *puÈkurwo
‘treasure’(3.4) takara˘ ta˘kara _takara _takara *taÈkara
‘mirror’(3.4) kagami˘ ka˘gami _kagami _kagami *kaÈgamyi
‘fan’ (3.4) o:gi˘ o :˘g i _afuki _o:gi *aÈpukyi

(f)
Tokyo Kyoto Myô Kagoshima PMJ
OOO˘ `OOO _OOO _OOO *_OOO
LHH(L) HHH(H) LLL(L) LLH ~ LLL(H) LLL(L)
‘calendar’(3.4) koyomi˘ `koyomi _koyomi -- * _k´y´myi
‘forest’ (3.4) hayasi˘ `hayasi _fayasi _hayasi * _payasi

(g)
Tokyo Kyoto Myô Kagoshima PMJ
O˘OO O˘OO OOÈO _OOO *OOÈO
HLL(L) HLL(L) LLH(H) LLH ~ LLL(H) LLH(H)
‘pillow’(3.5) ma˘kura ma˘kura makuÈra _makura *makuÈra
‘life’ (3.5) i˘noti i˘noti inoÈt i _inoti *i n´Èt i
‘tears’(3.5) na˘mida na˘mida namiÈda _namida *namyiÈta
‘firefly’ (3.5) h o˘taru h o˘taru -- _hotaru *potaÈru
‘figure’ (3.5) su˘gata su˘gata sugaÈta _sugata *sugaÈta
312 ACCENTUAL HISTORY

(h)
Tokyo Kyoto Myô Kagoshima PMJ
OO˘O O˘OO OOÈO _OOO *OOÊO
LHL(L) HLL(L) LLH(H) LLH ~ LLL(H) LLR ~ LLL(H)
‘heart’(3.5) koko˘ro k o˘koro kokoÈro _kokoro *k´k´Êr´
‘cousin’(3.5) ito˘k o i˘toko itoÈk o _itoko *itwoÊkwo

(i)20
Tokyo Kyoto Myô Kagoshima PMJ
OOO O˘OO OOÈO _OOO *OOÈO˘
LHH(H) HLL(L) LLH(H) LLH ~ LLL(H) LLH(L)
‘fat’ (3.5) abura a˘bura abuÈra _abura *abuÈra˘
‘pillar’ (3.5) hasira ha˘sira fasiÈra _hasira *pasiÈra˘

(j)
Tokyo Kyoto Myô Kagoshima PMJ
OOO _OOO OÈOO _OOO *OÈOO˘
LHH(H) LLH ~ LHH(H) LLH ~ LLL(H) LHH(L)
LLL(H)
‘eel’ (3.6) unagi _unagi uÈnagi _unagi *muÈnagi˘21
‘earthworm’(3.6) mimizu _mimizu miÈmizu _mimizu *miÈmizu˘

(k)
Tokyo Kyoto Myô Kagoshima PMJ
OOO OO˘O _OO˘O _OOO *OÈOOˆ
LHH(H) LHL(L) LHL(L) LLH ~ LLL(H) LHF ~ LHH(L)
‘medicine’ (3.7) kusuri kusu˘ri _kusu˘ri _kusuri *kuÈsuriˆ
‘whale’ (3.7) kuzira kuzi˘ra _kuzi˘ra _kuzira *kuÈdiraˆ
‘field’ (3.7) hatake hata˘ke _fata˘ke _hatake *paÈtakeˆ
‘behind’ (3.7) usiro usi˘ro -- _usiro *uÈsir´ˆ
‘washtub’ (3.7) tarai tara˘i _tara˘fi _tarai *taÈrapyiˆ

20
There are more examples: k ey ak i LHH(H) and sudare LHH(H) (Hirayama
1960: 221 and 386).
21
An Old Japanese cognate for ‘eel’ is munagyi (Martin 1987: 562).
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 313

(m)
Tokyo Kyoto Myô Kagoshima PMJ

O˘OO OO˘O _OO˘O _OOO *OOÈOˆ

HLL(L) LHL(L) LHL(L) LLH ~ LLL(H) LLF ~ LLH(L)
‘one person’(3.7) h i˘tori hito˘ri _fito˘ri _hitori *pyit´Èriˆ
‘helmet’(3.7) ka˘buto kabu˘t o _kabu˘t o `kabuto *kabuÈt oˆ
‘adult’(3.7) o˘tona oto˘na _oto˘na `otona *otoÈnaˆ
‘silkworm’ (3.7) ka˘iko kai˘k o _kafi˘k o `kaiko *kapyiÈkwoˆ22
‘reply’ (3.7) ta˘yori tayo˘ri _tayo˘ri `tayori *tayoÈriˆ
‘camellia’(3.7) tu˘baki tuba˘k i _tuba˘k i `tubaki *tubaÈkyiˆ
‘sickness’(3.7) ya˘mai yama˘i _yama˘fi `yamai * yamaÈpyiˆ23

As for (20a), Kagoshima cognates are normally high-register;

only ‘ yawn’ is low-register. Kagoshima high-register nouns
systematically correspond to traditional accent classes 3.1-3, and
Kagoshima low-register nouns to traditional accent classes 3.4-7.
Therefore, the accentuation of ‘ yawn’ in Kagoshima must be
irregular.

6.4 DEVELOPMENT OF THE ACCENT SYSTEM OF

MAINLAND JAPANESE DIALECTS FROM PROTO-
MAINLAND JAPANESE
In the previous section I have proposed a reconstruction of
Proto-Mainland Japanese accent as shown in (21). This section
will demonstrate how the accent systems of the modern mainland
Japanese dialects (Tokyo, Kyoto, and Kagoshima) as well as that
of Myôgishô have developed from the Proto-Mainland Japanese
accent.
In the present study I have claimed that the Kyoto accent did
not develop directly from the Myôgishô accent, although it is
widely believed that Kyoto dialect has developed from the
Myôgishô. It is obvious from the accentual correspondences that
Kyoto dialect is not a direct descendant of the Myôgishô dialect.

22
In Old Japanese, the cognate for ‘silkworm’ is k apy ik wo, consisting of two
morphemes (kapyi and kwo) (Martin 1987: 433).
23
In Old Japanese, for ‘sickness’ its cognate is yamapyi (Martin 1987: 572).
314 ACCENTUAL HISTORY

Therefore, I will account for the development of Kyoto accent

directly from Proto-Mainland Japanese, not from the Myôgishô.
Regarding Tokyo accent, I will also demonstrate that it a direct
descendant of Proto-Mainland Japanese and that it did not
develop from Myôgi-shô. The Kagoshima accent system may
also have developed directly from Proto-Mainland Japanese
accent, because there is no evidence or indication for its
development from Myôgishô.

(21) Proto-Mainland Japanese accent system

Monosyllabic Disyllabic
(a) *`oo [HH(H)] (a) *`OO [HH(H)]
(b) *o˘o [HL(L)] (b) *O˘O [HL(L)]
(c) *_oo [LL(H)] (c) *_OO [LL(L)]
(d) *oÈo [LH(H)] (d) *OÈO [LH(H)]
(e) *OO˘ [LH(L)]

Trisyllabic
(a) *`OOO [HHH(H)]
(b) *OOOˆ [HHF ~ HHH(L)]
(c) *OOO˘ [HHH(L)]
(d) *O˘OÈO [HLH(H)]
(e) *OÈOO˘ [LHH(L)]
(f) *OOÈO˘ [LLH(L)]
(g) *OOÈO [LLH(H)]
(h) *OOÊO [LLH ~ LLL(H)]
(i) *_OOO [LLL(L)]
(j) *OÈOO [LHH(H)]
(k) *OÈOOˆ [LHF ~ LHH(L)]
(m) *OOÈOˆ[LLF ~ LLH(L)]

6.4.1 Development of Myôgishô Accent from Proto-Mainland

Japanese
Most of the Proto-Mainland Japanese accent distinctions have
been retained in the Myôgishô dialect. Myôgishô and Proto-
Mainland Japanese monosyllabic accents are very similar, and
Myôgishô disyllabic accent is identical to those of Proto-
Mainland Japanese. For trisyllables, Proto-Mainland Japanese
underwent some changes.
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 315

Monosyllables: According to my hypothesis, the Proto-

Mainland Japanese monosyllabic accent system and that of
Myôgishô are almost identical. The only difference is in vowel
length. In the Myôgishô line, the Proto-Mainland Japanese long
vowel became short by undergoing Vowel Shortening.

(22) Myôgishô monosyllabic accent

PMJ V-shrt
(a) *`oo > `O

(b) *o˘o > O˘

(c) *_oo > _O

(d) *oÈo > ÈO

Disyllables: Just like its monosyllabic accent, the accent

system of Proto-Mainland Japanese disyllabic nouns remained
the same in the Myôgishô dialect. It did not undergo any
changes at all.

(23) Myôgishô disyllabic accent

PMJ
(a) *`OO = `OO

(b) *O˘O = O˘O

(c) *_OO = _OO

(d) *OÈO = OÈO

(e) *OO˘ = OO˘

Trisyllables: Unlike monosyllabic and disyllabic accent,

Myôgishô trisyllabic accent evolved by undergoing four
changes; i.e., Abductive Change, Contour Pitch Simplification
316 ACCENTUAL HISTORY

(C-smpl), Final Accent Deletion (F-dltn), Double Accent

Simplification (D-smpl).
(24) Myôgishô trisyllabic accent

abductive
PMJ change C-smpl F-dltn D-smpl
(a) *`OOO = `OOO = `OOO = `OOO = `OOO

(b) *OOOˆ
> `OOOˆ > `OO˘O = `OO˘O = `OO˘
(c) *OOO˘

(d) *O˘OÈO = O˘OÈO = O˘OÈO > O˘OO = O˘OO

(e) *OÈOO
> _OOO = _OOO = _OOO = _OOO
(f) *_OOO

(g) *OOÈO
> OOÈO = OOÈO
(h) *OOÊO
> OOÈO = OOÈO
(i) *OOÈO˘ = OOÈO˘ = OOÈO˘

(j) *OÈOO˘ = OÈOO˘ = OÈOO˘ > OÈOO = OÈOO

(k) *OÈOOˆ
> _OOOˆ > _OO˘O = _OO˘O = _OO˘O
(m) *OOÈOˆ

The first change, Abductive Change, caused a drastic shift by

merging accent classes. There are four mergers caused by
Abductive Change in the development of Myôgishô trisyllabic
accent. The first one put PMJ * /OOOˆ/ (24b) and * /OOO˘/ (24c)
together as one class (i.e., /`OOOˆ/). The second one is a merger
of PMJ * /OÈOO/ (24e) and * /_OOO/ (24f) as /_OOO/. The third
merger involves PMJ * /OOÈO/ (24g) and * /OOÊO/ (24h). All
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 317

became /OOÈO/. And the last one put two classes with final-
falling into one; i.e., PMJ * /OÈOOˆ/ and * /OOÈOˆ/ became
* /_OOOˆ/. As a result, with Abductive Change, the twelve
distinctions of Proto-Mainland Japanese merged as eight.
After Abductive Change, Contour Pitch Simplification
followed. This rule of accent change turned /_OOOˆ/ into
/_OO˘O/ (24k and m). When the application of Contour Pitch
Simplification was completed, /OOÈO˘/ (24i) and /OÈOO˘/ (24j)
underwent Final Accent Deletion and became /OOÈO/ and /OÈOO/
respectively. This rule deletes an accent on the final syllable of
trisyllabic words. In other words, it is not applicable to
monosyllables and disyllables. Because this rule was developed
in the Kyoto-type Japanese line, there are no final-accent
trisyllabic nouns in modern Kyoto-type Japanese.
Finally, PMJ * /O˘OÈO/ (24d) became /O˘OO/ by undergoing
Double Accent Simplification. This rule simply deletes the final
rising accent.

6.4.2 Development of Modern Kyoto Accent from Proto-

Mainland Japanese
Based on the proposal that Kyoto dialect developed from Proto-
Mainland Japanese, not from the dialect described in Myôgishô, I
demonstrate below the development of Kyoto accent from
Proto-Mainland Japanese. Its development is more complex than
that of Myôgishô. There are more stages of accentual changes
involved in the Kyoto line.
Monosyllables: In its evolvement, Kyoto monosyllables
underwent one change: Abductive Change.

(25) Kyoto monosyllabic accent

abductive
PMJ change
(a) *`oo = `oo

(b) *o˘o
> o˘o
(d) *oÈo

(c) *_oo = _oo

318 ACCENTUAL HISTORY

With Abductive Change, the four-way system of Proto-

Mainland Japanese accent evolved into three-way by merging
PMJ * /o ˘o/ (25b) and * /oÈo/ (25d) as /o˘o/. The Abductive
Change affected only tonic nouns. In addition, although Initial
Accent Gain played a role in the development of Kyoto
disyllabic and trisyllabic accent (see below), it did not affect its
monosyllables.
Disyllables: Kyoto disyllabic accent developed from Proto-
Mainland Japanese by undergoing Initial Accent Gain and
assimilation. First, by the rule of Initial Accent Gain, PMJ * /_OO/
(26c) became /O˘O/, resulting in merging with PMJ * /O˘O/ (26b).
When this happened, PMJ * /OÈO/ (26d) became /_OO/ in contrast
with * /`OO/ (26a). At this point, the rising accent (‘ È’ ) was no
longer distinctive. Later in the development, /_OO/ expanded its
domain for accent from the word to the phrase. Thus, its pitch
shape changed from LH(H) to LH ~ LL(H).

(26) Kyoto disyllabic accent

PMJ I-gain assml

(a) *`OO = `OO = `OO

(b) *O˘O
> O˘O = O˘O
(c) *_OO

(d) *OÈO ª _OO = _OO

[LH ~ LL(H)]

(e) *OO˘ = OO˘ > OOˆ

[LH(L)] [LF ~ LH(L)]

The next change, assimilation, turned PMJ * /OO˘/ into /OOˆ/.

This rule occurs at the end of words. Since none of the Kyoto
monosyllables and trisyllables is final-accent, this rule did not
take place in the monosyllables and trisyllables.
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 319

Trisyllables: In the development of the Kyoto trisyllabic

accent, there are five changes: Abductive Change, Contour Pitch
Simplification (C-smpl), Final Accent Deletion (F-dltn), Initial
Accent Gain (I-gain), and Double Accent Simplification (D-
smpl). They take place in order as listed.

(27) Kyoto trisyllabic accent

abductive
PMJ change C-smpl F-dltn I-gain D-smpl
(a) *`OOO ≡ `OOO = `OOO
> `OOO = `OOO = `OOO
(c) *OOO˘ = OOO˘ = `OOO˘

(b) *OOOˆ
(k) *OÈOOˆ > OOOˆ > OO˘O = OO˘O = OO˘O = OO˘O
(m) *OOÈOˆ

(e) *OÈOO
> _OOO = _OOO = _OOO > O˘OO
(f) *_OOO
> O˘OO
(d) *O˘OÈO = O˘OÈO = O˘OÈO = O˘OÈO
> O˘OÈO
(g) *OOÈO
(h) *OOÊO > OOÈO = OOÈO = OOÈO
(i) *OOÈO˘

(j) *OÈOO˘ = OÈOO˘ = OÈOO˘ > OÈOO ª _OOO = _OOO

In (27), under Abductive Change, there are three mergers. By

this change, the number of Proto-Mainland Japanese accent
distinctions was reduced to seven. The first one involves PMJ
* /OOOˆ/ (27b), * /OÈOOˆ/ (27k), and * /OOÈOˆ/ (27m). They all
became /OOOˆ/ [LHF ~ LHH(L)]. The second group of accent
classes that underwent Abductive Change includes PMJ * /OÈOO/
(27e) and * /_OOO/ (27f). Both merged as /_OOO˘/ [LHH(L)].
320 ACCENTUAL HISTORY

The last group involves PMJ * /OOÈO/ (27g), * /OOÊO/ (27h), and
*/OOÈO˘/ (27i). They merged as /OOÈO/.
At the second stage of the development, Contour Pitch
Simplification converted /OOOˆ/ to /OO˘O/ [LHL(L)]. This
change is followed by Final Accent Deletion, which deletes the
final accent of PMJ * /OOO˘/ (27c), and it merged with PMJ
* /`OOO/ (27a). As mentioned earlier, this rule is not applicable
to monosyllables or disyllables.
Another change contributing to the development of Kyoto
trisyllabic accent is Initial Accent Gain. This changed /_OOO/
(27e and f) and /OOÈO/ (27g, h, and i) to /O˘OO/ and /O˘OÈO/
respectively. The latter then merged with PMJ * /O˘OÈO/ (27d).
Furthermore, the change of /_OOO/ to /O˘OO/ caused /OÈOO/
(27j) to became /_OOO/ in contrast to /`OOO/.
At the final stage of the development, /O˘OÈO/ merges with
/O˘OO/ (27e and f) when it loses its rising accent by the Double
Accent Simplification rule. Kyoto trisyllables eventually have
four distinctive accent classes.
Finally, Proto-Mainland Japanese has locus and register
systems, which are typologically different from each other.
Similarly, Kyoto accent has both systems as well, although the
number of accentual distinctions is fewer than that of Proto-
Mainland Japanese. In short, typologically speaking, both the
Proto-Mainland Japanese and Kyoto accent systems are very
much alike.

