{ Continued from front flap } HAWAI‘I STUDIES ON KOREA

LANGUAGE AND LINGUISTICS

Of related interest
comparisons are not due to pure chance,
they can almost always be explained as

VOVIN
borrowings from Korean into a central The Role of Contact in the Origins
group of Japanese dialects from roughly of the Japanese and Korean Languages The Japonic (Japanese and Ryukyuan)

KOREO-JAPONICA
between the third and eighth centuries J. Marshall Unger portmanteau language family and the
2008, 224 pages
A.D. The remaining group of lexical (but Korean language have long been consid-
cloth: ISBN 978-0-8248-3279-7
not morphological) comparisons that ered isolates on the fringe of northeast
cannot be explained in this way is, he Despite decades of research on the reconstruction of proto-Korean- A R E - E VA L U AT I O N O F A C O M M O N G E N E T I C O R I G I N Asia. Although in the last fifty years many
Japanese (pKJ), some scholars still reject a genetic relationship. This
argues, too small to serve as proof of even specialists in Japonic and Korean histori-
study addresses their doubts in a new way, interpreting comparative
a distant genetic relationship. cal linguistics have voiced their support
linguistic data within a context of material and cultural evidence,
In this volume, a leading historical much of which has come to light only in recent years. for a genetic relationship between the two,
linguist presents a significant challenge to this concept has not been endorsed by
a view widely held by Japonic and Korean general historical linguists, and no signifi-

KOREO-JAPONICA
historical linguistics on the relationship Also in Hawai‘i Studies on Korea cant attempts have been made to advance
between the two language families and beyond the status quo. Alexander Vovin, a
offers material support for the skepti- Questioning Minds longtime advocate of the genetic relation-
cism long espoused by general histori- Short Stories by Modern Korean Women Writers ship view, engaged in a reanalysis of the
cal linguists on the matter. His findings Yung-Hee Kim known data in the hope of finding evi-
will both challenge and illuminate issues 2010, 248 pages, illus. dence in support of this position. In the
of interest to all linguists working with cloth: ISBN 978-0-8248-3395-4 process of his work, however, he became
language contact and typology as well as paper: ISBN 978-0-8248-3409-8 convinced that the multiple similarities
those concerned with the prehistory and Available for the first time in English, the ten short stories by modern between Japonic and Korean are the result
early history of East Asia. Korean women collected here touch in one way or another on issues of several centuries of contact and do not
related to gender and kinship politics. All of the protagonists are descend from a hypothetical common
women who face personal crises or defining moments in their lives as
Alexander Vovin is professor of East ancestor.
gender-marked beings in a Confucian, patriarchal Korean society.
Asian languages at the University of In Koreo-Japonica, Vovin carefully
Hawai‘i. reviews recent advances in the recon-
struction of both language families. His
jacket art: (left) Lanterns at the back of Nigatsudō Temple, Nara, Japan;
detailed analysis of most of the morpho-
(right) a Paekche royal tomb, Gongju, Korea. (Photos by author) logical and lexical comparisons offered
jacket design: Julie Matsuo-Chun so far shows that whenever the proposed

{ Continued on back flap }

ISBN 978-0-8248-3278-0 ALEXANDER VOVIN
UNIVERSITY 90000
of HAWAI‘I
PRESS 9 780824 832780
HONOLULU, HAWAI‘I 96822-1888 www.uhpress.hawaii.edu
KOREO-JAPONICA
HAWAI‘I STUDIES ON KOREA

Wayne Patterson
The Ilse: First-Generation Korean Immigrants, 1903-1973

Linda S. Lewis
Laying Claim to the Memory of May: A Look Back at the 1980 Kwangju Uprising

Michael Finch
Min Yŏng-gwan: A Political Biography

Michael J. Seth
Education Fever: Society, Politics, and the Pursuit of Schooling in South Korea

Chan E. Park
Voices from the Straw Mat: Toward and Ethnography of Korean Story Singing

Andrei N. Lankov
Crisis in North Korea: The Failure of De-Stalinization, 1956

Hahn Moon-Suk
And So Flows History

Timothy R. Tangherlini and Sallie Yea, eds.

Sitings: Critical Approaches to Korean Geography

Alexander Vovin
Koreo-Japonica: A Re-evaluation of a Common Genetic Origin

Yung-Hee Kim, translator

Questioning Minds: Short Stories by Modern Korean Women Writers
HAWAI‘I STUDIES ON KOREA

KOREO-JAPONICA
A Re-evaluation of a
Common Genetic Origin

ALEXANDER VOVIN

University of Hawai‘i Press, Honolulu

and
Center for Korean Studies, University of Hawai‘i
© 2010 University of Hawai‘i Press
All rights reserved
Printed in the United States of America
15 14 13 12 11 10 6 5 4 3 2 1

Library of Congress Cataloging-in-Publication Data

Vovin, Alexander.
Koreo-Japonica : a re-evaluation of a common genetic
origin / Alexander Vovin.
p. cm. — (Hawai‘i studies on Korea)
Includes bibliographical references and indexes.
ISBN 978-0-8248-3278-0 (hard cover : alk. paper)
1. Korean language—Grammar, Comparative—Japanese.
2. Ryukyuan language—Grammar, Comparative—Korean.
3. Japanese language—Grammar, Comparative—Korean.
4. Korean language—Ethnology. 5. Ryukyuan language—
Ethnology. 6. Japanese language—Dialects—Japan—Okinawa-
ken. 7. Comparative linguistics. I. Title. II. Series.
PL912.V59 2009
495.7'2—dc22
2008044590

The Center for Korean Studies was established in 1972 to coordinate

and develop resources for the study of Korea at the University of
Hawai‘i. Reflecting the diversity of the academic disciplines
represented by affiliated members of the university faculty, the
Center seeks especially to promote interdisciplinary and intercultural studies.
Hawai‘i Studies on Korea, published jointly by the Center and the University of
Hawai‘i Press, offers a forum for research in the social sciences and humanities
pertaining to Korea and its people.

University of Hawai‘i Press books are printed on acid-free

paper and meet the guidelines for permanence and
durability of the Council on Library Resources.

CAMERA - READY PAGES PREPARED BY THE AUTHOR

Printed by Sheridan Books, Inc.

For Yasha
CONTENTS
CHARTS....................................................................................................... IX
ACKNOWLEDGMENTS ........................................................................... XI
ABBREVIATIONS................................................................................... XIII

INTRODUCTION.......................................................................................... 3

CHAPTER 1: PROTO-KOREAN AND PROTO-JAPONIC

RECONSTRUCTIONS AND THEIR ROLE IN THE COMPARISON
OF THE TWO LANGUAGES.................................................................... 11
1.1 RECENT ADVANCES IN PROTO-KOREAN
RECONSTRUCTION........................................................................... 11
1.1.1 Middle Korean aspirates are from clusters ................................. 11
1.1.2 Proto-Korean did not have vowel harmony................................ 11
1.1.3 A new look at lenition in Middle Korean ................................... 12
1.2 RECENT ADVANCES IN PROTO-JAPONIC
RECONSTRUCTION........................................................................... 32
1.2.1 Six or seven, not four vowels in Proto-Japonic.......................... 32
1.2.2 Another look at PJ *b- and *d-.................................................... 36

CHAPTER 2: MORPHOLOGICAL COMPARISONS ............................ 45

2.1 NOMINAL MORPHOLOGY............................................................... 45
2.1.1 Case marking................................................................................ 45
2.1.2 Pronouns ....................................................................................... 62
2.2 ADJECTIVES ........................................................................................ 71
2.3 VERBAL MORPHOLOGY.................................................................. 73
2.3.1 The odd nature of the copula tö................................................... 73
2.3.2 Verbal markers ............................................................................. 77

CHAPTER 3: LEXICAL COMPARISONS .............................................. 92

3.1 DOUBLETS IN WESTERN OLD JAPANESE.................................. 92
VIII———Koreo-Japonica

3.2 WHITMAN’S LEXICAL COMPARISONS ....................................... 94

3.2.1 *p- ................................................................................................. 94
3.2.2 *b- ............................................................................................... 111
3.2.3 *t- ................................................................................................ 113
3.2.4 *d- ............................................................................................... 126
3.2.5 *k- ............................................................................................... 128
3.2.6 *c-................................................................................................ 161
3.2.7 *j- ................................................................................................ 169
3.2.8 *s-................................................................................................ 172
3.2.9 *z-................................................................................................ 186
3.2.10 *š-.............................................................................................. 186
3.2.11 *m- ............................................................................................ 190
3.2.12 *n- ............................................................................................. 203
3.2.13 *y- ............................................................................................. 218
3.2.14 *a- ............................................................................................. 222
3.2.15 *ü-, *ö-...................................................................................... 229
3.2.16 *u-, *o-...................................................................................... 231
3.2.17 *e- ............................................................................................. 233
3.2.18 *i- .............................................................................................. 236
3.3 STATISTICAL SUMMARY OF LEXICAL COMPARISONS
AND CONCLUSION ......................................................................... 237

REFERENCES ........................................................................................... 241

INDEX ........................................................................................................ 251
CHARTS
1: Contrast between voiced fricatives and voiceless stops in Middle
Korean verbs ......................................................................................... 12
2: t/l alternation in Class 6 verbs as compared to Class 1 verbs in
Middle Korean ...................................................................................... 12
3: Middle Korean voiceless p, s and voiced W, z in intervocalic
position.................................................................................................. 12
4: Some forms of the verb anc- ‘sit’ in Modern Korean and
Ceycwuto .............................................................................................. 17
5: Comparative chart of ‘nasal insertion’................................................ 18
6: Frequency of NC clusters in Middle Korean ...................................... 23
7: Evidence from Korean dialects for NC and LC clusters .................... 25
8: Consonant alternations in Nivx ........................................................... 30
9: NC clusters in Nivx dialects................................................................. 31
10: Reflexes of PJ *i and *e ....................................................................... 33
11: Examples of PJ *i and *e ..................................................................... 33
12: Reflexes of PJ *u and *o...................................................................... 33
13: Examples of PJ *u and *o .................................................................... 34
14: Examples of PJ *e and *o in Eastern Old Japanese and
Proto-Ryukyuan.................................................................................... 34
15: Evidence for pre-OJ *e and *o in loanwords in Proto-Ainu ............. 35
16: Proto-Japonic vocalism ........................................................................ 36
17: Psara vu- ~ v- corresponding to OJ u- ................................................. 37
18: Psara u- corresponding to OJ u- .......................................................... 38
19: Psara vu- ~ v- corresponding to OJ pu- ~ mu- .................................... 38
20: Reflexes of /wa/ in Shimao and Tokyo ............................................... 39
21: Primary case markers in Western Old Japanese from a
comparative perspective....................................................................... 62
22: Western Old Japanese personal pronouns from a
comparative perspective....................................................................... 65
23: Western Old Japanese interrogative pronouns from a
comparative perspective....................................................................... 67
24: Proto-Japonic and Middle Korean demonstrative pronouns ............. 68
25: Evolution of Japonic demonstrative pronouns according to
Frellesvig and Whitman (2003)........................................................... 68
X———Koreo-Japonica

26: Attributive forms in Western Old Japanese and Shuri....................... 82

27: Some comparative verbal morphology ............................................... 90
28: Lexical doublets in Western Old Japanese ......................................... 93
29: Proto-Japanese-Korean reconstructions based on the
correspondence Middle Korean -l- : Old Japanese -k-.................... 115
30: Apocope in Middle Korean and Early Middle Korean .................... 163
31: ‘Do’ and ‘white’ in Korean and Japonic........................................... 190
32: ‘Three’ and ‘four’ in Middle Korean ................................................ 211
33: ‘Four’ and ‘stop in’ in Middle Korean and Old Japanese................ 212
34: Middle Korean numerals.................................................................... 220
35: Potential Koreo-Japonic cognates ..................................................... 238
36: Reliable Koreo-Japonic cognates ...................................................... 238
ACKNOWLEDGMENTS
Although I discussed various parts of this book or some ideas found in it
with a number of colleagues, I am especially grateful to Stefan Georg. To
him I owe my vast improvement in comparative linguistics methodology
over the last eight years: if not for him, the very idea of writing this book
never would have been conceived. I am also very grateful to Anton
Antonov, Chris Beckwith, Robert Blust, Blaine Erickson, Bjarke Frellesvig,
Jiha Hwang, Katsunobu Izutsu, Juwen Kim, Ross King, Samuel E. Martin,
Masayuki Ōnishi, Toshiki Osada, Sven Osterkamp, S. Robert Ramsey,
Mitsuru Sakitani, Leon Serafim, Moriyo Shimabukuro, Jun’ichi Suda, John
B. Whitman, and J. Marshall Unger who either provided important
feedback, answered my questions, or checked necessary data in certain
texts that were inaccessible to me. All mistakes and shortcomings in this
book remain my responsibility. I also feel gratitude to Roy A. Miller, who
over the years relentlessly demonstrated in all his publications how Japonic
historical linguistics and philology should not be done.
Some explanations and ideas found in this book were tested during
seminars on Western Old Japanese, Eastern Old Japanese, Middle Korean,
and Old Korean with both older and younger generations of students at the
University of Hawai‘i at Mānoa: James Baskind, John Bentley, Timothy
Harris, Sukenari Hino, Kyuseek Hwang, Steven Ikier, John Kupchik,
Linda Lanz, William Matsuda, Marc Miyake, Matthew McNicoll, Matthias
Nyitrai, Kyoungwon Oh, Kerri Russell, Moriyo Shimabukuro, and Lina
Terrell.
Nancy R. Woodington deserves special credit for her excellent work as
the editor of this book. I am also extremely grateful to Ann Ludeman, UH
Press copy managing editor, for guiding me through the process of the
book formation.
Last but not least, my gratitude goes to the members of my family: my
mother Svetlana, my wife Sambi, and our son Yasha, who has put up for a
long time with konpyūtā ni haritsuichatta papa. This book is dedicated to
him.
ABBREVIATIONS
LANGUAGES AND DATA

Ainu
PA Proto-Ainu

Austronesian
PAN Proto-Austronesian

Chinese
EMC Early Middle Chinese
LMC Late Middle Chinese
MC Middle Chinese
Indo-European
PIE Proto-Indo-European

Japonic
CR Classical Ryukyuan
EMdJ Early Modern Japanese
EOJ Eastern Old Japanese
MdJ Modern Japanese
MJ Middle Japanese
OJ Old Japanese
OR Old Ryukyuan
PJ Proto-Japonic
PJK Proto-Japanese-Korean (Whitman)
PJN Proto-Japanese
PR Proto-Ryukyuan
WOJ Western Old Japanese
YG Yonaguni
Korean
EMdK Early Modern Korean
EMK Early Middle Korean
LMK Late Middle Korean (Whitman)
MdK Modern Korean
MK Middle Korean
NC Northern Ceycwuto
OK Old Korean
XIV———Koreo-Japonica

PK Proto-Korean
SC Southern Ceycwuto
SK Sino-Korean

Mongolic
MM Middle Mongolian
WM Written Mongolian

Tungusic
Ma. Manchu
PT Proto-Tungusic

LINGUISTIC TERMS

ABS absolutive
ACC accusative
ACT active
ADV adverbializer
ASSER assertive
ATTR attributive
CAUS causative
CL classifier
COM comitative
COMP comparative
CON conjunctive gerund
CONC concessive gerund
COND conditional gerund
CONJ conjunction
CONJC conjectural
COOP cooperative
COOR coordinative gerund
COP copula
DAT dative
DEB debitive
DES desiderative
DIM diminutive
DIR directive
DV defective verb
EMPH emphatic
EV evidential
EXCL exclamation
F falling pitch
FIN final verbal form
GEN genitive
GER gerund
H high pitch
HON honorific
 Abbreviations———XV

HUM humble
IMP imperative
INF infinitive
INTER interjection
INTL intentional
IRR irrealis
L low pitch
LF locative focus
LOC locative
MOD modulator
NEG negative
NML nominalizer
NOM nominative
PAST past tense
PERF perfective
PL plural
POL polite
POSS possessive
PREF prefix
PRES present tense
PRET preterite
PROG progressive
PROH prohibitive
PT particle
QUOT quotative
R rising pitch
REAL realis
RETR retrospective
SUB subordinative gerund
SUBJ subjunctive
SUP suppositional
TENT tentative
TERM terminative
TOP topic
TRANSF transferentive gerund
VOC vocative
VOL volitional

PRIMARY SOURCES

Japonic
BS Bussoku seki uta (ca. 756)
FK Fudoki (ca. 733)
GM Genji monogatari (ca. 1008)
IM Ise monogatari (late ninth century)
KJK Kojiki (712)
KK Kojiki kayō (712)
XVI———Koreo-Japonica

KKW Kokin wakashū (921)

MN Murasaki shikibu nikki (1010)
MS Makura no sōshi (1000)
MYS Man’yōshū (ca. 759)
NK Nihonshoki kayō (720)
NR Nihon ryōiki (early ninth century)
NS Nihonshoki (720)
OS Omoro sōshi (sixteenth century)
RK Ryūka (eighteenth-nineteenth centuries)
SM Senmyō (697-791)
USM Uji shūi monogatari (ca. 1213-22)
WMS [Ruiju] Wamyōshō (931-38)

Korean
HMC Hwungmin cengum (1445)
Hwungmwong Hwungmwong cahwoy (1527)
Kumsam Kumkangkyeng samka enhay (1482)
Kwukup Kwukuppang enhay (1466)
Kyeylim Kyeylim yusa Kolye pangen (1103)
Mwongpep Mwongsan hwasan pep.e yaklok enhay (1517)
Nammyeng Nammyeng cip enhay (1482)
Nayhwun Nayhwun (1475)
Nung Nungemkyeng enhay (1461)
Pak cho Pak thongsa chokan (1517)
Pennwo Penyek nwokeltay (1515[?])
Pep Pephwakyeng enhay (1463)
PT Pak thongsa (1517)
Samkang Samkang hayngsiltwo (1481)
SCH Sekpong chencamun (1583)
Sekpo Sekpo sangcel (1449)
SS Samkwuk saki (1145)
Twusi cho Twusi enhay chokan (1481)
Twusi cwung Twusi enhay cwungkan (1632)
WCK Welin chenkang ci kwok (1449)
Welin Welin sekpo (1459)
Welin se Welin sekpo se (1459)
Wenkak Wenkakkyeng enhay (1465)
Wunhay Hwungmin cengum wunhay (ca. 1750-81)
Yek.epo Yek.e yuhay po (1775)
YH Sincung yuhap (1576)
YP Yongpi echenka (1447)
Yukco Yukco potankyeng enhay (1496 [?])
KOREO-JAPONICA
INTRODUCTION
This book attempts to critically re-evaluate the relationship between
Korean and Japonic. 1 It is quite apparent that Korean and Japonic are very
similar, often to the extent that a word-to-word translation is possible from
one language to another. Such similarity is, however, purely typological
and cannot be used as evidence for a common genetic origin. The theory
that the two languages are genetically related was originally proposed in
the eighteenth century by Fujii Teikan, a Japanese scholar. The following
century saw very little scholarly activity on the matter, but the issue
became a subject of scholarly works once again between the late
nineteenth and early twentieth centuries, starting with Aston’s pioneering
study (1879), which was followed by Kanazawa (1910) and Ogura (1934).
All these publications supported the idea that Korean and Japonic are
genetically related, and it seems that the only person who held strong
reservations was Hattori (1959).
Despite the fact that there are many important publications on
comparative Koreo-Japonic, starting from Martin’s seminal work (1966),
with Whitman’s outstanding dissertation (1985) deserving a special
mention, I feel that proof of a genetic relationship between Korean and
Japonic is as lacking now as it was prior to 1966. In other words, although
many publications by Martin, Ramsey, Whitman, Serafim, Frellesvig, King,
Unger, and the present author have attempted to solve some particular
problems or proposed new and interesting etymologies, no substantial
progress has been made. Koreo-Japonic as a valid genetic family fails to
convince general linguists. The majority of Western linguists who are
engaged in the historical study of either Japanese or Korean, or both,
accept the genetic relationship between these two languages, while most
historical linguists in Korea and Japan are either skeptical or ambivalent. I
believe that this situation calls for a re-evaluation of all the progress that
has been made in reconstructing and comparing Proto-Korean and Proto-
Japonic that has led to the Koreo-Japonic hypothesis. To proceed with this
re-evaluation, I first need to define some fundamental notions, such as,
what a proto-language is, and what part of the Korean or Japonic heritage
is going to be treated as belonging to a respective proto-language.

Defining a proto-language
Sometimes one can observe a dangerous tendency in the field to treat data
from existing old languages as if they represent the respective proto-
languages. This is less of an issue on the Korean side, but on the Japanese
side the Western Old Japanese of the Asuka-Nara periods sometimes
1
See below on terminological difference between ‘Japonic’ and ‘Japanese’.
4———Koreo-Japonica

receives royal treatment, as though it stands in the same relationship to all

other known varieties of Japonic as Latin to all Romance languages. In
spite of the importance of Western Old Japanese to the history of Japonic,
we should not forget that it represents a very old stage of just one variety
of the family, namely Central Japanese, which is characterized by a certain
set of innovations that did not happen elsewhere. In this study I advocate
the position that a proto-language can be reconstructed only from a wide
range of data, including philological data from old languages, data from
modern languages and dialects, and internal reconstruction. Below I
provide exact definitions of how Proto-Korean and Proto-Japonic are
understood.
There are also areas traditionally neglected in the reconstruction of
Proto-Korean and Proto-Japonic. In Korean historical linguistics these
include almost all pre-alphabetic sources on Early Middle and Old Korean,
since they only occasionally find their way into the works on
reconstruction. In Japonic historical linguistics the same fate is shared by
the Eastern Old Japanese and the Old Ryukyuan languages. I try as much
as space allows to address these data in this study.

What can be called Proto-Korean?

There is less internal diversification in Korean than in Japonic, although
the Ceycwuto dialect on Ceycwuto island and the Yukcin dialect still
spoken in Northern Hamkyeng, as well as by immigrant communities in
China, Russia, Uzbekistan, and Kazakhstan, are very different from the
rest of Korean and from each other. Korean is also less fortunate than
Japanese in the respect that before 1443 (that is prior to the invention of
the alphabetic Hankul script) it is attested much more scantily than
Japanese, and even the attested data are often partially hidden under
semantographic Chinese script. Finally, even bits of data that happen to be
written phonetically are often poorly understood or misinterpreted due to
the application of outdated Chinese reconstructions like those of Karlgren. 2
This situation leads to the acceptance of certain phenomena found only in
Middle Korean as ‘Proto-Korean’. Such an approach is fallacious. R pitch,
for example, is found in Middle Korean, but it is by no means a Proto-
Korean feature. In this book I have tried to avoid equating Middle Korean
with Proto-Korean. In general, a phoneme, a morpheme, or a word is
considered to be Proto-Korean if it is supported by evidence present: (1) in
Middle Korean, Old Korean, and/or Early Middle Korean; (2) in any one
of the above and in the Ceycwuto dialect and/or in the Yukcin dialect; (3)
in modern standard Korean and in both the Ceycwuto and Yukcin dialects.
However, when no dialect data from Ceycwuto and/or Yukcin are
available, an exclusively Middle Korean phenomenon may be tentatively
treated as ‘Proto-Korean’.

2
Nevertheless, the tide is starting to turn, and pre-Hankul materials are slowly but steadily
finding the place they deserve in the domain of Korean historical linguistics.
 Introduction———5

What can be called ‘Proto-Japonic’ and ‘Proto-Japanese’?

Japonic represents a more diverse family than Korean, although it is in a
sense a portmanteau family, consisting of two languages: Japanese and
Ryukyuan, with a sharp boundary between the two to the north of the
Amami island group in the Ryukyuan archipelago. Following a suggestion
made by Leon Serafim more than ten years ago, I will use ‘Japonic’ as the
term for the whole family, while reserving ‘Japanese’ for ‘Japanese proper’,
i.e., all varieties of Japonic that are non-Ryukyuan. While the division
between Japanese and Ryukyuan is clear and rather self-evident, the
internal taxonomy of both Japanese and Ryukyuan is much less apparent,
especially in the case of Japanese.
Ryukyuan can be subdivided into Northern and Southern, with a major
boundary to the south of Kumejima island, which belongs to the Okinawa
island group. Within the Northern group, it is possible to speak about
Amami and Okinawan subgroups, although the border between them is not
clear: usually it is drawn to the north of Okinawa island (Hirayama 1966,
map #1), but some of the dialects found in the north of Okinawa island
share some substantial innovations with the Amami dialects, innovations
not found in the center or south of Okinawa island, such as, e.g., secondary
aspiration. The Southern group consists of the rather well-defined Miyako
and Yaeyama dialect subgroups, but in addition there is also a separate
dialect on the island of Yonaguni, and its position in South Ryukyuan, and
possibly even the whole Ryukyuan taxonomy, is not clear. It must be
added that all Ryukyuan dialects were and still are subject to strong
Japanese influence. Historically, this influence was strongest in the
Northern Ryukyus, especially in the southern part of Okinawa island in the
region adjacent to Shuri. The Southern Ryukyuan islands were
comparatively unaffected by Japanese influence until the end of the
nineteenth century.
Japanese dialects, with the major exceptions of the Tōhoku and
Southern Kyūshū dialects, as well as the dialect of the Hachijō islands, are
much closer to each other than the Ryukyuan dialects. There are reasons to
believe, though, that the present situation is rather recent, resulting from
more than a thousand years’ influence from Central Japanese. All known
literary varieties of Japanese except Eastern Old Japanese were based on
some dialect belonging to the Central Japanese dialect group. Due to the
existence of Eastern Old Japanese in the eighth century, we know that the
eastern border of Central Japanese roughly coincided with the eastern
border of the present-day Kansai region; however, during the course of
history Central Japanese expanded considerably to the west and east, while
the descendants of the Eastern Old Japanese dialect continuum receded to
Hachijō island. It is much less clear what the western border of Central
Japanese was in the eighth century, but the existence of a quite different
morphological residue in the modern dialects of Western Honshū means it
is probably safe to assume that this border more or less coincided with the
present-day border between the Kansai and Chūgoku regions. Nowadays
all Japanese dialects located in Shikoku and Honshū (except Tōhoku) can
6———Koreo-Japonica

be classified as Central Japanese. The dialects of Northern and Central

Kyūshū probably were originally different from Central Japanese as well,
but today they represent a kind of intermediate stage between Central
Japanese and Southern Kyūshū.
In Japonic, a phoneme, a morpheme, or a word is considered to be
Proto-Japonic (PJ) if it is found: (1) in Western Old Japanese and/or
Eastern Old Japanese (areas A and B) 3 and in at least one of the Southern
Ryukyuan dialects that escaped Middle Japanese influence, or, because the
Old Japanese corpus is somewhat limited; (2) in Middle Japanese and in at
least one of the Southern Ryukyuan dialects that escaped Middle Japanese
influence; and (3) in Ryukyuan and in at least one of the non-Central
Japanese dialects. A word or a morpheme is considered Proto-Japanese
(PJN) if it is found: (1) in Western Old Japanese and Eastern Old Japanese
(areas A and B); (2) in Middle Japanese and Eastern Old Japanese (areas A
and B).

Some fundamental typological differences

Although modern Korean and Japanese are remarkably similar
typologically, so that it is possible in most cases to provide a word-to-word
translation, it appears that the more we go into the past, the more
tantalizing differences we discover, even if we limit ourselves to the
history of the written language. Although typology cannot be used as
evidence in establishing genetic relationships between languages, the fact
that Korean and Japanese are typologically more similar now than they
were in the past suggests that convergence, not divergence, was at work.

Ergativity
Korean is historically ergative (King 1988), while Japanese is historically
nominative, although Western Old Japanese had a short-lived active
construction that is not attested in other branches of Japonic (Vovin 1997a;
see also the discussion of Western Old Japanese active marker -i in 2.1.1.1
below).
Passive
It appears that the morphological passive in Korean is quite young and that
it developed from a morphological causative (Whitman 2003: 3), which is
not surprising for a language that used to be ergative. Japanese, on the
other hand, had the passive throughout its history, although its use has
some peculiarities when compared to the Indo-European passive.

Ablaut
Even Modern Korean still has an active ablaut in its color terms and
onomatopoetic words (‘heavy’ isotopes with back vocalism and ‘light’
isotopes with original non-back vocalism, in Martin’s terminology, cf.
hayah- and huy- ‘be white’, phalah- and phwulu- ‘be blue’). There is an

3
Detailed language classification and explanations are provided below.
 Introduction———7

indication that ablaut was much more productive at the earlier stage of the
language, cf. MK nòlk- ‘be old (of things)’, nùlk- ‘be old (of humans)’;
múl ‘water’, mùlk- ‘be watery’, mòlk- ‘be clear’; púl ‘fire’, pùlk- ‘be red’,
pòlk- ‘be bright’. While some linguists claim that Old Japanese has a
similar system, e.g., kötö 2.3 ‘word’ and katar- A ‘to tell’, the data do not
stand up to scrutiny, as in this case: these two Japanese words belong to
two different accentual registers, and therefore cannot be related. There is
no reconstructable ablaut for Proto-Japonic.
CHAPTER 1: PROTO-KOREAN AND PROTO-JAPONIC
RECONSTRUCTION AND THEIR ROLE
IN THE COMPARISON

CHAPTER 2: MORPHOLOGICAL COMPARISONS

CHAPTER 3: LEXICAL COMPARISONS

1

PROTO-KOREAN AND
PROTO-JAPONIC
RECONSTRUCTIONS AND THEIR
ROLE IN THE COMPARISON OF THE
TWO LANGUAGES
1.1 RECENT ADVANCES IN PROTO-KOREAN
RECONSTRUCTION
Below, I will address the most important advances in the reconstruction of
Proto-Korean that are of great importance to its comparison with Japonic.

1.1.1 Middle Korean aspirates are from clusters

Ramsey (1991, 1993) and Yi Kimun (1991: 18) have demonstrated that the
Middle Korean aspirates ph, th, ch, and kh have a secondary origin
resulting from clusters of HC or CH type, where H stands for velar [k] or
pharyngeal [h]. Ramsey’s and Yi Kimun’s idea is based predominantly on
the analysis of gaps in the internal structure of Korean: there are no *CH or
*HC (where C stands for a stop obstruent and H for a velar) clusters in
Middle Korean, while other types of clusters are present. For example,
clusters such as pc, pt, and ps are extant, but *pk is not. In addition, there is
some purely historical evidence from pre-Hankul materials. Thus, e.g., MK
khú- ‘to be big’ is a later form of EMK huku- ‘id’. (Kyeylim #348), and
MK thó- ‘to ride’ developed from EMK hoto- ‘id’. (Kyeylim #99). In
addition, Old Japanese loanwords from some kind of Old Korean
(probably Paekche) also offer support for the Ramsey-Yi theory, e.g., OJ
pötökë ‘Buddha’, cf. MK pwùthyè ‘id’. and OJ patakë ‘dry field’, cf. MK
pàth ‘field’. It seems that this theory received general acceptance.

1.1.2 Proto-Korean did not have vowel harmony

Recently, Martin painstakingly demonstrated that Middle Korean vowel
harmony was a recent and rather short-lived phenomenon, since it did not
exist in Old Korean (2000: 1-23). An earlier suggestion about the absence
of vowel harmony in Old Korean was made in Vovin (1995: 226),
although without extensive argumentation. Thus, for example,
synharmonic variants -e of the infinitive -a, and -te- of the retrospective
-ta- represent a post-Old Korean innovation, as only -a and -ta- were used
in Old Korean.
12———Koreo-Japonica

1.1.3 A new look at lenition in Middle Korean

Lenition in Middle Korean is a much more controversial problem. Middle
Korean exhibits consonantal lenition of p > W, t > l [r], k > G, and s > z. I
briefly survey the environments for each type of lenition below, as well as
different explanations for the phenomenon’s conditions and origins.
Interested readers should consult Ramsey (1991), Martin (1996), and
Vovin (2003b) for more details.
Voiced fricatives W and z occur only in medial position, where they
contrast phonemically with voiceless p and s in intervocalic position:

Chart 1:
Contrast between voiced fricatives and voiceless stops
in Middle Korean verbs
Class 1 verbs Class 6 verbs
kwùpúmyén 1 ‘if [it] is bent’ kwùWúmyén ‘if [he] bakes’
ìsúmyén ‘if [he] exists’ nìzúmyén ‘if [he] joins’

Class 6 verbs show the morphophonemic alternations between p/W and

s/z: thus, :kwupta ‘[he] bakes’, but kwùWúmyén ‘if [he] bakes’; :nista ‘[he]
joins’, but nìzúmyén ‘if [he] joins’. A very similar contrast is found in the
alternation of t/l for Class 6 verbs, and its lack in Class 1 verbs:

Chart 2:
t/l alternation in Class 6 verbs as compared to Class 1 verbs in Middle
Korean
Class 1 verbs Class 6 verbs
mwùttá ‘[he] dyes’ :mwuttá ‘[he] asks’
mwùtúmyén ‘if [he] dyes’ mwùlúmyén ‘if [he] asks’

The voiced fricative G occurs in verbal stems only within the clusters
lG and zG, never in intervocalic position. All stems with these clusters are
Class 8 verbs, which do not have the k/G alternation in Middle Korean, e.g.,
kàlótá ‘[he] divides’, but kàlGómyén ‘if [he] divides’. I will return to this
problem later.
Both voiceless p, s and voiced W, z are attested in intervocalic position
in nouns.

Chart 3:
Middle Korean voiceless p, s and voiced W, z in intervocalic position
Voiceless Voiced
nàpwóy/nàpóy ‘butterfly’ tùWúy ‘kind of gourd used
as a jar’
nìpúl ‘quilt’ :syeWùl ‘capital’
:cyepì ‘swallow’ kòlòWí ‘drizzling rain’

1
The Middle Korean pitches are rendered in the following way: ` — low pitch, ´ — high
pitch, and : (before a syllable) — rising pitch.
 Proto-Korean and Proto-Japonic Reconstructions———13

àpí ‘father’ sàWí ‘shrimp’

tàsós ‘five’ còzó ‘kernel’
ìsúl ‘dew’ nàzí ‘shepherd's purse’
kàsóy ‘thorn’ kòzGáy ‘scissors’
kwùsúl ‘treasure’ kèzwúy/kèzGwúy ‘worm’
pèsús ‘mushroom’ kìzúm ‘weed’
pyèsúl ‘official rank’ mòzòm ‘heart’
kìsúlk ‘border’ mòzòlh ‘village’

Since the t/l alternation involves /l/, which also exists as a separate
phoneme, it is virtually impossible to decide which intervocalic -l- in
nouns may be a plain /l/, or the result of the t/l alternation, unless we have
doublets, e.g., pàtáh ~ pàlól ‘sea’, where it is obvious that -l- in the second
variant is a result of the t/l alternation.
Since the Middle Korean writing system did not have a special symbol
for G, the question of whether both G and k occurred in intervocalic
position in nouns can be answered only with the help of dialects that
differentiate the cases with G and W > Ø.
The problem that occupied the minds of many Korean language
historians is, in a nutshell: do the voiced W, z, G, and t/l alternation
represent a secondary development in Middle Korean, or do they reflect a
series of phonemes that have also been distinct in Proto-Korean? There are
two basic solutions, which we will call here the ‘lenition theory’ and the
‘voiced obstruents theory’. The lenition theory is a traditional explanation,
proposed in a number of classical works on Korean language history (Yu
Changton 1964, Yi Kimun 1987). It enjoys a certain amount of support in
Korea, and was also the choice of Ramsey (1978), who added new data.
Martin’s (1996) monograph dedicated to this problem presents the best
explication of the lenition theory. According to the lenition theory, voiced
W, z, G, as well as l in the t/l alternation, are secondary products of the
lenition of voiceless p, s, k, and t. Martin defines the environment for the
lenition as ... °Co/u (1996: 3). 2 The alternative voiced obstruents theory
was proposed by Ramsey, who suggested that MK W, t/l, G, and z are to be
reconstructed as voiced obstruents *b, *d, *g, and *z in Proto-Korean,
which were phonemically distinct from PK *p, *t, *k, and *s (1991: 225-
227). Let us review the weak points of these two opposing theories.

1.1.3.1 Weaknesses of the lenition theory

The major problem of the lenition theory as presented by Martin (1996) is
the lack of a satisfactory explanation why stems of Class 1 verbs do not
lenite, while stems of Class 6 verbs do, in the same environment, e.g.,
kwùpúmyén ‘if [it] is bent’, but kwùWúmyén ‘if [he] bakes’. Martin is, of
course, aware of the problem, so he postulates that the stems of Class 1
verbs were originally monosyllabic (*kwùp-), while stems of Class 6 verbs

2
Martin's raised circle ° in front of the syllable denotes high pitch and corresponds to the
acute sign ´ over a vowel in my notation.
14———Koreo-Japonica

were originally disyllabic (*kwùpú-), resulting in a different kind of

juncture (1996: 6-7). This explanation, however, is not without its own
problems.
First, as Ramsey (1991: 226) pointed out, native speakers are unlikely
to differentiate between different kinds of morphological boundaries in
kwùpúmyén ‘if [it] is bent’ and kwùWúmyén ‘if [he] bakes’, even if we
follow Martin and assume that the stems of Class 1 verbs are indeed
monosyllabic, and the stems of Class 6 verbs are disyllabic. Nor are the
native speakers likely to be aware of different types of junctures.
Second, the environment of ... °Co/u for lenition, suggested by Martin,
seems to have no universal phonological or phonetic motivation: why does
lenition occur only in front of a minimal vowel? 3
Third, there are cases of Sino-Korean vocabulary items with apparent
lenition occurring in other environments, e.g., molan ‘peony’ < LMC
məw´tan, chalyey ‘order’ < LMC tshz`tɦiaj`, kulan/kelan ‘Khitan’ < LMC
khjiaj`(khjiat) tan. Martin mentions these in his lenition monograph (1996:
20-21), but does not explain why lenition occurs here in a different
environment. To these examples I can add a Koreo-Chinese hybrid
compound táylwòng ‘bamboo pipe’ (LCT 1987: 205b), where the second
element -lwong comes from LMC tɦəwŋ ‘pipe’.
Fourth, there are a number of contradicting examples, where lenition
does not occur in the environment specified by Martin. He is, of course,
aware of the problem, and provides a list of these cases (1996: 55): kàsóm
‘breast’, sàsóm ‘deer’, tàsós ‘five’, yèsús ‘six’, pèsús ‘mushroom’, ìsúl
‘dew’, kìsúlk ‘border’, pyèsúl ‘government post, official rank’, kwùsúl
‘bead, treasure’, ìsólàs ‘cherry’, kwòsóy ‘coriander’, nàpóy ‘butterfly’. As
a solution, Martin proposes that these words originally had a cluster,
probably C[h], that would block lenition. Unfortunately, the last example
seems to contradict the C[h] cluster theory: *nàp[h]óy would result in
nàphóy, with aspirated -ph-, and not nàpóy. Furthermore, there are other
examples that, I believe, contradict Martin's rule: ètúy ‘where’, ètúW- ‘to
be dark’, kùtùy ‘you’ (although the accent environment in this example is
different), mwòtón ‘all’, :mwotóy ‘necessarily’, pwùtúl ‘bulrush’, pètúl
‘willow’, àtól ‘son’, yètúlp ‘eight’, yètún ‘eighty’. One can, of course,
claim that ètúy ‘where’ and kùtùy ‘you’ are compounds, but the other

3
Martin suggested to me that ‘precisely because the vowel is ‘minimal’ the phonological
environment suggests a weakening of a marking in the string’ (Martin, personal
communication). While the conditions for lenition are often idiosyncratic in various
languages, crosslinguistically it is quite clear that the prime environment for consonantal
lenition is intervocalic position V_V, with lenition in absolute initial position #_V
occurring more rarely, and rarely present in the protected environment _C (Lass 1984: 182).
Furthermore, a very well-known case of lenition of voiceless stops in Latin to voiced
fricatives in Spanish frequently involves cases when the second vowel in the environment
V_V is long: Lat. habēre ‘have’ > Sp. haber [aBer], Lat. natāre ‘swim’ > Sp. nadar
[naDar]. Long vowels are not minimal in any case, and this suggests that consonantal
lenition can occur in any intervocalic position, regardless of the quality of the following
vowel.
 Proto-Korean and Proto-Japonic Reconstructions———15

words are obviously not so. 4 Besides, both ètúy and kùtùy are old
compounds, so the question is: why does lenition affect certain Chinese
loans, but not these words?
Fifth, Martin’s restriction on lenition occurring only inside morphemes
seems to be contradicted by onset lenitions, discussed by Martin himself,
such as gerund -kwó => -Gwó or effective -ké/á => -Gé/á (1996: 30-34).
If lenition in the onset of a certain morpheme is possible, why should there
be a restriction preventing lenition in the offset of another morpheme?
These five are the main problems that beset the lenition theory in its
present form. Let me now turn to the weaknesses of the voiced obstruent
theory.
1.1.3.2 Weaknesses of the voiced obstruent theory
When I first learned of Ramsey’s alternative theory about the origin of MK
W, t/l, G, and z many years ago, it seemed to me that his reconstruction of
MK W, t/l, G, and z as Proto-Korean voiced obstruents *b, *d, *g, and *z
explained the facts much better than the traditional lenition theory. One
important argument in favor of Ramsey's explanation is that stems
containing voiced *-m- and *-n- occur only in Class 6, but not in Class 1
verbs. Since *b, *d, *g, and *z share the feature of voicing with *m and *n,
that seemed to be a powerful argument. However, I now tend to think that
this argument has its problems, too.
First, there is no internal evidence that Korean ever had voiced initial
obstruents, and a language that has a phonemic contrast between voiceless
and voiced obstruents only in the medial position is typologically odd. To
the best of my knowledge, there are no such languages attested.
Second, the voiced obstruent theory fails to explain compounds like
kòlòWí ‘drizzling rain’, consisting of MK kòlò ‘powder, flour’ and pí ‘rain’.
The only way for the voiced obstruent theory to explain the apparent
lenition of initial p- when it becomes medial is to claim that the medial
4
Martin indicated to me (personal communication) that mwòtón ‘all’ is the adnominal form
of the transitive verb mwòt- ‘gather’ and that the nouns pwùtúl ‘bulrush’, pètúl ‘willow’,
and àtól ‘son’ may have incorporated plural marker -tó/úl(h) at an earlier stage. I agree that
mwòt-ón ‘all’ can be explained with the traditional lenition theory as presented in Martin
(1996). But it can also be explained with the revised lenition theory I present later. However,
I disagree that any of the above nouns might have incorporated the plural marker -tó/úl(h) at
an earlier stage, as I think that the internal evidence contradicts this proposal. First, the
earliest attested MK shape of the plural marker is just –tólh, which is not affected by vowel
harmony, and the earliest attestation of -túlh is only in Sohak enhay (1586) (LCT 1987:
252). Martin has demonstrated that Old Korean was a language without vowel harmony,
and that vowel harmony developed only in Middle Korean (Martin 2000). Under this
scenario we would not expect that -tólh should follow the rules of vowel harmony in
lexicalized old compounds, while it does not follow these rules when used as a bound
morpheme until the end of the sixteenth century. Second, the plural marker -tólh includes
an unmistakable final -h (Yi 1961: 233) that is definitely not present at least in pètúl
‘willow’ and àtól ‘son’. The absence of -h in pwùtúl ‘bulrush’ is more difficult to ascertain,
since this word is predominantly attested as a gloss in dictionaries, and in the only
attestation where it appears followed by an instrumental case marker: pwutul-lwo (not
pwutulh-olwo). This attestation is from Twungchangkyeng hempang (1663) and is too late
to be used as definite evidence.
16———Koreo-Japonica

position in compounding preserves an original voiced obstruent, while in

the initial position it becomes voiceless. However, in this particular case
this argument is dubious in light of both internal and external evidence: the
word for ‘rain’ is transcribed in the Kyeylim yusa (#7) as piWi, and is likely
borrowed from Tungusic *pigi- ‘[to] rain’ (Vovin 2000: 146-147).
Third, it is surprising that there are only five -n stems, and three out of
these five belong to Class 6 (:sin- ‘to wear shoes’, :an- ‘to embrace’,
and :ten- ‘to wager, to bet’), which was pointed out by Martin (1996: 5).
The rarity of -n stems has to be explained, and the voiced obstruent theory
cannot account for it. Unless one can find a cogent explanation for this
imbalance, it seriously diminishes the force of Ramsey’s argument.
Fourth, if one follows Ramsey’s reconstruction, it turns out that
preservation of the high pitch accent on the second syllable of the stem will
be associated with a voiced obstruent. According to Ramsey (1991), Proto-
Korean disyllables were assigned an automatic L-H pitch. However, we
would expect that a voiced medial consonant would lower the pitch of the
following vowel. But this apparently did not happen, and the voiced
obstruent theory provides no answer for the oddity, since the feature
[+voice] is present in both Proto-Korean and Middle Korean. However, if
we assume that the medial consonants that became MK -W-, -l-, -z-, and
-G- were once voiceless, we have a plausible explanation if the ordering of
our rules is as follows: first, the pitch was phonologized, that is, it was no
longer a simple phonation, and only after that did voicing occur.
These facts, with the aforementioned Sino-Korean words that offer
historical evidence for lenition, now make me suspicious of Ramsey’s
voiced obstruent theory. Lenition theory, in spite of its obvious weaknesses,
seems to have more explanatory force. However, I think that it can be
further refined.

1.1.3.3 A revision of the lenition theory

The revision of the lenition theory that I attempt below is based on internal
considerations, which are further supported by typology. I believe that all
my predecessors were trying to answer the question: why does lenition
occur in Class 6 stems? I arrived at my solution by asking a different
question, namely: why is there no lenition in Class 1 stems?
I believe the answer to this question is rather simple: Class 1 stems
originally did not have a single consonant, but a consonant cluster, which
blocked lenition:

Class 1 kwùpúmyén ‘if [he] bends’ < *kwùCpú-

Class 6 kwùWúmyén ‘if [he] bakes’ < *kwùpú-

Therefore, I agree with Ramsey that the stems of both Class 1 and Class
6 verbs were disyllabic (1991: 227). The cluster in Class 1 blocked the
lenition, while in Class 6, where there was no cluster, lenition occurred.
The process of lenition was no longer productive when clusters in Class 1
 Proto-Korean and Proto-Japonic Reconstructions———17

verbs were simplified. This explanation demands that we prove the

existence of these clusters, not just posit them.

1.1.3.3.1 Where did the *NC clusters go?

The first argument in favor of a cluster in Class 1 stems, as well as in those
nouns that do not show lenition, comes from a skewed distribution of
medial consonant clusters in Middle Korean. The following medial clusters
are attested: -ks-, -ps-, -sk-, -lk-, -st-, -lth-, -lt-, -lph-, -lW-, -lp-,
-lm-, -lh-, as well as aspirated -ph-, -th-, -kh-, and -ch-, that go back to
earlier *CH or *HC clusters (Ramsey 1977, 1991: 231). What is striking is
the rarity of clusters starting with nasals -m-, -n-, or -ng-. It is believed that
there are only a few words that contain -nc- or -mc- clusters, like anc- ‘to
sit’, tenci- ‘to throw’, or memchwu- ‘to stay, to stop’. The traditional
explanation for these clusters is that they appear to represent a secondary
-n- or -m- insertion in earlier Middle Korean forms ac- ‘to sit’, teti- ‘to
throw’, and mec- ‘to stay, stop’ (Yi 1964, Ramsey 1978: 54-56). This
seems to present a problem for the new version of the lenition theory, since,
as noted by one reviewer, the nasal insertion and nasal deletion that I
propose here work in opposite directions. However, the situation with
nasal insertion may not be as straightforward as it may seem. It is true that
Middle Korean presents evidence for some of the forms without nasals. It
is also true that Early Middle Korean seemingly provides further evidence
for ac- ‘to sit’ without a nasal, because it is spelled without a nasal in the
Kyeylim yusa (Yi 1964: 346). Yi Kimun makes a strong point that in the
Ceycwuto dialect some forms of ac- ‘to sit’ do not have a nasal in them.
But he mentions that a stem anc- is also attested. He believes the nasal to
be a result of influence from modern Standard Korean (1964: 347-348).
Although the possibility of dialect mixing certainly cannot be excluded, Yi
Kimun provides no justification for the Ceycwuto dialect’s borrowing of
the stem anc- from mainland Korean. Certain paradigmatic forms of the
verb given in Kim Chwunghoy et al. (1995: 156, 160) are reproduced
below:
Chart 4:
Some forms of the verb anc- ‘sit’ in Modern Korean and Ceycwuto
Modern Korean Southern Northern
Ceycwuto Ceycwuto
imperative: anc-ala acila, ancila acila
adverbial: anc-key acila, ankkey, ankkey, akkey,
akkey acila
hortative: anc-ca ancca ancca, acca
imperative: anc- acipse acipse
useyyo
infinitive: anc-a anc-a anc-a
causative: anc-hi- acci- ancci-, acci-
18———Koreo-Japonica

The case of language diffusion proposed by Yi Kimun would involve

more forms with nasals in Northern Ceycwuto than in Southern Ceycwuto,
because that is how language diffusion would work. As a matter of fact, we
have approximately the same amount of forms with nasals and without
nasals both in Southern and in Northern Ceycwuto, and that calls for an
alternative explanation. I return to this alternative explanation below, but
for the time being I discount the Ceycwuto forms of the verb ‘to sit’
without nasals as supporting evidence for nasal insertion.
There are three other significant problems with ‘nasal insertion’,
namely, that it is claimed to occur only in front of -c-, that the nature of the
‘inserted’ nasal cannot be predicted, and that insertion itself is completely
unpredictable (Ramsey 1978: 55-56). The random nature of ‘nasal
insertion’ is odd, and it is not clear why it occurs, e.g., in icey ~ incey
‘now’, but never in the verbal stem ic- ‘to forget’. I will attempt to
demonstrate below that ‘nasal insertion’ is not limited to the environment
V_c, but can also occur in other cases. Before doing so, I wish to discuss
Ramsey’s chart (1978: 54-55), which was used to demonstrate the
secondary nature of ‘nasal insertion’. This chart, with added data from
Northern Hamkyeng based on Kim (1986) and numbers that I use for
reference, is given below:

Chart 5:
Comparative chart of ‘nasal insertion’
MK post-MK S. Ham- N. Seoul gloss
kyeng Hamkyeng
(1) kòchó- — kamchwú- kamchwu- ~ kamchwu- ‘to hide’
~kònchwó- komchi- ~
~kòmchwó- komchwu-
(2) ka:chí kanchi kkáchi kachi ~ kacchi ~ kkachi ‘magpie’
kkachi
(3) kwòthí- kwonchi 5 konchí- kochi- ~ kothi- kochi- ‘to repair’

(4) nèchwúl nenchwul nengkhwúli nengwul ~ nenchwul 6 ‘vine’

nengkhwul ~
nechwul ~
necchwul
(5) tètí- teci- tencí- tenci- 7 tenci- ‘to throw’

5
Ramsey (1978: 54) has konchi-, but it is apparently a typographical error for kwonchi-,
since post-MK *konchi- does not exist.
6
Ramsey lists nengkhwul as the Seoul form, but it seems to be a mistake. The Seoul form is
nenchwul (Martin et al. 1968a: 332), and if nengkhwul is provided in a standard language
dictionary it is treated as a dialect form (Martin et al. 1968a: 336).
7
Ross King informed me that in Yukcin it is teti- (King, personal communication).
 Proto-Korean and Proto-Japonic Reconstructions———19

(6) màchí — mangchí mangchi ~ mangchi ‘hammer’

machi 8
(7) mèchwú- memcho- memchwú- memchwu- memchwu- ‘to stop’

(8) àncò- ~ àcò- — anc(u)- anc- anc- ‘to sit’

(9) yèncò- ~ — enc(u)- enc- enc- ‘to place’

yècò-
(10) áncók ~ — ancúk acuk ~ acik acik ‘still, yet’
àcík ~ àncík
(11) èchí enchi — — enchi ‘saddle
blanket’ 9
(12) ícèy — íncey ~ icey icey ~ incey ‘now’
ícey
(13) cànchí- ~ — — — cachi- ‘to turn
càchí- down’
(14) hòWózá ~ honca honcá habunca ~ honca ‘alone’
hòwózá ~ hobunca ~
hòwòá ~ honca
hwòncá
(15) — esthyengi ~ heychéyngi enchayngi ~ enchengi ‘harelip’
enchyengi enchengi ~
encheyng ~
heychayngi 10

Looking at the above chart one can immediately notice that the Modern
Seoul data practically always agree with Northern Hamkyeng, and they
agree in most cases with Southern Hamkyeng as far as the presence or
absence of a nasal is concerned. This presents a double oddity for ‘nasal
insertion’. First, if ‘nasal insertion’ is random, why is there a rather
consistent agreement of ‘nasal insertion’ and lack of insertion between
three dialects; that is, why is the picture not completely chaotic? Second,
for the language diffusion explanation there is an even greater problem
than was the case for Ceycwuto: Seoul and Northern Hamkyeng
demonstrate better agreement with each other than either does with
Southern Hamkyeng. Besides these two general problems with ‘nasal
insertion’, there are also specific problems with the examples in the chart.
(1) Let me note for the sake of further discussion that ‘to hide’ in
Middle Korean itself shows a variation between forms with and without a
nasal, while all other dialects show a consistent form with a nasal -m-. The
form kònchó- is attested from 1463, while the form kòchó- appears slightly

8
The form machi without a nasal is attested only in Sucin, while the form mangchi with a
nasal is found all over North Hamkyeng, including Sucin (Kim 1986: 199).
9
Ramsey gives the gloss ‘worth’ (1978: 55), but this is apparently a typographical error.
10
Attested only in three localities. The most frequent form is enchayngi, attested in eight
localities.
20———Koreo-Japonica

earlier in 1447, and the form kòmchó- is attested only in 1573 (Nam 1997:
232, 234, 241). The Ceycwuto dialect has forms with and without the nasal
-m-: komchwo-, komchwu-, and kochwo-, all of them attested throughout
the island (Hyen et al. 1995: 21, 73, 76).
(2) Post-MK kanchi ‘magpie’ comes very close to being a hapax
legomenon, as it is found only twice in the same text (Twusi cwung XIV:
21b and XVI: 37a). However, two of the Ceycwuto dialect forms also have
the medial -n- in this word: kanchi, kkachi, kkanchi (Hyen et al. 1995: 19).
It is difficult to explain Ceycwuto forms with -n- as result of language
diffusion from post–Middle Korean, especially the form with the initial
fortis consonant kkanchi, as it does not occur in post–Middle Korean texts.
(3) Post-MK kwonchi- ‘to repair’ is a hapax legomenon attested only in
one sico poem (LCT 1987: 69, Nam 1997: 108). Ramsey’s Southern
Hamkyeng data are based on only two informants, so there is the
possibility of this being an individual peculiarity. The word appears only
as kwocci- in Southern Ceycwuto and as kwocci-/kwochi- in Northern
Ceycwuto without a nasal (Kim Chwunghoy et al. 1995: 79-80). Therefore,
it cannot be used as a valid example of ‘nasal insertion’.
(4) MK nèchwúl, post-MK, Northern Hamkyeng, and Seoul nenchwul
‘vine’, with -ch- vs. -kh- in Southern Hamkyeng and at times in Northern
Hamkyeng, present a problem, but the Northern Hamkyeng form nengwul
with -ng- and no other consonant seems to support an original velar rather
than an affricate in this word. The fortition [t >] c > k is known in the
history of Korean, but it seems to be limited to the position in front of the
vowel /i/ in four words: tisay ‘tile’ > kiwa, cis ‘feather’ > kis, timchoy
‘kimchee’ > kimchi, soncoy ‘rather’ > *sonci > sangki (Yu 1964: 95-96).
The palatalization k > c therefore seems to be a more probable change,
especially since most other dialect forms show -kh- rather than -ch- (Choy
1978: 824-825). The Ceycwuto dialect again has doublets with and without
the nasal: nechwul, nenchwul (Hyen et al. 1995: 97).
(5) MK tètí- and post-MK teci- agree with Yukcin teti- in terms of the
absence of a nasal, but the agreement of forms with a nasal between Seoul,
other Northern Hamkyeng, and Southern Hamkyeng is again difficult to
disregard. Ceycwuto dialect as usual has doublets with and without the
nasal: tenci-, teci- (Hyen et al. 1995: 132).
(6) MK màchí ‘hammer’ lacks the nasal -ng-, but all dialect forms have
a nasal. The word is not attested in the Ceycwuto dialect.
(7) MK mèchwú- ‘to stop’ lacks -m-, but all dialect forms exhibit this
nasal. The word is not attested in the Ceycwuto dialect.
(8) As for Middle Korean forms àncò- ~ àcò- ‘to sit’ (Ramsey 1978:
54), the second one is a ghost not attested in Middle Korean texts. Also,
the first one is better presented as MK ànc-, since *àncò- is a possible
Proto-Korean reconstruction, not the actual Middle Korean form. Yi
Kimun cites Middle Korean forms àns-, às-, and àz- ‘to sit’, noting that all
are rare (1964: 342). The problem is not that they are just rare: each of the
last two forms, às- and àz-, has a unique attestation in Middle Korean. In
addition, both às- and àz- are attested only in preconsonantal position _C,
 Proto-Korean and Proto-Japonic Reconstructions———21

so even if one wants to discard the explanation of a scribal mistake, the

expected simplification of a cluster is quite plausible in this position. The
form àns- is attested much more frequently than the other two, but it is also
confined to the preconsonantal position _C (Nam 1997: 1023). On the
other hand, the form ànc- is amply attested in prevocalic position _V
starting from the earliest text written in Hankul: pulk-un say kul-ul mul-e
chimsil iph-ey anc-o-n-i ‘a red bird holding a text in its mouth sat on the
door of the bedroom’ (YP 7). All the dialects, except Ceycwuto, which
shows the variation between anc- and ac-, have the form with a nasal. The
form without a nasal is attested as EMK acakala (阿則家口羅) ‘sit!’
(Kyeylim #317) as noted by Yi Kimun (1964: 346), but we should not
forget that the vocabulary in Kyeylim yusa was compiled by a Chinese, so
the possibility of error remains. The evidence from Cosen kwan yek.e
(1421) cited by Yi Kimun (1964: 346) is unlikely to be admissible either,
because the transcription is obviously corrupt: 阿格刺 /akela/. It would be
a stretch of the imagination to suggest that it represents *a(n)ckela,
especially since the text was apparently not compiled by a native Korean
speaker. In any case, the Cosen kwan yek.e data predate the earliest Hankul
texts by only 26 years, and it is highly unlikely that this text can have any
precedence over the combined evidence from Middle Korean and the
dialects, even when taken together with the inconclusive evidence from the
Ceycwuto dialect.
(9) The case of MK yèncò- ~ yècò- ‘to place’, cited by Ramsey (1978:
54) is quite similar to the case of ‘to sit’. MK yèc-, defined by Yi Kimun as
‘very rare’ (1964: 343), is another hapax legomenon, occurring only in
Sekpo (XIII: 17). Although it does occur in prevocalic position in the
infinitive form yec-e, a hapax legomenon is a hapax legomenon, and the
loss of the letter {n} is quite easy to imagine. There is also another hapax
legomenon, yès-, found in preconsonantal position (Nam 1997: 1089) and
identical to às- ‘to sit’ discussed earlier. Other than these two unique
attestations, both yènc- (in prevocalic position _V) and yèns- (in
preconsonantal position _C) are amply attested in Middle Korean (Nam
1997: 1085). Absence of any hint from any dialect for the secondary nature
of -n-, including the Ceycwuto dialect, where the word in question is not
attested, allows us to put safely to rest Yi Kimun’s explanation of a
secondary nasal in this word.
(10) MK àncók ~ àcík ~ àncík ‘still, yet’. There is a variation in Middle
Korean, and a discrepancy between the modern dialects. The word is not
attested in the Ceycwuto dialect.
(11) MK èchí ‘saddle blanket’ contrasts with both post-MK and Seoul
enchi which has a nasal. No attestation is found in the Ceycwuto dialect.
(12) MK ícèy ‘now’ and alternating forms icey ~ incey in the modern
dialects are indeed strong supporting evidence for the secondary nature of
-n- in this word, since the etymology is quite transparent: i ‘this’ + cèk
‘time’ + i, diminutive suffix.
(13) Middle Korean alternating forms cànchí- ~ càchí- ‘to turn
down/over’ and MdK cachi- ‘id’. are unlikely to offer strong support for
22———Koreo-Japonica

nasal insertion, because the nasal -n- presents itself in one of the Middle
Korean forms. It is worth noting that the Middle Korean form without the
nasal is slightly younger (attested from 1482) than the form with the nasal
(attested from 1449) (Nam 1997: 1196, 1198). We can disregard the
singular attestation of càchí- in Sekpo III, since there is no reliable textual
transmission of this particular volume of the Sekpo sangcel.
(14) On MK hòWózá ‘alone’ and other related forms in dialects Ramsey
writes: “The c in ‘alone’ is also not original, and again, n appears only
after the change z > c occurred” (Ramsey 1978: 55). The fortition z > c
indeed took place in Korean; Ramsey makes a reference to Yi Kimun
(1987: 38), but there are no examples of this fortition in intervocalic
position. On the contrary, this fortition took place exclusively after nasal
-m- or -n-, and Yi Kimun provides four examples: MK mwomzwo ‘self’ >
post-MK mwomcwo, MK swonzwo ‘personally’ > post-MK swoncwo, pre-
MK *namzin ‘man’ > MK namcin, pre-MK *samzil ‘third day of the third
month’ > MdK samcil (Yi Kimun 1987: 38-39). This fact, along with the
dialect forms, which all have -n-, clearly demonstrates that -n- in this word
is original and that MK hòWózá ‘alone’ represents a unique development
with loss of the nasal.
(15) On post-MK esthyengi ~ enchyengi ‘harelip’ Ramsey cites Yi
Kimun deriving this noun from the Middle Korean verb èhù- ‘to split’ (?).
Although such a semantic shift is possible, there are several problems with
this explanation. First, post-MK esthyengi is another hapax legomenon.
Without it, there is no evidence for a form without a nasal. Second, as far
as I can tell, MK èhù- ‘to split’ (?) is a ghost, as it does not exist in either
Middle Korean or post-Middle Korean texts. There is MK èhí- ‘to cut with
a knife, to carve’, but the semantic change will be more questionable.
Third, since reliable evidence for the form esthyengi does not exist, the
derivation from MK èhí- will face several phonetic difficulties. Finally, I
think that post-MK enchyengi ‘harelip’ is likely to be a borrowing from
Mongolic, cf. WM emcügür ‘person with a short upper lip’ (Lessing 1995:
331). Even if I am mistaken as to the direction of the loan, the nasal is
undeniably there, since Mongolian does not have any ‘nasal insertion’.
I believe that only one of the aforementioned cases presents
uncontroversial evidence for ‘nasal insertion’, namely (12) icey ~ incey
‘now’. In addition, the case of pànchywò ‘plantain’, where the nasal is
obviously secondary (< Sino-Korean phàchyò 芭蕉) also cited by Ramsey
(1978: 55), should be added, but the majority of the other cases are either
controversial or offer strong evidence for the original nature of the nasal
and therefore for an *Nc cluster. Among those, (5) is controversial; (2), (3),
and (10) present ‘nasal insertion’, but not in the dialects where we would
expect it according to the nasal insertion theory; and there is simply not
enough evidence for case (11). On the other hand, I believe that nine
examples, numbers (1), (4), (6), (7), (8), (9), (13), (14), and (15), present
rather uncontroversial evidence for an original nasal -n- or -m-.
In addition to the list from Ramsey (1978), there are other examples of
-nc- and -mc- clusters in Middle Korean with no evidence for forms
 Proto-Korean and Proto-Japonic Reconstructions———23

without nasals. The following list is based on Nam (1997). Numbers in

parentheses refer to the pages in that dictionary: kyèncwú-/kyèncí- ‘to
compare’ (79), kòmcòk- ‘to budge, to stir’ (241), nàmcòkì ‘above, more
than’ (271), nèncùsì ‘secretly, quietly’ (282), mwòncyè ‘first of all’ (572),
mòncí- ‘to touch’ (625), sòncóy ‘rather’ (944), wùmcùkí- ‘to move’ (1131),
wùmchú-/wùmchì- ‘to shrink back, to crouch’ (1131). Among these nine
words, mwoncye ‘first of all’ is especially interesting in light of OJ maNtu
‘id’. < *mantu, which clearly points to the original nature of the nasal there,
whether it is a cognate, or (more likely) a loanword from some Korean-like
language like Paekche. Altogether we have eighteen examples where the
clusters -nc- or -mc- appear to be original.
Besides words with -mc- or -nc- clusters, combinations of a nasal with
consonants other than -c- exist as well. One significant difference from -
Nc- clusters is that the other -NC- clusters are not attested at the end of
verbal roots. The following list is also based on Nam (1997). I have been
very conservative in its selection: obvious loanwords from Chinese are
excluded, as are cases where one can suspect a morphological boundary
between N and C:
kèntí- ‘to save’ (62), kyèntúy- ‘to endure’ (78), kwùngkùl- ‘be empty’ (157),
kùntày ‘chard’ (182), nànhwó-/nònhwó- ‘to divide’ (265, 347), nànkyés
‘arguably’ (264), tòngkóy-/tànkì- ‘to pull’ (368, 489), tàmpwóy ‘marten,
sable’ (373), túmpwùkí ‘a kind of bird’ (464), màntòlàmí ‘cockscomb’
(534), mèngGéy ‘yoke’ (555), mwòngkóy- ‘to dawdle’ (576), mwùngkúy-
‘to lump’ (594), mòyngkól- ‘to make’ (638), pàntwó ~ pàntwóy ‘firefly’
(649), pàntók- ‘be clear’ (650),11 pèmkúl- ‘be entangled’ (674), pèmpúy-
‘to hamper’ (675), pèngGúl- ~ pèngkúl- (677-678), pwòntwòkí ‘chrysalis,
pupa’ (694), sèngkúy- ‘be sparse, thin’ (834), àmkól- ‘be healed’ (1027),
ènték ‘hill’ (1061), yèmkúl- ‘to ripen’ (1088), cyèmkú- ~ cyèmkúl- ‘to get
dark’ (1227), cùmkéy ‘big tree’ (1266), cùmkéy ‘distance’ (1266). 12
Altogether we have twenty-seven words here, and the following chart
presents statistics regarding the distribution of clusters:

Chart 6:
Frequency of NC clusters in Middle Korean
-p- -t- -c- -k- -G- -h-
-m- 3 — 6 7 — —
-n- — 8 11 2 1 1
-ng- — — 1 7 1 —

The most frequent clusters are: -nc- (11), -nt- (8), -mk- (7), and -ngk- (7),
-mc- (6), and -mp- (3). These statistics demonstrate that while clusters with

11
Cf. also the variants pàntók hò- and pèntúk hò- ‘be clear’ (Nam 1997: 650, 671) and an
adverbial derivation pàntókì ‘clearly’ (Nam 1997: 650).
12
It might be tempting to see a morphological boundary between a nasal and what appears
to be a ‘suffix’ *-kol-/-kul-. But there is no basis for this segmentation because the elements
preceding *-kol-/-kul- do not present themselves as either independent or bound
morphemes elsewhere.
24———Koreo-Japonica

nasal + -c- tend to be the most frequent (18 total), they are only marginally
more frequent than clusters with nasal + -k- (16 total) and only twice as
frequent as clusters with nasal + -t- (9 total). The presence of nasals before
consonants other than -c- makes the nasal insertion theory highly
questionable. The skewed distribution toward the palatal and velar
articulations of the second consonant in the existing NC clusters suggests
that we are dealing here with residue of the process of nasal deletion *NC
> C. The occasional cases of real ‘nasal insertion’ as in icey ~ incey ‘now’
can be explained as hypercorrections that developed alongside the process
of nasal deletion *NC > C. The hypercorrection affects only -Nc- clusters
because the surviving -Nc- clusters probably were the last affected by the
nasal deletion.
There is a good piece of historical evidence supporting the claim that
existing NC clusters in Middle Korean are remnants of the process of nasal
deletion *NC > C rather than the result of ‘nasal insertion’. This piece of
evidence comes from the kwukyel script of the Koryo period. In particular,
forms such as Middle Korean verbs in -toy almost always appear as -ntoy
in kwukyel. There are a few other similar cases where -n- (albeit probably a
modifier) was dropped, e.g., EMK ho-ken-ton > MK hoketun (Ross King,
personal communication). I believe such developments demonstrate the
general trend toward nasal deletion.
Finally, the alternation between -nc- and -c- forms in the Ceycwuto
dialect needs to be explained in light of the theory of nasal deletion that I
propose here. I believe that the best explanation for the Ceycwuto facts is
that they represent a drift in the same direction as in Central Korean. On
the mainland the innovation NC > C seemed to originate in the Central
Korean area around Seoul. As we have seen above, other dialects seem to
preserve NC clusters better than Middle Korean. However, Central Korean
sits on top of a Japonic substratum, namely, it is located in the same area as
the so-called Koguryo place names recorded in the Samkwuk saki (1145).
Almost all dialects of insular Japonic underwent development *NC[-voice]
> C[+voice], with the major exception of the Tōhoku and Kōchi dialects. It
appears on the basis of the remaining ‘Koguryo’ glosses that peninsular
Japonic went in the same direction, losing completely any traces of a nasal
*NC[-voice] > C[+voice/-voice], e.g., ‘Koguryo’ sabu ~ sapuy (沙伏、沙
非) ‘red’ < *saNpu (cf. MJ sabi < OJ *saNpï < PJ *sanpu-i- ‘rust’),
‘Koguryo’ osïgam (烏斯含) ‘hare’ < *osaNkam (cf. WOJ usaNkî, EOJ
wosaNki ‘hare’ < PJ *wosanki). Although the issue merits further
investigation, there are some indications that the Ceycwuto dialect also
might have a Japonic substratum. For example, the vulgar word for
‘mouth’ in this dialect is kwulley, as opposed to Common Korean akari.
The former is likely to be connected to Japonic *kutuy ~ *kutu- ‘mouth’.
Also, some Ceycwuto place names seem to be of Japonic origin. For
example, a place in Antekmyen in Ceycwuto called Kamsan (柿山,
‘Persimmon mountain’) has the old spelling 神山 ‘deity mountain’ (Pak
1988: 210). The first character 神 ‘deity’ does not have the reading /kam/
in Korean. The most likely explanation is that it reflects OJ kamï ‘deity’, or
 Proto-Korean and Proto-Japonic Reconstructions———25

to be more exact, some Japonic form that is cognate to OJ kamï. The old
name of Ceycwuto, Tamna (聃羅, 儋羅), is a contraction of still earlier
Tanmura (漱牟羅) (Kwen 1994: 167). While Tanmura is meaningless in
Korean, it has a transparent Japonic etymology: Tanmura < tani mura (谷
村) ‘valley village’ or, less probably, < tami mura (民村) ‘folk village,
people’s village’.
It is important to note that the clusters -lt-, -lth- also are quite rare, and
there is no -lc- cluster in Middle Korean. My hypothesis on the basis of
this skewed distribution is that Class 1 verbs and nouns with no lenition
did not have just any kind of cluster, but mostly *NC clusters, although
some could be *LT clusters where the second consonant is a dental. The
*NC clusters simplified into *C after lenition had occurred, so the
remaining obstruent was unaffected by lenition.
Unfortunately, there is very meager evidence from the dialects
supporting the above claim. Since the time depth of Proto-Korean is
probably quite shallow, *NC clusters were probably simplified for the
most part at the Proto-Korean stage, leaving no trace in the language
except the aforementioned skewed distribution. Below I consider some
Class 1 verbs ending in a single obstruent, along with their dialect
counterparts, that may offer some additional support:

Chart 7:
Evidence from Korean dialects for NC and LC clusters
Gloss MK Yengyang 13 Reconstruction
‘hard, strong’ kwùt- kkol-da *kwultu-
‘be late’ nùc- nïrəsə *nulcu-
‘to ripen’ nìk- — 14 *nilko/u-
‘to close’ tàt- — 15 *tanto-
16
‘wide/broad’ nèp- nəlbəsə *nelpu-

There is no similar evidence for nouns that have non-lenited obstruents.

However, there are some glimpses suggesting that this might be the case
for nouns as well. For example, MK ètúy ‘where’ (note that lenition does
not occur here according to the condition proposed by Martin [1996])
obviously comes from ènú ‘which’ and túy ‘place’, therefore *enu tuy >
*entuy > ètúy. MK yètúlp ‘eight’, which also does not show lenition,
contrary to Martin’s rule, is probably a compound, consisting of yélh ‘ten’
and *twupul ‘two’, where a cluster blocked lenition before being
simplified: *yelh+twupul > *yel-twulp > yètúlp. The same can be said for
MK yètún ‘eighty’, which also historically includes yètúlp ‘eight’.

13
A dialect in Kyengsang Pukto province.
14
Cf. Kokseng (Cennam) ilgïnda (Choy 1978: 1430).
15
Cf. Cinsang (Cennam) tan-da (Choy 1978: 1323).
16
Note that the -lp- cluster in this MK word behaves in the same manner as a *-Np- cluster
would, exhibiting the loss of a sonorant in the V_C position. It is quite clear that MK nèp-
‘be wide’ is more innovative than MdK nelp- ‘wide’.
26———Koreo-Japonica

1.1.3.3.2 Where did the -n stems go?

As mentioned earlier, from the standpoint of distribution it appears strange
that there are eighteen stems ending with -m in Middle Korean, but only
five with -n, three of which, as Martin pointed out, are probably derived
from nouns (1996: 5). In addition, one may ask the question: why there is
not a single stem ending in -ng in the language?17
I believe that some of the *-n and *-ng stems may hide among the stems
belonging to accent Classes 3 and 4. First, all the verbs in these classes are
of the monosyllabic CV- type. Second, all the verbs in these classes exhibit
extremely irregular accentuation (Kim Wancin 1973: 57-61, Ramsey 1991:
232). The stems of these two classes in Middle Korean 18 could show either
H or L pitch depending on the following suffixes (Kim Wancin 1973: 57-
58, Martin 1995: 143). Kim Wancin suggested that these stems were
originally low-pitched, but there was a final element that he denoted as λ,
equaling it to l plus an unknown element X, that caused the pitch to shift to
high before certain suffixes (1973: 58). However, stems with -lC- clusters
are well attested. Although some dialect evidence that I cite below points
to [l] or [r] as a final consonant, from the point of view of a gap in
distribution, it is exactly -n and -ng stems that are missing. The stems of
monosyllabic verbs have L pitch before the suffixes -o/ulʔ, -o/uli, -o/un,
-o/uni, -o/umye, -kwo, -ta, and -key, but have H pitch before suffixes
-o/usi-, -no-, -zoW-, -te-, -a/e (Kim Wancin 1973: 58). At the present stage
of our knowledge of Korean historical morphology, such distribution
makes it difficult to find any segmental phonological conditioning of the
pitch changes, but there may be morphophonological conditioning: all the
suffixes affixed to stems that have H pitch are word non-final suffixes that
cannot conclude a verbal paradigmatic form, with the exception of the
infinitive suffix -a/-e. On the other hand, all suffixes affixed to stems that
have L pitch are word-final suffixes that can conclude a verbal
paradigmatic form.
Another puzzling feature of these verbs is that they do not exhibit
lenition in forms like indicative ka-ta ‘goes’ or gerund ka-kwo ‘goes and’
into *ka-la or *ka-Gwo, as would be expected according to the revised
hypothesis on lenition. The hypothesis maintains that simple stops lenite
(without juncture) in any intervocalic position, if the stems were really of
the CV- type in Proto-Korean. Since lenition does not happen, we can
expect that it was blocked in much the same way as in the stems of Class 1
verbs; in other words, there probably was a final consonant in the stem
which combined with the -ta or -kwo to form a consonantal cluster that
blocked lenition. From the gaps in distribution, it is likely that the
consonant in question was -n, as final -m and -l stems are attested in Class

17
Martin and Whitman have pointed out that *ng might not be a phoneme in Proto-Korean
(personal communication). They believe that MK ng is from earlier *nk. I agree that the
Middle Korean data indeed point in this direction, but some of the dialect data discussed
below might warrant a reexamination of the phonemic status of *ng in Proto-Korean.
18
The division between these two classes is based on modern dialect data, as in MK itself
they have similar accentual behavior (Ramsey 1991: 233).
 Proto-Korean and Proto-Japonic Reconstructions———27

2a verbs. The apparent exception is the verb ho- ~ hoy- ‘to do’ in this class,
which ends in a glide [y], a form that should not block the lenition of -kwo.
The Middle Korean gerund form is, however, ho-kwo, not the expected
hoy-Gwo. The absence of lenition can be attributed to the fact that lenition
did not happen at verbal morpheme boundaries after the minimal vowels o
and u. The same rule would explain why there is no lenition of *-ta > *-la
and *-kwo > *-Gwo in Proto-Korean after the stems of Classes 1, 2b, and 6,
which also end in a minimal vowel. None of the verb stems of Classes 3
and 4 ends in a minimal vowel, so we would expect lenition here, unless it
was somehow blocked. This brings us back to the idea of a final consonant
in the stems of these classes. A final consonant, incidentally, explains the
“puzzle of MK copula í-lá” (Martin 1998: 9), the only verb that has one of
its stems ending with a non-minimal vowel, after which we have to expect
the lenition of i-la < i-ta. 19
Let us turn our attention to some features of internal reconstruction that
offer some evidence in favor of the hypothesis that the stems of these verbs
once had a CVC- or CVCV- structure. At least one of the verbs in Class 3,
ho- ‘to do’, goes back to the stem hoy-, which ended in the glide [y]. 20 The
infinitive form in -a shows exactly this stem: hóy-á ‘does and..’. The verb
wo- ‘come’, which also belongs to Class 3, has an unexpected imperative
form wónàlá, instead of the expected *wala. Martin treats this -na- as a
suppletive form of the effective -ke/a- (1995: 146); however, it remains
unclear why this suppletive -na- never occurs in the imperatives of other
verbs. Furthermore, as Martin notes “the effective infinitive [-ke/a- -- A.V.]
is indeed different from the regular infinitive, since the effective infinitive
does not occur at the end of the clause or before most auxiliaries or before
the particle two” (1995: 144). However, Martin does point out an exception
to the last rule: there is the form wónàtwó ‘though one comes’ (Twusi 25:
23a) (1995: 150, note 4), again from the verb wo- ‘to come’. I suppose that
wón-à-twó really incorporates the regular infinitive -a, which can be
expected before the particle two. Also, as Martin noted, several forms of
the verb wo- ‘to come’ have the regular -ke/a- form for the effective (1995:
146, 150 note 5). But would a verb have both suppletive and regular forms
of the effective? Thus, I do not think that the imperative form wónàlá
‘come’ includes the suppletive form of the effective infinitive -ke/a-, as
happens in verbs ka- ‘to go’ and is- ‘to be’, which have ka-ke-la and is-ke-
la respectively. I believe that here we are dealing with an alternation of the
stem wo- with the stem won-, which is probably the original. I reconstruct
this verb as *won- ‘to come’.
In the Ceycwuto dialect, the verb wo- ‘to come’ shows an interesting
irregularity: although in recent materials there are infinitive forms wa <

19
One of the scenarios Martin proposed was that the copula stem was *il-, and, therefore
MK i-la < *ilta (1998: 11). It seems to me that there might not be a need for this scenario if
the above explanation is accepted: lenition of t > l is to be expected in intervocalic position.
20
Martin speculates that this hoy- is a metathesis from the earlier *hyo- (1995: 141). He
might be right, but his reconstruction of *hyo- is best supported by external evidence, and
there is little, if any, internal evidence that can prove this metathesis.
28———Koreo-Japonica

*wo-a (Northern Ceycwuto) and wan (Southern Ceycwuto), since both

Northern and Southern Ceycwuto share the irregular forms wora (Northern)
and woran (Southern) (Kim Chwunghoy 1995: 179), there is reason to
believe that the first set appeared under the influence of Standard Korean.
The older research on Ceycwu morphology describes only the form wora
(Yi Swungnyeng 1978: 69ff.). This form can be analyzed only as wor-a
‘come-INF’, with a final consonant in the stem. Why Ceycwu has -l rather
than -n in the stem is unclear, but since the dialect has regular -l verbs like
Middle Korean in Classes 1, 2, and 6, we can think of a possible
contamination of an older -n stem with -l stems.
Although the verb ka- ‘to go’, which also belongs to Class 3, does not
show any of the irregularities typical of wo- ‘to come’, there is a puzzling
hortative form in the Ceycwuto dialect: in addition to the expected kaca
and kakey ‘let us go’ we also find kəlla in Southern Ceycwuto and killa in
Northern Ceycwuto (Kim Chwunghoy 1995: 160). How are we to explain
these forms? In both cases -la looks like an imperative marker. But why is
-l- doubled? The possible sources should be *-nl- or *-ll-; therefore the
underlying forms are probably *kan-la or *kal-la. However, since there are
verbs with stems that end in -l in Middle Korean and in Ceycwuto, the
reconstruction of another -n stem here seems likely, based on a gap in the
distribution.
MK i- ‘to carry on the head‘, which belongs to Class 4, has the
infinitive form iŋ-ən in Southern Ceycwuto (Kim Chwunghoy 1995:
178).21 Thus, it appears that this verb once ended in a nasal.
MK twu- ‘to put’ (belonging to Class 3) in the Uylyeng subdialect of
Kyengsang Namto has the presumptive form tol-ci, the infinitive tól-a, and
the past tól-atta (Kim Chwunghoy 1993: 348). Again, there is some
evidence for a final consonant in this stem.
It must be noted that the available data on the paradigmatic forms of
verbs belonging to Class 3 and Class 4, and especially the Ceycwuto data,
are quite meager at present. Most of them come from the multivolume
edition Hankwuk pangen calyo cip [Collection of materials on Korean
dialects], which includes a volume on the divergent and, therefore,
important Ceycwu dialect, but it is the least studied dialect in South Korea.
No similar information is available for North Korean dialects, where we
have mostly word lists, included, e.g., in Kim Pyengcey’s dialect
dictionary (1980). Practically no paradigmatic forms are available. In
addition, some of these Middle Korean verbs, especially those belonging to
Class 4, do not have counterparts in the modern dialects. In light of these
facts, I believe that the above evidence for original final consonants is
promising, although not conclusive.

21
That form was given as a response to elicitation of the imperative form, which is
incorrect, as noted by compilers. Clearly, this is an infinitive form, as Southern Ceycwuto
has infinitives in -an/-en.
 Proto-Korean and Proto-Japonic Reconstructions———29

1.1.3.3.3 Why are there no -G stems?

The only verbal stems ending in -G in Middle Korean are those with
clusters -lG-, -zG-, and -mG- (there is only one stem with the last). Martin,
in my opinion, correctly reconstructs those as ending in *...lo/uk-,
*...so/uk-, and *...mo/uk- in Proto-Korean (1996: 28-30). Again, from the
standpoint of distribution it is strange that there are no stems in Middle
Korean ending just in -G. The solution which assumes that the stems are
not attested because there was no special orthographic device in Middle
Korean for G is not acceptable, because traces of G alternating with k
would be preserved in the modern standard language and dialects. A
question that arises in connection with this strange distribution of G in
stems is: why do -p- and -t- stems in Middle Korean lenite, while -k- stems
do not. It is even stranger, given that there is lenition of -k- > -G- in
suffixes, parallel to the /p/ and /t/ lenition in suffixes. Why are verbal stems
the exception?
It seems reasonable to suggest that non-leniting -k stems have the same
origin as non-leniting -p and -t stems, namely, that they go back to *Ck
clusters, the majority of which are probably *Nk clusters. Some internal
evidence points in this direction, for example, such near-doublets as MK
tòngkóy- ‘to pull’ and EMdK thwoki- ‘to pull strings’, MdK singkeW- ‘is
not properly salted’ (not attested in Middle Korean, but with possible
etymology: singk- *‘salty’ + ep[s]- ‘does not exist’) and MK swòkwòm
‘salt’. The problem with this interpretation is identifying what has become
of the genuine PK *-k stems in Middle Korean.
I think that the genuine PK *-k stems became -h stems in Middle
Korean, that is, the original Proto-Korean intervocalic *-k- lenited into -h-.
Below, I present some internal evidence in support of this claim.
Graphemically MK h and G are almost in complementary distribution:
G occurs only after /l/, /z/, /y/, and /i/, but never in initial or final position,
or before a consonant. On the other hand, MK h is attested in initial and
final positions, in intervocalic position, between a vowel and a following
consonant, and after /l/.
Second, there is a close connection between MK h and G, as
demonstrated by various allomorphs of the causative: -Gi-, -hi-, -ki-, -hwo-,
-Gwo-, -Gwu-: sal- ‘to live’, sal-Gi- ‘to make one live’; anc- ‘to sit’, anc-
hi- ‘to make one sit’; wolm- ‘to move (intr.)’, wolm-ki- ‘to move (tr.)’; nat-
‘to appear’, nat-hwo- ‘to make one appear’; noli- ‘to descend’, noli-Gwo-
‘to make one descend’; huli- ‘to be muddy’, huli-Gwu- ‘to make muddy’
(Yi Swungnyeng 1961: 333-335).
A problem that needs to be dealt with in conjunction with this
interpretation is that there are both -lh and -lG stems in Middle Korean.
However, -lh stems are found predominantly in Class 1, and -lG stems are
all confined to Class 8; that is, the verbs with -lh and -lG stems exhibit
different accentuation patterns, indicating that there probably was some
segmental difference between the two. What that difference was exactly is
difficult to determine at the present time. Besides internal evidence, we can
use other types of evidence to substantiate the claim further.
30———Koreo-Japonica

1.1.3.3.4 Typological evidence

The distribution of medial obstruents in Middle Korean bears a striking
resemblance to the situation in the Amur dialect of the Nivx language. The
Nivx language is famous for its system of consonant alternations, which
can be presented as follows:

Chart 8:
Consonant alternations in Nivx
p~v~b t~r~d č~z~ĵ k~γ~g q~R~G

ph~f th~ř čh~s kh~x qh~χ 22

As can be seen, unaspirated voiceless stops alternate with voiced

fricatives and voiced stops, while aspirated voiceless stops alternate with
voiceless fricatives. These alternations, of course, represent typical lenition.
The latter alternation should be of no concern to us, since Middle Korean
voiceless aspirates do not lenite. The conditions for alternation are as
follows:
(1) Initial stops alternate with fricatives after stops, vowels, or [j].
(2) Initial unaspirated voiceless stops alternate with voiced stops after
sonorants [l], [m], [n], [ń], and [ŋ]. Note that this alternation is allophonic,
since voiced stops in Nivx are not phonemes, and never occur in initial
position. This kind of distribution closely resembles Middle Korean:
unaspirated voiceless stops occur in initial or final position, but in medial
position we have either voiced fricatives, or phonetically voiced stops, as
in Modern Korean. I follow Martin in his assumption that Middle Korean
medial stops were phonetically voiced (1996: 45).
(3) Initial fricatives alternate with stops after fricatives.
These conditions for alternations in Nivx can be illustrated by the
following example: the allative case marker has allomorphs -toχ/-roχ/-doχ,
with the following usage: liγs-toχ ‘to the wolf’, čho-roχ ‘to the fish’, ytyk-
roχ ‘to the father’, kaŋ-doχ ‘to the dog’.
It has long been noted that transitive verbs in Nivx have fricative initials,
while intransitives have initial stops, e.g., čo- ‘to be bent’, zo- ‘to bend (it)’,
čhoγ- ‘to melt’, soγu- ‘to melt (it)’. Jakobson demonstrated that the initial
fricatives in transitive pairs originated as a result of the loss of initial object
prefixes i- or e- (1971: 272-274). This led to the natural conclusion that
intervocalic stops in Nivx are lenited to fricatives and that therefore
fricatives are secondary. (Initial fricatives are found only in transitive verbs,
exceptions are minimal [Burykin 1987: 187].)
In the Amur dialect of Nivx there is a contrast in medial position
between fricatives and stops (phonetically voiced). If we exclude voiceless

22
Those interested in a detailed description of Nivx consonantal alternations should refer to
Panfilov (1962, 1968), Jakobson (1971), Burykin (1987), Gruzdeva (1997).
 Proto-Korean and Proto-Japonic Reconstructions———31

fricatives lenited from voiceless aspirates from this opposition, we get

exactly the same situation as in Middle Korean: a contrast between voiced
fricatives and phonetically voiced stops. We already know that in Nivx,
medial voiced fricatives came from voiceless stops, but where did the
medial voiced stops in Nivx come from? In contrast to Korean, Nivx
turned out to be more fortunate in having a more archaic Eastern Sakhalin
dialect that sheds light on its origin. Consider the following chart:

Chart 9:
NC clusters in Nivx dialects
Gloss Amur Eastern Sakhalin Reconstruction
dialect dialect 23
‘ladle’ qob qomb *komp[V]
‘pepper’ jabar jambar *jampatV
‘baby’ ojdam ojndam *ojntam
‘barrel’ sidux sindux *sintukV 24
‘drink’ rad- rand- *e-tant-
‘here’ tug tuŋg *tuŋk

It is apparent that the medial stops in Nivx came from *NC clusters that
prevented the lenition of stops into voiced fricatives, but at the same time
the nasality of the first consonant in a cluster probably contributed to the
voiced phonetics of the medial stops. Thus, essentially we have in Nivx a
well-attested example of the schema I suggested for Korean.
Additional typological support for the origin of Middle Korean voiced
fricatives from the lenition of stops and for Middle Korean medial non-
leniting obstruents’ origin from *NC clusters can be drawn from the
history of Japanese and Manchu. Although there is no dental or velar
lenition in Japanese, there is a well-attested labial lenition, with
intervocalic *-p- leniting to voiced fricative -w- and then to -Ø- before the
vowels [i], [e], [o], and [u], and an intervocalic *-np- cluster resulting in
voiced -b- in the majority of modern dialects, e.g., PJ *sapa ‘marsh’ > MdJ
sawa, but PJ *sanpa ‘mackerel’ > MdJ saba. There is an apparent velar
lenition in Manchu, e.g., Ma. uhu- ‘to wrap’ < PT *uku-, Ma. ahūn ‘elder
brother’ < PT *aka, but Ma. akū ‘not’ < PT *anaka, Ma. dosi-ka ‘entered’
< Proto-Manchu *dosin-ka (Vovin 1997b: 273-274). The postvocalic loss
of *-n- in Manchu is also confirmed by other examples, e.g., Ma. nicuhe
‘pearl’, cf. Jin period Jurchen nincuhe ‘id’. 25, possibly borrowed from Old
Turkic jincüg ‘pearl’.