6.4.3 Development of Modern Tokyo Accent from Proto-

Mainland Japanese
This section will account for the development of Tokyo accent
directly from Proto-Mainland Japanese. Compared with the
developments of Myôgishô and Kyoto accents, Tokyo accent
involves a rather complex process of changes. In the Tokyo line
there are six changes in total, including phonemic and phonetic
changes. But they do not apply to all nouns -- they apply when
conditions are met for the changes. Monosyllables underwent
four of the changes, disyllables five, and trisyllables five. I will
discuss how the Tokyo accent system evolved with the changes.
Monosyllables: As shown below, a series of four changes
contributed to the development of Tokyo monosyllabic accent.
The first change was Abductive Change, which merged PMJ
* /`oo/ (28a) and * /o˘o/ (28b) as high-register atonic, and also
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 321

put PMJ * /_oo/ (28c) and * /oÈo/ (28d) into one class, i.e., /_oo/.
Thus, at this point, the four-way distinctions of Proto-Mainland
Japanese became two-way, and distinctions were made by
register only.
The second change is Initial Accent Gain, which is applicable
only to low-initial nouns. Therefore, /_oo/ became /o˘o/. /`oo/ did
not undergo Initial Accent Gain, but its initial pitch became no
longer distinctive; thus it became /oo/ [HH(H)] at this point.
At the third stage, Tokyo dialect developed a phonetic rule
that lowers the pitch of the initial syllable when the pitch is not
distinctive. Thus, /oo/ [HH(H)] resulted in LH(H).
Towards the end of its development from Proto-Mainland
Japanese, Vowel Shortening took place. It made Proto-Mainland
Japanese bimoraic monosyllables monomoraic, namely /oo/ > /O/
and /o˘o/ > /O˘/.

(28) Tokyo monosyllabic accent

abductive
PMJ change I-gain I-low V-shrt
(a) *`oo
> `oo ≡ oo > oo > O
(b) *o˘o [HH(H)] [LH(H)] [L(H)]

(c) *_oo
> _oo > o˘o = o˘o > O˘
(d) *oÈo [H(L)]

Disyllables: As demonstrated below there are five changes in

the development of Tokyo disyllabic nouns: Abductive Change,
Initial Accent Gain, Accent Shift, again, Initial Accent Gain, and
Initial Pitch Lowering.
PMJ * /OÈO/ (29d) and /OO˘/ [LH(L)] (29e) merged as /OÈO/
by Abductive Change. Similarity in their pitch shapes triggered
the merger - low pitch in the initial syllable and high in the
second syllable.
At the second stage the application of the Initial Accent Gain
rule on PMJ * /_OO/ resulted in merging with PMJ * /O˘O/. When
this happened, the initial pitch of /OÈO/ (29d and e) became
322 ACCENTUAL HISTORY

distinctive; i.e., /OÈO/ > /_OO/. By the time of the change, the
number of accent distinctions was reduced to three.
The third stage Accent Shift, turned /O˘O/ (29b and c) into
/OO˘/. By the following change (Initial Accent Gain), /_OO/ (29d
and c) became /O˘O/. Due to this, the initial pitch of PMJ * /`OO/
became no longer distinctive.
Finally, the following phonetic rule, Initial Pitch Lowering,
lowered the initial pitch of /OO/ [HH(H)] (29a).

(29) Tokyo disyllabic accent

abductive
PMJ change I-gain shift I-gain I-low
(a) *`OO = `OO = `OO = `OO ≡ OO > OO
[LH(H)]

(b) *O˘O = O˘O

> O˘O > OO˘ = OO˘ = OO˘
(c) *_OO = _OO [LH(L)]

(d) *OÈO
> OÈO ª _OO = _OO > O˘O = O˘O
(e) *OO˘

Trisyllables: As with disyllabic accent, Tokyo trisyllabic

accent evolved with five changes. They are Abductive Change,
Initial Accent Gain, Accent Shift, Double Accent Simplification,
and Initial Pitch Lowering. As noticed, the changes in disyllables
are not identical to those of trisyllables because Initial Accent
Gain did not take place twice in trisyllables. Instead, Double
Accent Simplification took place.
First, Abductive Change caused three mergers of accent
classes. The following six accent classes: * /`OOO/ (30a),
* /OOOˆ/ (30b), * /OOO˘/ (30c), * /_OOO/ (30i), * /OÈOO/ (30j),
and * /OÈOOˆ/ (30k) merged as high-level atonic. Notice that the
initial pitch height is distinctive. PMJ * /OOÈO/ (30g) and
* /OOÈOˆ/ (30m) also merged. They became /OOÈO/. Lastly PMJ
* /OÈOO/ (30e) and * /_OOO/ (30f) merged, becoming /_OOO/.
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 323

Because of the mergers, the number of original Proto-Mainland

Japanese distinctions was reduced to five.
At the second stage of the development, /OOÈO/ (30g and l)
gained an initial accent, resulting in merger with PMJ * /O˘OÈO/
(30d). /_OOO/ (30e and f) and /OOÊO/ (30h) also gained the
initial accent, i.e., /O˘OO/ and /O˘OÊO/. The rule of Initial Accent
Gain is applicable only to low-initial nouns, thus /`OOO/ did not
undergo the change. However, at this point, the initial pitch of
/`OOO/ is no longer distinctive because /_OOO/ became /O˘OO/.

(30) Tokyo trisyllabic accent

abductive
PMJ change I-gain shift D-smpl I-low
(a) *`OOO
(b) *OOOˆ
(c) *OOO˘ > `OOO ≡ OOO = OOO = OOO > OOO
(i) *OOÈO˘ [HHH(H)] [LHH(H)]
(j) *OÈOO˘
(k) *OÈOOˆ

(d) *O˘OÈO = O˘OÈO

> O˘OÈO = O˘OÈO > O˘OO = O˘OO
(g) *OOÈO
> OOÈO
(m) *OOÈOˆ

(e) *OÈOO
> _OOO > O˘OO > OOO˘ = OOO˘ = OOO˘
(f) *_OOO

(h) *OOÊO = OOÊO > O˘OÊO > OO˘ÊO > OO˘O > OO˘O
[HHL(L)] [LHL(L)]

The next sound change, Accent Shift, converted /O˘OÊO/

(30h) to /OO˘ÊO/. However, /O˘OÈO/ (30d, g, and m) did not
undergo the change because the final-rising accent prevents the
initial accent from shifting. On the other hand, /O˘OÊO/ is
324 ACCENTUAL HISTORY

phonetically HLR ~ HLL(H). Therefore, when the shift took

place, it was probably HHL(H) or HHR ~ HHL(H). As for
/O˘OO/ (30e and f), undergoing Accent Shift, it became /OOO˘/.
That is, the locus is placed on the final syllable, skipping the
second syllable. This type of accent shift is seen in the
development of Kagoshima Japanese, as well.
At the fourth stage, Double Accent Simplification deletes the
rising accent of /O˘OÈO/ and /OO˘ÊO/. They eventually became
/O˘OO/ and /OO˘O/ respectively.
Finally, the Initial Pitch Lowering rule lowered the initial pitch
of /OOO/ [HHH(H)] and /OO˘O/ [HHL(L)].
With regard to the ordering of changes, Abductive Change
precedes any other changes. Initial Accent Gain occurs before
and after Accent Shift. Double Accent Simplification should be
placed after Accent Shift. And Initial Pitch Lowering should take
place towards the end of the development.
Concerning typology of accent, as shown above, Tokyo
accent lost Proto-Mainland Japanese register system in the
course of its development. By the time the second application of
Initial Accent Gain occurred, the register distinctions completely
disappeared from the whole system -- trisyllables lost the register
distinction at the beginning of the development.

6.4.4 Development of Modern Kagoshima Accent from Proto-

Mainland Japanese
In the Kagoshima dialect, the distinctions of Proto-Mainland
Japanese accent were reduced to only two classes. This drastic
change resulted from an application of a series of four changes
(see below for details). While disyllables and trisyllables
underwent four, monosyllables underwent only two of them.
This section explains how Kagoshima Japanese developed a
simple accent system from the complex system of Proto-
Mainland Japanese in reference to these changes.
Monosyllables: As mentioned above, the development of
Kagoshima monosyllabic accent is simpler than that of disyllabic
and of trisyllabic accents. The accent system of Kagoshima
monosyllables evolved from Proto-Mainland Japanese accent by
undergoing Abductive Change and Vowel Shortening. The
development of monosyllables is as shown in (31).
First, Abductive Change merged PMJ * /`oo/ [HH(H)] (31a)
and * /o˘o/ [HL(L)] (31b) as /`oo/ [HL(L)], and it also put PMJ
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 325

* /_oo/ [LL(H)] (31c) and * /oÈo/ [LH(H)] (31d) into one class,
namely /_oo/ [LH(H)]. Notice that Kagoshima has only register
distinctions: all nouns belong to either high- or low-register. Due
to this change, a four-way system of Proto-Mainland Japanese
became two-way. At the following stage, under the effect of
Vowel Shortening, /`oo/ [HL(L)] and /_oo/ became F ~ H(L) and
R ~ L(H) respectively.

(31) Kagoshima monosyllabic accent

abductive
PMJ change V-shrt
(a) *`oo
> `oo > `O
(b) *o˘o [HL(L)] [F ~ H(L)]
[HL(L)]

(c) *_oo
> _oo > _O
(d) *oÈo [LH(H)] [R ~ L(H)]
[LH(H)]

Disyllables: The development of Kagoshima disyllabic accent

involves Accent Shift, Abductive Change, Domain Change (D-
chng), and Initial Pitch Lowering (I-low). They took place in the
order as listed.
First, Accent Shift shifted the accent of the initial-accent
nouns; /O˘O/ became /`OO/ [HH(L)]. Due to this change, a
merger of PMJ * /`OO/ [HH(H)] (32a) and * /O˘O/ (32b)
occurred. While all Proto-Mainland Japanese high-initial nouns
have become nouns with high-register accent, all Proto-Mainland
Japanese low-initial nouns were merged into /_OO/ [LH(H)] by
Abductive Change.
326 ACCENTUAL HISTORY

(32) Kagoshima disyllabic accent

abductive
PMJ shift change D-chng I-low
(a) *`OO = `OO
> `OO > `OO > `OO
(b) *O˘O > `OO˘ [HH(L)] [HL ~ [HL ~
HH(L)] LH(L)]

(c) *_OO = _OO

(e) *OO˘ ª _OO˘ > _OO > _OO = _OO
(d) *OÈO = OÈO [LH(H)] [LH ~
LL(H)]

Towards the end of the development, /`OO/ [HH(L)] and

/_OO/ [LH(H)] were treated phonetically by the speakers of the
language as penultimate-accent and final-high rising accent
respectively. Namely, the former became /`OO/ [HL ~ HH(L)]
and the latter /_OO/ [LH ~ LL(H)]. And finally, /`OO/ underwent
Initial Pitch Lowering, which turned it into HL ~ LH(L).
Trisyllables: The same changes observed in the development
of disyllabic accent also affected trisyllabic accent. First, PMJ
* /O˘OO/ (33d) became /OOO˘/ by the application of Accent
Shift, and it merged with PMJ * /OOOˆ/ (33b) and * /OOO˘/ (33c).
Low-initial words remained the same.
At the second stage, Abductive Change merged PMJ * /`OOO/
(33a) and /OOO˘/ (33b, c, and d) as /`OOO/ [HHH(L)], and all
low-initial words were also put into one class, i.e., /_OOO/ [LLH
~ LLL(H)].
Following Abductive Change, Domain Change took place. It
produced a phonetic change in high-register nouns (namely,
/OOO˘/ [HHH(L)]) to HHL ~ HHH(L): the speakers of the
Kagoshima dialect began to consider it as phonetically
penultimate accent. At the end, the high-register nouns lowered
the pitch except for the penultimate syllable; i.e., HHL ~
HHH(L) changed to LHL ~ LLH(L).
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 327

(33) Kagoshima trisyllabic accent

abductive
PMJ shift change D-chng I-low
(a) *`OOO = `OOO
> `OOO > `OOO > `OOO
(b) *OOOˆ [HHH(L)] [HHL ~ [LHL ~
(c) *OOO˘ > OOO˘ HHH(L)] LLH(L)]
(d) *O˘OO [HHH(L)]

(e) *OÈOO = OÈOO

(f) *_OOO = _OOO
(g) *OOÈO = OOÈO
(h) *OOÊO = OOÊO > _OOO = _OOO = _OOO
(i) *OOÈO˘ = OOÈO˘ [LLH ~
(j) *OÈOO˘ = OÈOO˘ LLL(H)]
(k) *OÈOOˆ = OÈOOˆ
(m) *OOÈOˆ = OOÈOˆ

Regarding the typology of accent, Kagoshima has the

distinctions between high and low register. All Proto-Mainland
Japanese high-initial nouns have become high-register and all
Proto-Mainland Japanese low-initial nouns have merged as low-
register in Kagoshima. The Proto-Mainland Japanese accent
system has lost a number of distinctions in the Kagoshima
dialect. As a result, the locus-register system of Proto-Mainland
Japanese became the register system in Kagoshima.

6.5 CONCLUSION
This chapter has discussed the reconstruction of Proto-Mainland
Japanese accent and also the development of Myôgishô, Kyoto,
Tokyo, and Kagoshima accent from Proto-Mainland Japanese.
Among those Mainland Japanese dialects, the dialect described
in Myôgishô underwent the simplest development. And as far as
the number of changes is concerned, the developments of Tokyo
accent are the most complex. However, the development of
Kagoshima shows the most drastic change from Proto-Mainland
328 ACCENTUAL HISTORY

Japanese accent. Namely, all distinctions in Proto-Mainland

Japanese were reduced to only two in Kagoshima.
Comparing changes involved in the development, Myôgishô
and Kyoto share a very similar series of changes. The only
difference is that Kyoto underwent Initial Accent Gain. Unlike
the others, Tokyo has developed the Initial Pitch Lowering rule.
Kagoshima has Domain Change in its development. It is not seen
in Myôgishô, Kyoto, or Tokyo.
From the point of view of typology, the locus-register system
of Proto-Mainland Japanese has been retained in the Kyoto-type
dialects of Mainland Japanese. However, Proto-Mainland
Japanese accent has evolved to the locus system of Tokyo and
the register system of Kagoshima.
 7

RECONSTRUCTION OF
PROTO-JAPONIC ACCENT

7.1 INTRODUCTION

I n this chapter I will establish correspondence sets comparing

accentual patterns of Proto-Ryukyuan and Proto-Mainland
Japanese cognates, and then reconstruct Proto-Japonic accent for
monosyllables, disyllables, and trisyllables. In the previous sections
of this book the Proto-Ryukyuan and Proto-Mainland Japanese
accent systems have been reconstructed as shown in (1) and (2)
below. Summarizing the characteristics of the accent systems,
Proto-Mainland Japanese accent is more complex than Proto-
Ryukyuan accent; in other words, Proto-Mainland Japanese has
more distinctions in accent for monosyllables, disyllables, and
trisyllables. From the viewpoint of typology, both Proto-Ryukyuan
and Proto-Mainland Japanese make use of locus and register
distinctions in their accent systems. I will also discuss how these
two different accent systems came to coexist in the system.

(1) Proto-Ryukyuan accent system

Monosyllables Disyllables
(a) *oo [HH(H)] (a) *`OO˘ [HH(L)]
(b) *`oo˘ [HH(L)] (b) *_OO˘ [LH(L)]
(c) *_oo˘ [LH(L)] (c) *ooOˆ [LHF ~ LHH(L)]
330 ACCENTUAL HISTORY

Trisyllables
(a) *`OOO [HHH(H)]
(b) *`OOO˘ [HHH(L)]
(c) *_OOO˘ [LHH(L)]
(d) *_OOO [LLH(H)]
(e) *OO˘O [LHL(L)]
(f) *OOOˆ [LHF ~ LHH(L)]

(2) Proto-Mainland Japanese accent system

Monosyllabic Disyllabic
(a) *`oo [HH(H)] (a) *`OO [HH(H)]
(b) *o˘o [HL(L)] (b) *O˘O [HL(L)]
(c) *_oo [LL(H)] (c) *_OO [LL(L)]
(d) *oÈo [LH(H)] (d) *OÈO [LH(H)]
(e) *OO˘ [LH(L)]

Trisyllabic
(a) *`OOO [HHH(H)]
(b) *OOOˆ [HHF ~ HHH(L)]
(c) *OOO˘ [HHH(L)]
(d) *O˘OÈO [HLH(H)]
(e) *OÈOO [LHH(H)]
(f) *_OOO [LLL(L)]
(g) *OOÈO [LLH(H)]
(h) *OOÊO [LLR ~ LLL(H)]
(i) *OOÈO˘ [LLH(L)]
(j) *OÈOO˘ [LHH(L)]
(k) *OÈOOˆ [LHF ~ LHH(L)]
(m) *OOÈOˆ [LLF ~ LLH(L)]

7.2 CORRESPONDENCES AND RECONSTRUCTION

This section displays correspondence sets between Proto-Ryukyuan
and Proto-Mainland Japanese accent for monosyllables, disyllables,
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 331

and trisyllables. Consequently, a reconstruction of Proto-Japonic

accent for each correspondence set is given.
Monosyllables: There are four sets of accent correspondences
between Proto-Ryukyuan and Proto-Mainland Japanese
monosyllables. Based on these, four accent distinctions for Proto-
Japonic monosyllabic nouns have been reconstructed. They are
*/`oo/, */o˘o/, */_oo/, and */oo˘/. I have reconstructed pitch accent
for the total of seven nouns. The first three classes have two nouns
for each class and the last has one (see below).
Proto-Japonic monosyllabic nouns are two moras long.
Comparing characteristics of Proto-Japonic accent with those of
Proto-Ryukyuan and Proto-Mainland Japanes, Proto-Japonic high-
register nouns (3a) are also high-level in both Proto-Ryukyuan and
Proto-Mainland Japanese. Proto-Japonic and Proto-Mainland
Japanese nouns in a correspondence set in (3c) are low-register
alone while Proto-Ryukyuan cognates are tonic. Furthermore, in
both Proto-Ryukyuan and Proto-Japonic accents are identical (3d).

(3) Reconstructed Proto-Japonic accent for monosyllabic nouns

(a)
PR PMJ PJ
*oo *`oo *`oo
HH(H) HH(H) HH(H)
‘hair’ (1.1)1 *khï: *`ke:y *`ka:y
‘blood’ (1.1) *ci: *`ti:y *`tu:y

(b)
PR PMJ PJ
*`oo˘ *o˘o *o˘o
HH(L) HL(L) HL(L)
‘leaf’ (1.2) *`pha:˘ *pa˘: *pa˘:
‘name’ (1.2) *`na:˘ *na˘: *na˘:

1
Numbers in parentheses indicate traditional accent classes.
332 ACCENTUAL HISTORY

(c)
PR PMJ PJ
*_oo˘ *_oo *_oo
LH(L) LL(L) LL(L)
‘tree’ (1.3) *_khï:˘ *_ki:y *_k´y:
‘eye’ (1.3) *_mï:˘ *_me:y *_may:

(d)
PR PMJ PJ
*_oo˘ *oÈo *oo˘
LH(L) LH(H) LH(L)
‘tooth’ (1.3) *_pha: *pÈa: *pa:˘

Disyllables: In a comparison of Proto-Ryukyuan disyllables

with corresponding Proto-Mainland Japanese cognates, eight sets
of accentual correspondences are recognized. These
correspondences enable us to reconstruct eight accent classes for
Proto-Japonic disyllables. For convenience, (4) is a summary of the
correspondences and reconstructed Proto-Japonic accent forms.
Reconstructed individual words for the correspondences are listed
in (5).

(4) Summary of correspondences and reconstructed Proto-Japonic

accent for disyllabic nouns
PR PMJ PJ
a *`OO˘ *`OO HH(H) *`OO HH(H)
b HH(L) *O˘O HL(L) *O˘O HL(L)
c *_OO˘ LH(L) *_OO *_ooO
d *ooOˆ LL(L) LLL(L)
LHF ~ LHH(L)
e *_OO˘ LH(L) *OÈO *OÈO LH(H)
f *ooOˆ LH(H) *ooÈO
LHF ~ LHH(L) LLH(H)
g *_OO˘ LH(L) *OO˘ *OO˘ LH(L)
h *ooOˆ LH(L) *ooO˘
LHF ~ LHH(L) LHH(L)
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 333

Comparing Proto-Ryukyuan and Proto-Mainland Japanese

disyllabic accent systems with Proto-Japonic in the correspondence
sets in (4), it is apparent that Proto-Japonic accent is much more
similar to that of Proto-Mainland Japanese than to Proto-Ryukyuan
accent. In fact, the only difference between these is vowel length in
the initial syllable in (4d, f, and h). The reconstruction of the vowel
length in Proto-Japonic is based completely on the evidence in
Proto-Ryukyuan forms. In other words, Proto-Japonic length was
retained in the Proto-Ryukyuan line, but not in the Proto-Mainland
Japanese line. (See the discussion on vowel length in chapter 5 of
this book.)

(5) Reconstructed Proto-Japonic accent for disyllabic nouns

(a)
PR PMJ PJ
*`OO˘ *`OO *`OO
HH(L) HH(H) HH(H)
‘nose’ (2.1) *`phana˘ *`pana *`pana
‘cow’ (2.1) *`/usi˘ *`usi *`usi
‘bird’ (2.1) *`thuri˘ *`t´ri *`t´ri
‘loins’ (2.1) *`khusi˘ *`k´si *`k´si
‘wind’ (2.1) *`khadi˘ *`kazey *`kansay
‘beard’ (2.1) *`pigi˘ *`pyinkey *`pyinkay

(b)
PR PMJ PJ
*`OO˘ *O˘O *O˘O
HH(L) HL(L) HL(L)
‘stone’ (2.2) *`/isi˘ *i˘si *e˘si
‘paper’ (2.2) *`khabi˘ *ka˘byi *ka˘mpyi
‘bridge’ (2.2) *`phasi˘ *pa˘si *pa˘si
334 ACCENTUAL HISTORY

(c)
PR PMJ PJ
*_OO˘ *_OO *_OO
LH(L) LL(L) LL(L)
‘flower’ (2.3) *_phana˘ *_pana *_pana
‘mountain’ (2.3) *_yama˘ *_yama *_yama
‘cloud’ (2.3) *_khumu˘ *_kumwo *_kumwo
‘dog’ (2.3) *_/inu˘ *_inu *_enu
‘mud’ (2.3) -- *_doro *_ntoro

(d)
PR PMJ PJ
*ooOˆ *_OO *_ooO
LHF ~ LL(L) LLL(L)
LHH(L)
‘bone’ (2.3) *phu:nïˆ *_pone *_po:ne

(e)
PR PMJ PJ
*_OO˘ *OÈO *OÈO
LH(L) LH(H) LH(H)
‘board’ (2.4) *_/ita˘ *iÈta *iÈta
‘shoulder’ (2.4) *_khata˘ *kaÈta *kaÈta

(f)
PR PMJ PJ
*ooOˆ *OÈO *ooÈO
LHF ~ LH(H) LLH(H)
LHH(L)
‘sea’ (2.4) */u:miˆ *uÈmyi *u:Èmyi
‘breath’ (2.4) */i:kiˆ *iÈkyi *i:Èkyi
‘boat’ (2.4) *phu:niˆ *puÈney *pu:Ènay
‘mortar’ (2.4) */u:siˆ *uÈsu *u:Èsu
‘chopsticks’(2.4) *pha:siˆ *paÈsiy *pa:Èsuy
‘needle’ (2.4) *pha:riˆ *paÈriy *pa:Èruy
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 335

(g)
PR PMJ PJ
*_OO˘ *OO˘ *OO˘
LH(L) LH(L) LH(L)
‘sweat’ (2.5) *_/asi˘ *ase˘ *ase˘
‘rain’ (2.5) *_/amï˘ *amey˘ *amay˘

(h)
PR PMJ PJ
*ooOˆ *OO˘ *ooO˘
LHF ~ LH(L) LHH(L)
LHH(L)
‘bridegroom’ *mu:khuˆ *mwokwo˘ *mwo:kwo˘
(2.5)
‘voice’ (2.5) *khu:ïˆ *k´we˘ *k´:way˘
‘shadow’ (2.5) *kha:gïˆ *kagey˘ *ka:nkay˘
‘bucket’ (2.5) *u:khïˆ *wokey˘ *wo:kay˘

A number of Proto-Japonic disyllabic cognates have been

reconstructed. However, the number of cognates for each category
varies. For instance, there are six nouns for PJ */`OO/ (5a) and
*/ooÈO/ (5f), but one for PJ */_ooO/ (5d). Although there is only
one cognate for the correspondence (5d), I consider it as a
distinctive accent class because the existence of two subclasses
among 2.3-5 nouns is supported by evidence in Ryukyuan dialects
(see 5.3.3 of chapter 5).
Trisyllables: For trisyllables, twelve Proto-Japonic accent
classes are reconstructed on the basis of the correspondence sets
shown in (6). Reconstructed trisyllabic nouns for each
correspondence set are listed under (7) below.
The Proto-Japonic trisyllabic accent system is much more
similar to that of Proto-Mainland Japanese than it is to Proto-
Ryukyuan accent. The differences between Proto-Japonic accent
and Proto-Mainland Japanese accent appear in (6i) and (6j). In
other words, Proto-Ryukyuan underwent more changes than Proto-
Mainland Japanese did. As a result, Proto-Mainland Japanese
336 ACCENTUAL HISTORY

retained more characteristics of Proto-Japonic accent than Proto-

Ryukyuan.

(6) Summary of accentual correspondences and reconstruction of

Proto-Japonic accent for trisyllabic nouns
PR PMJ PJ
a *`OOO *`OOO HHH(H) *`OOO HHH(H)
b HHH(H) *OOOˆ *`OOOˆ
HHF ~ HHH(L) HHF ~ HHH(L)
c *OOO˘ HHH(L) *`OOO˘ HHH(L)
d *`OOO˘ *O˘OÈO *O˘OÈO
HHH(L) HLH(H) HLH(H)
e *_OOO˘ *OÈOO *OÈOO
LHH(L) LHH(H) LHH(H)
f *_OOO LLH(H) *_OOO LLL(L) *_OOO LLL(L)
g *_OOO˘ *OOÈO LLH(H) *OOÈO LLH(H)
h LHH(L) *OOÊO *OOÊO
LLR ~ LLL(H) LLR ~ LLL(H)
i *OOÈO˘ LLH(L) *_OOO˘ LHH(L)
j *OO˘O LHL(L) *OÈOO˘ LHH(L) *OO˘O LHL(L)
k *OOOˆ *OÈOOˆ *OÈOOˆ
LHF ~ LHH(L) LHF ~ LHH(L) LHF ~ LHH(L)
m *OOÈOˆ *OOÈOˆ
LLF ~ LLH(L) LLF ~ LLH(L)

Looking into corresponding accent patterns, it is revealing that

high-level atonic accent (6a) is stable among all the accent types.
From Proto-Japonic on, this accent type remained unchanged in
both the Proto-Ryukyuan and Proto-Mainland Japanese lines of
development. In (6i) all the classes are final-accent, though the
phonetic pitch shapes vary.
The first four of the correspondences are high-initial and the rest
are low-initial. The high-initial and low-initial accent types
systematically correspond to Myôgishô high-initial and low-initial
classes respectively. Furthermore, some of the Myôgishô accent
classes correspond to more than one Proto-Japonic accent.
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 337

Myôgishô accent class 3.2 corresponds to both PJ */`OOOˆ/ (6b)

and */`OOO˘/ (6c). Myôgishô accent class 3.4 corresponds to PJ
*/OÈOO/ (6e) and */_OOO/ (6f). Myôgishô accent class 3.5
corresponds to PJ */OOÈO/ (6g), */OOÊO/ (6h), and */_OOO˘/ (6i).
Myôgishô accent class 3.7 corresponds to PJ */OÈOOˆ/ (6k) and
*/OÈOOˆ/ (6m). This suggests that these Myôgishô accent classes
result from a merger of Proto-Japonic accent distinctions.
Some Proto-Ryukyuan cognates are missing for the
correspondences in (7). This is simply because the data available to
me do not list them: ‘wing’ (3.2), ‘lizard’ (3.2), ‘interval’ (3.2),
‘forest’ (3.4), and ‘sickness’ (3.7). Notice that the cognate for
‘tears’ (3.5) is disyllabic, i.e., *na:da LLL(L).