1.1.3.3.5 Conclusion
The revised version of the lenition theory yields several important results:

23
The data on Eastern Sakhalin dialect are from Kreinovich (1972).
24
An apparent loanword from Ainu sintoko ‘container’, ‘barrel’, ‘big water pot’.
25
Marc Miyake provided me with the Jin period Jurchen form (personal communication).
32———Koreo-Japonica

(a) it better explains cases where there is no lenition, although expected

according to Martin’s (1996) ... ˚Có/ú rule;
(b) it demonstrates that the *NC clusters (the absence of which creates an
asymmetry in the Middle Korean and Proto-Korean consonantal systems)
originally underlie so-called regular verbs that have no lenition;
(c) it explains to a certain extent the irregularity of Class 3 and 4 verbs,
treating them as verbs that originally ended in a consonant;
(d) it permits a certain reconciliation of the lenition and voiced obstruent
theories. Although it disallows the reconstruction of voiced obstruents for
Proto-Korean, it maintains that the reason for lenition was rooted in the
consonantal structure of stems, and not connected to surrounding vowels
and their accent. Thus, I reconstruct PK *NC clusters for Middle Korean
non-leniting stops.

1.2 RECENT ADVANCES IN PROTO-JAPONIC

RECONSTRUCTION

1.2.1 Six or seven, not four vowels in Proto-Japonic

One of the few areas where Martin’s seminal 1987 reconstruction of Proto-
Japonic now seems to be superseded is the reconstruction of Proto-Japonic
vocalism. Namely, there is now ample evidence that Proto-Japonic had two
extra mid vowels *e and *o in addition to *a, *i, *ə, and *u, reconstructed
by Martin (1987: 67), following the lead of Ōno Susumu (1957: 162),
although in one case Martin seems to admit the reality of PJ *e, at least in
the case of PJ *me ‘water’ (1987: 483). The evidence for reconstructing
the Proto-Japonic mid vowels *e and *o is based on different kinds of data,
both comparative and philological. The most important piece of evidence
comes from comparative Ryukyuan and the reconstruction of Proto-
Ryukyuan. This evidence was first brought to light by Hattori (1976, 1978-
79), and further refined by Thorpe in his doctoral dissertation (1983). Both
linguists noticed that in some cases the Eastern Old Japanese data also
preserve primary PJ *e and *o, which were raised to *i and *u in Western
Old Japanese. The failure of others to notice this important fact can be
attributed to the general over-reliance on Western Old Japanese and
Middle Japanese data in Japanese historical linguistics. However, both
Western Old Japanese and Middle Japanese belong to Central Japanese,
which is just one of many Japonic dialectal groups. The observations and
reconstructions by Hattori and Thorpe concerning reflexes of PJ *e and *o
in Ryukyuan and Eastern Old Japanese were further developed and refined
by Serafim (1985, 1999a, 1999b), Whitman (2001), and Hino (2003).
Recently Miyake (2003b) provided one more important piece of
philological evidence, convincingly demonstrating that even within pre–
Western Old Japanese the raising of *e and *o to *i and *u took place
sometime before the Suiko period (592-628).
Since the issues crucial for the reconstruction of PJ *e and *o are
discussed in the aforementioned literature in great detail, I offer only a
short synopsis here. First, let us look at the evidence for PJ *e. It is
 Proto-Korean and Proto-Japonic Reconstructions———33

supported by a number of dialects in the Amami area in the north and the
Sakishima area in the south that faithfully preserve the distinction between
PJ *i and *e in postconsonantal position. 26 Even some Central Ryukyuan
dialects preserve the distinction after certain consonants, e.g., in Iejima PR
*me > mi, but PR *mi > ni. Only extreme Northern Amami dialects such
as Sani and Kikaijima seem to preserve the distinction in the absolute
initial position. The reflexes of PJ *i and *e in Western Old Japanese,
Proto-Ryukyuan, Amami, and Miyako are summarized in the chart.

Chart 10:
Reflexes of PJ *i and *e
PJ WOJ PR Amami Miyako
*i i *i i ï
*e i *e e, ï i

The following chart provides some examples of words with PJ *i and *e.
Chart 11:
Examples of PJ *i and *e
PJ WOJ PR Amami Miyako gloss
*pikar- pîkar- *pikari hikyayuN pïkaï ‘glitter’ (v.)
*mentu mîNtu *mezu mïzï mizï ‘water’
*piru pîru *piru hiru pïsuma ‘day’
*peru pîru *peru hïru piï ‘garlic’
*inu inu *inu iN iN ‘dog’
*entu- iNtu- *ezu ïdï, uduru Nzi ‘wh-’

PJ *o is supported by evidence from practically all Ryukyuan dialects

except those located in the extreme north of Amami island. For example,
PR *u is reflected as /u/ or /ï/ in Amami and Miyako and as /u/ or /i/ in
Okinawa; meanwhile PR *o is reflected as /u/ or /o/ in Amami, and as /u/
in Okinawa and Miyako (Thorpe 1983: 32). The reflexes of PJ *u and *o
in Western Old Japanese, Proto-Ryukyuan, Amami, Miyako, and Shuri are
summarized in the chart.

Chart 12:
Reflexes of PJ *u and *o
PJ WOJ PR Amami Miyako Shuri
*u u *u u, ï u, ï, i u, i
*o u, -ô *o u, o u u

The next chart provides some examples of words with PJ *u and *o.

26
The developments of *e and *o in Ryukyuan dialects are quite complex, exhibiting
various reflexes depending on environment in most dialects. My charts present the reflexes
in a simplified way. For details see Thorpe (1983).
34———Koreo-Japonica

Chart 13:
Examples of PJ *u and *o
PJ WOJ PR Amami Miyako gloss
*munay mune~muna- *mune munï Nni 27 ‘breast’
*moko mukô *moko muho 28 ‘bridegroom’
*kuray- MJ kure- 29 *kure kurïyuN fiiï ‘give’
*tuno tunô *tuno cïno cïnu ‘horn’
*tukor- tukuru *tukori cïkururi cïhuï ‘make’ (v.)
*yoru yôru *yUru yuru yuï ‘night’
*sirosi sirusi *sirosi sirusi sïrusï ‘mark, sign’

Next is a chart showing the agreement between the Eastern Old

Japanese and Proto-Ryukyuan data supporting the reconstruction of PJ *e
and *o.

Chart 14:
Examples of PJ *e and *o in Eastern Old Japanese and Proto-
Ryukyuan
PJ WOJ EOJ PR gloss
*esu isi, isu, isô osi, osu *esi ‘stone, rock’
*eki ikî oki *eki ‘breath’
*ye- yö- ye- *ye- ‘good’
*-o -u -o *-o attributive ending
*yoki yukî yoki *yoki ‘snow’
*noonsi niNsi noNsi *noozi ‘rainbow’
*mayo mayô mayo *mayU ‘brow’

I have discovered one more piece of evidence supporting the

reconstruction of PJ *e and *o, namely the old Japanese loanwords in
Proto-Ainu. It is quite apparent on geographical and phonological
grounds 30 that these words were borrowed from some old Tōhoku
dialect(s), adjacent to the Ainu homeland in the northern Honshū. Here are
the data comparing Proto-Japonic reconstruction with Old Japanese
loanwords in Proto-Ainu: 31

27
Also Nmi < Nni < *mïni.
28
Also moho.
29
Probably a semantic extension from OJ kure- ‘to go down, to set (of the sun)’.
30
It is possible to differentiate Old Japanese loanwords in Ainu from recent loanwords on
the basis of their phonology and distribution. For example, phonologically they exhibit OJ
/p/, and not /h/ (e.g., PA *pone ‘bone’ < OJ pone, cf. Obihiro Ainu huton < MdJ futon
‘futon’) and they reflect pre-Western Old Japanese diphthongs, e.g., PJ *pasuy > OJ pasi,
PA *pasuy ‘chopsticks’). Old loanwords also occur in all or almost all Ainu dialects,
including Sakhalin dialects, so a reconstruction of the PA archetype is possible. Recent
loanwords tend to occur in only one or two dialects.
31
On the basis of the list of Japanese loanwords in Ainu provided in Vovin (1990).
 Proto-Korean and Proto-Japonic Reconstructions———35

Chart 15:
Evidence for pre-OJ *e and *o in loanwords in Proto-Ainu
PJ WOJ PA gloss
*emo MJ imo *emo ‘potato’
*sirosi sirusi *sirosi 32 ‘mark, sign’
*taro MJ taru *ontaro 33 ‘barrel’

Thus, the conclusion that Proto-Japonic had six, not four, vowels seems
to be inescapable, since it is supported by several independent pieces of
evidence: Proto-Ryukyuan, Eastern Old Japanese, pre-Suiko period pre-
Old Japanese, and Tōhoku Old Japanese as reflected in Proto-Ainu. There
are other proposals that suggest reconstruction of an even greater number
of vowels. Hattori proposed to reconstruct a seventh Proto-Japonic vowel
*ü on the basis of the vocalic correspondence WOJ ï~ö : PR *i (Hattori
1978-79: 19). However, the evidence is limited to two examples, and
neither is exactly perfect. The examples are: WOJ pï ~ pö- ‘fire’ vs. PR *pi
‘id’. and WOJ mïna ‘all’ vs. OR mina ‘id’. OR mina is attested only in the
Haytong ceykwukki (1501), but the main problem lies on the Japanese side:
WOJ mïna is likely to be an old plural form in -na of WOJ morö ‘all’ (for
details see Vovin [2005a: 338-339]). The phonological history of the word
for ‘fire’ may be too complex: Eastern Old Japanese has a hapax
legomenon pu ‘fire’, with an irregular vowel correspondence of WOJ ï~ö :
EOJ u, so it seems too farfetched to base a reconstruction of an extra Proto-
Japonic vowel on one word with obscure vocalism.
Serafim’s (1999a) proposal to reconstruct another Proto-Japonic vowel
*O on the basis of the correspondence of WOJ ô to PR *o, e.g., PJ *mokO
‘bridegroom’ > WOJ mukô, PR *moko, seems to be better explained along
the lines suggested by Hayata (1998), that WOJ ô in word-final position
may just reflect PJ *o.
More recently, Frellesvig and Whitman (2003) made an interesting
observation that WOJ ö in combination with a following i results in two
different types of contractions: (a) opo (< *öpö) ‘big’ + isi ‘stone’ > opïsi
‘big stone’; (b) tönö ‘pavilion’ + iri ‘entering’ > töneri (< *tonëri)
‘retainer’, wo ‘small’ + inu ‘dog’ > wenu ‘puppy’. They reconstruct two
different Proto-Japonic vowels on this basis: in the case of (a) the vowel is
*i (*ipi ‘big’), but in the case of (b) the vowel is *ə (*tənə, *wə)
(Frellesvig and Whitman 2003: 4). They may be right, but for the time
being I remain skeptical and do not adopt this reconstruction for the
following reasons: first, the distribution of the newly proposed vowel *i is
extremely limited; second, there is no supporting evidence for this vowel
either in Ryukyuan, or in Eastern Old Japanese. Third, and most
importantly, opïsi ‘big stone’ is likely to be a ghost. It appears only once in
the Western Old Japanese corpus, spelled as 意斐志 /opïsi/ in KK 13. A
variant of KK 13 is attested as NK 8, where the sequence in question is

32
Note PR *sirosi ‘id’. in the preceding chart above.
33
Probably from Old Tōhoku *won-taro ‘small+barrel’.
36———Koreo-Japonica

written as 於費異之. While the overwhelming majority of modern

Japanese philologists read this sequence as opo isi ‘big stone’, there are
two considerable problems. First, the character 費 occurs as a phonogram
only once in Western Old Japanese and just in this sequence. Second, it is
hard to imagine that it would transcribe WOJ /po/, as the LMC phonetic
value of this character is /phjïj/, which should indicate *opï isi, and not
*opo isi. One should not forget, therefore, the dissenting voice of
Omodaka et al. who treated 於費異之 as opï isi ‘growing stone’ and not as
‘big stone’ (1967: 154). This also agrees well with Japanese mythology,
where stones are considered to be able to grow like trees and other plants.
Thus, the only example that Frellesvig and Whitman are able to cite in
favor of their speculative *ï is not reliable. Moreover, a projection of a
vowel that can be reconstructed only in Western Old Japanese has a more
than dubious proto-Japonic provenance, according to the definition of what
can be considered ‘Proto-Japonic’ provided in my introduction. It must be
mentioned that Frellesvig and Whitman try to further support their
reconstruction with external evidence from Korean, but I firmly believe
that any kind of external evidence cannot be used for an internal
reconstruction. This is even truer in the case when a genetic relationship
between languages remains hypothetical, as is the case with Korean and
Japonic.
Therefore, I accept the six-vowel reconstruction of Proto-Japonic
vocalism.

Chart 16:
Proto-Japonic vocalism
*i *u
*e *ə *o
*a
1.2.2 Another look at PJ *b- and *d-
It is assumed by most linguists working on the reconstruction of Proto-
Japonic today that OJ /w/ and /y/ go back to PJ *b and *d, although there
are several dissenting voices as well. I review the evidence for both
arguments below.

1.2.2.1 Sakishima initial b-: an innovation or an archaism?

In sharp contrast to the Northern and Central Ryukyuan dialects and an
overwhelming majority of Japanese dialects that have initial w-, the
Southern Ryukyuan dialects have initial b-. I have to add that two non-
adjacent regions in coastal Toyama also show or used to show initial b-,
although only before /a/ (see detailed discussion of the Toyama data
below). There are opposing opinions regarding the interpretation of this
correspondence and its subsequent Proto-Japonic reconstruction. The
majority of Western linguists seem to consider the lenition *b- > w- as the
likeliest explanation (Whitman 1985: 15-18, Martin 1987: 20, Serafim,
personal communication). Meanwhile, a more limited group of Western
 Proto-Korean and Proto-Japonic Reconstructions———37

scholars (Thorpe 1983: 105, Miyake, personal communication) and the

majority of Japanese scholars insist that the fortition *w- > b- took place in
Southern Ryukyuan.
Arguments for the lenition are predominantly typological: lenition is
easier to explain than fortition. The key argument in the opposition is
structural: if the lenition really took place in the rest of Japonic, we still
would find the initial *b- before /u/ in Southern Ryukyuan, but the syllable
*bu is nowhere in sight (Thorpe 1983: 105). There is also other evidence
suggesting that the fortition took place in Southern Ryukyuan, and not the
other way around, since initial w- in Chinese loans is replaced by b- as well,
cf. Miyako boo ‘king’ (child’s name) < Chin. wang ‘id’. (Miyako online
dictionary), Miyako tśaban ‘teacup’ < Chin. cha-wan ‘id’. (Karimata 1999:
56). However, I agree with Whitman that the case for PJ *b- based on
Southern Ryukyuan b-, and corresponding both to Northern and Central
Ryukyuan w- and to w- in most Japanese dialects, is stronger than the case
for PJ *d- based solely on the Yonaguni data (see below). Whitman argued
that PJ *bu in Miyako is reflected as /Qv/ on the basis of Miyako Ōura and
Uechi Qv and Miyako Nakasuji Qvi(i) ‘to sell’, corresponding to OJ ur-
‘id’. (cf. OJ aNpura ~ Miyako Nakasuzi aQva ‘oil’) (Whitman 1985: 17).
Since PJ *mb > PR *b, it seems to be a strong argument in favor of lenition,
but one example cannot prove anything. Nevertheless, it is possible to find
other examples besides ‘to sell’ in the Psara dialect of Miyako in which
/vu/ or /v/ corresponds to initial /u/ in OJ 34:
Chart 17:
Psara vu- ~ v- corresponding to OJ u-
Gloss Psara Old Japanese
‘be buried’ vdzmarü uNtumoru
‘quail’ vdzdza uNtura
‘inside’ vts uti
‘son’ vụta 35 uta
‘doubt’ (v.) vutagoo utaNkapu
‘move’ (v.) vụQtsü uturu
‘back’ vụra ura
‘hare’ (sign of the zodiac) vu u
‘sell’ Qv uru
‘hit’ (v.) vụtsụ 36 utu

In some cases, Psara initial /u-/ corresponds to OJ initial /u-/: 37

34
Data are from Shimoji (1979: 43-44).
35
Shimoji notes that the original Miyako word for ‘song’ is aagu (1979: 43).
36
Shimoji considers this word to correspond to MdJ butu ‘to hit’ (1979: 43) but I disagree,
because MdJ butu is a new word, and also because there are no other cases of Miyako vu-
corresponding to MdJ bu- except Miyako vụtsị̈ ‘whip’, which can correspond to either MJ
buti or muti ‘id’.
37
Data are from Shimoji (1979: 29-33).
38———Koreo-Japonica

Chart 18:
Psara u- corresponding to OJ u-
Gloss Psara Old Japanese
‘melon’ uuṛï uri
‘top’ ui upë
‘float’ ukabü ukaNpu
‘peep into’ ukagoo ukaNkapu
‘mortar’ usụ usu

It is impossible to claim that Miyako /v-/ represents a secondary

development. In addition, there are cases when Miyako /vu-/ corresponds
to OJ /pu-/ or /mu-/, clearly indicating that Miyako /v/ must be the result of
a lenited stop. 38
Chart 19:
Psara vu- ~ v- corresponding to OJ pu- ~ mu-
Gloss Psara Old Japanese
‘thick, fat’ vụda putô
‘joint’ vụsï pusi
‘whip’ (n.) vụtsị̈ MJ muti, buti
‘intimate’ vts mutu
‘angrily’ vụttī MdJ mutto

Thus, I think that the evidence from Psara cited above points toward
Proto-Japonic initial *bu-, and I tentatively reconstruct PJ *b- rather than
PJ *w-. The cases of secondary fortition in Miyako found in boo ‘king’
and tśaban ‘teacup’ probably represent analogical substitution, or
‘grandfathering’, in Leon Serafim’s terminology. It is necessary to note
that initial /wa-/ is found in Psara, but in most cases it is clearly secondary:
e.g., Psara waa ‘pig’ < PR *Uwa (Thorpe 1983: 316), Psara waabi ‘upper
side’ < PR *uwabe, Psara waari- ‘be chased’ < PR *oware-, and Psara
watśaku ‘mischief’, borrowed from Shuri ʔwacaku ‘id’. Cases such as
Psara waki ‘reason’, waza ‘deed’, and wadzụka ‘scanty’ (cf. MdJ wake,
waza, and wazuka) are more difficult to explain, but they might be recent
loanwords from Japanese.
The case for initial *b- rather than *w- may be further strengthened by
the second piece of independent evidence that comes from Japanese,
namely from two subdialects of the Toyama Hokuriku dialect found in two
pockets along the coast of the Sea of Japan. In Himi-shi on the western
side of Toyama Bay and around Uozu-shi on its eastern side one finds
initial /ba-/ corresponding to MdJ /wa-/ (Kawamoto 1973: 63-70; Shimono
1983: 321). These two areas are surrounded by Toyama subdialects that
behave exactly like Modern Japanese in that they have /wa-/ corresponding
to MdJ /wa-/. Since both initial /wa-/ and /ba-/ and medial /-wa-/ and /-ba-/
merge in Himi and Uozu as /ba-/ and /-wa-/, respectively, Kawamoto
38
Data are from Shimoji (1979: 43-44).
 Proto-Korean and Proto-Japonic Reconstructions———39

suggests that initial /w-/ underwent a fortition to /b-/ (1973: 75).

Structurally, this would be reasonable, but the geographic distribution
makes parallel innovation in Himi and Uozu unlikely. From the viewpoint
of linguistic geography, initial /ba-/ in Himi and Uozu looks instead like a
retention. The following data from Shimao-machi in Himi-shi is from
Kawamoto (1973: 64): 39

Chart 20:
Reflexes of /wa/ in Shimao and Tokyo
Gloss Shimao 40 Tokyo
‘trap’ bana wana
‘straw’ bara wara
‘fern’ barabi warabi
‘disposable chopsticks’ baruwasi waribasi
‘laugh’ (v.) baroo warau
‘bad’ barue warui
‘foam’ awa awa
‘wrinkle’ siwa siwa
‘telephone’ denwa denwa
‘banana’ banana banana
‘rose’ bara bara
‘midwife’ sanba sanba
‘bag’ kawan kaban
‘mackerel’ sawa saba
‘tobacco’ tawako tabako
‘buckwheat noodles’ sowa soba
‘firewood’ siwa siba

It is necessary to note that /-b-/ lenites into /-w-/ only in front of the
vowel /a/, cf. Shimao barabi ‘fern’ above. Therefore, Kawamoto’s rule
that /w-/ is fortified to /b-/ when there is no preceding vowel, and that /-b-/
is lenited to /-w-/ when there is one, works only when the following vowel
is /a/. Both rules are bizarre from the standpoint of human phonetics: Vba
should not be a better environment for lenition than Vbi or Vbu; and
certainly it is no less strange that /#wa-/ > /#ba-/. Of course, there is no b-
in Shimao or Uozu /__e, o, i. That makes the Toyama evidence for PJ *b-
overall weaker than the South Ryukyuan evidence, but I agree with Martin
that it is possible to imagine that lenition first affected *b- before high
vowels, then before mid vowels, and only in the last stage before the low
vowel /a/ (personal communication).

39
I have replaced Kawamoto’s phonetic transcription with phonological transcription.
40
In 2001 I tried to elicit more data from my sister-in-law and her parents, who are from
Shimao, but it looks like they rather consistently have /wa-/. It is possible that the /ba-/
corresponding to MdJ /wa-/ is rapidly disappearing in Shimao, or it might be an effect of
switching to Modern Japanese in front of a foreigner. I intend to do more field work with
Shimao informants in the future.
40———Koreo-Japonica

Therefore, I tentatively accept the lenition scenario for PJ *b- > w-.
Additional research must be done on the issue, because many problems,
especially Toyama b-retention, are far from being completely solved. It is
also necessary to find out whether other Miyako dialects (at least) show the
same consistent reflexes for PJ *b-/__u as Psara. If this is the case, then PJ
*b- has a firm basis in reality.

1.2.2.2 Yonaguni initial d- as an innovation

The Yonaguni dialect of the Ryukyuan language, spoken on Yonaguni, the
last island in the Ryukyuan archipelago and the closest to Taiwan, has a
number of phonological peculiarities that are not attested even on the
Sakishima islands nearest to Yonaguni. These peculiarities will place the
Yonaguni dialect in its special slot in the linguistic taxonomy of Southern
Ryukyuan, possibly even granting it the status of a separate language
rather than a dialect. However, the proper place of Yonaguni in Ryukyuan
will become clear only after its complex phonological history is
understood in greater detail. This section has the limited goal of solving the
problem of initial d- in Yonaguni (YG) corresponding to y- in all other
varieties of Japonic, for example, OJ yama ~ YG dama ‘mountain’, OJ yu
~ YG du ‘hot water’, OJ yôru ~ YG duru ‘night’, etc.
There are two points of view about initial d- in Yonaguni. Some
linguists, especially those who advocate the genetic relationship of Japonic
with Korean and/or ‘Altaic’, consider it to be an archaism, and support the
idea that proto-Japonic had *d- rather than *y- (Martin 1987: 20; Starostin
1991: 64, 135; Unger 1993: 46). Indeed, a number of suggested Koreo-
Japonic and Altaic-Japonic parallels, such as OJ yak- ‘to burn’ ~ MK thó-
< *toho- ‘id’., OJ yama ‘mountain’ ~ PT *daba- ‘cross over mountains’,
‘mountain pass’; OJ yworu ‘night’ ~ PT *dolbo ‘id’., crucially depend on
having an initial dental stop in proto-Japonic for their survival.
Other scholars question the validity of this reconstruction, and prefer to
view d- as an innovation, *y- > d-, that took place exclusively in Yonaguni
(Murayama 1981: 75-76; Thorpe 1983: 108; Whitman 1985: 19). My own
position for many years supported the lenition *d- > y- in all of Japonic
except Yonaguni rather than the fortition *y- > d- in Yonaguni. However, I
believe now that this position was mistaken, primarily motivated by my
adherence to the Altaic hypothesis and only superficial familiarity with the
Yonaguni internal data. As I am no longer a believer in ‘Altaic’, and have
a better understanding of the Yonaguni internal data, I think that the only
possible solution is to admit that Yonaguni underwent a fortition *y- > d-.

1.2.2.2.1 Internal evidence

Both Thorpe and Whitman argue for the secondary nature of YG d- on the
basis of the fact that PR *z went to /d/ in Yonaguni, e.g., PR *kezu
‘wound’ > YG kidi ‘id’. (Thorpe 1983: 91ff), (Whitman 1985: 19). A very
similar change of zy- > d- affected much more recent loans of mainland
Sino-Japanese elements: dama (MdJ zyama) ‘obstacle’, dutthu (MdJ
zyootoo) ‘first-class’, dudi (MdJ zyoozu) ‘skillful’, dudu (MdJ yoozyoo)
 Proto-Korean and Proto-Japonic Reconstructions———41

‘recuperation’, dubagu (MdJ zyuubako) ‘a tier of boxes’, dunŋaci (MdJ

zyuugatu) ‘tenth month’, dunsa (MdJ zyunsa) ‘policeman’.
There is also a fortition of r- > d- in Yonaguni: dukhunin (MdJ rokunin)
‘six people’, dukhunin (MdJ rokunen) ‘six years’, dusukhu (MdJ roosoku)
‘candle’, duya (MdJ rooya) ‘prison’.
Thus, one can come to the conclusion that initial voiced dental
affricates and taps all undergo fortition in Yonaguni. Thorpe (1983: 108)
also noted that in Yonaguni y > d in non-initial position after loss of an
original preceding high vowel, e.g., PR *iyaya ‘cliff’ > YG daya ‘id’. (cf.
Ishigaki iizaa, Taketomi iiyaa).
Yonaguni also underwent a fortition of voiceless fricative /s/ to affricate
/c/ before original high vowels /i/ and /u/ in both initial and medial
positions: chiruci < *sirusi ‘sign’, chima < *sima ‘island’, chiru < *siru
‘sap’, tuci < *tösi ‘year’, kwaci < *kwasi ‘sweets’, chii < *su ‘nest’, chici <
*susu ‘soot’, churi < *ksuri < *kusuri ‘medicine’. Fortition is very
widespread in Yonaguni.

1.2.2.2.2 Initial d- in Sino-Japanese loanwords

Whitman also argued for the secondary nature of d- in Yonaguni on the
basis of the loanword dasai ‘vegetables’ < Jpn. yasai ‘id’. < MC yacai ‘id’.
(Mandarin Chinese yecai ‘id’.) (Whitman 1985: 19). This is potentially a
very important argument because Yonaguni shows initial d- in a word that
has initial y- both in Japanese and its Chinese prototype. However, it was
the only case Whitman cited, and a single word could have possibly
undergone a process of phonological adjustment. In addition, yasai turns
out to be a new word in Yonaguni, because the old Yonaguni word for
‘vegetables’ is atakui (Takahashi 1986: 96). However, there are many
other old and recent loans that underwent a consistent replacement of y- to
d- in Yonaguni,41 and which also have y- in mainland Sino-Japanese and
its Chinese prototype. The following list contains all loanwords that can be
classified as recent borrowings of mainland Sino-Japanese or partially
Sino-Japanese words based on phonetic and/or semantic criteria: dagutuci
(MdJ yakudosi) ‘critical/unlucky year’, dakhunin (MdJ yakunin)
‘government official’, daguba (MdJ yakuba) ‘town hall’, dasai (MdJ yasai)
‘vegetables’, daguharai (MdJ yakubarai) ‘exorcism’, daci (MdJ yasi)
‘coconut palm’, dakkai (MdJ yakkai) ‘trouble’, dugu (MdJ yoku) ‘desire,
greed’, dudan (MdJ yudan) ‘negligence’, dudu (MdJ yoozyoo)
‘recuperation’.
Admittedly, one can still argue that in all these words initial d- was
substituted for mainland Japanese y- by analogy with the same
correspondence in native words, especially because initial y- seems to be
practically absent from the Yonaguni phonological system. In addition to
these relatively modern loanwords, we possibly also have several older
borrowings of mainland Sino-Japanese vocabulary.

41
All data below are taken from Ikema (1998).
42———Koreo-Japonica

YG dakkhan ‘kettle’ cannot be a direct loan from MdJ yakan, because

that would give *dagan in Yonaguni. Therefore, the Yonaguni word was
likely borrowed directly or indirectly from Middle Japanese at the time
when the latter still had final consonant -k: *yak-kan.
Similarly, YG dangasa ‘umbrella (lit. foreign umbrella)’ cannot be a
direct loan from MdJ yoogasa, because it would be reflected in Yonaguni
as *dugasa. The source of this word is a mystery, because it denotes an
object that did not appear in Japan prior to the sixteenth century, and by
that time MJ yaŋ > Early Modern Japanese yɔɔ. It is quite clear, though,
that it cannot be a very recent loanword. It is possibly a hybrid of Chinese
yang2san3 ‘steel-frame umbrella’ and Japanese yoogasa, which would
place it no later than the seventeenth century, when there was still frequent
contact with China.
YG dugunin ‘greedy person’ cannot be a loan from mainland Japanese,
because such a compound does not exist in the latter to the best of my
knowledge. Nor am I aware of *yokunin in any historical Japanese texts.
YG dudu ‘business’ does not look like a direct loan from mainland
Japanese yoozi either, because we would expect *dudi in Yonaguni.
YG duri ‘origin’ must be a rather old borrowing of Sino-Japanese yurai,
because it predates the raising of e > i in Ryukyuan (*e < *ai).
It is still possible to argue that Yonaguni has adjusted initial y- to d-
even in the early loanwords. Thus, the presence of initial d- in Sino-
Japanese loans cannot be used as a decisive argument against the lenition
d- > y-.

1.2.2.2.3 Typological arguments

Martin reconstructs both initial *b- and *d- 42 for Proto-Japonic on purely
typological grounds, as he believes that “a more natural hypothesis would
have the mainstream dialects lenite earlier stops” (1987: 20). However,
while he is right that the lenition d > y is more natural, the fortition y > d
or y > D is not unheard of either. For example, as witnessed by Chinese
loanwords in Vietnamese, Proto-Vietnamese *y- changed into dental stop
d- 43 by the time the quôc ngu’ writing system based on Portuguese
orthography was introduced in the sixteenth century (Mineya 1972: 97).
The Bunun language of Taiwan has consistently fortified PAN *y to ð, as,
for example, Bunun haðam ‘bird’ < proto-Austronesian *’ayam ‘id’.
(Zeitoun 2000: 116). The same process is well attested in the Turkic
language family, where Proto-Turkic *y- developed into palatalized d’- in
the Altai language, with intermediate ž- and j- well attested in other Turkic
languages, such as Kypchak and Bulghar (Serebrennikov and Gadzhieva
1986: 51-54). The ‘naturalness’ of lenition is countered by the fact that d-
is attested only in a single dialect and nowhere else in Japonic, which is
strange if Proto-Japonic indeed had *d- (see also 1.2.2.2.4 below).

42
Martin thinks that *d- was “surely palatalized and possibly affricated” (1987: 20).
43
In modern Standard Hanoi Vietnamese this initial d- > z-, but the stop articulation is still
preserved in dialects (Thompson 1987: 58).
 Proto-Korean and Proto-Japonic Reconstructions———43

1.2.2.2.4 Methodological arguments

I believe that the lenition theory suffers from three serious methodological
fallacies. First, it clearly violates the ‘majority rules’ principle of
comparative reconstruction: initial d- is attested only in Yonaguni, while
all other Japonic dialects have y-. Certainly, when typology can be used to
demonstrate that the proposed counterargument is typologically faulty,
namely, that y- > d- fortition does not happen, typology will override the
‘majority rules’ principle, but as we have seen above, in the case of y- > d-
fortition this is simply not true.
Second, the reconstruction of PJ *d- exclusively on the basis of the
Yonaguni data violates another important principle of comparative
reconstruction: the need for two independent pieces of evidence for a given
reconstruction. In this case we definitely lack the second independent piece
of evidence.
Third, the lenition theory takes Yonaguni d- at face value, without
examining the internal history of the Yonaguni dialect. When we do look
at it, we discover that Yonaguni underwent several fortitions. From this
point of view it is more likely that there was a systemic fortition y- > d-
rather than a lenition d- > y-, which goes against the major trend of
phonological development in Yonaguni.

1.2.2.2.5 Historical evidence for the fortition y- > d-

The most decisive argument in favor of the fortition y- > d- comes, not
surprisingly, from philology. In 1927, the prominent Ryukyuan philologist
and anthropologist Iha Fuyū published an article on the description of the
Ryukyuan islands in Sengcong taywang sillok [Records (of the reign) of
King Sengcong], written on the basis of an account provided by Koreans
shipwrecked on Yonaguni in 1477 and subsequently returned to Korea by
the Ryukyuan king Shō Toku in 1479 (Iha 1927: 54-56). This article is
primarily ethnographic and describes the various customs of the Sakishima
islands that the Koreans visited on their way to Shuri. A historical linguist
meanwhile will be more interested in how Yonaguni is written in Korean
itwu as 閏伊是麼. The last two characters clearly represent the word sima
‘island’, as the attached commentary makes clear: 其俗謂島為是麼 “in
their vernacular language ‘island’ is called 是麼” (Iha 1927: 56), and it is
consistent with the phonetic usage of these two characters in Korean itwu
writing. The first two characters represent the proper name of the island,
which is read [yun.i] in modern Sino-Korean. This fact did not go
unnoticed: Iha Fuyū comments that in the modern native name Dunan
‘Yonaguni’ the original /y/ must have changed to /d/, citing Ogura
Shinpei’s authority on the matter (Iha 1927: 56), who also equated 閏伊
[yun.i] with Yonaguni (Ogura 1935: 394). 44
More than half a century later, Murayama Shichirō also brought up this
fifteenth-century transcription *yun.i as evidence for the fortition of y- >
44
The difference in vocalism in the second syllable represents no problem, as in the Omoro
sōshi, the name of Yonaguni island is written as いにやぐに [iniyaguni] (OS XXI: 17) and
いねぐに [ineguni] (OS XI: 3).
44———Koreo-Japonica

d- in the Yonaguni dialect, also maintaining that this cannot be used as

evidence for reconstructing PJ *d- (1981: 74-76).
Nevertheless, Iha, Ogura, and Murayama overlooked one interesting
fact that makes the refutation of the lenition theory all the more spectacular.
The use of the modern Sino-Korean readings yun for 閏 and i for 伊 is, of
course, anachronistic. While the last character had the same Sino-Korean
reading i in the sixteenth century, the first character had the reading :zyun
with initial voiced fricative [z] (Hwungmwong I: 2a). Thus, Korean 閏伊
represents Yonaguni [zuni] or [zyuni], which is apparently an intermediate
stage of the fortition *y- > z- > d- in Yonaguni. Therefore, Yonaguni d- is
clearly an innovation, and cannot be used as evidence for reconstructing PJ
*d-.
2

MORPHOLOGICAL COMPARISONS
It goes without saying that common paradigmatic morphology represents
better proof of a genetic relationship than common basic lexicon, because
(a) it is much more stable than vocabulary, and (b) paradigmatic
morphology is never borrowed (except in the case of language mixing).
Surprisingly enough, much less has been done in the area of comparative
morphology between Korean and Japonic than in the comparison of their
basic lexicons. This is not to say that there have been no attempts to find
common morphological elements, for example, (Martin 1968, 1990;
Whitman 1985; Vovin 2001; Frellesvig and Whitman 2003). But all have
failed to demonstrate the existence of paradigmatic morphology beyond
listing certain isolated morphological parallels.

2.1 NOMINAL MORPHOLOGY

In this section, I will discuss case marking and pronouns.

2.1.1 Case marking

It is possible that case markers in Korean and Japonic have a different
morphological status. In Korean, case markers are clearly suffixes, because
they have variants depending on the preceding stem, exactly as in other
‘Altaic’ languages. However, in Japonic, at least historically they are likely
to be particles, since they are invariable and show no intimate connection
to a preceding stem. This circumstance alone makes it reasonable to
question whether we are comparing apples and oranges. Nevertheless, I
will review the known comparisons below as if this problem does not exist.

2.1.1.1 Non-nominative structures: case marker -i

Although Modern Korean is an accusative language, it is quite clear that it
used to be an ergative one, as convincingly demonstrated by Ross King
(1988), who showed that subjects of transitive verbs in dependent clauses
in Middle Korean are consistently marked by -i, while subjects of
intransitive verbs in dependent clauses are left unmarked.
While the active construction is attested in Western Old Japanese, with
subjects of active verbs optionally marked also by -i (Vovin 1997), this
construction has an apparently skewed distribution. The active case marker
-i is attested predominantly in Senmyō [Imperial edicts], where it occurs
twelve times in 62 texts, while it appears only five times in the whole
Western Old Japanese part (more than 4,100 poems) of the Man’yōshū,
comprising altogether 4,516 poems. There is only one attestation of -i in
the whole Eastern Old Japanese corpus of 400 poems, and that occurence
46———Koreo-Japonica

probably has a different explanation. The case marker -i is not attested in

the oldest Old Japanese poems in Kojiki and Nihonshoki. This skewed
distribution, with usage predominantly in imperial edicts, where the
Korean language influence must have been the greatest, suggests that -i
was borrowed from Korean, and that the active construction in Old
Japanese was short-lived. It was probably created as an imitation of
Korean patterns. Needless to say, -i is not attested in Ryukyuan. Examples
of -i in Western Old Japanese:
INA TÖ IP-EY-NTÖ KATAr-e KATAr-e tö NÖR-AS-E kösö Sipï-i pa
MAWOS-E SIPÏ-N-KATAR-I tö NÖR-U 1
no DV say-EV-CONC speak-IMP speak-IMP DV call-HON-EV PT Sipï-
ACT TOP say(HUM)-EV forced-DV(ATTR)-say-NML DV say-FIN
Though [I] say no, [you] command [me]: ‘Speak, speak!’ [but] the things
that Sipi says, [you] call a forced speech (MYS III: 237)

TAMA-NÖ WO-nö TAYE-NSI-i 2 IMÔ-tö MUSUNP-Î-TE-si

pearl-GEN cord-GEN tear-NEG/TENT-ACT beloved-COM tie-INF-
PERF(INF)-PAST/ATTR
The indissolubility of the pearly cord that tied [me] with [my] beloved
(MYS III: 481)

Kï-nö SEKÎ MORI-i TÖNTÖMË-TE-M-U kamo

Ki-GEN barrier guard-ACT stop(INF)-PERF-TENT-FIN PART
Will the barrier guard at Ki stop [me]? (MYS IV: 545)

UNAPÎ-WOTÖKÖ-i AMË APUNK-Î-TE...

Unapi-man-ACT sky look up-INF-SUB
The man from Unapi looked up at the sky and...(MYS IX: 1809)
IPÊ-N-AR-U IMÔ-i ONPONPOS-I-MISE-M-U
house-LOC-exist-ATTR beloved-ACT worry-INF-show-TENT-FIN
[My] beloved, who is at home, will worry [about me] (MYS XII: 3161)

PUNTIPARA-NÖ ASÖMÎ MARÖ-RA-i PUMÎ-WO OP-ÊR-U KAMË-WO

PÎTÖ-TU TATEMATUR-AKU TÖ MAWOS-i-TAMAp-u n-i...
PuNtipara-GEN retainer Marö-PL-ACT writing-ACC bear-PROG-ATTR
tortoise-ACC one-CL offer(HUM)-NML DV say(HUM)-INF-
grant(HON)-ATTR DV-INF
[They] said that the retainer Fujiwara Maro and others had offered a
tortoise bearing writing [on its back]... (SM 6) 3
1
Capital letters are used to transcribe parts of Japonic and Korean texts written
semantographically.
2
The -i in this example may be also interpreted as a quasi-free noun ‘one’, ‘thing’, ‘person’,
cognate to the Korean quasi-free noun -i with the same function (Martin 1987: 64).
3
Itabashi (1990), on the basis of his proposal that there is an accusative marker i in Old
Japanese, claimed that the i in the examples in SM 6-28 is also an accusative marker which
is used to mark a subject in indirect discourse. There is very little evidence for establishing
 Morphological Comparisons———47

KYAUPUKU-i KUNI-NÖ UTI-NÖ WONTA-NÖ KÖPORI-ni

KUNKANE-WO INTE-TAR-I TÖ MAWOS-I-te TATEMATUR-ER-I
Kyaupuku-ACT province-GEN territory-GEN WoNta-GEN district-LOC
gold-ACC exit-PERF-FIN DV say(HUM)-INF-SUB offer(HUM)-PROG-
FIN
Kyōfuku reported that gold had been found in the province, in the district
of WoNta (SM 12)

Naramarö Kômarö-RA-i SAKASIMA N-I AR-U TÖMÖ-NKARA-wo

iNsanap-î PÎK-Î-K-Î-TE
Naramaro Komaro-PL-ACT opposite DV-INF exist-ATTR companion-
kind-ACC instigate-INF bring-INF-come-INF-SUB...
Naramaro, Komaro and others instigated the rebels and brought them...
(SM 19)
Nakamarö-i ITUPAR-I-KANTAM-Êr-u KÖKÖRÖ-wo MOT-I-te IKUSA-
wo OKÖS-I...
Nakamaro-ACT lie-INF-deceit-PROG-ATTR heart-ACC have-INF-SUB
army-ACC raise-INF
Nakamaro, having a deceitful heart, raised an army and... (SM 28)

KOKU-WAU-i WAU-WI-ni IMAS-U TÖKÎ pa NPOSATU-nö

NSYAUKAI-wo UKË-yö tö NÖTAMAP-Î-te AR-I
country-king-ACT king-position-LOC exist(HON)-ATTR time TOP
Boddhisattva-GEN precept-ACC receive-IMP DV say(HON)-INF-SUB
exist-FIN
[Buddha] said that the king of a country, when [he] is on the throne, should
accept the Bodhisattva precepts (SM 28)

Nakamarö-i TANTASI-KÎ OMÎ tö s-i-te PANPER-I-t-u

Nakamaro-ACT loyal-ATTR retainer DV do-INF-SUB serve-INF-PERF-
FIN
Nakamaro served as a loyal retainer (SM 34)

Wakë-i MAWOS-I-te AR-I

Wake-ACT report-INF-SUB exist-FIN
[Prince] Wake reported [to us about Nakamaro’s plot] (SM 34)

NTAUKYAU-i TAMAPAR-I-te KARE-RA-Nka MATÔP-ÊR-U

KÖKÖRÖ-woNpa WOSIPË-MÎTINPÎK-Î-te...

an accusative marker i in Old Japanese (Vovin 2005: 112). Besides, why should a subject in
indirect discourse in Senmyō be marked by an accusative marker, while left unmarked
elsewhere? A comparison of examples in SM 6-28, with some previous examples from the
Man’yōshū which I did not exclude on other grounds, as well as examples in SM 34-62
below, demonstrates that i marks the subject whether it is a subject in indirect discourse or
not.
48———Koreo-Japonica

Dōkyō-ACT grant-INF-SUB he-PL-GEN go astray-PROG-ATTR heart-

ACC/EMPH teach(INF)-guide-INF-SUB...
Dōkyō will deign to teach and guide their hearts that went astray... (SM 35)

PUTA-RI N-Ö NSENSI-TATI-i ONANSI KÖKÖRÖ-wo MOT-I-te...

two-CL DV-ATTR meditation teacher-PL-ACT same heart-ACC have-
INF-SUB...
two meditation teachers have the same heart... (SM 41)
WA-NKA UTUKUSINPI-i AMË TUTI-nö MÎ-KÖKÖRÖ-wo
UNKWOKAS-I-matur-uNpë-kî KÖTÖ pa NA-SI
we-POSS virtue-ACT heaven earth-GEN PREF-heart-ACC move-INF-
HUM-DEB-ATTR thing TOP no-FIN
our virtues are not the ones that would move the hearts of heaven and earth
(SM 42)

POUKUN-i MONÖ MAWOS-Er-i

Hōkun-ACT thing say(HUM)-PROG-FIN
Hōkun was saying things (SM 44)

KÏ-NÖ KÔSAMÎ N-Ö ASÖMÎ-RA-i...IKUSA YANPUR-I KATE-WO

TUPÎYAS-I-te KAPÊR-I-MAWI-K-Î-T-U
Ki-GEN Kosami DV-ATTR retainer-PL-ACT...army destroy-INF food
supply-ACC exhaust-INF-SUB return(INF)-come(HUM)(INF)-come-INF-
PERF-FIN
retainer Ki no Kosami and others came back losing the army and
exhausting their food supplies (SM 62)

As I mentioned above, there is only one example of -i in Eastern Old

Japanese:

papa-i mor-e-Ntömo...
mother-ACT guard-EV-CONC
Though [my] mother guards [me]... (MYS XIV: 3393)

This is a poem from Simotupusa province, belonging to region A of

Eastern Old Japanese, which was the least influenced by the Western Old
Japanese literary language. However, this is the only example found in
Eastern Old Japanese, and it can be reanalyzed as containing not the active
case marker -i, but the i- prefix on the following verb that indicates the
directive-locative focus, thus:

papa i-mor-e-Ntömo...
mother LF-guard-EV-CONC
Though [my] mother guards [me] here... (MYS XIV: 3393)
 Morphological Comparisons———49

Therefore, the active case marker -i is limited to Western Old Japanese.

As mentioned above, there are only five cases of its occurrence in the
Man’yōshū. One of these (from MYS III: 481) is doubtful, as it may have
an alternative explanation. The remaining twelve cases are all confined to
the Senmyō [Imperial edicts]. If the active marker -i was indeed an
archaism, surviving only in Western Old Japanese and going back to the
remote times of ‘Proto-Koreo-Japonic’, it is strange that it should be
completely absent from the archaic songs of the Kojiki and the Nihonshoki,
as well as from the songs in the Jōgu Teisetsu. After the Nara period, -i is
occasionally used in the language of early Heian commentaries, but it
never appears in literary texts. After mid-Heian times it essentially
disappears altogether. Thus, we can clearly see that -i is an innovation
limited to Central Japanese; but even there it was short-lived. Such a
restricted usage within Japonic 4 can have only one explanation: it is a loan
from a variety of Old Korean.

2.1.1.2 Genitive case markers

There are different genitive markers in Korean and Japonic.

2.1.1.2.1 Middle Korean -s, Western Old Japanese -tu

In the important article “Morphological clues to the relationship of
Japanese and Korean”, Martin argues that the Middle Korean genitive –s,
also found in Old Korean as –cï, and the Old Japanese ‘locative genitive’ 5

4
There is a nominative case marker -i attested in the dialect of the southern part of Ōita
Prefecture in Kyushu. However, its function is clearly nominative, and not active or
ergative. The following examples are from Matsuda 1991:

kaziya-i sit-ty-or-u
blacksmith-NOM know-SUB-exist-ATTR
The blacksmith knows [it].

ame-i fur-u
rain-NOM fall-ATTR
It rains (lit. rain falls).

kuruma-i oo-i
car-NOM is many-ATTR
There are many cars.

Even if this -i in Ōita is related to WOJ -i, its usage is still limited to Japanese.
5
Martin also makes the important observation that the traditional label ‘locative genitive’ is
misleading, since OJ -tu is found not only in examples like nipa-tu tori ‘chicken (lit.: bird
of the yard)’, but also in asa tu ki ‘chives (lit.: mild-GEN onion)’ (1990: 488), which of
course has nothing to do with a ‘locative’ whatsoever. However, I doubt that the tu in asa tu
ki is a genitive case marker. In addition to asa tu ki ‘chives’ (not attested in Old Japanese),
we also have WOJ töwo tu pîtö ‘people from far [away]’ (MYS V: 857, 871), taka tu sima
‘high island’ (Norito), sikö tu okîna ‘stupid old man’ (MYS XVII: 4011), etc. It is easy to
see that tu in these cases is not a genitive case marker: it apparently has the function of a
copula, since ‘mild onion’ is not ‘onion of mild’, but ‘onion that is mild’; ‘high island’ is
not an ‘island of high’, but ‘island that is high’; ‘stupid man’ is not ‘man of stupid’, but
50———Koreo-Japonica

-tu are likely cognates (Martin 1990: 488). There is, however, a problem
that makes me doubtful about the proposed relationship: the
correspondence of OJ -s- to MK -t- is very problematic, because it is
supported by very few reliable etymologies. Martin lists this
correspondence as correspondence #14a and provides eight Koreo-Japonic
lexical comparisons that support it: ‘thing’, ‘bundle’, ‘dew’, ‘break’,
‘hatchet’, ‘husband’, ‘mouth’, and ‘one’ (1966: 212). Martin himself
placed a question mark after the last one, so I will not discuss it. Among
the remaining seven, only ‘thing’ (OJ kötö 2.3 ~ MK kes L) belongs to the
list of the most reliable etymologies (Martin’s Class I); ‘bundle’ and ‘dew’
are assigned less reliable status (Class II); and the remaining etymologies
are placed in the least reliable Class III. I will accept ‘thing’ as valid for
the time being, although it is necessary to mention that MK kes in this
equation has a potential rival: MK kwot H ‘thing’, which exhibits a much
more straightforward correspondence of t : t with OJ kötö. The other
etymologies are discussed below. 6
‘Bundle’: Martin compares MdK mus ‘bundle’, mukk- < musk- ‘tie into
bundles’ with MdJ musub- ‘tie into bundles’ < ?mutub- (1966: 227, #32).
Cf. MK mwús ‘bundle’, mwùs- L ‘bind into bundles’ (Nam 1997: 594), and
OJ musuNp- ‘tie‘. OJ and MJ mutuNp- does exist, but it means ‘be
intimate/be in good relationship with’ (JDB 1967: 727), so it does not
belong here. Therefore, OJ musuNp- < PJ *musunp-, and the comparison
with Korean reflects not the correspondence MK -t- : OJ -s-, but a more
reliable correspondence -s- : -s-. 7
‘Dew’: Martin compares MK ìsúl ‘dew’ with MdJ tuyu 8 2.5 ‘id’. (1966:
230, #63). We cannot internally justify the loss of initial i- in Japonic, nor
its accretion in Middle Korean. Therefore, this etymology is best
abandoned on phonetic grounds.
‘Break’: Martin compares MK pòsó- ~ pòzó- ‘break, smash’ with MdJ
but- ‘hit’ (1966: 227, #24). The latter is a new word, not attested in Old
Japanese or Middle Japanese, and possibly of an expressive nature. The
etymology is best abandoned on these grounds as well as because of its
vague semantics.

‘man who is stupid’, and ‘people from far away’ are not ‘people of far away’, but ‘people
who are from far away’. I believe it is necessary to differentiate this tu from genitive-
locative -tu in the same way we differentiate the defective verb n-ö from the genitive case
marker -nö. This tu probably represents one of the forms of the defective copula *t-
postulated by Frellesvig (2001).
6
I add Middle Korean to Martin’s Modern Korean data and Old or Middle Japanese to
Martin’s Modern Japanese data whenever applicable.
7
There is one obstacle to accepting this etymology as evidence for a genetic relationship.
Even if we assume that MK mwus- L ‘tie into bundles’ < PK *mwusu- LH, we cannot
easily explain the remaining -Np- in the PJ *musuNp-: it is likely to represent an obsolete
suffix, but what was its function? Thus, it remains a possible comparison, but it is not
absolutely proven.
8
Cf. OJ tuyu, which is likely to go back to PJ *tuyo on the basis of the Ryukyuan data. In
particular, the spelling tuyo is statistically more frequent than tuyu in the Omoro sōshi
(Hokama 1995: 137, 434).
 Morphological Comparisons———51

‘Hatchet’: Martin compares MK nàs ‘sickle’ with MdJ nata 2.1 ‘hatchet,
machete’ (1966: 233, #100). I am not aware of any attestations of this word
in Japanese prior to Early Modern Japanese, but it may be a pure
coincidence, since neither Old nor Middle Japanese texts deal with
agriculture. I accept it as a valid etymology supporting the correspondence
s : t, but it is necessary to note that this word may only be a loan, since it
refers to a certain object with a metal blade that did not exist during the
Lower Neolithic. Also, the Koreo-Japonic proto-language could not
possibly be less than 6,000 to 4,000 years old, when metallurgy did not yet
exist in East Asia.
‘Husband’ (or ‘man’, ‘male’): Martin compares MK :pes, :pet
‘companion’ and *‘husband’ (in kasi-pesi ‘wife and husband’) with OJ
wotö-kô ‘man’ (1966: 234, #114). The main problem here is that OJ wotö-
does not mean ‘man’, but *‘young’, cf. wotö-mê ‘maiden’ (lit. young
woman). Thus, this etymology must be abandoned due to the semantics of
the Japanese form.
‘Mouth’: Martin compares MK :kwut, kwùs ‘hollow, cavity’ with OJ
kuti ‘opening, mouth’ (1966: 237, #150). OJ kuti < *PJ kutuy as witnessed
to by the compound kutu-wa ‘bridle’ (lit. mouth-ring). Phonetically it is a
plausible match, but the semantics are questionable. However, I accept this
etymology as marginally acceptable.
Thus, the correspondence of MK -s- to OJ -t- can be traced on the basis
of the following etymologies: genitive marker, ‘thing’, ‘hatchet’, and
‘mouth’. We should provisionally accept it as it satisfies Meillet’s rule that
a given correspondence must be supported by at least three etymologies.
Let us look now at the distribution of -tu. There are plenty of examples
in Western Old Japanese. In Eastern Old Japanese, there are six examples
of -tu in the Aduma-uta (MYS XIV) (Mizushima 1984: 899), but -tu is
conspicuously absent from the Sakimori-uta (MYS XX). Let us look at
these examples in their entirety:

natu-sô-N-pîk-u una-kamî-N-kata-nö okî-tu SU-ni pune pa töNtömë-m-u

sa-yô pukë-n-i-kêr-i
summer-hemp-GEN-pull-ATTR sea-top-GEN-harbor-GEN offing-
GEN/LOC sandbank-LOC boat TOP stop-TENT-FIN PREF-night
deepen(INF)-PERF-INF-PRET-FIN
Let us stop the boat at the sand bank in the offing from the upper harbor on
the sea ‘that pulls like summer hemp’ (makura-kotoba). The night has
deepened! (MYS XIV: 3348)

This poem is from Kamitupusa Province, which belongs to Eastern Old

Japanese Area A. However, the poem does not have any typical Eastern
Old Japanese features, and what is even more interesting is that it faithfully
preserves Western Old Japanese vocalism without a single misspelling,
typical for normal Eastern Old Japanese texts. Thus, I am inclined to
disregard this poem as evidence for EOJ -tu.
52———Koreo-Japonica

pana-N-tir-ap-u könö muka-tu wo n-ö Wona-nö wo pîNsi-ni tuk-u-maNte

kîmî-Nka yö moNkamo
flower-GEN-fall-ITER-ATTR this opposite side-GEN/LOC peak DV-
ATTR Wona-GEN peak sandbank-LOC attach-ATTR-TERM lord-POSS
age PT
[I] wish [that my] lord live until the peak of Wona, which is the peak on
the opposite side, where the flowers fall, turns into a sandbank (MYS XIV:
3448)
The place of origin of this poem is not known, but again it looks like a
normal Western Old Japanese text, with no peculiarities typical of Eastern
Old Japanese. I do not think this poem constitutes evidence for EOJ -tu.

kôpïsi-k-êNpa k-îmas-e wa-Nka se-kô kakî-tu YANkï ure tum-î karas-i

ware tat-i-mat-am-u
miss-ATTR-COND come(INF)-HON-IMP I-POSS beloved-DIM fence-
GEN/LOC willow top pluck-INF wither-CAUS-INF I stand-INF-wait-
TENT-FIN
If [you] miss [me], please come, my beloved! Plucking the ends [of
branches] of the willows at the fence and letting them wither, I will wait
[for you] (MYS XIV: 3455)

The place of origin of this poem is not known either, and again it looks
like a normal Western Old Japanese text, with no peculiarities typical of
Eastern Old Japanese. I do not think this poem constitutes evidence for
EOJ -tu.
osö paya mo na-wo kösö mat-am-ê muka-tu wo-nö sipî-nö kô-yaNte-nö
ap-yi pa taNkap-aNsi
slow fast PT you-ACC PT wait-TENT-EV opposite-side-GEN/LOC peak-
GEN chinquapin-GEN DIM-branch-GEN meet-NML TOP differ-
NEG/TENT
Whether [you come] quickly or slowly, [I] will wait for you. It is no
different from the meeting of small branches of chinquapin trees on the
peak on the opposite side (MYS XIV: 3493)

The place of origin of this poem is not known either, but it has one
feature typical for Eastern Old Japanese, since matamë ‘will surely wait’ is
misspelled as matamê. In addition, the word yaNte ‘branch’, although it is
a hapax legomenon, clearly represents a dialectal form of WOJ yeNta
‘branch’. I think this poem offers some weak evidence for EOJ -tu.

osö paya mo kîmî-wo si mat-am-u muka-tu wo-nö sipî-nö sa-yeNta-nö tökî

pa suNk-u tömo
slow fast PT lord-ACC PT wait-TENT-FIN opposite side-GEN/LOC peak-
GEN PREF-branch-GEN time TOP pass-FIN CONJ
 Morphological Comparisons———53

Whether [you come] quickly or slowly, [I] will wait for [my] lord, even if
the time passes [between] the branches of chinquapin trees at the peak on
the opposite side (MYS XIV: 3493a)

This is a variant of the preceding poem, but it does not have any typical
Eastern Old Japanese features. I do not think this poem constitutes
evidence for EOJ -tu.

mawo-N-kömo-nö pu-nö ma tika-ku-te ap-an-ap-ê-Npa okî-tu ma-kamo-nö

naNkëk-î sö a-Nka s-uru
nettle-GEN-mat-GEN mesh-GEN space close-INF-SUB meet-NEG-ITER-
EV-CON offing-GEN/LOC INT-duck-COMP cry-NML PT I-POSS do-
ATTR
Because [we] do not meet being as close [to each other] as spaces between
meshes of a nettle mat, I cry like a real duck in the offing (MYS XIV: 3524)

This poem is clearly Eastern Old Japanese, because it includes a typical

Eastern Old Japanese sequence -anap- consisting of a negative marker -an-,
followed by iterative -ap- (Western Old Japanese has the opposite ordering
of these two morphemes). I believe this poem provides strong evidence in
favor of EOJ -tu.

Thus, of the six examples in the Azuma-uta (MYS XIV), only two poems
seem to be written in Eastern Old Japanese. The other four appear to be
normal Western Old Japanese texts. In addition, the occurrence of EOJ -tu
seems to be limited to just two contexts: okî-tu N ‘N of the offing’ and
muka-tu wo ‘peak on the opposite side’. Both contexts appear frequently in
Western Old Japanese poetry, and it is quite possible that they were just
imitated by Eastern poets. It is possible that -tu was ‘introduced’ to Eastern
Old Japanese texts by Opotomo-no Yakamoti, who collected and edited the
texts. In any case, EOJ -tu looks like a loan from Western Old Japanese to
me; I do not consider it evidence for reconstructing PJN *-tu.
After the Nara period, -tu becomes fossilized, remaining only in set
expressions, like wata-t[u] umi ‘sea’. Thus, one can clearly see that even
WOJ -tu was probably short-lived, although it appears in the oldest extant
texts of the Kojiki in contrast to active case marker -i.
There are no traces of -tu in Ryukyuan, and that brings us to the logical
conclusion that -tu must be a loan from Korean as well. However, we are
presented with a problem: why was Korean -s borrowed as WOJ -tu and
not, let us say, WOJ *-su? The answer, I believe, lies in Old Korean. The
genitive marker in Old Korean corresponding to the later MK -s is written
with the character 叱. This character has the EMC reading *tɕhit, which
indicates that the Old Korean predecessor of MK -s had an affricate
articulation. I believe that this Old Korean marker can be tentatively
reconstructed as *ci. Western Old Japanese, which had no affricates,
borrowed OK *-ci or its possible dialectal (Paekche?) variant *-ci as *-tu.
54———Koreo-Japonica

2.1.1.2.2 Middle Korean -oy ~ -uy, Japonic -nö

Although Ramstedt speculated that MK -oy ~ -uy < PK *-o-ńi ~ *-u-ńi, a
position that is impossible to justify on the basis of the Korean data, it is
apparent that these genitive markers are not related. What is important is
that Japonic -nö, in sharp contrast to WOJ -tu, has wide temporal and
geographic distribution: it is found consistently in almost all Japonic
dialects (with the exception of the few that replaced it with -ga), including
Western Old Japanese, Eastern Old Japanese, and Ryukyuan. It is also
attested throughout the whole written history of Japonic, from the seventh
century in Japanese and the sixteenth century in Ryukyuan to the present.
Middle Korean -oy ~ -uy, as well as its reflexes, enjoy a similar status in all
modern Korean dialects. The same marker is attested in Old Korean as
-(h)i (矣). Therefore, both markers can be safely reconstructed for their
respective proto-languages; on the proto-language level they turn out to be
different.
Even taking into consideration the age of ‘Koreo-Japonic’, with
estimates varying between 4,000 and 6,000 years, and the fact that both
languages are agglutinative, it remains unclear why the majority of Uralic
languages, which are no less agglutinative, with a Uralic proto-language
estimated to be much older than ‘Koreo-Japonic’, managed nevertheless to
have quite regular reflexes of their common genitive case marker *-n. The
same applies to the accusative case marker that I discuss below: most
Uralic languages have regular reflexes of Proto-Uralic accusative *-m.

2.1.1.3 Accusative markers: Middle Korean -lol ~ -lul ~ -ol ~ -ul ~ -l,
Old Japanese -wo
The basic form of the bewildering variety of accusative allomorphs found
in Late Middle Korean (-ol ~ -ul after consonantal stems and -lol ~ -lul ~ -l
after vowel stems) is probably just the -l that is found in the accusative
forms of the first and the second person singular personal pronouns: na-l
‘me’ and ne-l ‘thee’ (Yi Swungnyeng 1961: 220-221), later with the more
frequent nè-lúl (Martin 1992: 850). This -l also agrees well with the Early
Middle Korean accusative 乙 [ïr], which probably just renders [r]. I suspect
that this accusative originated as the grammaticalization of the Middle
Korean demonstrative pronoun tyé ‘that’, with a subsequent lenition of -t-
> -r-, similar to the case of the Hungarian accusative -t, which is also
believed to be a result of grammaticalization of Uralic *to ‘that’. The Old
Korean accusative 持 -γïr, 9 first noticed by Miller (1977: 160), is probably
only partially related to MK and EMK -r, as it consists of two morphemes:
-γï- and -r. Since OK -γ- can reflect both PK *-k- and *-p-, 10 I tried to
9
We have to reconstruct the Old Korean and Early Middle Korean forms with -t and -r ,
respectively, as their orthography is based on two different varieties of Middle Chinese: Old
Korean orthography is based on Early Middle Chinese, which still had final -t in the
syllables of entering tone, and Early Middle Korean orthography is based on Late Middle
Chinese (Chang’an dialect), where -t > -r.
10
Proto-Korean intervocalic *-k- is reflected as -G- in both Early Middle Korean and
Middle Korean, while Proto-Korean intervocalic *-p- is reflected as EMK -p- [-b-] or -Ø-
(rarely) and MK -W-[-B-] or -Ø-. See Vovin (1995) for details.
 Morphological Comparisons———55

connect this OK accusative with an OJ -wo on the assumption that OJ -w-

< PJ *-b- (Vovin 1995: 233). However, in spite of the possibility that OJ
initial w- < PJ *b- (see 1.2.2.1 above), we have no similar evidence for OJ
medial -w-, which should be reconstructed as PJ *-w-. It remains unclear
whether we can reconstruct OJ -w- as *-b-, in spite of the fact that PJ *b-
and *-w- are in complimentary distribution, simply because none of the
Japonic languages offers us any evidence for *-b-. Therefore, this equation
must now be abandoned. Since Proto-Korean intervocalic *-k- is regularly
reflected in Early Middle Korean and Middle Korean as -G-, but Proto-
Korean intervocalic *-p- has less stable reflexes as EMK -p- or -Ø- and
MK -W- or -Ø-, I suppose that Proto-Korean accusative should be
reconstructed as *-pï-tV, where *-tV represents an original demonstrative
pronoun (> MK tyé ‘that’), and *-pï is an original accusative marker. The
reconstructed form of the Proto-Korean accusative marker *-pï rules out
Miller’s comparisons of this marker with Proto-Turkic and Proto-Mongolic
accusative -*γ (Miller 1977: 166), as well as my own comparison with OJ
accusative -wo (Vovin 1995: 233).
The accusative marker in both Western and Eastern Old Japanese is
clearly -wo. Nohara indicates that there are two accusative case markers in
the Omoro sōshi: -yu (spelled -yo) and -fa, also appearing in the Ryūka
(Nohara 1986: 28-29). The first of these markers is also attested in Miyako,
for example, Shimoji -yu, Psara -yu (Nohara 1998: 331, 401). The expected
development of PJ *-w-/__o in Miyako is -Ø-, cf. OJ töwo ‘10’ ~ Psara tuu
‘id’. (Hirayama 1966: 298), and OJ awo ‘blue, green’ ~ Psara oo ‘id’.
(Hirayama 1966: 433). However, in one word we find the unexpected
reflex -z-: OJ iwo ‘fish’ ~ Psara ïzu ‘id’. (Hirayama 1966: 362). This -z-
could only be the result of a fortition of -y-, found in this word in other
Ryukyuan dialects: Nase ʔyu, Iejima ʔyuu, Hateruma ʔyuu ‘fish’
(Hirayama 1966: 362), Shuri ʔiyu ‘fish’ (RGJ 1976: 253). Therefore,
Miyako accusative -yu can be tentatively accepted alongside OR -yo as a
cognate of OJ -wo. Examples from Old and Classical Ryukyuan:

omoro-yo mioyas-i serumu-yo mioyas-i

sacred song-ACC present(HUM)-IMP prayer-ACC present(HUM)-IMP
Present the sacred song, present the prayer (OS VIII: 411)

TARU-yo URAm-e to-te NAky-u ga HAMA-CIDURI

who-ACC resent-INF DV-GER cry-ATTR PT shore-plover
Shore plovers, whom are you resenting and crying? (Ryūka, cited in RGJ
1976: 284)

Despite the fact that we can reconstruct PJ *-wo as an accusative

marker, we still cannot compare it with the tentative PK *-pï, because if PJ
*-wo was from *-bo, we would expect the reflex *-bu or *-vu in Miyako,
not -yu.
56———Koreo-Japonica

2.1.1.4 Dative-locative markers

It is significant that much more reliable parallels between Old Japanese
and Korean are found only among locative markers. Locative case markers
are more susceptible to borrowing than are nominative, genitive, and
accusative markers.

2.1.1.4.1 Old Korean -ra[ng], Old Japanese -ra

Old Korean has a locative marker -ra or -rang, 11 which appears in the
Hyangka texts four times. There is no equivalent in Middle Korean or any
other later variety of Korean. Old Korean examples:

MWORGAY-i phall-on NAli-rang KIRANG-uy CUs-i isi-swo-ra

sand-NOM blue-REAL/ATTR river-LOC [Hwa]rang KI[pha]-GEN image-
NOM exist-EMPH-FIN
In the river, where the sand is dark (lit.: blue, dark part of the spectrum)
the image of [Hwa]rang Ki[pha] is [right there] 12 (Hyangka IV: 4)

TWONGKYENG POLK-kuy TOLAL-rang PAM TUR-i NWOL-NI-ta-pi-

ka
eastern-capital bright-ATTR moon-LOC night enter-ADV play-go-IND-
FIN-TRANSF
In [the light of] the bright moon in the Eastern Capital [I] was playing
around at night, and... (Hyangka V: 1)

CHYEN SWU [CUMUN SWON?] KWANUM-ci ALPH-rang PIL-Gi-ki

SORP-wo-l TWU-no-hwo-ta
thousand hand Kwanum-GEN front-LOC ask-CAUS-NML(?) say(HUM)-
MOD-ATTR/IRR put-PRES-?-FIN(?)
[I] am saying [my] prayers in front of the Kwanum with a thousand hands
(Hyangka VII: 3)

TWUGwu ma-n NA-ra HOton-sa-un CWU-Si-kwo ki-no-wo-s-to-ra

two lack-ATTR/IRR I-DAT one-?-TOP give-HON-GER ?-PRES-MOD-?-
?-IMP
Please support [me] by giving just one [eye] to me, who does not have two
(Hyangka VII: 7)

The dative-locative case marker -ra is poorly attested in Old Japanese,

and does not occur in any later variety of Japonic. Most of the Japanese
sources provide only one example with this case marker: MYS XV: 3689
(cited below). I believe, however, that it also is preserved in some other
occurences. Most of the examples come from the Senmyō, where the
marker appears in the construction opo-myi-kötö-ra ma tö (great-HON-
deity-DAT according DV) ‘according to the Great Deity (i.e., the
emperor)’, which is frequently believed to be a petrified formation -rama
11
For a detailed discussion, see Vovin (2000: 150-153).
12
The image of Hwarang Kipha is represented by the reflection of the moon in the water.
 Morphological Comparisons———57

of unknown origin. However, Yamada Yoshio has convincingly

demonstrated, through his detailed analysis of the Senmyō examples, that
this formation is bi-morphemic, including euphonic -ra and ma ‘according
to’ (1954: 371-381). I disagree with his point of view that -ra is just a
euphonic device; since Yamada himself pointed out that imperial edicts
were read by imperial messengers, rather than by the emperors themselves,
it seems plausible to identify -ra as a dative-locative case marker in the
phrase opo-myi-kötö-ra ma tö (great-HON-deity-DAT according DV)
‘according to the Great Deity (i.e., the emperor)’. This identification is
further confirmed, contrary to Yamada’s opinion (1954: 375), by the fact
that the same -ra is found in the Nihonshoki and the Man’yōshū as an
apparent locative case marker.

yamatö pa kuni-nö ma-pô-ra ma

Yamato TOP land-GEN INT-top-LOC according
since Yamato is in the highest place of the country (NK 22)

kîkös-i-wos-u kuni-nö ma-po-ra Nsö ka n-i ka-ku n-i posi-kî mani-mani

sika n-i pa ar-aNsi ka
rule(HON)-INF-HON-ATTR country-GEN INT-top-LOC PT thus DV-INF
thus-INF DV-INF desire-ATTR according thus DV-INF TOP exist-
NEG/TENT PT
in the highest place of the country, where [the emperor] rules, it cannot be
thus according to what [you] wish in this way and that way, can it? (MYS
V: 800)

KÖNÖ YÔ-ra PA sa-YÔ PUKË-N-Urasi

this night-LOC TOP PREF-night deepen(INF)-PERF-SUP
It seems that the night (lit.: in this night) has grown deep (MYS X: 2224) 13
ipata-NÔ-ni yaNtôr-i s-uru kîmî ipê-N-pîtö-nö iNtu-ra tö ware-wo töp-
aNpa ika n-i ip-am-u
Ipata-field-LOC lodge-NML do-ATTR lord home-GEN-person-GEN
where-LOC DV I-ACC ask-COND how DV-INF say-TENT-FIN
[Oh, my] lord who lodged at the Ipata field. If people from [your] home
ask me where [are you], how should [I] answer? (MYS XV: 3689)

IyaPÎKÔ KAMÏ-nö puMÖTÖ-NI KÊPU-ra mo ka SIKA-nö PUS-Uram-u

Iyapiko deity-GEN foot [of the mountain]-LOC today-LOC PT PT deer-
GEN lie-TENT2-ATTR
Will the deer lie at the foot of [the mountain] of the deity Iyapiko today,
too? (MYS XVI: 3884)

13
I would like to thank Anton Antonov, who pointed out the examples from MYS X: 2224,
XI: 2763, and XVI: 3884 in his recent presentation (Antonov 2004). I disagree with his
analysis of nô-ra-ni in MYS XI: 2763 as ‘field-LOC-LOC’, as it seems to me that in the
context of this poem it should be taken as ‘field-PLUR-LOC’ or ‘field-DIM-LOC’.
58———Koreo-Japonica

AK-Î-TU MÎ-KAMÏ tö OPO-YA-SIMA-N-KUNI SIRASIMÊS-U

SUMÊRA-NKA OPO-MYI-KÖTÖ-ra ma tö NÖR-I-TAMAP-U OPO-MÎ-
KÖTÖ-wo UNKÖNAPAR-I-PANPÊR-U MÎKWO-TATI OPO-KÎMÎ-
TATI MÔMÔ TUKASA-NÖ PÎTÖ-NTÖMÖ AMEY-NÖ SITA-NÖ
TAMÎ MÖRÖ-MÖRÖ KÎK-Î-TAMAPË-YÖ
open-NML-GEN/LOC HON-deity DV great-eight-island-GEN-country
rule(HON)-ATTR emperor-POSS great-HON-deity-DAT according DV
say(HON)-INF-HON-ATTR great-HON-thing-ACC gather-INF-POL-
ATTR prince-PLUR great-lord-PLUR hundred office person-PLUR
heaven-GEN under-GEN people many-many listen-INF-HUM-IMP
princes, lords, officials of hundred offices, and the common people of the
land, all of you listen to the Great Edict according to the emperor — Great
Deity who rules the Great Country of Eight Islands as a Manifest Deity
(SM 1)
There is only one example of -ra in Eastern Old Japanese, but it is
found in a poem with typical eastern features (cf. the attributive in -wo
below):

KWO-ra pa ap-an-am-wo pyitö-ri nömiy s-i-te

girl-DAT TOP meet-DES-TENT-ATTR one-CL PT do-INF-GER
being absolutely alone, [I] wish to meet [this] girl (MYS XIV: 3405)

There is no dative-locative case marker -ra in Ryukyuan, and that

together with the low frequency and short-lived nature of attestations
makes OJ -ra a good candidate for an early loan from Korean.
2.1.1.4.2 Middle Korean dative -s-kuy, Japonic directive -[N]-kari
The Middle Korean parallel to the proto-Japonic directive *-kari (see
below) is not so obvious at first glance. I believe it is the Middle Korean
locative-directive -ay/-ey/-oy/-uy, which is well attested in Middle Korean
texts. The process r > Ø/__i in the history of Korean can be justified
internally, cf. OK NWUli ‘world’ > MK :nwuy ‘id’., OK NAli ‘river’ >
MK :nayh ‘id’., OK nyelim ‘lord’ > MK :nim ‘id’., so the reconstruction of
MK -ay/-ey/-oy/-uy as *-ari/*-eri/*-ori/-*uri does not present any problem.
The reconstruction of the initial consonant may be more problematic, but I
think it stands on firm ground due to the following data from Old Korean.
In at least two instances in the Hyangka texts the case marker that cor-
responds to Middle Korean -ay/-ey is written with the character 希 *xïy,
which starts with a velar /x/ in Early Middle Chinese. I tentatively recon-
struct this case marker as Old Korean -huy (with no additional forms
adjusting to vowel harmony, following Martin’s proposal about the lack of
vowel harmony in Old Korean [Martin 2000: 1-23]). The development of
Proto-Korean *-k- > -h- in all environments except after sonorants, where
*-k- > -G- takes place, is consistent with the new version of the lenition
theory (1.1.3.3 above). In combination with Proto-Korean *-l- loss
 Morphological Comparisons———59

preceding /i/, 14 this proposal allows a reconstruction of *-kuri as a Proto-

Korean locative-directive marker, with its development into -huy in Old
Korean and -ay/-ey/-oy/-uy in Middle Korean. It is also possible that this
Old Korean locative-directive -huy survives as a Middle Korean dative
marker -kuy, added to the locative-directive marker -oy/-uy or the genitive
marker -s, and -ngey (< *-n-key), added to the locative-directive markers
-oy and -uy (Yi Swungnyeng 1961: 206-210). The reconstruction of the
Proto-Korean form as *-(n)-kuri leaves a problem of vocalic cor-
respondence PK *u: PJ *a, with Proto-Japonic *-na-kari. Although this
correspondence is rare, there are other instances of it, cf. MK kùlyèkí ‘wild
goose’, OJ kari ?2.4 ‘id’.; MK tùlúh ‘field’, OJ ta 1.3a ‘rice field’ < *taa <
**tara. Examples from Old Korean Hyangka texts:

CIpwo PAhwo KOSO-huy TUL-um SWON AMSYWO NWOH-Osi-kwo

purple rock edge-LOC hold-NOM hand cow leave-HON-GER
leaving the cow that [I] lead at the edge of the purple rock (Hyangka II: 2)

ilGo NAli-ci COYWYek-huy LANG-la TUL-Gi-ki TAW-Osi-wo-n

MOSOm-oy KOs[o]-γïl CWOS-no-ra-cy-e
IlGo river-GEN pebble-LOC [Hwa]rang-DAT hold-PASS-NML like-
HON-MOD-PERF/ATTR heart-GEN limit-ACC follow-PRES-?-?-INF
following the limits of [my] heart, that is like being held by the [Hwa]rang
in the river bed (lit.: in the pebbles) of the river IlGo (Hyangka IV: 6)

The directive case marker -Nkari is not frequent in Western Old

Japanese: it occurs only eleven times, always in the Man’yōshū (vol. VII-
IX). It is quite possible that it was already on its way to extinction in Old
Japanese. -Nkari follows only personal pronouns and nouns denoting
people. None of the existing Old Japanese grammars discusses it, although
it is briefly mentioned in (JDB 1967: 232) and (Ōno 1990: 348). The
traditional etymology deriving -Nkari from -Nka ar-i POSS exist-NML
(Takagi et al. 1957-62.2: 209), (Martin 1990: 498) is controversial and not
universally accepted even among Japanese scholars (cf. Ōno 1990: 348). In
Western Old Japanese texts there is no uncontroversial phonetic spelling of
the case marker -Nkari; in most cases it is written with the character 許,
which probably represents a truncation from its Japanicized reading (kun-
yomi): (Npa)kari. The reading -Nkari, is, however, confirmed by phonetic
spellings found in the Eastern Old Japanese texts.

IMÔ-RA-NKARI WA-NKA YUK-U MÎTI

beloved-DIM-DIR I-POSS go-ATTR road
the road, on which I go to [my] beloved (MYS VII: 1121)

14
See the detailed discussion of this process in example (19) of Whitman’s lexical
comparisons in chapter 3.
60———Koreo-Japonica

KÖ-YÔPÎ ka KÎMÎ-NKA WA-Nkari K-ÎMAS-Am-u

this-night PT lord-POSS I-DIR come(INF)-HON-TENT-ATTR
Is it tonight that [my] lord will come to me? (MYS VIII: 1519)

IMÔ-Nkari YAR-AM-U MÔMÔNTI TA-WOR-Ana

beloved-DIR send-TENT-ATTR maple leaves hand-break-DES
I would like to break [with my hand] maple leaves to send to [my] beloved
(MYS IX: 1758)
There are four examples of the directive case marker -Nkari in Eastern
Old Japanese texts, all written phonetically.

imö-Nkari tö [i]p-ë-Npa
beloved-DIR DV say-EV-CON
as [I] said that I [would go] to [my] beloved (MYS XIV: 3356)

ika nar-u se-na ka wa-Nkari kö-m-u tö ip-u

how be-ATTR beloved-DIM PT I-DIR come-TENT-FIN DV say-ATTR
what kind of beloved [is he], who says that [he] will come to me? (MYS
XIV: 3536)

kökörö nömiy imö-Nkari yar-i-te wa pa kökö-ni s-i-te

heart PT beloved-DIR send-INF-GER I TOP here-LOC do-INF-GER
I [have] to stay here, sending just [my] heart to [my] beloved (MYS XIV:
3538)

iNtu-yu kamö kanasi-kî se-rö-Nka wa-Nkari kayôp-am-u

where-ABL PT beloved-ATTR husband-DIM-POSS I-DIR visit-TENT-
ATTR
From where will [my] beloved husband come to me? (MYS XIV: 3549)

Although MYS XIV: 3356 and XIV: 3549 do not have any typical
Eastern Old Japanese features, MYS XIV: 3536 includes the diminutive
suffix -na that occurs only in Eastern Old Japanese (Vovin 2005a: 212),
and MYS XIV: 3538 involves the usage of the unextended stem wa ‘I’ in
isolation, which is another typical Eastern Old Japanese feature. Thus, we
can rather safely conclude that EOJ -Nkari should be a native element, and
not a loan from Western Old Japanese.
In Ryukyuan, there is a directive case marker -kai attested in the Ryūka
(Nohara 1986: 27), as well as the locative-directive -nakai/-Nkai/-kai in the
Shuri dialect and the dative-locative -nakai in the Kumejima dialect
(Nohara 1986: 66-69, 126). The Kumejima dialect also exhibits -naa as a
locative. The following examples are from the Shuri and Kumejima
dialects:

Shuri
Naafa-Nkai ich-uN
 Morphological Comparisons———61

Naha-DIR go-FIN
[I am] going to Naha (Nohara 1986: 66)

ama-Nkai mayaa-nu u-N

there-LOC cat-GEN exist-FIN
There is a cat (Nohara 1986: 67)

Yamatu-Nkai ich-uN
Japan-DIR go-FIN
[I am] going to Japan (Nohara 1986: 68)

Kumejima
yaa-nakai ʔu-N na
home-LOC exist-FIN PT
[I] am at home (Nohara 1986: 126)

Although Ryukyuan -nakai/-Nkai is attested only in Central Ryukyuan,

it is unlikely to be a loan from Japanese, since there is no MJ -gari < OJ
-Nkari, and a direct loan from OJ into Ryukyuan is not realistic. Since in
Ryukyuan r > Ø/__i, Ryukyuan -nakai/-Nkai perfectly corresponds to Old
Japanese -Nkari. As implied by the Ryukyuan uncontracted form -nakai,
unless it is an innovation, the -N- part of Old Japanese -Nkari and
Ryukyuan -Nkai probably represents the locative case marker -na, attested
in Eastern Old Japanese and various Ryukyuan dialects. This further attests
to the authenticity of Ryukyuan -nakai/-Nkai and allows us to reconstruct
the directive marker as a bimorphemic PJ *-na-kari. This leaves *-kari as a
potential Proto-Japonic directive marker.
Nevertheless, we end up with only this one case marker that can be
reconstructed for Proto-Japonic and at the same time can be compared with
Korean. Korean and Japonic case marker paradigms are very different, and
it is impossible to reconstruct a single original paradigm that would
account for both systems. In some cases, as I demonstrated above,
similarities between Korean and Japonic can be better explained as loans,
since the case markers in question are not really Japonic, but specifically
Japanese or even Western Old Japanese. A single occurence of a directive
locative marker is unlikely to prove a genetic relationship.
The following chart includes primary case markers in Western Old
Japanese, with parallels in Eastern Old Japanese, Ryukyuan, Old Korean,
and Middle Korean:
62———Koreo-Japonica

Chart 21:
Primary case markers in Western Old Japanese from a comparative
perspective
WOJ EOJ Ryuk. OK MK
Active -i — — -i (erg.) -i (nom.)
Possessive -Nka -Nka -ga — —
Genitive -nö -no -nu — —
Genitive- -tu — — -cï (gen.) -s (gen.)
Locative
Dative- -ni -ni -ni — —
Locative
Locative -ra -ra — -ra(ng) —
Accusative -wo -wo -wo — —

As one can see from the chart, such Western Old Japanese case markers
as possessive -Nka, genitive -nö, dative-locative -ni, and accusative -wo are
attested in various branches of the Japonic family, but do not have any
parallels in Korean. Note, however, that those Western Old Japanese case
markers with Korean parallels are not attested in other branches of Japonic,
with the exception of dative-locative -ra, which has a singular attestation
in Eastern Old Japanese. Such a distribution certainly speaks in favor of an
areal rather than a genetic relationship between Korean and Japonic.