(7) Reconstructed Proto-Japonic accent for trisyllabic nouns

(a)
PR PMJ PJ
*`OOO *`OOO *`OOO
HHH(H) HHH(H) HHH(H)
‘mark’ (3.1) *`sirusi *`syir´syi *`syir´syi
‘smoke’ (3.1) *`khïbusi *`keyburyi *`kaympuryi
‘forehead’ (3.1) *`phitai *`pyitapyi *`pyitapyi
‘yawn’ (3.1) *`/akubi *`akubi *`akumpi

(b)
PR PMJ PJ
*`OOO *OOOˆ *`OOOˆ
HHH(H) HHF ~ HHF ~
HHH(L) HHH(L)
‘centipede’ (3.2) *`mukadi *mukadeˆ *mukanteˆ
‘two’ (3.2) *`futacï *futatuˆ *putatuˆ
‘wing’ (3.2) -- *tubasaˆ *tumpasaˆ
‘lizard’ (3.2) -- *tokageˆ *tokankeˆ
338 ACCENTUAL HISTORY

(c)
PR PMJ PJ
*`OOO *OOO˘ *`OOO˘
HHH(H) HHH(L) HHH(L)
‘interval’ (3.2) -- *apyida˘ *`apyinta˘
‘cherry’ (3.2) *`sakura *sakura˘ *`sakura˘

(d)
PR PMJ PJ
*`OOO˘ *O˘OÈO *O˘OÈO
HHH(L) HLH(H) HLH(H)
‘strength’ (3.3) *`cikara˘ *ti˘kaÈra *ti˘kaÈra
‘gold’ (3.3) *`khugani˘ *kwo˘gaÈney *kwo˘nkaÈnay

(e)
PR PMJ PJ
*_OOO˘ *OÈOO *OÈOO
LH(L) LHH(H) LHH(H)
‘bag’ (3.4) *_phukuru˘ *puÈkurwo *puÈkurwo
‘treasure’ (3.4) *_thakara˘ *taÈkara *taÈkara
‘mirror’ (3.4) *_khagami˘ *kaÈgamyi *kaÈnkamyi
‘fan’ (3.4) *_augi˘ *aÈpukyi *aÈpukyi

(f)
PR PMJ PJ
*_OOO *_OOO *_OOO
LLH(H) LLL(L) LLL(L)
‘calendar’ (3.4) *_khuyumi *_k´y´myi *_k´y´myi
‘forest’ (3.4) -- *_payasi *_payasi
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 339

(g)
PR PMJ PJ
*_OOO˘ *OOÈO *OOÈO
LHH(L) LLH(H) LLH(H)
‘pillow’ (3.5) *_makura˘ *makuÈra *makuÈra
‘life’ (3.5) *_/inuti˘ *in´Èti *in´Èti
‘tears’ (3.5) *na:da LL(L) *namyiÈta *namyiÈta

(h)
PR PMJ PJ
*_OOO˘ *OOÊO *OOÊO
LHH(L) LLR ~ LLLR ~
LLL(H) LLLL(H)
‘heart’ (3.5) *_khukuru˘ *k´k´Êr´ *k´k´Êr´
‘cousin’ (3.5) *_/itoko˘ *itwoÊkwo *_itwoÊkwo

(i)
PR PMJ PJ
*_OOO˘ *OOÈO˘ *_OOO˘
LHH(L) LLH(L) LHH(L)
‘fat’ (3.5) *_/abura˘ *abuÈra˘ *_ampura˘
‘pillar’ (3.5) *_phasira˘ *fasiÈra˘ *_pasira˘

(j)
PR PMJ PJ
*OO˘O *OÈOO˘ *OO˘O
LHL(L) LHH(L) LHL(L)
‘crow’ (3.6) *gara˘si *kaÈrasu˘ *kara˘su
‘eel’ (3.6) */una˘gi *uÈnagi˘ *una˘nki
‘earthworm’ (3.6) *mimi˘zi *miÈmizu˘ *mimi˘nsu
340 ACCENTUAL HISTORY

(k)
PR PMJ PJ
*OOOˆ *OÈOOˆ *OÈOOˆ
LHF ~ LHF ~ LHF ~
LHH(L) LHH(L) LHH(L)
‘medicine’ (3.7) *khusuriˆ *kuÈsuriˆ *kuÈsuriˆ
‘whale’ (3.7) *khudiraˆ *kuÈziraˆ *kuÈnsiraˆ
‘field’ (3.7) *hatakiˆ *paÈtakeˆ *paÈtakayˆ

(m)
PR PMJ PJ
*OOOˆ *OOÈOˆ *OOÈOˆ
LHF ~ LLF ~ LLF ~
LHH(L) LLH(L) LLH(L)
‘helmet’ (3.7) *khabutuˆ *kabuÈtoˆ *khampuÈtoˆ
‘silkworm’ (3.7) *kaiguˆ *kapyiÈkwoˆ *kapyiÈkwoˆ
‘sickness’ (3.7) -- *yamaÈpyiˆ *yamaÈpyiˆ

7.3 DEVELOPMENT OF PROTO-RYUKYUAN AND

PROTO-MAINLAND JAPANESE ACCENT FROM PROTO-
JAPONIC ACCENT
This section consists of two parts. The first part accounts for the
development of Proto-Ryukyuan accent from Proto-Japonic and the
second part is the development of Proto-Mainland Japanese accent
from Proto-Japonic. Both parts deal with monosyllabic, disyllabic,
and trisyllabic accent classes. A summary of the reconstructed
Proto-Japonic accent system is shown below:

(8) Proto-Japonic accent system

Monosyllabic
(a) *`oo [HH(H)]
(b) *o˘o [HL(L)]
(c) *_oo [LL(L)]
(d) *oo˘ [LH(L)]
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 341

Disyllabic
(a) *`OO [HH(H)]
(b) *O˘O [HL(L)]
(c) *_OO [LL(L)]
(d) *_ooO [LLL(L)]
(e) *OÈO [LH(L)]
(f) *ooÈO [LLH(H)]
(g) *OO˘ [LH(L)]

Trisyllabic
(a) *`OOO [HHH(L)]
(b) *OOOˆ [HHF ~ HHH(L)]
(c) *`OOO˘ [HHH(L)]
(d) *O˘OÈO [HLH(H)]
(e) *OÈOO [LHH(H)]
(f) *_OOO [LLL(L)]
(g) *OOÈO [LLH(H)]
(h) *OOÊO [LLR ~ LLL(H)]
(i) *_OOO˘ [LHH(L)]
(j) *OO˘O [LHL(L)]
(k) *OÈOOˆ [LHF ~ LHH(L)]
(m) *OOÈOˆ [LLF ~ LLH(L)]

7.3.1 Development of Proto-Ryukyuan Accent from Proto-

Japonic Accent
The accent system of Proto-Ryukyuan nouns is very different from
that of Proto-Japonic nouns. The Proto-Japonic accent system is
much more complex than that of Proto-Ryukyuan in terms of both
the number of accent distinctions and accent patterns. The
development of Proto-Ryukyuan accent from Proto-Japonic accent
will be explained below.
Monosyllables: There were two changes in the development of
Proto-Ryukyuan monosyllabic accent: Accent Shift and Abductive
Change. Due to these changes, the four-way distinctions of the
342 ACCENTUAL HISTORY

Proto-Japonic accent system became three-way in the Proto-

Ryukyuan line.
First, PJ */o˘o/ (9b) underwent Accent Shift, becoming /`oo˘/.
This change caused a secondary change to PJ */oo˘/ (9d). Namely,
word-initial register for final-accent nouns became distinctive: PJ
*/oo˘/ became /_oo˘/.

(9) Development of Proto-Ryukyuan monosyllabic accent

abductive
PJ shift change
(a) *`oo = `oo ≡ oo

(b) *o˘o > `oo˘ = `oo˘

(c) *_oo = _oo

> _oo˘
(d) *oo˘ ª _oo˘

Second, by Abductive Change the two low-register classes, i.e.,

PJ */_oo/ (9c) and /_oo˘/ (9d), merged as /_oo˘/. High-register
classes retained their distinctions. Moreover, it should be noted that
in Proto-Ryukyuan, disyllabic and trisyllabic high-register atonic
and high-register final accent classes merged due to the Abductive
Change, while in monosyllables they did not merge.
Disyllables: In the history of Proto-Ryukyuan disyllabic accent,
there are three stages of accentual changes. They can be ordered as:
Accent Shift, Abductive Change, and Assimilation (assml).
In the first stage, Accent Shift converted PJ */O˘O/(10b) into
/`OO˘/. The initial pitch also became distinctive in contrast with PJ
*/OO˘/ [LH(L)] (10g), whose initial pitch became distinctive as
well.
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 343

(10) Development of Proto-Ryukyuan disyllabic accent

abductive
PJ shift change assml
(a) *`OO = `OO
> `OO˘ = `OO˘
(b) *O˘O > `OO˘

(c) *_OO = _OO

(e) *OÈO = OÈO > _OO˘ = _OO˘
(g) *OO˘ ª _OO˘

(d) *_ooO = _ooO

(f) *ooÈO = ooÈO > _ooO˘ > ooOˆ
(h) *ooO˘ ª _ooO˘

In the second stage Abductive Change caused three mergers of

Proto-Japonic accent classes, causing the eight distinctions of
Proto-Japonic to merge into three. First, PJ */`OO/ (10a) and
/`OO˘/ (10b) merged as /`OO˘/. The second change involves three
classes, PJ */_OO/ (10c), */OÈO/ (10e), and /_OO˘/ (10g). They
were recategorized as /_OO˘/. The third merger includes words
with initial-syllable long vowel. PJ */_ooO/ (10d), */ooÈO/ (10f),
and */_ooO˘/ (10h) were turned into /_ooO˘/. As a result, there
were now three distinctions in register and vowel length.
In the final stage of the development /_ooO˘/ (10e, f, and h)
underwent Assimilation, which changed final accent (/_ooO˘/) to
final falling (/ooOˆ/). Note that this rule applied only to /_ooO˘/,
but not to the other final-accent nouns (/`OO˘/ and /_OO˘/). It also
did not take place in the development of Proto-Ryukyuan
trisyllabic accent. Taking these circumstances into consideration,
the only linguistic condition that seems to trigger Assimilation is
the initial-syllable vowel length or three-mora disyllabic final-
accent nouns in the Proto-Ryukyuan.
344 ACCENTUAL HISTORY

Trisyllables: As demonstrated in (11), the development of

Proto-Ryukyuan trisyllabic accent involved three types of changes -
- Double Accent Simplification, Accent Shift, and Abductive
Change. Note that Accent Shift took place twice and also that
Assimilation, seen in the development of Proto-Ryukyuan
disyllabic accent, did not take place.

(11) Development of Proto-Ryukyuan trisyllabic accent

abductive
PJ D-smpl shift change shift
(a) *`OOO = `OOO = `OOO
(b) *OOOˆ = OOOˆ = OOOˆ > `OOO = `OOO
(c) *`OOO˘ = `OOO˘ = `OOO˘

(d) *O˘OÈO > O˘OO > `OO˘O = `OO˘O > `OOO˘

(e) *_OÈOO = _OÈOO = OÈOO

(g) *OOÈO = OOÈO = OOÈO > OOO˘ ª _OOO˘
(h) *OOÊO = OOÊO = OOÊO
(i) *_OOO˘ = _OOO˘ = _OOO˘

(f) *_OOO = _OOO = _OOO = _OOO = _OOO

(j) *OO˘O = OO˘O ª _OO˘O = _OO˘O ≡ OO˘O

(k) *OÈOOˆ = OÈOOˆ = OÈOOˆ

> OOOˆ = OOOˆ
(m) *OOÈOˆ = OOÈOˆ = OOÈOˆ

With the application of the Double Accent Simplification rule,

PJ */O˘OÈO/ (11d) became /O˘OO/, and then underwent Accent
Shift. This shift resulted in /`OO˘O/. Due to this change, the initial
pitch of PJ */OO˘O/ [LHL(L)] (11j) became distinctive as well,
resulting in */_OO˘O/. At the end of the development another
Accent Shift eventually changed /`OO˘O/ to /`OOO˘/.
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 345

Next is Abductive Change. Just like its disyllables, Proto-

Japonic trisyllabic accent underwent three mergers. The first one
merged PJ */`OOO/ (11a), */OOOˆ/ (11b), and */`OOO˘/ (11c) as
high-register atonic, i.e., /`OOO/. Second, the rule put PJ */OÈOO/
(11e), */OOÈO/ (11g), */OOÊO/ (11h), and */_OOO˘/ (11i) into one
class, namely /OOO˘/. Accent classes involved in this merger are
either atonic or final-accent; their phonetic pitch shapes are similar
to each other, and they share the same word structure, with a long
vowel in the middle of words. The third merger involves PJ
*/OÈOOˆ/ (11k) and */OOÈOˆ/ (11m). These are distinctive accent
classes in Proto-Japonic, but are very similar both phonemically
and phonetically.
At the final stage, as mentioned earlier, another Accent Shift
took place. It converted PJ */`OO˘O/ (11d) into /`OOO˘/. Along
with this change, /_OO˘O/ (11j) lost its register distinction,
becoming /OO˘O/. However, the initial pitch of /OOO˘/ (11e, g, h,
and i) became distinctive in contrast with /`OOO˘/ (< PJ
*/O˘OÈO/) (11d). Namely, /OOO˘/ became /_OOO˘/.
With regard to the changes that took place in the development of
Proto-Ryukyuan accent, the changes can be seen in the following
order: Accent Shift > Abductive Change > Assimilation > Accent
Shift. Monosyllables underwent the first two, disyllables underwent
the first three, and trisyllables underwent three except Assimilation.
Trisyllables also underwent Double Accent Simplification, which is
not applicable to monosyllabic and disyllabic nouns.
Regarding typological change in accent, register distinctions
were kept throughout the development, although some classes lost
their register distinctions. Like Proto-Japonic, Proto-Ryukyuan is a
locus-register system.

7.3.2 Development from Proto-Mainland Japanese Accent from

Proto-Japonic Accent
Proto-Mainland Japanese accent system is very similar to Proto-
Japonic accent system. Therefore, it is natural that the number of
sound changes involved in the development of Proto-Mainland
346 ACCENTUAL HISTORY

Japanese accent is fewer than in Proto-Ryukyuan. In fact, there are

only two changes: Vowel Shortening and Accent Shift. The former
affects disyllables, and the latter monosyllables and trisyllables.
The following will reveal how Proto-Mainland Japanese accent
evolved from Proto-Japonic in detail.
Monosyllables: The development of Proto-Mainland Japanese
monosyllabic accent is rather simple, with only one change, namely
Abductive Change, which contributed to the development. When
this was applied, PJ *oo˘ (12d) became oÈo. The other Proto-
Japonic categories remained the same.

(12) Development of Proto-Mainland Japanese monosyllabic

accent
abductive
PJ change
(a) *`oo = `oo

(b) *o˘o = o˘o

(c) *_oo = _oo

(d) *oo˘ > oÈo

Disyllables: Compared with Proto-Japonic disyllabic accent, the

Proto-Mainland Japanese disyllabic accent system differs only in
the vowel length of the classes shown in (13d), (13f), and (13h).
Therefore, when Vowel Shortening took place, */_ooO/ (13d),
/ooÈO/ (13f), and */ooO˘/ (13h) merged with */_OO/ (13c), */OÈO/
(13e), and */OO˘/ (13g) respectively.
In the Proto-Mainland Japanese line, eight accent distinctions
became five. The register and locus distinctions of Proto-Japonic
were essentially retained as a system, because what the system lost
was its vowel length distinction.
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 347

(13) Development of Proto-Mainland Japanese disyllabic accent

PJ V-shrt
(a) *`OO = `OO

(b) *O˘O = O˘O

(c) *_OO
> _OO
(d) *_ooO

(e) *OÈO
> OÈO
(f) *ooÈO

(g) *OO˘
> OO˘
(h) *ooO˘

Trisyllables: The development of Proto-Mainland Japanese

trisyllabic accent involves Accent Shift only. As shown in (14),
Accent Shift caused a change in the system. More specifically, PJ
*/OO˘O/ [LHL(L)] (14j) became final-accent, distinguished from
*/OOÈO˘/ [LLH(L)] (14i) (< PJ *_OOO˘). When this change
occurred, the initial pitch of PJ *`OOO˘ [HHH(L)] (14c) became
nondistinctive, i.e., PJ *`OOO˘ > PMJ *OOO˘.
348 ACCENTUAL HISTORY

(14) Development of Proto-Mainland Japanese trisyllabic accent

PJ shift
(a) *`OOO = `OOO
(b) * OOOˆ = OOOˆ
(c) *`OOO˘ ≡ OOO˘
(d) * O˘OÈO = O˘OÈO
(e) * OÈOO = OÈOO
(f) *_OOO = _OOO
(g) * OOÈO = OOÈO
(h) * OOÊO = OOÊO
(i) *_OOO˘ ≡ OOÈO˘
(j) *OO˘O > OÈOO˘
(k) *OÈOOˆ = OÈOOˆ
(m) *OOÈOˆ = OOÈOˆ

7.4 CONCLUSION
I have presented my proposals in this chapter for a reconstruction
of Proto-Japonic accent. This is based on the Proto-Ryukyuan and
Proto-Mainland Japanese accents that I have reconstructed earlier
in this book. Dwelling more specifically on the reconstruction of
Proto-Japonic accent, the system of Proto-Japonic accent contains
four distinctions for monosyllables, eight for disyllables, and
twelve for trisyllables. The more syllables words have, the more
distinctions there are. As indicated above, contour pitches are seen
only at the end of words. As far as Proto-Japonic is concerned, they
do not occur in the middle of words. Proto-Japonic contour pitches
were retained in the Proto-Mainland Japanese line, but disappeared
in the Proto-Ryukyuan line. However, Proto-Ryukyuan developed
contour pitches on the final syllable of disyllabic words with the
initial-syllable long vowel (i.e., /ooOˆ/ [LHF ~ LHH(L)]).
A comparison of Proto-Ryukyuan, Proto-Mainland Japanese,
and Proto-Japonic makes it clear that the rising accent (namely ‘È’)
is another feature seen in a system where there are many
distinctions, such as Proto-Mainland Japanese and Proto-Japonic.
 RECONSTRUCTION OF PROTO-JAPONIC ACCENT 349