2.1.2 Pronouns
In this section, I will discuss Korean and Japonic personal, interrogative,
and demonstrative pronouns.
2.1.2.1 Personal pronouns
Unfortunately, Old Korean personal pronouns are known only in
semantographic writing. Therefore, we have no choice but to rely on the
information we can secure from Middle Korean and Korean dialects, but
the latter, to the best of my knowledge, do not provide any important
information for the reconstruction of Proto-Korean. 15 Essentially, we have
MK nà ‘I’, nè ‘thou’, wúlí ‘we’, and nèhùy ‘you’ at our disposal. Both MK
nà ‘I’ and nè ‘thou’ exhibit highly irregular accentuation in their
declension (Kim Wancin 1973: 62), which may point to a complex
phonological history beyond the limits of current knowledge.
In contrast to Korean, the picture in Old Japanese is more colorful, as
there are several different personal pronouns: first person wa ~ ware, a ~
are, na, and marö; and second person na ~ nare and mîmasi ~ imasi ~ masi.
There is no clear-cut distinction between singular and plural, as in Korean,
although in most cases the plural is expressed by the extended stems ware,
are, and nare (Vovin 2005a: 220, 234, 247). Among the aforementioned
personal pronouns, marö ‘I’ is almost a hapax legomenon, appearing twice

15
The data on the dialect forms of Korean pronouns na ‘I’ and ne ‘thou’ are provided in
Choy (1978): 240-241, 244-245.
 Morphological Comparisons———63

in one text that exists in two variants (Vovin 2005a: 244-245), so it will not
be discussed below.
Cognates of OJ wa- and a ‘I’ are attested throughout Japonic, but OJ na
appears only twice in Western Old Japanese and twice in Eastern Old
Japanese. It has no cognates in Ryukyuan, and does not occur in any other
varieties of Japonic. Therefore, I believe that it represents a loan from
Korean na ‘I’. Examples of OJ na ‘I’:

Western Old Japanese

na SE n-ö KÎMÎ
I beloved DV-ATTR lord
my beloved and lord (MYS XVI: 3885)

na OTÖ n-ö mî-kötö

I younger brother DV-ATTR HON-thing
my dear younger brother (MYS XVII: 3957)

Eastern Old Japanese

na se n-ö kô-Nka ya-sô sima kakur-i
I beloved DV-ATTR child-POSS eight-ten island hide-INF
my beloved hides [behind] eighty islands... (FK 8)

na se n-ö kô
I beloved DV-ATTR child
my beloved (MYS XIV: 3458)

Second person pronouns in Old Japanese present a more complicated

picture, where at first glance the genetic affiliation of doublets does not
seem to be as clear-cut as in the case of first person pronouns; therefore a
somewhat more detailed look at the textual examples is necessary.
The second person pronoun (mî-/i-)masi is fairly well attested in both
Western and Eastern Old Japanese, but not in Ryukyuan. The original form
of that pronoun is probably just masi, as already pointed out by Yamada
Yoshio (1954: 95), mî- and i- being prefixes that can possibly be identified
with an honorific prefix mî- (Vovin 2005a: 256). Examples:
Western Old Japanese
imasi mô WARE mô KÖTÖ NAR-UNPË-SI YA
you PT I PT thing be-DEB-FIN PT
Should [it] be something for both you and me? (MYS XI: 2517)

AMEY-NÖ SITA pa WA-NKA KÔ imasi-ni SANTUKË-TAMAP-U

heaven-GEN bottom TOP I-POSS child you-DAT give (INF)-HON-FIN
[I] give to you, my child, the land under the Heaven (SM 29)
mîmasi-nö TITI tö IMAS-U SUMÊRA-MÎKÖTÖ-nö mîmasi-ni TAMAP-
Î-si AMË-NÖ SITA
64———Koreo-Japonica

you-GEN father DV be(HON)-ATTR emperor-deity-GEN you-DAT give

(HON)-INF-PAST/ATTR heaven-GEN bottom
the land under Heaven that the emperor-deity, who is your father, gave to
you (SM 5)

mîmasi MÎKÔ-nö YÖPAPÎ-nö YÔWA-KÎ-ni

you prince-GEN age-GEN weak-ATTR-LOC
because you, prince, were in your infancy... (SM 5)
Eastern Old Japanese
masi mô are mô yöti-wo sö möt-er-u
you PT I PT same age-ACC PT have-PROG-ATTR
both you and I have [children of] the same age (MYS XIV: 3440) 16

imasi-wo tanöm-î papa-ni taNkap-î-n-u

you-ACC trust-INF mother-DAT become estranged-INF-PERF-FIN
[I] became estranged from [my] mother, trusting you (MYS XIV: 3359)
The second person pronoun na is fairly well attested in both Western
and Eastern Old Japanese. It also appears in Ryukyuan, but there are some
problems with its attestation that I discuss below.

Western Old Japanese

na kösö pa yö-nö naNka pîtö
you PT TOP world-GEN long person
you, [the most] long[-lived] man in the world (KK 71)

na pa kîk-as-u ya
you TOP ask-HON-FIN PT
Shall [I] ask you? (NK 62)

Eastern Old Japanese

se-na na-tö puta-ri sa-NE-TE kuyasi mo
beloved-DIM you-COM two-CL PREF-sleep(INF)-SUB regrettable PT
it is regrettable that [I], my beloved, slept with you, two [of us] (MYS XIV:
3544)

nare ya pîmo tök-aNs-u ne-m-u

you PT cord untie-NEG-INF sleep-TENT-ATTR
Will you sleep without untying the cords [of your garment]? (MYS XIV:
3370)

16
This is a variant of MYS XIV: 3440, where the second pronoun masi is used instead of
nare ‘you’. In the basic version we find nar:

nare mô are mô yöti-wo sö möt-er-u

you PT I PT same age-ACC PT have-PROG-ATTR
both you and I have [children of] the same age (MYS XIV: 3440).
 Morphological Comparisons———65

Ryukyuan
The second person pronoun na is attested comparatively well in Ryukyuan.
There are no attestations in the Omoro sōshi or the Ryūka: the Classical
Ryukyuan attestation naa ‘you’ seems to be confined to Ryukyuan plays
(Hokama 1995: 472). However, there is no lack of attestations in the
modern dialects of the Northern and Central Ryukyus, where forms such as
Koniya naṃ; Namizato naa; Sesoko naa, naN; Shuri naa, etc., are found
(Uchima and Arakaki 2000: 358). The only questionable attestation in
Sakishima is in Hateruma, but Hateruma has the aberrant form daa, which
is probably not related. In all these dialects this second person pronoun
seems to have a similar function to the Shuri pronoun naa, which is a
familiar pronoun used toward older people of lower status (RGJ 1976:
399). 17 An example from the Shuri dialect:

naa ya ʔiçi mooca ga

you TOP when come/PERF PT
When did you come? (RGJ 1976: 399)
The distribution in the Ryukyus (no attestations in Sakishima) and its
mild honorific nature suggest that Ryukyuan naa 18 is a loan from Japanese,
especially since Proto-Ryukyuan has another second person pronoun *ʔe
or *ʔo, with no skewed distribution. Besides, Koniya naṃ and Sesoko naN
indicate PR *namu, which suspiciously resembles MJ namudi ‘thou’ rather
than OJ na. This leaves masi and na competing for the status of a Proto-
Japanese pronoun. Although there has been a traditional comparison of OJ
na and MK ne, both second person pronouns, I suspect that na is a loan
from Korean ne, while masi represents a true Proto-Japanese (but not
Proto-Japonic) pronoun. It is also strange for genetically related languages
to share second person pronouns, 19 but not to share first person ones: OJ
wa- < PJ *ban and MK na certainly cannot be related.

Chart 22:
Western Old Japanese personal pronouns from a comparative
perspective
WOJ EOJ Ryuk. OK MK
I1 wa- wa- waN — —
I2 a- a- a- — —
I3 na- na — ? na

17
Uchima and Arakaki gloss their Ryukyuan dialect attestation as modern Japanese anata
‘you’ (Uchima and Arakaki 2000: 358), which is a semi-formal pronoun used to relatives
(e.g., wife to her husband), or to lower-ranking people, when some minimal degree of
respect is maintained.
18
There has ben a recent speculative attempt to relate OJ a ‘I’ and MK na ‘id’. as a reflex
of PA *a ‘I’ (Starostin, Dybo, and Mudrak 2003: 225). This brave proposal is nothing but
an attempt to reconstruct things from above: there is no way to prove that OJ Ø- goes back
to anything but *Ø-.
19
Second person pronouns are much less stable and more prone to borrowing than are first
person pronouns.
66———Koreo-Japonica

thou1 na- na- ?na ? ne

thou2 masi masi — — —

The situation with personal pronouns, presented in the chart above, is

reminiscent of the situation with case marking. With the exception of the
first person pronoun na, which has attestations in Eastern Old Japanese and
the second person pronoun na, which has limited and dubious attestations
in Ryukyuan, Western Old Japanese personal pronouns that have cognates
in Ryukyuan and/or Eastern Old Japanese lack parallels in Korean.
Meanwhile, WOJ na ‘I’ has a parallel in Korean, but no cognates in other
branches of Japonic. The situation again indicates an areal, not a genetic,
relationship.

2.1.2.2 Interrogative pronouns

Like Old Korean personal pronouns, Old Korean interrogative pronouns
are attested only in semantographic writing, so they are of no help in
reconstructing Proto-Korean archetypes. The Middle Korean interrogative
pronouns are nwú ‘who’, mùsú, mùsúm, mùsúk ‘what’, mùsús ‘which’, ètúy
‘where’, ènú ‘which’, :encèy ‘when’, and :estyé, :esté, :estyéy ‘how’. We
can see that there are different series of Middle Korean interrogative
pronouns based on different roots. MK nwú ‘who’ apparently stands in
isolation. MK mùsú, mùsúm, mùsúk ‘what’ and mùsús ‘which’ are based
on the same root, mùsú. Since MK ètúy does not show the lenition -t- > -l-
in the typical lenition environment …°Cu/o (Martin 1996: 2-3), it should go
back to PK *èntúy, with a nasal sonorant blocking the lenition (Vovin
2003b: 89-103). This reconstruction allows us to see that PK *èntúy
‘where’ is in fact bimorphemic, going back to ènú ‘which’ and túy ‘place’:
*ènú + túy > *èntúy > ètúy. Thus, MK ètúy ‘where’, ènú ‘which’,
and :encèy ‘when’ belong to the same series as ènú.
MK :estyé, :esté, :estyéy ‘how’, probably going back to *e-is-ti ‘how-
exist-ADV’, exhibits one more series *e-, which may be related to the ènú
series. But we do not have any internal evidence for segmentation of MK
ènú as *e-nu, and no reliable internal evidence for the segmentation of
MK :estyé, :esté, :estyéy ‘how’ as *e-is-ti.
The Japonic interrogative pronouns also look suspiciously hetero-
geneous, similar to their Korean counterparts. Western Old Japanese has
the following interrogative pronouns: ta- ~ tare ‘who’, nani ‘what’, ika
‘how’, iNtu ~ iNtuku ~ iNtuti ‘where’, iNture ‘which’, itu ‘when’, naNsö,
naNtö ‘why’, iku ~ ikura ~ ikuNta ‘how many, how much’. These are
paralleled by EOJ ta- ~ tare ‘who’, aN- ‘what’, ika ‘how’, iNtu ~ iNtuti ~
iNtusi ‘where’, iNture ‘which’, 20 itu ‘when’, aNse ‘why’, and PR *ta- ~
*taro ‘who’, *nau ‘what’, *e-ka ‘how’, *entu-ma ‘where’, *entu-re ‘which’,
*etu ‘when’, 21 *e-ku ‘how many, how much’. Thus, we can also establish
20
A hapax legomenon from MYS XX: 4392.
21
There is no internal Japonic evidence for the segmentation of PR *etu ‘when’, in contrast
to PR *e-ka ‘how’ and *e-ku ‘how many, how much’ (Vovin 2005a: 317, 330, 332),
contrary to Whitman (2001: 1).
 Morphological Comparisons———67

several series for Japonic interrogative pronouns. First, PJ *ta ‘who’ stands
in isolation. Second, we can reconstruct the *n-anu- series, 22 represented
by ‘what’ and ‘why’, although the second is only reconstructable for Proto-
Japanese, not for Proto-Japonic. Third, there is the iNtu- < *entu- series,
which can be compared with Korean *en- ‘wh-’, found in MK ènú ‘which’,
ètúy < *en-tuy ‘where’, and e:ncèy < *en-ce[k]i ‘when’. This *entu- series
looks like a loan from Korean, because *entu- cannot be segmented as *en-
tu- on the basis of internal Japonic evidence. Finally, there is the *e- series
that may be compared with MK :estyé, :esté, :estyéy ‘how’, but no reliable
internal evidence exists for the segmentation of the latter, so the
relationship between the PJ *entu- series and MK e:styé, e:sté, e:styéy
‘how’ remains obscure. All this is summarized in the following chart.

Chart 23:
Western Old Japanese interrogative pronouns from a comparative
perspective
WOJ EOJ PR OK MK
who1 ta- ta- *ta — —
who2 — — — NWUki nwú
what1 nani 23 aN- *nau — —
what2 — — — ? mùsú
‘wh-’ iNtu- iNtu- *eNtu- eno ènú
how ika ika *e-ka ? e:styé

The Korean and Japonic words for ‘wh-’ and ‘how’ could possibly be
related, but the segmentation problems on both sides suggest an areal,
rather than a genetic, relationship. No genetic relationship can be
established between PJ *ta ‘who’ and *n-anu- ‘what’ on the one hand and
MK nwú ‘who’ and mùsú ‘what’ on the other, which is strange in the
context of a suggested genetic relationship. As with case marking and
personal pronouns, comparison of Korean and Japonic interrogative
pronouns points toward an areal relationship.

2.1.2.3 Demonstrative pronouns

Although both Proto-Japonic and Middle Korean have systems of
demonstrative pronouns that include the three degrees: proximal, mesial, 24
and distal, they do not coincide.

22
On the reconstruction of Proto-Japonic archetype as *n-anu- see Vovin (2005a: 309-313).
23
< PJ *n-anu-.
24
Reconstruction of the Proto-Japonic mesial form is problematic, because the Japanese
and Ryukyuan data do not agree.
68———Koreo-Japonica

Chart 24:
Proto-Japonic and Middle Korean demonstrative pronouns
PJ MK
proximal *kə- í
mesial PJN *sə-, PR *ʔo- kù
distal *ka- tyé

The demonstrative systems in Japanese and Korean seem to have been

functionally much more different in earlier times than they are now.
Japonic formally differentiated between deictic and attributive forms of
demonstrative pronouns throughout its history: WOJ köre pa yama nar-i
‘this is a mountain’, but könö yama ‘this mountain’; Shuri kuree yama ya-
N ‘this is a mountain’, kunu yama ‘this mountain’. Middle Korean did not
differentiate between deictic and attributive forms: MK i-non mwoy [i]la
‘this is a mountain’, i mwoy ‘this mountain’. Modern Korean, however,
does differentiate between these two forms: i-ke[s-un] san i-ta ‘this is a
mountain’, but i san ‘this mountain’. This distinction probably originated
in Modern Korean under Japanese influence.
Mainland Japanese spatial demonstrative pronouns ending in -ko: koko
‘here’, soko ‘there’ look like partial borrowings from Korean (cf. Ryukuan
ku-ma, ʔu-ma): MK ye-kuy ‘here’, ke-kuy ‘there’.
Recently, Frellesvig and Whitman, in a paper presented at the XVIth
meeting of the International Conference on Historical Linguistics in
Copenhagen in August 2003, offered an interesting proposal that makes the
Middle Korean and Proto-Japonic systems of demonstrative pronouns
comparable if the oppositions ‘participant — non-participant’ and ‘speaker
— non-speaker’ are introduced into pre–Old Japanese and Old Japanese,
respectively. Their proposal for the development of the system of
demonstrative pronouns from Proto-Japonic to Middle Japanese is as
follows (Frellesvig and Whitman 2003: 4):
Chart 25:
Evolution of Japonic demonstrative pronouns according to Frellesvig
and Whitman (2003)
Proximal Mesial Distal
MK i ku tye
PJ *i *ki *si
pre- *i *ki *si
OJ.a
Participant Non-participant
pre- *ki *si
OJ.b
Speaker Non-speaker
OJ kö (~i) sö
Proximal Mesial Distal
MJ ko so ka
 Morphological Comparisons———69

In my opinion, their proposal faces several insurmountable problems

despite its seeming elegance. First, while MJ ko and so refer to areas near
the speaker or the non-speaker, respectively, WOJ kö and sö indicate only
general proximity or remoteness, without any regard to the speaker or the
addressee’s position (Vovin 2005a: 272-276, 285-288). Second, we have
no means to verify the suggestion that in pre-OJ.b the demonstratives
referred to ‘participant’ or ‘non-participant’, simply because we have no
texts in pre–Old Japanese to base our judgment on. Third, a change in the
history of Japonic from a tripartite system to a binary and then back to a
tripartite is circular. Fourth, proposing a system for Proto-Japonic that is
virtually identical with the Middle Korean system indicates a reliance on
the Middle Korean data in the first place, and on the premise that those
systems must have been more similar in the past than they are in Old
Japanese and Middle Korean. But the relationship between Old Japanese
and Middle Korean is not proven, so we do not know whether the systems
were identical or not. Therefore, in order to avoid another circular
argument, one must demonstrate on strictly internal grounds that it is
possible to reconstruct a Proto-Japonic system of demonstrative pronouns
identical to the Middle Korean system of demonstrative pronouns, and not
to attempt a derivation of the former from the latter. Fifth, the only
evidence for PJ and pre-OJ.a *i ‘this’ is based on OJ ima ‘now’, which is
analyzed as i-ma ‘this interval’. OJ ima ‘now’ is likely to be segmentable
on the basis of the comparison with Ryukyuan nama ‘id’., but we have no
internal evidence to prove that OJ i- and Ryukyuan na- are proximal
demonstratives. Sixth, and most important, the idea that pre-OJ.b and Old
Japanese had just a binary system of demonstratives is based on
Frellesvig’s suggestion that the Old Japanese distal demonstrative ka- ‘that
over there’ is an innovation (2003, personal communication). OJ ka- is
attested in Old Japanese texts very infrequently. The following are the
exhaustive examples:

Western Old Japanese

kîmî-Nka mî-pune kamö kare
lord-POSS HON-boat PT that
[Is] that [my] lord’s boat? (MYS XVIII: 4045)
TA SÖ KARE WARE-WO NA-TÖP-Î-SÖ
who PT that I-ACC NEG-ask-INF-do
Do not ask me: ‘Who is that?’ (MYS X: 2240)

TA SÖ KARE tö TÖP-ANPA
who PT that DV ask-COND
if [someone] asks: ‘Who is that?’ (MYS XI: 2545)

I also think that the following Western Old Japanese examples include
the stem ka- of this pronoun:
70———Koreo-Japonica

ap-as-i-si womîna ka möNka tö wa-Nka mî-si kwo-ra ka-ku möNka tö

meet-HON-INF-PAST1/ATTR woman that PT DV I-POSS see(INF)-
PAST1/ATTR girl-DIM thus-INF PT DV
[I] wish that [much] the woman [I] saw; [I] wish this [much] the girl I saw
(KK 42)

ka n-i ka-ku n-i posi-kî mani-mani sika n-i pa ar-aNsi ka

that DV-INF thus-INF DV-INF be desirable-ATTR according that way
DV-INF TOP exist-NEG/TENT PT
[acting] that [way] and this way, according to [your] desires, it cannot be
that way, can it? (MYS V: 800)

ka yuk-ë-Npa pîtö-ni itöp-aye ka-ku yuk-ë-Npa pîtö-ni nikum-aye

that go-EV-CON person-DAT avoid-PASS(INF) thus-INF go-EV-CON
person-DAT hate-PASS(INF)
when [they] go that [way], [they] are avoided by people, and when [they]
go this way, [they] are hated by people... (MYS V: 804)
Eastern Old Japanese
kanö kwo-rö-tö NE-Nsu ya nar-i-n-am-u
that girl-DIM-COM sleep-NEG/INF PT become-INF-PERF-TENT-ATTR
Will it become [so that I] will not sleep with that girl? (MYS XIV: 3565)

Such poor attestation is a puzzle, and it might be used as evidence in

favor of Frellesvig’s point of view. However, it is necessary to remember
that low frequency does not necessarily suggest innovation; it might also
indicate an archaism. In order to decide whether a form is an innovation or
an archaism we must look outside Old Japanese at other Japonic dialects
that are not within the same Central Japonic group.
The Ryukyuan data present an interesting picture in the comparative
perspective. Most of the Northern and Central Ryukyuan dialects as well
as Old Ryukyuan, have the distal demonstrative pronoun ʔari ~ ʔanu,
based on the stem ʔa-. This stem does not have a counterpart in Old
Japanese, but has parallels in Middle Japanese a- ~ are ‘that over there’
and later historical stages of Japanese. Only in dialects of the South
Ryukyus do we find the distal demonstrative pronoun kari ~ kanu, based
on the stem ka-, such as Agarinakasone kai, Yonaha kari, Tonoshiro kari
(Uchima and Arakaki 2000: 359), Psara kai ~ kari, Ikema kari, Ishigaki
kari (Hirayama 1966: 280), and Yonaguni kari (Hirayama 1967: 195).
Since direct borrowing from Old Japanese into Southern Ryukyuan is out
of the question, this distribution suggests that the ka- forms in the
Ryukyus are original, and that the ʔa- forms were probably spread by
language diffusion from Classical Japanese via Old Ryukyuan.
In any case, the fact that distal ka- is found in Old Japanese, Middle
Japanese, and Ryukyuan satisfies our requirement for considering a form
to be Proto-Japonic. Why it has such a low frequency in Old Japanese is
likely to remain a mystery, but we should not treat Old Japanese as if it had
 Morphological Comparisons———71

the same role for Japonic as Latin does for the Romance languages. In
addition, we should not forget that the development from Old Japanese to
Middle Japanese was not strictly linear: both Old Japanese and Middle
Japanese are based on geographically close but not quite identical dialects.
Thus, I reconstruct both Proto-Japonic proximal *kə and distal *ka on the
basis of Middle Japanese (also partially Old Japanese) and Proto-
Ryukyuan evidence. The reconstruction of the Proto-Japanese mesial is
more problematic, since PJN mesial *sə and PR mesial *ʔo do not
correspond to each other.
We probably can reconstruct tripartite systems of demonstrative
pronouns for both Korean and Japonic, although the reconstruction of the
mesial in the latter is problematic. However, it is quite clear that these
systems are not comparable; therefore, they cannot offer any evidence for a
genetic relationship.
2.2 ADJECTIVES
Korean does not really distinguish between verbs and adjectives, the latter
being in fact quality verbs: khu-ta ‘be big’, pulk-ta ‘be red’, etc. Modern
Japanese has both adjectives (like sizuka-na ‘quiet’) and quality verbs (like
aka-i ‘be red’). But in Old Japanese there are many adjectives that behave
like adjectives of the ‘Altaic’ languages, words that can modify a noun
without the attributive marker necessary for quality verbs in the later stages
of Japanese, e.g., awo ‘blue, green’, kata ‘hard’, kupasi ‘beautiful’, kusi
‘precious’, muna ‘empty’, naNka ‘long’, pîrö ‘broad’, pîsa ‘long’, putô
‘majestic, great’, sakasi ‘wise’, sira ~ sirô ‘white’, opo ‘big, great’, taka
‘high’, tökö ‘eternal’, töyö ‘abundant’, tô ‘sharp’, utukusi ‘beautiful’, waka
‘young’, etc. Some examples from texts:

awo na mö kîNpï pîtö-tö tömö n-i si tum-ë-Npa tanôsi-ku mö ar-u ka

green vegetables PT Kibi person-COM together DV-INF PT pick-EV-
CON delightful-INF PT exist-ATTR PT
Isn’t it delightful when [I] pick green vegetables together with the girl
from Kibi? (KK 54)

Kôsi-nö kuni-ni sakasi mê-wo ar-i tö 25 kîk-as-i-te kupasi mê-wo ar-i tö kîk-
ös-i-te
Kosi-GEN province-LOC wise woman-ABS exist-FIN DV hear-HON-INF-
SUB beautiful woman-ABS exist-FIN DV hear-HON-INF-SUB
[Okuni-nusi] heard that there is a wise woman in the northern province,
heard that there is a beautiful woman (KK 2)

könö kusi mî-tama

this precious HON-jewel
these precious jewels (MYS V: 814)

25
A declarative sentence followed by the citation particle tö must be treated as a main
clause because its predicate is always in final form.
72———Koreo-Japonica

pa-N-pîrö yu tu ma-tuNpakî
leaf-GEN-broad sacred DV INT-camellia
a sacred true camellia with broad leaves (KK 57)

sirô KAMÎ-maNte-ni OPO KÎMÎ-ni tukapë-matur-e-Npa

white hair-TERM-LOC great lord-DAT serve(INF)-HUM-EV-CON
when [I] serve the emperor (lit: great lord) until [I get] white hair (MYS
XVII: 3922)
In Middle Japanese marking adjectives with the attributive and final
predicate became obligatory, e.g., siro-ki nami white-ATTR wave ‘white
waves’ and nami siro-si wave white-FIN ‘waves are white’. Cf. OJ sira
namî ‘white waves’ (surviving only in Middle Japanese poetry). Although
Western Old Japanese has both attributive -kî and final predicate -si (also
attested in Eastern Old Japanese as -ke and -si), they can be reversed, as
pointed out by Martin (1987: 806-807). In other words, -kî can be used as a
marker of final predication, and -si as an attributive. Some examples:
naNTÖ SÖNÖ TAMA-NÖ TE-ni MAK-Î-NKATA-kî
why that jewel-COMP hand-LOC wind-INF-difficult-FIN
why is that [jeweled bracelet] difficult to wind around the arm that is like a
jewel? (MYS III: 409)

pana nipop-î ter-i-te tat-er-u pa pasi-kî ta-Nka tuma

flower be fragrant-INF shine-INF-SUB stand-PROG-ATTR TOP be
lovely-FIN who-POSS spouse
The [girl], who is standing, fragrant and shining [as] a flower, is lovely.
Whose spouse [is she]? (MYS XX: 4397)

ika n-i ka yuk-am-u karite pa na-si-ni

how DV-INF PT go-TENT-ATTR food supply TOP no-ATTR-LOC
because [I] do not have a food supply, how would [I] go? (MYS V: 888)
yapô ni yö-si i-kîNtuk-î n-ö mîya
eight hundred soil good-ATTR PREF-build-INF DV-ATTR palace
a palace built on an eight hundred [times] good soil (KK 100)

UMA-SI KUNI sö AKÎNTU SIMA yamatö-nö KUNI PA

beautiful-ATTR PT dragonfly island Yamato-GEN country TOP
Oh, [what a] beautiful country [is] Yamato, the land of dragonfly islands!
(MYS I: 2)
Neither -kî nor -si is attested in Ryukyuan, 26 which together with the
instability of their functions in Old Japanese, indicates that both represent a
Japanese innovation. It looks like they were introduced into the language
26
The meager evidence for these markers from Old Ryukyuan that Martin discusses and
rightfully rejects (1987: 806) is likely to represent loans from mainland Japanese and is
certainly not sufficient to establish a genetic relationship.
 Morphological Comparisons———73

under Old Korean influence. Moreover, it is possible that both were

borrowed from Old Korean.
OJ -si is likely to be a loan from the OK irrealis attributive marker -l
(phonetically probably voiceless [lh], as indicated by the character 尸 with
which it was written): 27

慕理尸心未行乎尸道尸
KUli-l MOSOm-i NYE-wo-l KIl
long for-ATTR/IRR mind-GEN go-MOD-ATTR/IRR way
the way that [my] mind, longing for [you], is going (Hyangka I: 7)

臣隱愛賜尸母史也
SIN-un TOSO-si-l Esi I-LA
retainer-TOP love-HON-ATTR/IRR mother be-FIN
Retainers are loving mothers (Hyangka III: 2)

Since WOJ -kî is paralleled by EOJ -ke, the reconstructed form has to
be PJN *-ke. It is possible that PJN *-ke is a loan from OK -kuy, the
attributive marker of quality verbs:

東京明期月良
TWONG-KYENG POLK-kuy TOLAL-la
Eastern Capital bright-ATTR moon-LOC
At the bright moon in the Eastern Capital… (Hyangka V: 1)

There are two obvious problems here. First, although OK 明期 POLK-

kuy ‘bright’ is clearly used with an attributive function here, this is the only
example where 期 -kuy occurs as the attributive marker of a quality verb in
Old Korean. Second, the -k- portion of -kuy may really belong to the stem
POLK- ‘bright’. In both of the aforementioned cases, 心未 MOSOm-i
‘mind-GEN’ in Hyangka I: 7, where the phonetically used character 未
renders both the final consonant of the stem and a suffix, and 月良
TOLAL-la ‘moon-LOC’ in Hyangka V: 1 above, where only the suffix is
written phonetically, there is no clear demarcation in writing between the
stem and the suffix.

2.3 VERBAL MORPHOLOGY

In this section I discuss the copula tö and some verbal markers, criticizing
my earlier work to a great extent.

2.3.1 The odd nature of the copula tö

The evidence for the copula tö outside Western Old Japanese and the later
stages of Central Japanese is slim at best, while the evidence for the
quotative tö, the comitative case marker -tö, and the subordinative gerund
27
There are other apparent loans from Old Korean into Japanese that reflect the
correspondence OK -l ~ OJ -si, e.g., MK kàlàp ‘oak’ ~ OJ kasi ‘id.’, MK kál ‘cangue’
~ MJ kasi ‘id.’ See #101-102 in Chapter 3 for details.
74———Koreo-Japonica

-te, which are sometimes viewed as extensions of the copula tö (Frellesvig

2001) is plentiful.

The first oddity

Let us start with the copula derivatives: nar-i (< *n-i ar-i DV-INF exist-
FIN) and tar-i (< *tö ar-i DV[INF] exist-FIN). If they were really
subjective and objective, as is usually claimed, it is odd that the distinction
between them is textual: tari is found almost exclusively in kunten texts,
while nari reigns supreme in literary texts. Why do they have this kind of
distribution?

The second oddity

There are thirty-two instances of the copula infinitive n-i and only three
cases of the copula tö in MYS XIV. In MYS XX there are twenty-one
examples of n-i and only three of tö.

The third oddity

The three clear cases of the copula tö in MYS XIV are found in the poems
MYS XIV: 3364, 3390, and 3400. Among these, MYS XIV: 3364 (Sagami)
and MYS XIV: 3390 (Hitati) are from Area B, and MYS XIV: 3400 (Sinano)
is from Area C. There are no ‘misspellings’ and no other typical features of
Eastern Old Japanese in MYS XIV: 3390 and 3400. In MYS XIV: 3364
particle mö is ‘misspelled’ as mô, but this type of misspelling was already
typical of Western Old Japanese by the middle of the eighth century.
Therefore, the chances are that all three poems were ‘edited’ and rewritten
in Western Old Japanese either by person(s) who collected them, or by
Yakamochi himself. MYS XX also has only three clear examples of the
copula tö. These are found in the poems MYS XX: 4375, 4404, and 4420.
Among these, only MYS XX: 4375 and 4420 are clearly written in Eastern
Old Japanese and not in Western Old Japanese. MYS XX: 4404 does not
have any Eastern Old Japanese features apart from the ‘misspelling’ of 家
/kê/ as 気 /kë/ in the verbal form tayenikëru ‘has broken off’. Although this
misspelling would be very unusual even for late Western Old Japanese, I
believe that it is not by itself sufficient evidence to classify this poem as
written in Eastern Old Japanese. Therefore, the presence of the copula tö in
Eastern Old Japanese is really verified by only two poems: MYS XX: 4375
and 4420. The revised combined ratio of n-i to tö in Eastern Old Japanese
then becomes 53 : 2. The pattern of distribution is obviously skewed, and I
think that the best explanation for these two cases of tö in Eastern Old
Japanese is that they represent borrowing or diffusion.

The fourth oddity

While locative-genitive -tu (as in nipa-tu töri ‘garden-LOC/GEN bird’) is
attested in Eastern Old Japanese songs, the attestations are not convincing,
because they are essentially limited to two repeatedly occurring contexts:
muka-tu wo ‘hill on the opposite side’ and okî-tu N ‘open sea-LOC/GEN
N’. They occur primarily either in poems from Areas B or C and/or poems
 Morphological Comparisons———75

that lack any typical Eastern Old Japanese features. This oddity should not
detain us further (for details see 2.1.1.2.1), because it is not relevant to
what we are discussing, but note that adnominal tu (as is in asa-tu ki), a
form of the copula t- does not appear in Eastern Old Japanese at all. Some
scholars cite momo Ntu sima ‘hundred islands’ from MYS XIV: 3367 as
momo tu sima, but this ignores the textual evidence: no MYS XIV
manuscript has tu there. It is consistently spelled with 豆 = [Ntu], and this
character is used as [Ntu] in numerous other examples in MYS XIV. I
suspect that the reading momo tu sima originated very late and was further
strengthened by Sasaki Nobutsuna’s authority. I think that -Ntu represents
a non-productive collective -Ntu, also found in yörö-Ntu ‘10,000/many’
and possibly in mîNtu ‘water’. MYS XIV: 3367, also a poem from Sagami,
has no Eastern Old Japanese features, just two unetymological spellings of
/mô/ and /mö/ typical of the Western Old Japanese of the period.
Possible conclusions are: (1) the copula tö never existed in Eastern Old
Japanese; (2) it spread there by diffusion from Western Old Japanese; (3) it
was on the brink of extinction in Eastern Old Japanese. I think that solution
(1) or (2) is the most plausible, since each is supported by the textual
distribution of tö within Eastern Old Japanese. The situation in Ryukyuan
is of paramount importance ― the existence of the copula tö in Ryukyuan
supports (3), its absence supports (1) and (2).

The fifth oddity

It may seem that copula tu exists in modern Shuri: RGJ gives two
examples (besides the comitative case marker tu) that can be recognized as
a copula (RGJ 1976: 525). However, I have examined several Shuri texts I
have, and while the comitative tu is frequent, strangely enough, the copula
tu is not present. I would like to add that the adnominal copula nu and
adverbial ni are present in these texts. That makes me think that the copula
tu is at least comparatively rare in Shuri. We also should keep in mind that
Shuri is the area probably influenced the most by mainland Japanese.

The sixth oddity

In Old Ryukyuan, as far as I can tell, the copula TO 28 is attested only once
in OS XIII: 854. All other examples come from Classical Ryukyuan Ryūka
and Kumiodori texts that were influenced much more by mainland
Japanese language and poetry patterns than were the Old Ryukyuan texts
in the Omoro sōshi. At the same time the quotatives TE and TETE (= tote)
frequently appear in Old and Classical Ryukyuan texts with various
spellings. These different spellings exhibit a bewildering variety: TE, TI (=
to) (Hokama 1995: 437), TETE, TETI, DETE, REITI (= tote) (Hokama
1995: 441). The absence of a variant TOTE suggests that they are not loans
from mainland Middle Japanese.

28
I transliterate Old Ryukyuan kana spellings with capitals.
76———Koreo-Japonica

The seventh oddity

The cleanest test for the existence of cognates of the Western Old Japanese
copula tö in Ryukyuan should certainly be based not on Old Ryukyuan or
Classical Ryukyuan, but on Southern Ryukyuan, which received the least
influence from Japanese. I decided to use Nevskii’s 1978 recordings on the
Miyako islands, done from the mid to late 1920s for this purpose, when the
Japanese influence on Miyako was still very weak (Okinawa only became
a prefecture in 1879, less than fifty years before Nevskii first set foot on
Miyako in 1924). I have not seen any traces of the copula tu there.
Furthermore, Shimoji Kazuaki, in his Miyako guntō go jiten, which has a
short sketch of the grammar, lists Miyako ti, quotational, and Miyako tu,
comitative case marker (Shimoji 1979: 172). But the most interesting item
is the following note: Miyako go de wa mizu to naru wo midzü ni narü to
iu (In the Miyako language [Japanese] mizu to naru ‘water copula to
become’ is said as mizu ni naru ‘water copula ni become’). I think one can
safely conclude that the copula tu does, or did not exist on Miyako. Of
course, one must look into other Sakishima dialects besides Miyako, but a
perusal of the comprehensive study by Nohara Mitsuyoshi (1998) reveals
that Ryukyuan tu as a copula appears only in the Northern and Central
Ryukyus. There are no traces of it to the south of Kumejima and Okinawa
islands. Even in the majority of the Northern and Central Ryukyuan
languages surveyed by Nohara, the only occurrence of the copula tu seems
to be limited to following adjectival stems. If this limitation were not
enough, in most cases it is further restricted to reduplicated adjectival
stems, like Nase magimagii tu ‘being extremely big’ (Nohara 1998: 65; for
data on other dialects see Nohara 1998: 207, 275, 297). So, where does this
leave us? I think we can reach the following conclusions:
(1) The copula to is present in Western Old Japanese, Middle Japanese,
and in all other varieties of Japonic that can be termed ‘Central Japanese’.
It is not really supported by evidence from Eastern Old Japanese and
Ryukyuan, therefore its Proto-Japonic status is more than doubtful. One
can argue for a parallel innovation in Eastern Old Japanese and Ryukyuan,
but I think that such a parallel innovation on the periphery of the family is
highly unlikely.
(2) The quotative to ~ tote and the comitative case marker to are, on the
other hand, amply supported by evidence from three different varieties of
Japonic: Western Old Japanese and/or Middle Japanese (Central Japanese),
Eastern Old Japanese (Eastern Japanese), and Ryukyuan. The same goes
for various forms of the copula n-. Now remember the first oddity, the
distribution of nari and tari, although tari is a compound of to+ari ― even
within Middle Japanese it is practically restricted to kunten texts.
My solution for this problem is the following:
(1) The copula tö in Western Old Japanese and Middle Japanese is an
innovation, and it is not related to the quotative tö or the case marker -tö,
or to the gerund -te (which is also attested throughout Japonic).
(2) Frellesvig (2001) made an important breakthrough in identifying the
Western Old Japanese copula tö with its Middle Korean counterpart ilwo-
 Morphological Comparisons———77

< *i[-]lwo-. However, in light of what I have said above, I differ with
Frellesvig on one point: the form is not a cognate, but an early loan from
some variety of Korean (possibly in the Kofun period, prior to the lenition
*-t- > -l- [-r-] in Korean) into the predecessor of Western Old Japanese.
The same pattern of attestation is found for the majority of other Koreo-
Japonic morphological markers that were perceived as cognates by many
of us, myself included, until recently. Namely, these ‘cognates’ are
normally limited to Western Old Japanese and its more or less direct
descendants in Japonic. This casts further doubt on the validity of Koreo-
Japonic as a family.

2.3.2 Verbal markers

The majority of verbal suffixes found in Proto-Japonic and Old Japanese
are secondary, as they represent grammaticalized auxiliaries that follow the
infinitive form of a verb. Meanwhile, in Korean we have real verbal
morphology, with suffixes directly attached to stems of verbs or to other
suffixes. For example, contrary to an attempt in Vovin (2001), most
comparisons of Old Japanese auxiliaries with Korean suffixes are nothing
but apples and oranges. Most of the few real affixes that can follow verbal
stems in Old Japanese lack reliable counterparts in Korean. For example,
OJ negative -an- potentially corresponds to MK ani ‘not’, but here we face
the opposite problem: MK ani is a negative particle, and not a suffix.
I now present the list of comparisons between Old Japanese and Middle
Korean verbal morphology given in Vovin (2001), with appropriate self-
critical commentary. The original comparisons with Tungusic are deleted,
since the discussion of the relationship of either Japonic or Korean with
Tungusic is outside the scope of the present work.

2.3.2.1 Negative marker *-an[V]-

Traditionally, WOJ -an-, -aNs- (<*-an-s-, where s- is a second stem of the
verb se- ‘to do’) ~ -n-, -Ns- , verbal negative marker, has been compared
with the Middle Korean negative marker ani (Whitman 1985: 244, Vovin
2001: 186). Examples:

Western Old Japanese

kökörö-yu mö omop-an-u apîNta-ni
heart-ABL PT think-NEG-ATTR interval-LOC
while [I] did not think even in (lit.: from) my heart (MYS V: 794)
Middle Korean
pwulhwuy kiph-un namk-on porom-ay ani mwuy-l-ssoy
root deep-ATTR tree-TOP wind-LOC not be bent-ATTR-place
because a tree with deep roots is not bent by the wind (YP 2)

However, there are problems with this comparison. Whitman argued

that MK àní should be directly compared with WOJ -ani, a negative
sentence non-final form, found almost exclusively after verb sir- ‘to know’:
78———Koreo-Japonica

sir-an-i (Whitman 1985: 244), but this cannot serve as evidence, because
WOJ -ani can be clearly analyzed as -an-i, where -an- is a negative, and -i
is an infinitive. 29 Meanwhile, it is not quite clear whether MK àní can be
segmented as *an-i, and the Middle Korean infinitives are -e/-a, not -i. No
less important is the fact that OJ negative suffix -an- belongs to verbal
inflective morphology, and it follows verbal roots, while MK ani, as
mentioned above, is a particle that precedes verbs. Therefore, this
comparison should be rejected on the basis of differences in morphology.

2.3.2.2 Nominalizer *-(V)m

In Vovin 1994 and Vovin 2008 I proposed a comparison of the 2.5 accent
class in Middle Japanese (LF) reflecting pre-PJN *-m on nouns
representing the names of the colors: awo LF ‘blue [color]’, kura LF ‘dark
[color]’, 30 with the Middle Korean nominalizer -(o/u)m (CV+[wo/wu]+m,
CVC+o/um, CVC+wu/wo+m), e.g., cwuk-um ‘death’ <= cwuk- ‘to die’,
twoW-um ‘help’ <= twoW- ‘to help’. Examples:

Middle Japanese:
ana kura ya
EXCL dark PART
Oh, is it dark? (GM 50)

Middle Korean:
kuli-m kuli-ki-yey
draw-NML draw-INF-LOC
when drawing a picture (Twusi cho 16.25)

Later, I also provided additional internal evidence directly pointing to

the existence of final *-m (Vovin 2008), which I briefly outline here.
Nouns of the 2.5 class have derived verbs with -m- as a part of their
stem:
(a) OJ wosa ‘elder’ 2.5, cf. OJ wosamë- ‘to rule’ < *wosamai- <
*wosam-(r)a-i-.
(b) OJ töNka ‘blame, offense’ 2.5, cf. OJ töNkamë- ‘to blame, to
reproach’ < *töNkamai- < *töNkam-(r)a-i-.
(c) OJ tutô(-ni) ‘early in the morning’ 2.5, cf. MJ tutomete ‘id’.,
believed to be a derivation from OJ tutômë- ‘to strive, to make an effort’,
although the semantic connection remains unclear.
(d) OJ paya ‘quickly’ 2.5, cf. MJ fayame- ‘to quicken’.

29
WOJ negative infinitive -an-i is also attested after verb ak- ‘to be satisfied’ and after
potential auxiliary -kate- ~ Nkate- (Vovin 2009: 706-708).
30
Cf. awo-si LL-L ‘is blue’, kura-si HH-H ‘is dark’. See Vovin (1994: 250) for a detailed
account.
 Morphological Comparisons———79

Examples (a) through (d) allow us to reanalyze the derivation of all

deadjectival and some denominal verbs in -me- (OJ -më-), such as
aratameru ‘to renew’ (< arata- ‘new’), atatameru ‘to warm up’ (<
atataka- ‘warm’, itameru ‘to hurt’ (< ita- ‘painful’), kiyomeru ‘to purify’
(kiyo- ‘pure’), sebameru ‘to narrow’ (< sem/ba- ‘narrow’), OJ naNkamë-
‘to lengthen the voice when reciting poetry’ (< naNka- ‘long’), etc. 31
Martin analyzes these verbs as consisting of stem + suffix *-ma- +
transitivity flipper *-Ci- (1987: 792). At first glance, his point of view
seems to be supported by the fact that intransitive verbs corresponding to
transitives in -meru all end in -maru, which Martin reconstructs as *-ma-
ra- (1987: 792): aratamaru ‘to be renewed’, atatamaru ‘to be warmed up’,
sebamaru ‘to get narrow’, etc. However, there are three arguments against
Martin’s analysis:
(1) *-ma- is a suffix whose function is unclear and it represents an extra
unnecessary entity, according to Occam’s Razor. It is, therefore, much
simpler to replace it with the nominalizer *-m to which verbalizer *-(r)a- is
added.
(2) Martin’s analysis of intransitives with -maru as having a stem in
*-ma-ra- produces yet one more suffix *-ra- with an unclear function. I
reanalyze -mar- as nominalizer *-m + verbalizer *-(r)a- + passivizing
suffix *-r-. Thus, we receive quite a symmetrical picture of how transitives
and intransitives are derived:

transitives: adjectival stem + *-m + *-(r)a- + *-i-

intransitives: adjectival stem + *-m + *-(r)a- + *-r-

(3) Under Martin’s analysis, it is unclear why his verbalizer *-ma- is

added to both nouns and adjectives. This strange fact can be easily
explained if *-m is treated not as a final phoneme of the nominal root, but
as a nominalizing suffix *-m after adjectives.
One might argue against the connection between the final *-m in nouns
and accent class 2.5, because not all nouns with denominal verbs in -meru
or -maru belong to the 2.5 class. Thus, OJ kîpa ‘brink’ 2.3, cf. OJ kîpamë-
‘to make go to the end’, kîpamar-, kîpam- ‘to reach the end’; OJ siwa
‘wrinkle’ 2.1, OJ siwam- 32 ‘to wrinkle’ (MdJ siwame-). The second case
looks like a real exception, but I will try to show later that there might be
some evidence for OJ kîpa ‘brink’ 2.3 with an original 2.5 accent.
(e) Some bird names belong to the 2.5 accent class, besides turu ‘crane’,
already mentioned above: sigi ‘snipe’, OJ taNtu ‘crane’, tobi ‘kite’,
toki/tuki ‘ibis’, tuku/zuku ‘owl’. Cf. suffix -më in other birds’ names:
suzume ‘sparrow’, WOJ kamamë (MJ kamome) ‘seagull’, sime ‘hawfinch’,
31
Of course verbs like homeru ‘to praise’ or hazimeru ‘to begin’ do not belong here.
32
Omodaka et al. list siwam- as a consonantal verb (JDB 1967: 376). Their examples, both
in logographic script in Old Japanese and in phonetic in Middle Japanese, provide no basis
for the conclusion that siwam- used to be a consonantal rather than a vowel verb in Old
Japanese. The earliest known attestation of this verb in the form siwabitarikeru ‘has
wrinkled’, which shows that it is either consonantal or upper-bigrade vowel, is in Konjaku
monogatari 26.2 (IKJ 1990: 695).
80———Koreo-Japonica

tubakurame ‘swallow’. It is interesting that none of the birds’ names found

in the 2.5 class has this suffix. Therefore, it is possible to speculate that the
2.5 accent class in these words originated as a truncation of the suffix -më.
Internal evidence pointing to some kind of nasal:
(f) OJ uso ‘a breath exhaled through a narrowed mouth’ (n.) 2.5, cf. OJ
usoNpuk-/usomuk- ‘to exhale a breath making one’s mouth narrower’ <
*usonpuka- < ?*usom-puk-(r)a-
(g) J nasu ‘eggplant’ 2.5, cf. MJ nasubi ‘id.’ 3.5b, with unclear
suffixation < *nasunpi < ?*nasumpi, which can be an alternation of an
earlier *nasum-i. 33
It is doubtful, however, that there are sufficient grounds to reconstruct
Proto-Japonic nominalizer *-m, because there is no uncontroversial
Ryukyuan evidence for this final *-m in Proto-Japonic. Of course, the 2.5
accent class is not preserved by any Ryukyuan dialect, but this cannot be
used against the Proto-Japonic nature of the nominalizer *-m. It is much
more revealing, though, that most of the -me- verbs in modern Japanese
and Old Japanese listed above do not have equivalents in Ryukyuan. Those
that do have equivalents look suspiciously like loans from mainland
Japanese. Thus, there is Shuri ʔaratamiyuN ‘renews’, corresponding to
MdJ aratameru ‘id’., but there is no adjective arata ‘new’ in Shuri, which
instead has miikuN ‘be new’. There is chiyumiyuN ‘purifies’,
corresponding to MdJ kiyomeru ‘id’., but the expected *chiyusa- ‘be pure’
does not exist. Moreover, Shuri has churasa- ‘be beautiful’, corresponding
to MJ kiyora ‘be spotlessly beautiful’, which is based on the same root
kiyo- and shows a different phonetic development as compared with
chiyumiyuN. There is Shuri nagamiyuN, but its meaning is suspiciously
identical with MJ and MdJ nagameru ‘to watch intently’, but not with OJ
‘to lengthen the voice when reciting poetry’. I was able to find usami- ‘to
rule’ and usamar- ‘to become quiet’ in the Psara dialect of the Miyako
dialect group in Sakishima (Shimoji 1979: 32), but these are apparently
late loans, since in Psara it is initial /bu-/ and not initial /u-/ that
corresponds genetically to OJ /wo-/, e.g., Psara buu, OJ wo ‘cord’, Psara
buu, OJ wo ‘hemp’, Psara bunu, OJ wono ‘ax’, Psara butu, OJ wo-pîtö
‘husband’, Psara budur- ‘to dance’, OJ woNtör- ‘to jump’. 34 The same
correspondence is typical for early loans: Psara bututuï < MJ wototofi ‘day
before yesterday’. 35

33
I am indebted to Blaine Erickson for pointing this out to me.
34
Psara data are from Shimoji (1979: 191-192).
35
Psara bututuï ‘day before yesterday’ cannot be a direct cognate of OJ wotö-tu pî ‘id’. for
two reasons. First, PJ *tu > Psara cu, e.g., Psara cuka, OJ tuka ‘mound’; Psara cunu, OJ
tunô ‘horn’; Psara cukï, OJ tukï ‘moon’, etc.; therefore, Psara bututuï reflects not OJ wotö-tu
pî ‘day before yesterday’, but MJ wototofi, with Psara /tu/ corresponding to Japanese /to/.
Second, the Old Japanese demonstrative wotö ‘that over there’ in OJ wotö-tu pî ‘the day
before yesterday (lit.: ‘that day over there’) does not have a cognate in Psara. Thus, Psara
bututuï ‘day before yesterday’ is an obscure compound like MJ wototofi and MdJ ototoi ‘id’.
 Morphological Comparisons———81

In the Eastern Old Japanese corpus, derived verbs in -me- occur twice:
saNtame- ‘to decide’ in MYS XIV: 3418 and katame- ‘to harden’ in MYS
XX: 4390. Both poems contain apparent Eastern Old Japanese features.
Therefore, the reflexes of the nominalizer -m are confined to Japanese,
and this limited distribution suggests that it is another loan from Korean.

2.3.2.3 Transitivity flipper *-gi-

Old Japanese: *-i- in -ë-/-ï- < *V-[C]i- < **V-Gi-, transitivity flipper, e.g.,
tuk(a)- ‘to be attached’, tukë- ‘to attach’ (< *tuka-Gi-), yak(a)- ‘to burn’
(tr.), yakë- ‘to burn’ (intr.) (<*yaka-Gi-).
Middle Korean: -Gi-, -hi-, -ki-, etc. transitivity flipper, 36 e.g., hel- ‘to
break’ (tr.), helGi- ‘to break’ (intr.), anc- ‘to sit’, anchi- ‘to make
somebody sit’.
Martin (1987 and 1990) suggested this Japanese-Korean comparison.
Examples:

Old Japanese:
omösirö-ki nô-woNpa na-yak-i-sö
pretty-ATTR field-ACC no-burn-INF-do
do not burn pretty field (MYS XIV: 3452)

yakë-m-u siNpa-N-kakî
burn-TENT-ATTR firewood-GEN-fence
a fence from firewood that will burn (KK 109)

Middle Korean:
ha-n mwul-ul he-no-n-i
big-ATTR thing-ACC break-PRES-PERF-NML
[he] broke the big thing (Kumsam I: 7)

nolh-i hel-Gi-ti mwot-ho-mye

blade-NOM break-PASS-PROH cannot-do-GER
blade could not be broken and... (Welin X: 70)

There are several problems with this comparison. First, on a

phonological level, we have no internal Japonic evidence that there ever
was any consonant in front of this *-i, let alone that the proposed
consonant was any kind of velar. Second, the reconstruction of the Korean
archetype as *-ki or *-hi oversimplifies the issue, because there are forms
that exhibit other vowels, e.g., -hwu-, -Gwu, -hwo, and -Gwo (Yi
Swungnyeng 1961: 333-335). Third, and most important, there is the
problem of functional difference, as recently outlined by Whitman (2003:
3). This comparison, therefore, should be rejected on phonological and
functional grounds.

36
To be more exact, in Korean it is a marker of a causative or a passive (Yi Swungnyeng
1961: 333-335).
82———Koreo-Japonica

2.3.2.4 Attributive *-VrV

Old Japanese: -uru, -u (< *-urô), attributive, e.g., kwop-uru pîtö ‘person
whom [I] love’, yöm-u uta ‘song that [I] compose’.
Middle Korean: -(o/u)lq (?< *[V]lV), imperfective attributive, e.g., ho-lq
salom ‘a person who will do/does’, cap-ulq ssyang ‘an elephant whom [I]
will catch’.
The comparison between Japanese and Korean is widely known.

Old Japanese:
tösi tukï pa naNkar-uru-N-kötö-si
year month TOP flow-ATTR-COMP-like-FIN
like the flowing of the years and months (MYS V: 804)

Middle Korean:
i KYENG nilk-ulq salom-on
this sutra read-ATTR/IRR person-TOP
the person who reads this sutra (Welin se 22b)

The major problem with this comparison is that PJ *-urô is a

grammaticalized auxiliary that originally followed the infinitive form. This
becomes apparent when we look at the Ryukyuan data.

Chart 26:
Attributive forms in Western Old Japanese and Shuri
Gloss WOJ Shuri
‘write’ kak-u kac-uru < *kak-i-uru
‘burn’ yak-u ’yac-uru < *yak-i-uru
‘row’ köNk-u kuuj-uru < *kug-i-uru
‘stand’ tat-u tac-uru < *tat-i-uru

Palatalization in Shuri takes place only before or after /i/. Therefore, the
only way to explain why Shuri palatalizes the last consonant in the stem is
to assume that the original form, to which the attributive form was attached,
is the infinitive form. Hattori, who was the first scholar to offer this
explanation, believes that the auxiliary in question was PJ *wor- ‘exist’
(1978-79). If he is right, then the grammaticalization of *wor- in Old
Japanese probably happened before the raising *o > u occurred, and
consequently /w/ dropped in front of /u/. The whole process of
development may be represented as *V-i-wor-ô > *V-i-wuru > *V-i-uru >
V-uru (nidan, sa-hen, na-hen, and ka-hen conjugations) > V-ru (ichidan
conjugation), and > V-u (yodan and ra-hen conjugations).
It is possible to conceive a completely different scenario, namely, that
the situation in Proto-Japanese was similar to Eastern Old Japanese and
different from Ryukyuan. Specifically, this presumes that Proto-Japanese
had the conclusive form -u (< *-um) and the attributive form -ô, which
were added directly to the stem of yodan and ra-hen verbs, exactly as in
 Morphological Comparisons———83

Eastern Old Japanese. 37 Meanwhile, nidan, ichidan, and other irregular

verbs had the same formation as found throughout the Ryukyuan verbal
paradigm: infinitive form + wor-ô. Under this scenario the spread of this
formation across the board in Proto-Ryukyuan then would be a Ryukyuan
innovation.
Both of these scenarios have weak points. It is hard to justify the
existence of a unique attributive form in -ô just for wor- ‘exist’ under the
first scheme. Under the second, it remains unclear why the formation of
the infinitive form + wor-ô affected only small groups of verbs in Proto-
Japanese, but was used throughout in Proto-Ryukyuan. Personally, I am
more inclined toward the second scenario. In any case, however, the
Korean attributive form cannot be compared to Japanese, since the Proto-
Japonic attributive is *-ô.

2.3.2.5 Infinitive *-i

Old Japanese: -î < *-i, infinitive, e.g., kîk-î ‘hears and…’, ‘hearing’; ip-î
‘says and…’, ‘saying’.
Middle Korean: -e/-ye/-a, infinitive, e.g., pat-a ‘receives and…’,
‘receiving’; kesk-e ‘breaks and…’, ‘breaking’.
This comparison was proposed in Vovin (2001: 190-191).
Examples:

Old Japanese:
turuNkî-tati kösi-ni tôr-i-pak-î satu-yumî-wo ta-niNkîr-i-mot-i-te
sword-long sword waist-LOC take-INF-wear.on.the.belt-INF hunting-
bow-ACC hand-squeeze-INF-hold-INF-PERF
[young lads], sashing swords at their waists and holding hunting bows in
their hands (MYS V: 804)

Middle Korean:
kil pes-e sswo-sy-a sey sal-ay ta ti-n-i
way take off-INF shoot-HON-INF three arrow-LOC all fall-PERF-NML
[He] swerved and shot, and all [three of his pursuers] fell from three
arrows (YP 36)

Serafim suggested that the infinitive should be reconstructed as *-e for

Proto-Ryukyuan (1985: 134), but the evidence he presented is not very
convincing. Even if we tentatively accept Serafim’s hypothesis that the
Proto-Japonic infinitive *-i is a result of raising from PJ *-e, the Middle
Korean infinitive -e/-a would still remain a dubious cognate, because Old
Korean clearly has *-a and not *-e in the pre-vowel harmony stage of
Korean, e.g., OK KESk-a ‘breaking and …’ (Hyangka IV: 4), TUl-a
‘entering and…’ (Hyangka V: 3), El-a ‘marrying and…’ (Hyangka VI: 2)

37
Frellesvig recently proposed an explanation very similar to this one, but the difference is
that he denies the existence of a separate attributive form in Proto-Japonic, and believes that
the form in -ô was originally a conclusive form that could be used in an attributive function
(Frellesvig 2003).
84———Koreo-Japonica

instead of the expected *kesk-e, *tul-e, and *el-e (cf. the corresponding
Middle Korean forms kesk-e, tul-e, and el-e).

2.3.2.6 Gerund -*mi

Old Japanese: -mî < *-mi, subordinative gerund of quality verbs, e.g.,
puka-mî ‘[because/when] X is deep’, ‘X is deep and’, taka-mî
‘[when/because] X is high’, ‘X is high and…’.
Middle Korean: -(o/u)mye, coordinative gerund, e.g., ka-mye ‘goes and’,
ho-mye ‘does and’, kiph-umye ‘is deep and…’.
This comparison was proposed in Vovin (2001: 191-192).
Examples:
Old Japanese:
KÖKÖRÖ-wo ITA-mî nuye-kô-TÖRI ura-NAKË-WOR-E-NPA
heart-ABS painful-GER nuye-little-bird PREF-cry-be-EV-GER
when [my] heart hurts and little nuye birds are crying (MYS I: 5)

ware-ni otör-er-u pîtö-wo opo-mî

I-DAT be worse-PROG-ATTR person-ABS many-GER
because there are many people who are worse than I (BS 13)

Middle Korean:
twomang-ay myeng-ul mit-umye nwolGay-yey ilhwum mit-un-i
escape-LOC mandate-ACC believe-GER song-LOC name believe-PERF-
NML
while fleeing [he] believed in the [Heavenly] Mandate, in the song [he]
believed [his] name (YP 16)

This is another likely borrowing from Old Korean, because of its

limited distribution in Japonic, and within Japanese itself. There are no
reflexes of this marker in Ryukyuan. Even in Central Japanese this marker
is found only in Old Japanese, 38 and only with quality verbs. Quality verbs,
as mentioned above, were a new development in Old Japanese, possibly as
a result of Korean influence. Meanwhile, in Korean, -mye occurs with both
action and quality verbs. Thus, it appears that the OJ gerund -mî was very
localized and short-lived in the history of Japanese.

2.3.2.7 Past marker *-ki

Old Japanese: past tense final marker -kî < *-ki, e.g., omöp-î-kî ‘[I]
thought/loved’.
Middle Korean: perfective marker -ke-, -ka- ~ -Ge- ~ -Ga-, e.g., tina-ke-n
‘passed’, ni-ke-n-i ‘the one who went’.
This comparison was proposed in Vovin (2001: 192-193).

38
The gerund -mi occurs in early Middle Japanese poetic texts, but there it is an apparent
archaism inherited from Western Old Japanese.
 Morphological Comparisons———85

Old Japanese:
myi-tat-as-i s-er-i-si isi tare mî-kî
HON-stand-HON-NML do-PROG-INF-PAST/ATTR stone who see(INF)-
PAST/FIN
who has seen the stone on which [she] took [her] stand? (MYS V: 869)
Middle Korean:
cwuk-taka sal-Ge-n poykseng
die-TRANSF live-PERF-ATTR people
people who were dying, but survived (YP 25)

The past marker -kî is found only in Western Old Japanese and Middle
Japanese, which both belong to Central Japanese; there are no Eastern Old
Japanese or Ryukyuan attestations. 39 Therefore, it is likely this marker was
borrowed from Old Korean as well. In addition, it must be noted that while
Middle Korean perfective -ke- is an inflectional marker, WOJ -kî is clearly
an auxiliary, because it follows the infinitive form. 40
2.3.2.8 Perfective *-ta-i
Old Japanese: -te- (< *-ta-i) perfective aspect marker, e.g., töNtömey-te-m-
u ‘[he] will stop’, orös-i-te-kyi ‘[he] has put down’.
Middle Korean: -te-, -ta-, retrospective marker, e.g., ho-ta-n salom ‘a
person who has done’, ka-te-n toy ‘a place where [he] has gone’.
This comparison was proposed by Martin (1995: 142).

Old Japanese:
yö-nö kötö nar-e-ba töNtömiy-kane-t-u mö
life-GEN thing be-EV-GER stop-cannot(INF)-PERF-FIN PART
[one] cannot stop life, alas (MYS V: 805)

Middle Korean:
wonol-s il-ol kituli-zop-te-n-i
today-GEN deed-ACC wait-HUM-RETR-PERF-NML

39
The attributive past form -si is attested in Eastern Old Japanese, but since it is the
suppletive form for final -kî, it does not prove the existence of -kî in Eastern Old Japanese.
The modal past form -kêr- < *-kî+ar- is found in several poems in MYS XIV (Mizushima
1984: 891-892), but none of these poems has any Eastern Old Japanese features. Even the
syllable /kê/ is consistently spelled etymologically with the character 家 ‘house’, with none
of the misspellings typical of Eastern Old Japanese texts. In addition, there is the Eastern
Old Japanese modal form -kar- < *kî+ar-, but it is attested only once in MYS XX: 4388. As
a hapax legomenon, it cannot offer strong evidence for the existence of *-ki in Eastern Old
Japanese. It is possible to see EOJ -kî in the reduced allomorph -k- that occurs in past
tentative -kêm-, but again it occurs in texts with no specific Eastern Old Japanese features.
40
Whitman suggested a comparison of the WOJ -kî with MK deverbal nominalizing suffix
-kí (Whitman 1985: 228-229). Although I agree with Whitman that WOJ -kî might be
originally a participial form, there are three problems that prevent me from accepting this
comparison. The first and the second are the same as in the case of comparison with MK
-ke-: WOJ -kî is an auxiliary and not a suffix, and it has no parallels in Eastern Old Japanese
and Ryukyuan. In addition, MK -kí is neutral in regard to tense, while WOJ -kî is not.
86———Koreo-Japonica

[they] waited for the deed of today (WCK 88)

Whether the Proto-Japanese perfective *-ta-i- (with reflexes attested

only in Old Japanese and Middle Japanese) is connected by origin with the
Japonic subordinative gerund marker -te (with reflexes amply attested in
all branches of Japonic) remains a problem. Although the overwhelming
consensus is in favor of this connection, I have my doubts, based on the
fact that in both Old and the Middle Japanese the perfective markers -te-
and -n- combine with different verbs, while at the same time both can be
followed by the subordinative gerund -te (Kolpakchi 1956: 120-121),
(Vovin 2003a: 305-306, 310). Since the gerund -te has a Proto-Japonic
provenance, but the perfective -te- can be confirmed only for Proto-
Japanese, it seems odd that Proto-Japanese would combine two perfective
markers -te- within the same word form. Even if I am wrong, and the
perfective -te- can indeed be derived from the gerund -te 41, the comparison
with Korean is no less problematic, because it compares a gerund marker
with an aspect marker. If I am right, we have a comparison of two aspect
markers, but on the Japonic side it has a limited attestation in Japanese, and
none in Ryukyuan. We are once again faced with the situation when the
auxiliary in Japonic is compared to the inflectional marker in Korean.
2.3.2.9 Perfective *-n-
Old Japanese: -n-, perfective aspect marker, e.g., watas-i-n-i-kêr-i ‘[we]
have crossed over’, tamap-î-n-i-kî ‘[he] has granted’.
Middle Korean: -(o/u)n, perfective/realis attributive, e.g., ka-n-i ‘he went’,
kel-un salom ‘person who walked’.
The comparison between Japanese and Korean is widely known. 42

Old Japanese:
kö 43-ra-ni sayar-i-n-u
child-PLUR-DAT be kept from-INF-PERF-FIN
[I] am kept from [leaving this world] by [my] children (MYS V: 899)
Middle Korean:
ma-pyeng-on mol tho-n pyeng i-Gwo
horse-troops-TOP horse ride-PERF/ATTR troops be-GER
Cavalry are troops that ride horses… (Welin I: 27b)

There are no traces of this marker in Ryukyuan (although there is

perfective -n- in Eastern Old Japanese), so the comparison on the Japonic
side is limited to Japanese, and is again between an auxiliary in Japanese
and an inflectional marker in Korean. I have a suspicion that Japanese
perfective -n- is a result of grammaticalization of either the copula n- ‘to

41
The derivation could only occur in this direction, because the gerund is Proto-Japonic,
and the perfective marker is only Proto-Japanese.
42
For example, in Whitman (1985: 240, #279).
43
This is an apparent misspelling in the text: it should be OJ kô ‘child’ with kō-rui /ô/.
 Morphological Comparisons———87

be’ or the verb in- ‘to go away’. At least typologically a model

infinitive+copula seems to be well justified as an underlying structure for
the perfective. 44 If so, the comparison with Korean should be rejected.

2.3.2.10 Tentative *-[V]m[V]

Old Japanese: -am-, a broad tentative marker which can indicate
probability, volition, etc., e.g., ika-m-u ‘[I] want/will probably go’, ‘[you]
should go’, kôpï-m-u ‘[I] will love’.
Middle Korean: -ma, intention marker, e.g., cwu-ma ‘I will give’, h-wo-ma
‘I intend to do’.
This comparison was proposed by Vovin (2001: 194-195).
Old Japanese:
opo-kî tô-yôri ukakap-î-te körös-am-u tö s-ur-aku-wo sir-an-i
big-ATTR door-ABL peek-INF-SUB kill-TENT-FIN DV do-ATTR-NML-
ACC know-NEG-INF
not knowing that [they] are peeking from the big door and intend to kill
[him] (NK 18)

Middle Korean:
na-y ne-tolye kolochy-wo-ma
I-NOM thou-DAT teach-MOD-INTL
I will teach you (PT I: 10)

In spite of the fact that tentative *-am- can be reconstructed on the

Proto-Japonic level, it is a weak comparison phonologically, since it
involves the supposition of a metathesis either in Old Japanese or in
Korean.

2.3.2.11 Nominalizer *-i

Old Japanese: -î < *-i, nominalizer, e.g., ik-î ‘going’, yöm-î ‘reading’.
Middle Korean: -i, nominalizer, e.g., ka-no-n-i ‘going’, wo-no-n-i
‘coming’.
This comparison was proposed by Martin (1995: 142).

Old Japanese:
ywo-N-pap-î-ni ari-kaywop-ase
night-GEN-crawl-NML-LOC PREF-go back and forth-HON/INF
[he] constantly goes to the night date [with a woman] (KK 2)

Middle Korean:
kwoc tywo-khwo yelum ha-no-n-i
flower good-GER fruit many-PR-PERF/ATTR-NML
[its] flowers are good and [its] fruits are many (YP 2)
44
The reader should keep in mind that the ‘infinitive’ for Japonic forms in -i is just a
customary term used in the field. Functionally, this ‘infinitive’ just corresponds to the
gerund.
88———Koreo-Japonica

Here Japonic and Korean change roles: it is quite clear that the Japonic
form (with reflexes amply attested in all branches) is an inflectional suffix
because it follows the verbal stem directly, but the Korean form is
historically a bound noun occuring only after an attributive.

2.3.2.12 Gerund *-ku/-ko

Old Japanese: -ku, quality verb infinitive and nominalizer, e.g., töpo-ku
‘far and’, ‘the fact that [it] is far’, omösirwo-ku ‘attractive and’, ‘the fact
that [it] is attractive’.
Middle Korean: -kwo, -Gwo, subordinative gerund, e.g., towoy-Gwo
‘becomes and..,’, ho-kwo ‘does and…’, kiph-kwo ‘deep and…’.
This Japanese-Korean comparison was proposed by Martin (1995: 148).

Old Japanese:
imô-tö nöNpor-e-Npa sagasi-ku mö ar-aNs-u
beloved-COM climb-EV-GER steep-GER PART be-NEG-FIN
when [I] climb with my beloved, [the mountain Kurapasi] is not steep at all
(KK 70)

Middle Korean:
nyeth-wo-si-kwo stwo kiph-i-si-n-i
shallow-CAUS-HON-GER again deep-CAUS-HON-PERF/ATTR-NML
[Heaven] made [the sea] shallow, and then made [it] deep again (YP 20)

Martin also adds the OJ nominalizer -ku used after verbs, e.g., ip-aku
‘the fact that [someone] says’, but I do not believe it belongs here, because
it is not just -ku, but –aku, as seen in forms such as mî-r-aku ‘the fact that
[someone] sees’ and s-ur-aku ‘the fact that [someone] does’. This makes it
clear that it is -aku and not -ku that historically follows the attributive form
of verbs, resulting in the following developments: ip-aku < *ip-u-aku, mî-
r-aku < *mi-ru-aku, and s-ur-aku < *s-uru-aku.
The reflexes of the quality verb infinitive -ku are amply attested
throughout different branches of Japonic. Overall, this might be a valid
comparison, but two reservations should be kept in mind. First, while the
gerund -kwo in Korean is used with both action and quality verbs, in
Japonic -ku is restricted to quality verbs. Second, for a perfect phonetic fit
we might expect that WOJ -ku would be from PJ *-ko, with the raising of
PJ *o > u typical of Central Japanese, but this is not the case, as the Proto-
Ryukyuan form is definitely *-ku, not *-ko.

2.3.2.13 Honorific *-s(V)-

Old Japanese: -as- ~ -ös-, honorific marker, e.g., kayôp-as- ‘to go back and
forth’ (hon.), tat-as- ‘to set out’, kîk-ös-/kîk-as- ‘to hear’ (hon.), omöp-ös-
‘ to think’ (hon.).
Middle Korean: -(o/u)si- ~ -(o/u)sy-, honorific marker, e.g., ka-si- ‘to go’
(hon.), cap-osi- ‘to catch’ (hon.).
 Morphological Comparisons———89

This comparison was proposed in Whitman (1985: #232), and repeated

with modifications in Martin (1995).

Old Japanese:
sakasi mê-wo ar-i tö kîk-as-i-te kupasi mê-wo ar-i tö kîk-ös-i-te
wise woman-ABS exist-FIN DV hear-HON-INF-SUB beautiful woman-
ABS exist-FIN DV hear-HON-INF-SUB
[he] heard that there is a wise woman, [he] heard that there is a beautiful
woman, and... (KK 2)

Middle Korean:
wuli sicwo-y kyenghung-ey sal-osy-a wang-ngep-ul yel-usi-n-i
we founder of the dynasty-NOM Kyenghung-LOC live-HON-INF king-
deed-ACC open-HON-REAL/ATTR-NML
When the founder of our dynasty lived in Kyenghung, [he] started [to do]
royal deeds (YP 3)

Martin’s comparison of Japonic and Korean forms largely depends on

his analysis of yodan verbs as vowel verbs uniformly ending in -a (with
occasional and rare -ö). As my morpheme breaks above indicate, I disagree
with his point of view and treat yodan verbs as consonantal. The problem
deserves a separate treatment, and its solution does not have a direct
impact on the present study. But it might be worthwhile to briefly mention
my two most important objections. First, it is not quite clear why only -a is
present as a stem-final vowel (with -ö being extremely rare, and sometimes
alternating with –a, as in the example above). This exclusive occurrence of
-a after consonantal verbs may call for a different explanation, namely, that
it was a separate morpheme. Second, the fact that addition of vowel-initial
suffixes to consonantal stems (for example, infinitive -i) did not result in
the expected monophthongization a+i > ë is bizarre. These two points,
however, can be easily reconciled if infinitive -i never followed -a.
Nevertheless, if -a or -ö in Western Old Japanese honorific -as-/-ös-
would historically represent a separate morpheme, Martin’s comparison
could possibly stand. This is exactly the approach taken by Whitman, who
posits OJ -s- as an honorific suffix (1985: 234-235). However, there are
two problems with this etymology. First, it seems that the Old Japanese
morpheme was really -as-, not just -s-. Otherwise such forms as OJ kês- ‘to
wear’ (honorific) < *kî-as- and mês- ‘to see’ (honorific) < *mî-as- are
difficult to explain: if the honorific suffix were really *-s-, we would
expect *kî-s- and *mî-s-. Second, we again face the problem of a limited
distribution in Japonic. There are no attestations of this honorific marker or
its reflexes in Ryukyuan (Serafim, personal communication), although it is
attested in Eastern Old Japanese. Thus, it appears that the honorific marker
-as-/-ös- is limited to Old Japanese and possibly to Middle Japanese. 45 The
explanation that it was borrowed seems preferable.
45
In Middle Japanese the honorific marker -ase-/-sase- formally coincides with the
causative marker -ase-/-sase-. It remains to be seen whether WOJ -as-/-ös- and MJ -ase-/-
90———Koreo-Japonica

To conclude this discussion on Koreo-Japonic comparative verbal

morphology, I present the following chart, where I put together the
morphological markers discussed above with some other morphological
markers in Japonic that do not have suggested Korean counterparts, but are
amply attested in all three branches of Japonic. I have not included the
markers that I rejected on the basis of insurmountable phonological and/or
morphological problems, such as transitivity flipper, attributive, and
infinitive.
Chart 27:
Some comparative verbal morphology
WOJ EOJ Ryuk. OK MK
Negative -an- -anap- -an- ?*anti ani
Gerund1 -te -te -ti — —
Adj. gerund2 -mî — — *-mi -mye
Final (verbs) -u -u *-um — —
Nominalizer1 -sa -sa -sa — —
Nominalizer2 (2.5) < *-m — — *-m -m
Nominalizer3 -i -i -i — -i
Past -kî — — *-ke- -ke-
Copula n- n- n- n- — —
Copula tö tö tö (<WOJ) — — il- <*i-tV-
Perfective1 -te- -te- — — -te-
Perfective2 -n- -n- — -n- -n-
Tentative -am- -am- ?-a — -ma
Adj. infinitive -ku -ku -ku -kwo -kwo
Honorific -as- -as- — -si- -si-

Looking at this chart, we can come to the following conclusions:

(1) Among the fifteen comparisons presented above, eleven have
suggested Korean etymologies, but among those eleven, only three have
clear reflexes in all branches of Japonic. Two of these three are
problematic, because they involve comparison of a Japonic inflectional
suffix to an auxiliary in Korean (negative) and comparison of a Japonic
inflectional suffix to a Korean bound noun (nominalizer3), respectively.
(2) Among the remaining eight attested only in Japanese (but not in
Ryukyuan) only four are attested in Eastern Old Japanese as cognates of
Western Old Japanese morphemes, and not as borrowings from Western
Old Japanese. Among these four parallels three are problematic, because

sase- represent the same or two unrelated morphemes. At the present time I am inclined to
see MJ -ase-/-sase- as a grammaticalization of se- ‘to do’, first as a causative, which then
subsequently started to be used as an honorific.
 Morphological Comparisons———91

two involve a comparison of a Japanese auxiliary to a Korean inflectional

suffix, and the remaining one presents phonetic problems.
(3) The remaining four verbal markers with suggested Korean cognates
are attested only in Western Old Japanese.
(4) It is revealing that two of the remaining four Western Old Japanese
markers, adjectival gerund -mî and the nominalizer reflected by accent
class 2.5, are mostly limited to quality verbs and have doublets or near-
doublets (subordinative gerund -te and nominalizer –sa, respectively) that
are amply attested in other branches of Japonic, but do not have any
Korean parallels.
This situation, similar to what we have already seen in the area of
nominal morphology, again suggests an areal rather than a genetic
relationship, with Western Old Japanese or, rather, Proto-Central Japanese,
heavily borrowing from Old Korean. Further confirmation for this proposal
comes from the fact that there is no evidence for a paradigmatic
commonality between Japonic and Korean verbal morphologies. Rather,
we have comparisons of isolated verbal markers, with the majority of them
running into the problems discussed above.
3

LEXICAL COMPARISONS
The Koreo-Japonic comparison has always been conducted mostly in the
area of vocabulary, with grammatical comparison playing only a secondary
role. This can be demonstrated by the fact that two seminal works that
attempted to prove the genetic relationship between these two languages,
Martin (1966) and Whitman (1985), dealt predominantly with lexical
comparisons. I believe that the lexicon represents the most unstable part of
a language, and that the ultimate proof of a genetic relationship can come
only from the demonstration of common paradigmatic morphology
(provided that the languages being compared do have morphology). But
because lexical comparison has played such a prominent role in Koreo-
Japonic comparative studies, below I discuss the Koreo-Japonic lexical
comparisons as presented in Whitman (1985). My choice of Whitman
(1985) rather than Martin (1966) is due to the fact that Whitman provided
many important updates to both the comparisons and the reconstruction of
Proto-Japonic and Proto-Korean. Although now we have a more recent
publication (Starostin, Dybo, and Mudrak 2003) that also tries to prove the
existence of a genetic relationship between Japanese 1 and Korean, albeit in
the framework of a greater ‘Altaic’, Whitman’s 1985 dissertation is clearly
superior to the pseudo-scholarly work by Starostin, Dybo, and Mudrak
(2003). 2

3.1 DOUBLETS IN WESTERN OLD JAPANESE

Before I proceed to the reevaluation of the Koreo-Japonic comparative
corpus in Whitman (1985), one general observation presented below is in
order. The structure of the Old Japanese lexicon is highly suspicious in the
respect that many of the basic vocabulary items have doublets. The issue
was discussed at length in Vovin (2007). Below is a comparative chart of
some of the doublets that have reliable etymological counterparts in
Korean with others that do not. The word without a Korean etymology is
listed first, with a subscript (1), and the word with a Korean etymology
second, with the subscript (2). The chart provides the distribution of the
words in question across various branches of Japonic.

1
However, not with Japonic, because Ryukyuan data is ignored.
2
I have dealt with Starostin, Dybo, and Mudrak (2003) elsewhere (Vovin 2005b).
 Lexical Comparisons———93

Chart 28:
Lexical doublets in Western Old Japanese
Gloss WOJ EOJ Ryuk. OK MK
father1 titi titi cici — —
father2 kasö — — kaso 3 —
mother1 papa papa fafa — —
mother2 omo amo/omo — — emi
head1 tub/muri — çiburu — —
head2 -mata 4 — — mati 5 mali
eye1 më me *me — —
eye2 na-mîNta 6 — nada — nwun
snow1 yukî yoki *yoki — —
snow2 na-dare 7 — — — nwun
sea1 umî umi *omi — —
sea2 wata — — — patah 8
many1 opö opo *opo — —
many2 mane- — —9 — manh-
morning1 tutômëte — *sutomete — —
morning2 asa asa — — achom
earth1 ni, nita 10 mizya 11 *mita — —
earth2 tuti tuti 12 — — twute- 13
house1 ya ya *ya — —
house2 ipê ipe 14 — — cip
root1 ne ne *ne — —
root2 mötö moto muutu(?) 15 — mith 16
17
arrow1 ya -ya *ya — —
arrow2 sa sa — — sal
treasure1 takara — *takara — —
treasure2 kusirö 18 — — — kwusul
3
A Paekche word.
4
Attested in ya-mata woröti ‘eight-headed serpent’; for a detailed discussion see Vovin
(2000: 144-145).
5
This is the Early Middle Korean form, attested in Kyeylim #161.
6
OJ na-mîNta ‘tear’, lit. ‘eye-water’.
7
Attested only in MJ: nadare ‘snow slide’.
8
Also MK palol ‘sea’.
9
Shuri maNdoo- ‘to be many’ is isolated in the Ryukyus. Segmentation problems arise if
one wants to relate it to WOJ mane-.
10
WOJ nita ‘dirty ground’.
11
Attested in the Hachijō dialect, but not in Eastern Old Japanese.
12
Single attestation in MYS XX: 4418.
13
Also attested as MK twuti-.
14
Also attested as EOJ ipa-ro ~ ipi ‘house’.
15
Limited attestations in the Ryukyus. Although the word does seem to occur in Miyako
and Yaeyama, it is predominantly used there as a classifier, not as an independent word.
16
The meaning is ‘base’.
17
Attested only as the second element of the compound satu-ya ‘hunting arrow’ (MYS XX:
4374).
18
‘Bracelet made of precious stones’.
94———Koreo-Japonica

As one can easily see, the second Western Old Japanese word among
the two doublets has parallels in Middle and/or Old Korean, but rarely has
a cognate in Ryukyuan and is only slightly better attested in Eastern Old
Japanese. This is exactly the same situation that we have already seen with
nominal and verbal morphology, confirming my general thesis that
Western Old Japanese morphemes and words with Korean parallels, but no
cognates in other branches of Japonic, are loans from Old Korean into
Central Japanese.
3.2 WHITMAN’S LEXICAL COMPARISONS
In his seminal dissertation Whitman presents 347 19 comparisons
between Korean and Old Japanese (1985: 209-246). Some of them include
morphological markers, but the majority are lexical. I discuss all of his
lexical comparisons below, except those that have been ruled out by recent
developments in the field, e.g., Whitman’s comparisons involving Middle
Korean initial aspirates. 20 Most of Whitman’s grammatical comparisons
were addressed above, but they are cross-referenced below in order to
preserve Whitman’s numbering. The discussion of individual etymologies
is followed by a statistical summary. I mark the obvious loans with an (L),
potential cognates with a (C), and rejected etymologies with an (R).
3.2.1 *p-
(1) (R) MK pá ‘place of’ ~ OJ -pa L ‘edge of, place of’, J ba ‘place of’
(Whitman 1985: 209). J ba does not have necessarily to be a contraction of
OJ nipa ‘garden‘, which Whitman believes to be a compound consisting of
ni ‘earth’ and pa ‘place’. It can be just a contraction of genitive -nö +*pa,
thus strengthening Whitman’s comparison, provided that one can prove the
existence of *pa ‘place’ in Japonic. However, I believe there are two
problems with this comparison. First, while -pa ‘edge’ certainly does exist,
cf. yama-nö pa-ni ‘at the edge of the mountains’ (MYS XV: 3623), I am
not aware of any instances of -pa ‘place of’ occurring even as the second
element of a transparent compound, let alone in independent usage in Old
Japanese. J ba, no matter what is its ultimate origin, is certainly attested
too late to support the comparison. Second, in spite of MK pá, found as a
translational equivalent of the Chinese character 所 ‘place’ (Hwungmwong
II: 8a), we have to keep in mind that 所 can be used as a nominalizer in
Classical Chinese, meaning ‘the matter, the fact’, rather than simply
‘place’. A perusal of textual examples in Middle Korean demonstrates that
MK pá has exactly the same usage as Classical Chinese 所: it occurs
exclusively as a nominalizer after verbs and is never used independently to
mean ‘place’:

19
The numbering in Whitman’s list shows 352 examples, but examples 196, 201, 220, 328,
and 344 are absent. I preserve Whitman’s numbering in order to avoid unnecessary cross-
listing.
20
Whitman suggested that MK initial aspirates ph-, th-, kh-, ch- go back to PK *op-/*up-,
*ot-/*ut-, *ok-/*uk-, *oc-/*uc- (1985: 176). This treatment is obsolete nowadays in light of
the Ramsey-Yi treatment of the origins of aspiration in Middle Korean (see 1.1.1).
 Lexical Comparisons———95

eli-n poyksoyng-i nilu-kwocye h-wo-lq pa-y isy-e-two

ignorant-ATTR/REAL commoner-NOM say-VOL do-MOD-ATTR/IRR
matter-NOM exist-INF-CONC
Although there are things that ignorant commoners want to say... (HMC 2a)

pwuthye nilo-sy-an pa-s pep-un i pep i-ra

Buddha speak-HON-ATTR/REAL matter-GEN law-TOP this law be-FIN
The law [of the matters] which the Buddha spoke of, is this law (Kumsam
II: 40b)

This comparison should be rejected: there is solid evidence only for OJ

pa ‘edge’, which ends up being compared with ‘matter’ in Korean.
(2) (R) MK pàk- ‘thrusts it into, inserts it, fills it’ ~ OJ pak- H ‘puts
(shoes, skirt, etc.) on by inserting legs’ (Whitman 1985: 209). Although the
verb is attested in both Western and Eastern Old Japanese, I believe that
the semantic tag Whitman provides is incorrect. There is a single Eastern
Old Japanese example where this verb refers to wearing shoes (MYS XIV:
3399). In all other Eastern and Western Old Japanese examples this verb
refers to wearing a sword, most often to WOJ tati (EOJ tasi/tati) ‘long
sword’, and in a few examples to turuNkî ‘double-edged sword’. Neither of
these swords was worn the way the samurai did in later periods, by
inserting it into the sash. In Japan from the Asuka through Nara period
swords were worn hanging from the belt, as can be seen not only from the
art of the period, but also by the following line in the Nihonshoki kayō,
where pak- is used together with tare- ‘hang down’:

opo tati-wo tare-pak-î tat-i-te

big long-sword-ACC hang-down(INF)-attach-INF stand-INF-SUB
standing with big long-sword hanging down (NK 89)
There is one more Eastern Old Japanese example where the verb pak- is
used for attaching a bowstring to a bow (MYS XIV: 3437). Unfortunately,
Old Japanese texts do not offer us any evidence for pak- used with skirts.
All I can say at the moment is that OJ mö ‘skirt’ was unlike a modern skirt,
into which the lower part of the body is inserted; instead it was wrapped
around the waist and legs. Therefore, I think that OJ pak- simply means
‘attach [to the body]’. In later periods of the development of the language
there are, of course, many examples of pak- attested exactly in the meaning
Whitman provides, but these are likely due to a secondary development:
we should not forget that styles of clothing and dress changed drastically
between the Asuka and Heian periods.
As far as the Ryukyuan evidence is concerned, Shuri has both hacuN
and hakiyuN ‘to wear’ (RGJ 1976: 199, 202). The latter is an apparent late
loan from Japanese, but the first one is likely to be a cognate. Both,
however, mean ‘to wear something at the sash’ or ‘to wear on the neck’.
Neither is used for wearing shoes, for which an unrelated verb, kunuN
(kum-), is used. This kum- is attested throughout the Ryukyus, while hak-
96———Koreo-Japonica

‘to wear shoes’ seems to be limited to Nase and Koniya in the Northern
Ryukyus (Hirayama 1966: 323). Isolated cases from the Northern Ryukyus,
where the Japanese influence was stronger than in the south, certainly
present more favorable evidence for a loanword scenario than for a genetic
relationship with J hak-. Thus, the Ryukyuan data supports the Proto-
Japonic semantic archetype as just ‘to attach [to the body]’. A comparison
with Korean should be ruled out on semantic grounds.
(3) (R) MK pàkwòní, LMK pakwulley ‘bamboo basket’ ~ OJ pakô HH
‘box’.21 Whitman notes that “this comparison must assume that MK -ni is
a suffix, perhaps the plural suffix -ni” (Whitman 1985: 209). I am not
aware of the existence of such a plural suffix in Korean. Whitman’s LMK
pakwulley is certainly Early Modern Korean, as it is not attested earlier
than 1677, although there is a predecessor, MK pákwùlGéy (Hwungmwong
II: 27a). The problem here is that neither MK pákwùlGéy nor EMdK
pakwulley have anything to do with bamboo baskets, since both mean
‘bridle’. We are left with MK pàkwòní, but we cannot really compare it
with Japonic, as the segment -ni is unaccounted for. Also, MK -k- < PK
*-nk-, so that for a genuine cognate with Japonic we would expect MK -h-
or -G-. I am inclined to dismiss this comparison on the basis of inadequate
morphological analysis and phonetic irregularity. In any case, this is a
cultural vocabulary item that is unlikely to provide any support for the
Koreo-Japonic genetic relationship.
(4) (R) MK pál ‘foot, leg’ ~ OJ paNkî ‘shin’, Ryukyuan (Sakishima)
pagi ‘leg’ (Whitman 1985: 209). There are three problems with this
etymology. First, MK pál does not mean ‘leg’, for which MK tàlí is used.
Second, we simply do not know whether ‘leg’ > ‘shin’ in Old Japanese, or
‘shin’ > ‘leg’ in Ryukyuan. Third, and most important, we have a phonetic
irregularity here: the correspondence of MK -l- to OJ -Nk-, I believe, is
unique, not supported by any other examples. Whitman reconstructs PJK
*parak here (1985: 209), but he also reconstructs PJK *tarak for his
comparison of ‘Koguryo’ tal ‘mountain’ and OJ taka- ‘high’ (1985: 214).
It is clear that the Japanese reflexes are different in both cases. I reject this
etymology on the basis of phonological irregularity and semantic
discrepancy.
(5) (R) MK pànól ‘needle’ ~ WOJ pari LH, EOJ paru ‘id’. 22 Whitman
also cites LMK palol (1985: 209), which is in fact an Early Modern
Korean hapax legomenon attested in the Ma kyeng enhay, dating from the
Inco period (1623-49). No hapax legomenon is a reliable source for
external comparison, especially when other sources, including modern
dialects, confirm MK pànól and not palol. 23 The comparison is not reliable,
since it rests on the speculation that WOJ pari < *parari < *panari,
according to Whitman’s *-r- loss law (Whitman 1990). While such a
development is certainly possible, it is not confirmed for this word by

21
This comparison is also found in Martin (1966: #271).
22
This comparison is also found in Martin (1966: #153).
23
The forms pa:l and pal attested in Kyengsang province (Choy 1978: 612), potentially
reflect EMdK palol, but they are isolated and rare.
 Lexical Comparisons———97

internal Japonic data. PJ *parari represents a reconstruction on the basis of

a Korean form that is itself a reconstruction from above. There is also the
problem of phonetic regularity; if we look at another of Whitman’s
comparisons, namely MK hànólh ‘sky’ and OJ swora ‘id’. (236), it is
immediately clear that the Middle Korean segment -ano- corresponds to
OJ -a- in the case of ‘needle’, and to OJ -wo- in the case of ‘sky’.
Therefore, I reject this etymology. See also (264) below.
(6) (R) MK pàs, pàsk < *pàsòk ‘exterior, outside’ ~ OJ pasi HH
‘outside edge’ (Whitman 1985: 209). Omodaka et al. (JDB 1967: 575)
define OJ pasi as ‘[small] inside corner (J katasumi), edge’, and the textual
evidence supports the definition. Even in modern Japanese hasi does not
necessarily mean ‘outside edge’ as, e.g., in the famous accent tongue
twister: hásì-wo motte hàsí-nò hàsí-wó wataru ‘[he] crosses the bridge on
the edge holding chopsticks’, as the walking is not done outside the
bridge’s railing. More important than the semantics is the fact that OJ pasi
‘edge, inside corner’ is a compound, including OJ pa ‘edge’, already
discussed above in (1) (cf. MJ pana ‘tip’, which although not attested
phonetically in Old Japanese, 24 testifies for the morphological boundary in
pasi). There is no similar internal evidence for segmenting MK pàsk <
*pàsòk ‘exterior, outside’ as *pa-sok. I reject this comparison on the basis
of incompatible morphological structure and semantics.
(7) (R) MK pàs- ~ pac- ‘hurries’ ~ OJ pasir- ‘runs, hurries’, pase-
‘makes it run’. Whitman maintains that “MK root appears in paspo- ‘is
hurried’ < pas/c- + -po- deverbal adjectival suffix; pach- ‘makes it hurry’
< pac-/pas- + -hi causative suffix” (1985: 209). There are numerous
problems with this comparison. First, MK pàch- is misglossed: it does not
mean ‘makes it hurry’, but simply ‘hurries’: Yu Changton glosses it as
pappi hata ‘hurries’ (LCT 1987: 373), and the textual examples he
provides fully support the gloss:
Pikwu-sung-ul pwo-si-kwo teGwuk pach-osi-n-i
Pikwu-monk-ACC look-HON-SUB even more hurry-HON-ATTR/REAL-
NML
[he] looked at the monk Pikwu and hurried even more (WCK 44)

pach-a mal-Gwo
hurry-INF do not-SUB
do not hurry, and... (Twusi cho I: 10)
The suggested causative meaning is not there, and with that the
proposed segmentation pach- < pac- + -hi- should be abandoned as well.
Consequently, the root is pach-, and no further segmentation is possible,
but MK pàch- < PK *pàcòh-, or, much more likely *pàhòc-, which is
probably from still earlier *pàkòc- (see velar lenition in 1.1.3.3 above).
The medial *-k- presents an insurmountable phonological difficulty, and

24
It is possible that there is one semantographic attestation in MYS XIX: 4217.
98———Koreo-Japonica

this alone should rule out the comparison with OJ pasir- ‘runs, hurries’,
pase- ‘makes it run’. However, there are problems on the Japonic side as
well that further invalidate this comparison. First, I fail to see what is
Whitman’s basis for glossing OJ pasir- as ‘hurries’: the word certainly
means ‘runs, moves quickly, falls down quickly’ (JDB 1967: 579). Second,
there is no internal Japonic evidence for segmenting *pasir- as *pas-ir,
because there is no suffix *-ir- in Japonic. 25 Therefore, the comparison
should be rejected on phonological and morphological grounds in addition
to the less than perfect semantics (‘hurry’ in Korean but ‘run’ in Japonic).
(8) (L) MK pàth ‘dry field’ ~ OJ pata LH, patakë LHL ‘id’. 26 Whitman
notes that OJ pata appears only in compounds (1985: 209). To the best of
my knowledge, the same is true in modern Japanese as well, despite the
fact that dictionaries frequently list hata as an independent word. But its
actual usage seems to be limited to compounds like ta-hata ‘wet and dry
fields’, or hata-kaesu ‘to dig up the field in preparation for sowing’. Here
we are dealing with an agricultural term, and the introduction of
agriculture to the Greater Manchurian region certainly postdates the
separation of any postulated Koreo-Japonic family. The word is not
attested in Eastern Old Japanese, but it is attested in Ryukyuan. The
attestations in Ryukyuan present a picture compatible with direct genetic
inheritance within Japonic, rather than borrowing from Japanese: Nase
hatee, Koniya hatehe, Kamezu hate, Ishigaki patagi, Ōhama patagi,
Taketomi hate, Kuroshima pataki, Hateruma pitegi, Yonaguni hatagi
(Hirayama 1966: 338; 1967: 309). Cf. the cognates of OJ takë ‘bamboo’ in
the same Ryukyuan dialects: Nase dee, Koniya dïxë, Kamezu dëë, Ishigaki
taki, Ōhama taki, Taketomi tai, Kuroshima taki, Hateruma taki, Yonaguni
tagi (Hirayama 1966: 352; 1967: 338). The presence of the word for ‘dry
field’ in both Ryukyuan and Japonic suggests that it may be either a
cognate to Korean or a loanword from the time of the mutual coexistence
of Japonic and Korean in Manchuria or on the Korean peninsula. Since this
is a cultural vocabulary item, even if it represents a cognate, it is not
indicative of a genetic relationship.
(9) (R)/(C) MK pàtó:li ‘wasp’ ~ OJ pati HH or LF ‘bee, wasp’. There
are no reliable Old Japanese phonetic attestations of pati, but the word is
attested in Middle Japanese as fati. 27 Whitman notes that the rising pitch
25
OJ pase- ‘makes run, runs’, cited by Whitman is sometimes considered to be a transitive
form of the intransitive pasir- (JDB 1967: 579). However, this is problematic for the
following reasons. First, the expected intransitive form of pase- should be *pasar-, not
pasir-. The exact morphological relationship between pase- and pasir- is unclear, and it is
likely that they are unrelated. Second, pase- also can be used as an intransitive verb ‘to run’,
further undermining the possibility of its relationship to pasir-. Third, examples of phonetic
attestation of pase- as a transitive verb in Old Japanese are lacking (all cases are limited to
much later kana glosses), and the first attestation known to me comes only from very late
Middle Japanese (USM III: 12). Meanwhile, although not indicated in JDB, the intransitive
pase- ‘runs’ is likely to be attested in Old Japanese in the expression ama-pase-Ntukapî
‘heavenly runner messenger’ (KK 2, KK 3).
26
This comparison is also found in Martin (1966: #79).
27
There are three attestations in Middle Japanese that I could trace: in Wamyōshō XIX: 25b,
MS 154, and Ruiju myōgishō (Mochizuki 1974: 436).
 Lexical Comparisons———99

on the last syllable in the Korean form is quite exceptional (1985: 209), but
I think that his source for accentuation is (LCT 1987: 360), which lists
only pátó:li (with H on the first syllable, and not L as in Whitman’s
citation). An accentual variant pátólí is also attested in addition to pátó:li
(Hwungmwong I: 24a). MK pátólí is not just any ‘wasp’, but a very
specific type, Ammophilia infesta ‘mud-dauber’. The Korean word cannot
be related to Old Japanese because it has a non-leniting -t- < *-nt-, thus
MK pátólí < PK *pantoli. Therefore, the traditional comparison with
MK :pel ‘wasp’ 28 < PK *pelu LH or *petu LH seems to be a better fit with
Japonic, provided the Middle Korean form is really from *petu. In
Ryukyuan, in addition to Shuri hacaa ‘wasp’ (RGJ 1976: 198), the word is
also attested with the meaning of ‘honey’ in the South Ryukyus in the
Miyako compound patsimidzi ‘honey’ (Nakamatsu 1987: 100), in
Kuroshima (Yaeyama) as patśi ‘honey’ (Nakamatsu 1987: 190), and in
Kawahira (Ishigaki) also as patsi ‘honey’ (Nakamatsu 1987: 229). The
word certainly has Proto-Japonic status, and it is possible that MK :pel <
*petu and MJ fati could be treated as cognates.
(10) (R) MK pólá- ‘desires it’ ~ OJ pör- 29 ‘id’. Whitman remarks that
“this comparison assumes that the Old Japanese vowel is /ô/ and results
from medial *-r- loss (textual evidence is lacking)’. The adjective pö-si
[presumably another typo for pôsi — A. V.] indicates that the stem final -r
in the Old Japanese verb is epenthetic. PJK *pora- (> pre-OJ *pua-r- > OJ
pwor-)” (Whitman 1985: 210). 30 This is certainly a Proto-Japonic word,
confirmed by Shuri husyaN ‘desirable’ (and other dialect forms) or even
better by Shuri volitional auxiliary -busaN < *pu-sa-N. It is not clear to me
on what basis both Whitman and Martin assume that the vowel in the Old
Japanese word is a kō-rui /ô/ and not an otsu-rui /ö/: Whitman specifically
says that there is no textual evidence. However, there is evidence; as
demonstrated by Bentley (1997), the β section of the Nihonshoki still
preserves the distinction between /pô/ and /pö/. OJ posi is attested twice in
the Nihonshoki kayō (NK 54 and NK 84), and OJ por- is attested there only
once (NK 123). Unfortunately, both NK 84 and NK 123 are found in the α
section of the Nihonshoki, where the distinction between /pô/ and /pö/ is
not maintained. This leaves us with the single attestation of OJ posi in NK
54 that belongs to the β section. In this song, OJ posi is spelled as 朋辭
/pösi/, which clearly indicates /ö/, and not /ô/. Admittedly, the evidence is
not very strong, since we have a single attestation, but this is typical of
almost all cases involving the contrast /ô/ : /ö/ after /p/, because only the
Kojiki and the β section of the Nihonshoki preserve it. So, if we, like
Whitman, are to take the evidence of the written sources for this contrast
into consideration, we have to admit this case. Thus, the Old Japanese form

28
This comparison is found in Martin (1966), #12.
29
The discussion makes it clear that it is a typographical error for pwor-.
30
This comparison is also found in Martin (1966), #62. Martin suggests that adjective pwo-
si < *pwora-si, or, as a second possibility, is an independent development of the verb and
the adjective: pwor- < *po-ra-, pwosi < *po-si (1987: 828).
100———Koreo-Japonica

is pösi, and it cannot be compared with MK pólá- for phonetic reasons:

according to Whitman, MK /o/ does not correspond to OJ /ö/ (1985: 129).
(11) (R) MK pòlí- ‘throws it away’ ~ OJ parap- ‘sweeps it away, drives
it out’ (Whitman 1985: 210). 31 This etymology is possible, but there are
two problems. First, in order for it to work, one must show that -ap- in OJ
parap- is an iterative suffix. OJ parara-ni ‘scattered’, cited by Whitman
apparently to demonstrate the much-needed morphological boundary, is
unlikely to be of any help due to its own unclear morphological
composition and dubious semantic connection to OJ parap-. Sweeping or
driving away does not necessarily include a repetitive action: in addition,
the verb without -ap- does not present itself in Old Japanese. Second, even
if PJ *par- is established, the next problem is to explain -i in MK pòlí- as a
suffix. I do not believe that any Middle Korean suffix -i could fit here. This
-i likely belongs to the root. An alternative explanation would be to claim
that PJ *par- lost final *-i, but again I see no way to substantiate such a
speculation with internal evidence. Therefore, I reject this etymology.
(12) (R) MK pòlò- ‘sticks it on, spreads it, pastes it’ ~ OJ par- ‘id’.
(Whitman 1985: 210). 32 MK pòlò- is a Class 8 verb, therefore it goes back
to PK *pòlòl-. This is a possible etymology only if one reconstructs PJ
*para- like Martin. If the reconstruction is just *par-, the etymology faces
the problem of accounting for the final segment in the Proto-Korean form,
as in etymology (11) above. There is also a significant semantic problem
which is hidden in Whitman’s glossing: MK pòlò- means sticking or
pasting some substance like oil or powder on a surface of something else,
but OJ par- indicates spreading of some object in the sense of stretching it.
This verb is applied in Old Japanese to spreading nets, sails, flags,
stretching bowstrings, etc. I believe that these two actions are very
dissimilar. Therefore, I am inclined to reject this etymology.
(13) (R) LMK pól ‘fathom’ ~ OJ pirö LL ‘id’. (Whitman 1985: 210).
The Korean word is actually attested in the Middle Korean of the fifteenth
century (Welin XXI: 16). OJ piro is not attested phonetically in Old
Japanese, so possible Old Japanese forms are *pîrö, *pîrô, *pïrö, and *pïrô.
Since MK /o/ does not correspond to OJ /î/, the only possibility that saves
the comparison would be to suggest that the first vowel was OJ /ï/. Even if
this speculation is true, one still needs to have OJ ï < *ui, not *öi, since
only OJ /u/, not /ö/ corresponds to MK /o/ (Whitman 1985: 129). Even if
we allow these two hypotheses, we still end up with PJ *puirô, which has a
remaining sequence -irô corresponding to MK -l. I do not see any way this
‘correspondence’ can be satisfactorily explained. In addition, outside
Central Japanese, there is hwiru ‘fathom’ attested in Shuri (RGJ 1976:
240), and Yaeyama pïru (Miyara 1981: 79). The latter form with /ï/ allows
us to confirm without any doubt that the Proto-Japonic vowel in the first
syllable was *i. Thus, any relationship between PJ *piro and MK pól can
be ruled out due to the irregularity. Therefore, I reject this etymology.

31
This comparison is also found in Martin (1966), #232.
32
This comparison is also found in Martin (1966), #219.
 Lexical Comparisons———101

(14) (L) MK pól ‘set, layer, occurrence of (counter)’ ~ OJ pê ‘layer, set

of’ < PJK *piro (Whitman 1985: 210). OJ pê is attested only in Western
Old Japanese. Since WOJ pê under Whitman’s scenario < PJ *pia, the
words in question cannot be cognates; as mentioned in (12) above, MK /o/
does not correspond to OJ /î/ < PJ *i. Since WOJ pê does not have any
cognates in Eastern Old Japanese and Ryukyuan, I think that the loanword
explanation is more likely, because it also does not involve phonetic
problems: WOJ pê (phonetically [pe]) was borrowed from the Old Korean
predecessor of MK pól with the expected loss of the final consonant in
Old Japanese. 33 Compare this with etymology (13) above: the same Middle
Korean sequence pól is reflected in Old Japanese as either piro or pê: such
a lack of regularity certainly speaks against the Koreo-Japonic hypothesis.
(15) (R) MK pòlh ‘arm’ ~ OJ pîNti ‘elbow’ < PJK *pidu(+-i) (Whitman
1985: 210). 34 In addition to the lack of correspondence between MK /o/
and OJ /î/, already discussed in (13) and (14) above, we have other
problems, of which the lack of semantic agreement is the most benign.
First, this comparison presents us with a correspondence of MK -l- and OJ
-Nt-, which is not supported by any other etymologies, cf. Whitman (1985:
184), where the same example is cited. The same correspondence appears
again in the list of etymologies, cf. (55) and (134) below. Second, OJ pîNti
< PJ *pinti (*pintu-i or *pintö-i are also possible) has an expected Korean
correspondence of PJ *-nt- to a non-leniting MK -t-, not MK -l- < PK *-t-.
Third, Whitman derives MK -lh from PK *-lVkV or *-lu/*-lo (1985: 175,
184). This reconstruction is obsolete nowadays, and the majority of
scholars in the field reconstruct MK -lh as *-lok/*-luk. This leaves us with
MK pòlh < PK *pòlòk, which creates another obstacle for comparison with
OJ pîNti: it is not clear whether PK *-k in this word is a suffix or belongs
to a root. With the exception of the initial p-, very little remains of this
etymology and it should definitely be rejected.
(16) (R) MK pùlk- ‘red’ ~ pòlk- ‘bright’ ~ J beni ‘rouge’, OJ pani HH
‘red clay’ (Whitman 1985: 210-211). beni is attested in Middle Japanese
(IKJ 1990: 1151). Although its etymology is unclear, as Whitman correctly
notes, it is likely a loanword, due to initial b- < *Np-, so we do not need to
discuss it further. OJ pani can be ‘yellow clay’ as well as ‘red clay’.
Whitman claims that OJ -n- corresponds to MK -lk- and that both are
regular reflexes of a Proto-Japanese-Korean cluster *-lg-, and he provides
four examples in addition to this one to support his proposal (Whitman
1985: 183):

MK mòlk- ‘clear, limpid, transparent’ ~ OJ muna ‘empty, vacant’ (254) 35

MK mòlòl ‘ridge, climax, gist’ ~ OJ muna-/mune ‘id’. (256)
33
It is not clear what kind of vocalism we have to reconstruct for this word in Proto-Korean.
Since the Ceycwuto dialect has pwul instead of expected *pol (Choy 1978: 1042),
apparently simply positing *pol on the basis of Middle Korean is out of the question.
Modern Korean has pel, and the majority of the other dialects have /e/ or /wu/, although
there are also dialects exhibiting /wo/, and even in one case /i/ (Choy 1978: 1042).
34
This comparison is also found in Martin (1966), #269.
35
Numbers in parentheses indicate the numbers of etymologies on Whitman’s list.
102———Koreo-Japonica

MK còlò ‘handle’ ~ OJ tunwo ‘horn, projection’ (184)

MK hòlk ‘earth, soil, clay’ ~ OJ su, MJ suna HH ‘sand, earth’ (238)

The second and third examples do not have an -lk- cluster in Middle
Korean, so they do not belong here. In addition, mòlòl, I trust, should be a
reconstructed form for MK mòlò ~ mòll-, which does not mean ‘gist’ or
‘climax’, but just ‘ridge, ridgepole, foundation’. The remaining two
examples, do not fare much better. OJ muna and munasi- are ‘empty’ in
the sense of ‘vain’ or ‘void’, but not in the sense of ‘vacant’, as the
following textual examples demonstrate: 36

yö-nö naka pa munasi-kî mönö tö sir-u tökî

world-GEN inside TOP empty-ATTR thing DV know-ATTR time
when [I] realized that the world is an empty thing... (MYS V: 793)
muna kötö mö oya-nö na tat-una
empty word PT ancestor-GEN name destroy-NEG/IMP
do not destroy the name of the ancestors [with] empty words (MYS XX:
4465)

It is not clear to me, what ‘vain’ or ‘void’ can have in common with
‘clear’.
Concerning the last example, neither OJ su nor MJ suna mean ‘earth’,
which seems to have been added to the comparison by Whitman, cf. JDB
1967: 378, IKJ 1990: 700. It is significant that OJ su ‘sand’ appears
without the following -na. 37 That demonstrates that suna- in the Middle
Japanese compound sunago ‘sand’ 38 must be a secondary derivation. I
believe that -na represents the rare OJ plural marker -na that I have
discussed elsewhere (Vovin 1994: 249, 253; 2005: 102-107). Thus, su-na
‘sand’ is morphologically complex, and it cannot be compared with MK
hòlk ‘earth’. The semantics also does not support the comparison.
Consequently, a correspondence of OJ -n- to MK -lk- becomes dubious,
because all of the other examples supporting it can be rejected. Therefore,
the correspondence itself and the comparison of MK pùlk- ‘red’, pòlk-
‘clear’ with OJ pani ‘red/yellow clay’ should be rejected.
(17) (R) MK pòlòl- 39 ‘hasty, quick’, spòlòl- 40 ‘fast’ ~ OJ paya- ‘fast’ <
PJK *po:ro(r)- (Whitman 1985: 211). I suppose that Whitman obtained his
meaning of MK pòlòl- as ‘hasty, quick’ from Yu Changton, who gives MK

36
It seems that the meaning ‘vacant’ developed in the Heian period. It is also used in Heian
period glosses for Old Japanese texts, but this certainly does not constitute any evidence.
37
There is also OJ su ‘sandbank’.
38
The form suna is actually attested only in Early Modern Japanese (IKJ 1990: 700), but it
can be traced to an earlier age, because various Ryukyuan dialects have sina ‘sand’ < PR
*suna.
39
The stems of this quality verb are actually pòlò- and pòll- in Middle Korean; *pòlòl- is a
Proto-Korean reconstruction.
40
The stems of this quality verb are actually spòlò- and spòll- in Middle Korean; *spòlòl- is
a Proto-Korean reconstruction.
 Lexical Comparisons———103

pòlò- as ‘poor, urgent, hard’ (LCT 1987: 350). It seems that the meaning
‘urgent’ is based on a single example from the Kwukuppang enhay (1466):

poll-a wumcuki mwot-kwo

urgent(?)-INF move(NML) unable-SUB
being unable to move urgently(?) (Kwukup I: 56a)

The Chinese text corresponding to this line is 拘急不得転側

‘restraining an urge, and not being able to move, and therefore’ (Kwukup I:
55b). It is clear that MK polla corresponds to Chinese 拘急 ‘restraining
urge’. But ‘restraining urge’ does not equal ‘urgent’, let alone ‘hasty,
quick’. Nam Kwangwu glosses MK pòlò- as ‘hard’ (kwutta), and then also
provides this example (1997: 750). I suspect that the phrase above just
means ‘could hardly move’, and that the meaning ‘poor’ listed by LCT for
this quality verb is a secondary semantic development of ‘hard’. This
excludes MK pòlò- from the comparison and leaves us only with MK
spòlò- ‘fast’, which fits nicely semantically with OJ paya-. The problem is
that we have no internal Korean evidence allowing us to segment the
prefix s- in MK spòlò-: it very well may be a part of the root. In addition,
MK -l- does not correspond to OJ -y-. Therefore, I reject this comparison.
See also (34), (40), and (104) below.
(18) (R) MK pòlp- ‘treads (on it)’ ~ OJ pum- ‘id.’ 41 Whitman further
notes that “like the preceding example, [this is] an instance of an original
*lC cluster realized in Old Japanese as a homorganic nasal” (1985: 211). I
trust that Whitman does not mean example (17), but example (16) above.
But the current one is the only example for the correspondence of MK -lp-
to OJ -m- provided among the correspondences for PJK *lC clusters
(Whitman 1985: 183). One example cannot justify the correspondence, and
even in a best-case scenario this etymology should be treated with
reservations.
(19) (R) MK póy ‘belly’ ~ OJ para ‘id.” 42 and (21) (R) MK pòy- ‘gets
pregnant with’ ~ OJ param- ‘id’. Whitman notes that the root is the same
in both cases (1985: 211). The etymology is possible, but it apparently
rests on the assumption of the development PK *polo-i > *pol-i > MK póy.
The loss of *-l- before /i/ in the history of Korean can be confirmed by
examples like OK NAli ‘river’ > MK :nayh ‘stream‘, OK NWUli ‘world’ >
MK :nwuy ‘id’., OK nyelim ‘lord’ > MK :nim ‘id’., but there are still two
problems. First, as one can see from these three examples, the resulting
pitch after the loss of *-l- and subsequent contraction is R, while MK póy
is H. Second, the sequence /li/ after a vowel is amply attested in Middle
Korean: phólì ‘fly (n.)‘, skwòlí ‘tail’, tàlí ‘leg’, tòlì ‘bridge’, etc. 43 Since
there is no loss of *-l- in these words, I believe that Middle Korean final -i
in these examples goes back to a falling diphthong *o/uy. Thus, we would

41
This comparison is also found in Martin (1966), #245.
42
This comparison is also found in Martin (1966), #233.
43
MK -l- does not disappear before -i at morphemic boundaries, either. Cf. MK tol-i,
nominative case for MK tol ‘moon’.
104———Koreo-Japonica

really need MK *poli < *polo[-]i ‘belly’ to prove the apocope in Middle
Korean, but MK póy ‘belly’ indicates PK *poli, which is not comparable to
OJ para, since MK /i/ does not correspond to OJ /a/. All these
circumstances make me rather doubtful about the validity of the etymology.
(20) (R) MK póy ‘boat’ ~ OJ puna-/pune LH ‘id’. (Whitman 1985:
211). 44 A similar kind of assumption as in (19) above is needed to justify
this etymology, this time involving the loss of *-n- before /i/ < *o/uy in
Korean. Unfortunately, in this case there is even less internal evidence in
Korean, since the loss of n- before /i/ or /y/ is known in the history of
Korean, but only in initial position and mostly after Middle Korean. A
possible way of saving this etymology is to claim the development *pono
> *pon > *po with *-i added later, since there are few Middle Korean
nouns with final -n, and probably this is what Whitman had in mind when
positing Proto-Japanese-Korean as *pon, but this is too speculative, and it
requires an explanation for the final vowel in Old Japanese. Besides, the
parallelism of development with (19) will be lost, and with that, I am afraid,
a big part of regularity will be lost as well. Whether OJ pë ‘prow of a boat’,
also mentioned by Whitman, is related to MK póy ‘boat’ or is a loanword
is a difficult question, but it is quite apparent that both OJ pune ‘boat’ and
pë ‘prow of a boat’ cannot be cognates of the Korean word. I am inclined
to see OJ pë (phonetically [pəy]) as a likely loan from Korean, which
seems to be phonetically confirmed by (22) below. It is attested in Eastern
Old Japanese (MYS XIV: 3559; MYS XX: 4359 [twice], 4389), but not in
Ryukyuan.
(21) (R) MK pòy- ‘gets pregnant with’ ~ OJ param- ‘id’. See (19)
above.
(22) (L) MK póyàm ‘snake’, OJ pëmî ‘id.’45 Whitman remarks that it is
a possible loan (1985: 211). This seems to be the case, as the word is
confined to Central Japanese, and not attested in Eastern Old Japanese and
Ryukyuan. 46 Also, it exhibits the same correspondence of MK /oy/ to OJ
/ë/ as in the word for ‘prow of a boat’, discussed in (20).
(23) (R) MK pwól ‘cheek’ ~ OJ popo LL ‘id’. 47 There are no Old
Japanese phonetic attestations. This comparison exhibits a number of
problems, outlined by Whitman himself (1985: 211), the main one being
the assumption that the word is a reduplication in Old Japanese. As there is
no internal evidence that supports this assumption, the comparison is best
set aside for now.
(24) (R) MK pwóm ‘spring’ ~ OJ paru LF ‘id’. Whitman notes, “this
comparison is supported by the falling tone of OJ paru LF (accent class
2.5), which indicates the loss of final *m after /u/” (Whitman 1985: 211).
However, not only is the correspondence of MK -Ø- to OJ -r- irregular

44
This comparison is also found in Martin (1966), #18.
45
This comparison is also found in Martin (1966), #315.
46
PR *pabu ‘snake’ is certainly not related, although it looks similar. Only few Amami
dialects have forms that could possibly be loans, or a native-loan hybrids, e.g., Yamatoma,
China çibu (Hirayama 1986: 646), Jikkiyo çibu: ‘snake’ (Hirayama 1966: 362).
47
This comparison is also found in Martin (1966), #42.
 Lexical Comparisons———105

(Vovin 1994: 250), to the best of my knowledge there are no other cases
where OJ -aru- corresponds to MK -wo-. I reject this etymology on the
basis of the irregularity of the correspondences.
(25) (R) MK pwòtólàp-/pwùtúlèp- ‘soft’ ~ OJ putô ‘fat’ < PJK *pütüra-
(Whitman 1985: 212). The semantics are highly speculative, because OJ
putô means ‘majestic, great, big, thick’, not ‘fat’, which is a much later
semantic development. OJ putô nöritö (MYS XVII: 4031) certainly means
‘great liturgy’, not ‘fat liturgy’. Also compare the following line:
uneNpï-nö mîya-ni mîya-N-pasira putô sir-i-tate-te
Unebi-GEN palace-LOC palace-GEN-pillar majestic rule-INF-place(INF)-
SUB
[they] placed in a majestic way palace pillars in the Unebi palace (MYS XX:
4465)
Pillars can be ‘majestic’, ‘great’, or even ‘thick’, but they are not ‘fat’,
and certainly cannot be ‘soft’, otherwise the palace would collapse.
There are other problems with this etymology besides its semantics.
First, MK pwòtólàp- ~ pwùtúlèp- ‘soft’ has a non-leniting -t-, therefore the
Proto-Korean reconstruction should be *pwòntólàp-/*pwùntúlèp-. PK *-nt-
corresponds to OJ -Nt-, not OJ -t-. Second, Whitman believes that the
Korean forms include the adjectival suffix -p (1985: 212), but the suffix
here might be -ap-/-ep-: cf. MK sìtúl- ‘emaciated, withered’ vs. MK
sìtúlép- ‘utterly exhausted, weary’ and MK pùskúlì- 48 ‘feels shy/ashamed’
vs. MdK pwukkulep- ‘id’. If so, the Proto-Korean form may be just
*pwòntól-~*pwùntúl-, and in that case PJ *putura (speculative itself,
because we do not know whether OJ ô in putô goes back to *ura in this
case or not) has an extra vowel -a that does not match to anything in
Korean. Thus, I reject this etymology on the basis of its phonetic
irregularity and improbable semantics.
(26) (R) MK pàhí-, pèhí- ‘cuts it, pares it’ ~ OJ paNk- ‘flays it’
(Whitman 1985: 212). OJ -Nk- (< PJ *-nk-) corresponds to MK -k- (< PK
*-nk-), not to MK -h- (< PK *-k-). Labeling OJ paNk- as ‘flays it’ is too
narrow; rather, it means ‘peels it off, strips it off’. At the same time, the
meaning ‘pares it’ seems to be added to Middle Korean to improve the
comparison: in all of the examples known to me the Middle Korean verb
means just ‘cuts it, chops it off’. Finally, the last segment -i in Middle
Korean is not matched to anything in Old Japanese. Thus, I reject this
etymology on the basis of its phonetic irregularity, unaccounted-for
segment, and imperfect semantics.
(27) (R) LMK pel, MK phélí ‘field, plain’ ~ OJ para LL, 49 Ryuk. paru
‘id.’ Whitman maintains that the aspiration in the Middle Korean word is
emphatic (1985: 212). LMK pel is certainly just EMdK pel, attested for the
first time in the Chengkwu yengen (1728). MK phélí is attested within the
expression cùn phélí (with variant cùn phélì) ‘marsh, wasteland’ only in
48
Also confirmed by OK PUSkuli- ‘id’. (Hyangka II.3).
49
This comparison is also found in Martin (1966), #170.
106———Koreo-Japonica

Hwungmwong (I: 5b, 6a) with several later attestations in Early Modern
Korean (Nam 1997: 1264-1265). The word cù-n is likely to be an
attributive form of MK cùl- ‘muddy’, so we can segment MK phélí out, but
there is no guarantee that it means ‘field, plain’. It may be just ‘spot’ or
‘place’. Modern Korean has both phel ‘wide expanse of land, vast plain,
silt, marsh’ and pel ‘field, plain’, the former considered by Martin to be a
para-intensive form of the latter (1967: 1744).50 However, the normal
Middle Korean word for ‘field’ is tùlúh (> MdK tul), and it is amply
attested in the texts (Nam 1997: 455) and dialects (Choy 1978: 78-79). It is
even more suspicious that the dialect attestations of pel are much narrower
than those of the modern dialect cognates of tùlúh (Choy 1978: 79, 100-
101). All of these limited attestations suggest that MdK pel and even MK
phélí are probably recent words, and therefore a reliable comparison
cannot be made with OJ para.
(28) (R) MK pyé ‘rice plant, grain of rice’ 51 ~ OJ po ‘ear of rice’ < PJK
*pye (Whitman 1985: 212). 52 It seems that Whitman assumes that OJ po is
pö with the otsu-rui vowel /ö/, at least that would follow from his list of
vocalic correspondences (1985: 129). Unfortunately, however, it is clearly
WOJ pô, with kō-rui /ô/, because in the Kojiki the character 穂 ‘ear of rice’
is used as the kungana for the word pô ‘top’ in the phrase NAMÎ-NÖ pô
‘top of the waves’ (Igarashi 1969: 106, 127). Since WOJ pô ‘top’ is also
spelled with the ongana 本 /pô/ in other cases, e.g., ma-pô (麻本)
‘pinnacle = true top’ (KK 30) and pô-tu ye (本都延) ‘top branches’ (KK
43), there is no doubt that ‘ear of rice’ was pô in Western Old Japanese.
Thus, the vowels in Korean and Japonic really do not match, so I reject this
etymology.
(29) (R) MK pyéch ‘moldboard of plow, cockscomb’ ~ OJ pitapi
‘forehead, brow’ 53 < PJK *pica + -pi [Japanese] body part suffix in kupîsu
~ kupîpîsu ‘heel’, otokapî 54 ‘chin’ (Whitman 1985: 212). There are no
phonetic attestations of OJ *pîtapî, so we do not know whether it was
*pîtapî or *pïtapï (JDB 1967: 612). I am puzzled by the body part suffix
-pi: I am not aware of any internal Japonic evidence that will allow us to
segment this suffix in any of the above words, and to the best of my
knowledge OJ *pita- ‘forehead’, *ku- ‘heel’, and *oto[N]ka- ‘chin’ are not
attested. The final -h in MK pyéch is supposed to reflect apocope in
Korean (Whitman 1985: 170-171), but this treatment is now obsolete. The
correspondence of MK /ye/ to OJ /î/ is not in Whitman’s list of vocalic

50
The accentuation pattern of MK phélí suggests another possibility: it may be a
nominalized form of MK phyé- ‘to spread’ < *phye-l-i. The alternation /ye/ ~ /e/ is well
known in the history of Korean.
51
The meanings of ‘non-glutinous rice’ or ‘grain of rice’ appear to be late and rare (Nam
1997: 679).
52
This comparison is also found in Martin (1966), #184.
53
I believe the word just means ‘forehead’; the source of Whitman’s ‘brow’ is not clear to
me.
54
Omodaka et al. give OJ otoNkapi (JDB 1967: 151), as there are no Old Japanese phonetic
attestations.
 Lexical Comparisons———107

correspondences (1985: 129). 55 These phonological and morphological

problems, along with the speculative semantics, require that I reject this
etymology. See also (31) below.
(30) (R)/(L) MK :pyel ‘star’ ~ OJ posi ‘id’. (Whitman 1985: 212). 56 The
comparison will work only if OJ posi < *pösi. If it is from *pôsi, the
comparison faces the same irregularity problem as (28) above. Recently
Whitman expressed doubts even about the correspondence of MK /ye/ to
OJ /ö/ (2004, personal communication). The correspondence of MK /l/ to
OJ /s/ may point to an early loan; see (102) below. Since the word is amply
attested in Ryukyuan (Hirayama 1966: 349; Hirayama 1967: 330), it may
be an early loan from the period when the languages coexisted on the
Korean peninsula. However, given the low credibility of the vocalic
correspondence, this etymology is best rejected for the time being.
(31) (R) MK pyèlwók ‘flea’ ~ OJ piru LF ‘leech’. Whitman notes that
the F pitch on the last syllable in Japanese indicates the loss of the final
[consonant ― A.V.] (1985: 212). Although there is supporting evidence
for F pitch reflecting the loss of final *-m in Proto-Japonic (Polivanov
1924: 152; Whitman 1985: 202; Vovin 1994: 250; and Vovin 2006), I am
not aware of the same evidence for final *-k, and it seems that Whitman
does not present any other examples. OJ piru ‘leech’ is not attested
phonetically, so we do not know whether it is *pîru or *pïru. Even if it
were *pîru, OJ /î/ does not correspond regularly to MK /ye/; see (29) above.
Finally, the only commonality between fleas and leeches is that they are
both bloodsucking creatures; otherwise they belong to two quite different
zoological phyla. Thus, I reject this etymology.
(32) (R) MK pyèth ‘sunshine, sun’, pích ‘light’ ~ OJ pî, pîru ‘sun,
daytime’. Whitman remarks that “the variety of vowels in the Middle
Korean forms indicates different suffixes” (1985: 212). Unfortunately, we
do not have any internal Korean evidence that would allow us to
reconstruct any different suffixes here, let alone to demonstrate cogently
that these two forms are related within Korean itself: I am not aware of the
suffixes *-eth and *-ch in any variety of Korean. For all practical purposes,
both Middle Korean words represent roots; we do not have any means to
analyze them further. Therefore, neither of them is compatible with OJ pî,
pîru ‘sun, daytime’. In addition, MK pyèth means ‘the sun’ only as heat,
not as the object in the sky, because the primary meaning of the word is
‘sunshine, sunrays’. Thus I reject this etymology on the basis of the
unaccounted segments in Korean and the problematic semantics.
(33) (C) MK púl ‘fire’ ~ OJ pï, po- < PJK *pör(+ -i) (Whitman 1985:
212). 57 The Proto-Japonic form is apparently *pö, as supported by PR *pi
‘fire’ (Thorpe 1983: 286) and the internal alternation ï ~ o after /p/ in Old
Japanese. I accept this etymology.

55
Yi Kimun argued that MK /ye/ < *i (1959: 131-137), but all his evidence is exclusively
external, and therefore not admissible.
56
This comparison is also found in Martin (1966), #220.
57
This comparison is also found in Martin (1966), #82.
108———Koreo-Japonica

(34) (R) MK pùlùl- 58 ‘calls, sings’ ~ OJ poye- ‘howls’ < PJK *poor-
(Whitman 1985: 212). The major problem here is the fact that MK -l- does
not correspond to OJ -y-, at least not among Whitman’s possible Old
Japanese correspondences to MK -l- (1985: 183-185). We also do not
know whether the Old Japanese form was *pôye- or *pöye-. I reject this
etymology on the basis of this irregularity. See also (17) above and (40)
and (104) below.
(35) (R) MK pwùk ‘shuttle’ ~ OJ pî ‘id’. < PJK *pü (+ -i/ -Vk)
(Whitman 1985: 212). Again, an irregularity is the problem: according to
Whitman’s own correspondences, MK /wu/ does not correspond to OJ /î/,
and PJK *ü is reflected in Old Japanese as /u/, not as /î/ (1985: 129). Thus,
I reject this etymology on the basis of irregularity in correspondences.
(36) (R) MK :pwul- < *pwùlú- ‘blow’ ~ OJ pï- ‘blows nose, farts’
(Whitman 1985: 212). There are problems with this comparison on the
Japonic side. First, it is not really clear whether the Old Japanese verb was
the upper bigrade pï- or the monograde pî-. The only pre-Middle Japanese
attestation is in the phrase PANA-wo sö PI-t-uru ‘blew nose’ (MYS XI:
2637), where the root itself is written semantographically, and the
paradigmatic form PI-t-uru tells us nothing about whether the verb is the
upper bigrade pï- or the monograde pî-. MJ fi- ‘blows nose, farts’ is an
upper monograde verb, which suggests OJ pî-, unless a shift from upper
bigrade to upper monograde can be cogently demonstrated. Omodaka et al.
provide the following philological explanation in favor of pï-, in which the
river Pï in Izumo province is spelled phonetically with the character 肥 /pï/
in the Kojiki, but semantographically with the character 簸 ‘winnow’ in the
Nihonshoki (JDB 1967: 629). This is good proof that the Middle Japanese
monograde verb fi- ‘winnow, fan’ (not otherwise attested in Old Japanese)
was an upper bigrade one in Western Old Japanese. Apparently both
Omodaka et al. and Ōno et al. believe that ‘winnow’ and ‘blow one’s nose’
represent one and the same verb (IKJ 1990: 1070). This is possible, but
blowing one’s nose or farting involves passing air through a tube, often
accompanied by a loud noise, while winnowing does not. Second, even if
the Old Japanese word really were pï- with an otsu-rui /ï/, we do not know
whether it goes back to PJ *ui or PJ *öi. The comparison would work only
in the former case, because OJ /ö/ does not correspond to MK /wu/. But I
suspect that it might be the latter: MJ fe ‘fart’ is likely to be connected with
OJ pï-. Of course, since MJ fe is not attested in Western Old Japanese, we
do not know whether it goes back to OJ *pê or OJ *pë. If it is *pê < *pia,
then it would point to OJ pi- rather than pï-, and a comparison with Korean
will not be possible. But if ‘fart’ is *pë, it could be only from *pay or *pöy.
The form *pöy will support a derivation of OJ pï- from PJ *pöy-, making a
comparison with Middle Korean untenable. Finally, I am afraid that both
the Korean and the Japonic forms are onomatopoetic roots, as evidenced
by their English translation ‘blow’ and other multiple crosslinguistic
examples.
58
The stems of this quality verb are actually pùlù- and pùll- in Middle Korean; *pùlùl- is a
Proto-Korean reconstruction.
 Lexical Comparisons———109

(37) (R) MK :pwut-/pwul ‘increases, swells’ < *pwùtú- 59 ~ MJ 60 putô

LL ‘fat’ < PJK *pütür (Whitman 1985: 212). The same Old Japanese word
was already compared with MK pwòtólàp-/pwùtúlèp- ‘soft’ in (25) above.
OJ putô certainly cannot be a cognate for both. As discussed in (25), OJ
putô originally meant ‘majestic, big, great, thick’, so here, it fits better
semantically with (25), although it is not perfect. But OJ /ô/ does not
correspond to PK *-u; therefore this etymology should be rejected.
(38) (R) MK pí ‘rain’ ~ OJ pî ‘ice, sleet’, also ‘(big) rain’ in pîsamë < pî
+ s (genitive) + amë LF ‘rain’ (Whitman 1985: 213). There are problems
on both the Korean and the Japonic sides of this etymology. First, as I
mentioned before (see 1.1.3.2), the Early Middle Korean word for ‘rain’ is
transcribed in the (Kyeylim #7) as piWi, so the Proto-Korean form was the
disyllabic *piwi < *pigi and is likely to be borrowed from Tungusic *pigi-
‘[to] rain’ (Vovin 2000: 146-147). Second, OJ pî is just ‘ice’; the texts do
not collaborate the meaning ‘sleet’ for OJ pî. Third, there is OJ pîsamë
‘hail (= ice rain)’, which can be derived from pî ‘ice’ + -samë 61 ‘rain’, and
there is also OJ pîsamë ‘big rain’, which I believe has a completely
different etymology: pîsa ‘long’ + amë ‘rain’. I think that this explanation
is preferable semantically, because Whitman’s pî-s amë ‘rain’s rain’ is
implausible as the derivation of ‘big rain’. I reject this etymology for all
these reasons.
(39) (R) MK pìlé, pìléy ‘cliff, bank’ ~ OJ pê ‘id.’ < PJK *pire (Whitman
1985: 212). I do not know Whitman’s source for the meaning ‘bank’; all of
the Middle Korean textual examples I am aware of involve only ‘cliff,
precipice’. There are a few examples in which OJ pê means ‘area close to
the seashore’ (JDB 1967: 647), but this seems to be a secondary derivation
of OJ pê ‘side‘. There are no texts where this Old Japanese word means
‘cliff’ or ‘precipice’. Finally, the etymology rests on the assumption that
Whitman’s *-r- loss law is applicable here, but this cannot be corraborated
by internal Japonic evidence. Therefore, I reject this etymology.
(40) (R) MK pìlóm ‘pigweed’ ~ OJ pîyu LF ‘id’. 62 Whitman notes that
the “falling tone on [the] last syllable indicates loss of final [*m] in OJ”
(1985: 213). This is certainly true, but nevertheless there are two problems
with this etymology. First is the non-correspondence of MK -l- to OJ -y-,
already mentioned in (17) and (34) above (see also 104 below). The second
problem is with the chronology of the Japanese word: it appears for the
first time in the herbalist book Honzōwamyō (918), so it is Middle, not Old,
Japanese. This leaves room for uncertainty whether its Old Japanese form
was *pîyu with a kō-rui /î/ or *pïyu with an otsu-rui /ï/. Thus, we have one

59
Whitman provides PK *pwùtúl-, but I see no basis for the final *-l in this verb, since it
belongs to Class 6, not Class 8.
60
I believe this is a typographical error for OJ.
61
There are several vowel-initial words in Old Japanese that acquire an initial s- when used
as the second element in some (but not all) compounds: amë/-samë ‘rain’, ine/-sine ‘rice
plant’, ipa/-sipa ‘rock‘. I doubt that it is possible to see a genitive -s there, as Whitman does,
since this genitive does not present itself in Old Japanese otherwise. I think that this /s/ is
rather a reflex of PJ *z- or *h-, or some other fricative.
62
This comparison is also found in Martin (1966), #165.
110———Koreo-Japonica

irregular correspondence and one uncertain correspondence, and I believe

that these constitute a sufficient basis for rejecting the etymology.
(41) (R) MK pìlús ‘beginning’ ~ OJ pîtö ‘one’ < PJK *pi:tö or *pitö
(Whitman 1985: 213). 63 At first glance, this looks like a plausible
etymology, but there are several problems. First, we have to account for -s
in Korean. We can do this, as there is the Middle Korean verb pìlús- ‘to
begin’ (LCT 1987: 412; Nam 1997: 738). Second, MK pìlús is actually not
a noun meaning ‘beginning’, but an adverb that means ‘from the start, at
first’. 64 Third, MK pìlús is attested only in three Middle Korean texts, the
Kumsam, the Nammmyeng, and the Twusi enhay. There are other more
widely attested forms of the same adverb: MK pìlúsé, pìlúswó, pìlúswú
(LCT 1987: 411-412; Nam 1997: 737-738). The first form is the original
one, as it represents the adverbialization of the infinitive pìlús-é. Thus, the
form pìlús is a contraction from pìlús-é, and the segment -s belongs to the
root of the Middle Korean verb. This together with the not quite perfect
semantics leads to the rejection of this etymology.
(42) (L) MK ptóy ‘dirt, filth, grime’ ~ OJ pîNti LL ‘mud’. Whitman
reconstructs PJK *pidö + -i (1985: 213), but it should be *pintui, since MK
/o/ corresponds to OJ /u/, not to OJ /ö/ (Whitman 1985: 129). The only
problem is that on the Japonic side the word is not attested outside Central
Japanese (WOJ pîNti and MJ fidi); there are no cognates in Ryukyuan and
Eastern Old Japanese. This distribution means that WOJ pîNti is likely a
loan from Korean.
(43) (R) MK :pwuy- ‘empty’ ~ OJ pï- ‘runs dry, drains’. Whitman
reconstructs PJK *pö:ri-, but the basis for this reconstruction is unclear to
me: Middle Korean clearly has /wu/, which is a reflex of PJK *ü, not *ö
(1985: 213). MK /wu/ should also correspond to OJ /u/, not /ö/, but OJ pï-
‘runs dry’ certainly goes back to PJ *pöy-, as witnessed to by the Old
Japanese transitive form pos- ‘dries it’. Thus, we are faced with MK /wu/ :
PJ *ö, which is an irregular correspondence. In addition, the R pitch on
MK :pwuy- indicates that the word was disyllabic in Proto-Korean.
Reconstructing the Proto-Korean shape of MK :pwuy- as *pwuli- is
certainly a possibility, but it cannot be confirmed internally. Therefore, I
believe Whitman’s reconstruction of *-r- for Proto-Japanese-Korean is
speculative, as it cannot be corraborated by internal evidence on either the

63
This comparison is also found in Martin (1966), #157.
64
Both Yu Changton and Nam Kwangwu list MK pìlús as a noun as well as an adverb
(LCT 1987: 412; Nam 1997: 738), but their actual examples (only two are given) reveal an
adverbial, not a nominal, function:

ku wo-m-i pilus eps-un cyencho lwo

that come-NML-NOM from the beginning not to exist-ATTR reason COP
That coming from the beginning was without a reason (Kumsam III: 59b)

nyeys yangco-non pilus eps-i wo-m-olwo

old appearance-TOP from the beginning not to exist-ADV come-NML-INSTR
as [he] came from the beginning without his former appearance (Nammyeng I: 75a).
 Lexical Comparisons———111

Korean or the Japonic side, cf. (19) above. Therefore, I reject this
etymology on the basis of this irregularity.
Etymologies (44) through (49) (Whitman 1985: 213) are all rejected,
since they involve an outdated treatment of MK ph- (see 1.1.1).

3.2.2 *b-
(50) (L) MK pàtáh, pàlól ‘sea’ 65 ~ OJ wata ‘id’. (Whitman 1985:
213). 66 The comparison is valid, but within Japonic the word wata ‘sea’ is
confined to Western Old Japanese; there are no Eastern Old Japanese or
Ryukyuan attestations. On the other hand, its doublet OJ umî ‘sea’ (< PJ
*omi) is attested in all branches of Japonic. Thus, I believe that WOJ wata
‘sea’ is a Korean loan on the basis of its distribution pattern in Japonic.
(51) (R) MK pólì- 67 ‘splits it, divides it’ ~ OJ war- ‘splits it’, ware- ‘is
split’ (Whitman 1985: 213). 68 The Korean form is misglossed, but it is
easy to see how the mistake originated. Both Yu Changton and Nam
Kwangwu gloss this Middle Korean verb as MdK paluta (剥), peyta (割)
(LCT 1987: 351; Nam 1997: 749). MdK paluta means ‘to peel, to split
(and turn inside out)’, but the character 剥 ‘to peel’ is used to clarify that
only that the first meaning applies here. MdK peyta means only ‘to cut’,
and the character 割 ‘to split’, ‘to cut’ was an unfortunate choice here
because it is ambiguous. Thus, it is apparent that Whitman based his gloss
on the meaning of the second character, ‘to split’, and/or on the meaning
‘to split (and turn inside out)’ of MdK paluta, which is not applicable here.
Of course our ultimate witnesses must be the texts themselves, and not the
glosses provided by modern linguists:

HWOY-non poli-l ssi [i]-ra

[name of the dish]-TOP cut-ATTR/IRR word be-FIN
Hwoy means ‘chopped [fish]’ (Nung I: 33b)

cey kwoki-l poly-e

self flesh-ACC cut-INF
cutting off his own flesh (Nung IX: 74b)

65
The Middle Korean forms are doublets that present an interesting puzzle: according to the
lenition theory adopted here, MK pàtáh < PK *pantah, but MK pàlól < PK *patol.
According to Martin’s lenition theory, both should certainly reflect PK *patV, since
Martin’s condition for lenition, ...˚Có/ú, for MK pàtáh is not present (Martin 1996: 2).
However, since this appears to be the only example of such a doublet, I believe that there
might be a different explanation, such as an irregular development or an interdialectal
borrowing.
66
This comparison is also found in Martin (1966), #190.
67
Yu Changton gives pólì- (LCT 1987: 351), but Nam Kwangwu provides different
accentuation: pòlì- (1997: 749).
68
This comparison is also found in Martin (1966), #218.
112———Koreo-Japonica

chiwu-y solh-ol poli-no-n tos-ho-n PWUK nyek polom-i

cold-NML flesh-ACC peel-PRES-ATTR/REAL like-do-ATTR/REAL
North side wind-NOM
wind from the North that is like cold flaying flesh (Twusi cwung X: 40b)

Thus, we can see that MK pòlì- ‘peels, cuts’ does not mean ‘split’.
Besides, the final segment -i in Middle Korean is either impossible to
match with Old Japanese (assuming that the Proto-Japonic original form is
the consonantal verb *bar-), or we have to deal with an irregular
correspondence of MK -i to PJ *-a (assuming the Proto-Japonic
reconstruction *bara-). Thus, this etymology should be rejected on
semantic and phonetic grounds, as well as on its segmentation problems.
(52) (R)/(L) MdK pey ‘hemp, flax, linen’ ~ OJ wo ‘id’. < PJK *ber (+ -i)
(Whitman 1985: 213). As far as I know, OJ wo refers only to ‘hemp’ as the
plant and to its fiber, never to ‘hemp cloth’. The Korean word refers to
both, but this is not so significant compared to other problems that this
comparison presents. The word is actually attested in Middle Korean as
pwóy (Nam 1997: 698), and this creates a serious problem for the
comparison. We certainly have no way of knowing (and proving) whether
OJ wo < *wô or *wö, but neither OJ /ô/ nor OJ /ö/ corresponds to MK /wo/
(Whitman 1985: 213). Thus, we again have an irregular correspondence. In
addition, as I noted above in (19), there is no solid internal evidence for the
reconstruction of PK *-l- < PJK *-r- before MK /i/ or /y/; therefore the
reconstruction of Proto-Japanese-Korean with *-r is at best speculative.
That leaves us with a correspondence of initial consonants only. However,
even this correspondence is very doubtful, because Whitman presents only
three examples, including this one. As I demonstrated above, one of the
remaining two etymologies is a likely loanword (50), and another one (51)
cannot stand on semantic grounds. The etymology probably should be
rejected, although there is a slim possibility that it is a loanword. 69
All of the comparisons discussed above (1-43 and 50-52) represent
etymologies involving Proto-Japanese-Korean initial *p- and *b-.
Whitman has presented forty-three etymologies supporting PJK *p- and
only three etymologies supporting PJK *b-. I exclude the six etymologies
(44 through 49) that I rejected as obsolete. Even without my rejection of all
Proto-Japanese-Korean etymologies with *b- as supporting a genetic
relationship, there is obviously a skewed pattern. If my rejections are
accepted, we see a gap: no reliable Korean etymologies for Old Japanese
words with initial w-. Such gaps should not exist in a true genetic
relationship, but they are usually found in an areal one. Below I will
demonstrate that there are other gaps as well.

69
I believe that in the case of MK pwóy, and possibly OJ wo as well, we might be looking
at one of the most famous Eurasian Wanderwörter (Arabic bazz, Turkic böz, Chuvash pir,
Mongolian bös, Nanai boso, etc. ‘cloth, linen’) on which so much ink has been spilled
before that there is no need to add more here.
 Lexical Comparisons———113

3.2.3 *t-
(53) MK :ta ‘all’ ~ OJ taNta ‘only, just, exclusively’. Whitman notes:
“This comparison assumes the Old Japanese form is reduplicated. A few
other Old Japanese forms give evidence for the existence of this morpheme:
tamura ‘gathering, collection of similar types’ < ta +mura ‘group’; also
-Nta in kököNta ‘this much, this many’ < kökö ‘here’ + -Nta, ikuNta ‘how
much’ < iku ‘how, to what extent’ + -Nta, sapaNta ‘much, great(ly)’ <
sapa ‘much, great(ly)’ + -Nta” (1985: 214). This comparison has multiple
problems. First, the R pitch of MK :ta shows that the word was disyllabic
at one time: *taCV, but we have no means of knowing what the second
syllable was, and, therefore, any external comparison becomes
hazardous. 70 Second, assuming that OJ taNta is reduplicated is one thing,
but proving it is an entirely different matter, and the internal evidence does
not seem to support the reduplication hypothesis. Third, Whitman
compares ‘all’ with ‘only’, but these have opposite meanings: ‘all’ is
inclusive, while ‘only’ is exclusive. Fourth, the compounds Whitman cites
are unlikely to offer strong evidence for OJ *ta ‘all’ for the following
reasons: the compound tamura indeed probably includes mura ‘group’, but
the internal Japonic evidence does not tell us anything about ta-.
Explaining this ta- as ‘all’ on the basis of Korean is methodologically
unjustified, since it involves a ‘reconstruction from above’. Not only
tamura itself is attested only in Western Old Japanese; mura ‘group’ is
found in Eastern Old Japanese once, but only as a part of the makura-
kotoba mura-tama (MYS XX: 4390). Ryukyuan attestations are limited to
Shuri buri, Nakijin burii (used only in attributive position), and Psara murï
in murï ta ‘gathering’ (ta is ‘person’) (Shimoji 1979: 212). But none of
these three can be treated as a cognate, because they all exhibit phonetic
irregularities: Shuri and Nakijin have b- instead of the expected m-, and
Psara -ï instead of the expected -i (cf. Psara kuri ‘this’ [Shimoji 1979: 78]
< *kure). In addition, all three apparently reflect the Japanese form mure,
which started to be used as an attributive in the Middle Japanese period,
not WOJ mura. The Psara form murï was apparently borrowed from some
Ryukyuan source that had *muri, with *e > i, hence Psara has -ï (a regular
reflex of *i), but not -i (a regular reflex of *e). Thus, it looks like mura
‘group’ is confined to Western Old Japanese, and it is probably itself a
loan from Korean (cf. MK mwúl ‘group’). Other compounds cited by
Whitman involve not -ta, but -Nta, with -Nt- left unexplained. One
possible alternative is that -Nta is a collective suffix, cf. also OJ na-myi-
Nta ‘tears’ (= eye-water-COLL) (Vovin 2005a: 328). Thus, this etymology
is beset by too many problems to be accepted.
(54) (C)/(R) MK tàhí- ‘makes a fire, heats it with fire’; OJ tak- ‘id’.
(Whitman 1985: 214). There is only one Middle Korean attestation of this
verb:

70
It is possible that MK :ta ‘all’ is derived from the verb tàGò- ‘to exhaust, to use up’
(Nam 1997: 355), but one would rather expect *tàGó- with H pitch on the second syllable.
This, together with the less than perfect semantics, makes me somewhat suspicious of this
internal Korean etymology, although I do not reject it completely.
114———Koreo-Japonica

pwup twutuli-mye pul tahi-key ho-mye

drum beat-GER fire burn-CAUS do-GER
making [them] beat drums and make fires (Kwukup I: 15)

All other attestations belong to Early Modern Korean, so it may be that a

comparison with MK thó- < *tòhó- ‘to burn’ would be stronger, in spite of
slightly more divergent semantics. This is certainly a very minor point. The
Japonic forms, cognate to OJ tak- are attested in Ryukyuan, e.g., Shuri
tacuN (RGJ 1976: 505), Nakijin tachuN (Nakasone 1983: 230), Iejima
thachuN (Oshio 1999: 195), Psara taksu (Shimoji 1979: 135), all meaning
‘to cook on the fire’. The verb is not attested by itself in Eastern Old
Japanese, but there is the compound tak-î-N-KÏ ‘firewood’ (MYS XIV:
3433), 71 which includes tak-. Thus, although this appears to be a good
etymology at first glance, we are again faced with the problem of matching
MK -i to PJ *-Ø or *-a. This may be avoided only if we opt for the
reconstruction of PJ as *taka- instead of *tak- and compare it with MK
thó- ‘to burn’. However, even in this case we still get only a fifty percent
probability of being correct, since MK thó- can be either from PK *hoto-
or from PK *toho-.
(55) (L) MK :tal ‘rush’ 72 ~ OJ taNte HH, tara LF ‘smartweed’
(Whitman 1985: 214). Certainly the Middle Korean word cannot be related
to both OJ taNte and tara. OJ taNte < *tante should be out of the equation,
because OJ -Nt- does not correspond to MK -l-, whether the latter reflects
PK *-t- or *-l-; see also (15) above. Thus, only OJ tara ‘edible type of
smartweed’ remains, but there are problems here as well. First, one would
expect the Middle Korean to be something like *talom or *talam, to
account for the F pitch in OJ tara LF, but apparently there is no trace of the
final nasal *-m in Korean. Second, there is a problem with distribution in
Japonic: tara is not attested either in Eastern Old Japanese or in Ryukyuan.
Thus, WOJ tara may be a loan from Korean, but it is certainly not a
cognate.
(56) (R) MK tàlàk ‘upper story, loft’ ~ OJ taka- ‘high’ < PJK *tarak.
Whitman also adds Koguryo tal (達) ‘mountain’ here (1985: 214), but
unless one can cogently demonstrate that the Koguryo transcriptions were
based on Tang period Northwestern Chinese readings, which had final *-r
instead of EMC *-t, reading Koguryo 達 ‘mountain’ as tal is anachronistic,
as it is based on a much later Sino-Korean reading tal (which, incidentally,
is derived from Tang period Northwestern Chinese, see Martin [1997]).
Thus, Koguryo 達 ‘mountain’ is likely to be just *tat, and has nothing to
do with this etymology. This brings us back to MK tàlàk ‘loft’, but I
believe that the comparison is faulty due to the problematic
correspondence of MK -l- to OJ -k-. Martin seems to be the first who
suggested this correspondence, supporting it with six etymologies (Martin
1966: 211), of which Whitman retained three: this one, (66), and (144),
71
MYS XIV: 3433 appears to be written in pure Western Old Japanese, so it may not be a
valid attestation.
72
Phragmites japonica.
 Lexical Comparisons———115

with an addition of a fourth (65). Theoretically it is possible to provide an

explanation for such a correspondence, but the explanation must involve
regular phonetic development in all cases. In addition, all etymologies that
are used to support such an unusual correspondence must be absolutely
impeccable. Unfortunately, neither of these two conditions is maintained in
these four cases.

Chart 29:
Proto-Japanese-Korean reconstructions based on the correspondence
Middle Korean -l- : Old Japanese -k-
MK OJ PJK
(56) ‘high, loft’ tàlàk taka *tarak
(65) ‘attach, hang’ tól- tukë- *tokar
(66) ‘moon’ tól tukï *tokol
(144) ‘hang’ :kel- kak-/kakë- *keker-

The Proto-Japanese-Korean reconstructions for examples (65), (66), and

(144) are consistent, but for the present case Whitman presents quite a
different reconstruction. However, all four comparisons (and
reconstructions) suffer from some problems; therefore, they are far from
impeccable. I will analyze these problems for (65), (66), and (144) below.
The comparison of MK tàlàk ‘loft’ with OJ taka- ‘high’ is based
apparently on the assumption that OJ taka- < PJ *taraka- via *-r- loss. But
*-r- loss cannot be justified here on the basis of the internal Japonic
evidence. The final segment -a in taka- also is left unexplained. Finally,
there is another irregularity, since another etymology, PJK *parak (which
is very similar to *tarak), is reconstructed as a protoform for OJ paNkî and
MK pal; see (4) above. I reject the present etymology, because the
suggested correspondence is irregular and unique.
(57) (R) MK tàlì ‘leg’ ~ OJ te ‘hand, arm’ (Whitman 1985: 214). I do
not know of any cognates in any language family that mix up ‘leg’ and
‘hand’. The implied semantic is probably ‘limb’, but can we imagine any
language that has a word for ‘limb’ and no words for ‘hand’ and ‘leg’?
Such latitude in semantic comparison is not credible, and I reject this
etymology. For the same semantic weakness in Whitman’s comparisons
involving body parts, cf. also (219) and (250) below.
(58) (R) MK tàlóG- 73 ‘differs’ ~ OJ taNkap- ‘id’. < PJK *tarog-.
Whitman comments: “The Old Japanese suffix is verbalizing -p of yurup-
‘loosens, softens’ < yuru- ‘loose, gentle’ + -p and parap- (11)” (1985: 214).
There are several problems with this etymology. First, the adjectival root
*taNka- does not present itself, therefore the segmentation of -p in taNkap-
is not substantiated by internal Japonic data. The situation is the same for
parap-, see (11) above. Second, in spite of the fact that OJ taNkap- is
attested in both Western and Eastern Old Japanese as well as in Middle

73
The stems of this quality verb are actually tàló- and tàlG- in Middle Korean; *tàlóG- is a
Proto-Korean reconstruction.
116———Koreo-Japonica

Japanese, and its doublet MJ tiNkap- appears for the first time in the
Kagerō nikki (ca. 974), it is quite clear that the Proto-Japonic form was
*tinkap- for two reasons: (a) in Ryukyuan and in divergent Japanese
dialects only reflexes of *tiNkap- are attested; (b) the change *taNkap- >
*tiNkap- is unmotivated, but the opposite represents a well-known i-
breaking in the first syllable. Third, the Proto-Korean sequence *-loG-
does not correspond regularly to PJ *-nk- (> OJ -Nk-). Therefore, there are
sufficient grounds to reject this etymology.
(59) (C) MK :tam- ‘fills it’ ~ OJ tamë- ‘id’., tamar- ‘fill up with
(water)’. Whitman notes: “Intransitive OJ tamar- is clearly attested;
transitive tamë- appears in only one unclear attestation (MYS XIII: 3227)”
(1985: 215). Let us look at the segment of MYS XIII: 3227, where OJ
tamë- ‘to fill’ occurs:

MÎWO PAYA-MÎ OPÎ-tamë KATA-KÎ IPA-MAKURA

navigating channel fast-GER grow(INF)-fill(NML) difficult-ATTR rock-
pillow
rock bed, where it is difficult for reeds to grow and fill [the water], because
[the current in] the navigating channel is fast (MYS XIII: 3227)

I think that it is quite clear that OJ tamë- means ‘to fill’ here.
Unfortunately, it is a hapax legomenon, which does not present a solid
basis for external comparisons. But it can prove that segmentation of final
-r in WOJ tamar- (not attested in Eastern Old Japanese) is correct. There
are cognates of both OJ tamë- and tamar- in Ryukyuan, e.g., Shuri
tamiyuN, tamayuN (RGJ 1976: 509), Psara tamirü, tamarü (Shimoji 1979:
139-140). Therefore, I accept this comparison.
(60) (L) MK :tamón, :tamóyn ‘only, just’ ~ OJ -Ntamï ‘just, about’
(Whitman 1985: 447). There is also an Old Japanese variant -Ntamë (JDB
1967: 447). The best way to explain this variation is to assume that both
are derived from *Ntamöi, since *öi can become OJ /ï/ or /ë/. There are
two phonetic problems: MK /t/ does not correspond to OJ /Nt/, and MK /o/
does not correspond to OJ /ö/. In addition, the Japanese word is not attested
outside Western Old Japanese. I believe that OJ -Ntamï/-Ntamë is a
possible loanword from Korean to Western Old Japanese, or to its
predecessor, but definitely not a cognate due to its limited distribution.
(61) (R) MK tàt- ‘closes it, shuts, cuts off’ ~ OJ tat- ‘cut off, interrupt’
(Whitman 1985: 215). 74 There are two problems with this comparison.
First, MK tàt- has a non-leniting -t- and goes back to PK *tàntó-, and PK
*-nt- does not correspond to OJ -t- < PJ *-t-. Second, to the best of my
knowledge, MK tàt- means only ‘to close’, and the overwhelming majority
of textual examples refer to closing doors (Nam 1997: 370). The meaning
‘cuts off’ seems to have been brought in to improve the comparison.
Therefore, I reject this etymology.

74
This comparison is also found in Martin (1966), #46.
 Lexical Comparisons———117

(62) (R) MK táy ‘bamboo’ ~ OJ takî HH ‘id’. < PJK *tagi/*taki

(Whitman 1985: 215). 75 To the best of my knowledge only OJ takë
‘bamboo’ exists, while OJ takî is ‘waterfall’ (JDB 1967: 413, 416-417). OJ
takë < PJ *takay. The etymology is easily rejected on the basis of a major
irregularity: MK -Ø- does not correspond to OJ -k-. One would expect
something like MK *tàhí or *tàhóy for a cognate.
(63) (L) MK tó, tóy (< tó + -i nominative suffix) ‘place’ ~ OJ tô, tö, te
‘place‘. This etymology is followed by a long discussion at the end of
which Whitman says: “The vowel correspondence for MK tó(y) : OJ -te is
regular; for non-suffixed forms we would expect *ta. The occurrence of
both tô and tö in Old Japanese may have something to do with this
irregularity. There is also evidence for an earlier morpheme *ta in forms
like OJ sita ‘below’ < si ‘down’ + ta (viz. simô ‘downward’ < si ‘down’ +
mô ‘direction’), and könata, etc., ‘here, this way’ < kö ‘this’ + -na genitive
+ ta” (Whitman 1985: 215). There are multiple problems with this
etymology. First, as Whitman himself admits, neither OJ tô nor OJ tö
corresponds regularly to MK tó. Thus, they must be ruled out from the start.
The occurrence of both tô and tö cannot be used as an explanation for the
irregularity in suggested external comparison, because this irregularity is
easily explained internally: the original form is tô, which is regressively
assimilated to tö in the word tökörö ‘place’, the only case in which /tö/
occurs (JDB 1967: 485). In addition, it is now well known that the /tô/ :
/tö/ contrast collapsed in Late Western Old Japanese (Mabuchi 1972: 131)
or at least started to collapse, and there are other examples in Late Western
Old Japanese where original /tô/ was substituted with /tö/, as in the case of
nöritô ‘liturgy’ > nöritö (Bentley 2001: 6-7). Second, the form -te,
occurring in usirö-N-te ‘behind’ and omö-te ‘front side’, cited by Whitman,
indeed corresponds regularly to MK tóy, but the problem is that MK tóy is
morphologically divisible into MK tó and -i, while te is not. In other words,
it can only correspond regularly to MK tóy, but not to MK tó. Connecting
OJ -te to OJ -ta is speculative, but even if one accepts it, then -te should
consist of *ta + nominative suffix -i. But this ‘nominative’ suffix -i in Old
Japanese (if its existence can be cogently shown in this case in the first
place), as I demonstrated above in the morphology section, is a loan from
Korean. This suggests a loanword rather than a genetic relationship. Third,
the existence of -ta in compounds like sita and könata is not a particularly
strong case. In spite of the existence of both OJ sita and simô, the first
element *si does occur in an independent form. Also, OJ mô ‘side’ is
supported by EOJ wote mô ‘that side’ and könö mö ‘this side’ (MYS XIV:
3361), ta-nö mô ‘field side’ (MYS XIV: 3523), as well as by ya mo ‘eight
directions’ and yö mo ‘four directions’ in Western Old Japanese, the latter
attested several times. No comparable evidence can be provided for -ta. I
have demonstrated elsewhere that the alleged genitive marker -na does not
exist in Old Japanese (Vovin 1994: 249, 253; 2005a: 102-107); therefore
könata is likely to be an irregular development of *könö kata ‘this side’. In

75
This comparison is also found in Martin (1966), #7.
118———Koreo-Japonica

addition, the contrast between OJ sita and simô is not really between
‘below’ and ‘downward’, since OJ simô indicates the lower part of the
stream or the ‘down part’ more removed from the central part, while OJ
sita does not have these connotations (JDB 1967: 353). Thus, I reject this
etymology with the exception of OJ -te ‘place’, which could be a loan from
Korean.
(64) (L) MK tó in tó-lwó ‘because’ (+ instrumental -lwó) ~ OJ ta
‘because’. Whitman notes: “OJ ta also occurs in the apparent compound
tamë ‘because’, which eventually replaces it; the provenance of -më in the
latter form is unclear” (1985: 215). There are problems with this
etymology. First, MK tó in tó-lwo is likely to be MK tó ‘place’, discussed
above in (63). This internal etymology is strengthened by two facts: (a) it
never occurs by itself, but only within tó-lwó, so the meaning ‘because’ is
likely to be triggered by the instrumental -lwó; (b) very much unlike OJ ta,
it is found exclusively after the attributive forms of verbs, predominantly
after ile-n or kule-n ‘be such-ATTR’ (Nam 1997: 478). OJ ta ‘for’ can
occur after: (a) a NOUN + genitive nö, (b) the attributive form of a verb
with the following possessive Nka. Second, OJ ta has only Western Old
Japanese attestations. There are no attestations in Ryukyuan and the single
attestation in Eastern Old Japanese is problematic, as it has at least two
possible explanations:

(A)
pîk-ô pune-nö siri pîk-asi-mö yö kököNpa kô-Nka ta n-i
pull-ATTR boat-COMP buttocks pull-ADJ-EXCL PT PT with such
difficulty child-POSS for DV-INF
[I] am in the mood of pulling the buttocks [of my husband] like a tow boat!
[It] is with such a difficulty for children (MYS XIV: 3431)

(B)
pîk-ô pune-nö siri pîkasi mö yö kököNpa kô-Nkata n-i
pull-ATTR boat-COMP buttocks pull-ADJ-EXCL PT PT with such
difficulty come-difficult DV-INF
[I] am in the mood of pulling the buttocks [of my husband] like a tow boat!
As [he] comes [here] with such a difficulty (MYS XIV: 3431)

I believe that reading (A), which is required for establishing ta in

Eastern Old Japanese, makes much less sense than reading (B). A possible
setback for reading (B) is that in this Eastern song, which otherwise
faithfully indicates the kō/otsu distinctions, we would expect the stem of
the verb ‘come’ to be spelled as kö-, not as kô-. However, since this is an
Azuma song, where the preservation of kō/otsu is problematic, the
objection is a minor one. Therefore, WOJ ta and tamë coexist only in
Western Old Japanese texts. Even if my first objection above is rejected, it
is likely that WOJ ta is a loanword from Korean. Thus, I am inclined to
accept this etymology as a possible loan (with reservations about the
 Lexical Comparisons———119

probable internal explanation of MK tó- outlined above), but not as

evidence for genetic relationship between Korean and Japonic.
(65) (R) MK tól- ‘hangs it, attaches it, fixes it (on)’ ~ OJ tukë- ‘attaches
it, fixes it (on)’. Whitman comments: “OJ tar- ‘hangs it, dangles it down’,
tare- ‘hang it, dangle it down’ and tur- ‘hangs, strings’ have also been
compared with MK tól-. However, the semantic fit with OJ tukë- is better;
the phonological fit with OJ tukë- is also perfect if we assume a Proto-
Japanese-Korean protoform *tokar (cf. [66])” (1985: 215-216). I believe
the problem here lies in the assumption that the Proto-Japanese-Korean
protoform is *tokar-, which can be based only on another assumption
about the development of MK tól- from PK *tokal-. But such a Proto-
Korean form cannot be substantiated by the internal Korean evidence: we
would expect PK *tokal- to become MK *tohal-. or *tohol-. Therefore, I
reject this etymology. See also (56) above and (66) below.
(66) (R) MK tól ‘moon, month’ ~ OJ tukï, tuku- ‘id’. 76 Whitman
indicates that the reconstruction of OK TOlal-i ‘moon’ (spelled 月羅理,
etc., in Hyangka) proposed by Kim Wancin may be dubious, because tól
would not be an expected outcome of PK *tòlòl. He cites as a supportive
example MK mòlò ‘ridge’ < PK *mòlòl (1985: 216). However, Whitman’s
objection can be overruled, because the phonological shapes of MK tól and
mòlò are not identical: MK mòlò ‘ridge’ has a minimal vowel in the second
syllable; MK tól ‘moon’ does not. The former has L pitch on the first
syllable, and the latter has H. The morphophonology is also different: MK
mòlò has an alternating stem mòll-, but MK tól does not. In addition, while
we do not know how to read the first syllable of OK TOlal (and probably
never will, since it is hidden behind semantographic spelling), the
remaining part of the word -lal is not a reconstruction, but a philological
reality that is confirmed by Old Korean texts (Hyangka IV: 2, IX: 1, XIII:
5). Whitman further presents the ingenious idea that the Old Korean form
is actually *TOClal with a cluster *Cl resulting from the loss of
unaccented *ò or *ù, because while lC clusters are attested in Middle
Korean, *Cl clusters are not (1985: 216). The idea is really brilliant, but
the problem is that there is no second piece of evidence we can use to
verify this, so I am afraid it must remain speculative and cannot be used as
evidence for establishing an external etymology. Furthermore, for the sake
of regularity in this comparison, we should expect the Old Japanese form
to be *tuka, not WOJ tukï, EOJ tuku < PJ *tuko, 77 since it is OJ /a/ that
should correspond to OK /a/ in TOlal (provided Whitman’s cluster
reconstruction is disallowed). Therefore, I reject this etymology: it works if
we use an unverifiable hypothesis that is not possible to prove or accept an
irregularity. See also (56) and (65) above.
(67) (L) MK -tólh/-túlh (nominal plural suffix) ~ OJ -tati ‘id’.
(Whitman 1985: 217). Actually, as pointed out above, there is only MK
-tólh (see 1.1.3.1, especially note 4). The form -tulh is attested essentially
only in Early Modern Korean, with a single Middle Korean attestation
76
This comparison is also found in Martin (1966), #143.
77
Final *o is reconstructed on the basis of PR *tukoya ‘moon’.
120———Koreo-Japonica

from 1586. The problem with this comparison lies in the limited usage and
distribution of OJ -tati. The usage of the Old Japanese suffix -tati is
restricted to animate nouns. For both Old and Classical Japanese it is
believed to be an honorific plural marker used exclusively with the names
of deities and with the titles of princes, princesses, and other relatives of an
emperor or empress (Yamada 1954: 537). However, this is not completely
true. In the Senmyō, the suffix -tati occurs with the second person pronoun
twice (SM 17, 37), directed to lower-ranking courtiers (SM 17) and
provincial governors (SM 37); it also occurs with the words omî ‘noble’,
and possibly with asômî ‘retainer’, all social inferiors to the emperor and
princes. Since all Senmyō edicts represent the emperor’s words, -tati can
hardly be taken as an honorific marker. Yet, its usage seems to be
restricted to gods and members of the court. The plural suffix -tati is not
attested in Eastern Old Japanese. There is plural marker -taa in Shuri and -
caa in Nakijin, but it seems that these do not correspond regularly to OJ -
tati. In addition, there is a single attestation of -tati in the Omoro sōshi (OS
XII: 740), but this is likely to be a loan from Japanese. Given its
distribution in Ryukyuan, and lack in Eastern Old Japanese, it seems that
the plural marker -tati is confined to Western Old Japanese. Limited
distribution and narrow semantics make -tati a perfect candidate for a loan.
Therefore, the genetic comparison of -tati with the Middle Korean plural
marker -tólh should be revised in the light of the data: -tati in all
probability represents a loan from the Old Korean predecessor of MK -tólh
in a form predating the lenition t > l and the vowel reduction a > o. This
conclusion can be further supported by the fact that MK -tólh is a neutral
marker of plurality, while Old Japanese -tati significantly narrowed the
meaning, using it only for deities and members of the elite. This can be
expected, since the Japanese nation was essentially founded by people
from the Korean peninsula.
(68) (R) MK tòlì ‘bridge, ladder, stair’ ~ OJ ti ‘road’ < PJK *tor (+ -i)
(Whitman 1985: 217). Unfortunately, we do not know what Proto-Japonic
form underlies OJ ti: *tö-i, *tu-i, or *ti. Only the second form *tu-i would
work for this comparison, but there is only a one in three chance that it is
correct. One must also assume the loss of *-r- in Old Japanese, but it
cannot be confirmed internally. In addition, roads surely go across bridges,
but they do not climb ladders. Even the semantic shifts ‘road’ > ‘bridge’
and ‘bridge’ > ‘road’ are not very convincing. Thus, this etymology should
be rejected: it is based essentially on the initial consonant, and it has
problematic semantics.
(69) (R) MK tóm- ‘sinks’ ~ OJ siNtum- ‘id’. Whitman comments: “This
comparison assumes that the additional syllable in the Old Japanese verb is
directional si- ‘down” (Whitman 1985: 217). There is no internal Japonic
evidence to support this segmentation; see also (63) above on the
speculative nature of OJ *si ‘down’. In addition, the correspondence of
Middle Korean initial t- to OJ Nt- is irregular. It is possible that, if the
segmentation above were proved, OJ *-N- could be demonstrated to be
morphologically separate from *-tum-, but apparently this has not been
 Lexical Comparisons———121

done. I reject this etymology due to the speculative nature of the

segmentation proposed for the Old Japanese form.
(70) (R) MK tech ‘trap, snare’ ~ OJ töNti- ‘closes it’ < PJK *tedö-.
Whitman also adds OJ töNtömë- ~ töNtömï- ‘stops it, catches it’ (1985:
217). As a matter of fact, there are no Middle Korean attestations, only
EMdK tes (from 1690 in the Yek.e yuhay), in addition to MdK tech (LCT
1987: 217; Nam 1997: 399). Since the Early Modern Korean attestations
are all Korean glosses for Chinese characters, we can only speculate that
MdK tech reflects the original form. We can further speculate that MdK
tech goes back to PK *tecuh or *tehuc. Only the first form can be
compared (with some difficulties) to OJ töNti-, but the dialect data, such as
the form tek attested in Kyengsang Namto and Cenla Namto, as well as
thel or thul (if related at all) attested in various modern dialects (Choy
1978: 585), unmistakably point to PK *tehuc. The case is further undercut
by the difficulty of reconstructing a proper Proto-Korean vocalism for this
word at the present stage of our knowledge, since dialects show
considerable variation: e ~ a ~ wo ~ u (Choy 1978: 585). These difficulties
largely invalidate the comparison with Japanese, and the etymology must
be rejected.
(71) (L) MK :twochóy ~ :twochwóy ‘axe’ ~ OJ tuti LH ‘mallet’ < PJK
*tücu (+ -i) (Whitman 1985: 217). A Western Old Japanese pre-contraction
form tutui is also attested (KK 10, NK 9). MK :twochwóy should go back
to PK *twòhòcwóy or *twòcòhwóy, as indicated by the R pitch on the first
syllable and the cluster -ch-. In the case of PK *twòhòcwóy, a genetic
comparison would be impossible, as we would expect an Old Japanese
form like *tukutui. In the case of PK *twòcòhwóy it would be very
difficult, because final *-hwoy in Korean would have to be a suffix, an
assumption for which there is no internal evidence. The easiest explanation
is to assume that this is another loan from Korean into Japonic that
postdates the contraction of PK *-hòc- or *-còh- into -ch-. The second
piece of evidence that supports the loanword scenario is based on the
possibility that the Proto-Japonic word had a long vowel in the first
syllable, as demonstrated by an Old Japanese loanword in Proto-Ainu
*tuuti (Vovin 1993a: 150). The vowel length certainly reflects later Korean
R pitch; which would not be expected in the case of a genuine cognate.
(72) (L) MK twòk ‘pot’ ~ OJ tukî HL ‘id’. < PJK *tuk (Whitman 1985:
217). MK twòk is really a ‘jar’, and OJ tukî is more likely to be a ‘saucer’,
‘shallow cup/plate’, at least in the Asuka-Nara periods (JDB 1967: 461),
but this certainly does not invalidate the comparison. The problem is that
OJ tukî is attested only in Western Old Japanese and Middle Japanese (MJ
tuki), and there are no attestations in Eastern Old Japanese or Ryukyuan. 78
This distribution strongly suggests that WOJ tukî is a loan from Korean.
(73) (R) MK tè ‘more’ ~ OJ -tö ‘and, also, plus’. Whitman further
comments that OJ -tö is postpositional (1985: 217). OJ -tö is surely just a
78
There is an isolated Shuri haimaa sakazici ‘sake cup’ (RGJ 1976: 689), but it is likely to
be a loan from Japanese, because it is: (a) isolated in Ryukyuan and (b) found only within a
compound which also includes native haimaa with the same meaning.
122———Koreo-Japonica

comitative case marker, which can roughly be translated into English as

‘with, and’. It has no meaning ‘plus’, which is used solely to make the
comparison look better. Given the fact that functionally and syntactically
the Middle Korean word (which is found exclusively as a preposition) is
completely different, I do not think that much remains of this etymology.
(74) (R) MK tyé ‘that’ (distal) ~ OJ sö ‘that’ (medial) (Whitman 1985:
217-218). 79 This was discussed in the section on demonstrative pronouns
above (2.1.2.3).
(75) (C) MK tùl- ‘holds it, raises it’ ~ OJ tör- ‘takes it, grasps it’ < PJK
*tör- (Whitman 1985: 218). 80 The etymology seems valid, but there is one
slight problem. As Bentley has demonstrated, there are two Old Japanese
verbs: tôr- ‘to hold, to support’ and tör- ‘to take, to pick up, to capture’
(Bentley 1999). MK tùl- corresponds regularly to OJ tör-, but it is a better
semantic fit with OJ tôr-. Tentatively I accept this etymology.
(76) (R) MK tùlí- ‘falls’ (? = tùlí- ‘hang down’) ~ OJ tir- ‘falls’ < PJK
*tör- (Whitman 1985: 218). It seems that the meaning ‘falls’ for MK tùlí-
is a misglossing by Yu Changton (LCT 1987: 249). It is found only once in
the following example, where tùlí- is used as a translational equivalent of
the Chinese character 落 ‘to fall down’:

TUNG-s kaci-non pwontoy cey tuly-e-ys-kwo

wisteria-GEN branch-TOP naturally itself hang down-INF-exist-GER
Wisteria branches are hanging down naturally by themselves (Twusi cho
XV: 15a)

It is quite clear that the wisteria branches are ‘hanging down’, not
‘falling down’. In numerous other examples MK tùlí- also refers only to
‘hanging down’ (LCT 1987: 249; Nam 1997: 455). Notice that the
progressive form of the verb is unlikely to be used with a punctuated action
like ‘to fall’. This is a minor problem, though, but there is clearly a major
one: MK /u/ cannot correspond regularly to OJ /i/, regardless of the
provenance of the latter; only MK /uy/ can correspond to OJ /i/ in cases
where it is from *ï < *öi. Cf. (75) above, where PJK *tör- is reconstructed
on the basis of a different (and regular) correspondence. I reject this
etymology on the basis of its irregularity.
(77) (L) MK túmúl- ‘rare’ ~ OJ tömö-si- ‘scarce, longed for’ < PJK
*tömör- (Whitman 1985: 218). ‘Longed for’, or to be more exact,
‘enviable’, certainly represents a secondary development in Old Japanese,
so semantically the etymology is valid. There are no obvious phonological
problems either, although the Middle Japanese form is tobosi-, possibly
indicating an original *-np- cluster, which was simplified in Old Japanese
to -m-. However, since this is a problem that has no generally accepted
solution, and opinions differ on whether one has to reconstruct *-np- or
*-m- here, I accept this etymology, but it is likely a loan: the word is
attested in Eastern Old Japanese, but not in Ryukyuan.
79
This comparison is also found in Martin (1966), #237.
80
This comparison is also found in Martin (1966), #106.
 Lexical Comparisons———123

(78) (R) MK stwó ‘and, also, too, again, while’ (adverb) ~ OJ -tutu
‘while at the same time as’ (postverbal particle conjunction joining two
clauses) < PJK *totu (Whitman 1985: 218). Besides EOJ -tutu, which is
frequently used, there are two specific Eastern Old Japanese forms that
each appear once: -tusi (MYS XX: 4386) and -toto (MYS XX: 4421). The
latter, if it were not a hapax legomenon, could possibly offer some more
support to Whitman’s etymology, but in addition to functional problems,
there is an insurmountable problem in its irregularity: MK s- does not
correspond to OJ t-. Of course, some linguists argue that MK s- in sC
clusters is not to be taken at face value, but even if it were so (which is
unlikely, because we should not forget that hankul in the fifteenth century
represented a phonetically oriented writing system), there is no way to
demonstrate that this MK st- < PK *tt-. I reject this etymology on the basis
of its irregularity.
(79) (R) MK twùlG- ‘circles, turns’ ~ OJ tuNk- ‘succeeds, continues’ <
PJK *türk- (Whitman 1985: 218). Besides the Middle Korean variant with
stems twùl- ~ twùlG-, there is also the variant twùlù- (Nam 1997: 439). 81
This variation points to the Proto-Korean form *twùlùG- or *twùlùk-.
However, PK *-luG- does not correspond regularly to OJ -Nk- (< PJ *-nk-),
as was pointed out in (58) above. I reject this etymology on the basis of its
irregularity and rather far-fetched semantics.
(80) (C) MK twúlwúmí ‘crane’ ~ OJ turu LF ‘id’. < PJK *türüm (+ -i)
(Whitman 1985: 218).82 I accept this etymology as a possible cognate.
(81) (L) MK twùté, twùtí-, twùtú- ‘earth’ ~ OJ tuti LL ‘id’. 83 Whitman
presents a long supporting discussion from which I quote: “The MK
morpheme appears in twùtén, twùtúlk (MdK twuteng, twutek, twuleng)
‘bank, levee’ and twùtícwúy, twutecwuy ‘mole’ < twùté + cwúy ‘rat’ (i.e.,
‘earth rat’) (Martin 1966, #71) ... Old Japanese ‘bank, embankment, levee’
is tutumî LLH, a nominalization from the verb tutum- ‘to heap up dirt’.
This verb is itself derived from the original noun root tutu- LL with the
verbalizing suffix -m; it allows us to recover the original second syllable
vowel. Since we assume OJ tuti LL [tutï] < pre-OJ *tutu LL + -i, the
phonological fit here is perfect” (Whitman 1985: 218). It may not be as
perfect as it seems, because the Middle Korean forms have a non-leniting
-t- < PK *-nt- (MdK twuleng, which has -l-, does not count, because it
occurs later and may invite different explanations). MdK twuntek ‘low hill,
mound, hillock’, which preserves original *-nt- may not be completely
irrelevant. Thus, we have PK *twùnté-, *twùntí-, *twùntú- which could
correspond regularly to OJ *tuNti < PJ *tuntu-i, but not to OJ tuti < PJ
*tutu-i. Another problem is that among the Middle Korean forms twùté-,
twùtí-, twùtú-, the last one is obviously secondary because it includes a
minimal (i.e., reduced) vowel /u/; but it is the only one that can correspond
regularly to OJ tuti < *tutu-i, because neither MK /e/ nor MK /i/

81
An l-doubling stem tull- seems to be attested only in Early Modern Korean texts (Nam
1997: 439).
82
This comparison is also found in Martin (1966), #55.
83
This comparison is also found in Martin (1966), #71.
124———Koreo-Japonica

corresponds to PJ *u in the second syllable. Irregularities certainly rule out

the possibility of a genetic relationship. In addition, the distribution of OJ
tuti ‘earth’ seems to be limited to Central Japanese. There is EOJ tuti,
attested only once in MYS (XX: 4418). More important, there is only one
isolated attestation in Ryukyuan: Ishigaki tsïtsï ‘earth’ (Hirayama 1966:
341). Since the word is isolated, it is likely to be a loan. Otherwise
throughout the Ryukyus one finds a completely different word for ‘earth’:
*mita (Hirayama 1966: 341; 1967: 315). Therefore, OJ tuti represents
another loan from Korean into Central Japanese which postdates the loss of
postvocalic nasals in Korean: PK *twùnté > twùtú > pre-OJ *tutu-.
(82) (R) MK :twuyh ‘rear, behind’ ~ OJ usirö ‘id.’, siri ‘rear, rear end’
< PJK *(ü)tyü:r > pre-OJ *tiri > siri. Whitman adds: “Cf. OJ yöri ‘behind,
after, from’; yuri ‘behind, after’ < ? *dyür” (1985: 218). This etymology is
apparently based on the assumption that MK :twuyh < *twulih. Although
such a reconstruction is certainly possible, it cannot be proved internally in
this particular case; cf. (19) above. There are three other irregularities as
well. First, MK /wu/ does not correspond to OJ /i/, regardless of the
provenance of the latter. Only MK /wuy/ can correspond to OJ /i/ if it is
from *ï < *ui. MK /wuy/ is certainly present in Middle Korean :twuyh, but
in that case OJ -r- has nothing to correspond to in the Middle Korean form.
Second, Whitman claimed that Middle Korean initial ty- corresponds to
Old Japanese initial s-, cf. (74) above and (84) below, but there is no ty- in
MK :twuyh. Third, MK -y (< *-i) does not correspond regularly to OJ /ö/.
Finally, OJ yöri is a ghost; the correct form is -yôri, the ablative case
marker, which has nothing to do with this etymology. Thus, this etymology
is rejected on the basis of multiple irregularities.
(83) (L) MK twúy, ptwúy ‘miscanthus reed’ ~ OJ ti (< *tï) ‘id’. ~ tu- in
tu-N-pana ‘miscanthus flower’ (Whitman 1985: 218). The
correspondences are impeccable, but there are no cognates of OJ ti in
Eastern Old Japanese and/or Ryukyuan. I believe that it is a loan into
Central Japanese from Korean.
(84) (R) MK :tywoh- ‘good’ ~ MJ suk- ‘like, be fond of’ < PJK *tyuk-
(Whitman 1985: 218). 84 There are several problems here that give me
pause. First, MJ suk- originally meant ‘be lustful, be passionate, be
fanciful’, and the broader meaning ‘to like’ does not seem to be attested
before Early Modern Japanese. Cf. the following example from Ise
monogatari:

mukasi-no waka-bito fa sar-u suk-er-u mono-omof-i-wo namu s-iker-u

old times-GEN young-person TOP be such-ATTR be passionate-PROG-
ATTR thing-think-NML-ACC PT do-RETR-ATTR
Young people of the past had such passionate thoughts (IM XL)

84
This comparison is also found in Martin (1966), #125.
 Lexical Comparisons———125

The semantics of the comparison are rather far-fetched. Second, there

are no Eastern Old Japanese attestations, and in Ryukyuan the verb seems
to be attested only in Nakijin: sicuN ‘to like’ (Nakasone 1983: 185), the
meaning being suspiciously close to later Japanese, but not to Middle
Japanese. There is also Shuri sichi (RGJ 1976: 465) and Iejima sichii
(Oshio 1999: 173), but these are nouns, exactly like Modern Japanese suki,
and corresponding verbs do not exist. The Shuri usage 85 is exactly like that
of Modern Japanese: sichi-na Qchu ‘person whom [I] love’ (cf. MdJ suki-
na hito). The verb is also attested in Old Ryukyuan, once in the Omoro
sōshi in the passive form sukar- (OS V: 267), and once in the Ryūka, also
in the passive form sukar- (RK 864). In both Old Ryukyuan cases we again
witness the late meaning ‘to like’. In addition, all Ryukyuan attestations
are confined to the area around the island of Okinawa, and the absence of
the word in the Sakishima dialects is conspicuous. Thus, I think that all of
the cases above are loans from Japanese, some of them quite recent. Thus,
MJ suk- ‘be lustful, be passionate, be fanciful’ seems to be limited to
Central Japanese. However, I doubt that in this case we are dealing with a
loan from MK :tywoh- ‘good’. These doubts are based not only on the
improbable semantics, but also on the phonologically odd shape of the
Middle Korean word itself. To the best of my knowledge, the shape tywo
in the first syllable is not just rare in native Middle Korean words, it occurs
only in MK :tywoh-: there are no other native Middle Korean words that
have tywo-. 86 This is a fact that cries out for an explanation, and it has to be
explained before any external comparisons are made. I suspect that the
Proto-Korean form of MK :tywoh- was something like *tiCwoko-,
although there might be other possible solutions. In any case, the lack of a
proper Proto-Korean reconstruction makes a comparison with MJ suk-
doubtful, and consequently the suggested correspondence of MK ty- to OJ
s-, supported by two more rejected examples, (74) and (87), becomes
untenable.
(85) (R) MK tìh- ‘pounds it (with mortar)’ ~ OJ tuk- ‘id’. < PJK *tük-
(Whitman 1985: 218). This comparison must be rejected, because it
involves an irregular correspondence of MK /i/ to OJ /u/. Cf. (79), (80),
and (81) above, where Whitman reconstructs PJK *ü on the basis of the
correspondence of MK /wu/ to OJ /u/. See also (86) below.
(86) (R) MK tìk- ‘sticks it (into), points it, pierces it’ ~ OJ tuk- ‘pierces
it, sticks in (a cane)’ < PJK *tük- (Whitman 1985: 219). Here we have the
same irregularity problem as in (85) above. Note also that there is another
irregularity in this example. If in (85) MK -h- corresponded to OJ
-k-, here we have not MK -h-, but MK -k- corresponding to OJ -k-, but
both comparisons yield the same Proto-Japanese-Korean reconstruction
*tük-. As was mentioned above in the section on lenition in Middle Korean,
MK -k- < PK *-nk- (see also [3] above). Therefore, MK tìk- < PK *tinkV-

85
No comparable data are available on the Iejima usage.
86
There are certainly numerous hypercorrected spellings with tywo- for the etymological
cwo- in Early Modern Korean, but these can be disregarded.
126———Koreo-Japonica

cannot correspond to OJ tuk-. Since two out of the three correspondences

are irregular, I reject this etymology.
(87) (R) MK típ ‘straw, cut grass’ ~ OJ siNpa HH ‘grass, turf’
(Whitman 1985: 219). 87 No Proto-Japanese-Korean reconstruction is
provided. The accusative form tiph-ul (PT I: 22), attested in the first
edition (1515) of the Pak thongsa, makes it clear that the underlying form
of the Middle Korean word is actually típh. This is also supported by MdK
ciph ‘straw’. Therefore, MK típh should go back to PK *tikup or *tipuk.
The former Proto-Korean form, *tikup, would rule out any possibility of
comparison with OJ siNpa, but even the form *tipuk might present a
difficulty, since we do not know whether final *-k is a suffix or belongs to
the root. The correspondence of MK ti- to OJ si- is also suspicious, even if
we treat it as a variant of the already rejected correspondence MK ty- to OJ
s-, see (84) above. If we treat it as a separate correspondence, then it also
becomes irregular, because there are no other examples. There is a further
semantic problem: OJ siNpa is a general name for different weeds that
grow beside a road or in a wasteland. MK típh, as far as I can tell on the
basis of textual examples, is either ‘straw’ or ‘edible grass’. A weed is
certainly not an edible grass. These problems force me to reject this
comparison.
Etymologies (88) through (93) (Whitman 1985: 219) are all rejected,
since they involve an outdated treatment of MK th- (see 1.1.1).

3.2.4 *d-
(94) (R) MK :twolh ‘stone’ ~ OJ isi HL, isô HH ‘id’. < PJK *dil2o (+ -
ga) (Whitman 1985: 219). 88 This famous etymology rests on the
assumption that OJ isi ‘stone’ < PJ *yisi < *disi. The reconstruction of *d-
for Proto-Japonic was refuted above (see the section on *b- and *d- in
Proto-Japonic). This etymology is plagued by problems on both the Korean
and the Japonic sides. PJ *disi HL ‘stone’ presents several problems. The
presence of initial *d- or *y- at first glance cannot be proved or disproved,
but the vowel of the first syllable in Proto-Japonic is *e-, which can be
convincingly reconstructed on the basis of the Ryukyuan evidence. Proto-
Ryukyuan treats primary *e and secondary *e < PJ *ay, *ia in the same
way, so we would expect a preservation of PR *d- or *y- in front of
primary or secondary *e, cf. OJ yeda ‘branch’, ye ‘handle’, and PR *yUda
‘branch’, *yUe ‘handle’ (Thorpe 1982: 267, 293). The Ryukyuan evidence
might seem sufficient to exclude any initial *d- or *y-, but it is always nice
to have a second independent piece of evidence, and such a piece can be
found in Eastern Old Japanese. Before we consider this evidence, we must
consider another Old Japanese word included by Whitman in the
etymology, isô HH ‘rock‘, which is connected with OJ isi HL ‘stone’. The
seemingly unusual i : ô correspondence can be easily explained on the
basis of OJ isu ‘rock’ that is used as a doublet for isô in the name of the
shrine: Isô-nö kamî ~ Isu-nö kamî ‘top of the rock’, also written in
87
This comparison is also found in Martin (1966), #225.
88
This comparison is also found in Martin (1966), #224.
 Lexical Comparisons———127

semantographic spelling as 石上 ‘top of the stone’ (JDB 1967: 76). Thus,

OJ isi ‘stone’ is probably < *isï < *isuy, and OJ isô ‘rock’ < *iso.
Unfortunately, isô has no reflexes in Ryukyuan, but it is attested in Eastern
Old Japanese. Eastern Old Japanese has isi ‘stone’, which is attested twice
in the Azuma uta (MYS XIV: 3398, 3425). However, with the exception of
a dialectal word teko ‘girl’ in MYS XIV: 3398, none of these poems has
any other Eastern Old Japanese dialect features. We really do not know
whether isi in these poems is a genuine Eastern Old Japanese word, or
whether it is used under Western Old Japanese influence. EOJ isô ‘rock’ is
attested four times (MYS XIV: 3563; MYS XX: 4324, 4328, 4338), but the
same word also appears as osi in MYS XIV: 3359 and as osu in MYS XIV:
3385. Now, if the word in question had an initial *d- or *y- in Proto-
Japonic, why do these Eastern Old Japanese words not show up as *yosi
and *yosu? The answer seems to be quite straightforward: Proto-Japonic
has *esuy HL-L ‘stone’ and *eso HH-H ‘rock’, without any initial
consonant. Thus, PJ *esuy HL ‘stone’ cannot be related to MK :twolh,
because the correspondence of MK t- to OJ Ø- is irregular. The
correspondence of MK /wo/ to OJ /i/ < PJ *e is also irregular. Finally, the
reconstruction of MK :twolh is likely to be *twòlóh, where -h is likely a
reflex of the diminutive suffix -(a)k that follows both nominal and verbal
stems (Martin 1992: 416, 592). I believe that Martin is right and there is an
internal etymology for this word: MK :twol- ‘turn’ (intr.) < PK *twòló
‘pebble’ < ‘that which rolls’ (1996: 36). Even if all these facts were not
true, the correspondence of MK /wo/ to OJ /i/ is still irregular. Thus, this
etymology must be rejected.
(95) (R) MK :ti- ‘forges it’ ~ OJ ire- ‘id’. < PJK *dira- (Whitman 1985:
219). There are several problems with this etymology. First, the basis for
Whitman’s claim that it was a lower bigrade verb *ire- in Old Japanese is
unclear to me. There is only one Late Western Old Japanese attestation in
the form of a gloss, partially written in man’yōgana, as 伊ル /iru/ (NR II:
17), but it is impossible to tell the conjugational pattern of the verb on the
basis of this form alone. All other attestations are later, with the earliest
found in the Shinsenjikyō (898-901) as i-mono ‘cast thing’ (JDB 1967:
106), which demonstrates that the verb is upper monograde, with a root i-.
This is confirmed by all later attestations. Second, in contrast to (94) above,
where we can certainly say that there was no PJ *y- in front of OJ /i/, in
this case, as far as I can tell, we have no evidence for the presence or
absence of *y-. It is possible that OJ i- < PJ *yi-, but it is also possible that
it is from *i-. With no evidence for the initial consonant, the comparison
with Korean is a comparison based on one phoneme. And, of course,
MK :ti- goes back to a disyllabic form *tiCV-. Not only do we not have
any evidence for reconstructing the exact phonological shape of this word
in Proto-Korean, we have no evidence that OJ i- contracted from an earlier
disyllabic form. Therefore, this etymology must be rejected.
128———Koreo-Japonica

(96) (R) MK túl- ‘enter’ ~ OJ ir- ‘id’. < PJK *dirV- (Whitman 1985:
219). 89 The Proto-Japonic form is *er-, on the basis of PR *er-, but as in
(94) and (95) above we have no evidence to support PJ *yer-, which is
badly needed for this comparison. On the contrary, PR *er- suggests that
there was no PJ *y- in this word, although unlike (94) there is no
corraborating evidence from Eastern Old Japanese this time that can
completely rule it out: the only attestation is EOJ ir-. In addition, MK /u/
does not correspond regularly to OJ /i/. Thus, out of three correspondences,
one is highly questionable at least, and one is irregular. Therefore, there are
sufficient grounds for rejecting this etymology.
The comparisons involving numbers (53-87) and (94-96) represent
etymologies involving Proto-Japanese-Korean initial *t- and *d-. Whitman
presents 35 etymologies supporting PJK *t- and only three supporting PJK
*d-. The ratio is very similar to the ratio of etymologies with *p- and *b-
discussed above. Even if I had not rejected all Proto-Japanese-Korean
etymologies with *d- as supporting a genetic relationship, there is
obviously a skewed pattern. If my rejections are accepted, there is a gap
again: no reliable Korean etymologies for Old Japanese words with initial
y-. 90 Thus, in addition to OJ w-, we have another gap in correspondences.

3.2.5 *k-
(97) (R) MK -ká interrogative suffix ~ OJ -ka id. < PJK *ka. Whitman
adds: “In both languages -ka is preceded by a noun or a nominalized
(attributive) verb form” (1985: 220). At first glance the etymology seems
impeccable, but there is one problem. PJ *ka is an interrogative particle
used in wh- questions. 91 MK ká ~ Gá is the interrogative particle that
occurs only in general questions:

hanolh-i kolhoy-Gi-si-n-i nwupi cywong an i-ntol haytwong lyenmin-ul

nic-osi-l i-ngi-s ka
heaven-NOM choose-CAUS-POL-ATTR/REAL-NML quilt monk not be-
GER Korea people-ACC forget-HON-ATTR/IRR be-POL-NML PT
Heaven has made the choice. Were it not for a monk in the quilted robe,
would [Heaven] forget the Korean people? (YP 21)

thwukwu sey sal-i nyey two stwo is-te-si-n ka

helmet three arrow-NOM old.times PT still exist-RETR-HON-
ATTR/REAL PT
Did it happen even in the past that three arrows [all hit] the helmet? (YP
89)

89
This comparison is also found in Martin (1966), #76.
90
Whitman also presents four etymologies with OJ y- reflecting PJK *j- and four
etymologies with OJ y- reflecting PJK *y- (190, 199-202 [there is no 201 — A.V.], 313-
317). See the discussion under these number.
91
WOJ ka is predominantly used in wh-questions, although sometimes it may occur in
general questions. PR *ka occurs only in wh-questions.
 Lexical Comparisons———129

This problem is further aggravated by the fact that the interrogative

particle for wh- questions in Middle Korean is kwó ~ Gwó (An and Yi
1990: 204-205). This leaves the comparison with Japanese rather lame: at
best it is an etymology based on one phoneme, /k/, common to both Middle
Korean particles ká ~ Gá and kwó ~ Gwó. In addition, it leaves
unexplained the vowel /a/ in WOJ ka; and we should not forget the
functional difference. Thus, this etymology should be rejected.
(98) (R) MK kàch ‘leather, skin’ ~ OJ kasa HH ‘blemish, hardening of
skin, scab’ < PJK *kaso (Whitman 1985: 220). MK kàch < PK *kàcók, cf.
MdK kacwuk, which actually is closer to the protoform. The final *-k may
be a suffix, but we have no definite internal evidence that allows us to
segment it. There are no phonetic attestations for OJ kasa (JDB 1967:
184); the word is effectively found only in Middle Japanese. This is a very
minor point. The crucial problem is that it does not mean ‘blemish,
hardening of skin, scab’, but ‘swelling, boil, abscess’. The word is not
attested in Eastern Old Japanese, and Ryukyuan attestations are few and
far-between: Urazoe-Kowan kasa ‘swelling, boil’ (Nakamoto et al. 1995:
77), Shuri kasa ‘malignant swelling’ (RGJ 1976: 311), Nakijin ka[a]saa
‘swelling, boil’ (Nakasone 1983: 97), Psara kasa ‘general name for skin
diseases’ (Shimoji 1979: 51). The Proto-Japonic semantic archetype seems
to be ‘swelling’, and semantically the comparison with Korean is not
possible. Therefore, I reject this etymology.
(99) (R) MK kàcí ‘kind, sort, variety’ ~ OJ kata HL ‘shape, form, mold’
(Whitman 1985: 220). Phonetically the comparison may be viable, but
there is one serious obstacle: one has to explain final MK -i as a suffix. I
fail to see any internal evidence for this. In addition, there is no semantic
coherence. Therefore I reject this etymology.
(100) MK kácí ‘branch’ ~ OJ ka, kai, kaNti ‘oar’ < PJK *kadi
(Whitman 1985: 220). 92 The reconstruction of the archetype of the Old
Japanese word is difficult. First, OJ kaNti is not an ‘oar’, but a ‘rudder’.
Rudders on Old Japanese boats were just ‘big oars’, so this may not be so
important. The problem is that OJ kaNti ‘rudder’ belongs to the 2.2 (HL)
accent class, while OJ kai ‘oar’ is 2.4 (LH). Thus, the register is not
compatible. Consequently, OJ kaNti ‘rudder’ and OJ kai ‘oar’ are not
related, and OJ kaNti should be taken out of the equation. Second, OJ kai is
a phonologically aberrant form, since VV clusters were not permitted in
Old Japanese. Martin reconstructs PJ *kaCi (1987: 433). I have recently
demonstrated that the form was actually kayi as late as in Western Old
Japanese (Vovin 2009: 420ff).Thus, a comparison with Korean is out of
the question. Moreover, there is OJ ka ‘oar’,93 also cited by Whitman,
which appears by itself written phonetically only once in Old Japanese
(MYS XX: 4408), but WOJ ka-kô ‘oarsman, rower’ appears in phonetic
script three times (MYS XV: 3627; MYS XX: 4331, 4408). While it is

92
This comparison is also found in Martin (1966), #22.
93
Omodaka et al. gloss this word as ‘rudder’ (JDB 1967: 170), but the textual basis for this
is unclear to me. Yaswo ka in MYS XX: 4408 is much more likely to be ‘eighty oars’, rather
than ‘eighty rudders’. Takagi et al. also read ka as ‘oar’ in this poem (1962: 441).
130———Koreo-Japonica

possible that OJ ka is a truncation of OJ kai, it is also likely that WOJ ka is

the original form, since it appears in a compound, and compounds tend to
preserve older forms. A possible explanation for the kayi is that it
represents an obscure compound. Whether the original form was ka or kayi,
the comparison with Korean is impossible. Not to mention the fact that
branches cannot be used as oars ― you simply would not be able to move
your boat! Therefore, I reject this etymology.
(101) (L) MK kàlàp, 94 LMK kal 95 ‘oak’ ~ OJ kasi ‘id’. < PJK *kal2
(Whitman 1985: 220). 96 This might look like a perfect comparison, but
WOJ kasi and its cognates are attested only in Central Japanese. 97 Thus, by
the distribution criteria, it must be a loan from Korean into Central
Japanese.
(102) (L) MK kál (probably kálh, cf. MdK khal) ‘pillory’ ~ OJ kasi, MJ
kase ‘id’. (cf. 104 ‘oak’) < PJK *kal2o (+ -i). Whitman comments that the
word is a likely loan into Japanese (1985: 220). I agree with him
completely, but two corrections are in order. First, it is necessary to correct
the semantics: ‘pillory’ seems to be a European invention, and this
meaning is attested only for Modern Korean. The word in question in both
Middle Korean and pre-modern Japanese actually means ‘cangue’, a
wooden board with holes for the head and/or hands and feet. The
instrument itself was apparently borrowed into Japan from the mainland,
since binding prisoners with ropes was the more usual practice, and the use
of the cangue was restricted. Second, there are no phonetic attestations of
this word in Old Japanese (JDB 1967: 186). We can hypothesize that it
was there only on the basis of later glosses of Nara period texts written
entirely in Chinese. Phonetically, the word is attested for the first time in
Middle Japanese as kasi. As far as I can tell, the form kase given as Middle
Japanese by Whitman actually appears for the first time only in Early
Modern Japanese. It is likely that (101) above and this etymology either
represent parallel loans from Korean into Central Japanese or two cases of
parallel semantic development after the word for ‘oak’ was borrowed: MJ
kasi 2.3 ‘cangue’ is called so, because it was made from kasi 2.3 ‘oak’,
known for the hardness of its wood, and MK kál ‘cangue’ was also made

94
Appears only within the compound kalap-namwo or kalap-namk- ‘oak-tree’ (Nam 1997:
5; LCT 1987: 19).
95
Whitman’s Late Middle Korean is certainly the Early Modern Korean form.
96
This comparison is also found in Martin (1966), #156.
97
There is a Shuri word kaasya ‘broad tree leaf used for food wrapping’ (RGJ 1976: 298),
which probably represents a loan from MJ kasifa ‘oak tree leaf’ (used for similar purposes),
since it appears only in the Shuri dialect. It also has an irregular accent B (Martin 1987:
441); we would expect A for a true cognate (OJ kasipa is 3.2a). As far as I can tell, the
practice of using oak tree leaves as cups or wrapping for food is textually supported only
for the Asuka and Nara periods. A direct loan from Western Old Japanese into Shuri is
impossible, but the absence of the word kasi ‘oak’ anywhere in Ryukyuan, as well as the
isolated nature of the word in question in Shuri and its irregular accent, still strongly
suggest that it is a loan.
 Lexical Comparisons———131

from kal 98 ‘oak’. In any case, the word for ‘cangue’ allows us to date the
correspondence of MK /l/ to OJ /s/ to no later than the beginning of the
first millennium. This may have far-reaching consequences for
etymologies like ‘star’ (30), that are traditionally accepted as evidence for
a genetic relationship.
(103) (R) MK :kal- ‘plows it, cultivates it’ ~ OJ karasukî LLLL ‘plow’
< kara (?) + sukî ‘plow’ < PJK *kara-. Whitman comments: “The first
morpheme in the Old Japanese form is often identified as kara- ‘foreign’
(Korean, Chinese), but there is no clear support for this identification”
(1985: 220). There are two problems with Whitman’s etymology. First, the
Old Japanese verb *kar- ‘to plow’ does not present itself, so the proposed
identification is speculative and exclusively external; it is based on Korean,
and therefore involves circular logic. Second, I am unaware of the
existence of any tatpurusha-type compounds in Japonic in which a plain
verbal stem can modify a following noun. Finally, there is wonderful
archeological evidence supporting the traditional interpretation: U-shaped
karasukî 99 have been excavated only from the cites dating from around the
seventh century A.D. (Takayanagi and Takeuchi 1974: 218), so there is no
way they can go back to the period of ‘Proto-Koreo-Japonic’. The
traditional identification of kara- in OJ karasukî ‘plow’ as ‘Korean’ or
‘Chinese’ is certainly correct, and I reject this etymology.
(104) (R) MK kàlàp, LMK kal ‘oak’ ~ OJ kaya, kape ‘torreya’ < PJK
*kara (? + p < niph) ‘leaf’ (Whitman 1985: 220). Certainly, MK kàlàp,
LMK kal ‘oak’ cannot be a cognate for both OJ kasi ‘oak’ (see [101] above)
and OJ kaya, kape Torreya nucifera, but since I demonstrated that OJ kasi
‘oak’ is likely a loan from Korean, let us review this etymology as well.
First of all, OJ kaya and kape certainly cannot be etymologically related. It
is not quite clear what kind of tree OJ kape is (JDB 1967: 213), so it is
better to exclude it, especially considering that tying it to MK kàlàp will
certainly involve Whitman’s *-r- loss law, which cannot be proved
internally for this particular case. OJ kaya is indeed Torreya nucifera, a
kind of yew tree, but oaks and yews belong to two completely different
species of trees. Finally, MK -l- does not correspond to OJ -y-, as shown in
(17), (34), and (40) above. Therefore, I reject this etymology.
(105) (R) MK :kalkí ‘mane’ ~ OJ ka-/kë ‘hair’ < PJK *kar (+ -ki
[diminutive]) (Whitman 1985: 220). I am unaware of the existence of any
diminutive -ki in Korean. Since MK :kalkí < PK *kàlókí, it is possible that
-i is a suffix here, cf. MdK [meli-]khal ‘hair on the head’, likely to go back
to PK *kalok. OJ ka- ‘hair’ fits only with the first syllable of the Korean
word, and it is not clear how we can account for the second. This
etymology has to be rejected.
(106) (R) MK :kalm- ‘hides it, puts it away’ ~ OJ kömï- ‘id’. < PJK
*kerömö-. Whitman comments: “The vocalism in this comparison is

98
Unfortunately, there is no information on accent, since the word appears for the first time
only in the Mwulpwo, a Korean text that was compiled in the late eighteenth or early
nineteenth century.
99
For the picture see JDB 1967: 230.
132———Koreo-Japonica

troublesome, since MK and K[orean] do not evince an alternant in /e/”

(1985: 220). It certainly is, and I also have to add that OJ kömï- (attested
only once, in KK 1), as well as its much more frequent variant kömë-, does
not provide any internal evidence for *-r- loss. With these two
observations we can safely put this etymology to rest.
(107) (L) MK kálmwó ‘axle, axle guard’ 100 ~ OJ karimo ‘axle guard’ <
PJK *karimo. Whitman also remarks that it is likely a loan (1985: 220),
and I agree with him.
(108) (R) MK kám-, MdK ka:m- ‘winds it, coils it’ ~ OJ karamë- ‘binds
it’ < PJK *karamU-. Whitman remarks that “the long vowel in [Md]K
ka:m- indicates the original long vowel necessary for this comparison”
(1985: 220). I believe that it is extremely hazardous to posit a long vowel
in Proto-Korean on the basis of a single reflex with a long vowel in
Modern Korean, not to mention a ‘reconstruction’ of a much needed PK
*-l- for this comparison on the same basis. There is no regular
correspondence of MK -Ø- to OJ -r-, and this, with the less than perfect
semantics allows me to reject this etymology.
(109) (L) MK kàmá ‘cooking pot’ ~ MJ kama HH ‘id’., OJ kama- in
kamaNtô ‘cooking place, hearth’ < kama + tô ‘place’ < PJK *kama.
Whitman admits that this etymology may represent a loan (1985: 220).
There are no attestations of the word in Eastern Old Japanese. The
attestations in Ryukyuan suggest a loanword scenario: the reflexes of kama
are found throughout the Northern and Central Ryukyus, but they are very
sparse in Sakishima. As far as I can tell, only Higashi Nakasone on Miyako
Island has ukama ‘cooking pot’, and there are two attestations of kamado
meaning not ‘hearth’ but ‘cooking pot’: Tonoshiro kamadu and Hateruma
kamdu (Uchima and Arakaki 2000: 384). These sparse attestations in the
South Ryukyus and the semantic shift ‘hearth’ => ‘cooking pot’, with
kama not attested separately strongly suggest that the loanword scenario is
correct, and that the Ryukyuan words are loans from Japanese. This leaves
us only with Central Japanese kama, which also, by my distribution criteria,
makes it likely to be a loanword from Korean.
(110) (L) MK kàph- ‘returns it’ ~ OJ kapë- ‘exchanges it’ < PJK *kapo-.
I see little textual basis for glossing MK kàph- as ‘to return’, since in
Middle Korean texts the verb certainly does not mean ‘to return’ in general,
but rather means ‘to pay [back], to reward’:

ZYELOY-s UNHYEY-lol kap-soW-a-za ho-l-i (i)-la

Tathagata-GEN kindness-ACC repay-HUM-INF-GER do-ATTR/IRR-
NML (be)-FIN
[we] have to repay the kindness of the Tathagata (Sekpo XXIII: 21b)

100
Yu Changton glosses MK kálmwó as swuley-uy kwultay ‘axle of a vehicle’ (LCT 1987:
29), but in Hwungmwong cahwoy it glosses the Chinese character ‘axle guard’
(Hwungmwong II: 26b).
 Lexical Comparisons———133

na-y ne-ykey pit-ul kaph-wo-m-i

I-NOM thou-DAT debt-ACC pay-MOD-NML-NOM
the fact that I repay [my] debt to you (Nung VI: 91b)

The difference in meaning is a relatively minor point, but I think that it

actually strengthens the comparison on its semantic side. At first glance the
comparison seems to be impeccable, but there is one problem. MK kàph-
can go back either to PK *kakop- or to PK *kapok-. The former case rules
out a genetic relationship between the Korean and the Old Japanese forms
from the start, but even in the second case we would have some difficulty
explaining the correspondences of PJ *kapay- to PK *kapok-. Namely, PK
*ok does not neatly correspond to PJ *ay, even if one postulates a
speculative PJ *a-Ci in PJ *kapa-Ci- following Martin (1987: 701). In this
scheme PJ *-C- belongs to the suffix, and PK *k still belongs to the root,
so there is definitely a segmentation problem. It is possible to speculate
that the root of the Proto-Japonic verb also once ended in *-k, but this is
impossible to prove on the basis of the internal evidence. The cognates of
OJ kapë-, as well as of kap- ‘to buy’, which are also likely to be related
(both belonging to accent class A [Martin 1987: 701, 706]), do not appear
in Eastern Old Japanese texts, but they are attested throughout the Ryukyus.
This would indicate the possibility of a genetic relationship, but problems
with the correspondences make me think that a loanword scenario
represents an easier explanation, since it does not require any speculative
hypotheses. If it is a loan, it must be an old one, going back to the period of
linguistic coexistence on the Korean peninsula.
(111) (R) MK kàphól ‘case, sheath‘, MdK kkaphwul ‘skin’ ~ OJ kapa
LL ‘skin’ < PJK *kapo (+ -ol diminutive). Whitman adds: “The original
K[orean] root is recoverable from [Md]K kkaptayki ‘shell, skin’ < kaph +
tayki ‘thingy’. Considerable etymological confusion with SK kap ‘armor’
has also occurred here” (1985: 221). At the present stage of our knowledge,
the proposed Proto-Japanese-Korean reconstruction does not account for
the aspirated /ph/ in both Middle and Modern Korean. I am not sure that
SK kap ‘armor’ could cause any etymological confusion, because MK
kàphól ‘scabbard, sheath’ (which, incidentally, does not mean ‘case’ in
general) is not exactly armor. Armor protects the body of a warrior, while
scabbards do not. It is interesting that MK kàphól means only ‘scabbard,
sheath’, while MdK kkaphwul means ‘skin’. I must also mention that MdK
kkaphwul ‘skin’ also means ‘film, coat, skim‘, as in, e.g., nwun-kkaphwul
‘eyelid’, so its archetype meaning seems to be essentially *‘cover’. Sheaths
in medieval Korea were made from leather and/or wood, but MdK
kkaphwul certainly is never used in the sense of ‘leather’. The semantic
development from ‘sheath’ => ‘cover, skin’ and not vice versa is bizarre. I
wonder whether etymological contamination indeed took place, but under
a different scenario from Whitman’s. We must keep in mind that MdK
kkaptayki ‘shell, skin’ also has an alternative form kkepteyki, which is
obviously etymologically connected with MK kèpcìl and MdK kkepcil
‘skin, bark, husk, etc.’. MK kèpcìl and MdK kkepcil are semantically much
134———Koreo-Japonica

closer to OJ kapa, but cannot be its cognates due to the lack of a regular
correspondence between MK /e/ and OJ /a/. Note that there is no MK
*kapcil or MdK *kkapcil, so it is likely that MdK kkaptayki ‘shell, skin’ is
just a secondary development from MdK kkepteyki, being the heavy
isotope of the latter. Thus, it may not be useful in recovering the original
root for MK kàphól ‘sheath’ and MdK kkaphwul ‘skin’. I believe that MK
kàphól ‘sheath’ may have an unrelated internal etymology which has
nothing to do with ‘skin’. Crosslinguistically an etymology for the word
‘sheath’ is very frequently derived from the word for ‘sword’ or ‘knife’,
and Modern Korean is no exception: MdK khal-cip ‘sheath’ < khal ‘knife,
sword’ + cip ‘house’. 101 I believe that MK kàphól may be a very similar
derivation: MK kalh ‘knife, sword’ + -pol. We could expect the loss of -l-
in this position, because the Middle Korean cluster -lph- is extremely rare.
The final -h of MK kalh then produced aspiration on the first consonant of
the second component of the compound. This last component may be
difficult to explain, but I think that a predecessor of EMdK and MdK pwul
‘scrotum’ may be involved here. The form -pol may contain a reduced
minimal vowel /o/, which changed from a [-RTR] vowel to a [+RTR]
vowel under the influence of /a/ in the first syllable of the compound.
Using the word ‘scrotum’ for a knife’s cover may seem ludicrous at first
glance, but we should not forget that knives and swords are often used as
phallic symbols. When the compound became obscure (no doubt under the
influence of Confucianism), it could be just perceived as a ‘cover for a
knife, sheath’, one more step toward becoming a *‘cover’ in general,
which, as we have seen above, is the semantic archetype of MdK kkaphwul.
I must also add that ‘sheath’ in English seems to have undergone the same
semantic evolution. Therefore, I reject this etymology due to the more
realistic internal explanation for the Middle Korean word.
(112) (R) MK kàpóyép- 102 ‘to be light’ ~ OJ karu- (+ -si) ‘id’.
(Whitman 1985: 221). No Proto-Japanese-Korean form is provided, and I
doubt that these two forms can be reconciled phonetically, with the
exception of the first syllable, /ka/, in both languages. Therefore, this
etymology is also rejected.
(113) (L) MK káps ‘value, worth, price’ ~ OJ kapi ‘id’. < PJK *kap-
(Whitman 1985: 221). This is the same root as in (110), so I disregard it for
statistical purposes.
(114) (L) MK kát ‘hat’ ~ OJ kasa ‘sunshade, hat’ < PJK *kasa
(Whitman 1985: 221). The correspondence of MK /t/ to OJ /s/ is not
regular, but there is a later Middle Korean variant kás attested from 1586
in the first edition of Swohak enhay. All other attestations of kas belong to
Early Modern Korean and Modern Korean (Nam 1997: 35). Under normal
conditions, the preference should be given to earlier attestations, but since

101
Cf. also Russian nozhny ‘scabbard’ < nozh ‘’knife’.
102
I presume this is a typographical error for MK kàpóyyáW- (LCT 1987: 21; Nam 1997:
10). There are also the Early Modern Korean forms kapwoyyeW- and kapuyyep- (Nam 1997:
9), but they are attested much later. The word should also be presented with final -W, since
the last consonant of the stem is a leniting /p/.
 Lexical Comparisons———135

the phonological history of dentals, and especially final -t in Korean is very

complex, I tentatively allow a later attestation to be taken into
consideration. OJ kasa ‘sunshade, hat’ is not attested in Eastern Old
Japanese, and the Ryukyuan attestations are mostly concentrated in the
Okinawa region: Yamatoma hasaa (Hirayama 1986: 175), Nakijin hasaa
(Nakasone 1983: 382), Iejima hasaa (Oshio 1999: 301), Shuri kasa (RGJ
1976: 311). The single attestation in Sakishima is Psara -kasa, which
occurs only in the compounds ama-kasa ‘umbrella’ and kuba-kasa ‘hat
made from palm leaves’ (Shimoji 1979: 51). A different word sana ‘hat,
sunshade’ is much more typical for Sakishima and is attested in Ishigaki,
Taketomi, and Hateruma (Miyara 1980: 299). This kind of distribution in
the Ryukyus suggests that hasaa/kasa is a loan from Japanese. Therefore,
Central Japanese kasa itself is likely to be a loan from Korean.
(115) (R) MK kól ‘reed, rush’ ~ OJ kaya LH ‘general term for long
grasses, thatch’ < PJK *ko:ra (Whitman 1985: 221). The correspondence
of MK -l- to OJ -y- is irregular; therefore, I reject this etymology.
(116) (R) MK kòláp-103 ‘itchy’ ~ OJ kayu- ‘id’. < PJK *koryo-.
Whitman provides the following commentary: “Final /p/ in MK is the
adjectival suffix. LMK [i.e., Early Modern Korean — A.V.] also attests
kolyap- and kolyep- (the latter the K form of the adjective 104). MK in
general does not attest *yo, but there is strong evidence (including a direct
reference in the Hwunmin cengum enhay 1446, where it is reported for
‘dialects and children’s speech’) that the sequence existed at a slightly
prior stage of the language. LMK /le/ is sometimes the outcome for MK
/lo/; since the a/e alternation is otherwise extremely exceptional in just the
second syllable of a stem, the LMK alternation points to pre-MK *kolyop-,
which provides a perfect fit for OJ karu-” (Whitman 1985: 221). I believe
that the reconstruction of PK *yo in the second syllable of MK kòláp-
solely on the basis of Early Modern Korean forms that are famous for their
inconsistent and messy spelling is potentially hazardous, but in this case I
think Whitman is right: his reconstruction can be further supported by the
Ceycwuto forms kolop-, kolyop-, and kolyep- (Ceycwu pangen yenkwuhoy
1995: 71). Nevertheless, there are two problems with this etymology. First,
the correspondence of MK -l- to OJ -y- is irregular; see also (115) above.
That alone would warrant the rejection of this etymology. The
reconstruction of PK *-ly- here hardly helps, because the correspondence
of PK *-ly- to OJ -y- is not supported by other examples. Second, there is
no direct internal evidence that final /p/ in Middle Korean is an adjectival
suffix, because a corresponding verb or noun without it does not occur. At
present I prefer to reject this etymology, but it could be saved if more
examples for PK *-ly- corresponding to OJ -y- were presented along with
internal evidence for segmentation of an ‘adjectival suffix -p’ in Korean.
(117) (R) MK kólàs ‘foxtail’ ~ OJ karasi LHL ‘mustard plant, rape’ <
PJK *ko:ras (+ -i) (Whitman 1985: 221). I do not know Whitman’s source
103
The word should also be presented with final -W, since the last consonant of the stem is
a leniting /p/.
104
The Modern Korean form is actually kalyep-, not kolyep-.
136———Koreo-Japonica

for the meaning ‘rape’ for OJ karasi; to the best of my knowledge it means
only ‘mustard plant, mustard grain, product of mustard plant’. I see no
reason to reject the traditional etymology of the Old Japanese word, which
treats it as a lexicalized form of the conclusive form kara-si 105 of kara- ‘be
spicy, salty’ (JDB 1967: 229). The accent classes of the adjective kara- (B)
and of the noun karasi (3.7) are congruent, because both are low-register
classes. This, alongside the fact that the mustard plant was probably
imported from China, further strengthens the traditional etymology. Finally,
foxtails and mustards belong to two different kinds of grassy plants. I have
no comment on the taste of a foxtail, because it happens to be on the menu
of cattle, not of humans. With this additional comment on its semantics, I
reject this etymology.
(118) (R) MK kóláy ‘wild walnut’ ~ OJ kuru-/kuri LL ‘chestnut’, MJ
kurumi ‘walnut’ < PJK *koro (+ -i). Whitman remarks: “The final vowel in
MK is problematic: the expected form would be *kólóy. However, the
transcription of /oy/ ~ /ay/ is fairly variable in MK. MJ kurumi is a
compound of kuru + mi (OJ mï) ‘fruit, nut’, suggesting the existence of an
original root pre-OJ *kuru- < PJK *koro (+ -i) ‘nut tree’” (1985: 221).
There are several problems here. First, the claim for the transcriptional
variation /oy/ ~ /ay/ in Middle Korean is not substantiated. There is the
Early Modern Korean form koloy attested in sico (Nam 1997: 226), but it
cannot be used as evidence, because MK /o/ had been lost as a phoneme by
that time, and the form koloy probably just represents a written
hypercorrection. I think that MK /oy/ and /ay/ are definitely not
interchangeable within Middle Korean itself; cf. MK :say ‘bird’ and sáy
‘new’, on the one hand, and MK sóy ‘east’, on the other (Nam 1997: 813,
950). Only in Early Modern Korean can one see the hypercorrected
spelling soy ‘bird’ (Nam 1997: 950). Thus, MK kóláy ‘wild walnut’ does
not correspond regularly to the Old Japanese form, and that alone allows
for the rejection of this etymology. Second, while Whitman’s etymology of
MJ kurumi is certainly attractive, without an attestation in Old Japanese it
is impossible to verify whether -mi in MJ kurumi is indeed from OJ mï
‘fruit’, with the otsu-rui vowel /ï/. 106 MJ kurumi can be just a lexicalized
nominalization in -i from kurum- 107 ‘to wrap it up’. In addition, walnuts
and chestnuts belong to two different tree species, so the resulting
archetype may be only ‘a kind of nut’, which is always suspicious.
Therefore, on the basis of all these considerations, I reject this etymology.
(119) (R) MK kòlí- ‘hides it, conceals it, obscures it’ ~ OJ kë-, kiye-
‘disappears’ < PJK *kor-. Whitman comments: “The identity of the first
syllable vowel in OJ kiye- is unclear; this comparison predicts kuye-. The
related transitive form is kët- ‘extinguishes it, makes it disappear’” (1985:
105
See Martin (1987: 806-809) for the lack of a clear demarcation between the attributive
and conclusive forms of adjectives in Old Japanese, as well as for examples of the usage of
the ‘conclusive’ form in the attributive function.
106
Both kuri and kurumi are low-register words, a fact that may give additional support to
Whitman’s etymology.
107
The accentuation of MdJ kurum- is unclear (Martin 1987: 717), and it is attested only
from Early Modern Japanese, which may invalidate this alternative etymology.
 Lexical Comparisons———137

221). As Whitman himself points out, the correspondence of vowels in the

first syllable is irregular. Besides, the loss of *-r- in this word in Old
Japanese cannot be confirmed by internal Japonic evidence. In addition,
the morphophonological history of these two Old Japanese words, if they
are to be reconciled as one etymon, is rather complex. We do not really
know whether OJ ket- was really kët- or kêt-: the word is not attested
phonetically in Old Japanese, and the reasons Omodaka et al. list it as kët-
(JDB 1967: 281) are unclear to me. OJ -t- as a transitivizing suffix in ket-
cannot be solidly supported: there is Old Japanese transitive panat- ‘to
release, to let go’, correlating with intransitive panare- (JDB 1967: 586-
587), but that is the only other case I can think of, and the analogy is not
perfect. 108 The form panare- may suggest that OJ kë- was once *këre-, but
in that case the much needed *-r- turns out to be in the wrong place for a
comparison with Korean. The reconciliation of first syllable vowels in OJ
kë- and kiye- is only possible on the assumption that /i/ < *ï < *öi, because
the only common source for OJ /ï/ and /ë/ is *öi. However, that also creates
an irregular vowel correspondence with Korean, because in that case MK
*keli- would be needed. Finally, I think that there is rather solid evidence
for OJ *kîye-, because the Proto-Ryukyuan form is *kiyasi- ‘to extinguish’
(Thorpe 1983: 283), with PR *-i- pointing to PJ *-i- (PR *-e- would be
needed to support OJ /ï/). Given one irregular correspondence and another
correspondence that is based on a tentative *-r- loss, the only commonality
that remains is the initial consonant. The situation is exacerbated the by
lack of a precise Proto-Japonic reconstruction, and I believe that at present
the etymology should be abandoned.
(120) (R) MK kòlk-/kúlk- ‘scratches it, scrapes it’ ~ OJ kak- ‘id’. 109 <
PJK *korak. Whitman claims that MK kòlàk ‘finger, stick’ represents the
uncontracted nominal form of this root (1985: 221). I think that a
connection between ‘finger’ or ‘stick’ and ‘scratching’ is quite speculative:
one usually scratches with a nail, not a finger. Fingers and sticks may be
used for other actions, like poking or pointing. Even if Whitman is right
that the uncontracted form was PK *korak-, that leaves us with problems
of regularity in the correspondences already discussed in (4) and (56). If
the nominal is to be excluded, we face another irregularity in the
correspondences, also addressed above in (16). Thus, I reject this
etymology.
(121) (R) MK kòlò, kòl, kòll- 110 < *PK kòlòk ‘powder’ ~ OJ kô ‘id.’,
MdJ kona ‘id.’ Whitman comments: “The otherwise mysterious relation
between OJ kô, later J -ko and [Md]J kona is explained by this comparison.
OJ kô is derived from PJK *korok by loss of final *k followed by medial
*-r- loss” (1985: 222). In my opinion it is hazardous methodologically to
explain the internal relationship between OJ kô and MdJ kona on the basis

108
MJ kes- and EMdJ fanas- also appear later, with OJ ket- being completely supplanted by
kes- in the modern language, while both MdJ hanas- and hanat- still coexist.
109
This comparison is also found in Martin (1966), #189.
110
Rather kòlG-, because kòll- is attested only once in the form kòll-í (Welin I: 29a), and all
other attestations point to kòlG-.
138———Koreo-Japonica

of an external comparison. -na in MdJ ko-na certainly has an internal

explanation as an obsolete plural/collective marker (Vovin 1994: 249, 253;
2005a: 102-107); see also (16) above. The comparison is based on the
assumption of medial *-r- loss in Japonic, which cannot be internally
verified for this word. In addition, it is based on another assumption, that
the vowel /ô/ in OJ kô evolved from PJ *ua, which is only one of three
possibilities, the other two being *au and primary PJ *o. Again, we have
no internal Japonic evidence to confirm that PJ *ua underlies OJ /ô/.
Furthermore, there is an additional problem. I will accept for a moment
that OJ kô < PJ *kua < pre-PJ *kura, but then the vowels in PJK *kòlòk
and pre-PJ *kura do not correspond regularly, because according to
Whitman’s set of vocalic correspondences OJ /a/ < PJ *a corresponds to
MK /o/ only in long syllables; otherwise (as in this case) the expected Old
Japanese correspondence is /u/. Ceycwuto cognates of MK kòlò are kulwu,
kulu, kulul, kolwu, and kolu (Choy 1978: 471-472) essentially confirming
the Proto-Korean reconstruction *kòlòk and excluding the possibility of
any long vowel in the second syllable: both vowels in Proto-Korean appear
to be minimal. Thus, I reject this etymology on the basis of its irregularity.
(122) (R) MK kólp ‘matches (with), pits against, competes’ ~ OJ
kuraNpë- ‘id’. 111 < PJK *kora:b- (Whitman 1985: 222). I believe that
Whitman marked the accent of the Middle Korean incorrectly following
LCT 1987: 12, where it is mistakenly cited as kólp- with H pitch. The form
is actually MK :kolW- with R pitch, as the textual evidence makes clear, cf.
e.g., Pep II: 17b. Textual attestations further confirm that the last
consonant of the stem was MK -W-, not -p- (LCT 1987: 12; Nam 1997:
240). That allows us to reconstruct PK *kòlóp-, with a leniting *-p-. OJ
kuraNpë-, on the other hand, goes back to PJ *kuranpai-, with a cluster
*-np-, which corresponds regularly to Middle Korean non-leniting -p- <
*-np-, but not to leniting -p- < *-p-. This irregularity allows me to reject
the comparison. In addition, the semantics of the comparison are
questionable, since the predominant meaning of MK :kolW- seems to be
‘matches with, stands side by side’ (LCT 1987: 12; Nam 1997: 240),
although in some Middle Korean texts the meaning ‘to be hostile’ is
attested as well, e.g., Hwungmwong III: 27b. The basic meaning of OJ
kuraNpë- is just ‘to compete’ (JDB 1967: 274).
(123) (R) MK kóm- ‘shuts it, closes (eyes), joins it’ ~ OJ kum- ‘joins it’
< PJK *kom-. Whitman adds: “[Md]K ka:m- < MK kóm- has become
specialized in the meaning of ‘close the eyes’, but Yongpi echenka (1447,
IV: 118) has the example hota-ka spolli kom-key ho-myen ‘then quickly
make them meet’” (1985: 222). There is a serious problem with this
comparison. First of all, the phrase cited by Whitman above does not
appear in Yongpi 118. It is found in Nung IV: 118b, and I give the whole
context below:

111
This comparison is also found in Martin (1966), #50.
 Lexical Comparisons———139

tye SEY yey-s salom-i nwun-ey pwo-m-ol mwoyhwo-taka ho-taka spolli

kom-key ho-myen
that world old-GEN person-NOM eye-LOC see-NML-ACC concentrate-
TRANSF do-TRANSF quickly close-CAUS do-COND
if an old person from that world concentrates on what [he] sees in [his]
eyes and then quickly makes [them] close (Nung IV: 118a-118b)

It is quite clear that in this phrase MK kóm- refers also to closing the
eyes. Just by browsing dictionaries of Middle Korean, one can find another
example in another volume of the same text, stwo kom-wo-m-ay ‘when
again joins’, which, judged solely by its Chinese translational equivalent
且合 provided in dictionaries (LCT 1987: 12; Nam 1997: 241), may
indicate that MK kóm- may have other meanings besides ‘to close the
eyes’. However, this is an illusion that shatters if we look at the text itself:
nwun-i NUNG-hi twuluhhye pwo-nwos-ta stwo kom-wo-m-ay
eye-NOM able-ADV turn around(INF) look-EXCL-TRANSF again close-
MOD-NML-LOC
when eyes ably looked around and again closed (Nung I: 61a)

All other examples listed in LCT 1987: 12-13 and Nam 1997: 240-41
also involve MK kóm- used only in reference to closing the eyes. Thus, we
can come to the inescapable conclusion that MdK ka:m- did not narrow its
meaning to ‘to close the eyes’, as Whitman suggests, but continues to exist
in the same and only meaning as it exhibited in Middle Korean. OJ kum- is
a hapax legomenon, attested in the compound kum-i-ne- ‘to sleep
embracing [each other]’ (KK 91), and it appears to mean ‘embrace,
entwine‘. Middle Japanese attestations from the tenth and eleventh
centuries are not numerous, either, and they mostly involve the
nominalized form kumi ‘cord made from threads plaited together’. The
verb itself appears in the meaning ‘to plait, to weave’ (KKJ 1969: 304;
Miyajima 1971: 105). As far as its semantic development, I could trace the
meaning ‘to join’ in the sense ‘to put together’ only in the very late Middle
Japanese texts. The few available Ryukyuan attestations also point to the
primary meaning ‘to entwine, to plait’: Nakijin khumiN (Nakasone 1983:
137), Iejima kunyuN (Oshio 1999: 112), Shuri kunuN (RGJ 1976: 334). It
is interesting that Yaeyama fumuN, which is a genetic cognate of Japanese
kum-, means both ‘to weave, to plait’ and ‘to be accepted within a group’
(Miyara 1980: 531), while Yaeyama kumuN, which is a loan from Japanese,
means only ‘to join’ (Miyara 1980: 283). Historically the semantics of
Korean and Japanese words appear to be very different, and, therefore, I
reject this comparison.
(124) (R) MK kóm- ‘bathes’ ~ OJ kum- H ‘draws water’ < PJK *kom-
(Whitman 1985: 222). I believe that ‘bathing’ and ‘drawing water’
represent two quite different activities. When ‘bathing’, you either soak in
water or pour it over yourself, while ‘drawing’ represents the opposite
140———Koreo-Japonica

process. Since the semantics are difficult to reconcile, I reject this

etymology.
(125) (R) MK kònòlh/kùnùl ‘shade, shadow’ 112 ~ OJ kaNka-/kaNkë LF
‘id’. < PJK *ko:golo (+ -i). Whitman also adds OJ kaNkayak- ‘shines,
reflects’ (1985: 222), which is completely unrelated; it is in fact OJ
kakayak- (JDB 1967: 174-175), which acquired secondary voicing
(kagayak-) only in Early Modern Japanese (IKJ 1990: 272). The
correspondence of MK -n- to OJ -Nk- is surely irregular, and, as a matter
of fact, there is another non-conditioned correspondence of OJ -Nk- to MK
-l-; see (4) above. In addition, Whitman also reconstructs PJK *-g- for MK
-Ø- corresponding to both OJ -k- and -Nk-; see (62) above and (145) below.
This etymology can be rejected on the basis of its irregularities.
(126) (L) MK :kos ‘edge, border, brink’ ~ OJ kata HL ‘shore, side’ <
PJK *koco or *kaco (Whitman 1985: 222). This etymology involves an
irregular correspondence of MK -s- to OJ -t-, which is not listed by
Whitman among his correspondences for coronal obstruents (1985: 163-
167). However, this correspondence may be one of those correspondences
typical of Korean loans in Japanese; see the discussion of MK -s ~ OJ -tu
in the section on morphology. I believe that two different Old Japanese
words are involved here: (a) OJ kata ‘direction’, rather than ‘side’, see
JDB 1967: 190; and (b) OJ kata ‘part of the shore line covered by high
tide’ (JDB 1967: 190). The words are homonyms: both belong to accent
class 2.2 (Martin 1987: 442), but their semantics are difficult to reconcile.
Only the second can be compared to Korean without stretching its
semantics beyond credibility. There is also EOJ kata ‘lagoon’, attested
once as an independent word (MYS XIV: 3551), and once as a place name
(MYS XIV: 3549. The only Ryukyuan attestation is in the Shuri compound
kata-baru ‘part of the shoreline covered by high tide’ (RGJ 1976: 313),
which is likely to be a loan from Japanese due to its isolated nature. Thus,
unless we are dealing here with a chance resemblance, the irregular
correspondence and distribution pattern suggest that the Japanese word
may be another loanword from Korean.
(127) (R) MK -kwo (verbal gerundive suffix) ~ OJ -ku (adjectival
gerundive suffix) < PJK *ku (Whitman 1985: 222). Discussed in 2.3.2.12
above.
(128) (R) MK kòzòlàkí ‘awns and bits of rice or barley husks’ ~ OJ
kusô ‘ordure, trash, chaff’ < PJK *kosora. Whitman adds: “MK -ki is a
diminutive suffix. MK kòzòlh ‘autumn, harvest’ is very likely cognate”
(1985: 222). MK kòzòlh ‘autumn’ certainly cannot be a cognate, because
OK 秋察 /KOcal/ ‘autumn’ (Hyangka XI: 5) clearly indicates that MK -z-
in this word goes back to PK *-c-, and not *-s-, and the former does not
correspond regularly to PJ *-s-. There are two other problems with this
etymology. First, the Korean diminutive is really -i, not -ki. In addition, a
word without this ‘diminutive’ does not present itself, so there is no
positive internal evidence that the segmentation proposed by Whitman is

112
The actual accentuation is kónólh ~ kúnúlh, with H, not L, pitch.
 Lexical Comparisons———141

correct. Second, the primary and basic meaning of OJ kusô is simply

‘excrement’, and that is too far from ‘awns and bits of rice or barley husks’.
It appears as ‘chaff, trash’ only in some Old Japanese compounds, but
there are also other compounds, like OJ kusô-kaNtura ‘stinky vine’ (JDB
1967: 260) where certainly only ‘excrement’ can be involved. There are
many crosslinguistic examples with ‘excrement’ used periphrastically for
‘trash’, but I am not aware of the opposite development in any language.
Therefore, I reject the etymology on the basis of the problem with its
morphological segmentation and its unrealistic semantics.
(129) (L) MK kòzòm ‘stuff, material’ ~ OJ kusa LH ‘id’. < PJK *kosom
(Whitman 1985: 222). OJ kusa also means ‘kind, variety’ (JDB 1967: 255).
One would expect accent class 2.5 (LF) in Japanese to account for Middle
Korean final -m in a genuine cognate. However, the distribution of the
word strongly suggests that it is a loan from Korean into Western Old
Japanese, since it is not attested anywhere in Japonic except Western Old
Japanese.
(130) (R) MK kwóh ‘nose, protuberance’ ~ OJ kukî ‘peak, outcropping’
< PJK *kuki (Whitman 1985: 222). I am afraid that ‘protuberance’ is a
meaning added to improve the semantics of the comparison: the Middle
Korean word clearly means basically ‘nose’, and such semantic extensions
as ‘tip, end, protuberance’, etc. are clearly secondary. Unfortunately, it is
even worse that OJ kukî is cited with a completely wrong meaning. It
certainly means neither ‘peak’ not ‘outcropping’, but ‘narrow path between
two peaks or cliffs’ or ‘cave’ (JDB 1967: 253). I am afraid that no
persuasive connection with either ‘nose’ or even ‘nasal cavity’ can be
demonstrated. I would also like to add that it is very suspicious that the
words for body parts in Korean and Japanese match poorly, which is
usually not the case in the uncontroversial language families, cf. the word
for ‘nose’ in Indo-European, Turkic, Mongolic, or Finno-Ugric as well as
(57) above where Whitman compares ‘leg’ in Korean with ‘hand’ in
Japanese. I believe that such cases of loose semantic comparison are more
likely to hurt rather than help the Koreo-Japonic theory.
(131) (R) LMK 113 kwohay ‘stork’ ~ OJ kuNkuNpî, kuNpî ‘id’.
(Whitman 1985: 222). The Korean word is actually attested in the Middle
Korean of the early sixteenth century as kwòhày (Hwungmwong I: 15a). I
believe that glossing it as MdK hwangsay ‘stork’ (LCT 1987: 68; Nam
1997: 105) is a mistake, because in the Hwungmwong cahwoy the word is
used as a translation equivalent of the Chinese character 鵠 ‘swan‘, with a
further note that it is a 水鳥 ‘waterfowl’ (Hwunngmwong I: 15a). A swan
is an aquatic bird, but a stork is not. The Old Japanese form clearly appears
as kukupî, 114 with MJ kofu and kofi also attested (JDB 1967: 254), but the
form kuNpî is a hapax legomenon, and it appears only in the Kojiki, where
there is no absolutely consistent distinction between -p- and -Np-, so this
attestation may also render OJ kupî; see JDB 1967: 254 for details. Like
113
This is Early Modern Korean.
114
This actually improves the comparison, because MK -h- can correspond only to OJ -k-,
but not OJ -Nk-.
142———Koreo-Japonica

MK kwòhày ‘swan’ all of the Old Japanese forms also mean ‘swan’.
However, there are several problems with this etymology. First, the Middle
Japanese forms kofu and kofi suggest that OJ kukupî may be a partial
reduplication: ku-kupî. Second, even if it is not, the segmenting suffix -pî
in Japanese, which would be necessary for this comparison to survive, is
not viable on the basis of the internal Japonic evidence. Third, EMdK -a-
in the second syllable cannot regularly correspond to OJ -u-; OJ *kuka-
would be necessary according to Whitman’s own correspondences (1985:
129). 115 I believe there are too many problems to allow recognition of this
comparison as a valid etymology.
(132) (L) MK kwòkwòlí ‘stem, stalk’ ~ OJ kuku-/kukï ‘id’. < PJK
*kukur(+ -i) (Whitman 1985: 222). EOJ kuku- (in compounds) is also
attested twice (MYS XIV: 3406, 3444), but its distribution in Ryukyuan is
more indicative of a loan, since the word is attested only in the Northern
and Central Ryukyus: Yamatoma kuuchii, Shirana kuuki, Tokunoshima
kuki (Hirayama 1986: 245), Nakijin gucii (Nakasone 1983: 129), Iejima
gucii (Oshio 1999: 108), Shuri guci (RGJ 1976: 191). The Northern
Ryukyuan reflexes point to PR *kukuy rather than *kokoy, but only the
latter could be expected in the case of a genuine cognate. In addition, the
Middle Korean non-leniting -k- < PK *-nk- does not correspond regularly
to OJ -k-, as OJ -Nk- is needed. I believe that this etymology represents a
loan from Korean into Japanese, with a subsequent spread into Northern
and Central Ryukyuan.
(133) (R) MK :kwol ‘valley’ ~ OJ kura ‘id’. 116 < PJK *ku:r (Whitman
1985: 222). As far as I can tell, OJ kura ‘valley’ is a ghost. There is WOJ
kuratani, a hapax legomenon (MYS XVII: 3941) possibly meaning ‘deep
valley with soaring cliffs on both sides’, but its exact meaning is unclear.
OJ tani means ‘valley’ by itself. Assigning the meaning ‘valley’ to kura- in
this obscure compound goes back to Motoori Norinaga, but Omodaka et al.
point out that the basis for this solution is unclear (JDB 1967: 273). I agree
with them: kura- in WOJ kuratani could be anything, and cannot be used
as a basis for external comparison.
(134) (R) MK kwòlày ‘whale’ ~ OJ kuNtira LHL ‘id’. < PJK *kudori (+
-ra plural or diminutive) (Whitman 1985: 223). This example involves a
dubious correspondence of MK -l- to OJ -Nt-; see also (15) and (55) above,
where this correspondence is discussed and rejected. In addition, MK -ay
simply cannot correspond regularly to OJ -ira. Thus, I reject this
etymology on the basis of its irregularities.

115
There is also MK kwòhwày (pitch accent marks are tentative, as I do not have access to
this particular edition) in the edition of Sincung Yuhap done by the Ansimsa Temple (Nam
1997: 105), which might be looked upon as a possible solution to this particular problem.
But in addition to being a hapax legomenon, it also creates another problem: the remaining
MK -ay still does not correspond to anything in OJ kukupî.
116
This comparison is also found in Martin (1966), #248.
 Lexical Comparisons———143

(135) (C) MK kwó:lwop- 117 ‘troublesome, hard, painful’ ~ OJ kuru-si-

‘id’. < PJK *ku:ru- (Whitman 1985: 223). The earliest Korean attestations
are from the 1586 edition of the Swohak enhay. The etymology may seem
acceptable, but it largely depends on seeing MK -W- < PK *-p- as an
adjectival suffix. In addition, the adjectival suffix is not -p-, but -ap-/-ep-
(see [25] above). Therefore, the morphemic analysis is difficult to prove,
especially given the fact that the root *kwolwo- does not present itself in
Korean. It is only a marginally acceptable etymology.
(136) (R) MK kwòmá ‘lover, concubine‘; [Md]K koma ‘little one, child’
~ OJ kô H ‘child, lover’, Shuri kwaa ‘child’ 118 < PJK *kuma (Whitman
1985: 223). This etymology must be rejected, because, as Yi Kimun
demonstrated, MK kwòmá ‘concubine’ is a loan from Middle Mongolian
quma ‘concubine’ (1991: 246-247). For information on the spread of this
Mongolian word into Turkic and Persian, see Doerfer 1963: 414-415.
(137) (L) MK :kwomá ‘paying respect’ ~ OJ kuma ‘offering to [the]
gods’ < PJK *ku:ma (Whitman 1985: 223). There are two problems with
this etymology. First, MK :kwomá ‘paying respect’ should go back to PK
*kwòCómá due to the R pitch on the first syllable. Second, WOJ kuma
‘offering to [the] gods’ does not have any cognates in Eastern Old
Japanese and Ryukyuan. Both of these facts strongly suggest that it is a
loan from Korean into Western Old Japanese that took place after the
contraction in Korean had occurred.
(138) (L) MK :kwom ‘bear’ ~ OJ kuma LL ‘id’. < PJK *ku:mo
(Whitman 1985: 138). 119 The cognates of WOJ kuma ‘bear’ are not
attested in Eastern Old Japanese and the Southern Ryukyus, but this time
that fact cannot be used as evidence against a genetic relationship, because
bears are not endemic to the Central and Southern Ryukyus. However,
Paekche kuma (久麻) ‘bear’ shows the same raising of *o to *u as WOJ
kuma does, and it is further confirmed by Paekche mure ‘mountain’
(borrowed into Western Old Japanese as mure ‘mountain’), as compared
with MK mwòlwó and :mwoy ‘mountain’ (see [260] below). This leaves us
with two choices: either WOJ kuma is a loanword from Paekche, or, if one
wants to insist on a Koreo-Japonic genetic relationship, that both Paekche
and Western Old Japanese must have independently undergone the same
innovation. The loanword scenario presents a simpler and more elegant
solution.

117
Whitman gives MK kwólwúp-, but this is an obvious typographical error for kwólwóp-,
since *kwólwúp- is not attested (it also contradicts the rules of Middle Korean vowel
harmony). Furthermore, since this word exhibits a leniting -p-, it should be cited as MK
kwólwóW-. Finally, Nam Kwangwu shows the accentuation of this word as HL (Nam 1997:
96), Yu Changton as HH (LCT 1987: 63), and in the Wuli mal khun sacen it is given as HR
(Hankul hakhoy 1999.2: 4886). I do not have access to a facsimile of Swohak enhay (1586),
where the word is attested for the first time, so I cannot verify either of these accentuations,
but I suspect that Hankul hakhoy’s version is the correct one.
118
The Proto-Ryukyuan form is *kuwa ‘child’ (Thorpe 1983: 271), which would add
further problems to this comparison.
119
This comparison is also found in Martin (1966), #10.
144———Koreo-Japonica

(139) (L) MK :kwop- ‘pretty, beautiful’ ~ OJ kupasi ‘id’. < PJK *ku:pa
or *ku:po (Whitman 1985: 223). There is just one Eastern Old Japanese
attestation in the compound ka-N-kupasi- ‘having a beautiful fragrance’
(MYS XX: 4371), but there are no Ryukyuan cognates. Thus, I believe that
this is a loan from Korean into Japanese.
(140) (R) MK kwòp- ‘double, increase twofold’ ~ OJ kupapë- ‘add’ <
PJK *kup- (Whitman 1985: 223). There are two problems with this
etymology. First, MK kwòp- has a non-leniting -p-, therefore it goes back
to PK *kwònpó-. PK *-np- does not correspond to OJ -p-. Second, OJ -pë-
in kupapë- must be explained as a suffix in order for this etymology to
work, but I do not see any internal Japonic evidence to support such a
segmentation. Therefore, I reject this etymology.
(141) (L) MK kwóc ‘stake’ ~ OJ kusi LL ‘id’. < PJK *kuc (+ -i)
(Whitman 1985: 223). We would expect pre-WOJ *kusuy ‘stake, skewer’
on the basis of MJ kusu-nuk- H ‘to skewer’, although as Martin points out,
the register is incongruent (1987: 466). There are no cognates of this word
in Eastern Old Japanese, and Ryukyuan attestations are extremely limited.
There are only Shuri guusi (RGJ 1976: 195) and Yaeyama gui (Miyara
1980: 271). 120 The latter is apparently irregular, since something like *fusi
or *fuci would be expected in Yaeyama, depending on the dialect (Miyara
does not specify locations for this word). Thus, Yaeyama gui can be
ignored as a possible loan from Shuri, but Shuri guusi is isolated in
Ryukyuan, and itself likely represents a loan from Japanese. This brings
me to the conclusion that WOJ kusi < *kusuy is a loan from Korean. It also
probably preserves the syllabic structure of the Korean word, which could
be reconstructed as *kwócòy.
(142) (R) MK kù ‘that’ (medial demonstrative) ~ OJ kö ‘this’ (proximal
demonstrative) < PJK *kö (Whitman 1985: 223). 121 This was discussed
above in the section on demonstrative pronouns (2.1.2.3).
(143) (R) MK kèch ‘outer appearance, exterior’ ~ OJ kata HL ‘shape,
form’ < PJK *kecö/*kaco. Whitman adds: “The MK vowel requires PJK
front/back alternants. Note that OJ kata HL ‘side’ also appears as kötö with
front vocalism” (Whitman 1985: 223). There are multiple problems with
this etymology, and the dubious semantics is the least among them. I am
not aware of any OJ kötö with front vocalism meaning ‘shape, form’, and I
could not find either in JDB, or in all of the Old Japanese texts that I have
in my Old Japanese database. Unless a verse or line with OJ *kötö ‘shape,
form’ is provided, I am afraid it leaves us only with OJ kata ‘id’. Finally,
MK kèch may go back to PK *kèkúc or PK *kècúk. Only the latter could
remotely resemble OJ kata, but we do not have internal Korean evidence
for the suffix *-k, and in addition PK *u in the second syllable does not
correspond regularly to OJ /a/ < PJ *a. Therefore, I reject this etymology.