In Proto-Ryukyuan the rising accent is not distinctive. Proto-

Mainland Japanese inherited the feature directly from Proto-
Japonic, but Proto-Ryukyuan lost it.
As far as accent typology is concerned, both Proto-Ryukyuan
and Proto-Mainland Japanese, like Proto-Japonic, make use of
locus and register. It seems that level-pitch atonic, such as low-
level and high-level, are in general more stable than other types of
accents.
Regarding vowel length, Proto-Japonic vowel length was kept in
the Proto-Ryukyuan, but it was lost in Proto-Mainland Japanese. In
Proto-Japonic, the vowel length was a phonemic distinction.
Finally, on the developments of Proto-Ryukyuan and Proto-
Mainland Japanese accent, the development of Proto-Ryukyuan
accent is complex compared with that of Proto-Mainland Japanese.
In the Proto-Mainland Japanese the changes that took place are
Abductive Change in monosyllables, Vowel Shortening in
disyllables, and Accent Shift in trisyllables. On the other hand, in
the Proto-Ryukyuan line, Proto-Japonic accent underwent drastic
changes. One of the main changes is Abductive Change, which
reduced the number of accent distinctions. Needless to say, other
changes involved in the development of Proto-Ryukyuan played
crucial roles, as well.
 8

CONCLUSION

I n this book, I have proposed a reconstruction of Proto-Japonic

accent for monosyllabic, disyllabic, and trisyllabic nouns. In
order to pursue this, the Proto-Amami, Proto-Okinawa, Proto-
Miyako, and Proto-Yaeyama accent systems were reconstructed
first. Second, based on these protoforms, I presented a
reconstruction of Proto-Ryukyuan accent. Proto-Mainland Japanese
accent was reconstructed on the basis of three modern mainland
Japanese dialects (i.e., Tokyo, Kyoto, and Kagoshima) and the
dialect of Myôgishô. Finally, applying the comparative method to
Proto-Ryukyuan and Proto-Mainland Japanese, Proto-Japonic
accent was reconstructed. My reconstruction for Proto-Japonic
accent has four accent distinctions for monosyllabic nouns, eight
for disyllabic nouns, and twelve for trisyllabic nouns. From the
viewpoint of accent typology, Proto-Japonic accent has a register-
locus accent. In register system, the initial pitch of a word is
distinctive. As for locus, I claim that locations of both pitch rise
and pitch fall are distinctive. Furthermore, there are two types of
pitch fall and rise: a pitch falls between moras or syllables (i.e., ‘˘’),
a pitch falls within a syllable (i.e., ‘ˆ’), a pitch rises between moras
or syllables (i.e., ‘È’), and a pitch rises within a syllable (i.e., ‘Ê’).
I have also accounted for the development of the accent system
of each dialect or language presented in this book, including the
Proto-Amami, Proto-Okinawa, Proto-Miyako, Proto-Yaeyama, and
Proto-Mainland Japanese accent. My explanation is based on the
‘natural’ accent changes outlined in the chapter 4 of this book.
 CONCLUSION 351

Compared with previous studies of Japonic accent, my

reconstruction differs on the following points.
First, on methodology, my reconstruction is pursued solely by
the comparative method applied to modern dialects/languages and
Myôgishô. Failing to use the comparative method systematically,
Okuda and McCawley have reconstructed Proto-Japonic accent.
Matsumori’s reconstruction is based on the phenomenon of down
step; i.e., she did not use the comparative method.
Second, this book has pointed out that it is crucial to determine
natural accent change in order to reconstruct an accent system of a
proto-language. Rules of accent change have been investigated by
taking ‘naturalness’ into account, and formalization of ‘natural’
accent changes for the Japonic is pursued.
In the studies of Jpanonic accentology, rules of accent change
were formalized by Kindaichi in 1947. This was the first attempt
(and only one that I know) to formalize the rules of accent change
in the Japanese language. These rules of Kindaichi’s have been
thoroughly gone over with the viewpoint of ‘naturalness’, and I
have concluded that according to his rules it is difficult to account
for the accentual history of the Japonic because the rules are not
consistent - more specifically under Kindaichi’s rules accent shifts
both rightward and leftward.
With regard to descendant-ascendant relationship, scholars have
believed that Kyoto is a direct descendant of the dialect of
Myôgishô, and because of this, it has been difficult to account for
the accentual history of the Japonic accent. However, on the basis
of accent this book demonstrated that the Kyoto dialect cannot be
treated as a direct descendant of the Myôgishô dialect. This
discovery has allowed me to account for the more straightforward
development of dialects from their protoforms.
On typological difference in accent, I have discussed typology of
accent and explained how accent typology shifts one type from
another or vice versa. In general, the study of Japonic accent in
Japan has not put much focus on typology of accent.
Regarding subcategories of accent classes, Hattori demonstrated
that there are 2.3-5a and 2.3-5b subclasses in Amami and Okinawa
352 ACCENTUAL HISTORY

Ryukyuan. Regarding this matter I have pointed out that the same
subclasses are seen in Miyako and Yaeyama, and proposed that
these distinctions should go back to Proto-Ryukyuan and Proto-
Japonic. The reconstruction of the distinctions is only possible
when one includes all Ryukyuan dialects.
Concerning the question of the relationship between accent and
vowel length, Hattori and Martin suggested that there is a
correlation between a long vowel and initial pitch height. I have
presented more evidence to support this hypothesis, and based on
this I reconstructed word-initial low register in relation to the
initial-syllable long vowel. Without Ryukyuan data, this is not
possible. I must point out that I have also reconstructed initial-low
register which does not correlate with vowel length.
Certainly, the reconstruction of Proto-Japonic accent presented
in this book is not the final word in the research on Japonic accent.
Looking back into the accentual history of the Japonic languages
even earlier than Proto-Japonic (e.g., Koreo-Japonic), it raises more
questions: e.g., what kind of accent earlier Japonic had and how did
it evolve?
Vovin (2000) presents an intriguing argument on an earlier
accentual system of the Japonic languages in relation to Korean
accent. Pointing out a correlation between Korean vowel reduction
(corresponding to unstressed syllable in earlier Korean, which had
lexical stress) and the Japanese locus accent, Vovin suggests that
Proto-Koreo-Japonic was a stress language. Furthermore, he
assumes that pre-Proto-Japonic had stress and Proto-Japonic
developed register due to vowel length and voiced/voiceless
consonant distinctions. Looking into my reconstruction of Proto-
Japonic, in fact, this assumption seems to be quite possible. This is
at present an open question, and as this is beyond a scope of this
book, I leave it for future research.
Finally, needless to say, this book could not be possible without
taking advantage of the studies done earlier by the scholars
mentioned in this book. As their studies have done, I hope that this
book will contribute to the studies of the accentual history of the
Japonic languages.
 REFERENCES

Akinaga, Kazue. 1966. ‘Sanagi akusento no teigi suru mono [What

Sanagi accent defines]’. Kokubungaku Kenkyû 33, pp.74-88.
Ak ina ga, Kazu e, e d. ( unde r the sup erv isio n o f Kind aich i
Haruhiko). 1997. Meikai nihongo akusento jiten dai 2-han
[Concise dictionary of the Japanese language accent, second
edition]. Tokyo: Sanseidô.
Am ha, Azeb. 1996. ‘Tone-accent and prosodic domains in
Wolaitta’. Studies in African Linguistics 25-2, pp.111-138.
Anttila, Raimo. 1989. Historical and comparative linguistics:
S econd revised e dition . Am sterd am/Ph ilade lph ia : Joh n
Benjamins Publishing Company.
Blust, Robert. 1988. ‘Sketches of the morphology and phonology
of Bornean languages 2: Mukah (Melanau)’. Pacific linguistics
series A-78, pp.151-216.
Chao, Y. R. 1965. A grammar of spoken Chinese. Berkeley:
University of California Press.
Chao, Y. R., and Yang, L. S. 1962. Concise dictionary of spoken
Chinese. Cambridge: Harvard University Press.
Cheng, Chin-Chuan. 1973. A synchronic phonology of Mandarin
Chinese. Monographs on linguistic analysis 4. The Hague:
Mouton.
Halim, Amran. 1975. Intonation in relation to syntax in Bahasa
Indonesian. Jakarta: Proyek Pengembangan Bahasa dan Sastra
Indonesia dan Daerah, Departemen Pendidikan dan Kebudayaan.
Haraguchi, Shôsuke. 1977. The tone pattern of Japanese: An
autosegmental theory of tonology. Tokyo: Kaitakusha.
Has him oto, Oi-Kan Yu e. 1 97 2. Phon o log y o f C a nt onese .
Cambridge University Press.
Hattori, Shirô. 1931a. ‘Kokugo sho-hôgen no akusento gaikan [An
overview of the accent of the national-language dialects]’.
Hôgen 1-1, pp.11-32.
Hattori, Shirô. 1931b. ‘Kokugo sho-hôgen no akusento gaikan 2
[An overview of the accent of the national-language dialects 2]’.
Hôgen 1-3, pp.14-24.
354 REFERENCES

Hattori, Shirô. 1931c. ‘Kokugo sho-hôgen no akusento gaikan 3

[An overview of the accent of the national-language dialects 3]’.
Hôgen 1-4, pp.11-27.
Hattori, Shirô. 1932a. ‘Kokugo sho-hôgen no akusento gaikan 4
[An overview of the accent of the national-language dialects 4]’.
Hôgen 2-2, pp.1-12.
Hattori, Shirô. 1932b. ‘Kokugo sho-hôgen no akusento gaikan 5
[An overview of the accent of the national-language dialects 5]’.
Hôgen 2-4, pp.18-26.
Hattori, Shirô. 1933a. ‘Kokugo sho-hôgen no akusento gaikan 6
[An overview of the accent of the national-language dialects 6]’.
Hôgen 3-6, pp.5-18.
Hattori, Shirô. 1933b. ‘Akusento to hôgen [Accent and dialect]’.
Kokugo Kagaku Kôza VII: kokugo hôgengaku, pp.41-71. Tokyo:
Meiji Shoin.
Hattori, Shirô. 1937. ‘Genshi nihongo no ni-onsetsu meishi no
akusento [The accent of disyllabic nouns in proto-Japanese]’.
Hôgen 7-6, pp.44-58.
Hattori, Shirô. 1951. ‘Genshi nihongo no akusento [The accent of
proto-Japanese]’ in Terakawa et al. (eds.), Kokugo akusento
ro nsô (sy mp o siu m o n J ap a ne s e a cc e nt ) , Tok y o: Hô s ei
Daigaku Shuppankyoku, pp. 45-65.
Hattori, Shirô. 1959. Nihongo no keitô [The lineage of the Japanese
language]. Tokyo: Iwanami Shoten.
Hat to r i, Shirô. 1 967 . ‘ Ain ug o n o on ’in k ôzô to ak use n to
[Phonological structure and accent of Ainu]’. Onsei no kenkyû
13, pp.207-223.
Hattori, Sh irô. 197 9a. ‘Niho n sogo ni tsuite 21 [On Proto -
Japanese]’. Gekkan gengo 8.11, pp.97-107. Tokyo: Taishûkan.
Hattori, Shirô . 1979b . ‘Nihon sog o ni tsuite 2 2 [On Proto -
Japanese]’. Gekkan gengo 8.12, pp.100-114. Tokyo: Taishûkan.
Hayata, Teruhiro. 1973. ‘Accent in old Kyoto and some modern
Japanese dialects’. Gengo no kagaku 4, pp.139-180. Tokyo:
Tôkyô Gengo Kenkyûjo.
Hayes, Bruce. 1995. Metrical stress theory: Principles and Case
Studies. Chicago and London: University of Chicago Press.
 REFERENCES 355

Hirayama, Teruo, ed. 1935. Zenkoku akusento jiten [the dictionary

of the national language accent]. Tokyo: Tôkyôdô Shuppan.
Hirayama, Teruo. 1951. Kyûshû hôgen onchô no kenkyû: Kyôtsûgo
keihango tono hikaku kôsatsu [A study of Kyushu dialects: A
comparison with Standard Japanese and the dialects of Kyoto
and Osaka]. Tokyo: Gakkai no Shishinsha.
Hirayama, Teruo. 1960. Zenkoku akusento jiten [Dictionary of the
national language accent]. Tokyo: Tôkyôdô Shuppan.
Hirayama, Teruo, ed. 1982. Zenkoku hôgen jiten Vol. 2 – Jintai goi
no bubun taikei [Dictionary of Japanese dialects - names of body
parts]. Kadokawa Shoten.
Hirayama, Teruo. 1986. Amami hogen kisogoi no kenkyû [A study
of Amami Ryukyuan basic vocabulary]. Tokyo: Kadokawa
Shoten.
Hirayama, Teruo, ed. 1992. Gendai nihongo hôgen dai-jiten
[Dictionary of Modern Japanese dialects]. Tokyo: Meiji Shoin.
Hirayama, Teruo and Nakamoto Masachie. 1964. Ryûkyû Yonaguni
hôgen no sôgôteki kenkyû [A comprehensive study of Yonaguni
Ryukyuan dialect]. Tokyo: Meiji Shoin.
Hirayama, Teruo, Ôshima Ichirô and Nakamoto Masachie. 1966.
Ryûkyû hôgen no sôgôteki kenkyû [A comprehensive study of
Ryukyuan]. Tokyo: Meiji Shoin.
Hirayama, Teruo, Ôshima Ichirô and Nakamoto Masachie. 1967.
Ryûkyû sakishima hôgen no sôgôteki kenkyû [A comprehensive
study of Sakishima Ryukyuan]. Tokyo: Meiji Shoin.
Hyman, Larry M. 1975. Phonology: Theory and analysis. Harcourt
Brace Jovanovich College Publishers.
Hyman, Larry M. and Russell G. Schuh. 1974. ‘Universals of tone
rules: evidence from West Africa’. Linguistic Inquiry 5-1, pp.81-
115.
Inaba, Seiichirô. 1996. Initial binary feet and surface ternary feet
in metrical theory: An integrated computational analysis of
loanwords and nominal compounds in Japanese. University of
Hawai’i Ph.D. dissertation.
Jeong, Yeon-chan. 1985. ‘On the functional load of the tonemes in
Korean phonology’. Korea Journal 25-6, pp.5-13. Seoul.
356 REFERENCES