120
In the index volume Miyara also cites another word gudzï (Miyara 1981: 166), but it is
not found in the main text of his dictionary, and gudzï, even if it does exist, also presents
irregular correspondences.
121
This comparison is also found in Martin (1966), #240.
 Lexical Comparisons———145

(144) (R) MK :kel- ‘hangs it up’ ~ OJ kak-/kakë- ‘id’. < PJK *këkër-
(Whitman 1985: 223). 122 This etymology involves an irregular
correspondence of MK -l- to OJ -k-; see also (56), (65), and (66) above.
Since this etymology shares only the initial syllable, I reject it.
(145) (L) MK :kel- ‘thickens, congeals, freezes’ ~ OJ kör- ‘id’. < PJK
*kegöre-. Whitman adds: “OJ also has köNkör- and the parallel /y/ stem
verbs köye- and köNköye- in this meaning. The forms in köNkör/y-
represent the original stem shape, since the loss of the medial voiced
obstruent in pre-MK would account for MK length” (Whitman 1985: 223).
The Middle Korean word certainly means ‘thickens, congeals’, but I fail to
see any examples in the texts that would support the meaning ‘freezes’.
But since ‘freezing’ appears to be a secondary meaning in Old Japanese as
well, this is a minor philological point. While OJ kör- is attested in
phonetic spelling, OJ koye- and koNkor- are not (JDB 1967: 289, 313-
14). 123 In addition, koNkoye- is attested only in Early Modern Japanese
(IKJ 1990: 474; KKJ 1969: 341). That creates a reasonable doubt as to the
antiquity of forms with -Nk- in Japanese, as the first reliable attestation is
only from the late Heian period. However, this actually helps to improve
the etymology for the following reasons. First, MK :kel-, as we know it
nowadays, does not result from *kegel-, but from PK *kèlú-. Second, OJ
-Nk- does not correspond to MK -Ø-, but to MK -k- < PK *-nk-. Cf. also
(125) above. This leaves us with WOJ kör-, which is not attested in Eastern
Old Japanese. In Ryukyuan, there is the isolated Yaeyama kooruN (Miyara
1980: 292), but this looks like a late loan from Japanese, since /u/ but not
/o/ will be expected in the first syllable in Yaeyama. 124 Therefore, I think
that this etymology demonstrates another case of borrowing from Korean
into Central Japanese.
(146) (R) MK :kel- ‘rich, thick’ ~ OJ kôye- ‘rich, fertile, fat’ 125 < PJK
*keböre-. This etymology is followed by a lengthy discussion in which
Whitman admits that a correspondence of MK /e/ to OJ /ô/ is unacceptable.
He further correctly indicates that MK :kel- ‘rich, thick’ is the same
etymon as MK :kel- ‘thickens, congeals, freezes’ in (145) above (Whitman
1985: 224). These two facts should be sufficient to rule out this etymology,
as the same Proto-Japanese-Korean etymon could not have two different
reflexes in Old Japanese. But since I have established that the Old Japanese
‘cognate’ kör- in (145) above is likely a Korean loan in Central Japanese, I
am going to give this etymology the benefit of the doubt for a moment. Let
us assume that while (145) represents a loan relationship, the present
etymology represents the genuine cognate. In this case, the irregularity of
vowel correspondences in the first syllable still presents an insurmountable
problem; this difficulty alone could serve as a basis for rejection. Whitman

122
This comparison is also found in Martin (1966), #98.
123
Omodaka et al. write that OJ koNkor- appears as köNkör- (JDB 1967: 289), but since the
first phonetic attestation they cite is from the Ruiju myōgishō (1081), a late Heian
dictionary, it is risky to assign vocalism just on the basis of OJ kör-.
124
The vowel length remains unexplained. It may be a late loan from MdJ kooru ‘to freeze’?
125
This comparison is also found in Martin (1966), #78.
146———Koreo-Japonica

attempts to prove the validity of this etymology by bringing in OJ köpor-

‘to freeze, to congeal’, with the following note: “A verb in this set with a
medial voiced labial obstruent would produce OJ *köwor-/*köwoye-;
simplification of the /owo/ sequence (which is almost entirely restricted to
onomatopoetic expressions) would produce OJ kôye-. In fact there is direct
evidence for this Old Japanese verb form in the reduplicated
onomatopoetic phrase köworö-köworö ‘sound of liquid substance
congealing as stirred’” (Whitman 1985: 224). This citation opens a
Pandora’s box of unsolvable problems. First, OJ köpor- is very different
from OJ kör- in (145) above. The former means only ‘to freeze’ of ice,
water, rivers, snow, etc. (JDB 1967: 308). It never had the meaning ‘to
thicken, to congeal, to harden, to become stiff’, which is the primary
meaning of the latter. Second, it is inconceivable to have the development
köpor- > kör- in Old Japanese. Third, the suggested development from
*köwoye- (unattested) > OJ kôye- cannot be confirmed by any other
internal examples. Moreover, it is absolutely unrealistic: OJ /ö/ and /ô/
never combine within the same morpheme, but this proposed development
could work only on the assumption that the unattested *köwoye- goes back
to pre-OJ *köwôye-. Fourth, although there are indeed only WOJ töwo
‘ten’, töwom- ‘to bend’, wowor- ‘to overgrow’ reflecting an /owo/
sequence in Western Old Japanese, not only /owo/, but almost any /VwV/
sequence is rare. Fifth, and most important, OJ köworö-köworö does not
mean ‘sound of liquid congealing as stirred’, although this definition is
given in IKJ 1990: 527 without any discussion. Let us consider the
following example:

uk-î-si aNpura oti-naNtusap-î mîna köworö köworö n-i

float-INF-PAST/ATTR oil fall(INF)-churn-INF water churning churning
DV-INF
churning down the floating oil, churning-churning the waters (KK 100)

It is quite clear that köworö-köworö in this poem refers to a deity

churning or stirring the water to raise the foam that will eventually form an
island. No congealing is involved. This is further supported by OJ
naNtusap- ‘to stir up, to knock up’. All of this combined counterevidence
allows me to reject this etymology.
(147) (R) MK kèlí ‘road, street’ ~ OJ kaNtô HL, kanatô ‘gate,
entranceway’ < PJK *ker. Whitman provides another lengthy commentary
on this etymology, which I abbreviate below: “OJ kantô is derived from
kanatô by contraction and voicing assimilation (/nt/ > /nd/ > /d/). Kanatô is
itself a compound usually thought to derive from ka (?) + -na (genitive) +
tô (suffix meaning ‘place’), parallel to mînatô ‘harbor’ < mî ‘water’ + -na
+ tô. Here we compare the first element of this compound, ka- to MK kèlí
< *kèl + -i (nominative)” (Whitman 1985: 224). I have already commented
on the non-existence of genitive -na in Old Japanese (Vovin 1994: 249,
253; 2005a: 102-107); see also (63) above. A contraction of *naC > *nC >
*C[+voice] is very rare in Japanese; certainly there is no parallel form
 Lexical Comparisons———147

*mîNtô for mînatô ‘harbor’, and OJ mîNtu ‘water’ (cf. also WOJ mîna ‘id’.)
is probably the only somewhat parallel case I can think of. Similarly, there
are no alternants such as *taNkökörö for tanakökörö ‘palm of the hand’ or
*mîNkîpa for mînakîpa ‘waterfront’. Thus, the parallelism as suggested by
Whitman may not be there, or at least it may not be so obvious without
additional proof. The greater problem, of course, lies with a comparison of
MK kèlí ‘road’ with OJ ka- (?) which is not attested independently, and for
which we can verify neither the meaning of the word nor its very existence
in the language. As far as I can tell, no existing Japonic idioms have *ka
‘road’. Treating -i in MK kèlí ‘road, intersection’ as a nominative case
marker is also an unsubstantiated claim, because in Middle Korean texts
the word never appears as *kel- followed by a different case marker. A
couple of examples are in order:

kil-s keli-lul lim-hoy-a-se

road-GEN intersection-ACC face-do-INF-GER
facing the intersection of the roads (Twusi cho VIII: 21a)
keli-yey pakphuyngi thi-l ahoy
street-LOC toy top hit-ATTR/IRR boy
boy(s) who play with toy tops in the street (Pak cho I: 17a)

These examples demonstrate that -i in MK kèlí is not a nominative case

marker, but part of the root. Finally, I think that OJ kaNtô < kanatô has a
simple internal Japanese etymology. There is no reason to explain -tô in
kanatô or kaNtô as OJ tô ‘place’: after all, a ‘gate’ is a kind of ‘door’, and
not just any kind of ‘place’. Thus, using OJ tô ‘door’ rather than OJ tô
‘place’ is much more realistic for this etymology. Second, the difference
between simple doors and gates in ancient Japan was that the latter always
had some metal implements, while the former did not necessarily have
them. Therefore, OJ kanatô/kaNtô ‘gate’ is simply kana-tô ‘metal door’,
the etymology also suggested in JDB 1967: 203. In view of this simple
internal explanation, any external comparison has to be abandoned.
(148) (R) MK kèmúy ‘spider’ ~ OJ kumo LF ‘id’. 126 < PJK *kerümö.
Whitman adds the following commentary: “The initial vowel
correspondence in this well-known comparison is irregular, and there is no
Kojiki spelling available to verify expected final /ö/ (rather than /ô/) in OJ.
Kyūshū has kobu, with the vowel expected from the MK comparison, but
with a medial consonant that raises additional problems. Shuri has kuubaa,
with the same medial consonant. The length of the Shuri form may indicate
contraction in both syllables” (1985: 224). Due to recent advances in
Japonic historical phonology, several comments must be made on Japonic
‘spider’. First, although there is no phonetic attestation of WOJ kumô in
Kojiki, the sequence CuCö is not attested at all in Western Old Japanese.
The phonemes /u/ and /ö/ do not occur in neighboring syllables of the same

126
This comparison is also found in Martin (1966), #214.
148———Koreo-Japonica

morpheme. Therefore, by default the word must be *kumô, because CuCô

is well attested. Second, 久毛 ‘spider’ is attested phonetically as kumô with
the last syllable written as 毛 /mô/ in MYS V: 892. Bentley has
convincingly demonstrated that book five of the Man’yōshū has a very
high percentage of etymologically correct spellings of /mô/ and /mö/ at
least in nouns (Bentley 1997), so now we have a philological confirmation
in addition to the structural point mentioned above. Thus, from now on I
will write the word in Western Old Japanese as kumô. The vocalism of the
second syllable of this spelling represents an irregular correspondence with
Korean. Third, the vowel length in the first syllable of Shuri kuubaa
‘spider’ indicates the primary Japonic vowel length, and has nothing to do
with contraction (Hattori 1979a, 1979b; Shimabukuro 2002: 203), 127 so the
Proto-Japanese-Korean reconstruction *kerümö is completely ad hoc.
Fourth, the Ryukyuan forms indicate PR *koobo ‘spider’, 128 with a
primary PR and PJ *o, but this does not improve the vocalic
correspondence in the first syllable, because OJ /ö/ < PJ *ə is needed for
regularity. Fifth, the problem with WOJ -m- vs. PR -b- < PJ *-np-, as
pointed out by Whitman himself, still remains. There are no cognates of
WOJ kumô ‘spider’ in Eastern Old Japanese, but in the sources on the
Hachijō dialect, the sole modern survivor of Eastern Old Japanese, the
word kubona[me] ‘spider’ is attested in three different sources, while kumo
‘big spider’ appears in just one (Kokuritsu kokugo kenkyūjo 1950: 344).
This probably indicates that both PR *koobo and Hachijō kubona[me]
preserve the same archaism with the reflex -b- of the original Proto-
Japonic cluster *-np-, while WOJ kumô with medial -m- is an innovation.
In addition, a change in the medial position from *-np- > -m- is
typologically more viable than the other way around. This leads me to
reconstruct it as PJ *koonpo ‘spider’, and I believe that the only part it
shares with MK kèmúy ‘id’. is the initial consonant. Thus, I reject this
etymology as a likely chance resemblance.
(149) (R) MK kèp(h)- ‘shell, husk’ ~ OJ kapî ‘shell(fish), husk (of rice,
etc.)’ < PJK *kep. Whitman adds: “MK kèp(h)- appears as kèphí (with
nominative suffix -i absorbed with high tone), kèphùl, and possibly as the
first syllable of MK kèpwúp, [Md]K kepwuk ‘turtle’” (1985: 225). The first

127
Shimabukuro managed to find a strong internal correlation supporting a reconstruction
of vowel length in Ryukyuan on the basis of two, not just one, Okinawan dialects, thus
taking care of Martin’s objection that Hattori’s data are not consistent (Martin 1987: 252-
253). Namely, Shimabukuro demonstrated that the initial accent in the Nakijin dialect
regularly corresponds to the vowel length in Shuri for the subcategories 2.3a-2.5a, while
there is no such correspondence in the cases where vowel length is absent in Shuri
(subcategories 2.3b-2.5b).
128
Thorpe reconstructs PR *kobu (Thorpe 1983: 333), but I supply the vowel length in the
first syllable on the basis of the recent work by Shimabukuro (2002). I also change the
second syllable final vowel from *u to *o, because that is what some of the Northern
Ryukyu dialects that were not incorporated in Thorpe’s study indicate, e.g., Oku khumo
‘spider’ (Uchima and Arakaki 2000: 393). This is also further supported by the fact that at
least some occurrences of Western Old Japanese final /ô/ reflect primary PJ *o as well; in
these cases we have two independent pieces of evidence supporting the Proto-Japonic
reconstruction *koonpo ‘spider’.
 Lexical Comparisons———149

problem is that OJ kapî is nothing but a ‘shellfish‘, while ‘husk’ is OJ

kaNpî (JDB 1967: 210-211), so these two words should not be confused.
Only the former can be phonetically comparable with Korean. Comparison
of Korean words also runs into a number of significant problems. First,
MK kèphí may go back either to PK *kèpùkí or *kèkùpí, and similarly MK
kèphùl can be derived from both PK *kèpùkùl or *kèkùpùl. Only the forms
*kèpùkí or *kèpùkùl can be remotely compared to OJ kapî, which reduces
the credibility of the comparison down by half. Second, I cannot see any
internal Korean justification for analyzing -i in MK kèphí as a nominative
case marker, similar to the case MK kèlí ‘street’ discussed in (147) above.
As far as I can tell, MK kèphí ‘husk, bark’ appears just four times in two
Middle Korean texts, and of those, two are confined to the same passage
and context in a single text:

kephi pasky-e... kephi paski-kwo

husk peel off-INF... husk peel off-GER
peeling off the husk... peeling off the husk (Kwukup I: 41a/b)
kephi as-kwo
bark peel off-GER
peeling off the bark (Kwukup II: 59a/b)

PAK-on namwo-s kephi (i-)ra

PAK-TOP tree-GEN bark (COP)-FIN
PAK is the bark of the tree (Pep I: 220a/b)

In none of these examples does MK kèphí appear as the subject of the

sentence, so it is safe to conclude that -i belongs to the stem. Here, the
existence of MK kèphùl ‘bark, husk, bamboo sprouts’ can potentially
indicate a morphological boundary, but the different accentuation pattern
and the nonexistence of the suffix *-ul in Korean leave too many internal
problems to be solved before we can involve any of these in external
comparisons. All these phonetic and morphological problems on the
Korean side, plus the less than perfect semantics on the Japonic side
prevent me from accepting this etymology as a valid one.
(150) (L) MK kès ‘thing, the one, the fact that’ ~ OJ kötö LL ‘id’. 129 <
PJK *ketö (Whitman 1985: 225). See also the discussion above under the
genitive case marker in the morphology section, where MK -s- ~ OJ -t-
was treated as a loanword correspondence. The fact that prevents me from
embracing this etymology as a genuine Koreo-Japonic cognate is the
unique situation in Ryukyuan, where kutu ‘thing’ is attested as an
independent word ‘thing’, but is never used as a nominalizer ‘the one, the
fact that’, 130 which is one of the most fundamental functions of koto in
Japanese and kes in Korean. Since it is much easier to borrow an
129
This comparison is also found in Martin (1966), #239.
130
For this purpose Ryukyuan languages use the bound noun si, which appears to be
completely unrelated.
150———Koreo-Japonica

independent word than a nominalizer, and since the correspondence of MK

-s- to OJ -t- appears in other loanwords from Korean to Japanese, I think
that this is another loan from Korean to Japanese with a subsequent spread
in its function as an independent noun to the Ryukyus.
(151) (L) MK :ket- 131 < [PK *]kètú- ‘walks’ ~ OJ kati LH ‘walking’ <
PJK *ketö. Whitman comments: “OJ kati LH appears to be the
nominalized (continuative) form of a verb *kat- or *kati-, of which it is the
only survivor. MK :ket- is historically a thematic verb stem (Ramsey 1977),
suggesting that OJ kàtí LH is to be associated with bigrade thematic
conjugation (*kati-) as we would expect from the final syllable /ti/ (=[tï])”
(Whitman 1985: 225). This is exactly where the problem for a genetic
comparison lies: Besides WOJ kati there are Middle Japanese attestations
(KKJ 1969: 230), and possibly Early Modern Japanese attestations as well,
albeit with a slight semantic shift (IKJ 1990: 304). There is also EOJ kasi
‘walking’ attested once in MYS XX: 4417, which contradicts Whitman’s
suggestion that OJ kati is derived from a thematic verb: we would not
expect a palatalization /ti/ > /si/ in Eastern Old Japanese in the case of kasi-
< *katï-. Needless to say, there are no cognates in Ryukyuan. It is also
highly suspicious that both WOJ kati and EOJ kasi survived in one and the
same form, with no other paradigmatic forms attested. This suggests a
much simpler explanation than a genetic comparison with Korean: WOJ
kati was borrowed from Korean as a set form, and then re-borrowed into
Eastern Old Japanese as kasi.
(152) (C) MK :key ‘crab’ ~ OJ kani HH ‘id’. 132 < PJK *keni (Whitman
1985: 225). Since the Middle Korean sequence [C]Vni within the same
morpheme is practically unattested, a reconstruction of MK :key as PK
*keni seems to be viable, but we must keep in mind that the Ceycwuto
dialect has the form kəŋi ‘crab’, and other forms with -ŋ- sporadically
appear in other dialects (Choy 1978: 946-947). This medial -ŋ- is unlikely
to be from PK *-nk-, but the reconstruction of the Proto-Korean form is
not quite clear. Tentatively, I accept this etymology, but it still may be
rejected in the future once we have a better understanding of the Proto-
Korean reconstruction.
(153) (R) MK kyél ‘wave, grain, texture’ ~ MJ (Wamyōshō) kisa ‘wood
grain’ < PJK *kil2a (Whitman 1985: 225). There are several problems with
this etymology. First, there is a problem with the correspondence of MK
/ye/ to OJ /î/; see (29) above on that issue. In addition, since the only given
attestation is MJ kisa, we do not even know whether it was OJ *kîsa or
*kïsa. Second, we should not forget that MJ kisa (木佐) ‘wood texture
pattern’ is a hapax legomenon, attested only once (WMS XX: 22b).
Furthermore, the word is glossed in the Wamyōshō as 木文 ‘wood texture
pattern’, and the Modern Japanese translation 木目 /mokume/ ‘wood
grain’, as far as I can tell, appears only in IKJ 1990: 358. It is also signifi-
cant that the Wamyōshō provides the explanation that the origins of MJ
131
Whitman gives MK kèl-, but this is probably a typographical error, since MK :ket- is a
-t- leniting verb belonging to verbal accent Class 6, with alternating stems kèlú- and :ket-.
132
This comparison is also found in Martin (1966), #54.
 Lexical Comparisons———151

kisa ‘wood texture pattern’ is in its resemblance to the pattern found on the
kisa (MdJ akagai) ‘ark shell’ (WMS XX: 22b). A greater problem may lie
on the Korean side, since MK kyél clearly just means ‘wave’, and nothing
else (LCT 1987: 56; Nam 1997: 80). In Early Modern Korean kyel also
acquires the meaning of ‘wavy pattern’ (LCT 1987: 56), which still
survives in MdK namu-s kyel ‘wavy pattern of lines on the bisected section
of a tree’. 133 It is quite clear that these later usages ultimately are derived
from MK kyél ‘wave’. Thus, this comparison boils down to a comparison
of ‘ark shell’ with ‘wave’, and it also includes an irregular correspondence
of OJ /i/ to MK /ye/. Therefore, I reject this etymology.
(154) (R) MK kyèth ‘side (of)’ ~ OJ kata HL ‘one side of, half of,
direction’ < PJK *kyato. Whitman adds that Old Japanese also has kötö
(Whitman 1985: 225). I believe that OJ kata means only ‘direction, side’,
while OJ kata- ‘one’ is a completely different etymon, also surviving in
modern compounds such as: kata-miti ‘one way’, kata-asi ‘one foot’, kata-
ude ‘one arm’, kata-omoi ‘unreciprocated love’, kata-oya ‘one parent’,
kata-gawa ‘one side’, kata-toki ‘single moment’, etc. This kata- ‘one’ is
certainly a Korean loanword: although Middle Korean has hònàh (MdK
hana) ‘one’, pre-fifteenth-century Korean materials clearly indicate earlier
*xata-: EMK xatun (Kyeylim #19), OK HAton ‘one’ (Hyangka VII: 6, 8;
XI: 7). I am not aware of the existence of any OJ kötö ‘side, direction’,
either. A lack of regularity in the vocalic correspondences in the first
syllable between MK kyèth ‘side (of)’ ~ OJ kata HL ‘side, direction’
certainly calls this etymology into question. Cf. also (126) above, where a
different Middle Korean ‘cognate’ of OJ kata is used. Thus, I reject this
etymology.
(155) (R) MK kùlùh ‘stump, counter for trees’ ~ OJ kö-/kï L ‘tree’ <
PJK *kör (? + -Vk diminutive) (Whitman 1985: 225). This etymology
certainly depends on positing the loss of *-r- for OJ kö-, a loss that cannot
be verified internally for Japonic in this case. I cannot see any Middle
Korean textual evidence for Whitman’s claim that in addition to meaning
‘stump’ MK kùlùh is also used as a counter for trees. 134 Although I may be
wrong, I am not aware of any crosslinguistic semantic shifts ‘tree’ >
‘stump’ or vice versa. In uncontroversial families the word for ‘tree’
generally tends to retain its semantics, cf. English tree and Russian derevo
‘id’. Therefore, I reject this etymology on the basis of both phonetic and
semantic problems.
(156) (R) MK kyèzúlh ‘winter’ ~ OJ kisaraNki ‘second [lunar — A.V.]
month, month of ice melting’ < PJK *kisereg (Whitman 1985: 225). 135 OJ
*kisaraNki is not attested phonetically in Old Japanese (JDB 1967: 240),
so for all practical purposes we have to deal with MJ kisaragi (Vovin
133
I would like to express my gratitude to my late colleague Hwang Jiha for explaining the
exact meaning of this compound.
134
MdK kulwu ‘stump’ can be used as a counter for trees, but without direct supporting
evidence from Middle Korean, this is likely to be a late secondary development.
135
Whitman presents an additional discussion for this etymology, but it is very speculative
and represents an attempt to save this particular etymology in a unique way, so I omit it
here.
152———Koreo-Japonica

1993c: 116). We have no indication which Old Japanese vowels, /î/ or /ï/,
were originally in the first and last syllables, and this provides all kinds of
opportunities for wild speculation. In addition, it is quite clear that MJ
kisaragi is too long to represent a single Proto-Japonic root; therefore, it
must be a compound of some sort. We have no internal Japonic evidence
to break this compound, although speculations are possible. I presented
one, that now seems very unlikely, in Vovin 1993c. Names for lunar
months are likely to have much later origins than the alleged Koreo-
Japonic proto-language, and in any case I doubt that any ice really melts
during the harsh Korean winters. This is all the more so in Manchuria,
where the likely motherland of any supposed Koreo-Japonic proto-
language would be located. Therefore, I reject this.
(157) (L) MK kús ‘certainly, without fail (emphatic adverb)’ ~ OJ kösö
‘indeed, verily, without fail’ < PJK *kösö (Whitman 1985: 225). Besides
WOJ kösö, EOJ and MJ koso are also well attested, but there are no
cognates in Ryukyuan. Therefore, I treat this etymology as a likely loan
from Korean to Japanese.
(158) (R) MK kùz- < *kùsùl-; cf. [Md]K kkul- ‘draw, drag, pull’ ~ OJ
köNsi ‘pull out by the roots’ < PJK *közör- (Whitman 1985: 225). My
impression from the textual examples in Western Old Japanese cited by
Omodaka et al. is that the meaning of OJ köNsi- is rather ‘to dig out the
roots’ (JDB 1967: 294), resulting in a significant semantic difference. The
etymology might still be tempting as a loanword, because there are no
other Japonic attestations besides WOJ köNsi-. However, since PK *-s- can
correspond only to OJ -s- < PJ *-s-, and not to OJ -Ns- < PJ *-ns-, it is best
abandoned altogether as a chance resemblance.
(159) (R) MK :kwul ‘chimney’ ~ OJ kuri- in kuri-ya LL-L ‘cookhouse’
< PJK *kü:r (Whitman 1985: 226). WOJ kuriya ‘cookhouse, kitchen’ is an
obscure compound, with no internal evidence that kuri- there means
‘chimney’, and not, let us say ‘oven’ or ‘fireplace’. There are no other
attestations outside Western Old Japanese, and the etymology can be safely
discarded.
(160) (R) MK kwùlèk ‘basket’ ~ OJ kô L ‘id’. < PJK *kür (+ -Vk
diminutive) (Whitman 1985: 226). There are several problems with this
etymology. First, MK kwùlèk ‘basket’ is essentially a hapax legomenon: it
appears only once in (Twusi cho XXII: 11a), and then in several Early
Modern Korean anthologies of sico, but in the same line of the same text
(LCT 1987: 82; Nam 1997: 138). Furthermore, MK kwùlèk, judging by its
only one available attestation in Middle Korean seems to mean not just any
‘basket’ in general, but a ‘square basket for fruits’, since it translates the
Chinese character 筺 of the same meaning. Since Korean *kwul ‘basket’
does not present itself, segmenting diminutive *-ek in MK kwùlèk is
speculative at best. Finally, the most serious problem lies with OJ kô,
which may go back to PJ *ko, *kau, and *kua. The tentative PJ *kua may
be comparable with MK kwùlèk ‘basket’, but only assuming *-r- loss in
Japanese, for which there is no internal evidence. This leaves us with only
 Lexical Comparisons———153

a one in three chance that this etymology is credible. Given all the other
problems mentioned above, it is best rejected.
(161) (R) MK kwùlí ‘copper’ ~ OJ ku-/ki ‘yellow, reddish-gold color’ <
PJK *kür (Whitman 1985: 226). WOJ ki ‘yellow’ is not attested
phonetically, but on the basis of WOJ ku- occurring in the compound ku-N-
kane ‘gold’ (lit.: yellow metal) we can speculate that it was WOJ *kï. The
comparison with Korean has a significant problem, however: there is no
internal evidence for the Korean suffix -i in MK kwùlí. It must represent a
root, as does pre-OJ *kuy > *kï < OJ ki. One then would have to match
pre-OJ *-y to MK -li, which is not possible without apriori assumption of
the *-r- loss in pre-Old Japanese. Therefore, I reject this etymology.
(162) (R) MK kwùlú- ‘stamps feet, treads noisily’ ~ OJ kuwe- ‘kicks,
steps hard’ < PJK *küre- (Whitman 1985: 226). MK -l- certainly does not
correspond regularly to OJ -w-. Even if one assumes *-r- loss in this Old
Japanese word, finding Korean correspondences for the unaccounted
segment -we- in the Old Japanese word presents another problem.
Therefore, I reject this etymology.
(163) (R) MK kwùlwúm ‘cloud’ ~ OJ kumô ‘id’. 136 < PJK *kürümü.
Whitman adds: “The final vowel in [Western] Old Japanese is secondary;
Azuma [that is, Eastern Old Japanese — A.V.] has the expected /u/” (1985:
226). As far as I can tell, EOJ *kumu ‘cloud’ is not attested at all. EOJ
kumo (spelled as kumô or kumö) is attested as an independent word in MYS
XIV (3510-3512, 3514, 3516, 3518, 3520, 3522) and in MYS XX (4380,
4421). In addition, it also appears in the compound kumo-wi ‘sitting place
of clouds’ in MYS XIV: 3441 and MYS XX: 4355. The basis for
Whitman’s statement that final /ô/ in WOJ kumô is secondary is unclear to
me. It appears instead to be original, since in addition to EOJ kumo it is
also supported by PR *kumo ‘cloud’ (Thorpe 1983: 272), which
demonstrates that WOJ /ô/ in kumô reflects a primary PJ *o. There is no
internal Japonic basis for the justification of *-r- loss except to compare it
with Korean, and this leaves only the initial syllables looking alike in
Korean and Japonic. Therefore, I reject this etymology.
(164) (R) MK :kwup- ‘roasts it, toasts it, 137 burns wood to make
charcoal’ ~ MJ kube- ‘feeds wood into fire’ < PJK *kürübe- (Whitman
1985: 226). MK :kwup- ‘to bake, to roast’ is a Class 6 verb that contains
leniting MK -p- < PK *-p-, not < PK *-np-. MJ kube-, on the other hand, is
from PJ *kunpai-. Therefore there is an irregular correspondence of PK
*-p- to PJ *-np-. I am unaware of any Middle Korean examples in which
the word would mean ‘to burn wood into charcoal’: this seems to be a
semantic innovation in Modern Korean. Without this meaning, the
comparison with MJ kube- becomes even less plausible; therefore, I reject
this etymology.
(165) (R) MK kwúp ‘hoof’ ~ OJ kuNpîsu HHL ‘heel’ < PJK *küb(+i) (+?
sü). Whitman adds: “The OJ form also appears as kîNpîsu HHL, kupîpîsu,
and kîpîpîsu. The forms with first syllable /î/ show sporadic fronting of
136
This comparison is also found in Martin (1966), #47.
137
The semantic tag ‘to bake’ needs to be added as well.
154———Koreo-Japonica

PJK *ü. I assume the forms in -pîpîsu involve reduplication, interestingly

with /p/ rather than /Np/ to avoid a Lyman’s law violation. I cannot
identify the final syllable in the OJ form, although the MK comparison and
the (otherwise internal) reduplication of /pî/ suggest that it is a suffix”
(1985: 226). First of all, the Middle Korean form cannot be used as
evidence for internal segmentation of a suffix in proto-Japonic, because it
involves ‘reconstruction from above’, which is methodologically
unacceptable. Second, OJ kuNpîsu is attested only once phonetically in Old
Japanese. In addition, it is spelled 久比須 /kupîsu/ (JDB 1967: 267), so it
is likely that it is just kupîsu, and not kuNpîsu. As far as I can tell, the form
kîNpîsu is attested for the first time in the Wamyōshō as 岐比須 /ki[N]pisu/
(WMS III: 14b), also with the voiceless sign 比 /pi/, so it is another hapax
legomenon, and in addition not from Old, but from Middle Japanese. 138
Besides, it is marked as a vulgarism in the Wamyōshō, so I doubt it can be
used as evidence for ‘sporadic fronting’ of PJK *ü: most likely, it is just a
sample of sloppy vulgar pronunciation. The form kipipisu appears twice in
only one manuscript of the Shinsen jikyō (898-901) (JDB 1967: 226), so it
is again a Middle, and not an Old Japanese attestation, but such a limited
attestation can quite easily indicate simple scribal errors. Thus, the only
reliable Old Japanese form is WOJ kupîpîsu. In addition, I cannot think of
any other examples in Japanese involving only partial reduplication of the
last syllable of the root in front of another suffix. Therefore, I think that the
traditional explanation of MJ kup/bisu, as a contraction of WOJ kupîpîsu
(IKJ 1990: 413) is more likely than Whitman’s suggestion of reduplication.
That leaves no evidence for a suffix -su or -pîsu in WOJ kupîpîsu. Taking
the less than ideal semantics into account as well, this comparison should
be rejected.
(166) (L) LMK kwus ‘something demonic’, MdK kwus ‘exorcism
ceremony’ ~ OJ kusi ‘something mysterious or ghostly’ < PJK *küs.
Whitman notes that the earliest appearance of the Korean word in the form
kwus-kes ‘demon, spirit’ < kwus + kes L ‘thing’ is in the reprint of the
Twusi enhay of 1632 (Whitman 1985: 226). Actually the word is not only
attested in Early Modern Korean (Whitman’s LMK), the earliest attestation
of MK kwús ‘[exorcism] ceremony’ occurs in Middle Korean:

emi PHYENGSENG-ay simpang kwus spwun culki-si-l-ssoy

mother whole life-LOC shamaness ceremony only enjoy-HON-
ATTR/IRR-because
Because mother only enjoys shamanistic ceremonies her entire life (Welin
XXIII: 68)

The paucity of Middle Korean attestations can be explained by the very

nature of the Middle Korean texts: most of them are either Buddhist or
Confucian in content. In addition, the disdain of Yi dynasty rulers for
shamanistic rituals is well known. Considering these two factors, we

138
There are other (later) Middle Japanese attestations of kubisu.
 Lexical Comparisons———155

should not expect frequent attestations. However, the problem of

perceiving this etymology as a Koreo-Japonic cognate lies on the Japonic
side. WOJ kusi ‘strange, mysterious’ is not attested outside Western Old
Japanese; and its variant WOJ kususi- appears only in Western Old
Japanese and Middle Japanese. 139 Since there are no Eastern Old Japanese
or Ryukyuan attestations, WOJ kusi is a perfect candidate for a loan from
Korean.
(167) (L) MK kwùsúl ‘public office, government post’ ~ OJ kîsi ‘title
for the clan leader; specifically for Korean (Paekche and Silla) diplomats
and military officials, and for heads of families of Korean extraction’.
Whitman comments: “The OJ term is thus clearly a loan from OK (Silla)”
(1985: 226). I agree: this is indeed a loan.
(168) (R) MK kwùsú- ‘(pleasantly) odorous’ ~ OJ kusa- ‘smelly’.
Whitman adds: “The final vowel in OJ indicates PJK *küse-, with normal
weakening of the second unaccented vowel in MK” (1985: 227). The
accentuation of the Middle Korean word is, however, kwùsù-, not kwùsú-,
as seen in the texts, see e.g., twòthóy kìlúm-kwá kwùsùn :kes-kwá-lól kwù-é
‘roasting pig’s fat and pleasantly odorous things’ (Kwukup II: 43b). This
suggests that MK kwùsù- is a verb belonging to accent class 8, which has a
complex Middle Korean morphological and phonological history (Ramsey
1991: 236) that needs to be clarified before any external comparisons are
attempted. We really do not know whether the second vowel in this word
is the result of normal weakening of the second unaccented vowel or not.
Also, it is certainly not methodologically viable to reconstruct PJK *küse-
on the basis of OJ kusa- and then to reconstruct MK kwùsù- < PK *kuse-
on the basis of that, because it is a tautological reconstruction from above.
Therefore, until the phonological and morphological history of MK kwùsù-
is clarified, we are facing an irregular vocalic correspondence between
Middle Korean and Japonic in the second syllable. Also, given the
semantic differences (which are, however, possible to explain), it seems to
me that this etymology should be rejected for the time being.
(169) (L) MK kwùsúl ‘bead, precious stone’ ~ OJ kusirö ‘decorative
armband, bracelet’ < PJK *küsiro (Whitman 1985: 227). There is also
WOJ kusi, attested in the compound kusi mî-tama ‘precious stone’ in MYS
V: 813 and 814. 140 It is quite possible that the same word occurs in KK 39
and 49 in reference to precious sake. JDB 1967: 257 defines it as a word
praising sake. The problem with establishing a Koreo-Japonic etymology
is again connected with the distribution of both kusirö and kusi in Japonic:
both are attested exclusively in Western Old Japanese. Thus, I believe that
this is another loan from Korean into Western Old Japanese.

139
The Middle Japanese attestations are quite rare and completely disappear after the mid-
Heian period: there is one attestation in the Murasaki shikibu nikki, one attestation in the
Genji monogatari, and two attestations in the Makura-no sōshi (Miyajima 1971: 102).
140
There is a tendency to explain this kusi as ‘mysterious’ (Takagi, Gomi, and Ōno. 1959:
71), discussed above in (166). However, the context of the poem would also allow this kusi
to be interpreted as ‘treasure, precious thing’.
156———Koreo-Japonica

(170) (R) MK kwút ‘hollow, pit, cave’ ~ OJ kutu-/kuti ‘mouth’ 141 < PJK
*kütü (Whitman 1985: 227). Aside from the somewhat doubtful semantics,
not a major problem here, there is a problem in the regularity of the
correspondences. MK kwút cannot be from PK *kwut[w]u, because with
intervocalic PK *-t- we would expect a lenition *-t- > -l-, and the form MK
*kwúl, not kwút. MK kwút can be either originally monosyllabic from PK
*kwút, or from PK *kwunt[w]u. In both cases we will have a problem. In
the first case, we would need to explain the correspondence of a
monosyllabic Proto-Korean word to a disyllabic word in Japonic: in other
words, the second syllable vowel in Proto-Japonic corresponds to nothing
in Proto-Korean. In the second case, we will have an irregular
correspondence of PK *-nt- to PJ *-t- (PJ *-nt- should be expected). Thus,
I dismiss this etymology on the basis of apparently irregular
correspondences.
(171) (R) MK kwùt- ‘hard, solid, adamant, unyielding’ ~ OJ kîta- ‘id’.
in kîtasi ‘hardened salt’ < kîta + si(po) ‘salt’ < PJK *kyüt-/*kyut-
(Whitman 1985: 227). Whitman comments: “OJ kîtamë- ‘chastises, deters
through punishment’ is also thought to be cognate with these forms (IKJ
1990). If MJ kitu-si ‘cruel, hard, tough’ is also, we have some indication
that the second vowel in OJ kîta- is not original” (1985: 227). I do not
quite understand his logic here: not only do kîta- and kitu- have a
significant semantic discrepancy, MJ kitu- certainly belongs to a different
dialect which is later than Old Japanese. This is not to say that later forms
cannot preserve more archaic forms; they certainly can, but there is no
evidence to claim that MJ kitu-si has an original vowel in the second
syllable, while OJ kîta- does not. Whitman adds a lengthy discussion that
also involves OJ kata- ‘hard’ and a number of ‘Altaic’ forms that need not
detain us here. Whitman believes that OJ kata- ‘hard’ and MK kwùt- ‘id’.
are “doubtlessly cognate, 142 but the vowel correspondence [in the first
syllable ― A.V.] is irregular; OJ kita- indicates the survival of a regular
correspondence for MK kwùt- in J[apanese]” (1985: 227). But according to
Whitman’s own set of vocalic correspondences, OJ -î- does not correspond
regularly to MK -wu- (1985: 129). The reconstruction of PJK *kyut-
/*kyüt- is, therefore, completely teleological, as it is produced by simple
juxtaposition of the Old Japanese and Middle Korean vowels that do not
match. In addition, WOJ kîtasi is not attested in Eastern Old Japanese or in
Ryukyuan, and even its attestation in Western Old Japanese is not
completely unproblematic. It is only properly attested phonetically for the
first time in the Heian period dictionary Wamyōshō (931-938), with
Western Old Japanese “attestations” being phonetic glosses added to the
original texts at much later dates. Finally, segmentation of WOJ kîtasi as
kîta-si is based on the assumption that -si is a truncation of OJ sipo ‘salt’.
Unfortunately, the proposed *kita- ‘hard’ does not appear in other
compounds, and its connection with kîtamë-, which simply means ‘to
punish, to chastise’ and not ‘to deter through punishment’, is tenuous at
141
This comparison is also found in Martin (1966), #150.
142
This comparison is also found in Martin (1966), #99.
 Lexical Comparisons———157

best. Furthermore, to the best of my knowledge, OJ sipo ‘salt’ does not

undergo truncation in other compounds, such as kata-sipo, which also
means ‘hard salt’. Thus, chances are that WOJ kîtasi represents a
specialized term for ‘hard salt’ with an obscure internal structure. All these
problems make me reject this etymology.
(172) (R) MK kwùwùl-, kwùùl-, continuative kwùllé- < *kwùlúl-; MdK
kwulu- (-ll-), :kwul- ‘roll’ ~ OJ kururu-ni ‘rolling, revolving’, kuruma
‘wheel, cart‘, kuru, MJ kururu ‘pivot door swings on’ < PJK *kürü(r)-
(Whitman 1985: 227). Whitman adds: “The MK continuative forms and
the [Md]K verb clearly indicate original *kwùlùl-, 143 although loss of
medial */l/ evinced in the MK citation forms kwùwùl- and kwùùl- is
exceptional. OJ kururu-ni and MJ kururu preserve both *r’s; however, the
fact that the former is onomatopoetic and the second later than OJ kuru
may suggest that final -ru has been restored by reduplication” (1985: 227).
There are numerous problems with this etymology. Let us start with the
Korean data. First, Whitman overlooked the fact that both MK kwùwùl-
and kwùùl- are in fact kwùGwùl- and kwùGùl-, which can be deduced on
the basis of Middle Korean spelling. Second, the Middle Korean form
kwùGùl- appears late, with the first and only Middle Korean attestation in
the first edition of the Pak thongsa (1515); all of the other attestations
belong to Early Modern Korean (Nam 1997: 146). MK kwùGwùl- is
attested in fifteenth-century texts (Nam 1997: 145). Therefore, we have
here the typical change of /wu/ > /u/ in the second syllable, which took
place in the history of Korean. Third, as demonstrated by the form kwupul-,
which is found in Hamkyeng Namto, and in Kyengsang (Kim Pyengcey
1980: 431) MK kwùGwùl- < pre-MK *kwùWwùl- < PK *kwùpwùl-.
Fourth, to the best of my knowledge, the Middle Korean continuative form
kwùllé for this verb does not exist. The continuative form is in fact
kwùGwùl-é ‘roll and’ (Nammyeng I: 12a). There is an Early Modern
Korean continuative form kwuGulle, attested in the Hancheng munkam
(Nam 1997: 146), compiled during the reign of King Yengco (1725-1776),
but this is late by any standard, and in any case the form is still not *kwulle.
Thus, if there is any evidence at all to treat this verb as l-doubling on the
basis of this late form and its morphological behavior in Modern Korean,
the Proto-Korean form is likely to be reconstructed as *kwùpwùlúl-, not
*kwùlùl-. If we give more weight to the Middle Korean evidence alone,
the reconstruction is likely to be simply *kwùpwùl-. In any case, there is
no “exceptional loss” of */l/, at least not in the place that is needed for
comparison with Japanese. Turning now to the Japanese side of the
comparison, it first must be noted that OJ kuru and MJ kururu refer not to
the swinging of a door pivot, but to a device whereby a door’s pivot is
inserted into a groove or attached to the door itself (JDB 1967: 275).
Therefore, the connection with ‘rolling’ may be far-fetched. WOJ kururu-
ni should be removed from the equation as well, not only because it is
onomatopoetic, but also because it is a hapax legomenon, which is in

143
Cf. *kwùlúl- given with different accentuation at the beginning of this entry.
158———Koreo-Japonica

addition known to us only from a later phonetic gloss in the Nihonshoki

(JDB 1967: 276). That leaves us only with WOJ kuruma, which
incidentally is only ‘cart‘, not ‘wheel’. The etymology for this word
baffled the minds of many generations of Japanologists, and for good
reason, as the reader will see below. The word is not attested in Eastern
Old Japanese, but this is probably accidental. It is attested throughout
Ryukyuan (Hirayama 1966: 337), but this is no surprise, since it refers to
an object that was likely borrowed from Japan. The fact that in most
Ryukyuan dialects this word appears completely unaltered by any phonetic
developments probably speaks against its antiquity in Ryukyuan.
Nevertheless, the possibility does remain that we are dealing with a
genuine Proto-Japonic word here, unless one can present some evidence to
the contrary. I believe that two independent pieces of such evidence do
exist. First is the historical evidence: it seems that kuruma as an object
appeared for the first time only in the Nara period (Takayanagi and
Takeuchi 1974: 299). It probably would not be a stretch to believe that it
might have been present in the Asuka period as well, but the absence of
any archaeological evidence of its existence for earlier periods, and
especially for Yayoi, is conspicuous. Second, there is the linguistic
evidence. If WOJ kuruma is a native word, it surely must be a compound,
because it is simply too long. However, viewing this in the most favorable
light, it must be an obscure compound. Yet, an obscure compound
describing a cultural object from a relatively recent period, with no clues
on how to analyze it, preserved in any variety of Japonic, is a strange
phenomenon. Thus, it is likely that it is not a native compound, but a
loanword. The appearance of kuruma in Japan only in the Asuka or Nara
periods also is highly suggestive of a loanword. The only places it could
have come from would be either China, or, more likely, Korea. As far as I
can tell, no plausible sources can be identified immediately. However, with
some degree of speculation, Whitman’s etymology may ultimately work
for a loanword. Let us posit that a variety of Old Korean from which
Western Old Japanese borrowed kuruma had already lenited intervocalic
*-p- to zero in *kwùpwùl- ‘to roll’, pretty much as in Modern Korean. The
ending -ma in kuruma may be then explained as a reflex of a cognate of the
Middle Korean nominalizer -m. Therefore, this variety of Old Korean
might have a word like *kurum ‘cart (= one which rolls)’, but it also
remains unclear under this scenario why Western Old Japanese added -a as
an echo-vowel, and not -u. Ultimately, this scheme has too many
speculations that cannot be proven, and I prefer to reject the etymology
altogether, although it may be possible that a loanword relationship exists
between Korean *kwùpwùl- ‘to roll’ and WOJ kuruma ‘cart’.
(173) (R) MK -kí (deverbal nominalizing suffix) ~ OJ -kî (past tense
suffix for verbs) < PJK *-ki (Whitman 1985: 228-229) ― discussed and
rejected above (see 2.3.2.7, especially footnote 40).
(174) (L) MK kíc ‘share, division’; ‘collar, lapel (of clothing)’ ~ OJ
kîNta LL ‘dividing point, seam, crease’; ‘counter for pieces of cloth’ < PJK
*kija (Whitman 1985: 229). Phonetically the comparison seems to be
 Lexical Comparisons———159

impeccable, but there are other problems. First, there are no phonetic
attestations of kîNta in Old Japanese; the first one appears only in the
Wamyōshō (931-938), where the word is written as 岐多 (WMS V: 26a),
which might be /kîNta/, but is more likely to reflect /kîta/. Because there
are no phonetic spellings attested before the Wamyōshō, the word might
also be OJ kï[N]ta with an otsu-rui /ï/. This would present a problem for
the etymology. Since the quality of the vowel in the first syllable is not
known, I will write it as WOJ ki[N]ta. Second, the source for the meanings
‘seam, crease’ that Whitman cites is unclear to me: Omodaka et al. give
only ‘dividing line, divide; counter for pieces of cloth and plots of rice
fields’ (JDB 1967: 241-242), and the examples they provide confirm their
glossing. Third, and most important, WOJ ki[N]ta is not attested either in
Eastern Old Japanese or in Ryukyuan. Due to its limited distribution, it is
likely to be an old loanword from Korean. Both Nam Kwangwu and Yu
Changton treat MK kíc ‘share’ and MK kíc ‘collar, lapel’ as two
homonymous words (Nam 1997: 219; 144 LCT 1987: 120). I believe they
are right: I fail to see the connection between the two. It can be further
seen that WOJ ki[N]ta is likely to be a loanword from Korean *kica
‘share’ for semantic reasons as well: ‘plot of land’ is certainly semantically
narrower than ‘share’ in general.
(175) (R) MK :kil- ‘long‘, kìlúy ‘length’ ~ OJ kî ‘unit of length, inch’ <
PJK *kiro (Whitman 1985: 229). Whitman adds: “The MK nominal form
is very likely original, since the contraction of the second syllable in PJK
*kiro would account for the MK accent. OJ also has kî- in the meaning of
‘far, distant’ in compounds, often anomalously interpreted as the
continuative form of the verb ku ‘come’: kîpanare- ‘go far away (from
home or point of origin)’ < kî + panare- ‘separate, part’; kîpênar- ‘be far
separated’ < kî + pênar- ‘be separated’” (1985: 229). I believe that
Whitman’s treatment of the Middle Korean data is correct, but I find
several problems on the Japanese side. First, there is a semantic problem.
Omodaka et al. indicate that originally OJ kî designated a measure of
length equal to the length of a joint on the index finger (JDB 1967: 236). I
have grave doubts that this unit of length could possibly have any
associations with ‘length’, let alone with ‘far away’: ‘going far away to the
extent of an index finger joint’ is obviously an oxymoron. Second, glossing
OJ kîpênar- as ‘be far separated’, stretches the limits of credibility. There
are only three poems in the Man’yōshū where the word is attested in the
same context:

asi-pîkï n-ö yama kîpênar-i-te

foot-low DV-ATTR mountain ?-INF-SUB
doing X [with?/to?/on?/in?] low-base mountains (MYS IV: 670; XVII:
3969, 3981)

144
Nam Kwangwu cites both words incorrectly as having L pitch.
160———Koreo-Japonica

While Omodaka et al. speculate that a possible meaning can be ‘to be

separated’ (but without any connotation of ‘far’), they still consider it to be
unclear (JDB 1967: 245). WOJ kîpanare- ‘to be separated from a place
where one was originally’ is attested only twice in the Western Old
Japanese corpus: 145

Nara-wo k-î-panare ama-N-sakar-u pîna-ni ar-e-Ntö

Nara-ACC come-INF-be separated(INF) heaven-LOC-be separated
village-LOC exist-EV-CONC
although [I] am in a village that is far from Heaven, having left Nara
behind (MYS XVII: 4008)

iya töpo n-i kuni-wo k-î-panare iya taka n-i yama-wo kôye-suNkiy
oh distant DV-INF province-ACC come-INF-be separated(INF) oh high
DV-INF mountain-ACC cross(INF)-pass(INF)
oh, [I] left [my home] province far behind; oh, [I] crossed high mountains
(MYS XX: 4398)
The contexts in both poems support the traditional semantic
interpretation of WOJ kîpanare- as ‘to leave behind one’s original place, to
be separated from a place where one was originally’, but there is one
problem with the traditional analysis: the infinitive form k-î ‘come-INF’ +
panare- ‘be separated’. For the movement of a speaker from an original
place to a new place where he is located at the moment of speech, we
would expect *panare-k-î- (first leave behind, and then come), not the
attested k-î-panare (first come, and then leave behind), which does not
make any sense. If this is what Whitman meant by ‘anomalous
interpretation’, then his criticism is certainly justified, but for the reasons
indicated above, his own proposal to view kî- in kîpanare- as ‘unit of
length, inch’ is not acceptable semantically. It seems safer to assume that
kî in kî-panare- is an element with an unclear meaning. This leaves us with
WOJ kî ‘measure of length equal to the length of an index finger joint’,
which has no cognates in Eastern Old Japanese and Ryukyuan, and
therefore could qualify only as a loan from Korean. However, there are
two problems even with the loanword solution. First, WOJ kî refers to a
very specific unit of length, while MK kìlúy is length in general. Under
these circumstances, WOJ kî can be etymologically just about anything, cf.
English inch < Old English ynce ‘twelfth part of a foot’, related to Old
High German unza, Gothic unkja, and Latin uncia ‘twelfth part’ (Onions
1978: 468). Second, the borrowing of a word similar to MK kìlúy would
result in WOJ *kï with the otsu-rui vowel /ï/, not kî with the kō-rui vowel
/î/. Finally, even if these two obstacles are removed, the etymology will
still rest on Whitman’s medial *-r- loss law, which cannot be
independently corroborated in this case. In the light of all these problems, I
reject this etymology.
145
Both poems below were composed by members of the Opotomo clan. MYS XVII: 4008
belongs to Opotomo-no Ikenusi, and MYS XX: 4398 to Opotomo-no Yakamochi.
 Lexical Comparisons———161

(176) (R) MK kìph- ‘deep’ ~ OJ kîpa LH 146 ‘edge, extremity’, kîpam-

‘take to the outermost limit’ 147 < PJK *ki:po- (Whitman 1985: 229). In
addition to the less than perfect semantics, there are more problems on the
Korean side. First, we do not know whether MK kìph- ‘deep’ goes back to
PK *kipuk- or PK *kikup-. In the case of PK *kikup-, the comparison
would not be possible, and the etymology loses half its credibility. Second,
even in the case of PK *kipuk-, the vowels in the second syllables of
Japanese and Korean words do not correspond regularly to each other.
Therefore, I prefer to reject this etymology.
Etymologies (177)-(180) (Whitman 1985: 230) are all rejected, since
they involve reconstruction of PJK *g- on the basis of a ‘correspondence’
of MK n- to OJ k-. Refer to Martin (1991: 273) and Vovin (1993b: 339-
340) for critiques of this pseudo-correspondence.

3.2.6 *c-
(181) (R) MK càlí ‘seat, place, location’ ~ OJ -te nominal suffix
expressing place or direction in omote ‘front’ < omo ‘face’ + -te, patate
‘boundary’ < pata ‘edge’ + -te < PJK *car (+ -i) (Whitman 1985: 230).
Whitman has already proposed a different etymology for OJ -te: MK tóy
‘place’, which I accepted as a Korean loan in Western Old Japanese (see
[63] above). Since one and the same Japanese morpheme cannot have two
different Korean etymologies, this one is rejected.
(182) (L) MK cás ‘fort, castle’ ~ OJ sasi HL (or LL) ‘id’. < PJK *cas(+
-i). Whitman adds that is “an early loan, OJ sasi appears associated with
Korean locations in Nihonshoki” (1985: 230). It must be added that OJ sasi
is found only in Western Old Japanese, and since it is indeed a loan, there
is no need to reconstruct a Proto-Japanese-Korean form.
(183) (C)/(R) MK cólà- 148 ‘suffices’ ~ OJ tar- ‘id’. < PJK *córà-
(Whitman 1985: 230). 149 Although there is no cognate in Eastern Old
Japanese, there are attestations in Ryukyuan, including Southern Ryukyuan:
Shuri tayuN (RGJ 1976: 506), 150 Yaeyama taruN (Miyara 1980: 368),
Psara taro: (Shimoji 1979: 141). There are two problems, though. First,
under Whitman’s vowel correspondences rules, OJ /a/ corresponds to MK
/o/ only in Old Japanese long syllables (Whitman 1985: 129), but OJ tar-
belongs to accent type A (Martin 1987: 764); therefore, the first vowel in
146
Whitman gives an incorrect accentuation: both MJ kifa LL and PJ *kipa 2.3 indicate that
it should be the LL (2.3) accent class.
147
This comparison is also found in Martin (1966), #61.
148
There is a discrepancy in accentuation provided by different dictionaries: cólà- (LCT
1987: 633), còlà- (Nam 1997: 1287), cólá- ~ cólà- (Hankul hakhoy 1999.2: 5358). The data
on accentuation from fifteenth-century texts: cólà- (Sekpo VI: 11b; Twusi cho XVI: 56,
XXV: 17), cólá- (Twusi cho XXV: 25b, Samkang/Hyoca 26) demonstrate that there
apparently was a variation between HH and HL patterns, thus Hankul hakhoy’s
presentation is a correct one. I thank Ross King for providing me accentual data from the
Samkang Hayngsiltwo and the first edition of volumes XVI and XXV of the Twusi enhay,
which I do not have at my disposal.
149
This comparison is also found in Martin (1966), #229.
150
There is also Shuri tariyuN (RGJ 1976: 513), which looks like an early loan from
mainland Japanese tariru.
162———Koreo-Japonica

OJ tar- could only be short in Proto-Japonic. Second, unless one accepts

the Martin-Unger reconstruction of consonantal verbs as having *[C]VCa-
structure in Proto-Japonic, it is necessary to account for a final vowel in
Korean that has no correspondence in Japanese. Therefore, this etymology
is only marginally acceptable.
(184) (R) MK còlò ~ còl.- ~ còl.l- < *còlòk ‘handle’ ~ OJ tuka LL ‘id’.
< PJK *coloko (Whitman 1985: 230). Whitman’s còl.- ~ còl.l- should
really be rewritten as còlG-. There are three problems with this etymology.
First, since OJ tuka is LL, the vowel in the first syllable in Proto-Japonic
should be long. According to Whitman’s vowel correspondences, one
would expect OJ /a/ here, not /u/ (Whitman 1985: 129), cf. the opposite
case in (183) above. Second, the etymology rests on the assumption of *-r-
loss in Old Japanese, which cannot be verified on the basis of independent
evidence. Third, even if the first two objections are rejected, the final
vowel -a in OJ tuka has no correspondence in Korean. Therefore, I reject
this etymology.
(185) (R) MK cómk- ‘sinks, is immersed’ ~ OJ tuk- ‘is immersed, gets
wet’ < PJK *comok- (Whitman 1985: 230). There are two problems with
this comparison. First, OJ tuk- means ‘to soak in’ rather than ‘to be
immersed’ (JDB 1967: 464). Second, Middle Korean shows a variation
between the forms còmò- ~ còm- ‘is immersed’ without any traces of *-k-
in the root, and còmkì- ~ cómk- ‘id’. with a velar. Such a variation would
imply that the velar in the second variant is secondary. Third, and most
important, this comparison is based on the assumption that Proto-Japonic
intervocalic *-m- is lost in Old Japanese after PJ *-u-, but this assumption
largely rests on external comparative evidence from Korean based on five
other examples; see (136) above and (186), (212), (293), and (330) below,
with two more dubious internal examples (Whitman 1985: 25). Therefore,
the logic in establishing the correspondence of MK -m- to OJ -Ø- appears
to be circular and unwarranted by the internal Japonic evidence. Note that
while the Old Japanese sequence /tuk/ corresponds to MK /comk/ in this
example, it also corresponds to PK *colok in (183), which further
diminishes the credibility of both comparisons. I reject this comparison on
the basis of irregularity of correspondences.
(186) (R) MK cómk- ‘is locked, is joined’ ~ OJ tuk- ‘is attached’ < PJK
*comok- (Whitman 1985: 230). Like (185) above, this comparison is based
on the assumption that Proto-Japonic intervocalic *-m- is lost in Old
Japanese after PJ *-u-, but this assumption largely rests on external
comparative evidence from Korean based on four examples (Whitman
1985: 131, 139, 141, 143), with two dubious internal examples (Whitman
1985: 25). The same objections as in (185) apply here; therefore, this
comparison is rejected.
(187) (R) MK cwòch- ‘follows it, keeps step with it, obeys it’ ~ OJ
tutuk- ‘follows it, continues it’ < PJK *cucuk- (Whitman 1985: 230). First,
a minor correction is needed: WOJ tu[N]tuk- is an intransitive verb; the
transitive one is WOJ tu[N]tukë- (JDB 1967: 469-470). The first two
syllables of WOJ tutuk- are spelled in the texts as 都都 or 都追, which
 Lexical Comparisons———163

would imply [tutu], but the word is not attested in the Nihonshoki kayō, the
only text that consistently differentiates between voiceless and pre-
nasalized voiced stops; therefore, those spellings can reflect [tuNtu] as well.
The word is not attested in Eastern Old Japanese, but the majority of
Ryukyuan attestations, with the exception of Koniya t’ïtïkijum ‘to continue
(tr.)’, point to PR and PJ *tuntuk- with *-nt- rather than *-t-: Nase
tśidʒikirjuN, Hentona tʒidZiki:N, Iejima sizikijuN, Psara tsïdzïkiï, Ikema
tsïdzïki:, Ishigaki tsïdzïkiruN, Hateruma śimiśikiruN, etc. ‘to continue’
(Hirayama 1966: 382). It is further important to note that Ryukyuan forms
clearly indicate PJ *tuntuk- with *-u-, not PJ *tontok- with *-o-, which
creates a problem for the Korean etymology, as at the present stage of our
knowledge we would expect PJ *-o- corresponding regularly to MK
/wo/. 151 In addition, Hateruma śimiśikiruN is likely to indicate that the
Proto-Ryukyuan and Proto-Japonic form was something like *tumutuk- or
*tumituk-, which rules out the comparison with Middle Korean cwòch-,
since we have internal Korean evidence that allows us to account for the
loss of a nasal before /c/ in Middle Korean, but not for the whole /mV/
syllable. Finally, there is a problem on the Korean side, too: MK cwòch-
can go back either to PK *cwocok- or to PK *cwokoc-. The second
possibility would rule out the comparison completely, so the comparison
loses half of its credibility. Too many problems beset this etymology to
make it acceptable.
(188) (L) MK cèk ‘time’ ~ OJ tökî LL ‘time’ < PJK *cek (Whitman
1985: 230). 152 This etymology looks impeccable at first glance, but there is
a problem: according to Whitman’s vowel correspondences, OJ /ö/
corresponds to MK /e/ either in the final position or in words with /ö/ in
another syllable. Neither of these two conditions applies here, so we have
an irregular correspondence. If we assume that Whitman’s Proto-Japanese-
Korean reconstruction *cek is correct, it is unclear how Old Japanese
acquired its final /-i/. Furthermore, MK cèk belongs to a small class of
nouns that have L pitch. This suggests that in all likelihood MK cèk is a
result of an apocope: MK cèk < PK *cekV LH, with V = I, as witnessed by
OJ tökî. However, PK *ceki or *cokV would result in MK *cèh, not cèk,
the latter being an expected development from PK *cenki, unless the
postvocalic nasal sonorant loss in Korean occurred after apocope took
place. Nonetheless, ordering the rules as (1) apocope and (2) postvocalic
sonorant loss is highly unlikely, because Early Middle Korean materials
offer evidence for the pre-apocope stage of the language.

Chart 30: Apocope in Middle Korean and Early Middle Korean

Gloss EMK MK
‘rain’ piWi 153 pí

151
In Whitman’s system of vocalic correspondences, OJ /u/ corresponds to both MK /wo/
and /wu/. However, nowadays we know that WOJ /u/ has two Proto-Japonic sources: *u
and *o.
152
This comparison is also found in Martin (1966), #242.
153
Kyeylim yusa, #7.
164———Koreo-Japonica

‘day’ nacay 154 nác

‘earth, ground’ holki 155 hòlk

However, there is no evidence in Early Middle Korean for forms that still
include postvocalic nasal sonorants, so the process of simplification of
*NC clusters must predate apocope. Thus, MK cèk < PK *cenkV or *cenki
LH, yielding one more irregular correspondence in this etymology: PK
*-nk- : PJ *-k-. Therefore, due to irregularities in the correspondences, OJ
tökî is likely to be an old loan from Korean, probably borrowed during the
time of mutual coexistence on the Korean peninsula, since PJ *töki is well
attested within Japonic (in both Eastern Old Japanese and Ryukyuan).
(189) (R) MK cyéc ‘breast, milk’ ~ OJ ti ‘id’. < PJK *ce (+ -i)
(Whitman 1985: 230). 156 Whitman notes that Jpn. titi is a reduplication,
which is probably true. The reconstruction of PJK *ce is doubtful, because
it does not account for the final -c in Middle Korean (there is no internal
evidence that this word is from *cyecye), and the reconstruction of PJK *e
on the basis of the correspondence of MK /ye/ to OJ /i/ is also left
unexplained. But most important is the fact that this is certainly a nursery
word that cannot be used for proving genetic relationships, cf. English tits,
Russian tit’ki, German Tzitzen, among others.
(190) (R) MK cyècéy ‘market, fair’ ~ OJ iti LH ‘id’. < PJK *jicey
(Whitman 1985: 231). 157 Whitman comments: “Assuming a voiced PJK
initial, the expected form is OJ *yoti or *yati if we assume PJK *jyecey.
The MK vowel must then result from breaking of *i in K[orean]. Note that
OJ final /ti/ ([tuy]) confirms the reconstruction of the final syllable.
Raising and fronting of the mid vowel in the first syllable may have
occurred after merger of pre-OJ *tuy and ti, indistinct at the OJ stage”
(1985: 231). There are several problems with this etymology. First, even
the assumption that the pre-Old Japanese form was *yiti, let alone non-
existent OJ *yati or *yoti, is highly speculative: we simply have no internal
evidence for the *y- initial rather than an absence of any initial consonant.
Second, the assumption that OJ /ti/ is from *tï is also speculative. Although
Whitman seems to believe that any OJ /ti/ is from *tï (1985: 38), this
seems to be contradicted by his statement above about the merger of pre-
OJ *tï and ti. This assumption is also problematic in the light of Eastern
Old Japanese palatalization /ti/ => /si/, e.g., WOJ iNtuti ‘where’ ~ EOJ
iNtusi ‘id’. 158 The third problem is on the Korean side: although MK

154
Kyeylim yusa, #42.
155
Kyeylim yusa, #53.
156
This comparison is also found in Martin (1966), #25.
157
This etymology should really have been moved to the PJK *j- list.
158
Whitman believes that this palatalization occurred after PJ *uy > EOJ /i/. However,
Eastern Old Japanese presents evidence for the palatalization /ti/ => /si/ in an environment
where one cannot suspect the development of *tuy > *ti > si, namely, the palatalization of
final -t in consonantal verbs to -s in front of the infinitive suffix -i. For example, WOJ tat-i
‘stand and’ ~ EOJ tas-i ‘id’. (MYS XIV: 3395; XX: 4372, 4423), WOJ möt-i ‘hold and’ ~
EOJ mos-i (MYS XX: 4420), WOJ panat-i ‘let go and’ ~ EOJ panas-i (MYS XIV: 3420). I
 Lexical Comparisons———165

cyècéy ‘market’ is attested, it seems to be the later form of the word, as it is

attested once in 1475 (Nayhwun III: 13), and all other attestations are from
the sixteenth and seventeenth centuries (Nam 1997: 1220; LCT 1987: 656).
Meanwhile, the earlier form is MK cyècáy ‘market’, not cited by Whitman.
It is amply attested in fifteenth-century texts, starting from YP 6. It is quite
apparent that MK cyècáy => cyècéy under the pressure of recently
developed vowel harmony (Martin 2000). Thus, under Whitman’s vocalic
correspondences, MK -ay cannot possibly correspond to OJ -i, even if the
latter were from *ui (Whitman 1985: 129). I reject this etymology due to
its irregularities in correspondences and speculative reconstruction. WOJ
iti does not have any cognates attested in Eastern Old Japanese or
Ryukyuan, but I think that the phonetic problems with this etymology are
too great to recognize it even as a loan from Korean into Central Japanese.
(191) (R) MK cùl- ‘muddy, mushy’ ~ MJ doro LL ‘mud’ < PJK *cör-
(Whitman 1985: 231). Whitman comments: “MJ doro LL is not attested in
Old Japanese, although töwo- (in onomatopoetic expressions) ‘soft, gentle,
fluid (movement)’ is very likely related. There is good evidence that the
voicing in the Middle Japanese initial consonant is secondary: note MJ
toroke- ‘softens, melts’” (1985: 231). First of all, I fail to understand how
WOJ töwowo n-i, and possibly other forms cited in JDB 1967: 511, 159
which refer to an onomatopoetic description of bending, as far as I can tell
from the textual examples that Omodaka et al. cite, can be related to MJ
doro ‘mud’ either phonetically (there is no correspondence of WOJ -w- to
MJ -r-), or semantically (bending mud would be quite a difficult
enterprise). Second, the earliest attestation of MJ toroke- is in the Konjaku
monogatari (IKJ 1990: 936), which is practically on the borderline with
Early Modern Japanese, so we only have a very late Middle Japanese
attestation. The semantic connection with ‘mud’ is not completely
impossible, but remains tentative in the best-case scenario, so using it to
prove the secondary nature of d- in MJ doro is speculative and not reliable.
Third, MJ doro is attested only in very late Middle Japanese: it appears for
the first time as doro LL in the Myōgoki, published in 1270 (Martin 1987:
391). Given the size of the Middle Japanese corpus, the fact that doro is
not attested in any other texts strongly suggests it is a late innovation.
However, the word is attested throughout the Ryukyus, and the forms in
various Ryukyuan dialects strongly suggest that it is not a recent loan from
Early Modern Japanese. Thus, PR *doro has to be reconstructed (Thorpe
1983: 309), and it confirms the original nature of voiced d- in EMJ doro,
contrary to Whitman. Martin reconstructs PJ *ntoro ‘mud’ (1987: 391). I
suggest that the reconstruction be revised as PJ *nVtoro or *mVtoro, since
initial consonant clusters are prohibited by Japanese phonotactics. In any
case, the comparison with MK cùl- does not seem feasible, simply because
we have no internal evidence for the morphological analysis of

believe that the traditional point of view that OJ /ti/ resulted from the merger of pre-OJ *tî
and *tï (< *tui, *töi) is correct.
159
Other forms, such as töwoyör-, töwowo-töwowo, n-i töworap- appear to be either unique
attestations or forms with unclear meanings (JDB 1967: 511).
166———Koreo-Japonica

*nVtoro/*mVtoro; nor do we know what the constituents of this (likely)

compound are. Thus, I reject this etymology on morphological grounds.
(192) (R) MK cùl- ~ cùl.é- ~ cùl.l- < *cùlók- ‘strikes home, crosses
over, traverses’ ~ OJ toNkë- ‘completes it, carries it through’ < PJK
*cörök- (Whitman 1985: 231). WOJ toNkë- is not attested phonetically in
Old Japanese texts, 160 and its meaning is rather ‘to accomplish, to achieve,
to carry out’. This seems to be somewhat far-fetched vis-à-vis its Middle
Korean counterpart, which does not mean ‘to cross over, to traverse’,
contrary to Whitman. Actually, MK cùlù- ~ cùlG- means ‘to cut off, to die
prematurely, to make a short-cut’. 161 Cf. the following examples:

mot atol-i culG-e eps-un-i

elder son-NOM cut off-INF be non-existent-ATTR/REAL-NML
eldest son prematurely died (Welin se 14a)
culG-e w-a
making short-cut-INF come-INF
[he] came making a short-cut (Yukco I: 106) 162

The semantics alone invalidate this comparison, but it is also necessary

to note that PK *-luk- cannot possibly correspond to PJ *-nk-. Therefore, I
reject this comparison on semantic and phonetic grounds.
(193) (R) MK cwùk- ‘to die’ ~ OJ tukï- ‘is used up, is exhausted, is
expended’ < PJK *cük- (Whitman 1985: 231). Although phonetically the
comparison looks possible, I am somewhat reluctant to accept it, because
WOJ tukï- refers to exhausted words, feelings, or provisions, but not to the
cessation of life. In addition, the word seems to be confined to Western
Old Japanese, with no attestations in Eastern Old Japanese, and with a
single attestation in Ryukyuan: Shuri çikusjuN (RGJ 1976: 152), which is a
counterpart of the Western Old Japanese and Middle Japanese transitive
form tukus-. In the best-case scenario, this is a loanword, but I still have
doubts about the semantics.
(194) (L) MK cwúl ‘rope’ ~ OJ tuna LL ‘id.’, tura HL or tura HH ‘line,
string‘, turu HL ‘string’ < PJK *cür (Whitman 1985: 231). OJ tuna has to
be ruled out from the start, because MK -l- does not correspond to OJ -n-.
It appears that there is WOJ tura ‘line’ in the sense of ‘row’ (not ‘rope’),
which has to be excluded for semantic reasons, and in addition there are
WOJ turu ‘bowstring, string of a musical instrument’ and EOJ tura
‘bowstring’ (JDB 1967: 479-480). 163 It seems that Middle Japanese has

160
Therefore, we do not know whether the vowel in the first syllable was WOJ /ö/ or /ô/: in
the second case the comparison will be ruled out, so the etymology has only 50 percent
credibility before anything else is taken into consideration.
161
The Proto-Korean reconstruction should be rather *cùlúk-, not *cùlók-, as Whitman
presented it.
162
Cited according to LCT 1987: 678. I do not have access to this text.
163
Omodaka et al. believe that tura ‘bowstring’ also appears in Western Old Japanese (MYS
VII: 1329; JDB 1967: 479), but since the word in question is written in this poem
 Lexical Comparisons———167

only turu ‘bowstring, string of a musical instrument’ while tura means

exclusively ‘row’ (KKJ 1969: 591-592). There are scattered attestations in
Ryukyuan, all meaning ‘bowstring, musical instrument string’: Nakijin
ciruu (Nakasone 1983: 286), Shuri çiru (RGJ 1976: 163), Iejima siru
(Oshio 1999: 180), Yaeyama tsïru (Miyara 1980: 377). This probably
means that we can reconstruct PJ *turu ‘bowstring, string of a musical
instrument’, but its exact relationship to MK cwúl ‘rope’ remains unclear.
The narrowed meaning in Japonic in contrast to Korean, and the fact that it
is a cultural item, strongly suggest that this word represents an early loan
from Korean into Japonic. The fact that most Japonic forms likely have an
echo-vowel -u further strengthens this possibility. One more argument in
favor of a loan can be made on the basis of the fact that there is OJ napa
‘rope’, which seems to be native Japonic, so WOJ turu represents a kind of
doublet with restricted semantics. I will treat this comparison as an early
borrowing.
(195) (R) MK chwúm ‘spit’ ~ OJ tuNpa ‘id’. 164 < PJK *cümba,
Whitman notes: “K[orean] aspiration is problematic” (1985: 231). In
addition, there is the alternative Middle Korean form chúm ‘spit, saliva’,
attested once in Samkang cwung 16. Although it is a hapax legomenon, it
may possibly indicate that MK chwúm is to be analyzed as the nominalized
modulator form ch-wú-m of the verb *chu-, which did not survive in the
texts. I think that the Korean aspiration is not just problematic, but rules
out this comparison as a cognate, because MK chwúm < PK *cukwum or
*kucwum. Neither of those is comparable with OJ tuNpa, which, I believe,
comes close to being a ghost. MJ tufa ‘saliva’ is attested only once in the
Shinsenjikyō (IKJ 1990: 863), an obvious truncation of *tupaki, which is
not phonetically attested in Old Japanese. However, MJ tufaki is amply
attested (JDB 1967: 473; IKJ 1990: 863), demonstrating that EMJ and MdJ
tubaki underwent secondary voicing. Since there is also late MJ and EMJ
tu ‘saliva’ (IKJ 1990: 843), the internal etymology of the Japanese word
becomes quite clear: it is tu ‘saliva’ + pak- ‘to spit’ + nominalizer -i. The
Ryukyuan forms that are not loans from mainland Japanese tubaki show
the reduplicated form of tu: Kamezu tsïdu, Iejima tutupe, Ikema and
Ishigaki tsïtsï, etc. (Hirayama 1966: 374). Thus, the comparison can be
rejected on phonetic and morphological grounds.
(196) This number is absent from Whitman’s list.
(197) (R) MK -cil ‘act of doing’ ~ OJ sirö HH ‘object used for
something, place for doing something’ (Whitman 1985: 231). No Proto-
Japanese-Korean reconstruction is provided, but Whitman adds the
comment: “Both MK and OJ morphemes are nominal suffixes. MK -cil
appears in compounds such as panolcil 165 ‘sewing’ < panol ‘needle’ + -cil.
Compounds formed by the OJ suffix are usually concrete: aNsirö ‘netting,

semantographically as 弦, it may represent turu, not tura. EOJ tura is a hapax legomenon,
attested only in MYS XIV: 3437.
164
This comparison is also found in Martin (1966), #216.
165
This is not attested in Middle Korean to the best of my knowledge. There is EMdK
panocil ‘sewing’ (with a loss of -l- in panol), which is attested in Yek.epo 41.
168———Koreo-Japonica

fish weir’ < amî ‘net’ + -sirö” (1985: 231). One obvious problem is that
these suffixes are very different functionally. In addition, I believe that
Whitman’s definition of OJ -sirö as ‘object used for something, place for
doing something’ is imprecise. It really means ‘substitute, something used
instead of or like something else’. Thus, for example, the accurate
definition of WOJ aNsirö < *amî-sirö is ‘fish weir made from bamboo and
wood and used like a net (WOJ amî) to catch fish’ (JDB 1967: 23).
Therefore, this specific meaning of WOJ -sirö is not really compatible with
MK -cil ‘act of doing’. Another problem is that the survival of this
etymology depends on the assumption that OJ si- < *ti-. This is an
assumption which is usually made on the basis of observation that there are
many more si- in Old Japanese than ti-. This is true, but it is impossible to
prove the shift *ti- > si- just on the basis of this statistical observation. We
have already seen above that the second part of this assumption, which
explains all OJ /ti/ as derived from *tuy, does not work (see [190] above).
Therefore, in the light of these semantic and phonetic problems, I reject
this etymology.
(198) (R) MK :cyek- ~ :cyak- ‘little’ ~ OJ sukô(-sikî) ‘id.’, 166 sukôsi
‘little, few’ (adv.), sukunasi ‘id’. < PJK *cyokor- (Whitman 1985: 231-
232). Whitman adds: “The MK stem vowels may simply be the usual
adjectival ablaut variants, but this comparison assumes that the original
stem vowel is *o, assimilated to /a/, /e/, or /wo/ in MK after /y/. The MK
long vowel may indicate an original second syllable, which is also
supported by the second vowel /wo/ in OJ sukô, and the relation between
these two forms and OJ sukunasi. We may reconstruct pre-OJ *sukur- and
hypothesize that the alternants sukunasi and sukô- result from suffixation
of the adjectival formant -si at an early (sukunasi) and later (sukô-sikî)
period. Note that the Old Japanese adjective sukô-sikî shows the regular
inflection of the -sikî class of adjectives, usually thought to be of relatively
late derivation, while sukunasi (adverb) is a frozen form. Attachment of -si
to *sukur- in the case of sukunasi shows the same behavior of stem-final
*[-]r in compounds that we observed with OJ mî- < PJK *möl ‘water’:
minakuti, minamoto. Attachment of -si(kî) in the case of sukô-sikî occurs
prior to vowel epenthesis after a stem-final sonorant and medial *-r- loss,
giving *sukur- > *sukura- > *sukua- > *sukô- + -si(kî)” (Whitman 1985:
231-232). I believe that this comparison has numerous problems, and the
discussion cited above represents an attempt to save an etymology in a
unique way, which is methodologically unacceptable. First of all, even if
we take the data presented at face value, there is the problem of the
correspondence of MK c- to OJ s-. This problem was already addressed
above (see [190] and [197]), but here it is further aggravated by the fact
that in this case there are no conditions for OJ /s-/ to palatalize: unlike the
etymology discussed in (197), the following vowel is /u/, not /i/. Second,
Whitman’s assumption that the original stem vowel was *o can be rejected
on the basis of Ceycwuto cwok- ‘to be little’ (Ceycwu pangen yenkwuhoy

166
This comparison is also found in Martin (1966), #128.
 Lexical Comparisons———169

1995: 503), which does not support Whitman’s *cyokV-, since Ceycwuto
is the only dialect that preserves original *o as /o/, even in combination
with preceding /y/. Furthermore, non-leniting MK -k- surely goes back to
PK *-nk-, which does not correspond to OJ -k-. Thus, we have to
reconstruct Proto-Korean alternants *cyènkú- ~ *cyànkó- with a usual
adjectival umlaut. However, in this case neither MK /e/ nor MK /a/
regularly corresponds to OJ /u/. This is the third irregular correspondence
we are dealing with in this etymology. Finally, if we are dealing with the
Proto-Korean variant *cyèkú-, PK *-u should correspond to OJ /ö/, not to
/ô/ or /u/, and that leaves us with no regular correspondences at all. All of
these irregularities are sufficient to cause rejection of this etymology.

3.2.7 *j-
(199) (L) MK cyèmúl-, cyémkúl 167 ‘(day) comes to a close, gets dark’ ~
OJ yamï ‘darkness’ < PJK *jyemör (Whitman 1985: 232). Whitman notes:
“Final /ï/ in the Old Japanese form is the expected outcome for the
continuative (nominalized) form of a thematic verb stem in /ö/. Old
Japanese attests no verb *yamö- or *yamï-; however, yamë- ‘cease, end’ is
quite possibly related” (1985: 232). Semantically, this seems to be a stretch.
Rather, it is possible that OJ yamï should not be separated from WOJ
yömö- in yömö-nö kuni and MJ yomi ‘Hades, land of the dead‘. OJ yamï
belongs to accent class 2.3, and WOJ yömö- likely belongs to the same
accent class (Martin 1987: 573, 576). Yamï ‘darkness’ is attested once in
the Eastern Old Japanese corpus (MYS XX: 4436), but this poem is written
in pure Western Old Japanese, so this single attestation probably should be
discarded. There are scattered attestations of yami in the Northern and
Central Ryukyus: Izena, Kijoka, Arumi, Tomigusuku yami, Kumejima
ye:mi (Uchima and Arakaki 2000: 458), Shuri ’yami (RGJ 1976: 277). The
isolated Southern Ryukyuan forms, Hateruma yoiN, Ishigaki yo:N (Miyara
1981: 342), Tonoshiro yoN (Uchima and Arakaki 2000: 458) are probably
not related to yami due to a discrepancy in the first syllable vowels. This
distribution suggests that both WOJ yamï and yömö could only be loans
from Korean, probably from different dialects. In addition, the infinitive of
MK cyèmúl- is cyemulG-e, indicating PK *cyemuluk(u)-. However, MK
cyèmkùl- indicates PK *cyem(u)kul(u)-. One of these variants was
probably derived by metathesis from the other; unfortunately, we do not
know the direction of this change, nor do we have any internal Korean
evidence for the segmentation of these protoforms. Such evidence is
needed for a genetic comparison with Old Japanese. However, a loan from
a Korean form resembling the Middle Korean infinitive cyemulG-e is
certainly possible. This is further supported by the fact that only nominal
forms are found in Western Old Japanese. Thus, I will treat this etymology
as a loan from Korean into Central Japanese, which was in turn borrowed
by Northern and Central Ryukyuan.

167
Whitman’s cyémkúl- appears to be a typographical error for cyèmkùl- (LCT 1987: 658;
Nam 1997: 1227).
170———Koreo-Japonica

(200) (R) [Md]K ciley ‘lever, handspike’ 168 ~ OJ ye ‘handle, projection

for grasping’ < PJK *jire(+ -i) (Whitman 1985: 232). K ciley ‘lever,
handspike’ is not attested in Middle Korean, which immediately makes it
suspect. I am not aware of any dialectal attestations, either. Kim Minswu
suggested an internal etymology from MK tìlò- ~ tìlù- ‘to insert, to prick,
to stab’ (Kim 1997: 959), but this seems to be too distant semantically.
Intuitively I feel that MdK ciley must be some undetected loan, but nothing
except Chinese ji1li3 (機理) ‘mechanism’ comes to mind, and this also may
be far-fetched. The lack of an etymology may be a minor problem, but it
requires an explanation nonetheless. The problems on the Japonic side are
more significant. Theoretically, OJ ye can go back either to PJ *yai or to PJ
*yia. The latter fits better with MdK ciley, but it is likely that PJ *y- /___-i-
would be lost before the monophthongization *ia > /e/ took place, since
this loss preceded monophthongization. Also, positing PJ *yia contradicts
the correspondences proposed in (202) below, where MK /ci-/ corresponds
to OJ /i-/ < PJ *yi, and not to OJ /ye/. This leaves us with PJ *yai, but PJ
*-a- apparently cannot correspond regularly to MdK -i-. In addition, *-r-
loss in pre–Old Japanese is needed to make this etymology valid. However,
there is no internal independent evidence, so reconstructing PJ *yari would
be an ad hoc assumption. Finally, Korean /e/ does not regularly correspond
to OJ /i/, so the etymology should be rejected.
(201) This number is absent from Whitman’s list.
(202) (L) MK cìp ‘house’ ~ OJ ipê, ipî (Azuma), ipa (Azuma) 169 ‘id’. <
PJK *jip. Whitman comments: “The alternations in the OJ second syllable
suggest alternant outcomes of a compound, most likely *ip + ya ‘house,
building” (1985: 232). Some minor corrections to the data are in order
before I proceed to the discussion of this etymology. The word ipê ‘house’
is attested in both Western and Eastern Old Japanese, but its distribution in
the latter is interesting. EOJ ipê is attested in five Azuma texts (MYS XIV)
and in four Sakimori texts (MYS XX). However, out of five poems in MYS
XIV, it appears in only one poem (MYS XIV: 3423) with Eastern Old
Japanese features. The other four poems represent typical Western Old
Japanese texts (MYS XIV: 3481, 3532, 3354, 3542). The situation in the
Sakimori corpus is the opposite: ipê appears in three poems with Eastern
Old Japanese features (MYS XX: 4384, 4388, 4415), and in one poem that
is written in pure Western Old Japanese (MYS XX: 4347). Neither EOJ ipî
nor EOJ ipa appears in the Azuma corpus, which seems to be more
influenced by Western Old Japanese than is the Sakimori corpus;
occurrences are all confined to MYS XX: ipî is found in MYS XX: 4343
and ipa in MYS XX: 4416, 4423, 4427. In addition, there is also EOJ iparo
‘house’ found in MYS XX: 4406, 4419 and the compound ipa-Npîtö
‘people of the house’ in MYS XX: 4375. EOJ ipî, ipa, iparo, and ipa-Npîtö
are all found in texts that exhibit other significant Eastern Old Japanese
features. I believe that this peculiar distribution indicates that EOJ ipê is a
loan from Western Old Japanese. The main Eastern Old Japanese form is
168
Whitman did not present a gloss for MdK ciley.
169
This comparison is also found in Martin (1966), #113.
 Lexical Comparisons———171

undoubtedly ipa ‘house’, because EOJ ipî is a hapax legomenon. The

correspondence of WOJ /ê/ (< PJ *ia) to EOJ /a/ is attested in other cases,
most of them involving the morphological boundary between /i/ and /a/.
Examples are: MJ ayafuke ‘dangerous (evidential form)’ < *ayapu-kî-a ~
EOJ ayapoka ‘id.’ (MYS XIV: 3539), WOJ siNkëkêku ‘dense (of
vegetation, nominalized form)’ < *siNkë-kî-aku ~ EOJ siNkekaku ‘id’.
(MYS XIV: 3489), WOJ töpokê ‘distant (evidential form)’ < *töpö-kî-a ~
EOJ topoka ‘id’. (MYS XIV: 3473), WOJ okêre ‘put (progressive
evidential form)’ < *ok-î-ar-e ~ EOJ okare ‘id’. (MYS XIV: 3556), WOJ
kêru ‘wear (progressive attributive form)’ < *kî-ar-u ~ EOJ karu ‘id’.
(MYS XX: 4431), WOJ ipêru ‘say (progressive attributive form)’ < *ip-î-
ar-u ~ EOJ iparu ‘id’. (MYS XIV: 3512), WOJ nerapêri ‘have one’s eye on
(progressive final form)’ < *nerap-î-ar-i ~ EOJ nerapari ‘id’. (MYS XIV:
3529), WOJ omöpêri ‘think (progressive final form)’ < *ömöp-î-ar-i ~ EOJ
[o]mopari ‘id’. (MYS XIV: 3526), WOJ nöreru ‘tell (progressive
attributive form)’ < *nör-î-ar-u ~ EOJ noraro ‘id’. (MYS XIV: 3469), WOJ
pareru ‘bud (progressive attributive form)’ < *par-î-ar-u ~ EOJ pararo ‘id’.
(MYS XIV: 3546). There are, however, a couple of cases in addition to
WOJ ipê ~ EOJ ipa ‘house’ in which the existence of the morphological
boundary is not obvious: WOJ pê, directive case marker < *pia ~ EOJ pa,
id. (MYS XX: 4428), WOJ yeNta ‘branch’ < *yaiNta ~ EOJ yaNte ‘id’.
(MYS XIV: 3493). For the reasons outlined in (200) above, it seems that
this last example is different from others, because WOJ yeNta ‘branch’ can
go back only to PJ *yaiNta, but not to PJ *yianta, because the latter would
result in WOJ *eNta, not *yeNta. However, it is quite clear that all of the
Eastern Old Japanese examples listed above include the contraction *ia > a
(with the exception of ‘branch’ where the contraction *ai > a occurred),
while their Western Old Japanese counterparts all exhibit the
monophthongization *ia > ê (again with the exception of ‘branch’, where
the monophthongization *ai > *ë > e occurred). Therefore, Whitman’s
speculation that WOJ ipê ‘house’ resulted as a contraction of the
compound *ip+ya ‘house+house’ appears to be unfounded internally,
especially if we take into consideration the fact that Japanese had no
nominal roots ending in a consonant (with the possible exception of *-m or
*-y at some stage of its recorded or reconstructed history). Another
problem is that while cognates of WOJ ya ‘house’ are attested in both
Eastern Old Japanese and Ryukyuan, cognates of WOJ ipê ‘house’ are
found only in Eastern Old Japanese. This distribution strongly suggests
that WOJ ipê and EOJ ipa ~ ipî are loans from Korean. But this
explanation, of course, faces the same problem that was mentioned in
(190): we have no internal evidence for reconstructing PJN *yipia with
initial *y-. However, the distribution of the word in Japanese and the fact
that MK cìp ‘house’ has L pitch, indicating possible Proto-Korean
disyllabic *cipu LH, where *u stands for a reduced [-back] vowel,
strengthens the possibility that it is a loanword. Furthermore, isolated EOJ
ipî ‘house’ may also provide some evidence for a morphological breakup
of the tentative PJ *yipia as *yipi-a, although this is much more
172———Koreo-Japonica

speculative. Therefore, I treat this etymology as a Japanese loan from

Korean.
The comparisons involving numbers (181-195), (197-200), and (202)
represent etymologies involving Proto-Japanese-Korean initial *c- and *j-.
Whitman has presented 15 etymologies supporting PJK *c- and only 4
etymologies supporting PJK *j-. The ratio is somewhat similar to the ratio
of the aforementioned etymologies with *p- ~ *b- and *t ~ *d-. Although
these numbers look slightly better, they are weighted disproportionately in
favor of etymologies involving voiceless *c-. Even without my rejection of
all Proto-Japanese-Korean etymologies with *j- as supporting genetic
relationship, the distribution is obviously skewed. If my rejections are
accepted, once again, there is a gap: no reliable Korean etymologies for
Old Japanese words with initial y-. 170 The existence of this gap is further
reinforced by the fact that MK c-, which reflects an alleged PJK *j-, is
always found before a following /i/ or /y/.

3.2.8 *s-
(203) (R) MK sàks ‘wages, pay’ ~ OJ sakî ‘good fortune, prosperity’
(Whitman 1985: 232).171 There are two problems with MK sàks ‘wages,
pay’. To the best of my knowledge, this word appears only three times in
Middle Korean, twice on two sides of the same leaf in the first edition of
the Pak thongsa (1517), and once in the Hwungmwong cahwoy (1527). It
seems that in both examples in the Pak thongsa it means not just ‘wages,
pay’, but has a much more restricted meaning of ‘payment to the porter’:

tye sak pat-ol salom-a

that payment to the porter receive-ATTR/IRR person-VOC
Hey, you who will receive the payment to the porter! (PT I: 11a) 172

wuli ancik saks-ul hyeyGali-cye

we still payment to the porter-ACC count-VOL
We are still going to calculate payment to the porter (PT I: 11b) 173
The gloss in the Hwungmwong cahoy for sak pat-ol, lit. ‘receiving sak’
is clearly ‘be hired’ or ‘be rewarded’ (Hwungmwong II: 2a). 174 It seems
that out of the four other examples attested in Early Modern Korean texts,
170
Whitman also presented three etymologies with OJ y- reflecting PJK *d- and four
etymologies with OJ y- reflecting PJK *y- (190, 199-202 [there is no 201], 313-317). See
the discussions above and below under these numbers that reject these etymologies as
evidence of a genetic relationship.
171
No Proto-Japanese-Korean reconstruction is provided.
172
The Chinese phrase 那挑脚的 ‘[You], the one who carries loads!’ of which this Korean
phrase is a translation, makes it quite clear that indeed payment to the porter is meant, since
Chin. 挑脚 tiao1jiao3 means ‘to carry loads/luggage’.
173
Again, the Chinese phrase 商量脚錢着 ‘[We,] the merchants are calculating payment to
the porter’ makes it clear that the intended meaning is ‘payment to the porter’, because
Chinese 脚錢 jiao3qian2 means ‘payment to the porter’.
174
The Chinese equivalents are 雇 ‘to hire, to reward’ and 傭 ‘to hire, to be hired, to be
rewarded’.
 Lexical Comparisons———173

in two cases the meaning of EMdK saks is also ‘payment to the porter’, as
far as I can judge on the basis of examples given in LCT 1987: 435, but the
other two examples involve also ‘payment for hire’, and one more may be
just ‘payment’ in general (Nam 1997: 801). Thus, the archetype of the
meaning in Middle Korean seems to be ‘payment for hire’, not just
‘payment’. The second problem appears to be even more serious. MK sàks
needs to be analyzed as sàk-s in order to be comparable with OJ sakî, and
it does look like a deverbal noun in -s from the verb sàk-. Such a verb
exists in Middle Korean, but its general meaning is ‘to disappear’. That
would seem to be a dead end, and MK sàks ‘payment for hire’ would
remain morphologically indivisible. However, I think there is a possible
solution: while MdK sak- means ‘to wear thin, to become rotten, to be
digested’, its transitive counterpart saki- (not attested in Middle Korean)
includes the meaning ‘to consume money’. Hiring a worker is certainly
going to consume your money, so it is quite possible that the internal
etymology of MK sàks is ‘spent money, consumed money’ < *sàk- ‘to be
consumed’ + -s, nominalizer. This, however, will create a semantic
problem in a comparison with OJ sakî ‘good fortune, prosperity’: one does
not acquire good fortune or prosperity by spending money. In addition, and
most important, MK sàk- with a non-leniting -k- goes back to PK *sank-,
and this also makes the comparison phonetically incompatible with OJ sakî,
because PK *-nk- can correspond only to OJ -Nk-, not to OJ -k-. Therefore,
I reject this comparison due to its phonetic and semantic problems.
(204) (L) MK sál ‘arrow’ ~ OJ sa ‘id’. < PJK *sar. Whitman also states
that “OJ ya ‘arrow’ is apparently unrelated” (1985: 232). I agree with him,
and I also must add that yaa ‘arrow’ is amply attested in Ryukyuan,
including the Southern Ryukyuan dialects: Izena, Irabu, Psara, Iriomote,
Taketomi, Hateruma ya: (Nakamatsu 1987: 39, 70, 148, 196), Shuri ʔiya 175
(RGJ 1976: 253), etc. Ya ‘arrow’ is also attested in Eastern Old Japanese in
the compound satu-ya ‘hunting arrow’ (MYS XX 4374). Due to this
distribution, we can safely conclude that the word is native and can be
reconstructed as PJ *ya. It is a different story with OJ sa ‘arrow’, which is
mostly Western Old Japanese, with a single attestation in Eastern Old
Japanese (MYS XX: 4430), but has no cognates in Ryukyuan. The
distribution of OJ sa ‘arrow’ strongly suggests that it is a loan from Korean.
(205) (L) MK sàm ‘hemp’ ~ OJ asa LL ‘id’. 176 < PJK *asam. Whitman
notes that “the MK accent, rare for a monosyllable in /-m/, betokens the
original first syllable” (1985: 232). I agree with Whitman that the Proto-
Korean reconstruction should be *asam. In Japonic, asa ‘hemp’ is attested
in Western Old Japanese, as well as in Eastern Old Japanese (but only in
compounds and in poems that have no apparent Eastern Old Japanese
features: MYS XIV: 3454 [twice], 3484), and in Ryukyuan. However, the
Ryukyuan attestations are mostly confined to the Northern and Central

175
The Shuri form ʔiyaa, also found in some other Ryukyuan dialects, probably represents
a compound originating from the infinitive form ʔi- of the verb ʔiyuN ‘to shoot’ + yaa
‘arrow’ (RGJ 1976: 253).
176
This comparison is also found in Martin (1966), #104.
174———Koreo-Japonica

Ryukyus, with three limited and scattered attestations in the South

Ryukyus: Psara (ʔ)asa ‘hemp’ (Hirayama 1966: 353), Ōura ʔasa, Sonai
(ʔ)asa (Hirayama 1967: 339). Such a distribution strongly suggests that
Japonic asa ‘hemp’ is a loan from Korean *asam into Japanese, which then
spread into the Northern and Central Ryukyus. This is further supported by
the fact that Western Old Japanese has a doublet wo ‘hemp’, 177 also
attested in Middle Japanese and in Southern Ryukyuan: Ikema, Ishigaki,
Hateruma buu ‘hemp’ (Hirayama 1966: 353), Uechi, Nakasuji, Ōbama,
Taketomi, Kuroshima, Yonaguni buu (Hirayama 1967: 339).
(206) (R) MK :sam- ‘takes it (as), makes it (into), adopts it, assumes (a
role, office)’ ~ OJ sama HH ‘method, way of being, situation’ < PJK
*sama (Whitman 1985: 232). Although at first glance this comparison
seems to be phonetically impeccable, there are certain problems. First, I
believe that it is overextended semantically. To the best of my knowledge,
the Middle Korean textual examples of :sam- indicate only the meaning ‘to
make it (into/by)’ (Nam 1997: 808; LCT 1987: 439). OJ sama ‘way, state,
condition, appearance’ involves more serious problems. First, its
independent phonetic attestations in Western Old Japanese are limited to
compounds or collocations akara-sama n-i ‘suddenly’, ka-ku sama n-i ‘in
this way’, kari-sama ‘euphemeral’, saka-sama n-i ‘against’, and yökö-
sama n-i ‘horizontally on the side’. Second, the last two compounds also
have the variants saka-sima n-i and yökö-sima n-i, which suggest that the
first syllable vowel /a/ in WOJ sama may be a result of regressive vowel
assimilation. If this is true, the comparison with MK :sam- becomes
problematic. In addition, WOJ yökö-sima ‘horizontally on the side’ is also
attested as WOJ and MJ yökö-si ‘id’.(JDB 1967: 794), which may
represent a truncation, but more likely indicates that sama < sima is
bimorphemic in origin: *si + ma. Quite conspicuously, there is no sama in
Eastern Old Japanese, and attestations in Ryukyuan seem to be limited to
Shuri zama (RGJ 1976: 597). 178 Finally, WOJ sama is attested only as a
nominal, and MK :sam- only as a verb, with no reverse derivations. All of
these problems make this etymology more than doubtful, so I reject it.
(207) (L) MK sàsól ‘bamboo branch, stick (for drawing lots)’ ~ OJ sasa
‘bamboo grass, small bamboo’ < PJK *sasar (Whitman 1985: 233). MK
sàsól is better defined as ‘piece of split bamboo’ (Nam 1997: 797-798),
which might have different uses: for divination, drawing lots, book pins,
etc. MK sàsól looks suspiciously like a verbal attributive irrealis form, and

177
WOJ wo ‘hemp’ is attested phonetically only in the compound wo-kë ‘[hemp] container’
(JDB 1967: 831-832), but it is also used independently or in compounds written
logographically as 麻 in several poems in the Western Old Japanese section of the
Man’yōshū (MYS VI: 1056, IX: 1807, XI: 2687, XIII: 3272), where this character is likely
to be read as /wo/ and not /asa/ on the basis of the poetic meter. It also appears written
phonetically as /wo/ in the aforementioned MYS XIV: 3484, which looks like a typical
Western Old Japanese text, although it is found within the Eastern Old Japanese corpus.
178
In addition, a unique compound with -zama is attested only in Shuri as si-zama ‘situation,
way of doing’ (RGJ 1976: 487) and in Nakijin as si-zaama ‘id’. (Nakasone 1983: 184).
They are frequently used with a negative connotation. Their narrowed meaning and
confinement to the island of Okinawa clearly indicates that they are loans from Japanese.
 Lexical Comparisons———175

since in Middle Korean the attributive irrealis form could be used as a

nominalized form, I suspect that MK sàsól is historically the attributive
irrealis *sas-ol, which just means ‘a split one’. It is true that there is no
Middle Korean verb *:saz- ‘to split’ to underlie such a derivation, but MK
sàsól may be etymologically connected to MK sàhól-, MdK ssel- ‘to split’.
However, this internal etymology is not completely unproblematic, due to
irregularities in the correspondences. The distribution of sasa in Japonic is
skewed: there is no lack in attestations of WOJ and MJ sasa (JDB 1967:
327), (Miyajima 1971: 131). EOJ sasa is attested twice in poems with
Eastern Old Japanese features (MYS XIV: 3382, XX: 4431). The only
Ryukyuan attestation is in the Northern Ryukyus: Yamatoma sasa
(Hirayama 1986: 314), which is certainly a loan from Japanese. A possible
secondary semantic connection with ‘bamboo’ in Korean, plus the skewed
distribution in Japonic, strongly suggest that this is a loan from Korean to
Japanese.
(208) (R) MK sàwó- ‘rough, wild, fierce’ ~ OJ sawa ‘id., noisy’ < PJK
*sabo (Whitman 1985: 233). Whitman adds: “The OJ morpheme appears
in the reduplicated phrase sawa-sawa-ni ‘violently’ and in the verb sawak-
‘is violent, is noisy’. The MK stem appears as the adjectival verb sàwónàp-
< sawo + (?) na- ‘become’ + -p- (adjectival suffix). Sawo- appears in
LMK as an independent verb stem” (1985: 233). There are numerous
problems with this etymology. First, the meaning ‘to be violent’ for WOJ
sawak- seems to be poorly supported: clear examples indicate only ‘to be
noisy’, and the examples that Omodaka et al. argue indicate some
appearance of physical unrest (JDB 1967: 344) are limited to two poems
that do not include phonetic spellings for the word in question (MYS III:
262, VI: 927), but are also open to alternative interpretations. I also believe
that WOJ sawa-sawa-ni, being an onomatopoetic form, cannot really be
used in comparative linguistics. The word certainly does not mean
‘violently’, but just ‘noisily’, as far as we know from the two examples in
Western Old Japanese, where sawa-sawa-ni is usually interpreted as
‘noisily’ (JDB 1967: 344; Tsuchihashi 1957: 76). Second, MK sàwónàp- is
really MK sàGwónàW-, which contains a velar fricative -G-. MK -G-
mostly reflects PK *-k-, although in some cases it can go back to *-p-, but
in this particular case the velar solution is preferable, given the Northern
Kyengsang saanap- (Choy 1978: 1220), which can reflect only PK
*sakwonap-. PK *-k- does not correspond to WOJ -w-. Third, Whitman’s
morphological analysis of MK sàGwónàW- as sawo + (?) na- ‘become’ +
-p- (adjectival suffix) is teleological, because there is no verb *na- ‘to
become’ in Korean. Fourth, the LMK sawo- that he cites cannot be used as
supporting evidence for his morphemic analysis, because it is a hapax
legomenon, attested only in the Chiltay Manpep (1569) (LCT 1987: 433;
Nam 1997: 799). This occurrence is strongly contradicted by both MK
sàGwónàW- and MdK sanap-, as well as by attestations in dialects (Choy
1978: 1220). Thus, I reject this etymology on the basis of its multiple
phonological, morphological, and semantic problems.
176———Koreo-Japonica

(209) (R) MK :say ‘bird’ ~ OJ saNkî ‘heron’ 179 < PJK *sagi (Whitman
1985: 233). This etymology must be rejected, because MK -Ø- does not
correspond regularly to OJ -Nk- < PJ *-nk-. MK *saki < PK *sanki would
be expected. In addition, the semantics of the comparison is dubious, since
it compares a generic word ‘bird’ with the name of a particular species of
birds.
(210) (R)/(L) MK -sò nominalizer for preceding verbal expression ~ OJ
-sa nominalizing suffix for adjectives; nominalizer for preceding verbal
expression < PJK *-so (Whitman 1985: 233). Whitman comments: “OJ -sa
occurs as a productive nominalizing suffix for inflected adjectives in -si
such as tanô-si 180 ‘enjoyable’, tanô-si-sa ‘enjoyment’, kuru-si ‘painful’,
kuru-si-sa ‘painfulness’. It also occurs after the attributive ending of verbs
of motion such as kapêr- 181 ‘return home’: kapêr-u-sa ‘returning home’
(MYS XV: 182 3614). This usage is exactly parallel to the usage of the
attributive form of verbals in the sense of ‘act of doing, state, fact of being’
(1985: 233). However, -sa in Western Old Japanese does not exclusively
nominalize adjectives ending in -si, as Whitman suggests, although his
statement holds for Eastern Old Japanese, where -sa is found only after
adjectives ending in -si. In Western Old Japanese -sa is also found as a
nominalizer after other adjectives as well: na-sa (MYS XVII: 3928)
‘nonexistence’ < na- to be non-existent’; NANKA-sa (MYS VI: 985) 183
‘length’ < naNka- ‘long’; PARUKË-sa (MYS VIII: 1494, 1550; X: 1952)
‘distance’ < parukë- ‘distant’; SAYAKË-sa ‘clearness’ (MYS III: 314; VII:
1076, 1112, 1201; IX: 1724, 1737; X: 2141), 184 < sayakë- ‘clear’;
suNpêna-sa ‘uselessness’ (MYS IV: 757; V: 796; XVIII: 4106), <
suNpêna- ‘useless’; TAPUTÔ-sa ‘awe’ (MYS XIX: 4254, 4255, 4266), <
taputô- ‘awesome’; YÖ-sa ‘goodness’ (MYS X: 2073, 2233) < yö- ‘good‘.
The function of WOJ -sa as a nominalizer for adjectives is found
throughout the Japonic family, including Ryukyuan. More important, the
nominalizing function of -sa for verbs as in kapêr-u-sa-ni ‘when [I] return
home’, cited by Whitman, does not appear to be nominalizing; on the
contrary, traditionally it is believed to be found not after the attributive, but
after the final verbal form with verbs that have morphologically different
attributive and final forms. This can be seen in the constructions k-u sa
come-FIN sa ‘when [one] comes’ (MYS III: 281; IX: 1784; XX: 4514) and
wasur-u sa forget-FIN sa ‘when [one] forgets’ (MYS XI: 2580). 185 Clearly,
179
This comparison is also found in Martin (1966), #14.
180
Whitman has tano-si instead of tanô-si in this case and the following, but, it is quite
clear that it should be tanô-si rather than tanö-si (JDB 1967: 434).
181
Whitman has kaper- instead of kapêr- in this case and the following, but it is quite clear
that it should be kapêr- rather than kapër- (JDB 1967: 215).
182
I have added the book number to the citation.
183
One more uncertain case is found in (MYS IV: 546).
184
Two more uncertain cases are found in (MYS VII: 1102, 1159).
185
Besides kapêr- ‘to return’, kö- ‘to come’, and wasure- ‘to forget’, -sa is attested after
yuk-u ‘go-FIN’, but yuk- ‘to go’ is also a consonantal verb that like kapêr- ‘to return’ does
not differentiate between final and attributive forms. Further supporting evidence comes
from Eastern Old Japanese, where -sa is attested once after the final form pus-u of EOJ
puse- ‘to lay, to put’ (MYS XIV: 3484).
 Lexical Comparisons———177

the nominalizations would require the attributive forms k-uru ‘come-

ATTR’ and wasur-uru ‘forget-ATTR’, with resulting forms *k-uru sa and
*wasur-uru sa. In addition, these constructions introduce temporal
dependent clauses ‘when V’ (with one exception found in MYS IV: 656),
which is too narrow a context for a nominalization in general. It is unclear
why a nominalizer would follow a final form rather than an attributive
form or a stem of a verb (cf. the behavior of the nominalizers –[a]ku and
-[y]i). I believe that it is appropriate to treat the vowel /u/ immediately
preceding /sa/ as belonging to the suffix itself. Therefore, I prefer to
segment this nominalizer as -usa (Vovin 2009: 776-779). I believe that the
phonetical, functional, and formal differences between the OJ -usa
following the root or the contacted root form and the Middle Korean
nominalizer -sò are too significant to allow the comparison, and I believe it
is just a chance similarity. This leaves us with only Japonic nominalizer -sa
found after adjectival stems. However, there is another discrepancy: while
the Japonic nominalizer -sa follows an adjectival stem or root, MK -sò as a
nominalizer appears only after the attributive irrealis in -l(ʔ), never after a
verbal root or even after an attributive realis in -n, so there is an important
formal difference. We should also keep in mind that the verbalization of
adjectives in Japonic is a very recent process. It is quite clear that pre-Old
Japanese adjectives did not have any or most of the verbal-like markers
they later developed, and that they were for the most part indistinguishable
from nouns. In Ryukyuan, -sa is a necessary part of any adjectival form,
which raises serious doubts about its antiquity as well. Finally, we should
keep in mind that in both branches of Japonic -sa occurs only with
adjectives (quality verbs), while on the Korean side there is no such
limitation. Therefore, I strongly suspect that this etymology is a chance
similarity. Japonic -sa may be an early loan from Korean, dating back to
the period of coexistence on the Korean peninsula, a period that
contributed significantly to the ‘Koreanization’ of the Japonic adjectival
system.
(211) (R) MK sólh ‘flesh’ ~ OJ sisi LL ‘id’. < PJK *sil2o (+ -i)
(Whitman 1985: 233). 186 There are two problems with this comparison.
First, MK -o- does not regularly correspond to OJ -i-, according to
Whitman’s correspondences (Whitman 1985: 129). Second, WOJ and EOJ
sisi does not mean ‘flesh’, but ‘game animal’, in particular referring to deer
or boars (JDB 1967: 352). In Ryukyuan, sisi means ‘meat’ (Hirayama
1966: 325), but this may be a result of secondary semantic development:
*game animal > *edible animal > meat, especially since the opposite
semantic development is hard to imagine. Therefore, I reject this
etymology.
(212) (R) MK sómóy ~ sómáy 187 ‘sleeve’ ~ OJ sôNtë 188 ‘id’. Whitman
adds: “The OJ form is a compound of *sô + të ‘hand, arm’. The original

186
This comparison is also found in Martin (1966), #83.
187
Actually MK sómày (LCT 1987: 422; Nam 1997: 936).
188
Actually OJ sôNte, because we do not know whether it is *sôNtë or *sôNtê, since there
is no /ê/ : /ë/ contrast after coronals in Old Japanese.
178———Koreo-Japonica

morpheme *sô is related to the MK form by medial *-m- loss from PJK
*swómó or *swómá” (Whitman 1985: 233). There are several problems
with this etymology. First, there is no internal evidence in Japonic that OJ
sôNte ‘sleeve’ is a compound. Whitman’s assumption that it originally
consisted of the morphemes *sô + te ‘hand, arm’ is based exclusively on a
circular comparison with Korean. There is no morpheme *sô ‘sleeve’ in
Japonic. In addition, I am not aware of any crosslinguistic examples where
a word for ‘sleeve’ is derived from ‘sleeve’ + ‘hand’. On the contrary,
there is some typological evidence that ‘sleeve’ can be just a derivation
from the word ‘hand, arm’, cf. Russian ruka-v ‘sleeve’ < ruka ‘hand, arm’.
Second, the rule of *-m- loss (although recently Whitman seems to have
abandoned the idea, personal communication) in its classic formulation
involves only cases after the high back vowel /u/ (Whitman 1985: 25),
which is clearly not the case here. Third, the alleged rule of *-m- loss
leaves OJ -Nt- instead of -t- unexplained, which should be expected if the
nasal was really lost. Fourth, MK /o/ does not correspond regularly to OJ
/ô/, according to Whitman’s own vocalic correspondences (Whitman 1985:
129). Due to all these problems I reject this etymology. I believe that OJ
sôNte ‘sleeve’ may be a loan from unattested Old Korean *swontoy <
swon ‘hand’ + toy ‘place’, that is ‘the thing where you put your hands’.
Although this etymology is speculative, since OK *swontoy is not actually
attested, it involves regular phonetic correspondences and is typologically
compatible. 189 See also (136, 185-186) above and (293, 330) below on the
alleged *-m- loss.
(213) (L) MK sòy- ‘sours, tastes like acid‘, swùy- ‘turns sour’ ~ OJ
suyur- ‘food goes sour’ < PJK *sueru-/*sürü- (Whitman 1985: 233).
Although at first glance this comparison looks good, there are problems. I
presume that Whitman’s swùy- ‘turns sour’ with L pitch is a typographical
error: it is actually MK :swuy- with R pitch (LCT 1987: 479; Nam 1997:
890), indicating the disyllabic origin of this word. The alternation between
MK /o/ and /wu/ is not expected either, so the phonological history of this
word may be very complex: it may go back to something like PK *swuyu-
or *swuCuy-, and data from Korean dialects strongly favor the second
possibility: Hamkyeng Pukto sikwu-, Hamkyeng Namto sikhwu-,
Kyengsang Pukto saykwup-, Kangwen sikep-, etc. (Choy 1978: 1226-1227).
Thus, the Proto-Korean reconstruction should be something like
*swukwuy-, and this rules out any possibility of a genetic connection
between MK sòy- ~ :swuy-, on the one hand, and OJ suyur-, on the other,
since PK *-k- does not correspond regularly to OJ -y-. Also, there are no
phonetic attestations of suyur- in Old Japanese. The first phonetic
attestation is found only in Middle Japanese (WMS XVI: 11b). There is
also a disagreement among Japanese scholars as to whether the verb in
question is a consonantal verb suyur- or a vowel verb suye- (JDB 1967:
394), since even in Middle Japanese it is attested only twice: in WMS,
mentioned above, and in the Iroha jirui shō (1180). Nevertheless, the

189
Cf. (63) above, where OJ- te ‘place’ is treated as a loan from Korean toy ‘id’.
 Lexical Comparisons———179

attributive of the progressive form suyur-er-u ‘turn sour-PROG-ATTR’

that appears in WMS XVI: 11b strongly suggests that it was a consonantal
verb in Middle Japanese, in contrast to Modern Japanese monograde sue-
‘to turn sour/bad’, which seems to be a secondary development from a
vowel bigrade verb. As always, the distribution of this verb in Japonic is
revealing: it is not attested in either Eastern Old Japanese or in Ryukyuan,
except in three Central Ryukyuan dialects: Iejima tsiiyuN 190 (Oshio 1999:
221), Nakijin siiruN (Nakasone 1983: 173), and Shuri siiyuN ‘to turn sour’
(RGJ 1976: 470). All of these apparently represent loans from Japanese
after MJ suyur- became a monograde verb. Together with the phonetic
impossibility of explaining this etymology as having a Koreo-Japonic
genetic heritage, this distribution makes MJ suyur- a perfect candidate for
a loan from Korean that took place in a period postdating the lenition of
PK *-k- to *-Ø- in some Korean dialects.
(214) (R) MK skí- ‘inserts it, thrusts it into’ ~ OJ suk- ‘id’. < PJK *sok-
(Whitman 1985: 233). I really cannot find any attestations of OJ suk- with
the meaning ‘to insert, to thrust’, except MJ suk- 191 ‘to eat/drink, to gulp
down something’ (JDB 1967: 384), but this seems to be too far-fetched,
and I have no other choice but to dismiss this etymology as a ghost.
Another decisive blow to this etymology comes from the fact that MK skí-
is a later form of the word, also attested as MK pskí- in slightly earlier
texts (LCT 1987: 112; Nam 1997: 785).
(215) (R) MK sywó ‘cow’ ~ OJ usi HH < PJK *syu (Whitman 1985:
233). 192 I am afraid that the resemblance between MK sywó and OJ usi is
accidental: there are no other examples where MK s- would correspond to
OJ *Vs-. In addition, OJ -i- certainly cannot be derived from *io.
Therefore, this etymology is really based on one phoneme /s/, so I reject it.
(216) (R) MK spól- ‘sucks it’ ~ OJ sup- H ‘id’. < PJK *sopor-
(Whitman 1985: 233). In both languages these are clearly onomatopoetic
words, since crosslinguistically ‘sucking’ words tend to include /s/ and/or
/p/ as well: English suck, German saugen, Russian sosat’, Malay isap,
Mandarin Chinese xi1 < MC *xiəp, etc. Even if this word were not
onomatopoetic, this etymology still faces the problem of explaining -ol- in
MK spól- as a suffix, otherwise this comparison will not work. Since there
is no such suffix in Korean, the etymology can be safely rejected.
(217) (R) MK swól ‘pine’ ~ OJ su- ‘cryptomeria’ < PJK *sur. Whitman
adds: “Both forms occur in compounds with the word ‘tree’: MK
swónámk- < swól + námk- ‘tree’; OJ suNkï ‘cryptomeria’ < su- + kï ‘tree’
(1985: 233). Apart from the observation that the only commonality
between pines and cryptomerias is that both are evergreens, everything else
is tantalizingly different: East Asian pines are mostly crooked, but
cryptomerias are straight; cryptomerias also dwarf pines by at least 15
meters. This etymology seems ingenious, but unfortunately in its Japanese

190
Initial ts- in Iejima is irregular. In addition, both Iejima and Nakijin also have a verb
pumichuN ‘to go sour, to rot’, that appears to be ‘native’.
191
No Old Japanese phonetic attestations.
192
This comparison is also found in Martin (1966), #300.
180———Koreo-Japonica

part it is completely ad hoc: there is no internal Japonic evidence that OJ

suNkï ‘cryptomeria’ is a compound consisting of mysterious *su + kï ‘tree’.
Luckily, for our purposes, cryptomerias also grow in the Ryukyus: Nakijin
sizii (Nakasone 1983: 181), Iejima siji (Oshio 1999: 170), Shuri siji (RGJ
1976: 487), Yaeyama sïgï (Miyara 1980: 307), sïg’i (Miyara 1981: 286),
Hatoma s’ig’i (Miyara 1980: 323). Nakijin sizii, Shuri and Iejima siji are
particularly important. They cannot possibly be reflexes of PJ *siNköi,
which would not undergo a process of the palatalization *g > /j/. Thus,
Nakijin /zii/ and Shuri and Iejima /ji/ can correspond only to OJ /Nkî/ or
/Nkï/ < *nkui; they cannot correspond to OJ /Nkï/ < *nköi. Since the root
of the word ‘tree’ in pre-Old Japanese is *köy, not *kuy (cf. OJ kö-nö pa
‘tree leaves’), any attempt to see WOJ suNkï ‘cryptomeria’ as a compound
including kï ~ kö- should be abandoned. 193 Thus, WOJ suNkï represents an
indivisible word, and Whitman’s etymology should be rejected.
(218) (R) MK swólá ‘dish, vessel’ ~ OJ suwe LL ‘pottery vessel
manufactured chiefly by Korean potters’ < PJK *sura (+ -i) (Whitman
1985: 234). The semantics of WOJ suwe, which is also attested only in
Middle Japanese and Western Old Japanese, clearly indicates that it must
be a loan from Korean. However, even for a loan, the correspondence of
MK -l- to WOJ -w- is problematic. Therefore, I reject this comparison.
(219) (R) MK swón ‘hand, arm’ ~ OJ sune ‘shin’ < PJK *sun (Whitman
1985: 234). MK swón means only ‘hand’, while ‘arm’ is MK pòlh, and that
makes the semantic side of the comparison even weaker. Cf. also
Whitman’s other semantically unacceptable body parts comparisons, (57)
and (250). The Ryukyuan data: Nase, Koniya sïnï, Hentona śini, Iejima
sinii, Ishigaki sïni, etc., ‘shin’ (Hirayama 1966: 379) indicate PR and PJ
*sune, not *sone, which also represents a problem for this comparison,
because PJ *sone with *o would be expected as a cognate. Therefore, I
reject this etymology.
(220) This number is absent from Whitman’s list.
(221) (R) MK swóy ‘metal’ ~ OJ suNsu ‘metal bell, chimes’, MJ suzu
‘tin’ < PJK *suzu (Whitman 1985: 234). In addition to the dubious
semantics, 194 MK -Ø- does not correspond regularly to OJ -Ns-, so the
etymology can be rejected.
(222) (L) MK :se-, :sek-, :sey-, :seyh ‘three’ ~ OJ sakî- 195 in sakîkusa
(unidentified plant glossed ‘three branches’ 196) < sakî + kusa ‘grass plant’.
193
Another obstacle to this etymology is its vocalism: Ryukyuan forms clearly indicate PR
and PJ *u in the first syllable, not *o.
194
As far as I can tell, the word suzu with the meaning ‘sake jar made of tin’ appears for the
first time in the Edo period text Genkei keiki (IKJ 1990: 694), so it is definitely not Middle
Japanese, and can be discounted for comparative purposes.
195
There is only oblique evidence in favor of sakî- vs. sakï-, as once in the Nihonshoki the
name of this plant is spelled as 福草 ‘happiness-grass’ in contrast to the usual
semantographic spelling 三草 ‘three-grass’ (JDB 1967: 323). Since the Western Old
Japanese word for ‘happiness’ is clearly /sakî/ (JDB 1967: 322), we can presume that the
reading is /sakî/.
196
It is quite clear from textual evidence that WOJ sakîkusa referred to some kind of plant
that either had three stalks or whose leaves had tripartite shape. Cf.
SAKÎ-KUSA-NÖ NAKA-NI-WO ne-m-u tö UTUKUSI-ku si-Nka katar-ap-ë-Npa
 Lexical Comparisons———181

Whitman notes: “Sakî- in this word may represent a loan” (1985: 234).
Since sakî- occurs only in sakî-kusa and only in Western Old Japanese, it
can only be a loan from Korean. In addition, it is also necessary to add that
numerals usually correspond as a set if they go back to the same proto-
language. If they are borrowed as independent numerals, they are also
borrowed as a set. The fact that WOJ sakî- ‘three’ is not an independent
numeral but a part of the compound, speaks strongly in favor of a
loanword relationship. Finally, I would like to note that, like many other
loans from Korean to Japanese, WOJ sakî- ‘three’ is extremely helpful in
reconstructing the earlier shapes of Korean, as it provides another strong
and independent piece of evidence that MK :seyh is a metathesized form of
OK *seki, which is also supported by MK :sek- found in combination with
certain classifiers. 197
(223) (R) MK sèk-, K sak- ‘rots, ferments’ ~ OJ saka-/sakë ‘wine‘. Cf.
also OJ sakar- ‘be at peak, be in heat’, sak- ‘bloom’ < PJK *sakar-
(Whitman 1985: 234). There are problems with this etymology. First,
Middle Korean non-leniting -k- goes back to PK *-nk-, which does not
correspond to Old Japanese voiceless -k-. Second, the semantic connection
between ‘rotting’ and ‘blooming’ escapes me. I would also add that the
connection between ‘rotting’ or ‘fermenting’, and ‘rice wine’ is very weak:
‘rice wine’ is ‘brewed’, not ‘rotted’ or ‘fermented’. Third, even in Japonic
internally OJ sak- ‘bloom’ belongs to accent class B, and MJ sakar- 198 ‘to
flourish’, which has nothing to do with ‘heating’, to accent class A. In spite
of their close semantics, this accent difference makes their etymological
connection dubious. Thus, I reject this etymology.
(224) (L) MK sùmúy- ‘soaks, permeates’ ~ OJ sim-, söm- ‘id.’, simë-,
sömë- ‘lets soak, dyes’ < PJK *söme- (Whitman 1985: 234). 199 This
etymology appears to be a difficult case, which I will try to untangle below.
MJ söm- is not attested in Old Japanese, and even in Middle Japanese
examples are rare (JDB 1967: 406). WOJ sim- is attested phonetically only
once (MYS XX: 4445); in all other cases, where the word is written
logographically we simply do not know whether it is sim- or söm- (JDB
1967: 369). In addition, all the examples of consonantal verbs sim- and
söm- I was able to trace in Middle Japanese texts seem to be limited to the
cases with a following infinitive -i or final -u. In this position, we cannot
say with certainty whether the verb is indeed consonantal: it may also be
an upper bigrade vowel verb. Söm- seems to disappear after late Middle
Japanese, and the only continuation of sim- beyond Middle Japanese is
clearly in the form of the vowel verb simi-, which indicates that the alleged

three-grass-COMP middle-LOC-ACC(?) sleep-TENT-FIN DV beautiful-INF he-POSS

speak-ITER-EV-CON
as he repeatedly said in a beautiful manner: ‘Let [three of us] sleep [together] like saki-
kusa (MYS V: 904).
197
It seems that the same generalization can be made about the identical alternation of
MK :nek- and :neyh ‘four’ < PK *neki, although comparable evidence from Western Old
Japanese is lacking.
198
There are no phonetic attestations of sakar- in Old Japanese (JDB 1967: 321).
199
This comparison is also found in Martin (1966), #205.
182———Koreo-Japonica

consonantal sim- could actually be the bigrade simï-. The only evidence
that speaks against it is the aforementioned hapax legomenon, simînisi ‘has
permeated’ in (MYS XX: 4445), but I seriously distrust such limited
evidence. The restricted attestation of söm(ï)- allows me to remove it from
further consideration; therefore, I limit the following discussion to WOJ
sim(ï)- ‘to permeate’, and simë-, sömë- ‘to dye’, none of them attested in
Eastern Old Japanese with the exception of EOJ sömë-, which has a single
attestation in MYS XX: 4424. In spite of their semantic similarity, I have
grave doubts that WOJ sim(ï)- and simë-, belonging to accent class A, are
in fact related to WOJ sömë-, which belongs to accent class B. In any case,
WOJ sim(ï)- and simë- cannot be related to MK sùmúy-, because MK /u/
does not regularly correspond to WOJ /i/ (Whitman 1985: 129). They can,
however, be loans from Korean. This leaves only WOJ sömë- ‘to dye’ < PJ
*sömöi- as a potential cognate of MK sùmúy-. The distribution in the
Ryukyus presents a very interesting picture: parallels to WOJ sömë- are
amply attested throughout the Northern and Central Ryukyus: Yuwan
sumïyuN; Izena, Namizato, Sesoko sumiN; Oku, Benoki, Kijoka, Arumi,
Uezu, Maejima, Tomigusuku, Tokashiki sumiiN; and there are also two
isolated attestations found in two neighboring localities on Miyako Island
in the Southern Ryukyus: Higashi Nakasone and Yonaha sumiï (Uchima
and Arakaki 2000: 411). This distribution is highly reminiscent of a
loanword gradually spreading from north to south. Thus, on the basis of its
distribution in Ryukyuan, I conclude that WOJ sömë- ‘to dye’ is a
loanword from Korean, which subsequently spread to the Ryukyus.
However, cognates of WOJ simë-, which, as we have seen above, cannot
be a cognate to MK sùmúy-, are found in the Ryukyus with a distribution
that clearly points to a genetic relationship to WOJ simë-. With the
exception of two attestations in the southern part of Okinawa Island, Henza
and Kowan śimiiN (Uchima and Arakaki 2000: 411), all other examples
are found deep in the southern part of the archipelago: Tonoshiro sïmiruN;
Hateruma śimiruN (Uchima and Arakaki 2000: 411); Yaeyama sïmuN,
sïmaruN (intr.), sïmi(ru)N (tr.) (Miyara 1981: 282). Therefore, I think that
we must reconstruct PJ *simöi- on the basis of WOJ simë- and its South
Ryukyuan cognates. However, it still cannot be genetically linked to MK
sùmúy- due to the lack of a regular correspondence for the vowel in the
first syllable, as mentioned above. I believe that we are dealing here with
two different layers of loans from Korean to Japonic. The early layer is
represented by Korean *simuy-, borrowed as PJ *simë-, which still
probably reflects the original Korean vocalism prior to vowel reduction *i
> MK /u/. This early loanword would have been borrowed at the time of
mutual coexistence on the Korean Peninsula, or at least before the final
separation of Ryukyuan from Japanese. The later layer is represented by
WOJ sömë- and the successive loans from MJ some- into Ryukyuan.
(225) (C) MK sèl- ‘clears it, washes it off’ ~ OJ saras- ‘id., leaves (to
be bleached) to wind and rain’ < PJK *sar-. Whitman adds: “The original
stems do not match in transitivity: OJ has sar- ‘is left to, is washed by
wind and rain’” (1985: 234). I am afraid that the Old Japanese data are not
 Lexical Comparisons———183

presented quite accurately. Omodaka et al. list three meanings for WOJ
saras-: 200 (1) ‘expose to wind and rain’, (2) ‘dry and bleach by exposing to
the sun’, (3) ‘to wash [out] (by waves)’ (JDB 1967: 343). Neither the first
nor third meanings is supported by examples in phonetic spelling, but in
any case the underlying meaning seems to be ‘to expose (to the elements)’,
not ‘to wash’ or ‘to clean’. OJ sar- is a hapax legomenon attested as a kana
gloss in NR III:1. 201 Since it is a kana, not a man’yōgana gloss, it cannot
possibly be as old as Old Japanese. There is sare- ‘be exposed (to
elements)’, ‘be polished’, attested in Middle Japanese (IKJ 1990: 580),
which, however, supports the existence of an intransitive verb. Cognates of
saras- and sare- are attested in Ryukyuan, including the South Ryukyus:
Nakijin saraasuN (Nakasone 1983: 756); Iejima saraśun, sariyuN (Oshio
1999: 134); Shuri sarasyuN, sariyuN (RGJ 1976: 460); Psara sarasu,
sarirü (Shimoji 1979: 86); Yaeyama sarasïN (Miyara 1981: 255). It is a
possible etymology, but its semantics are not exciting.
(226) (L) MK sèp, K seph ‘brush, firewood’ ~ OJ siNpa LL or LH ‘id’.
(Whitman 1985: 234). 202 No Proto-Japanese-Korean reconstruction is
provided. MK sèp is likely to originate from a disyllabic form, due to the
rare low pitch on a monosyllabic noun. Since the distinction between MK
-p- and -W- is neutralized in the final position, it is difficult to say whether
it is originally from PK *senpo or *sepo. Only PK *senpo can be
genetically related to OJ siNpa, thus the etymology has only 50 percent
credibility. In addition, in order to be related to MK sèp, OJ siNpa must be
from PJ *seNpa, or the regularity of the correspondences will be violated.
Cf. (227) below, where the correspondence is MK /ye/ < *i : OJ /i/ < *i.
There are no attestations in Ryukyuan, so we cannot tell whether the word
had *e or *i in the first syllable. The single attestation in Eastern Old
Japanese occurs in a poem with only one possibly typical Eastern Old
Japanese feature (MYS XIV: 3355). Therefore, on the basis of the
distribution of the word in Japonic, I think that it represents a
comparatively early loan from Korean to Japanese, prior to the lowering
of PJ *e > WOJ /i/.
(227) (L) MK :syem ‘island’ ~ OJ sima ‘id’. < PJK *sima 203 (Whitman
1985: 234). 204 WOJ sima has cognates amply attested throughout
Ryukyuan (Hirayama 1966: 351, 1967: 334). EOJ sima also appears three
times (MYS XIV: 3367, XX: 4355, 4374), although one of these poems has
no typical Eastern Old Japanese features (MYS XIV: 3367). Therefore,
there is no doubt that here we are dealing with a Proto-Japonic lexical item.
However, the necessary condition for this comparison on the Korean side
is the assumption that Proto-Korean underwent the breaking PK *i > MK
/ye/ that was originally suggested by Yi (1959: 131-137). However, Yi’s

200
Saras- is attested once in the Eastern Old Japanese corpus, but in a poem without any
Eastern Old Japanese features (MYS XIV: 3373).
201
The context of the kanbun text indicates ‘exposing to the sun’.
202
This comparison is also found in Martin (1966), #30.
203
For the reconstruction of the vocalism in the first syllable, cf. (228) and (262) below.
204
This comparison is also found in Martin (1966), #117.
184———Koreo-Japonica

assumption is based mostly on external data of questionable nature, so it is

unreliable for the purpose of establishing Koreo-Japonic cognates.
Whitman does not include a correspondence of MK /ye/ : OJ /i/ in the list
of his vocalic correspondences, either (Whitman 1985: 129); cf. also (29)
above. Therefore, this correspondence is irregular, and the comparison
should be treated as an early loan dating back to the period of mutual
coexistence on the Korean peninsula. The direction of the loan is likely
from Korean into Japonic, unless strong internal evidence indicating *i >
/ye/ in the history of the Korean language is provided.
(228) (R) MK :syey- ‘white, it whitens’ ~ OJ sirô- LL ‘id’. < PJK
*si:lV- (Whitman 1985: 234). There are two problems with this etymology.
The first is the irregular correspondence of MK /ye/: OJ /i/; see also (227)
above: two examples are apparently insufficient to prove the
correspondence. Second, even if we suppose that MK :syey- goes back to
PK *siri-, PK *i in the second syllable would not correspond regularly to
OJ /ô/ or even /a/, if we take OJ sira ‘white’ into consideration as well.
These two irregularities lead me to reject this etymology.
(229) (R) MK skwól ‘(grass for) pasture, fodder’ ~ OJ suNka-/suNkë LL
‘sedge’ < PJK *sugar (Whitman 1985: 234). This etymology has several
problems. First, as I indicated in (188) above, the loss of postvocalic nasals
in Korean must have occurred prior to vowel apocope. There is no similar
direct evidence regarding vowel syncope, but indirectly the intervocalic
consonant lenition *[C]V1kV2 > *[C]V1hV2, followed by the syncope of
*V1 in *CV1hV2, that led to the formation of the Middle Korean aspirates
/ph/ < *pV1h-, /th/ < *tV1h-, /kh/ < *kV1h-, /ch/ < *cV1h-, speaks in favor
of vowel syncope occurring after postovocalic nasal loss leading to the
origin of *Ck clusters: *[C]V1Nk V2 > *CV1kV2 > *CkV2-. The obvious
problem here is that we do not know the exact development of the PK
*sV1kV2- and *sV1NkV2- sequences. I suspect that while PK *sV1kV2- >
*sV1hV2- resulted in MK sV2- or ssV2-, 205 PK *sV1NkV2- > *sV1kV2-
produced MK skV2-. It is not impossible, though, that both PK *sV1kV2-
and *sV1NkV2- sequences merged as MK skV2-. Even if the latter solution
turns out to be correct, the etymology can have only 50 percent credibility,
because PK *-nk- is needed to make it compatible with OJ sunkë. Second,
OJ /a/ does not correspond regularly to MK /wo/, according to Whitman’s
vocalic correspondences (Whitman 1985: 129). Third, I have grave doubts
that ‘sedge’ can really be used as fodder. Therefore, I reject this etymology.
(230) (R) MK swúm- ‘hides, lurks in’ ~ OJ sum- ‘resides, makes one’s
nest in (of animals)’ < PJK *süm- (Whitman 1985: 234). The semantic
discrepancy is too great to entertain this etymology; in addition, MK
swúm- with rare H pitch on the verbal stem is likely to indicate PK *Vsum-
LH, which would completely invalidate the comparison.

205
Note that MK ss- is the only ‘double’ consonant that occurred in the native Middle
Korean vocabulary. Although in some cases it seems to be a late alternative of MK ps-,
there are other cases like MK ssú- ‘to write’, where psu- appears only in much later texts
and seems to be a hypercorrection of MK ssú-.
 Lexical Comparisons———185

(231) (L) MK swùsk, K swuch ‘charcoal’ ~ OJ susu LF or LH ‘soot’ <

PJK *süsük (Whitman 1985: 234). 206 There are two problems with this
comparison. First, although ‘charcoal’ and ‘soot’ are both black in color,
the similarity ends here: while ‘charcoal’ is a product designed for burning,
‘soot’ is the by-product of burning. Second, while MK swùsk could be a
development from *swuc(V)k, since the cluster -ck is not attested in
Middle Korean, it is more difficult to imagine the opposite development,
with the fortition -s- > -c-: *swusuk > *swucuk > swuch, which would be
responsible for a form with an affricate /-c-/ in Modern Korean. There are
several Modern Korean dialects, including Ceycwuto, with the form /sut/
‘charcoal’, which is not complicated by the suffixation found in Middle
Korean and Modern Korean (Choy 1978: 642) and is spelled in modern
standard orthography as {swus}. However, due to a lack of paradigmatic
forms, it is not possible to conclude whether these reflect underlying
/swus/ or /swuc/. If the underlying form is /swuc/, OJ susu can only be an
early loanword from Korean, in spite of the fact that the word is attested
throughout Ryukyuan (Hirayama 1966: 342, 1967: 317).
(232) (L) MK -si-, verbal honorific suffix ~ OJ -s-, id. < PJK *-s-. This
etymology is discussed in the section on verbal morphology (2.3.2.13). I
treat it as a likely loan from Korean into Japanese due to its morphological
problems and limited distribution in Japonic.
(233) (R) MK :sit-/sìlú- < *sìtú- ‘loads it, puts it in’ ~ OJ sute- ‘throws
it away, discards it’ < PJK *süter-. Whitman comments: “The intransitive
variants of this root strongly support the comparison: MK sìtúl- ‘withered,
emaciated’, sìtúlép- ‘tired, exhausted’ (note that the accent on the second
syllable blocks loss of the weak vowel in these two forms), OJ sutare-
‘worn out, useless’” (Whitman 1985: 235). There are several problems
with this comparison. First, neither MK sìtúl- ‘withered, emaciated’, nor
sìtúlép- ‘tired, exhausted’ can be an intransitive form of MK :sit-/sìlú- <
*sìtú- ‘loads it, puts it in’, because the former two include the non-leniting
MK -t- < PK *-nt-, while the last has the leniting MK -t- < PK *-t-.
Therefore, MK sìtúl- ‘withered, emaciated’, sìtúlép- ‘tired, exhausted’
cannot be related either to OJ sute- or MJ sutare- 207, because Middle
Korean non-leniting -t- < PK *-nt- does not correspond regularly to OJ -t-
< PJ *-t-. In addition, while ‘loading’ can make one ‘exhausted’, it is
highly unlikely that it will make one ‘withered’ or ‘emaciated’. In short,
semantically it is stretched beyond any credibility: we would expect the
intransitive variant of ‘to load’ to be simply ‘to be loaded’. Second, I can
see no connection between OJ sute- (accent class A) ‘to throw away’ and
MJ sutare- ‘to decline, to become useless’ (accent class B) either
phonetically or semantically. Third, OJ /u/ does not correspond regularly to
MK /i/, according to Whitman’s vocalic correspondences (Whitman 1985:
129). Therefore, I reject this etymology.

206
This comparison is also found in Martin (1966), #208.
207
Not attested phonetically in Western Old Japanese (JDB 1967: 389).
186———Koreo-Japonica

3.2.9 *z-
Whitman presents only two etymologies to support PJK *z-, and both
have very different reflexes in Old Japanese. While this Proto–Japanese-
Korean phoneme should be rejected, 208 this does not automatically
invalidate the etymologies that I discuss below.
(234) (R) MK :sel ‘new year, year of age’ ~ OJ -Nsö in kîNsö ‘last year’
< kö ‘this’ + sö < PJK *se:r/*ze:r. Whitman adds: “OJ voicing may be
secondary due to rendaku” (Whitman 1985: 235). First, OJ kîNsö meaning
‘last year’ is a ghost: there is WOJ kîsö/kîNsö 209 and EOJ kiso ‘last night’,
but ‘last year’ is only WOJ köNsö (JDB 1967: 241, 295). The fact that
there are OJ kî[N]sö ‘last night’ and WOJ köNsö ‘last year’ makes
Whitman’s interpretation of OJ -[N]sö as ‘year’ very doubtful. It is much
more realistic to assume that both OJ kî[N]sö ‘last night’ and WOJ köNsö
‘last year’ are compounds, and at present their etymologies are beyond our
knowledge. Therefore, I reject this etymology.
(235) (R) MK :sil ‘thread’ ~ OJ itö ‘id’. < PJK *zitör? (Whitman 1985:
235). The correspondence of MK /s-/ to OJ /Ø-/ is unique for this example,
so it has to be treated as irregular. In addition, although MK :sil ‘thread’
must have originated from a disyllabic structure, we have no way of
knowing whether the Proto-Korean form was *sìlú or *sìtú, and only the
second form could be comparable to OJ itö. Therefore, I reject this
etymology.

3.2.10 *š-
As the reader will see below, the reconstruction of PJK *š- based on the
alleged correspondence of MK h- to OJ s- is questionable considering the
etymologies that are supposed to support it. This new Proto–Japanese-
Korean phoneme also has a unique distribution (it occurs only in initial
position), and this also makes it dubious.
(236) (R) MK hànólh ‘sky’ ~ OJ sôra LF 210‘id’. < PJK *šanora
(Whitman 1985: 235). In addition to the unrealistic correspondence of MK
h- to OJ s-, this etymology largely depends on the assumption of *-r- loss
in Japonic, which can be assumed only on the basis of the Korean form in
the first place; it cannot be verified internally. See also (5) above for more
detailed discussion and (264) below. I reject this etymology.
(237) (R) MK hò- ‘does’ ~ OJ su- ‘id’. < PJK *šo- (Whitman 1985:
235). This etymology is potentially acceptable, but certainly not as it is
presented. First, the Middle Korean verb exhibits the variation of two
stems: ho- and hoy-, but assigning any inherent pitch for them is premature,
because this verb belongs to a verbal class with extremely irregular
accentual behavior (Kim Wancin 1973: 57; Ramsey 1991: 232). The stem

208
Martin also believes that Whitman’s PJK *z- is unconvincing (Martin 1991: 273).
209
WOJ kîNsö ‘last night’ is likely to be a ghost: it tentatively occurs only once spelled as
伎賊 in MYS II: 150. The problem is that the man’yōgana sign 賊 that has an alleged
reading /Nsö/ (JDB 1967: 896) does not occur anywhere else in the Western Old Japanese
corpus.
210
OJ sôra should be LH (2.4), not LF (2.5) (Hirayama 1989: 432; Martin 1987: 530).
 Lexical Comparisons———187

of the Old Japanese verb is clearly se, not su-, but we have no internal
Japonic evidence indicating whether this OJ se- is from PJ *sia- or PJ *sai-.
Martin speculated that the Middle Korean form hoy- might be a result of
metathesis hoy- < *hyo- (1996: 36). The Ceycwuto dialect has evidence for
ho- (Ceycwu pangen yenkwuhoy 1995: 601), and hoy- (Kim Chwunghoy
et al. 1995: 176), but not for *hyo-, as far as I can tell. Since Ceycwu is the
only Korean dialect that preserves PK *ya as /yo/, 211 Martin’s speculation
seems to be unsupported by data, and we have to assume that MK hoy- <
*hoy-, not *hyo-. It is possible to suggest a common origin for OJ se- and
MK ho-/hoy- only if we can cogently demonstrate that OJ se- is from PJ
*sia- and that MK hoy- is from PK *hyo-. However, because it is
impossible to prove that OJ se- goes back to PJ *sia on the basis of internal
Japonic data, and because the development of MK hoy- < *hyo- seems
unlikely as well, the genetic comparison of the two items is, even in the
best case, a hypothesis that is built on two other unproven hypotheses.
Therefore, for the time being this etymology must be rejected. See also
(244) below.
(238) (R) MK hòlk ‘earth, soil, clay’ ~ OJ su, MJ suna HH, sunago
‘sand, earth’ (Whitman 1985: 235). Discussed and rejected in (16) above.
(239) (R) MK hò, 212 hòn-, hònòlh 213 ‘one’ ~ OJ su- L ‘plain, unadorned’
< PJK *šon (Whitman 1985: 235). In addition to the vague semantics,
there are two significant problems with this comparison. First, OJ su ‘plain,
unadorned’ is a loan from MC swo ‘plain, simple, unadorned’ (Vovin
1993b: 340). Second, the original nature of the nasal in MK hònàh is
dubious in light of EMK xatun (Kyeylim #19) and OK HAton ‘one’
(Hyangka VII: 6, 8; XI: 7). Therefore, I reject this etymology.
(240) (R) MK hwò- in hwòwák 214 ‘mortar’ ~ OJ usu LH ‘id’. < PJK *šu
(+ -Vk diminutive). Whitman comments: “The MK form is a compound of
hwo + -pak (suffix forming vulgar nouns). Note the prothetic [sic] vowel
in the OJ form from the monosyllabic PJK root *so” (1985: 236). There
are three problems with this etymology. First, the assumption that OJ usu
‘mortar’ has a ‘prothetic’ initial /u-/ is ad hoc: it is not only unsupported by
internal Japonic evidence, but is directly contradicted by it, since there are
plenty of Japonic words with initial /s-/ that have no such /u-/. Second,
there is no evidence for the segmentation of the suffix -pak ‘forming
vulgar nouns’ in MK hwò[G]wàk, because the suffix is actually -ppak
(Martin 1992: 758) and is not attested in Middle Korean. The Middle
Korean form is likely to have been hwòGwàk with -G-, and not hwòwàk.
Third, it remains unclear why ‘mortar’ should be a ‘vulgar’ noun.
Therefore, I dismiss this etymology due to these phonetic and
morphological problems.
211
Kim Cwuwen has presented persuasive argumentation that Ceycwu /yo/ in fact reflects
not PK *yo, but PK *ya (Kim Cwuwen 1994: 275-296).
212
A ghost form, as far as I can tell.
213
A typographical error for MK hònàh ‘one’ (LCT 1987: 730; Nam 1997: 1450).
214
Actually, the correct Middle Korean form is hwòGwàk (LL, not LH) (Hwungmwong II:
11a). See also LCT (1987: 756) and Nam (1997: 1416), which also treat MK hwòGwàk as
LL.
188———Koreo-Japonica

(241) (R) MK hèlí ‘waist, small of the back’ ~ OJ sö-/se ‘(anatomical)

back’ < PJK *šer. Whitman adds: “OJ sö- (in compounds) is an exception
to the normal vowel alternation of nouns in /-ë/, where /a-/ ~ /ë/ is expected.
This exception is explained by the fact that the original Proto–Japanese-
Korean vowel was *e” (1985: 236). However, today we know that WOJ /ë/
had two different sources: *ai (more frequent) and *öi (less frequent), so
the Proto-Japonic vowel is clearly *ö. This creates a problem for the
regularity of the phonetic correspondences, since according to Whitman’s
rules WOJ /ö/ corresponds to MK /e/ only when found in final position or
in a word with another /ö/ (Whitman 1985: 129). However, WOJ /ö/ is not
found in final position in this word, because it is always followed by the
second element of a compound. In addition, the etymology rests on the
assumption of *-r- loss in Old Japanese, which cannot be verified
internally. This leaves us only with the correspondence MK /h-/ : OJ /s-/,
which this etymology is supposed to prove. However, since everything else
is problematic, I reject this etymology.
(242) (L) MK hyé ‘tongue’ < hyel (Yi Kimun 1959 215) ~ OJ sita LL <
PJK *šita or *šito 216 (Whitman 1985: 236). I have demonstrated elsewhere
that the Early Middle Korean word for ‘tongue’ that is transcribed in
Kyeylim yusa (#169) as 蝎 (LMC xiat) should be hyet with final /t/ rather
than /l/ (Vovin 2000: 145-146). The correspondence of MK /ye/ to OJ /i/ is
irregular, as mentioned in (227) above. Similar to the case of MK :syem ~
OJ sima ‘island’ (227), cognates of WOJ sita ‘tongue’ (not attested in
Eastern Old Japanese) are found in Ryukyuan, although cognates of WOJ
sita ‘tongue’ are not attested throughout the Ryukyus, but in multiple
locations in the Southern Ryukyus (Hirayama 1966: 365; 1967: 363),
making the Proto-Japonic status for *sita beyond any doubt. Since the
correspondence of MK /ye/ to WOJ /i/ is irregular, I suspect that similar to
WOJ sima ‘island’, WOJ sita ‘tongue’ represents an early loan from
Korean. 217 WOJ /s-/ in this word is likely to be explained as a
palatalization of Korean /h-/ preceding /i/, a phenomenon which is
widespread in northeast Asia well beyond Japonic and Korean. Such
Korean dialect forms as settε, settεgi, and settïi ‘tongue’ (Choy 1978: 426)
exhibit the same kind of palatalization. Therefore, we do not need to
reconstruct PJK *š- here.
(243) (R) MK hùn-hò- ‘is plentiful’ < hùn + hò- ‘do’ ~ OJ sön- ‘id’. in
söNtar ‘is plentifully adequate’ < sön + tar- ‘is adequate’; sönapë-
‘completely supplies’, sönapar- ‘is completely supplied’ < sön + apë-
‘joins it, makes it meet’ < PJK *sön (Whitman 1985: 236). There are
problems with this comparison. First, assigning the meaning ‘adequate’ to
*sön- in söNtar- is speculative at best; see other suggestions listed in

215
Probably a typographical error for Yi Kimun (1957: 399-403).
216
For the reconstruction of the vocalism in the first syllable, cf. (227) above and (262)
below.
217
Notice, though, that the parallelism is incomplete: MK :syem ‘island’ has R pitch,
indicating disyllabic structure, while MK hyé ‘tongue’, with H pitch, provides no internal
evidence for disyllabic structure. Both OJ sima ‘island’ and sita ‘tongue’ are LL (2.3).
 Lexical Comparisons———189

Martin (1987: 754), which are also speculative. One also should not forget
that WOJ söNtar- is a hapax legomenon, attested phonetically only once in
the Bussoku seki ka:

mîsô-ti amar-i puta-tu n-ö katati yasô kusa tö söNtar-er-u pîtô

thirty-CL exceed-INF two-CL DV-ATTR mark eighty type DV be
complete-PROG-ATTR person
a person, who is complete with thirty-two marks and eighty [lesser sign]
types (BS 2)

Luckily for us, this text refers to the very well-known definition of the
Buddha, who is said to have been endowed with thirty-two marks of
wisdom and eighty of noble birth. Thus, Whitman’s gloss ‘plentifully
adequate’ may be not quite appropriate. ‘Plentifully complete’ or
‘plentifully supplied’ is probably closer. Finally, we have no real internal
evidence for segmenting WOJ söNtar- into *sön-tar-, and the same applies
to sönapë- ‘prepare, make offering to the gods’ (JDB 1967: 403) as well.
Phonetic attestations of WOJ sönapë- are wanting, so writing /ö/ in this
verb is based on sönapar- ‘be supplied’, which has a single Western Old
Japanese phonetic attestation that is, of course, a hapax legomenon (JDB
1967: 402). As if the issue were not muddied enough, there is also WOJ
sônap- with kō-rui /ô/, which is not a hapax legomenon, but whose
meaning is not quite clear: it probably means ‘be completely supplied’
(JDB 1967: 402). None of these words is attested in Eastern Old Japanese,
but there are attestations in the Southern Ryukyus: Psara sunairü (Shimoji
1979: 106), Yaeyama sunairuN (Miyara 1980: 345). Therefore, there must
be a Proto-Japonic etymon, but it is unclear how to reconstruct it. Thus,
several problems on the Japonic side invoke reasonable doubt in the
reliability of the comparison with Korean. Second, MK hùn-hò- ‘is
plentiful’ also appears to be problematic, as it has a very un-Korean look:
derived verbs with the dummy verb ho- ‘to do’ usually have roots of
Chinese origin. I believe that such is the case here, although Korean
dictionaries normally treat the word as native. I suspect that MK hùn is just
the Sino-Korean reading of the Chinese character 很 ‘very, extreme’.
Therefore, I reject this etymology.
(244) (R) MK húy-/hóy- ‘white’ ~ OJ sirô-(si) LL, sira- LL (in
compounds) ‘id’. < PJK *šori (+rV) (Whitman 1985: 236). We have
already seen a different comparison of OJ sirô- ‘white’ with MK :syey-
‘white, it whitens’ (see [228] above). Methodologically it is impossible for
the same Old Japanese word to correspond genetically to two different
Middle Korean words, but since I have already rejected the comparison
with MK :syey- in (228), I am going to give this comparison the benefit of
the doubt, and evaluate it on its own merits. However, a correspondence of
either MK /o/ or /u/ to OJ /i/ is irregular, but the assumption can be made
that the same kind of metathesis occurred in MK hóy- < *hyo-, as
described for MK hóy- ‘do’, compared with OJ se- ‘do’ in (237) above.
But there is no internal Korean evidence for this metathesis, and a simple
190———Koreo-Japonica

comparison of Middle Korean (including hypothesized Proto-Korean)

forms with Old Japanese (including hypothesized Proto-Japonic) forms
still reveals a lack of regularity in the correspondences:

Chart 31:
‘Do’ and ‘white’ in Korean and Japonic
Gloss MK PK OJ PJ
‘do’ hóy- *hyo- se- *sia-
‘white’ hóy- *hyo- sira- *sira-

In particular, it remains unclear why Japonic keeps an *-r- in the second

case, but loses it in the first. Therefore, I reject this etymology on the basis
of the irregularity in the correspondences.
In conclusion, a survey of Whitman’s etymologies in support of the PJK
*š- demonstrates that the reconstruction of this phoneme is not warranted.

3.2.11 *m-
(245) (R) MK màc- ‘goes to meet, receives it’ ~ OJ mat- ‘awaits it’ <
PJK *mac- (Whitman 1985: 236). 218 Phonetically the comparison is
impeccable. OJ mat- ‘to wait’ has reliable cognates amply attested in
Eastern Old Japanese and throughout Ryukyuan (Hirayama 1966: 398;
1967: 427), but the semantic side of the comparison is very weak:
‘waiting’ does not necessarily imply ‘receiving’ or ‘meeting’. Old
Japanese texts frequently offer evidence for ‘waiting in vain’. Therefore, I
reject this comparison.
(246) (R) MK màc- ‘is correct, matches (something), hits the mark’ ~
OJ ma- ‘the true one’ (nominal prefix), masa ‘correct(ness)’ < PJK *mac-
(Whitman 1985: 236). 219 As Whitman himself comments, -sa in OJ masa
is a nominalizing prefix (1985: 236). Thus, we have a morphemic
boundary in Old Japanese between ma- and -sa, but no evidence for the
same boundary in MK màc-. Even if we disregard this boundary, the
correspondence of MK -c- to OJ -t- in (245) above and of MK -c- to OJ -s-
in this etymology brings up the issue of regularity of correspondences:
certainly both cases cannot go back to PJK *mac-, as suggested by
Whitman. Therefore, I reject this etymology on the basis of its irregularity.
(247) (R) LMK mah, K cang-ma ‘long rain, rainy season’ ~ OJ ama- ~
amë LF ‘rain’ < PJK *o:mago (Whitman 1985: 236). Whitman’s LMK is
in fact Early Modern Korean. EMdK mah is a hapax legomenon attested
only in Yun Sentwo’s collection Kwosan yukwo (1791) (Nam 1997: 523;
LCT 1987: 296). Even if we place our faith in this single attestation of
Korean mah in the late eighteenth century, the final -h might be a suffix -h
found in a number of other nouns, therefore the reconstruction of PJK
*o:mago with -go is unwarranted. Moreover, given the enormous size of
the Middle Korean corpus, the lack of attestation of mah in Middle Korean

218
Also in Martin (1966), #139.
219
This comparison is also found in Martin (1966), #53.
 Lexical Comparisons———191

is puzzling. In addition, a correspondence of OJ am- to MK m- seems to be

supported by only one more example (see [251] below), which brings a
reasonable doubt to the correspondence. Finally, OJ ama-/amë ‘rain’
appears as the second element of compounds with initial s-, e.g., paru-
samë ‘spring rain’ (JDB 1967: 601), kô-samë ‘small rain’ (JDB 1967: 291).
These compounds introduce another problem to the comparison with
Korean, despite the fact that we do not know the exact value of this initial
consonant in Old Japanese, which surfaces only in compounds: 220 How do
we explain the loss of the whole first syllable *Ca- in Korean? All these
problems seem to cry for internal explanations before we can allow this
comparison.
(248) (R) MK :mal- negative imperative ~ OJ mana HL ‘id’.
(uninflected) < PJK *mara (Whitman 1985: 236). There are no phonetic
examples of mana attested in Old Japanese, although it does appear in
Middle Japanese (JDB 1967: 685; IKJ 1990: 1197). However, the fact that
MJ mana is essentially used as a prohibitive interjection, while MK :mal-
is a verb with a full set of forms, should create a reasonable doubt to the
possibility of a genetic relationship in the first place. This doubt is further
strengthened by the fact that MK -l- does not regularly correspond to OJ
-n-. The suggestion by Omodaka et al. that MJ mana is a hybrid of MC 勿
/mjut/ or MC 莫 /muw’/ plus the Old Japanese prohibitive particle -na
(JDB 1967: 685) should be rejected due to the phonetic differences
between the Middle Chinese forms and OJ /ma/ as well as the fact that OJ
-na is not really a ‘particle’. 221 There are no cognates of MJ mana in
Eastern Old Japanese or Ryukyuan, and this causes further doubt about the
archaic nature of the Middle Japanese word. It is also difficult to imagine a
specific Korean form from which it could be borrowed, so the comparison
is best rejected.
(249) (L) MK málh ‘stake, post’ ~ OJ mara LL ‘phallus’ < PJK *mara
(Whitman 1985: 236). A tentative etymology connecting WOJ mara
‘penis’ with WOJ mar- ‘to have a bowel movement, to have diarrhea’
(JDB 1967: 692) should be discarded as a folk etymology. Nevertheless,
WOJ mara ‘penis’ is not attested in Eastern Old Japanese, Middle
Japanese, and Ryukyuan. Therefore, it might have a connection with MK
málh ‘stake, post’, but only as a loanword due to its distribution in Japonic.
(250) (R) MK màlí/mèlí ‘head’ ~ OJ ma-/më L ‘eyes’ < PJK *mar or
*mer (Whitman 1985: 237). This comparison is dependent on the
assumption of *-r- loss in Japonic, which cannot be verified internally in
this case. In addition, MK màlí < *mati, as demonstrated by EMK 麻帝
/mati/ (Kyeylim #161). Therefore, MK -l- here is from PK *-t-, and
comparison with a hypothetical PJ *-r- is not applicable in any case.
220
Both PJ *hama- and *zama- have been suggested, but both appear to be speculative.
221
Traditional grammar treats -na as a negative imperative particle following the final
(shūshikei) forms of verbs, but this is apparently not true in light of the Ryukyuan evidence:
cf. Shuri final kachuN ‘writes’ < *kak-i-uN, with a negative imperative kak-una ‘do not
write’. The lack of palatalization in the latter form clearly shows that we have a Shuri (and
Old Japanese) inflective negative imperative form -una < PJ *-ona that directly follows
verbal roots.
192———Koreo-Japonica

Finally, the semantic side of the comparison, like some of Whitman’s other
comparisons involving body parts, is forced. Although cases of
compounding like ‘face’ = ‘nose’ + ‘eye’ are widely attested in the
languages of the world, especially in the languages of Southeast Asia, to
the best of my knowledge, semantic shifts like ‘head’ => ‘eye’ or ‘eye’ =>
‘head’ do not occur. See also (57) and (219) above. Therefore, I reject this
etymology.
(251) (R) MK más ‘flavor, food’ ~ OJ uma-(si) ‘flavorful, delicious’,
ama(-si) ‘sweet’. 222 Whitman notes: “The OJ a/o alternation points to
original *o. The MK verb is a nominalization in /-s/ from PJK *oma”
(1985: 237). Some minor corrections to Whitman’s data are necessary
before discussing this etymology at length. First, I cannot see any
alternation a/o in Old Japanese, as there is an alternation a/u, if it is an
alternation at all; quite possibly we are looking at two completely different
words here, since WOJ uma- is ‘good, excellent, delicious’ and WOJ ama-
is ‘sweet, delicious(?)’.223 In addition, it is not quite clear from the
comparative Japonic perspective, whether WOJ uma- goes back to PJ
*uma-, *oma-, or *muma- (Vovin 2005a: 57-58), so it is better left out of
the equation. Second, MK más is certainly not a nominalized verb, since
no verbal form is present in any variety of Korean. Thus, segmenting final
-s as a nominalizing suffix in MK más is more than questionable. Besides
these preliminary considerations, even if we concentrate just on the
comparison of WOJ ama- ‘sweet’ (which has cognates attested throughout
Ryukyuan) with MK más, there is the same irregular correspondence of
MK m- to OJ am- seen in (247) above. And there is the problem of
semantic discrepancy: ‘flavor, food’ vs. ‘sweet’. Although it is possible to
bridge this difference, given the other phonetic and morphological
problems that this comparison involves, it is best discarded.
(252) (L) MK màth ‘garden, plot of land for agricultural purposes’ ~ OJ
mati HL ‘id’. < PJK *mato (Whitman 1985: 237). First of all, WOJ mati
has another meaning, ‘measurement of land’, but this is phonetically
attested for the first time only in the Saibara uta (JDB 1967: 680), which
can be called an Old Japanese text only if we stretch the definition. ‘Plot of
land’ seems to be an archetypal meaning. There are no attestations in
Eastern Old Japanese. The meaning ‘section of a city, town’ is attested in
Central Japanese starting from Early Middle Japanese, and it survives up to
this day. In the latter meaning the word was also borrowed into virtually
every Ryukyuan dialect, but the secondary semantics, as well as the
phonetics, clearly indicate that it is a loan from Japanese. In the Southern
Ryukyus both Ikema and Ishigaki have matśi (Hirayama 1966: 313)
instead of *matsï, which should be expected if it were a real cognate.
Therefore, the etymology can be accepted, but only as a loan from Korean
into Japanese.

222
This comparison is also found in Martin (1966), #284.
223
There are no phonetic attestations of WOJ ama- in the meaning ‘tasty, delicious’ (JDB
1967: 42).
 Lexical Comparisons———193

(253) (L) MK mól ‘edible seaweed’ 224 ~ OJ mö, më ‘id’. 225 < PJK *mor.
Whitman adds: “This comparison assumes OJ mô. The two OJ forms are
interesting, since they show a variant without absorption of -i (PJK *mol >
pre-OJ *mura > OJ mô), and the one with it (*mol > *mal+i > më)” (1985:
237). There are multiple problems with this comparison. First, there is no
evidence in favor of WOJ mô vs. mö based on textual attestations: although
the word is written as 毛 /mô/ in several cases, none of them in texts that
faithfully preserve the distinction between /mô/ and /mö/. Omodaka et al.
avoid the identification and read the word as /mo/ with an unclear /o/ -type
vowel (JDB 1967: 737). Second, the alternation mo ~ më in Western Old
Japanese rather points towards pre-WOJ *mö, since WOJ /ë/ can go back
to *öi, but it cannot go back to *ôi. Third, Whitman’s comparison again
depends on postulating *-r- loss for Japanese, which cannot be verified
internally. Fourth, the suggested development of PJK *mol to pre-OJ
*mula is speculative: not only does it involve reconstruction of an
otherwise unproven *-r-, it also adds another unwarranted vowel *-a,
apparently based on the necessity of deriving WOJ *mô from *mua.
However, as we have seen above, the data do not warrant WOJ mô < *mua,
so the argument for reconstruction of pre-WOJ *mura is completely
circular. Fifth, if we decide that pre-WOJ *mö is the most likely form
underlying both WOJ mo and më, then MK /o/ no longer corresponds
regularly to WOJ /ö/ in Whitman’s system of vocalic correspondences
(Whitman 1985: 129). To the best of my knowledge, WOJ mo ~ më
‘[edible] seaweed’ does not have any cognates in Eastern Old Japanese,
and Ryukyuan. It is likely to be another loanword from Korean.
(254) (R) Discussed and rejected in (16) above.
(255) (R) MK máh ‘yam’ ~ OJ umö ‘id’. 226 < PJK *omogo. Whitman
comments: “This comparison assumes original [P]K *móh. Note that MK
/o/ does not survive in monosyllabic open syllable noun stems (including
those in -h)” (1985: 237). First, MK máh is a monosyllabic noun with a
closed syllable: the fact that final -h disappears in certain environments,
including positions where it is isolated, means that this noun does not
really have an open syllable. Second, MK /o/ perfectly ‘survives’ in MK tó
‘place’ and sò ‘thing, matter’, so the reconstruction of MK máh as *móh
appears to be unmotivated. Third, I see no grounds for Whitman’s /ö/ in his
WOJ umö: the word is attested phonetically quite late, so Omodaka et al.
wisely transcribed it as umo with an ‘unknown’ type of /o/ (JDB 1967:
133). Moreover, since there are no cases in which WOJ /u/ and /ö/ combine
within the same root, the word is likely to be *umô, with a kō-rui vowel
/ô/. 227 Fourth, in any case neither WOJ /ö/ nor WOJ /ô/ corresponds

224
As far as I can tell from the textual examples, MK mól is ‘seaweed’, but not necessarily
‘edible’ (Nam 1997: 627; LCT 1987: 289).
225
Not all types of seaweed designated by WOJ mo ~ më are edible, but many are.
226
Also ‘taro’ in Old Japanese.
227
Ryukyuan dialects also indicate PR *umo > Oku, Kijoka, Sesoko ʔumu; Henza, Kowan,
Maejima, Kumejima, etc. ʔmmu; Higashi Nakasone, Yonaha mm ̣(Uchima and Arakaki
2000: 366), therefore Japanese imo attested from Middle Japanese (IKJ 1990: 135) is
194———Koreo-Japonica

regularly to MK /a/ or /o/. Fifth, this etymology rests on the assumption of

a correspondence of WOJ /um-/ to MK /m-/. We have already seen two
cases above where WOJ /am-/ was claimed to correspond to MK /m-/ in
(247) and (251). However, the regularity of this correspondence becomes
even more questionable due to the present comparison, where we have not
WOJ /am-/, but WOJ /um-/: since in other etymologies Whitman suggests
a straightforward correspondence of OJ /m-/ to MK /m-/, the uncon-
ditioned loss of an initial vowel in Middle Korean is not convincing. See
also (265) below, where one more initial vowel is unconditionally lost in
Middle Korean before /m-/. Due to these multiple problems, I reject this
etymology.
(256) (R) Discussed and rejected in (16) above.
(257) (R) MK mòyn- (prenoun) ‘nothing but, lots, many, all’ ~ OJ mïna
‘all’ < PJK *moyn (Whitman 1985: 237). There is one serious problem
with this etymology. MK mòyn- does not appear to have the meaning ‘lots,
many, all’. It is rather ‘bare, nothing but, naked’, as one can deduce from
the numerous textual examples cited in Nam (1997: 636-637) and LCT
(1987: 295). This prenoun also appears rather late in Middle Korean: the
first attestation is around 1515 (Penwo I: 25). This late appearance and
exclusively adnominal usage suggests that MK mòyn is a secondary
development. I think it is the grammaticalized realis attributive form mòy-n
of the Middle Korean verb mòy- ‘to tie, to bind, to restrict, to confine’.
Therefore, any analogy with OJ mïna is superficial, so I reject this
etymology.
(258) (R) MK mòzòm ‘heart’ ~ OJ muna-/mune HL ‘chest, heart’ < PJK
*mosom (Whitman 1985: 237). MK /-z-/ < PK *-s- does not correspond
regularly to OJ /-n-/ < PJ *-n-, so this etymology can be safely rejected.
(259) (L) MK mwóm ‘body’ ~ OJ mu-/mï ‘id’. < PJK *mum (Whitman
1985: 237). 228 The etymology is phonetically impeccable, but it clearly
represents a loan from Korean into Japanese, because cognates of WOJ mï
‘body’ are not found in Ryukyuan. There is only one attestation in Eastern
Old Japanese (MYS XIV: 3485).
(260) (L) MK :mwoyh ‘mountain’ ~ OJ mure ‘id’. < PJK *mùrág(+ -i).
Whitman adds: “OJ mure mainly occurs in place names, and is identified
(JDB 1967: 730-731) as a possible Korean loan” (1985: 237). I agree with
Omodaka et al.: OJ mure ‘mountain’ is limited to Western Old Japanese,
with no attestations in Ryukyuan and Eastern Old Japanese; therefore it
can only be a loan. The -l- in the original Korean form is confirmed by MK
mwòlwó ‘mountain’, attested once in the commentary to the Yongpi
echenka (YP IV: 21b).
(261) (R) MK mèk- ‘embraces it, cherishes it’ ~ OJ mak- ‘embraces as a
wife’ < PJK *mek-. Whitman comments: “The glosses cited make this
comparison straightforward, but OJ mák- ‘enfold, roll up, encircle’ is
usually identified as cognate with mak- ‘embrace’. MK mèk- ‘eats it, takes

apparently a secondary form. Therefore, it is not possible to argue for original PJ *imö >
WOJ umo.
228
This comparison is also found in Martin (1966), #19.
 Lexical Comparisons———195

it’ is likely cognate with the MK verb and thus should complete the
semantic range of PJK *mek-” (Whitman 1985: 237). There are several
problems with this etymology. First, MK mèk- means ‘to take’ only in the
sense ‘to eat‘, no ‘embracing’ is involved, therefore, I fail to see any
connection with MK mèk-, which incidentally has nothing to do with
physical ‘embracing’ or ‘encircling’, either. It means ‘to embrace, to
harbor, to express, to have (a feeling), to wear (a smile)’. Cf. the following
examples:
syelp-kwo ayGwatp-un ptut-ul mek-e kaskasulwo sa-n-i noni
sad-GER resentful-REAL/ATTR feeling-ACC harbor-INF barely live-
REAL/ATTR-NML 1psQUOT
I say, [I] am sad, and harboring resentful feelings, [I] barely live (Sek VI:
5a)
nguysim-ul mek-wu-m-ul myen thi mwot ho-mye
doubt-ACC have-MOD-NML-ACC allow do/INF cannot do-GER
it cannot be allowed to have any doubts (Wenkak II.2.1: 49b)

nimkum-i wuzwum-ul mek-usy-a

king-NOM smile-ACC wear-HON-INF
the king, wearing a smile (Twusi cho XVI: 27a)

I think the semantic gap is too great to accept this etymology. Relying
on the English glosses, which can involve ‘embrace’ in both cases, is
rather misleading.
(262) (R) MK :mel- ‘far, distant, remote’ ~ OJ mara-/mare- HH (non-
inflected adjective) ‘rare, far off’ < PJK *mere/*mara. Whitman comments:
“The OJ adjective mare has the sense of ‘rare, seldom seen’, but the
original meaning of the gloss is clear from OJ marapîtö ‘visitor from afar’
< mara + pîtö ‘person’” (1985: 238). I wonder on what source Whitman
bases his definition ‘visitor from afar’. Omodaka et al. simply gloss it as ‘a
person who came to visit from outside/another place, guest’ (JDB 1967:
692), and the textual evidence I am aware of does not indicate distance.
WOJ marapîtö is certainly just ‘guest‘, lit. ‘a person [who is] rare[ly seen]’,
as one normally does not see guests every day. Therefore, the connection
between the Korean and the Japanese words is semantically forced, and I
reject this etymology.
(263) (L) MK :myel ‘lizard-tail’ (Smilax) ~ OJ mîra ‘leek’ (Allium
porrum) < PJK *mi:ra 229 (Whitman 1985: 238). One of the most difficult
things in comparative linguistics is to correctly compare the names of
species of plants and animals so they do not become random comparisons
of unrelated flora and fauna. However, in this particular case, Whitman is
absolutely right: both plants not only belong to the Liliaceae family (Jpn.
229
It is not clear why Whitman reconstructs long PJK *i: in this case, but does not in two
other cases that have the same irregular correspondence of MK /ye/ to OJ /î/ or /i/. Cf. (227)
and (242) above.
196———Koreo-Japonica

Yuri-ka), but also to the Alliaceae genus (Jpn. Negi-zoku), in spite of their
somewhat confusing definitions in the bilingual Japanese-English and
Korean-Japanese dictionaries. This fact, however, does not provide us with
a guarantee that we are dealing with a Koreo-Japonic cognate here. The
correspondence of MK /ye/ to WOJ /î/ is irregular. WOJ 230 mîra ‘leek’ also
appears as mira and nira in Middle Japanese (JDB 1967: 718), with the
latter form surviving to this day. It is also attested in the Central Ryukyus:
Shuri ciri-bira (RGJ 1976: 162), Nakijin ciribiraa (Nakasone 1983: 285);
and in the South Ryukyus: Aragusuku bï:dza, Ishigaki bi:ra, Hateruma
bïra, Taketomi, Kurojima, Hatoma b’ira, Yonaguni n:da: (Miyara 1981:
217). The distribution points to a Koreo-Japonic cognate, but the problem
here is in the irregular correspondences of not just the initial consonant
(with m- ~ n- variation in Japanese and b- ~ n- in Ryukyuan), but also with
the reflexes of the vowel in the first syllable, where only Aragusuku and
Hateruma /ï/ seem to correspond regularly to WOJ /î/. The other South
Ryukyuan data do not. Due to these irregularities I believe that this is
another early loan from Korean to Japanese, which subsequently spread to
Ryukyuan.
(264) (R) MK myènól, myènólí 231‘wife, son’s wife, daughter-in-law’;
also ámh ‘female’ ~ OJ mê, mîna ‘woman‘. Whitman comments: “OJ mîna
appears in womîna ‘young woman’ < wo ‘little’ + mîna and omîna ‘old
woman’ < o (? ‘great’) + mîna. Both of these forms have alternate
traditional readings womuna and omuna, suggesting original PJK *myonor
‘woman’.… I suggested in (3.1.5.2.) [Whitman 1985: 152 — A.V.] that
MK myènól represents something close to the original form with J mê
derived through loss of medial *-n- from a protoform of the shape *mi:nar.
Another possibility suggested by MK ámh is a protoform *mi:ga, with
breaking of *i and metathesis in K[orean] to give MK ámh” (Whitman
1985: 238). However, this etymology is beset by multiple problems. First,
through a standard application of the comparative method, OJ mê ‘woman’
cannot possibly be related to both MK myènólí ‘daughter-in-law’ and ámh
‘female’ at the same time: only one can be a genuine cognate. So one can
discuss these two etymological proposals separately, but not together. The
relationship to MK ámh ‘female (of animals)’ involves a metathesis, which
cannot be cogently demonstrated on the basis of the internal Korean data,
and in addition, MK -h is in all likelihood just a suffix there. Thus, this
etymology can be rejected from the start. Moving on to Whitman’s second
proposal, MK myènólí does not mean ‘wife’ at all, but only ‘daughter-in-
law (son’s wife)’ (LCT 1987: 316; Nam 1997: 557). Thus the meaning of
‘wife’ cited by Whitman seems to be ‘imported’ to improve the
comparison with OJ mê ‘woman, wife’. Third, omina ‘old woman’ does
not appear in phonetic attestations until the Heian period, and omuna ‘id’.
appears only as a late kana gloss for the word written semantographically
in the Nihonshoki (JDB 1967: 164). In any case, the contraction opo- ‘big,
great’ > o- is not feasible for Old Japanese from the viewpoint of historical
230
No attestations in Eastern Old Japanese.
231
Also MK myènúlí (LCT 1987: 316; Nam 1997: 557).
 Lexical Comparisons———197

phonology, so neither MJ omina nor omuna can be used as reliable

evidence for the segmentation of -mîna in OJ womîna ‘[young] woman’.
Likewise, womuna ‘woman’ is not attested in Old Japanese, and appears
for the first time only in Middle Japanese. 232 In addition, OJ -mê ‘woman’
is attested as such as the second element in compounds, e.g., wotö-mê
‘young woman’, making Whitman’s claim that OJ -mîna is an archaic form
of OJ mê, preserved as such without undergoing *-n- loss in the compound
wo-mîna, even more doubtful: most likely, the two forms are completely
unrelated. Finally, the suggested *-n- loss (Whitman 1985: 148-152) is
very speculative. In the case of the two other etymologies that allegedly
include it (see [5] and [236] above), the Japonic forms at least include
another *-r- to which allegedly the preceding *-n- is assimilated before
being lost according to *-r- loss. In this case we do not have any evidence
for a Japonic word containing *-r- at all. Due to all these problems, I reject
this etymology.
(265) (R) MK múkèp- ‘heavy’ ~ OJ omo-(si) ‘id’. < PJK *ömö-.
Whitman notes: “MK -kap-/-kep- ‘like’ forms adjectives from noun,
adjective, and verb stems. OJ omosi- is a so-called ‘si-adjective’ where the
adjectival suffix -si has been incorporated into the original stem. The fact
that omo- is the original stem is attested by MJ omor- ‘get heavy, weigh’,
omori ‘weight’” (1985: 238). 233 First of all, OJ omo- is not a ‘si-adjective’:
although Old Japanese phonetic attestations are unfortunately lacking,
omo-kî ‘heavy-ATTR’ in OMO-KÎ UMA-NI-ni ‘to the heavy horse load’
(MYS V: 897) can be read only as omo-kî, not as omosi-kî from
considerations of poetic meter. This can be further supported by later
Middle Japanese phonetic attestations. This fact actually improves
Whitman’s comparison, but the phonetic problem of unconditioned vowel
loss in the Middle Korean form, which has already been addressed in (247),
(251), and (255) above, remains. Another problem is that, while the
adjective-forming Middle Korean suffix -kàp-/-kèp- is mentioned in the
traditional grammar (LCT 1987: 30, 49; Nam 1997: 35, 68), it likely
presents a misanalysis, since the suffix in question should be -àp-/-èp-, 234
as correctly analyzed by Martin (1992: 482). Thus the Middle Korean root
should be not *mu-, but muk- ‘heavy’ < PK *munk-, which, of course,
cannot be compared to OJ and MJ omo- ‘heavy’ for obvious phonetic
reasons. Therefore, I reject this etymology; cf. (320) below where OJ asi-
‘bad’ (which is a -si- adjective) is treated as if it were a root.
(266) (R) MK múl ‘water’ ~ OJ mî-/mîna/mîNtu ‘id’. < PJK *mör
(Whitman 1985: 238). 235 This etymology is attractive, since it involves
such a basic notion and, unlike many others, is semantically
straightforward. However, there is also a grave problem of phonetic

232
Omodaka et al. provide womuna-mê ‘woman, wife’ (JDB 1967: 840), but the form is not
attested phonetically before Middle Japanese.
233
This comparison is also found in Martin (1966), #103.
234
More exactly, -àW-, -èW-.
235
This comparison is also found in Martin (1966), #257.
198———Koreo-Japonica

irregularity: WOJ /î/ < PJ *e 236 does not correspond regularly to MK /u/,
according to Whitman’s own vocalic correspondences (Whitman 1985:
129). This irregularity becomes apparent if one compares it to (33) above,
where the correspondences are regular, or even to (265) above and (267)
below, where, in spite of other problems, the regularity of vocalic
correspondences is maintained. Since MK múl ‘water’ includes three
segments, and OJ mî- includes only two, of which only initial /m-/
represents a regular correspondence, I believe that this comparison, at the
present stage of our knowledge, has no better chance of being correct than
a comparison of either word with Hebrew mayim ‘water’. Thus, I reject
this etymology on the basis of phonetic irregularity.
(267) (R) MK mùl- ‘bites (of animals)’ ~ OJ mor- ‘(birds) pick, pluck
berries’ < PJK *mör-. Phonetically this etymology works only if WOJ
mor- is from *mör-, but since there is only one attestation in late Western
Old Japanese (MYS XVI: 3872), it is impossible to tell whether the word is
pre-OJ *mör- or *môr-, making the possibility that this etymology is
accurate only 50 percent. Later attestations are also quite restricted: there is
only one late Middle Japanese attestation in the Hōjōki (KKJ 1969: 878;
JDB 1967: 752), and the word seems to be otherwise restricted to the
modern Japanese dialects of Western Honshū and Shikoku (JDB 1967:
752). The prospects of its being an inheritance even from Proto-Japanese,
not to mention Proto-Japonic, are more than slim. Given the semantic
discrepancy (no bird can bite) I believe that the best solution is to reject
this etymology.
(268) (R) MK mùlùl- 237 ‘it returns, goes back’; also transitive ‘returns
it’ ~ OJ mï ‘turning’, mö- ‘turns, goes around’ < PJK *mör-. 238 Whitman
comments: “Unger (1977) demonstrates the identity of the stem vowel in
OJ mö-, citing compounds such as mötöpor- ‘goes around, goes by’ < mö-
+ töpor- ‘passes’” (1985: 238). There are problems with both the Korean
and Japonic suggested cognates. First, MK mùlù-/mùlùl- rather means ‘to
retreat, to go back, to turn around’ (LCT 1987: 337; Nam 1997: 597).
Second, its comparison with Japonic largely rests on the assumption of *-r-
loss in Old Japanese, which cannot be verified internally. Third, WOJ mï-
means ‘to go around [a certain object]’. The reconstruction of the stem of
WOJ mï- as *mö- with the vowel /ö/ rather than /u/ on the basis of mö- in
WOJ mötöpör- is doubtful, because the first element in a verbal compound
in Old Japanese appears in its infinitive form, and not the stem form. Thus
*mïtöpör- < *mö/ü-î-töpör- should have been expected, not mötöpor-. Since
neither WOJ mï- and mötöpör- is attested in other branches of Japonic,
they could be loanwords from Korean, but the lack of internal Japonic
evidence and the semantic difference favor rejecting this etymology.
(269) (R) MK mùlùl- ‘ripens and gets soft’ ~ OJ moye-/moyï- ‘sprouts,
buds’ < PJK *mölör-. Whitman comments: “Although most often attested

236
Proto-Ryukyuan has clearly *medu ‘water’, amply supported by attestations throughout
the Ryukyus, e.g., Koniya mït, Iejima midzi, Psara, Hateruma midzï (Hirayama 1966: 343).
237
More exactly, MK mùlù-/mùlùl-, a verb belonging to accent Class 8 (Ramsey 1991: 236).
238
This comparison is also found in Martin (1966), #183.
 Lexical Comparisons———199

as a thematic stem in -ë- (Whitman’s -A- — A.V.), the continuative form

[moi] < moyU (+ -i) (continuative) appears in MYS [XVIII 239]: 4111. Here,
Martin (1966) cites OJ moro ‘fragile, easily broken’ 240 (Whitman 1985:
238-339). A quick comparison with (268) above reveals a serious problem
in the regularity of the correspondences, because while the Middle Korean
forms are the same, the Old Japanese forms are very different. Since (268)
was rejected, we have to judge this etymology on its own terms, but the
problem in the regularity of correspondences still remains: MK /-l-/ is not a
regular correspondence to OJ /-y-/ according to Whitman’s system of
consonantal correspondences (Whitman 1985: 183-185). We also do not
know whether OJ moye- was from *möye- or *môye-: in the second case
there would be one more irregular correspondence, which further
diminishes the credibility of this comparison. The continuative moyi-
indeed appears in (MYS XVIII: 4111), but it is a hapax legomenon.
Relying on a hapax legomenon to reconstruct the vowel /ö/ in the second
syllable of the Japonic verb is more than questionable. In addition, even
the hapax legomenon moyi- does not indicate that the form was actually
*möyö-i (necessary for the comparison to survive) rather than *môyu-i. All
of these problems, in combination with the very questionable semantics,
lead me to reject this etymology.
(270) (R) LMK 241 mulus ‘all, in general’ ~ OJ morö HH (prenoun) ‘all,
the bunch of’ < PJK *mörö (+ -s substantivizer) (Whitman 1985: 239).
There are three problems with this etymology. First, it is strange that
EMdK mulus is not attested in Middle Korean. The dialectal attestations in
Korean seem to be quite limited (Choy 1978: 1099). It is even stranger that
this word appears in the 1656 and 1736 reprint editions of the Nayhwung,
although apparently not in the original edition in 1475 (LCT 1987: 338;
Nam 1997: 598). Since I do not have access to the original edition at the
moment, I am not in a position to say what older Middle Korean word was
replaced by EMdK mulus in the reprints. In any case, such a late
appearance suggests that it is an innovation. Second, although a
substantivizer (or nominalizer) -s in Middle Korean certainly exists, it is
added to verbal stems, and in this case Middle Korean verb *mulu- with an
appropriate meaning does not occur. Thus, there is a clear problem with
the morphological segmentation. Third, I do not think that EMdK mulus
really means ‘all’: textual usages point to ‘in general’ or ‘generally
speaking‘. Given these problems, I reject this etymology.
(271) (R) MK mùsú- ‘what, which’ (prenoun); mùsés ‘what’ (noun) ~
OJ mosi HL (adverb; introduces a question, conjecture, or hypothetical
statement) < PJK *mösö- (Whitman 1985: 239). To the best of my
knowledge, WOJ mosi (not attested in Eastern Old Japanese) never

239
I have added the volume number.
240
Martin indeed introduces comparison with MdJ moro- ‘fragile’ (Martin 1966, #207), but
he does not mention that the word is attested in Old Japanese. As a matter of fact, it is likely
to be well attested only starting from Middle Japanese (KKJ 1969: 878), since the only Old
Japanese attestation in MYS V: 902 is not in phonetic script.
241
Whitman’s Late Middle Korean is an Early Modern Korean form.
200———Koreo-Japonica

introduces a question, but only a conjecture or hypothetical statement. This

certainly makes a comparison with MK mùsés ‘what’ much more
problematic from a functional point of view. 242 One would expect a much
more straightforward comparison if the languages were genuinely related,
cf. German was and Russian chto, both meaning ‘what’ and reflecting
quite regularly (in spite of their dissimilarity today) the same PIE *k΄we
‘what’. In addition, WOJ mosi can potentially go back not only to PJ
*mösö-i (the option chosen by Whitman), but also to *mösi, *môsi, and
*môsu-i. This leaves no more than 25 percent phonetic credibility for this
etymology. Finally, MK mùsúk and mùsúm ‘what’ include non-leniting -s-
< *-ns-, which cannot correspond regularly to WOJ -s-. Since there is also
a functional problem, as outlined above, it seems natural to reject this
etymology.
(272) (R) MK psí- ‘steams it’ ~ OJ mus- ‘id’. < PJK *mös-. Whitman
adds: “Loss of original vowel in MK” (1985: 239). Even given the loss of
the original vowel in Middle Korean (which can only be /u/, since MK
/wu/ does not undergo reduction, and since only MK /u/ can be a reflex of
PJK *ö), it is not quite clear why Whitman reconstructs the same Proto-
Japanese-Korean vowel *ö for the correspondence of MK /Ø/ < PK *u : OJ
/u/ in this example and for the correspondence of MK /u/ : OJ /ö/ in (267-
271) above. Clearly, according to Whitman’s correspondences, OJ /u/ does
not correspond to MK /u/ (Whitman 1985: 129). Even more important is
the fact that while the initial consonantal cluster *ms- is not attested in
Middle Korean, there is no internal Korean evidence suggesting that PK
*mVs- > MK ps-. In addition, there are no other correspondences of MK
ps- to OJ mus- presented by Whitman. Finally, this etymology also leaves
final -i in MK psí- unaccounted for in its Old Japanese comparison.
Therefore, multiple irregularity problems call for the rejection of this
etymology.
(273) (L) MK mwúl ‘crowd, throng, group’ ~ OJ mura-/mure HH ‘id’.
< PJK *mure (+ -i) (Whitman 1985: 239). Discussed in (53) above and
accepted as a loan from Korean into Japanese.
(274) (R) MK mwùth ‘dry land’ ~ OJ mutu (name of a province in the
far northeast of Honshū); mîti HH ‘road, area’ (Whitman 1985: 239). The
comparison with OJ mîti ‘road, way’ (the meaning ‘area’ is not attested at
all, but the expected semantic extension is sometimes ‘direction’ [JDB
1967: 705]) is untenable, because it is quite clear from compounds like OJ
ti-mata ‘road fork‘, Tagîma-ti ‘road to Tagima’, etc., that the original
Japanese word for ‘road, way’ is just *ti, and that mî-ti is a
morphologically complex word containing the beautification prefix mî-. I
am afraid that comparing an actual Middle Korean word with a Japanese
place name of unknown meaning, has very low credibility. Thus, I reject
this etymology.