Kamei T., Ôfuji T., and Yamada T, eds. 1964. Nihongo no rekishi
[A history of the Japanese language], Vol. 5. Tokyo: Heibonsha.
Kenstowicz, Michael. 1994. Phonology in generative grammar.
Blackwell.
Kindaichi, Haruhiko. 1937. ‘Gendai sho-hôgen no hikaku kara
mita Heian-chô akusento: Toku ni ni-onsetsu meishi ni tsuite
[The accent of the Heian period in comparison with modern
dialects of Japan: focusing on disyllabic nouns]’. Hôgen 7-6,
pp.1-43.
Kindaichi, Haruhiko. 1947. ‘Gochô henka no hôsoku no tankyû
[An investigation on laws of lexical pitch change]’. Tôyôgo
kenkyû, pp.9-52.
Kindaichi, Haruhiko. 1951. ‘Kotoba no senritsu [Prosody in
language]’. Kokugogaku 5.
Kindaichi, Haruhiko. 1954. ‘Tôzai ryô-akusento no chigai ga
dekiru made [Until the differences between Eastern and Western
accent came into existence]’. Bungaku 22-8. [References are to
Kindaichi, Haruhiko, 1975, Nihon no hôgen: Akusento no
henkan to sono jissô, pp.49-80. Tokyo: Kyôiku Shuppan.]
Kindaichi, Haruhiko. 1960. ‘Kokugo no akusento no jidaiteki
henka [Diachronic change in accent of the national language]’.
Kokugo to kokubungaku 37-10, pp.29-46.
Kindaichi, Haruhiko. 1966. Meikai Nihongo akusento jiten (11-han
hakkô) [A concise dictionary of Japanese accent (11th edition)].
Tokyo: Sanseidô.
Kindaichi, Haruhiko. 1967. Hihongo on’in no kenkyû [A study of
Japanese phonology]. Tokyo: Tôkyôdô Shuppan.
Kindaichi, Haruhiko. 1973. ‘Hikaku hôgen-gaku to hôgen chiri-
gaku [Comparative dialectology and dialect geography]’.
Kokugo to kokubun-gaku 50-6, pp.1-19.
Kindaichi, Haruhiko. 1974. Kokugo akusento no shiteki kenkyû:
Genri to hôhô [Historical study on accent in the National
language: Principles and methods]. Tokyo: Hanawa Shobô.
Kindaichi, Haruhiko. 1975. Nihon no hôgen: Akusento no henkan
to sono jissô [Dialects of Japan: The way accent is and the way it
has changed]. Tokyo: Kyôiku Shuppan.
 REFERENCES 357

Kindaichi, Haruhiko. 1984. ‘Nihon sogo no akusento to Ryûkyû

hôgen [Proto-Japanese accent and the Ryukyuan dialects]’.
Sophia linguistica 17, pp.3-25.
Ko k uritsu Ko kug o Ken kyûjo, ed. 1 96 3. Oki na w ag o jiten
[Dictionary of Okinawan]. Tokyo: Ôkurashô Insatsukyoku.
KOS = Kagoshima-ken Ôshima Shichô Sômu-ka, ed. 1999. Heisei
10 nendo Amami guntô no gaikyô [General information on
Amami islands]. Kagoshima: Kagoshima-ken Ôshima Shichô
Sômu-ka.
Kyûgakkai Rengô Amami Ô shima Kyôdô chôsa Iinkai, ed. 1959.
Amami: Shizen to bunka [Amami: Nature and culture]. Tokyo:
Nihon Gakujutsu Shinkôkai.
Leben, W. R. 1971. ‘The morphophonemics of tone in Hausa’ in C.
W. Kim & H. Stahlke (Eds.), Papers in African linguistics.
Edmonton: Linguistic Research Inc., pp.201-218.
Leben, William R. 1978. ‘The representation of tone’. In Fromkin
(ed.). Tone: A linguistic survey. New York: Academic Press.
Lee, Sang-ak. 1985. ‘Perturbation and decay of tonal system in the
late Middle Korean’. Korea journal 25-6, pp.14-18. Seoul.
Maddieson, Ian. 1976. ‘Tone reversal in Ciluba: A new theory’ in
Larry Hyman (ed.), S tudies in Bantu Tonology (Southern
California Occasional Papers in Linguistics 3), pp.141-165.
Martin, Samuel E. 1966. ‘Lexical evidence relating Japanese to
Korean’. Language 42, pp.185-251.
Martin, Samuel E. 1975. ‘Problems in establishing the prehistorical
relationships of Korean and Japanese’. Proceedings of the
International Symposium Commemorating the 30th Anniversary
of Korean Liberation, pp.149-172. Seoul: National Academy of
Sciences.
Martin, Samuel E. 1987. The Japanese language through time.
New Haven and London: Yale University Press.
Martin, Samuel E. 1996. Consonant lenition in Korean and the
Macro-Altaic question. Honolulu: University of Hawai’i Press.
Matsumori, Akiko. 1998. ‘Ryûkyû akusento no rekishiteki keisei
katei: Ruibetsu goi 2-hakugo no tokui na gôryû no shikata wo
tegakari ni [Analysis of the formation process of Ryukyuan
358 REFERENCES

accent systems: Based on the distinctive patterns of merger for

2-mora classified words]’. Gengo kenkyû 114, pp.85-114.
Matsumori, Akiko. 1993. ‘Nihongo akusento no so-taikei saiken no
kokoromi: Iwayuru ‘kakôshiki akusento’ no seiritsu ni kansuru
kôsatsu wo moto ni shite [Towards a reconstruction of proto-
systems of Japanese tones: Based on the analysis o f the
downstep phenomenon in some Japanese dialects]’. Gengo
Kenkyû 103, pp.37-91.
M c C a r th y, J o h n J . a n d Al a n S . P r i n c e . 1 9 9 5 . ‘ P r o s o di c
morphology’ in John A. Goldsmith (ed.), The handbook of
phonological theory, pp.318-66. Cambridge: Basil Blackwell.
McCawley, James D. 1968. The phonological component of a
grammar of Japanese. The Hague: Mouton & Co.
McCawley, James D. 1977. ‘Accent in Japanese’ in Hyman (ed.),
S tudies in s tress an d accen t (S COP IL 4). Lo s An geles:
University of Southern California Department of Linguistics.
McCawley, James D. 1978. ‘What is a tone language?’ in Victoria
A. Fromkin (ed.). Tone: A linguistic survey, pp.113-131. New
York:Academic Press.
Miller, Roy Andrew. 1 97 1. Japanese a nd the o th er Alta ic
languages. Chicago: University of Chicago Press.
Mochizuki, Ikuko. 1974. Ruiju myôgi-shô yonshu shôten-tsuki
wakun shûsei [Collection of Myogi-sho text glosses marked with
symbols for tones]. Tokyo: Kasama Shoin.
Murayama, Shichirô. 1985. ‘Nihongo katsuyô kigen ni tsuite no
oboegaki [Notes on the origin of verbal inflection in Japanese]’.
In Shibata, et al. (eds.). Nihongo no keitô: Kihon ronbunshû 1.
Osaka: Izumi Shoin.
Nakasone, Seizen. 1983. Okinawa Nakijin-hôgen jiten [Dictionary
of Nakijin Okinawan dialect]. Tokyo: Kadok awa Shoten.
Nihon Hôsô Bunka Kenkyûjo, (ed.). 1971. Nihongo hatsuon
akusento jiten (dai 9-satsu hakkô) [Dictionary of accentual
pronunciation in the Japanese language (9th printing)]. Tokyo:
Nihon Hôsô Shuppan Kyôkai.
 REFERENCES 359

Numamoto, Katsuaki. 1979. ‘Heian-jidai ni okeru nichijô kango no

akusento [Accent of common Chinese loanwords in the Heian
period]’. Kokugo kokubun 48-6.
Od de n , D a v i d . 1 9 8 5. An a c c e nt u a l a pp r oa c h t o t o n e i n
Kimatuumbi, African Linguistics, Essays in memory of M.W.K.
Semikenke. Didier L. Goyvaerts (ed.). John Benjamins
Publishing Company.
OGJ. 1963. Okinawago jiten [see Kokuritsu Kokugo Kenkyûjo,
ed.]
Ôhara, Takamichi. 1942. ‘Ruiju-myôgi-shô no akusento to sho-
hôgen akusento to no taiô kankei: shu to shite san-onsetsu
meishi ni tsuite [Correspondences between accent in Ruiju
myôgi-shô and accent in dialects of Japanese: focusing on
trisyllabic nouns]’. In Nihon Hôgen Gakkai (ed.). Nihongo no
akusento. Tokyo: Chûô Kôron-sha.
OKCS = Okinawa Kikaku-kaihatsu-bu Chiiki-ritô-shinkô-kyoku
Shichôson-ka, ed. 1998. Jûmin kihon daichô jinkô no gaikyô
[General information on the population of the inhabitants].
Okuda, Kunio. 1975. Accentual systems in the Japanese dialects: A
generative approach. Tokyo: Bunka Hyôron Publishing.
Onseigaku dai-jiten [Dictionary of phonetics]. 1976. Tokyo:
Sanshûsha.
OPOCVB = Okinawa Prefecture and Okinawa Convention &
Visitors Bureau. 1996. Okinawa magazine. Okinawa.
Poser, William. 1990. The phonetics and phonology of tone and
intonation in Japanese. MIT Unpublished Ph.D. dissertation.
Ramsey, Samuel Robert. 1978. Accent and morphology in Korean
dialects: A descriptive and historical study. Seoul: The Society
of Korean Linguistics.
Ramsey, S. Robert. 1979. ‘The Old Kyoto dialect and the historical
development of Japanese accent’. Harvard journal of Asiatic
studies 39-1, pp.157-175.
Sakurai, Shigeharu. 1978. Nihongo no senritsu [Prosody in
Japanese]. Tokyo: Sôgunsha Shuppan.
360 REFERENCES

Schadeberg, Thilo C. 1977. ‘Anticipation of tone: Evidence from

Bantu’. In Kotey, Paul F.A., and Haig Der Houssikian (eds.).
Language and linguistic problems in Africa, Proceedings of the
VII Conference on African Linguistics. South Carolin a:
Hornbeam Press.
Schuh, Russell G. 1978. ‘Tone Rules’. In Victoria A. Fromkin (ed.),
Tone: A linguistic survey, pp.221-256. Academic Press.
Serafim, Leon A. 1993. ‘Initial obstruent voicing in Late Proto-
Japonic’. Unpublished ms.
Serafim, Leon A. 1994. ‘Vowel length in Japanese and Ryukyuan’.
Unpublished ms.
Shibatani, Masayoshi. 1990. The languages of Japan. New York:
Cambridge University Press.
Shimabukuro, Moriyo. 1997. ‘Word-initial low register in Proto-
Japanese’ in Sohn and Haig (eds.), Japanese Korean linguistics
6, pp.135-141. Standford: CSLI Publications.
Sh imabukuro, Moriyo. 1 99 8a. ‘A recon structio n o f Pro to -
Ryukyuan accent’. In David J. Silva (ed.), Japanese Korean
linguistics 8 , pp.361-374. California: CSLI Publications.
Shimabukuro, Moriyo. 1998b. ‘Intrinsic pitch of long vowels and
short vowels in Japanese’. Presented at the Seventh Annual East-
West Center Participants’ Conference, in Honolulu.
Shimabukuro, Moriyo. 1999. ‘Ainugo Kankokugo Nihongo no
akusento ni kansuru ichi kôsatsu [Thoughts on accent of Ainu,
Korean, and Japanese]’. Presented at the 14th Annual Okinawa
Gaikoku Bungakkai Conference, in Okinawa.
Sneddon, J.N. 1984. ‘Proto-Sangiric and the Sangiric languages’.
Pacific linguistics B-91.
Thomason, Sarah G. and Terrence Kaufman. 1988. Language
contact, creolization, and genetic linguistics. California:
University of California Press.
Thorpe, Maner L. 1983. Ryukyuan language history. University of
Southern California Ph.D. dissertation.
Tokugawa, Munemasa. 1958. ‘Yoron-tô to Okinoerabu-tô no hôgen
ni tsuite [On Yoron and Okinoerabu dialects]’. Jinrui Kagaku 10.
Kyûgakkai Rengô.
 REFERENCES 361

Tokugawa, Munemasa. 1962. ‘Nihon shohôgen akusento no keitô

shiron: “Rui no tôgô” to “chiriteki bunpu” kara miru [Towards a
family tree for accent in Japanese dialects: from the point of
view of merger of categories and geographical distribution]’.
Gakushûin Daigaku Kokugo Kokubun Gakkaishi 6. [References
are to Shibata, Takeshi, et al. (eds.). 1978. Niho no gengogaku 6:
hôgen, pp.543-570.]
Uemura, Kouji. 1959. ‘Amami hôgen no akusento – Okinoerabu-
shima ü Tokuno-Ôshima no bu [Accent of Amami dialects:
Okinoerabu island and Tokuno- shima]’. Cultural Science
Reports of Kagoshima University 8, p.17.
Uemura, Yukio. 1972. ‘Ryûkyû no hôgen [The dialects of the
Ryukyus]’ in Hokama Shuzen (ed.), Okinawa bunka ronsô 5,
pp.80-102. Tokyo: Heibonsha.
Umegaki, Minoru. 1957. ‘Ôsaka hôgen akusento henka no keikô
[Tendency of accentual change in Osaka dialect]’. Kinki hôgen
sôsho 6, pp.39-52.
Uwano, Zendô. 1977. ‘Nihongo no akusento [Japanese accent]’.
Iwanami Koza Nihongo 5: On’in, pp.281-321.
Uwano, Zendô. 1985. ‘Kagawa-ken Ibuki-jima hôgen no akusento
[Accent of Ibuki-jima in Kagawa Prefecture]’. Nihon Gakushiin
Kiyô 40-2, pp.75-179.
Vance, Timothy. 1987. An introduction to Japanese phonology.
New York: State University of New York Press.
Vovin, Alexander. 1993a. ‘Long vowels in Proto-Japanese’. Journal
of East Asian linguistics 2, pp.125-134.
Vovin, Alexander. 1993b. A recon stru ction o f pro to -A inu.
Leiden•New York• Köln: E.J. Brill.
Vovin, Alexander. 1997. ‘The origin of register in Japanese and the
Altaic theory’ in Sohn and Haig (eds.), Japanese Korean
linguistics 6. Stanford: CSLI Publications.
Vovin, Alexander. 2000. ‘On the great vowel shift in middle Korean
and position of stress in proto-Korean’. Korean Linguistics 10,
pp.61-78.
362 REFERENCES

Wada, Minoru. 1958. ‘Fukuzatsu na akusento taikei no kaishaku

[Interpretation of complex accent systems]’. Kokugo gaku 32,
pp.72-90.
Wada, Minoru. 1966a. ‘Nihaku-meishi itsutsu no rui o itsutsu ni ii-
wak eru akusento [Disyllabic nou ns with fiv e accentua l
distinctions]’, Gengo kenkyû 50, pp.128-30.
Wada, Minoru. 1966b. ‘Dai ichi-ji akusento no hakken: Ibuki-jima
[Discovery of the first-stage accent system: Ibuki-jima]’.
Kokubo kenkyû 22, pp.24-8.
We n c k , Gü n t h er. 1 9 5 9 . J a p a n i s c h e P h o n e t i k : B a n d I V.
Erscheinungen und Probleme des Japanischen Lautwandels.
Otto Harrassowitz, Wiesbaden.
Wh itm a n, Jo h n B . 1 98 5. T h e pho no lo gi ca l b asis f o r t h e
comparison of Japanese and Korean. Harvard University Ph.D.
dissertation.
Whitman, John B. 1994. ‘The accentuation of nominal stems in
proto-Korean’ in Kim-Renaud (ed.), Theoretical issues in
Korean linguistics, pp.425-39.
Yamamoto, Takeshi. 1970. Kôchi-ken no rekishi [History of Kôchi
Prefecture]. Tokyo: Yamakawa Shuppansha.
Yanagita, Kunio. 1927. ‘Kagyûkô [A study of words meaning
‘snail’]’. Jinruigaku zasshi 42-4-7. [References are to Kagyûkô.
1930. Tokyo: Tôkô Shoin.]
 APPENDIX

JAPONIC COGNATES AND

RECONSTRUCTED FORMS

The following are in order: Gloss, Myôgishô accent class in

parentheses, Proto-Japonic (PJ) form, Proto-Mainland Japanese
(PMJ) [Tokyo, Kyoto, Kagoshima, Myôgishô]: Western Old
Japanese (WOJ), Eastern Old Japanese (EOJ), Middle Japanese
(MJ), Proto-Ryukyuan (PR), Proto-Amami (PA) [Naze, Kamishiro,
Kametsu], Proto-Okinawa (PO) [Nakijin, Shuri, Aguni]: Old
Ryukyuan (OR), Proto-Miyako (PM) [Ikema, Ôura, Uechi], Proto-
Yaeyama (PY) [Ishigaki, Sonai, Kuroshima], Yonaguni (Y).
Western Old Japanese, Eastern Old Japanese, Middle Japanese,
and Old Ryukyuan forms are extracted from the following sources:

WOJ: Omodaka, et al., eds. 1967 (Jidaibetsu kokugo daijiten:

jôdai-hen.)
EOJ: Mizushima 1984. (Man’yôshû azumauta no kokugogaku
teki kenkyû)
MJ: Saeki and Mabuchi, eds. 1969. (Kogo-jiten Kôdansha
kogo jiten.)
OR: Okinawa kogo daijiten henshû iinkai, ed. 1995.
(Okinawa kogo daijiten.)
364 APPENDIX

adult (3.7) PMJ *otoÈnaˆ [o˘tona, oto˘na, `otona, _o to˘n a]:

MJ otona.
bag (3.4) PJ *puÈkurwo, PMJ *puÈkurwo [hukuro˘, hu˘kuro,
_huk uro, _fukuro]: WO J p ukurwo, MJ huk uro,
PR *_p h ukuru˘, PA *hukuruˆ [huku˘ro, huku˘ru, hu˘Qku],
PO *p h uku˘ru [_p h uku :ru, _h ukuru, --], PM *_fukuru ˘
[fukuru˘, fukuru˘, fukuru˘], PY *_hukuruˆ [_hukuru, hukuÈru,
huku˘rï], Y _ku:ru.
beard (2.1) PJ *`pyinkay, PMJ *`pyinkey [hige, `hige, `hige,
`fige]: WOJ pige, MJ hige, PR *`pigi˘, PA *higi [higi, -- ,
sïgi], PO *p h i˘zi [`pizi:, `hwizi, --], PM *`pïgi˘ [higi˘,
pïgi˘, pïgi], PY *`pïni˘ [`pïni, pini, pini], Y `Ngi.
behind (3.7) PMJ *uÈsir´ˆ [siro , usi˘ro, _usiro, ---]: WOJ usiro,
MJ usiro.
bird (2.1) PJ *`t´ri, PMJ *`t´ri [tori, `tori, `toi, `tori]: WOJ tori,
EOJ tori, MJ tori, PR *`t h uri˘, PA *t h uri [t h uri, `t h ui, t h uI],
PO *t h u˘i [`t h ui, `tui, --]: OR tori, PM *`tuï˘ [tui˘, tuï˘, tuï],
PY *`turï˘ [`turï, --, turï].
blood (1.1) PJ *`tu:y, PMJ *`ti:y [ti, `ti:, `ti, `ti]: WOJ ti,
MJ ti, PR *ci:, PA *ci: [ci, `ci:, ci:], PO *ci˘: [`ci:, `ci:, --],
PY *`cï:˘ [`cï:, ci:, si:], Y `ci:
board (2.4) PJ *iÈta, PMJ *iÈta [i˘ta, _ita, _ita, iÈta]: WOJ ita,
MJ ita, PR *_/ita˘, PA */ita˘ [/ita, _/ita, --], PO */ica
[_hica:, _/ica, --]: OR ita ~ ica, PM *_icya˘ [icya˘, icya˘,
icya˘], PY *_ita˘ [_ita, iÈta, ica], Y _ita.
boat (2.4) PJ *pu:Ènay, PMJ *puÈney [hu˘ne, _hune, _hune,
fuÈne]: WOJ pune, EOJ pune, MJ hune, PR *phu:niˆ, PA
*hu:˘nï [hu˘nï, hi:ni˘, hu˘nï], PO *phu:˘ni [phu˘ni(:), _huni,
huni˘]: OR hune ~ huni, PM *funi [funi, fu˘ni, funi˘], PY
*funi [_huni, huÈni, fu˘ni], Y nniˆ.
 APPENDIX 365

bone (2.3) PJ *_po:ne, PMJ *_pone [hone˘, ho˘ne, _hone, _fone]:

WOJ pone, MJ hone, PR *p h u:nïˆ, PA *hu˘nï [hu˘nï, -- ,
hu˘nI], PO *p h u:˘ni [p h u˘ni(:), _huni, --], PM *puni [puni,
pu˘ni, puni˘], PY *puni [_puni, puÈni, pu˘ni], Y huniˆ.
breath (2.4) PJ *i:Èkyi, PMJ *iÈkyi [i˘ki, _iki, _iQ, iÈki]: WOJ ikyi,
EOJ iki, MJ iki, PR */i:kiˆ, PA */i˘ki [/i˘ki, `/iki, /I˘k h I],
PO */i:˘ci [/i˘ci(:), _/i:ci, --]: OR iki, PM *ikï [icï, i˘kï, ikï˘],
PY *ikï [_ikï, i˘ki, i˘ki], Y itiˆ.
bridegroom (2.5) PJ *mwo:kwo˘, PMJ *mwokwo˘ [mu˘ko, muko˘,
_muko, muko˘]: WOJ mwokwo, MJ muko, PR *mu:khuˆ, PA
*mu˘kwa [mo˘ho, muQkwa˘, (muQkwa)], PO *mu:ku˘
[mu˘hu(:), _mu:ku, muku˘]: OR moko, PM *muku [muku,
mu˘ku, muku˘], PY *muku [_muku, muÈku, mu˘ku], Y
muguˆ.
bridge (2.2) PJ *pa˘si, PMJ *pa˘si [hasi˘, ha˘si, `hasi, fa˘si]:
WOJ pasi, MJ hasi, PR *`p h asi˘, PA *hasi [hasi, `hasi, hasi],
PO *p h a˘si [`p h asi:, `hasi, --]: OR hasi, PM *`pasï˘ [hasï˘,
pasï˘, pasï], PY *`pasï˘ [`pasï, pasi, pasi], Y `haci.
bucket (2.5) PJ *wo:kay˘, PMJ *wokey˘ [o˘ke, oke˘, woke˘]:
WOJ wokey, EOJ woke, MJ woke, PR *u:k h ïˆ, PA *wu˘k h ï
[u ˘hï, wui˘, / ukï˘], P O *u :k h i˘ [h u˘k h i(: ), _ u :k i, --] ,
PM *wu:ki [wu:ki, u:˘ki, u:ki˘], PY *u:ki [_u:ki, uÈgi, u:˘ki],
Y ugiˆ.
calendar (3.4) PJ *_k´y´myi, PMJ *_k´y´myi [koyomi˘, `koyomi,
-- , _koyomi]: WOJ koyomyi, MJ koyomi, PR *_khuyumi, PA
*kuyumi˘ [kuyumi, -- , --], PO *khuyumi [_khuyumi:,
_kuyumi, kuyumi], PM *_kuyum˘ [kuyun˘, kuyun˘, kuyum˘],
Y kuyumi.
366 APPENDIX

centipede (3.2) PJ *mukanteˆ, PMJ *mukadeˆ [mukade, muka˘de,

`mu kade, `mu ka˘d e]: W OJ mu kade, MJ mu kade,
PR * `mukadi, P A *mukaziˆ [muka˘zi, - - , muka˘re] ,
PO *_mukazi [_muka:zi, _nkazi, nkazi], PM *`mmkazi˘
[nkadi˘, nkazi˘, mmkazi], PY *_mukazaˆ [_ngaza, mo:Èza,
nka˘za], Y Nkadiˆ.
cherry (3.2) PJ *`sak ura˘, PMJ *sak ura˘ [sakura, `sak ura,
`sakura, `saku˘ra]: WOJ sakura, MJ sakura, PR *`sakura,
PA *sakura [sakura, -- , sakura], PO *sa˘kura [saku:˘ra,
`sakura, --]: OR sakura.
chopsticks (2.4) PJ *pa:Èsuy, PMJ *paÈsiy [ha˘si, _hasi, _hasi,
faÈsi]: WOJ pasyi, MJ hasi, PR *pha:siˆ, PA *ha˘si [ha˘si,
hasi˘, --], PO *pha:˘si [--, _ha(:)si, ha:˘si], PY *pasï [_pasï,
pa˘si, pa˘si], Y haciˆ.
cloud (2.3) PJ *_kumwo, PMJ *_kumwo [ku˘mo, ku˘mo, _kumo,
_kumo]: WOJ kumwo, EOJ kumo, MJ kumo, PR *_k h umu˘,
PA *kh umu˘ [kumu, -- , kumo˘], PO *kumu [_kumu:, _kumu,
--]: OR kumo , PM * _fumu ˘ [h mm u˘, fumu ˘, f umu ˘] ,
PY *_fumu˘ [_humu, huÈmu, fumu], Y _mmu.
cousin (3.5) PJ *_itwoÊkwo, PMJ *itwoÊkwo [ito˘ko, i˘toko,
_itoko, itoÈko]: WOJ itwokwo, MJ itoko, PR *_/itoko˘,
PA */itoko˘ [/itoko, -- , --], PO */icuku: [_hici(:)ku:,
_/icuku, --], PM *icufu [icyufu, icï˘fu, icufu˘], PY *_icikuˆ
[_icihu, icihu, isiku LHH], Y itigu˘.
cow (2.1) PJ *`usi, PMJ *`usi [usi, `usi, `usi, `usi]: WOJ usi,
MJ usi, PR *`/usi˘, PA */usi [/usi,`/usi, /usi], PO */u˘si
[`husi:, `/usi, --]: OR usi, PM *`u sï˘ [usï˘, usï˘, usï],
PY *`usï˘ [`usï, usi, usi], Y `uci.
curtain (3.2) PMJ *`tobari˘ [tobari, `tobari, `tobari,`toba˘ri]:
WOJ tobari, MJ tobari.
 APPENDIX 367

dog (2.3) PJ *_enu, PMJ *_inu [inu˘, i˘nu, _in, _inu]: WOJ inu,
MJ inu, PR *_/inu˘, PA */in˘ [/in, -- , /in˘], PO */in [-- ,
_/in, --], PM *_in˘ [in˘, in˘, in˘], PY *_inu˘ [_in, iÈnu, in],
Y _inu.
earthworm (3.6) P J * m i m i˘ n su , P M J *m i È mi zu ˘ [ m i m iz u ,
_mimizu, _mimiQ, miÈmizu]: MJ mimizu, PR *mimi˘zi,
PA *më(C)ë˘za [më:za˘, -- , mï˘da], PO *mimi˘zi [_miminza,
_ mi miz i, --] , PM *mi :mi zï [- -, mi: ma˘ zï, m i:m izï ˘] ,
PY*mi:mi˘zï [_mi:mizï, mimin˘ci, mi:ma˘zi], Y --.
eel (3.6) PJ *una˘nki, PMJ *uÈnagi˘ [unagi, _unagi, _unaQ,
uÈnagi], PR */una˘gi, PA */una˘gi [unagi˘, -- ,/una˘gi],
PO */una˘zi [_/una:zi, _/nnazi, --], PM *unagï [unazï,
una˘gï, unagï˘], PY *_u:nagiˆ [_u:nai, o:Èni, una˘gi].
eye (1.3) PJ * _may:, PMJ *_me:y [me˘, _me:, _me, _me]:
WOJ mye, EOJ me, MJ me, PR *_mï:˘, PA *mï:˘ [mï, _mi:,
mï:˘], PO *mi: [_mi:, _mi:, --]: OR me, PM *_mi:˘ [mi:˘,
mi:˘, mi:˘], PY *_mi:˘ [_mi:, --, mi:], Y _mi:
fan (3.4) PJ *aÈpukyi, PMJ *aÈpukyi [o:gi˘, o:˘gi, _o:gi, _afuki]:
WOJ apukyi, MJ ahugi, PR *_augi˘, PA */o:˘gi [/o˘:gi ~
/u˘gi, /o:gi˘, o˘gi], PO */o:zi [_/o:zi, _/o:zi, --], PM *_augï˘
[auzï, o:˘gï, wa:gï˘], PY *ongï [_ongï, onÈgi, on˘gi].
fat (3.5) PJ *_ampura˘, PMJ *abuÈra˘ [abura, a˘bura, _abura,
abuÈra]: WOJ abura, MJ abura, PR *_/abura˘, PA */abura˘
[/abura, -- , /amba˘], P O */anda [_/anda: _/anda, --],
PM *_aQva˘ [aQva˘, aQva˘, aQva˘], PY *_aburaˆ [_aba,
abuÈra, aQva], Y _anda.
field (3.7) PJ * paÈtak ayˆ, PMJ *paÈtak eˆ [hatake, h ata˘ke,
_hatake, _fata˘ke]: WOJ patakye, MJ hatake, PR *hatakiˆ,
PO *phata˘ki [--, _hataki, --], PM *patagi [--, pata˘gi, --], PY
*pata˘ki [_patagi, --, pata˘ki], Y hatagiˆ.
figure (3.5) PMJ *sugaÈta [su˘gata, su˘gata, _sugata, sugaÈta]:
WOJ sugata, MJ sugata, PA *sïgataˆ [sigata, --, siga˘ta].
368 APPENDIX

firefly (3.5) PMJ *potaÈru [ho˘taru, ho˘taru, _hotaru , ---]:

WOJ potaru, MJ hotaru.
flower (2.3) PJ *_pana, PMJ *_pana [hana˘, ha˘na, _hana, _fana]:
WOJ pana, EOJ pana, MJ hana, PR *_p h ana˘, PA *hana˘
[hana, _hana, hana˘], PO *p h ana [_p h ana:, _hana, hana]:
OR hana, PM *_pana˘ [hana˘, pana˘, pana˘], PY *_pana˘
[_pana, paÈna, pana], Y _hana.
forehead (3.1) PJ *`pyitapyi, PMJ *`pyitapyi [hitai, `hitai,
`hitai, `fitafi]: WOJ pyitapy i, MJ hitahi, PR *`p h itai,
PO *p h i˘ c e: [` p h i ce: , ` h wice :, h ic e˘ : ], PM *` f ut a i˘
[futai˘, futai˘, futai], PY *futai [`hutai, hute, futai], Y `tai.
forest (3.4 ) PJ * _payasi, PMJ *_payasi [hayasi˘, `h ay asi,
_hayasi, _fayasi]: WOJ payasyi, EOJ payasi, MJ hayasi.
gold (3.3) PJ *k wo ˘n ka:Ènay, P MJ *kwo˘gaÈn ey [ko˘g an e,
`kogane, `kogane, ko˘gane], PR *`k h ugani˘, PA *kh ugani
[ k h u g a ni , - - , - - ] , P O * k h u ˘ g a n i [ - - , ` k u g a n i , - - ] :
PM *`kugani˘ [kugani˘, kugani˘, kugani], PY *`kuganiˆ
[`kugani, kuga˘ni, kugani], Y `khuNani.
hair (1.1) PJ *`ka:y, PMJ *`ke:y [ke, ke:˘, `ke, `ke]: WOJ kye,
MJ ke, PR *khï:, PA *kh ï: [kh ï, `hi:, kï:], PO *k hi˘: [`kh i:, `ki:,
ki˘:], PM *_ki:˘ [ki:˘, --, --], PY *`ki:˘ [`ki:, ki:, ki:], Y `kh i:
heart (3 .5) PJ *k´k´Êr´, PMJ *k´k´Êr´ [ko ko˘ro, ko˘koro,
_kokoro, kokoÈro]: WOJ kokoro, EOJ kokoro, MJ kokoro,
PR *_k h ukuru˘, PO *k h uku˘ru [_k h uku:ru, _kukuru, --]:
OR kokoro ~ kukuru, PY *_kukuruˆ [_kukuru, kukuÈru,
kuku˘ru], Y _kuguru.
helmet (3.7) PJ *kh ampuÈtoˆ, PMJ *kabuÈtoˆ [ka˘buto, kabu˘to, ]:
WOJ kabuto, MJ kabuto, PR *khabutuˆ.
interval (3.2) PJ *`apyinta˘, PMJ *`apyida˘ [aida, `aida, `aida,
`afi˘da]: WOJ apyida, EOJ apida, MJ ahida.
leaf (1.2) PJ *pa˘:, PMJ *pa˘: [ha, ha:˘, `ha, fa˘]: WOJ pa,
EOJ pa, MJ ha, PR *`ph a:˘, PA *ha: [ha, `ha:, ha:], PO *pha˘:
[`pha:, `hwa:, ha˘:], PM *`pa:˘ [ha:˘, pa:˘, pa:], PY *`pa:˘
[`pa:, pa:, pa:], Y `ha:
 APPENDIX 369

life (3.5) PJ *in´Èti, PMJ *in´Èti [i˘noti, i˘noti, _inoQ, inoÈti]:

WOJ inoti, MJ inoti, PR *_/inuti˘, PA */inuci˘ [/inuci, -- ,
/inucï˘ ~ /inoci˘], P O * /inu˘ci: [_ /inu ci:, _ nuci, --]:
OR inoci ~ nuci ~ wonoci ~ wenoci, PM *nnucï [nnucï,
nnu˘cï, mnucï˘], PY *inu˘cï [_nucï, inu˘ci, nu˘ci], Y nuti˘.
lizard (3.2) PJ *tok an keˆ, PMJ *tokageˆ [tokage, tok a˘ge,
`kokage, `toka˘ge]: MJ tokage.
loins (2.1) PJ *`k´si, PMJ *`k´si [kosi, `kosi, `kosi, `kosi]:
WOJ kosi, MJ kosi, PR *`khusi˘, PA *khusi [khusi, -- , khusI],
PO *khu˘si [`husi:, `kusi, --], PM *`kusï˘ [kusï˘, kusï˘, kusï],
PY *`kusï˘ [`kusï, kusi, kusi], Y `khuci.
mark (3.1) PJ *`syir´syi, PMJ *`syir´syi [sirusi, `sirusi, `sirusi,
`sirusi]: WOJ syirusyi, MJ sirusi, PR *`sirusi, PA *sirusi
[sirusi, `sirusi, sirusi], PO *si˘rusi [siru:˘si, `sirusi, si˘rusi],
PM *`sïrusï˘ [sïrusï˘, sïrusï˘, sïrusi], PY *sirusï [`sirusï,
sirusi, sirusi], Y `ciruci.
medicine (3.7) PJ *kuÈsuriˆ, PMJ *kuÈsuriˆ [kusuri, kusu˘ri,
_kusui, _kusu˘ri]: WOJ kusuri, MJ kusuri, PR *k h usuriˆ,
PA *khusuriˆ [khusu˘ri, -- , kusu˘i], PO *khusu˘i [_khusui,
_kusui, --], PM *fusuï [Qfui, Qsu˘ï, fusuï˘], PY *_fusurïˆ
[_husirï, husiÈri, fusi˘rï], Y cu:riˆ.
mirror (3.4) PJ *kaÈnkamyi, PMJ *kaÈgamyi [kagami˘, ka˘gmi,
_kagan, _kagami]: WOJ kagamyi, EOJ kagami, MJ kagami,
PR *_khagami˘, PA *khagami˘ [kagan, _hagani, kagami˘], PO
*khaga˘mi: [_khaga:mi:, _kagan, --], PM *`kagam˘ [kagan˘,
kagan˘, kagam], PY*kaga˘nï [_kangan, kaga˘n, hanga˘rï], Y
_kaNan.
mortar (2.4) PJ *u:Èsu, PMJ *uÈsu [u˘su, _usu, _usu, uÈsu]:
WOJ usu, MJ usu, PR */u:siˆ, PA */u˘si [/u˘si, /usi˘, /u˘sï],
PO */u:˘si [/u˘si(:), _/u:si, /u:˘si], PM *usï [usï, u˘sï, usï˘],
PY *usï [_usï, u˘si, u˘si], Y uciˆ.
370 APPENDIX

mountain (2.3) PJ *_yama, PMJ *_yama [yama˘, ya˘ma, _yama,

_yama]: WOJ yama, EOJ yama, MJ yama, PR *_yama˘,
PA *yama˘ [yama, _yama, yama˘], PO *yama [_yama:,
_yama, yama]: OR yama ~ ya:ma, PM *_yama˘ [yama˘,
yama˘, yama˘], PY *_yama˘ [_yama, (yamana), yama],
Y _dama.
mud (2.3) PJ *_ntoro, PMJ *_doro [doro˘, do˘ro, _doro, _doro].
name (1.2) PJ *na˘:, PMJ *na˘: [na, na:˘, `na, na˘]: WOJ na,
EOJ na, MJ na, PR *`na:˘, PA *na: [na, `na:, na:], PO *na˘:
[`na:, `na:, --]: OR na ~ mya, PM *`na:˘ [na:˘, na:˘, na:],
PY *`na:˘ [`na:, na:, na:], Y `na:
needle (2.4) PJ *pa:Èruy, PMJ *paÈriy [ha˘ri, _hari, _hai, faÈri]:
WOJ pari, MJ hari, PR *p h a:riˆ, PA *ha:˘ri [ha˘ri, _ha:ri,
ha˘i], PO *p h a:i˘ [p h a˘i, _ha:i, ha:˘i], PM *paï [pai, pi˘ï,
p iï˘], PY * pa:rï [_ parï, pa˘ri ~ p a˘: ri, p a˘rï], Y h aiˆ .
nose (2.1) PJ *`pana, PMJ *`pana [hana, `hana, `hana, `fana]:
WOJ pana, MJ hana, PR *`p h ana˘, PA *hana [hana, `hana,
hana], PO *p h a˘na [`p h ana:, `hana, ha˘na], PM *`pana˘
[hana˘, pa:˘, pa:], PY *`pana˘ [`pana, pana, pana].
one person (3.7) PMJ *pyit´Èriˆ [hi˘tori, hito˘ri, _hitori, _fito˘ri]:
WOJ pyitori, EOJ hitori, MJ hitori. OR hitori.
paper (2.2) PJ *ka˘mpyi, PMJ *ka˘byi [kami˘, ka˘mi, `kan,
ka˘mi]: WOJ kamyi, MJ kami, PR *`khabi˘, PA *kh abi [k habi,
-- , kabi], PO *k h a˘bi [`habi:, `kabi, ha˘bi], PM *`kabï:˘
[kabi:˘, kabï˘, kabï], PY *`kabï˘ [`kabï, kabi, habi], Y `k habi.
pillar (3.5) PJ *_pasira˘, PMJ *fasiÈra˘ [hasira, ha˘sira, _hasira,
fasiÈra]: WOJ pasira, MJ hasira, PR *_phasira˘, PA *hasiraˆ
[hasi˘ra ~ ha˘rya ~ ha˘ra, -- , ha˘ra], PO *phasi˘ra:
[_phasira:,_ha:ya, ha:˘ya], PM *para [hara, para˘, para˘], PY
*para: [_para:, paÈra, pa˘ra], Y hira˘.
 APPENDIX 371

pillow (3.5) PJ *makuÈra, PMJ *makuÈra [ma˘kura, ma˘kura,

_makura, makuÈra]: WOJ makura, EOJ makura, MJ makura,
PR *_makura˘, PA *makura˘ [makura, -- , m a Q k w a ˘ ] ,
PO *maQkwa [_maQka:, _maQkwa, --], PM *_maQfa˘
[maQfa˘, maQ˘fa, maQfa˘], PY *_mafuraˆ [_mahwa, mo:Èra,
maQ˘fa], Y _magura.
rain (2.5) PJ *amay˘, PMJ *amey˘ [a˘me, ame˘, _ame, ame˘]:
WOJ amye, EOJ ame, MJ ame, PR *_/amï˘, PA */amï˘
[/amï, _/ami, /amˆ˘], PO */ami [_/ami:, _/ami, /ame]:
OR ame ~ ami, PM *_ami˘ [ami˘, ami˘, ami˘], PY*_a:mi˘
[_a:mi, aÈmi, ami], Y _ami.
sea (2.4) PJ *u:Èmyi, PMJ *uÈmyi [u˘mi, _umi, _un, uÈmi]:
WOJ umyi, EOJ umi, MJ umi, PR */u:miˆ, PA */u˘mi
[/u˘mi, `/uni, /un˘], PO */u:mi˘ [/u˘mi(:), _/umi, /umi˘]:
OR umi ~ mi:, PM *_in˘ [in, i˘m, im˘].
shadow (2.5) PJ *ka:nkay˘, PMJ *kagey˘ [ka˘ge, kage˘, kage˘]:
WOJ kagye, MJ kage, PR *k h a:gïˆ, PA *k h a˘gï [k h a˘gë,
haga˘, ka˘gï], P O *kh a:gi˘ [ ha˘gi(:), _ka:gi, --] ,
PM *kagi [kagi, ka˘gi, kagi˘], PY *kai [_kai, ka˘i, ha˘i],
Y khaNiˆ.
shoulder (2.4) PJ *kaÈta, PMJ *kaÈta [ka˘ta, _kata, _kata, kaÈta]:
WOJ kata, EOJ kata, MJ kata, PR *_khata˘, PA *khata ˘ [khata,
_kata, khata˘], PO *khata [_hata:, _kata, kata], PY *_kata˘
[_kata, kaÈta, hata], Y _kata.
sickness (3.7) PJ *yamaÈpyiˆ, PMJ *yamaÈpyiˆ [ya˘mai, yama˘i,
`yamai, _yama˘fi]: WOJ yampyi, MJ yamhi.
silkworm (3.7) PJ *kapyiÈkwoˆ, PMJ *kapyiÈkwoˆ [ka˘iko,
kai˘ko, `kaiko, _kafi˘ko]: WOJ kapyik wo, MJ kahiko ,
PR *kaiguˆ.
372 APPENDIX

smok e ( 3. 1 ) P J *` k aympu r yi , PM J * `k ey bu ryi [ k emu ri,

`kemuri, `kemui, `kemuri]: WOJ kyeburyi, MJ keburi,
PR *`khïbusi, PA *khïbusi [khibusi, `hibusi,
k h ï bu si ] , P O *k h i ˘ b us i [ k h i bu : ˘ s i, ` k ib u s i, k i n˘ s i] ,
PM *`kifusï˘ [kyu:sï˘, kiQfu˘, kiQfu], PY *kibusï [`kibusï,
kibusi, kibusi], Y `khibunci.
stone (2.2) PJ *e˘si, PMJ *i˘si [isi˘, i˘si, `isi, i˘si]: WOJ isi,
EOJ ishi, MJ isi, PR *`/isi˘, PA */isi [/isi, `/isi, /IsI],
PO */i˘si ~ *hi˘si [`/isi: ~ `hisi:, `/isi, /i˘si]: OR isi,
PM *`isï˘ [isï˘, isï˘, isï], PY *`isi˘ [`isi, isi, isi].
strength (3.3) PJ * ti˘k aÈra, PMJ * ti˘kaÈ ra [t ik ara ˘, ti ˘k ara ,
`tikara, ti˘kara]: WOJ tikara, MJ tikara, PR *`cikara˘,
PA *cikyara [cikyara, -- , cïkyara], PO *ci˘kara [cika:˘ra,
`cikara, --], PY *`cïkaraˆ [`cïkara, cika˘ra, sïkara], Y `sikara.
sweat (2.5) PJ *ase˘, PMJ *ase˘ [a˘se, ase˘, _ase, ase˘]: WOJ ase,
MJ ase, PR *_/asi˘, PA */asï˘ [/asi, -- , /asI˘], PO */asi
[_hasi:, _/asi, /ase], PM *_asi˘ [asi˘, asi˘, asi˘], PY *_asi˘
[_asi, aÈsi, asi], Y _asi.
tears (3.5) PJ *namyiÈta, PMJ *namyiÈta [na˘mida, na˘mida,
_nanda, namiÈda]: WOJ namyita, MJ namida, PR *na:da
LL(L), PA *nada˘ [nada, --, --], PO *nada [_nada:, _nada, --],
PM *_nada˘ [nada˘, nada˘, nada˘], PY *_na:da˘ [_nada,
na:Èda, nada], Y _nuda.
tooth (1.3) PJ *pa:˘, PMJ *paÈ: [ha˘, ha:˘, _ha, Èfa]: WOJ pa,
MJ ha, PR *_p h a:, PA *ha:˘ [ha, _ha:, ha:˘], PO *ph a: [_p h a:,
_ha:, ha:], PM *_pa:˘ [ha:˘, pa˘:, pa:˘], PY *_pa:˘ [_pa:, pa˘:,
pa:], Y _ha:
treasure (3.4 ) PJ *taÈkara, PMJ *taÈkara [takara˘, ta˘kara,
_takara, _takara]: WOJ takara, MJ takara, PR *_t h akara˘,
PA *thakara˘ [thakara, _takara, --], PO *thaka˘ra [_thaka:ra,
_takara, --]: OR takara, PM *_takara˘ [takara˘, takara˘, --],
PY *_takaraˆ [_takara, takaÈra, takara], Y _tagara.
 APPENDIX 373

tree (1.3) PJ *_k´y:, PMJ *_ki:y [ki˘, _ki:, _ki, _ki]: WOJ kyi,
EOJ ki, MJ ki, PR *_k h ï:˘, PA *kh ï:˘ [k h ï, _hi:, kï:˘], PO *k h i:
[_khi:, _ki:, ki:], PM *_ki:˘ [ki:˘, ki:˘, ki:˘], PY *_ki:˘ [_ki:,
ki˘:, ki:], Y _khi:
two (3.2) PJ *puta:tuˆ, PMJ *futatuˆ [hutatu˘, huta˘tu, `hutaQ,
`futa˘tu]: WOJ putatu, EOJ hutatu, MJ hutatu, PR *`futa:cï,
PA *ta:cï [ta:ci, `ta:ci, ta:ci], PM *`futa:cï˘ [futa:cï˘, futa:cï˘,
futa:cï], PY *futa:cï [`huta:zï, huta:ci, futa:ci], Y `ta:ci.
v oice (2.5) PJ *k´:way˘, PMJ *k´we˘ [ko˘e, k oe˘, _koe ,
kowe˘]: WOJ kowe, EOJ kowe, MJ kowe, PR * k h u :ïˆ,
PA *k h u˘i [k h u˘i, hui˘, ku˘i], PO *ku:˘i [hu˘i, _kwi:, --],
PM *kui [kui, ku˘i, kui˘], PY *kui [_kui, ku˘i, ku˘i], Y k h uiˆ.
washtub (3.7) PMJ *taÈrapyiˆ [tarai, tara˘i, _tarai, _tara˘fi]:
MJ tarahi.
whale (3.7) PJ *kuÈnsiraˆ, PMJ * kuÈziraˆ [k uzira, kuzi˘ra,
_kuzira, _kuzir˘a]: WOJ kudira, MJ kudira, PR *k h udiraˆ,
PA *kuziraˆ [kuzi˘ra, guzi˘ya, gun˘za], PO *guzi˘ra [_gunza,
_ guzira , --] , PM * kuzir a [f uQ za˘ , ku zi˘r a, ku zira ˘] ,
PY *_kuziraˆ [_kuzira, kuziÈra, kuzi˘ra], Y khudiraˆ.
wind (2.1) PJ *`kansay, PMJ *`kazey [kaze, `kaze, `kaze, `kaze]:
WOJ kazye, EOJ kaze, MJ kaze, PR *`k h adi˘, PA *k h azï
[hazye, `hazi, k h adI], PO *k h a˘zi [`hazi:, `kazi, --]: OR kaze
~ kazi ~ kasu ~ kase ~ kadi, PM *`kadi˘ [kadi˘, kazi˘, kazi],
PY *`kazi˘ [`kazi, kazi, kazi], Y `khadi.
wing (3.2) PJ *tumpa:saˆ, PMJ * tubasaˆ [tub asa, tuba˘sa,
`tubasa, `tuba˘sa]: WOJ tubasa, MJ tubasa.
yawn (3.1) PJ *`akumpi, PMJ *`akubi [akubi, `akubi, _akuQ,
`akubi]: MJ akubi, P R *`/akubi, PA */akubi [/aku bi,
`/akubi, /akubi], PO *ha˘kubi [hak u:˘bi, `/akubi, --],
PM *`afugï˘ [afuzï˘, afukï˘, afugï], PY *akubï [`akubï, aburi,
aubi], Y `agui.