242
In addition, MK mùsés appears to be a rather late form from the sixteenth century, while
MK mùsúk and mùsúm ‘what’ are attested earlier, in the fifteenth century (LCT 1987: 339;
Nam 1997: 599-600).
 Lexical Comparisons———201

(275) (R) MK :mil- ‘pushes it, shoves it, tide rises’ ~ OJ mît- ‘tide,
moon is full’ < PJK *mitür- (Whitman 1985: 239). There are two problems
with this comparison. First, MK :mil- is clearly from PK *mìlú-, but not
PK *mìtú- (with a leniting /-t-/), so there is an irregular correspondence of
MK /-l-/ < PK *-r- to OJ /-t-/. Second, to the best of my knowledge,
MK :mil- means only ‘to push, to shove, to put off, to delay’ (LCT 1987:
347; Nam 1997: 616). Certainly, there is MK :ilmúl ‘tide’, attested as early
as (YP 67, 69), but this can easily be analyzed as :mil-múl ‘pushing water’.
Omodaka et al. note that there are many cases when OJ mît- ‘to be full’
refers to a tide or to the moon, but it is apparent from the examples they
cite that, at least in Western Old Japanese, 243 its usage is not limited to
these two cases (JDB 1967: 707). Thus, even the semantic side of the
comparison appears forced. The phonetic irregularity and the semantic
differences seem to be sufficient to reject this comparison.
(276) (R) MK mìlù ‘dragon’ ~ OJ mï ‘serpent’ (Whitman 1985: 239-
240). Whitman adds: “The vowel correspondence is at first glance irregular.
I have included this comparison because it may cast further light on the
sources of OJ /ï/. We would expect something like MK *muli; note that the
Aen kakpi (1819) gives mili, probably from milu + -i > miluy > mili. This
derivation suggests a parallel (although historically unrelated) source for
OJ mï: medial *r loss from PJK *mirö would give *myö > pre-OJ *mö >
*mö + -i (nominative) > mï. Note that in contrast to OJ pëmî, mï refers to
mythical serpents and the snake in the twelve-animal cycle. OJ mïtuti
‘serpent spirit’ is glossed as 蛟 ‘dragon’, 244 龍名 ‘name of the dragon’ in
the Shinsen jikyō’” (Whitman 1985: 239-240). There are several problems
with this comparison. First, the vowel correspondence is indeed irregular,
and suggesting another unique hypothetic development for OJ /ï/, which
occurs only in this case does not help, but rather further diminishes the
credibility of the comparison. Second, MJ mituti ‘water dragon’ is not
attested phonetically in Old Japanese, and the suggestion that it is a
compound, consisting of mï ‘snake’ + -tu genitive-locative + ti ‘spirit’ is
just a kokugogaku (traditional Japanese philology) etymology provided in
JDB 1967: 706. 245 Thus, we cannot ascertain that the first syllable of this
word is /mï/, and, consequently that MJ mituti ‘water dragon’ has any
connection with OJ mï. Third, if we put aside the alleged connection with
MJ mituti, the only usage attested for WOJ mï is ‘snake’ in the twelve-year
animal cycle. It is also the only usage found in Middle Japanese (KKJ
1969: 820). I see no philological basis for assigning the meaning ‘serpent’
to this word. Although the comparison of a ‘dragon’ with a ‘snake’ is not
impossible, the very fact that WOJ mï is used only as an animal cycle word,
suggests that it means just ‘snake’, and not ‘dragon’, because the dragon is

243
Within the Eastern Old Japanese corpus, mit- is attested only in MYS (XIV: 3366, 3549).
The last of these two poems does not include any typical Eastern Old Japanese features.
244
Actually, ‘sea serpent’ or ‘water dragon’.
245
Another etymology suggests mî- ‘water’ as the first element. Although this also has
problems, it is preferable because of the meaning ‘water dragon’, and also because WOJ -tu
indicates belonging to some location (see above in the chapter on morphology).
202———Koreo-Japonica

already present in the animal cycle list. Finally, the comparison again
depends on the assumption of *-r- loss in Old Japanese, which in this case
cannot be verified independently. Therefore, I reject this etymology.
(277) (L) MK mìs, mìt, K mich ‘and, with, also’ < mich- ‘reaches, is
equal to’ ~ OJ mîta~muta ‘and, with’ (noun preceded by noun in the
genitive) < PJK *mito (Whitman 1985: 240). I disagree with Whitman’s
derivation of MK mìs 246 ‘with, and, also’ from MK mìch- ‘to reach’; a bare
verbal root is unlikely to be grammaticalized. Therefore, I will deal only
with MK mìs. There are problems on the Japonic side of the etymology as
well, as only muta occurs in Western Old Japanese (JDB 1967: 726), while
mita is attested once, and only once, in Eastern Old Japanese (MYS XX:
4394). In addition, this hapax legomenon is not without its own
philological problems, since a number of manuscripts have mini instead of
mita (JDB 1967: 703-704). The prevailing point of view among Japanese
philologists is that mini is a mistake for mita, and that mita should be taken
as the Eastern Old Japanese form of WOJ muta (Takagi et al. 1962: 434;
JDB 1967: 703-704; Mizushima 1974: 216; Mizushima 2003: 628-629;
Nakanishi 1984: 1461; Kojima et al. 1975: 404; Omodaka 1974.20: 127;
Satake et al. 2003: 424). I will accept this point of view, since mini does
not make any sense in the context of the poem:
yumî-nö mîta sa-ne ka watar-am-u naNka-kë könö yô-wo
bow-GEN with PREF-sleep PT cross-TENT-ATTR long-ATTR this night-
ACC
Will [I] pass this long night with [my] bow? (MYS XX: 4394)

What is more important is that, to the best of my knowledge, there are

no cognates of mîta or muta in Ryukyuan. Thus, due to distribution criteria,
the Western and Eastern Old Japanese forms are likely to be loanwords
from Korean. I assume that EOJ mita reflects a more archaic form, while
WOJ muta underwent vowel labialization after a labial.
(278) (L) MK míth 247‘base, bottom’ ~ OJ mötö LL ‘id’. < PJK *mötö.
Whitman adds that he cannot account for the Middle Korean vowel
(Whitman 1985: 240). The correspondence is indeed irregular, and both
the Middle Korean and Japanese words also mean ‘root’. This is important,
because Old Japanese also has a doublet ne ‘root’. While OJ ne ‘root’ is
attested throughout Japonic, WOJ mötö has much more limited attestations
of its cognates. Besides EOJ moto, there are scattered attestations in
Ryukyuan, but those found in Miyako and Yaeyama in the Southern
Ryukyus involve mutu attested as a classifier, not as an independent word.
This indicates that it must be a loan into Ryukyuan from Japanese.
Therefore, I treat this etymology as another loanword from Korean to
Japanese.

246
Dictionaries provide different accentuation: mís (Nam 1997: 618), mìs (LCT 1987: 348).
I am unable to verify which one is wrong, because I do not have access to the texts. The
form mit is attested only in Early Modern Korean (Nam 1997: 615; LCT 1987: 347).
247
The correct accentuation is with L pitch: mìth (LCT 1987: 349; Nam 1997: 615).
 Lexical Comparisons———203

Whitman’s comparisons involving Proto-Japanese-Korean initial *m-

(245-278) do not contain any good etymologies (apart from loanwords)
that could support the hypothesis of the Koreo-Japonic genetic relationship.
Therefore, I believe, we have another significant gap in addition to those
already mentioned. The absence of good etymologies with *m- certainly
speaks in favor of a loanword, rather than a genetic, relationship.

3.2.12 *n-
(279) (R) MK -n/-ún/-ón (past/perfective attributive suffix) ~ OJ -n
(perfective suffix) < PJK *-n (Whitman 1985: 240). This comparison was
discussed above in the section on verbal morphology (2.3.2.9) and rejected.
(280) (R) MK ná, first person pronoun ~ OJ na, second person pronoun
< PJK *na/*ne. Whitman adds: “OJ na occurs as a second person pronoun,
although it is widely hypothesized that it originated as a first person
pronoun. The Old Japanese pronoun is a perfect comparison in sound and
meaning to MK né (second person pronoun). Noteworthy in this regard are
the following facts: (1) the Middle Korean first and second person pronoun
are front/back vocalic alternants of each other; (2) J pronouns have
frequently undergone the shift from first person/reflexive to (intimate)
second person: thus onöre (first person pronoun, ‘self’) => second person
(derogatory)” (1985: 240). This comparison presents a number of
significant problems. First, MK nè ‘thou’ as a comparison with OJ na ‘id’.
is found again in (298) below, this time with a L accent. 248 This seemingly
goes along with Whitman’s line of argumentation, presented above, but the
line itself is faulty, as the reader will shortly see below. Second, I am not
aware of any ‘wide hypothesizing’ that OJ na ‘thou’ originated from OJ na
‘I’ outside the kokugogaku tradition, e.g., JDB 1967: 512. It is quite clear
that OJ na ‘I’ (not attested anywhere else in Japonic and short-lived even
in Old Japanese) is a straightforward loan from Korean (see the brief
discussion in the pronouns section of the morphology chapter above and a
more detailed discussion in Vovin [2005a: 245-246]). Third, Whitman’s
argument that OJ onö-/onöre ‘self’ was an exclusive first person reflexive
pronoun that shifted to being a second person intimate/derogatory pronoun
goes against the textual evidence. As the following examples clearly
demonstrate, WOJ onö-/onöre is clearly just a reflexive pronoun, not
limited to the first person: 249

onö-Nka wo-wo nusum-î si-se-m-u tö

yourself-POSS cord-ACC steal-INF die-CAUS-TENT-FIN DV
[they] are going to steal your own [life-]cord and murder [you] (KK 22)

248
Whitman does mention that the accent of MK nè ‘thou’ is problematic (1985: 242, #298).
Due to the extremely irregular variations of accent in personal pronouns (Yi Swungnyeng
1961: 220-221; Kim Wancin 1973: 61-66), I believe that it is not just problematic, but
almost impossible to reconstruct.
249
For more examples, also including Eastern Old Japanese, see Vovin (2005a: 266-269).
204———Koreo-Japonica

onö-Nka MIY si itapasi-kêre-Npa

myself-POSS body PT ill-EV-CON
as my own body became ill (MYS V: 886)

onö-Nka op-êr-u onö-Nka NA OP-Î-te

himself-POSS carry-PROG-ATTR himself-POSS name carry-INF-SUB
[every one of] themselves carrying their own names (MYS XVIII: 4098)

Therefore, there is no evidence for a shift from a first person pronoun to

a second pronoun within Japonic. Fourth, OJ na ‘thou’ is also a loan from
Korean; see (298) below and the more detailed discussion in the pronoun
section of the morphology chapter above. Fifth, in spite of the fact that MK
na ‘I’ and ne ‘thou’ look like [+back]/[-back] alternants of the same word,
there is no internal Korean evidence to support a shift from a first person
pronoun to a second person pronoun, with the first person pronoun
remaining intact and not replaced by something else. Sixth, this is not the
only possible speculative way to analyze first and second person pronouns
that start from the same consonant: typological evidence from Chukchee-
Koryak, which has *g-ə-m ‘I’ and *g-ə-t ‘thou’ suggests that MK na ‘I’
and ne ‘thou’ can also be analyzed as *n-a and *n-e, where *-a and *-e
represent real personal pronouns, and the initial element *n- some kind of
pronominal marker. All these problems, as well as the fact that Whitman
also compares MK ne ‘thou’ and OJ na ‘id’. below in (298), strongly call
for the rejection of this etymology.
(281) (R) MK -ná (adversative) ‘(whether)… or, although’ ~ OJ -na
‘whether… or, both… and’ (linking time expressions) < PJK *-na.
Whitman adds: “OJ -na links time expressions in phrases like asa-na yupu-
na ‘whether morning or night, both morning and night’” (Whitman 1985:
240). Even at first glance there is an apparent functional difference: MK
-ná is a verbal suffix, and OJ -na occurs only after nouns. But there is a
much more serious problem for Whitman’s comparison on the Japanese
side, because OJ -na 250 more frequently than not also occurs in cases that
cannot be treated as ‘adversative’: asa-na [a]sa-na 251 ‘in the mornings,
every morning’. Cf. the examples:

ISE-nö AMA-NÖ ASA-na YUPU-na-ni KANTUK-U

Ise-GEN fisherman-GEN morning-PLUR evening-PLUR-LOC 252 dive-
FIN
Fishermen of Ise dive [into the sea] in the mornings and in the evenings
(MYS XI: 2798)

250
The traditional explanation for this OJ -na is that it is a locative case marker. However, a
comparison with the examples below clearly demonstrates that it is not.
251
The uncontracted spelling asa-na asa-na is attested in Middle Japanese poetry, e.g., see
KKW 513.
252
If -na were a locative case marker, as tradition has it, it is incomprehensible why it
would have another locative case marker -ni after it, as in yupu-na-ni ‘in the evenings’. This
example proves that -na is not a locative case marker.
 Lexical Comparisons———205

naNtesikô-Nka pana n-i môNkamö na asa-na [a]sa-na MÎ-m-u

carnation-POSS flower DV-INF PT PT morning-PLUR morning-PLUR 253
see-TENT-FIN
[I] wish [you] were a carnation flower, so [I] would see [you] every
morning (MYS XVII: 4010)

asa-na [a]sa-na aNkar-u pîNpari n-i nar-i-te-si ka

morning-PLUR morning-PLUR rise-ATTR skylark DV-INF become-INF-
PERF-PAST/ATTR PT
every morning, [I would] have become a skylark, flying up (MYS XX:
4433)

I believe that this OJ -na is a plural marker. See Vovin (2005a: 102-107)
for a more detailed discussion and more examples. Thus, I reject this
etymology.
(282) (R) MK ná-254 ‘becomes, comes out, is born’ ~ OJ nar- ‘id’. 255 <
PJK *na-. This etymology faces two problems. The first is semantic: MK
ná- does not mean ‘to become’, it is simply ‘to exit, to go/come out, to be
born’ and OJ nar- does not mean ‘to go/come out’ or ‘to be born’, it just
means ‘to become’. Certainly, various speculative connections between
these two verbs can be proposed (Martin 1966: 226), but presenting them
as verbs with identical semantics means adjusting their actual meanings to
make the comparison work. The second problem is morphological: OJ -r-
in nar- does not correspond to anything at all in MK ná-, and this calls for
an explanation, since there are plenty of Middle Korean verbs that end in -l.
However, no cogent explanation for this fact is provided, and it does not
seem that one is even remotely possible. Given these problems, it is
necessary to reject this etymology.
(283) (R) MK nàh- ‘produces it, gives birth to it’ ~ OJ nas- ‘id’.
Whitman remarks: “Both MK and OJ verbs are original causative
derivations from PJK *na- ‘become, come into being’ (282)” (1985: 241).
First, MK -h- < *-k- does not correspond regularly to OJ -s-. Second, OJ
-s- in na-s- is clearly a causative marker, while MK -h- belongs to the root.
This is another case of adjusting the meaning to make the comparison
work: OJ nas- primarily means ‘to make, to make something to be
something else’, and the meaning ‘to give birth’ is clearly secondary (JDB
1967: 512). It is also somewhat speculative, because it is not attested
phonetically in Old Japanese or in any other variety of Japonic. Therefore,
I reject this etymology.
(284) (R) MK ánh ‘inside’ ~ OJ naka LH ‘id’. < PJK *na:ka (Whitman
1985: 241). This etymology requires a metathesis either in Korean or

253
More evidence against -na as a locative case marker comes from this and the next
example. To the best of my knowledge, there are no cases of reduplication in Old Japanese
that would have a locative case marker on both members of a reduplicated form.
254
MK na- ‘to go out’ belongs to accent class 3, where all verbs exhibit extremely irregular
accentuation. Determining MK pitch as H may be, therefore, premature.
255
This comparison is also found in Martin (1966), #11.
206———Koreo-Japonica

Japonic, but neither can be proven internally. In addition, there are three
examples in Western Old Japanese, where na occurs alone with the
meaning ‘inside’ (JDB 1967: 512). These suggest that OJ naka is a
compound consisting of na ‘inside’ + -ka ‘place’. Since -ka ‘place’ is well
attested in other compounds, such as sum-î-ka ‘place of living’, umî-N-ka
‘sea place’, oku-ka ‘deep place, place in the back’, the internal etymology
appears to be preferable to a speculative external explanation that involves
metathesis. Therefore, I reject this etymology.
(285) (R) MK nàlhwó- ‘slow, gradual, gentle’ ~ OJ naNkï- ‘quiets down,
still’, naNkô (uninflected adjective) ‘soft, gentle’, naNku(si) HL ‘peaceful,
placid’ < PJK *nargu- (Whitman 1985: 241). The first minor problem with
this etymology is in the semantics: as far as I can tell, MK nàlhwó- means
just ‘to be slow, to be late’ (LCT 1987: 137; Nam 1997: 267), while the
basic meaning of WOJ naNkï- is ‘to become quiet/tranquil’ (JDB 1967:
520). Second, although EOJ naNkoya ‘quiet’ is attested once in the Eastern
Old Japanese corpus (MYS XIV: 3499), there are no traces of this word in
Ryukyuan. Third, and most significantly, MK nàlhwó- < PK *nalokwo-,
because MK -h- reflects PK *-k- in intervocalic position. This creates a
significant problem for Whitman’s claim that MK -lC- corresponds to Old
Japanese ‘voiced’, i.e., prenasalized voiced obstruents (Whitman 1985:
183). Cf. also Whitman’s other comparisons above, where MK -lh-
corresponds as well to OJ -Nt- (15) or to OJ -s- (94). Therefore, although
the distribution could speak for an early loan from Korean into Japanese,
the problems in the regularity of the correspondences as well as in the
semantics lead me to reject this etymology.
(286) (R) MK nàmwò ~ nàmk- (< *nàmwòk) ‘tree’ ~ MJ 256 nagi ‘sacred
tree’ (Whitman 1985: 241). MJ and MdJ nagi is an evergreen tree of the
cypress type with leaves resembling bamboo. It was believed to have the
ability to divert calamities and the influences of evil forces, and this fact,
together with the limited attestation of the word in Japonic (only Central
Japanese), invites other explanations instead of hypotheses that the name
of this tree was directly inherited from a Koreo-Japonic proto-language, or
even that any Korean tree became ‘sacred’ on Japanese soil. The lack of
attestation in Old Japanese is particularly telling, and I believe that the
word has rather straightforward internal etymology: MJ nagi < *na-n[ö]-kï,
where the last syllable represents a fusion of genitive case marker -nö and
the word ki ‘tree’ (WOJ kï). The first syllable is the MJ na ‘ceremony for
expelling demons’ < MC *nan (儺). Therefore, I reject this etymology.
(287) (R) MK náp ‘lead’ ~ OJ namari LHL ‘id’. Whitman mentions
that this is originally a loan from Middle Chinese, but does not indicate the
source. He also refers to Koguryo, again without providing a source
(Whitman 1985: 241). The Chinese origin for the word for ‘lead’ in either
Korean or Japanese is problematic, since Mandarin qian1 does not
regularly reflect EMC *yen (鉛) ‘lead’, and neither could possibly be the
source for either MK náp or OJ namari ‘lead’. Pseudo-Koguryo *namur
256
Whitman has MK instead of MJ but this is probably a typographical error, as MK (=
Middle Korean) compared with MK (= Middle Korean) does not make any sense here.
 Lexical Comparisons———207

(乃勿) ‘lead’ does indeed resemble OJ namari, but it does not provide a
connection to MK náp ‘id’. The correspondence of MK -p- to OJ -m- is
problematic; in addition, the remaining segment -ari in OJ namari is left
unexplained. I reject this etymology as a chance similarity.
(288) (R) MK nàsk- ~ nàksk- ‘fishes’, náks ‘fishing’ ~ OJ na ‘fish (for
food)’. Whitman notes that “the MK root originates as a compound,
possibly from *na + tik- ‘catches with hook or spear’” (Whitman 1985:
241). There are several problems with this comparison. First, there is no
internal Korean evidence to justify the development *tik- > *sik- > *sk-
that is needed for this etymology. Second, it seems that MK nàsk- is a
simplification of MK nàksk-, since besides its appearance in the first
edition of the Twusi enhay, it occurs only in later texts (LCT 1987: 139;
Nam 1997: 272). The Middle Korean verb nàksk- is also supported by the
nominal form nàks ‘fishing’. MK nàks ‘fishing’ looks like a nominalized
form in -s from an unattested verb *nak-, but this still leaves the final *-k
in *nak- unexplained. Third, and most significant, OJ na does not mean
exclusively ‘fish’. This word refers to vegetables as well (JDB 1967: 511),
and the archetype meaning seems to be ‘fresh food used as a side dish’.257
As a good illustration of this point, note the plural marker -na written with
both 魚 ‘fish’ and 菜 ‘vegetable’ signs in the following example:
伊勢乃白水郎之朝魚夕菜尓潜
ISE-nö AMA-NÖ ASA-na YUPU-na-ni KANTUK-U
Ise-GEN fisherman-GEN morning-PLUR evening-PLUR-LOC dive-FIN
Fishermen of Ise dive [into the sea] in the mornings and in the evenings
(MYS XI: 2798)
I reject this etymology on the basis of its semantic, morphological, and
phonetic problems.
(289) (R) MK :nayh ‘river’ ~ OJ na- ‘water’ in namî ‘wave’, naNta
‘open sea, difficult crossing’, naNtuk- ‘be soaked in water, get wet’ < na-
+ tuk- ‘touch’ < PJK *na-. Whitman adds that ‘OJ namî ‘wave’ is often
assumed to be a compound of na + mî ‘water’ (Whitman 1985: 241). I
cannot really understand how the compounding of *na ‘water’ and mî
‘water’ is going to produce ‘wave’, and -Nta part in OJ naNta ‘open sea’ is
left unexplained altogether. I consider these two etymologies to be
teleological. The existence of OJ *na- ‘water’ cannot be completely ruled
out due to OJ naNtuk- ‘be soaked in water.’ This is likely to go back to
*na-ni tuk-, which allows us to segment *na ‘water’ on the basis of
structural analysis, although it remains somewhat speculative. However,
the comparison still remains unlikely due to the problem on the Korean
side: MK :nayh ‘river’ certainly goes back to PK *narih, as confirmed by
OK NAli (汀理, 川理)‘river’. Cf. the following examples:

257
Cf. MdJ sakana ‘fish’, which is historically a compound consisting of saka- ‘rice wine’
+ na ‘fish or vegetables used as a side dish for drinking’.
208———Koreo-Japonica

MWORGAY-i phall-on NAli-rang KIRANG-uy CUs-i isi-swo-ra

sand-NOM blue-RERF/ATTR river-LOC [Hwa]rang KI[pha]-GEN image-
NOM exist-EMPH-FIN
In the river, where the sand is dark (lit.: blue — dark part of the spectrum)
the image of [Hwa]rang Ki[pha] is [right there] 258 (Hyangka IV: 4)

ilGo NAli-s COYWYek-huy LANG-la TUL-Gi-ki TAW-Osi-wo-n

MOSOm-oy KOs[o]-γïl CWOS-no-ra-cy-e
IlGo river-GEN pebble-LOC [Hwa]rang-DAT hold-PASS-NML like-
HON-MOD-PERF/ATTR heart-GEN limit-ACC follow-PRES-?-?-INF
following the limits of [my] heart, that is like being held by the [Hwa]rang
in the riverbed (lit.: pebbles) of the river IlGo (Hyangka IV: 6)

This leaves the *-rih in PK *narih unaccounted for. Even if one accepts
Whitman’s *-r- loss law, the regularly corresponding word in Old Japanese
should be *ne < *nai, not *na. Therefore, I reject this etymology.
(290) (R) MK nòlí- ‘goes down’ ~ OJ ne- ‘lies down, sleeps’ < PJK
*nor- (Whitman 1985: 241). This etymology depends on Whitman’s *-r-
loss law to explain OJ ne- as pre-PJ *nari-, but in addition to the less than
perfect semantics (OJ ne- predominantly refers to sleeping or to a couple
sleeping together [JDB 1967: 551] rather than to the simple fact of lying
down, which is normally referred to by the Old Japanese verb pus-), it
appears that the root of the Old Japanese verb is just ne-. This leaves the
-li- segment in the suggested Middle Korean cognate unexplained, unless
one accepts the speculative *-r- loss. I reject this etymology on
morphological and semantic grounds.
(291) (R) MK nòlk- ~ nùlk- ‘old’ 259 ~ OJ naNka- ‘long’ < PJK *norg-.
Whitman remarks that OJ naNka- refers to length in both time and space
(1985: 241). In addition to the problematic semantics, the major problem
with this example is the lack of regularity in the correspondences:
Whitman claims elsewhere that OJ -n- corresponds to MK -lk-, and that
both are regular reflexes of a Proto-Japanese-Korean cluster *-lg- 260 (1985:
183); cf. examples (16), (238), and (254) above. In addition, in one more
case he presents an example where MK -lk- corresponds to OJ -k- (120).
Since the conditioning of these split correspondences is not explained, I
consider them all irregular. Finally, since the peculiar ablaut in the Middle
Korean forms is also left without explanation, it makes the correspon-
dences in the first syllable at least somewhat questionable. Therefore, I
reject this etymology.
(292) (R) MK nóm ‘person, other’ ~ OJ na F ‘name, person, one’ < PJK
*nom (Whitman 1985: 241). There are three problems with this
comparison, two on the Japonic side and one on the Korean. First, I believe
that the interpretation of MJ na 1.2 as a monosyllabic accent class

258
The image of Hwarang Kipha is represented by the reflection of the moon in the water.
259
MK nòlk- ‘old’ refers to objects, while MK nùlk- ‘old’ refers to human age.
260
Here, the reconstruction presented is PJK *-rg- rather than *-lg-. However, it appears
that Whitman has no other examples for PJK *-rg-.
 Lexical Comparisons———209

corresponding to disyllabic 2.5 (where the final falling pitch is believed to

reflect pre-PJN *-m) is mistaken, because there is no such falling pitch that
goes through the vowel in Middle Japanese: it is just H plus L on the
particle. In the modern Kyoto dialect the word naa ‘name’ is HL as well,
not HF, but the automatic lengthening of one-mora words to two-mora
words in Kyoto is a late phenomenon, and in any case the accentual
contour of Kyoto naa HL ‘name’ (1.2) is very different from Kyoto, e.g.,
saru LF ‘monkey’ (2.5), because the former belongs to the H register, but
the latter to the L register. Therefore, the reconstruction of pre-Proto-
Japanese final *-m for OJ na ‘name’ cannot be proven. Second, I am
unaware of any textual evidence in Old Japanese for the meanings ‘person,
one’ for OJ na. The word clearly has a primary meaning ‘name’ and a
derived meaning ‘fame, reputation’ (JDB 1967: 512); it is apparent that
‘person’ and ‘one’ are added to make the comparison look better. Third,
moving now to the Korean side, I believe that there is a significant
philological mistake. MK nóm is amply attested in the meaning of ‘other’,
but in the meaning ‘person’ it is a hapax legomenon appearing only in
(Welin I: 45a) (LCT 1987: 128; Nam 1997: 351). On the other hand, MK
nwóm ‘person, fellow (pejor.)’ is also amply attested in Middle Korean
texts (LCT 1987: 159; Nam 1997: 306). MdK nam ‘other’ and nom ‘fellow,
bastard’, as well as dialect data, also treat these as separate words (Choy
1987: 242, 247), and the singular usage in the Welin sekpo is in all
probability a scribal error. MK nwóm ‘person, fellow’ can be taken out of
the equation simply because MK /wo/ does not correspond to OJ /a/
according to Whitman’s own vocalic correspondences; OJ /u/ would be
expected (1985: 129). Therefore, we are left only with MK nóm ‘other’.
According to Whitman’s vocalic correspondences, MK /o/ can correspond
to OJ /a/ only if the latter is long (1985: 129), but this is impossible to
prove. Moreover, since OJ na ‘name’ belongs to the H register, it is highly
unlikely. As this comparison is further aggravated by the far-fetched
semantics (‘other’ vs. ‘name’), it should be abandoned.
(293) (R) MK nòmòlh ‘vegetable’ 261 ~ OJ ne L ‘root, shoot (of plant)’ <
PJK *nom (+ -il / -Vk). Whitman adds the following discussion: “If MK
nòmòlh is monomorphemic, this comparison requires both medial *m loss
after *o and final *r loss to derive the OJ form. However, the MK noun is a
possible compound of *nòm + -álh ‘grain, berry, small round object’”
(1985: 241). Let us deal separately with these two conflicting solutions, as
they both present various problems. First, I agree with Whitman that MK
nòmòlh should be treated as monomorphemic, because we have no internal
Korean evidence allowing us to analyze it as a compound. Second, as
already discussed above (see [212]), there is no internal justification for
*-m- loss here, as /u/ in Old Japanese would be needed in front of *-m-.
Third, we really do not know whether OJ ne ‘root’ is from PJ *nai, or from
PJ *nia. The second possibility rules out the etymology completely, but
even the first one has a problem with the vocalic correspondences. As

261
More precisely, ‘greens, herbs’.
210———Koreo-Japonica

mentioned in (292) above, MK /o/ can correspond to OJ /a/ only if the

latter is long (Whitman 1985: 129). Since OJ ne has a L register, this might
be the case, but taking into consideration the recent demonstration by
Shimabukuro that Proto-Japonic words with LOW register can be
reconstructed both with and without vowel length (Shimabukuro 2002:
203), we cannot be certain. Ryukyuan evidence, which is the only basis for
determining the existence of vowel length, is lacking for monosyllabic
words. Fourth, I am unaware of the meaning ‘shoot (of plant)’ for OJ ne: to
the best of my knowledge the textual evidence supports only ‘root’.
Therefore, the comparison of MK nòmòlh ‘greens, herbs’ with OJ ne ‘root’
is semantically weak. Finally, the explanation of MK nòmòlh as a
compound *nòm- + -álh ‘grain, berry, small round object’ is completely
teleological, because we have no internal evidence for *nòm- in Korean.
PK *nom- seems to have been reconstructed solely on the basis of OJ ne,
which constitutes circular reasoning and represents a reconstruction ‘from
above’. Due to these phonological, morphological, and semantic problems,
I reject this etymology. See also (136, 185-186, 212) above and (330)
below on the rejection of *-m- loss.
(294) (R) MK nwòh- ‘puts it, lets it go’ ~ OJ nuk- H ‘pulls it out, sticks
it in place’ < PJK *nuk- (Whitman 1985: 241). The phonology of this
comparison is fine, but it has semantic problems. First, MK nwòh- really
means ‘to put on the surface, to let go’, which contrasts with MK nyèh- ~
nèh- ‘to put inside, to insert’. MK nyèh- ~ nèh- would be a much better
match, but it has incompatible vocalism with OJ nuk-. Therefore, I reject
this etymology for the reason of semantics.
(295) (R) MK nwòh ‘rope’ ~ OJ nusa LL ‘Shintō offerings of rope or
chains of paper’ < PJK *nus (Whitman 1985: 241). There are two
problems with this etymology. The first is the unrealistic nature of the
correspondence MK -h- : OJ -s-, see (236-244). Second, WOJ and EOJ
nusa is not really a rope in any sense: it is pieces of cloth or paper cut in a
zigzag pattern and used either as offerings or as the top part of a
purification wand. I reject this etymology on phonetic and semantic
grounds.
(296) (R) MK nwòlós ‘role, post, official position’ ~ OJ nusi LH or LF
‘master, owner of, occupant of a position’ < PJK *nurus. Whitman adds:
“Ryukyuan (Shuri) nuusi attests to the long first syllable vowel from
medial *r loss’” (1985: 242). While Shuri nuusi really points to PJ *nuusi
(see Shimabukuro 2002: 203 and 293 above), it offers no internal evidence
for *r loss. 262 In addition, as far as I can tell from textual examples, MK
nwòlós, in contrast to MdK nolus, only means ‘play, game, amusement
party’ (LCT 1987: 155; Nam 1997: 299). MdK nolus, on the other hand,
may mean ‘role, party, work, duty, office’ (Martin et al. 1967: 341), but all
of these meanings in Modern Korean clearly represent a secondary

262
See the discussion about the lack of internal Japonic evidence for *-r- loss in similar
cases analyzed in (5), (39), (56), (104), (106), (119), (121), (155), (161), (236), etc. above.
 Lexical Comparisons———211

semantic development. MK nwòlós has nothing to do with posts or official

positions, 263 so I reject this etymology on phonetic and semantic grounds.
(297) (C) MK nwón ‘paddy field’ ~ OJ nô ‘field’ < PJK *nun (Whitman
1985: 242). 264 OJ nô indicates only ‘wild, uncultivated field’, but this is a
minor semantic difference. The word is attested in Eastern Old Japanese,
as well as in Ryukyuan, including the Southern Ryukyus, so I accept this
etymology.
(298) (L) MK nè (second person pronoun) ~ OJ na ‘id’. < PJK *ne
(Whitman 1985: 242). I demonstrated above that this is a Korean loan in
Japonic (see the pronoun section in the morphology chapter above). Cf.
also (280).
(299) (R) MK nèlù-, nèp-, MdK nelp- < ?*nèlùp- ‘broad, wide,
spacious’ ~ OJ nöNpï- ‘it stretches, unfolds, elongates‘, nöNpë- ‘stretch it,
lengthen it’ < PJK *nerpU- / *nerpe- (Whitman 1985: 242). 265 The Proto-
Korean form is clearly *nèlp-; see section 1.1.3.3.1 on lenition, especially
note 19 on p. 27. The major problem here is the irregularity of the
correspondences: there are no other examples in which PK *-lp-
corresponds to OJ -Np-. In addition, Whitman claims in (18) that PK *-lp-
corresponds to OJ -m-. Therefore, I reject this comparison due to the
irregularity in the correspondences.
(300) (R) MK :ne, :neyh ‘four’ ~ OJ yö- ‘id’. < PJK *nye (Whitman
1985: 242). In dealing with MK :neyh ‘four’ we should not overlook the
special form :nek that occurs with some classifiers. Note that the Middle
Korean forms for ‘three’ and ‘four’ are almost parallel:

Chart 32:
‘Three’ and ‘four’ in Middle Korean
Gloss MK free form MK bound forms
‘three’ :seyh :se-, :sek-, :sey-
‘four’ :neyh :ne-, :nek-

As I demonstrated in (222) above, MK :seyh is a metathesized form

derived from PK or OK *seki. By the same token, we have to reconstruct
MK :neyh as *neki. This creates a significant problem for Whitman’s
comparison, because the remaining *-ki has no analogue in OJ yö; nor do
we have internal Korean evidence that it was a special numeral ‘suffix’.
Another major problem is that in this etymology MK n- corresponds not to
OJ n- as above, but to OJ y-. Whitman’s solution is to reconstruct an initial
Proto-Japanese-Korean cluster *ny-, but the practice of creating extra
entities in reconstruction by juxtaposing two elements is unlikely to be
methodologically credible in comparative linguistics. Whitman provides
only one other example with MK n- corresponding to OJ y-: see (308)

263
I suspect that Whitman uncritically accepted the incorrect glossing of MK nwòlós as
MdK nolus in LCT (1987: 155). Cf. the correct glossing as MdK noli ‘play’, cangnan
‘game’ in Nam (1997: 299).
264
This comparison is also found in Martin (1966), #281.
265
This comparison is also found in Martin (1966), #261.
212———Koreo-Japonica

below. Numerals, especially lower numerals are typically either borrowed

or inherited as a set: we cannot expect Middle Korean and Old Japanese to
share the numerals for just ‘four’ and ‘eight’ (see [316] below) in their
respective numerical systems. Therefore, I reject this etymology on the
basis of phonetic and morphological problems, as well as on
methodological grounds.
(301) MK nyé- ‘comes and goes, stops in’ ~ OJ yör- ‘stops in, comes
toward’ < PJK *nye- (Whitman 1985: 242). Starting from this etymology,
Whitman introduces yet another correspondence: MK ny- to OJ y-. Three
other examples are given in support of it (302-304), but as I show below,
none of them is acceptable. The first issue here is regularity. Let us
compare this etymology with the one found in (300):

Chart 33:
‘Four’ and ‘stop in’ in Middle Korean and Old Japanese
gloss MK OJ PJK
‘four’ :neyh yö *nye
‘stop in’ nyé yör *nye

One can clearly see that the identical Proto-Japanese-Korean recon-

structtions *nye ‘four’ and *nye ‘stop in’ result in quite different
descendant forms in Middle Korean. There are other problems with this
comparison as well. First, -r in OJ yör- is left unexplained: it does not
correspond to anything in the Middle Korean form, nor is it present in the
stipulated Proto-Japanese-Korean reconstruction. Second, MK nyé- ‘comes
and goes, stops in’ has a rare H pitch on the stem (the majority of Middle
Korean monosyllabic verbs have a default L pitch) which indicates that in
Proto-Korean the verb must have been disyllabic, with MK nyé- resulting
from aphaeresis: PK *Vnye- LH > MK nyé-. Therefore, it becomes even
more difficult to compare MK nyé- to OJ yör-, and I must reject this
etymology due to all these problems of phonetic regularity. In any case,
MK nyé- really means ‘to go and come back’ or just ‘to go’ rather than ‘to
stop in’, which becomes important in my discussion of (302).
(302) (R) MK nyèh- ‘puts in’ ~ OJ yöse- ‘brings toward’ < PJK *nye-
sV-. Whitman remarks that “both MK and OJ verbs are transitive
derivatives in *-sV-” (1985: 242). Presumably he means that they are
transitive derivatives of verbs in (301). While this is certainly true of OJ
yöse-, the transitive counterpart of OJ yör-, the claim that MK nyèh- is a
transitive derivative of MK nyé- ‘to go [and come back]’ is much more
problematic, not just semantically, but also because of the two verbs’
different accent patterns. The monosyllabic transitive derivative of MK
nyé- would be expected to have the form *nyéh- with H, and not L pitch.
See also (301) above on the archetypal form of MK nyé-, which precludes
a comparison with OJ yör-, and consequently with its transitive derivative
yös-. Transitive derivatives in Middle Korean are formed with -i-, -hi-, -ki-,
-Gi-, -hwo-, -wo-, -Gwo-, -Gwu-, but never with -h- alone (Yi Swungnyeng
1961: 333-335). Finally, there is no internal evidence that final -h in MK
 Lexical Comparisons———213

nyèh- ‘to put inside’ does not belong to the root of the verb. Therefore, I
reject this etymology.
(303) (R) MK nyèkh ‘area around, vicinity’ ~ OJ yökö HH ‘side,
vicinity’ < PJK *nyeke (Whitman 1985: 242). There are several problems
with this etymology. The least serious one is with its semantics. I believe
that glossing OJ yökö ‘side, horizontal’ as ‘vicinity’ is not correct. Omoda-
ka et al. also list another meaning ‘wrong direction, abnormal, unjust,
contra-’ (JDB 1967: 793), but this most likely derives from a different
word, homophonous with yökö ‘side’. Comparison of ‘side’ with ‘vicinity’
is not impossible, but it is scarcely convincing. But the phonetic problems
are much more serious: MK nyèkh, which has L pitch rare for monosylla-
bic nominal stems, should probably go back either to PK *nyènkúkú or PK
*nyèkúkú, depending on whether final MK -kh originated from *-nkuku or
*-kuku. The first option rules out the comparison in a genetic sense
altogether, since MK -k- < PK *-nk- cannot correspond to OJ -k-. The
second option involving PK *nyèkúkú is more plausible for comparison
with OJ yökö ‘side’. Given the fact that the word is attested in Southern
Ryukyuan ― Yaeyama yuku (Miyara 1981: 347), Psara yuku (Shimoji
1979: 232), Higashi Nakasone, Yonaha, Tonoshiro yuku, Hateruma yugu
(Uchima and Arakaki 2000: 462) ― it could be treated as a cognate, but if
one agrees with my rejection of the only two other etymologies above
involving a correspondence of MK ny- to OJ y-, this comparison also
becomes questionable. Therefore, I reject this etymology.
(304) (R) MK :nyey ‘long ago, olden times’ ~ OJ inisi- ‘id’. Whitman
remarks that “both MK and OJ nouns are deverbal nominalizations of
*inye” (1985: 242), but this *inye- is nowhere to be found in his lists. It is
possibly a typographical error for PJK *ine-, discussed in (305). Even so,
there are many problems with this comparison. First, to the best of my
knowledge, there is a little inaccuracy: OJ inisi- ‘olden times’ is a ghost.
The word is really WOJ inisipê (JDB 1967: 88), which has a transparent
internal etymology: in- ‘go away’ + -i-, infinitive + -si, past attributive +
pye ‘side’. In addition, there is also WOJ inisi-kata ‘past’ (JDB 1967: 88).
Both words are certainly derived from OJ in- ‘to go away, to depart‘, but
they are not nominalizations; at best they might be compounds. Second,
and much more serious, is that this time Whitman claims that MK /ny/
corresponds to OJ /n/ and not OJ /y/, as in (301-303). Taking into account
the proposed correspondence of MK /n/ to OJ /y/ in (300) and (308), we
arrive at three different possibilities:
(1) MK n : OJ y
(2) MK ny : OJ y
(3) MK ny : OJ n

Such multiple correspondences strongly indicate an irregularity. Third, I

do not see any specific nominalization marker in MK :nyey ‘olden times’
that is identical in shape to MK :nyey- ‘to go’. If the former were a
nominalization of the latter, we would expect a form *:nyey-m, with the
214———Koreo-Japonica

nominalizer -m. Since MK :nyey means ‘olden times’, if it were indeed

derived from MK :nyey- ‘to go’, we would expect an even more complex
form, involving the past or retrospective markers, not just a bare verbal
stem. Fourth, both MK :nyey ‘olden times’ and MK :nyey- ‘to go’ have R
pitch, indicating that the words were originally disyllabic. Since a R pitch
is by origin a combination of L+H pitches, we know that it was the syllable
on the right, not a syllable on the left, that underwent reduction. This is
further supported by the presence of the triphthong /yey/ in these words.
The tentative Proto-Korean form would be something like *nyeCi-, with a
consonant /C/, whose phonetic value we cannot determine. Consequently,
neither of the Middle Korean words can be compared with OJ in- (written
by Whitman as inV- on this occasion but as in- in [310] below) ‘to go
away’, since the initial OJ i- and the extra syllable on the right in Middle
Korean are not explained either phonologically or morphologically.
Therefore, both this etymology and the following one, (305), should be
rejected; cf. also (310) below.
(305) (R) MK :nyey- ‘go’ ~ OJ inV- ‘id., depart’ < PJK *ine- (Whitman
1985: 242). Discussed and rejected in (304) above.
(306) (R) MK nùc- ‘late, behind time’ ~ OJ nöti LL ‘after’ < PJK
*nöc-. 266 Whitman offers the following explanation: “The comparison
must assume that OJ nöti is a nominalization in -i (ren’yōkei: continuative)
of a verb stem *nöt- now lost” (1985: 242). Since the verb *nöt- is not
present in any variety of Japonic, the comparison presented here is mere
speculation. This is further aggravated by imperfect semantics and a lack
of internal evidence for the suggested morphological segmentation of OJ
nöti. It is methodologically unacceptable to build a hypothesis on the basis
of another unproven hypothesis, so I reject this etymology.
(307) (R) MK núlí- 267 ‘makes it into, -like’ ~ OJ nör-, ni- ‘resembles it’
< PJK *nöri- (Whitman 1985: 242). There are two problems here. The
minor one is that WOJ nör- ‘to be like, to resemble’ is not attested phone-
tically in Old Japanese texts. Its first phonetic attestation nor- cannot be
definitively established until Middle Japanese (JDB 1967: 570). OJ ni- ‘to
be like, to resemble’ does not have a phonetic attestation in Western Old
Japanese either, again appearing in phonetic script only in Middle Japanese.
In addition, since both OJ *ni- and MJ ni- are vowel monograde verbs, we
simply do not know whether the pre-Old Japanese form was *ni- or *nï-.
This, in turn, makes relating MJ nor- and MJ ni- more dubious. The
comparative case markers WOJ -nö, -ni, -nasu and EOJ -nö, -nösu, -nasu,
which are usually presumed to be connected with this verb etymolo-gically,
may offer some evidence in favor of pre-OJ *nï- < PJ *nöy-, but the form -
nasu presents another problem in the area of vocalism. The internal history
of this word family in Japonic is not completely understood, although
some significant and important steps have already been taken toward elu-
cidating it (Hendriks 1992, 1994). The major problem is that MK núlì-
does not mean ‘makes it like’. There are no textual examples supporting
266
This comparison is also found in Martin (1966), #122.
267
The correct accentuation is núlì- (LCT 1987: 166; Nam 1997: 320).
 Lexical Comparisons———215

this meaning, and I have grave doubts that it even means ‘makes it into’. I
believe the problem can be solved if we ignore the modern Korean glosses
in the dictionaries and turn again to the texts. Yu Changton glossed the
word as MdK chita (LCT 1987: 166), but there are many homonymous
verbs chita in modern Korean (Martin et al. 1967: 1654-1655). As far as I
can tell, none of them mean ‘makes it into’. Nam Kwangwu glossed the
word as MdK nulita, nulettulita ‘to hang, to let hang down’ (Nam 1997:
320). Look at some examples from actual Middle Korean texts:
KUM nwoh-olwo kilh-ul nuli-kwo
gold rope-INSTR road-ACC build-GER
constructing a road by a golden rope (Sekpo IX: 10b-11a)

This first example may give the impression that Whitman’s definition is
right, but let us look at another one:

hhwang-kum-olwo nwo moyngkol-a kil-s kaz-ay nuli-kwo

yellow-metal-INSTR rope make-INF road-GEN side-LOC hang-GER
making a rope of gold and hanging [it] on the sides of the road (Pep III:
59a)

This example certainly agrees with the definition provided by Nam 1997:
320. It is further supported by its Chinese equivalent, 黄金爲縄以界道側
‘making a rope of gold in order to delineate the sides of the road’, also
found in Pep III: 59a. A very similar example is also found in a different
part of the same text:
hhwang-kum-olwo nwo moyngkol-a ku kyeth-uy nuli-kwo
yellow-metal-INSTR rope make-INF that side-LOC hang-GER
making a rope of gold and hanging [it] on that side (Pep II: 32) 268

The Chinese equivalent again says 黄金爲縄以界其側 ‘making a rope of

gold to delineate that side’ (LCT 1987: 166). What is probably meant in all
three cases is that the roadside or side of something else is marked by the
hanging rope. Let us look at one more example:

LAN-un namwo nuly-wu-n KWOW -LAN i-Gwo

balustrade-TOP wood construct-MOD-ATTR/REAL high-balustrade be-
GER
balustrade is a high balustrade constructed with wood (Welin X: 51b)

Once again, we encounter the possible meaning ‘to construct’, but I

believe that as in the above examples, the function of a balustrade is to
mark off or delineate the edge of something. Of course, we may be dealing
268
Cited on the basis of LCT 1987: 166; I was not able to locate the example myself on the
given leaf in the text, as the location given by Yu Changton probably includes a
typographical error.
216———Koreo-Japonica

with two different homonymous verbs here, one being ‘to construct’, and
another ‘to hang something in order to mark a boundary’, but the more
likely solution is that the archetype meaning is ‘to put or hang something
down in order to mark the boundary’. In light of this newly established
meaning, the comparison with OJ nör-, ni- ‘to resemble’ has to be rejected.
(308) (R) MK :nwuy ‘world, generation’ ~ OJ yö ‘id’. < PJK *nyö
(Whitman 1985: 242). This is the second example in which Whitman
claims that MK n- corresponds to OJ y-. The only other one was discussed
and rejected in (300) above. An additional discussion of how this ‘corres-
pondence’ is irregular is presented in (304) above. I am afraid that the
meaning of MK :nwuy is edited to make it look more like Old Japanese: it
certainly does not mean ‘generation’. The basic meaning is ‘world’, al-
though there is also a definition ‘time’. With the meaning ‘time’ the word
is predominantly attested in Early Modern Korean texts, but not in Middle
Korean texts (LCT 1987: 165; Nam 1997: 317). A much more serious
problem for this comparison is that MK :nwuy has R pitch which points to
its disyllabic origin: PK *nwuCi LH (cf. a similar discussion in [304]
above). We are extremely fortunate in this case, because the word is attes-
ted in mixed semantographic-phonographic writing in Old Korean as 世理
/NWUli/ (Hyangka XIII: 8). Thus, we can establish that the lost syllable on
the right was -li, which does not correspond to anything in OJ yö, and no
internal Korean evidence would allow us to explain it away as a suffix.
Finally, according to Whitman’s vocalic correspondences, MK /wu/ does
not correspond to OJ /ö/, as OJ /u/ is needed instead (Whitman 1985: 129).
Since this etymology presents numerous irregularities, it has to be rejected.
(309) (R) MK ní ‘tooth’ ~ OJ ki ‘fang’ < PJK *gi (Whitman 1985: 242).
This etymology is apparently misplaced in Whitman’s list. It has to appear
with the etymologies (177-180), all rejected, that involve the
reconstruction of PJK *g- on the basis of the ‘correspondence’ of MK n- to
OJ k-. For a detailed critique of this alleged correspondence refer to Martin
(1991: 273) and Vovin (1993b: 339-340). Although OJ ki ‘fang’ is not
attested phonetically in Western Old Japanese (the first phonetic attestation
is MJ ki and Middle Japanese compound kiba ‘fang’ 269 [JDB 1967: 237]),
Martin reconstructs PJ *kuCi on the basis of palatalization in Shuri
ciiba, 270 lack of aspiration in Kusigwa Kwiiba, and Yaeyama gïï-baa (1987:
450). Therefore, there is no regular correspondence between the Middle
Korean and Old Japanese vowels, so the etymology can be rejected.
(310) (R) MK ní- ‘goes’ ~ OJ in- ‘id’. < PJK *ni- (Whitman 1985: 242).
Whitman claimed earlier that OJ in- (presented this time as inV-) is related
to MK :nyey- ‘to go’, see (304) and (305) above. Obviously, OJ in- cannot
be related to both MK :nyey- ‘to go’ and MK ní- ‘to go’, unless one
manages to demonstrate that these two words in Middle Korean are
derived from the same root. However, even in this case, the number of
etymologies will be reduced by one. Earlier, I rejected the comparison
between MK :nyey- ‘to go’ and OJ in-, but for the sake of argument let me
269
MJ kiba ‘fang’ has a transparent internal etymology: *kï-N-pa ‘fang-DV(ATTR)-tooth’.
270
PJ *köCi > Shuri kii, without palatalization.
 Lexical Comparisons———217

now address the comparison of MK ní- and OJ in-, ignoring MK :nyey-. At

first glance a comparison looks possible, because MK ní- in Whitman’s
presentation appears to have H pitch that is rare for monosyllabic verbal
stems. It should indicate the loss of an initial vowel: PK *Vni- LH > MK
ní-. However, the problem is that while the dictionary form ní-tá ‘to go’
indeed has H pitch on the stem, in reality this verb belongs to verbal
Accent Class 4, which includes verbs with highly irregular accentuation
(Ramsey 1991: 232-233). Therefore, taking the H pitch of the dictionary
form ní-tá at face value is premature. We cannot prove that there was a PK
*Vni-, which would be needed for a valid comparison. There are also other
problems. OJ in- is an irregular verb that exhibits paradigmatic forms
typical of both consonantal and vowel classes. It is not clear whether it was
originally a consonantal or a vowel verb. We can surmise PJ *in- if the
verb was consonantal or PJ *inu[-]i- or *inö[-]i- if the verb originally
belonged to a vowel paradigm. However, in all of these cases a comparison
with MK ni- ‘to go’ (even if it is proved one day that it is from PK *Vni-)
still faces some significant problems. If OJ in- is originally a consonantal
verb, how do we explain final -i in MK ni-, which corresponds to nothing
in the Old Japanese form? If OJ in- is originally a vowel verb, then neither
PJ *u nor PJ *ö in the second syllable corresponds to MK /i/. We are again
faced in this etymology either with irregularities or with unexplained
problems, and this leads me to reject it.
(311) (R) MK -(n)ún, -(n)ón (topic marker) ~ OJ -nö (genitive suffix,
nominative suffix in embedded clauses) < PJK *n (Whitman 1985: 242-
243). Middle Korean has the following forms for the topic marker: -nón,
-nún, -ón, -ún, and -n (Yi Swungnyeng 1961: 149). Of those, the first two
are clearly secondary, as only one sign, 隱 /on/, is used in Old Korean after
both vowel and consonantal stems. This creates an obvious phonetic
problem: the Middle Korean topic and Old Japanese genitive can be related
only through metathesis. In addition, the ‘nominative’ function of Old
Japanese genitive -nö is really just an extension of the genitive, with
similar cases widely attested in various SOV languages. To the best of my
knowledge, I know of no languages that developed their genitive markers
from topic markers, or vice versa. Therefore, I reject this etymology.
(312) (R) MK nìlùl- ‘tells it, informs’ ~ OJ nör- ‘id., declares it’ < PJK
*nirör- > pre-OJ nyör- > OJ nör- (Whitman 1985: 243). There are several
problems with this etymology. First, it rests on an assumption of *-r- loss
in Old Japanese for which there is no internal Japonic evidence. Second,
the pre-Old Japanese form nyör- is completely teleological; there is no
internal evidence supporting it either. Third, I believe that Whitman
confuses MK nìlù- ~ nìlò- ~ nìl- ~ nìlG- ‘to tell, to speak’ (LCT 1987: 170-
171, 174-175; Nam 1997: 326, 328, 335) with MK nìlú- ~ nìlúl- ‘to reach’
(LCT 1987: 171-172; Nam 1997: 326-327): I see no textual evidence for
the extended stem *nìlùl- in Middle Korean for the first of these verbs.
Fourth, MK /i/ certainly does not correspond to OJ /ö/. Fifth, the variants
nìlò- ~ nìl- ~ nìlG- appear in Middle Korean much more frequently than
the variant nìlù-. The change of /o/ to /u/ in the second syllable is an
218———Koreo-Japonica

expected and very well documented change in the history of Korean, while
the opposite change did not take place, so MK nìlò- must be a more archaic
variant. This creates a problem for Whitman’s vocalic correspondences,
because MK /o/ does not correspond to PJ *ö if we reconstruct Proto-
Japonic as *nörö-. On the other hand, if we reconstruct PJ *nör-, we face a
morphological problem: as the variants nìl- ~ nìlG- indicate, the Proto-
Korean form was probably *nìlòk-, making PK *-ok an unaccounted-for
segment. Therefore, due to these various problems, I reject this comparison.
Among the thirty-two etymologies representing Proto-Japanese-Korean
initial *n- (279-312),271 I was able to find one possible cognate and one
loanword; the remaining etymologies were rejected. The almost complete
lack of good Koreo-Japonic etymologies reflecting PJK *n- speaks
strongly against a genetic relationship.

3.2.13 *y-
(313) (L) MK yélh ‘ten‘, yèléh ‘a large number, many’ ~ OJ yöröNtu
‘id., ten thousand’ 272 < PJK *yere(l). Whitman adds: “Final -Ntu in the OJ
form is probably the genitive suffix -tu generally attached to numerals used
as substantives” (Whitman 1985: 243). There is one major problem with
this comparison. MK yélh ‘ten’ and yèléh ‘a large number, many’ are not
etymologically related. In the Ceycwuto dialect, which still keeps reflexes
of PK *ye and *yo distinct, we find SC yora 273 and NC yorai ‘a large
number, many’, but both SC and NC yel ‘ten’ (Kim et al. 1995: 95, 173).
Thus, we have to reconstruct PK *yora or *yara ‘a large number, many’.
PK *yora/*yara cannot be a cognate of WOJ yöröNtu, because PK *o and
*a do not correspond regularly to OJ /ö/, according to Whitman’s vocalic
correspondences (Whitman 1985: 129). This leaves only MK yélh as a
candidate for a possible genetic cognate, but the comparison of ‘ten’ with
‘great number’ is semantically weak. The final -Ntu in WOJ yöröNtu
cannot be equated with the genitive-locative suffix -tu for three reasons: (1)
-tu attached to numerals used as substantives is not really a genitive, but a
classifier; (2) it is used only with the lower numerals 1-9; and (3) -Ntu
includes a voiced prenasalized consonant /Nt/, while both genitive -tu and
classifier -tu exhibit voiceless /t/. I believe that this -Ntu is the collective
suffix -Ntu found also in OJ mî-Ntu ‘water’. Since there are no cognates of
WOJ yöröNtu in Ryukyuan and Eastern Old Japanese, distribution strongly
suggests that it is a loan from some Old Korean dialect where PK *yo/*ya
> ye, as in Middle Korean. The collective suffix -Ntu must have been
added after the borrowing took place.
(314) (L) MK :yel- ‘ties it together, weaves it’ ~ OJ yör- ‘braids it,
twists it, twines it together’ < PJK *yel- (Whitman 1985: 243). The
problem with this seemingly impeccable Old Japanese candidate for a
cognate is that it has limited distribution: it is not attested in Eastern Old

271
As indicated earlier, etymology (309) is misplaced, and is not counted here.
272
The same etymology is found in Martin (1987: 577).
273
There is also a Southern Ceycwuto doublet form yele (Kim Cwunghoy et al. 1995: 173),
but since it is isolated in Southern Ceycwuto, it is probably a loan from Modern Korean.
 Lexical Comparisons———219

Japanese and Ryukyuan. Therefore, I treat WOJ yör- as a loan from

Korean.
(315) (R) MK yèlp- ‘thin, weak’ ~ OJ yôwa- ‘weak’ < PJK *yorob-.
Whitman adds: “Cf. OJ yuru- ‘loosen, soften’” (Whitman 1985: 243). The
major problem here is the vocalism: MK yèlp- < PK *yelp-, as seen in SC
and NC yelp- (Kim Chwunghoy et al. 1995: 137). Even if one accepted
Whitman’s treatment of OJ yôwa- ‘weak’ and OJ yuru- ‘loose’ as related
words, 274 admitting OJ yuru- as internal evidence for *-r- loss, one can still
reconstruct only PJ *yurawa- > OJ yôwa- ‘weak’. PJ *u does not
correspond to MK /e/ according to Whitman’s vocalic correspondences
(Whitman 1985: 129). Therefore I reject this etymology on the basis of its
irregularity. In addition, I have to mention that glossing MK yèlp- as
‘weak’ is somewhat misleading, because it is an English translational
equivalent for this word used to describe the quality of drinks, e.g., ‘weak
tea’.
(316) (R) MK yèt- < MK yòt- ‘eight’ in MK yètólp ‘eight’, yètón
‘eighty’ ~ OJ ya H < PJK *yo(t). Whitman comments: “ MK yotolp
actually occurs in the Hwungmin cengum haylyey (1446), where it is
described as a dialect pronunciation. Yi Kimun (1964: 121-123) notes that
the modern Ceycwu pronunciation [yodap] supports early MK *yotolp and
gives additional evidence for a general shift of earlier *yo to /ye/” (1985:
243). There are several problems with this etymology. First, Whitman’s
data have to be slightly corrected. The dialectal MK yotolp which
Whitman cites is mentioned not in the Hwungmin cengum haylyey, but in
the Hwungmin cengum wunhay (Wunhay, p. 209), 275 a work by the
eighteenth-century Korean philologist Shin Kyengcwun (1712-1781). In
addition, the MK yètólp ‘eight’ and yètón ‘eighty’ that Whitman cites are
actually later forms that are typically found in Early Modern Korean texts.
The earlier Middle Korean forms are yètúlp ‘eight’ and yètún ‘eighty’,
respectively (LCT 1987: 561; Nam 1997: 1078). Second, in both MK
yètúlp ‘eight’ and yètún ‘eighty’ we have a non-leniting MK -t-, which
goes back to a cluster. This cluster, as I discussed earlier in the section on
lenition, is most likely *-nt-, but it can also be *-lt-. Third, Whitman’s
proposed morphological segmentation, which segments out MK yèt- <
*yòt- out of both MK yètúlp ‘eight’ and yètún ‘eighty’, is difficult to
justify. If we segment MK yètúlp as *yèt-úlp, then what is the remaining
part *-úlp? To the best of my knowledge, there is no such suffix in Korean.
The morphology of Middle Korean numerals represents an extremely

274
In spite of the fact that both adjectives belong to Accent Class B, there are two
formidable problems that make their common etymological origin highly unlikely. First,
there is a phonetic problem: we would expect OJ *yura-, not yuru-, in order to relate it to
OJ yôwa- < PJ *yurawa-: the vowels in the second syllables do not match. Second, we
would need to justify the morphological segmentation of *yurawa as *yura-wa-. This seems
to be an impossible task since, to the best of my knowledge, there is no internal Japonic
justification for the much needed suffix *-wa-.
275
The only manuscript of Shin Kyengcwun’s work to survive the twentieth-century
hostilities on the Korean Peninsula. I quote it on the basis of a 1938 edition by Cosen e
hakhoy, reprinted in the Wenpon kwuke kwukmunhak chonglim series (1985, vol. 13).
220———Koreo-Japonica

difficult problem that merits its own investigation. I try to offer a solution
here. Let us compare Middle Korean digit numerals with its tens:

Chart 34:
Middle Korean numerals
MK digits MK tens
1. hònáh < *hotan 10. yélh < *yel[u]h
2. :twulh < *twupulh 276 20. súmúl
3. :seyh < *seki 277 30. syèlhún
4. :neyh < *neki 278 40. màzón (cf. MdK mahun) < *mason
5. tàsós 50. :swuyn
6. yèsús < *yonsus 279 60. yè:sywuyn < *yonsywuyn
7. nìlkwúp 70. nìlhún
8. yètúlp < *yon/ltulp 80. yètún < *yon/lton
9. àhwóp 90. àhón

The derivation of the tens from digits may seem quite idiosyncratic, but
I believe we can see a certain pattern. First, with the exceptions of 1~10
and 2~20, which clearly represent unrelated roots, and 5~50, which is also
likely to involve unrelated roots, the teens are derived from the shortened
forms of digits. Second, we can see that the numerals 30, 40, 70, 80, and
90 end in -zon > *son, -hun, -on, or -un. By a logical typological approach
to numerical systems, we are forced to conclude that all these variants go
back to the same archetype with the meaning ‘ten’, although apparently
have no connection with MK yélh ‘ten’. I believe that the phonetic shape
of this archetype can be reconstructed conditionally as *-son 280 ‘ten’ and I
will try to prove this hypothesis below.
First, WOJ -sô ‘ten’, found only within the tens, is a likely loan from
some variety of Old Korean (Vovin 2005a: 367). Therefore, it confirms the
initial consonant as /s/. Second, we should notice that the numerals :swuyn
‘50’ and yè:sywuyn ’60’, which were not listed above, also end in -n like
other teens, and that they both have syllables with R pitch that indicates a
disyllabic origin, as MK :twulh ‘two’ < *twupulh has. Thus, it is quite
possible that :swuyn ‘50’ is from something like *swuy-son and yè:sywuyn
‘60’ is from *yonsywuy-son. Third, since the cluster -ls- is extremely rare
in Middle Korean, and the clusters -ts-, -nts-, -lts-, and -hs-, do not exist,
we would expect certain morphophonological changes to occur on the
morphemic boundaries when the shortened forms of digit numeral roots
ending in -l, -t, and -h would be followed by our tentative *-son. Keeping

276
This reconstruction is possible on the basis of EMK twupul, written as 途孛 (MC du-
bwot) in Kyeylim #20.
277
See (222) above for the justification of this reconstruction.
278
See (300) above for the justification of this reconstruction.
279
PK *yonsus with *o is supported by Ceycwuto forms such as SC and NC yoset, although
the isolated SC yeset also exists (Kim Chwunghoy et al. 1995: 94).
280
This is just an approximation. The exact reconstruction of this morpheme cannot be
proposed at the present time due to multiple problems.
 Lexical Comparisons———221

this in mind, I propose the following developments for tens:

40. *nè-són > *nà-són > *mà-són > mà-zón

50. *swùy-són > *swùy-zón > *swùy-ón > :swuyn
60. *yòsywùy-són > *yèsywùy-zón > *yèsywùy-ón > yè:sywuyn
70. *nìl-són > *nìl-hón > nìl-hún
80. *yòlt-són > *yèlt-ón > *yèt-ón > yètún
90. *àh-són > àhón
The numeral syèlhún ‘30’ is absent from the above list, because it is
difficult to explain how it obtained its /l/. It may be either a result of
analogy, or more likely, a very complex case involving the two-layered
compounding of the short form se ‘three’ + yélh ‘ten’ + *-són ‘ten’, thus:

30. *sèkí-yélh-són > *:se-yélh-són > *s-yèlh-són > syèlhún

Now we are ready to go back to the problem of MK yètúlp ‘eight’ and

yètún ‘80’. Both have a non-leniting MK -t- which can go back to the
Proto-Korean clusters *-lt- or *-nt-. I suspect that MK yètúlp ‘eight’ is
really derived by counting in reverse: namely it is from yélh ‘ten’ + EMK
twupul, the latter being metathesized with a consecutive loss of the vowel
in the second syllable and vowel reduction in the first: twupul > *twulup >
-tulp. There is, of course, a strong argument against this etymology. As we
have seen above, MK yètúlp < *yoltulp, while MK yélh < *yélh. This
discrepancy in vocalism is difficult to explain. Despite its irregularity, the
advantage of this internal etymology over Whitman’s external etymology
is that it does not involve an ad hoc morphological segmentation of MK
yètúlp ‘eight’ as *yèt-úlp. In addition, as I mentioned in (300) above, lower
numerals are normally inherited or borrowed as a set, and we cannot
expect that Japonic and Korean share just ‘four’ and ‘eight’. Thus,
regardless of whether my internal etymology for MK yètúlp ‘eight’ can be
proven or not, I reject Whitman’s comparison of it with OJ ya- ‘id’. on
morphological and typological grounds.
(317) (R) MK yèwúy- ‘gets thin, becomes emaciated, grows haggard’ ~
OJ yase- ‘id’. < PJK *yese-. Whitman comments: “MK [there is a lacuna
in the text — A.V.] should result from a loss of a medial consonant,
according to this comparison /z/, but *[yèzwúy-] is not attested” (195: 243).
It is no wonder that speculative MK *yèzwúy- is not attested, as the
evidence from dialects indicates that the mysterious consonant was *-p-.
Choy Hakkun cites such forms as Kyengsang Pukto yebinda, Hamkyeng
Namto yəbinda, Cenla Namdo yabunda, etc. (1987: 1420-1421). Thus, MK
yèwúy- ‘gets thin, becomes emaciated, grows haggard’ (which could be
pronounced [yèGwúy-]) goes back to *yepwuy-. I reject the comparison
with OJ yase- ‘id.’ due to the impossible correspondence PK *-p- : OJ -s-.
Among the five etymologies presented by Whitman for PJK *y- (313-
317), I have rejected three and accepted two as loans from Korean into
222———Koreo-Japonica

Japanese. This leaves no good Koreo-Japonic etymologies reflecting PJK

*y-, creating another gap that strongly speaks against a genetic relationship.

3.2.14 *a-
(318) (R) MK a- ‘my’ ~ OJ a (first person pronoun) < PJK *a
(Whitman 1985: 244). Whitman claims that MK a- ‘my’ “appears as the
first member in the following Middle Korean kin terms, all referring to the
speaker’s kin: à-pá:nim ‘father’ (polite), à-pí ‘father, male’ (cf. OJ pî
‘male’ 281), à-zó ‘younger brother’ (cf. OJ se male relative’ 282), à-sóm ‘re-
latives’, à-cómi ‘aunt’, à-càpí ‘uncle’, LMK 283 a-ci ‘wet nurse’. This is
certainly an ingenious and interesting proposal, but there are multiple ob-
stacles against identification of Korean a- in kinship terms with the first
person pronoun. First, Whitman’s claim that these kinship terms all refer to
the kin of the speaker is simply not true. Cf. the following examples, in
which the kinship terms acomi and acoma-nim (polite) ‘aunt’ do not refer
to speaker’s kin:

acomi-lul ceh-osy-a
aunt-ACC fear-HON-INF
[Xuan-zong] feared [his] aunt, and… (YP 99)
acoma-nim-on Ttay.qoy.ttwoM-lul nilu-si-n-i
aunt-HON-TOP Mahâprajâpati-ACC call-HON-ATTR/REAL-NML
[Buddha’s] aunt was called Mahāprajāpati (Sekpo VI: 1a)

Second, there is no independent internal Korean evidence for a- ‘my’,

nor is there any internal evidence for its segmentation as a separate
morpheme in the above terms. It seems that the existence of OJ a ‘I’ is the
only justification for such a segmentation, and this is a reconstruction from
‘above’.284 Third, MK àtól ‘son’ and MK ákí ‘child, baby’, which also start
with /a-/, are not on the above list, because they are needed for other
etymologies, where they are compared with OJ atô ‘footstep’ and WOJ
aNkî (intimate informal second person pronoun); see (322) and (325)

281
To the best of my knowledge, there is no textual evidence for OJ pî ‘male’, and
Whitman does not indicate his source. There is OJ pî ‘spiritual power’, found sometimes in
deities’ names (JDB 1967: 603), but it can hardly be equated with ‘male’.
282
OJ se originally meant ‘elder brother’, but it could be used by females to refer to their
husbands or lovers, cf. OJ imô ‘younger sister’, that was used by males to refer to their
spouses or lovers. Modern Korean also offers an interesting parallel in this respect: oppa
‘elder brother (of a woman)’ is used as a term of address to a boyfriend by his girlfriend.
283
This is actually Early Modern Korean.
284
Let me provide the following hypothetical example illustrating why such a methodology
is unacceptable. Let us claim that Russian and Ket (Yeniseian language family, Central
Siberia) are related. Yeniseian has a third person singular possessive prefix d- ‘his’. What
prevents us from segmenting this d- as a possessive prefix in the following Russian kinship
terms all starting with d-: ded ‘grandfather’, diadia ‘uncle’, dever’ ‘brother-in-law
(husband’s brother)’, doch’ ‘daughter’, and ditia ‘child’, representing them as d-ed, d-iadia,
d-ever’, d-och’, and d-itia, and claiming that this d- is the same prefix as found in Ket?
Certainly, there is no internal Russian evidence for such a morphological segmentation.
 Lexical Comparisons———223

below. It remains unclear why this a- ‘my’ cannot be segmented in àtól

‘son’ or ákí ‘baby’ with the only justification being that they have different
Japonic etymologies. Fourth, there are other Middle Korean kinship terms
that do not start with a-, e.g., stól ‘daughter’, émí ‘mother’, nwù[G]ùy <
*nwupuy ‘sister’, etc. Why do these kinship terms lack a- ‘my’? Finally,
Korean never had any nominal prefixes, but Whitman’s suggested analysis
necessitates treating a- as a possessive prefix. This would be a unique
nominal prefix and against the general structure of Korean. Therefore, on
the basis of numerous problems, I reject this etymology.
(319) (R) MK ách ‘reason’ ~ OJ (mainly Azuma) aNto ‘how, why’ <
PJK *ecö (Whitman 1985: 244). The form aNto is not ‘mainly Azuma’,
but predominantly Azuma, as it is attested in Eastern Old Japanese texts
seven times, but only once in the entire Western Old Japanese corpus (MYS
XV: 3639). This makes it a hapax legomenon in Western Old Japanese,
and taking into consideration the imperfect history of the textual
transmission of MYS XV, it can be probably safely disregarded as a
Western Old Japanese form. Whitman follows the Omodaka et al. glossing
of aN-tö as ‘how, why’ instead of ‘what’ (JDB 1967: 32). Yet the meaning
‘how’ does not occur in the texts at all, and in only one example out of
seven does aN- mean ‘why’, while it certainly means ‘what’ in the
remaining six. Some Eastern Old Japanese examples are presented below;
for the others, see Vovin (2005a: 309-310):

wa-Nka se-kô-wo aN-tö kamö ip-am-u

I-POSS beloved-DIM-ACC what-DV PT say-TENT-ATTR
What shall [I] say about my beloved? (MYS XIV: 3379)
ak-an-u-wo aN-tö ka a-Nka se-m-u
satisfy-NEG-ATTR-ACC what-DV PT I-POSS do-TENT-ATTR
since it was not enough [for me], what should I do? (MYS XIV: 3404)

NA pa aN-tö ka [o]môp-u
you TOP what-DV PT think-ATTR
What do you think? (MYS XIV: 3494)

aN-tö ka taye se-m-u

why-DV PT break(NML) do-TENT-ATTR
why should [we] break up? (MYS XIV: 3397)
Given the fact that nani ‘what’ appears in the entire Eastern Old
Japanese corpus only twice, and both times in poems that have no typical
Eastern Old Japanese features (MYS XIV: 3373 and XX: 4323), it probably
can be disregarded as an Eastern Old Japanese form (Vovin 2005a: 309). I
treat EOJ aN-tö as a combination of aN-, a cognate of WOJ nani ‘what’
and PR *nau ‘id.’ and the defective verb tö ‘to say’. I present a detailed
discussion along these lines in Vovin 2005a: 309-314, also demonstrating
the reconstruction of the common Proto-Japonic form as *n-anu. Thus,
224———Koreo-Japonica

EOJ aN- ‘what’ cannot be compared to MK ách ‘reason’, and I reject the
etymology.
(320) (R) MK àchí- in àchíet-, àchyét- ‘hate’ < *àchí- + :et- ‘get’ ~ OJ
asi- ‘bad, evil, hateful’ < PJK *aci (Whitman 1985: 244). First, the data
require a minor correction. As far as I can tell, MK àchíet- is a ghost, the
actual form is MK àchíètpù-, a hapax legomenon attested only in Sekpo
XIX: 7a. 285 Therefore, we essentially have to deal with MK àchyét- ‘to
hate, to dislike’, but the basis for Whitman’s internal analysis of this form
as consisting of *àchí- (which does not present itself otherwise) plus :et-
‘to get, to receive’ remains unclear to me. The MK àchyét- form ‘to hate,
to dislike’ is clearly a transitive verb, and its derivation from ‘to get, to
receive’ seems somewhat dubious, since we would expect it rather in the
case of an intransitive verb. Besides, given the fact that MK :et- ‘to get’
has R pitch, the proposed contraction L+H+R > L+H seems phonetically
unnatural as well. Finally, OJ asi- ‘bad’, which belongs to the traditional
-siku class of adjectives, certainly consists of the root a- + petrified suffix
-si- found in other adjectives belonging to the same class. 286 Needless to
say, the basic meaning of a-si- is just ‘bad’, as it is the antonym of yö-
‘good’ (JDB 1967: 20), and neither ‘evil’ nor ‘hateful’ is its primary
meaning. The proposed etymology boils down to a one-phoneme
etymology with dubious semantics: the monophonemic Old Japanese root
a- is compared to MK *àchí-, which is itself segmented ad hoc, and
certainly cannot be segmented any further. Therefore, I reject this
etymology.
(321) (L) MK àchóm ‘morning’ ~ OJ asa LF ‘id’. 287 < PJK *asom.
Whitman notes that he cannot account for the aspiration in the Middle
Korean form (Whitman 1985: 244). The correspondence of MK -ch- to OJ
-s- is indeed irregular. If we look at the distribution of OJ asa ‘morning’ in
Japonic, we can clearly see that while the word is present in both Western
and Eastern Old Japanese, as well as in Middle Japanese, it is not attested
in Ryukyuan, with a single exception: Shidōke ʔasa ‘morning’ from the
island of Kikaijima in the Northern Ryukyus (Hirayama 1966: 284). The
isolated nature of Shidōke ʔasa ‘morning’ and a lack of any attestations in
the Southern Ryukyus strongly suggests that the Shidōke form is a loan
from mainland Japanese. The lack of Ryukyuan attestations indicates that
WOJ and EOJ asa is a loan from Korean. 288
(322) (L) MK ákí, àhóy ‘baby, child’ ~ OJ aNkî (intimate informal
second person pronoun) < PJK *agi (Whitman 1985: 244). First, it is
necessary to note that MK ákí ‘baby, child’ < PK *ánkí and MK àhóy
‘child’ < PK *àkóy cannot come from the same etymon because of
differences in their consonantism, vocalism, and pitch patterns that could
not be easily explained if these words were truly related. Phonetically only

285
The form àchyétpù- is amply attested (LCT 1987: 522; Nam 1997: 1015-1016).
286
Cf. the treatment of this adjectival -si- in (265) above.
287
This comparison is also found in Martin (1966), #144.
288
The native Japonic word for morning is reflected in WOJ tutômëte (MJ tutomete)
‘morning’ and PR *sutomete, which is amply attested in the Southern Ryukyus.
 Lexical Comparisons———225

MK ákí ‘baby, child’ < PK *ánkí can correspond regularly to WOJ aNkî,
because WOJ -Nk- and -î do not correspond regularly to MK -h- and -oy,
respectively. It is also important to know that only MK ákí is used as a
term of endearment, ‘baby’ (the same is true of the Modern Korean as
well). For WOJ aNkî Whitman adopts the gloss provided by Omodaka et al.
(JDB 1967: 7), but ignores the etymological commentary that follows,
which claims that WOJ aNkî is essentially a contraction of a-Nka kîmî ‘my
lord’, an explanation which dates back to Yamada Yoshio (1954: 32-33).
Such a contraction would be most unusual from the viewpoint of Western
Old Japanese phonology (Vovin 2005a: 236). There are only two examples
where WOJ aNkî is clearly attested in phonetic writing. One is found in the
following poem from the Kojiki, where the rebellious prince Osikuma-nö
mîkô addresses his commanding general Isapî-nö sukune after they have
been defeated by forces of the empress Jingū:
iNsa aNkî purukuma-Nka ita te op-aNs-u pa nipô-N-töri-nö apumî-nö umî-
ni kaNduk-î se-na wa
hey child Purukuma-POSS painful hand cover-NEG-INF TOP nipo-
DV(ATTR) bird-COMP Apumi-GEN sea-LOC dive-NML do-DES I
Come, baby, if [we are] not to receive heavy wounds from Purukuma, I
wish [that we] dive into the sea of Apumi like nipo birds (KK 38)

After the prince sang this song, both he and his general committed suicide
by drowning. Tsuchihashi notes the parallelism of iNsa aNkî ‘hey, child’ in
this poem with iNsa kô-Ntömö ‘hey, children’ in KK 43, but he believes
that WOJ aNkî is more honorific (Tsuchihashi 1957: 60). However, since
the prince is addressing his commanding general, who is clearly of lower
status than the prince himself (as is also indicated by their respective titles
mîkô and sukune), I cannot see anything ‘honorific’ in this usage.
The second phonetic appearance of WOJ aNkî occurs not in a poem, but
in the Chinese prose of the second volume of the Kojiki. Here the emperor
Ōjin addresses his son Opö-saNsakî-nö mîkötö (the future emperor
Nintoku):

爾天皇詔佐邪岐阿藝自佐至藝五字以音如我所思
er tianhuang zhao SaNzakî aNkî zhi yan (zi 佐 zhi 藝 wu zi yi yin) ru wo suo si
then emperor declare SaNsakî child PT speech (from 佐 to 藝 five characters by
289
sound) like I NML think
Then the emperor deigned to say: ‘SaNsakî, [my] child ([read] five characters
from 佐 to 藝 by their sound), [your] words are like what I thought’ (KJK II:
69b)

Of course, it is impossible for an emperor to address his own son as ‘my

lord’, because a prince is lower on the social ladder than the emperor.
Therefore, it is likely that in the first example, from KK 38, WOJ aNkî is

289
This is an inserted commentary.
226———Koreo-Japonica

used as a term of endearment meaning ‘baby’, while in the second example

it is used in its primary meaning ‘child’, although it seems likely that an
element of endearment is also present. These examples clearly show that
there is no honorific usage involved. The rarity of WOJ aNkî in the texts
also speaks strongly against a pronominal usage, even as a ‘fond informal
second person pronoun’, as Whitman suggests. Finally, we have to look at
the distribution of WOJ aNkî ‘baby, child’. The biggest problem for
Whitman’s etymology is that the word is not attested in any variety of
Japonic except Western Old Japanese. Moreover, WOJ aNkî in both of the
above examples occurs within a small chronological window: during the
regency of the empress Jingū and the reign of her son emperor Ōjin. After
this they never resurface. Both Jingū and Ōjin are known for their Korean
connections, and it is highly likely that the dynasty Ōjin founded was of
Korean origin (Ledyard 1975: 237-254). In short, everything points to the
fact that WOJ aNkî ‘baby, child’ is another loan from Korean.
(323) (R) MK :al- ‘below, before’ ~ OJ asi LL ‘foot, below’ < PJK
*al2i (Whitman 1985: 244). There are problems with this etymology on
both the Korean and the Japonic sides. First, Whitman conflates two very
different Middle Korean words: :aláy ‘before’ and àláy ‘below’ (LCT
1987: 516; Nam 1997: 1002-1003). Since besides MK :aláy ‘before’, there
is also its doublet form :alóy ‘before’ (LCT 1987: 515; Nam 1997: 1005)
and a related MK àlph ‘before, front’ (LCT 1987: 527; Nam 1997: 1026),
the segmentation of MK :aláy ‘before’ as :al-áy is probably justified
despite the word’s obscure phonological history. But, it is hardly possible
to compare ‘before’ with ‘foot’, so we are left with only MK àláy ‘below’,
although the semantics of this comparison are far from perfect. There is no
similar evidence for the segmentation of MK àláy ‘below’ as *àl-áy, 290
although one might argue that -áy is a petrified locative case marker. Such
an argument could seemingly be supported by the fact that MK àláy
‘below’ is not followed in Middle Korean texts by case markers, including
locative case markers. However, a strong counterargument can be made.
First, there is a single attestation of MK àlá ‘below’ (Mwongpep 21a). It is,
of course, a hapax legomenon, which I would normally not use as evidence,
but in this particular case it is strongly supported by the Middle Korean
compound àlà-wùh ‘below and above‘, which is amply attested (LCT 1987:
516; Nam 1997: 1002). Therefore, I think that MK àláy ‘below’ represents
a contraction of MK àlá ‘below’ + -áy, locative case marker. A
comparison of MK àlá ‘below’ with OJ asi ‘foot, leg’ runs into an
irregular correspondence of the vowels in the second syllable, since MK /a/
does not correspond to OJ /i/, whether the latter is from *i or from *ui or
*öi. However, as I mentioned earlier, there are also problems on the
Japonic side. First, OJ asi only means ‘foot, leg’, and I am not aware of
any textual evidence for the meaning of ‘below’ that Whitman mentions.
Second, the root of this word is just a ‘foot, leg’, attested as such in
290
The form aloy is attested only in Early Modern Korean. Since in Early Modern Korean
there was no longer a phonemic distinction between /a/ and /o/, EMdK aloy cannot be
claimed a doublet.
 Lexical Comparisons———227

Eastern Old Japanese (MYS XIV: 3387), and in compounds in both

Western and Eastern Old Japanese: WOJ and EOJ a-N-kakî ‘gallop’ (lit.
foot’s scratching the ground’s surface) (JDB 1967: 2), WOJ a-na-suwe
‘feet end’ (JDB 1967: 34), WOJ a-N-pumî ‘stirrup’ (lit. foot’s stepping
place) (JDB 1967: 39), WOJ a-yupî ‘strings to tie hakama under knees’ (lit.
leg-tier) (JDB 1967: 52), etc. A comparison of OJ a ‘foot’ and MK àlá
‘below’ is an etymology based on one phoneme. It leaves the second
syllable in the Korean word unaccounted for and involves very vague
semantics. Therefore, I reject this etymology. 291
(324) (R) MK ani (negative) ~ OJ -an-, id. < PJK *-an (Whitman 1985:
244). This etymology was discussed in the section on verbal morphology
(2.3.2.1) and rejected.
(325) (R) MK àtól ‘son’ ~ OJ atô ‘foot, trace, footprints, after,
successor’ < PJK *ator (Whitman 1985: 244). This etymology has
problems on both the Korean and the Japonic sides. First, MK àtól ‘son’
has non-leniting MK -t- < PK *-nt-, which cannot correspond regularly to
OJ -t- (see the chapter on lenition in Middle Korean). Second, OJ atô does
not mean ‘successor’; this meaning is simply added to make the words
look closer. Omodaka et al. assign two meanings to the word: (1) ‘foot, leg,
footbase’; (2) ‘footprint, trace’ (JDB 1967: 32). Martin glosses it as
‘footprint, trace, behind’ (1987: 387). The last meaning ‘behind’ is
certainly a semantic development of ‘footprint’ or ‘trace’. Third, we have
an interesting problem: in Western Old Japanese texts the word appears in
two different shapes: atô, with a kō-rui /ô/, and atö, with an otsu-rui /ö/.
Omodaka et al. suggest that the original form was atô, and that the form
atö is the later form due to an early merger of /ô/ and /ö/ after /t/ (JDB
1967: 32). However, there is a recent study by Bentley which indicates that
the traditional position on the early nature of the orthographic confusion of
/tô/ and /tö/ is mistaken (Bentley 1999). 292 There are only three examples
of atô spelled phonetically in Western Old Japanese. One appears in the
Nihonshoki kayō, one in the Man’yōshū, and the third is a phonetic gloss in
the Chinese text of the Nihonshoki:

291
It is also necessary to mention that asi (but not a) has a limited distribution in Ryukyuan:
only Shuri ʔasi ‘pig’s leg, leg’ (RGJ 1976: 124) and Nakijin ʔasi (Nakasone 1983: 8) are
attested, but the former is limited to idiomatic expressions, and the latter occurs in the
compound ʔasi’yuucaa ‘four-legged [animal]’. These are clear loans from Japanese.
Consequently, OJ a may not be a Proto-Japonic word.
292
In this article, Bentley convincingly demonstrates that tôru ‘to hold’ and töru ‘to take, to
grab’ were two different verbs in Western Old Japanese, which later merged in one
phonetic form toru.
228———Koreo-Japonica

ware ir-i-mas-i atô tôr-i tuma-N-tôr-i s-i-te 293

I enter-INF-HON-INF foot hold-INF wife-DV(INF)-hold-INF do-INF-
SUB
I entered inside, and holding [her] foot, [I] held [her] as [my] spouse,
and… (NK 96)
脚辺阿度陛
jiao bian (atô pê)
foot side ([read] atô pê) (NS I) 294

kaNse-nö mî-ye-n-u-Nka Nkötö-ku atô mô na-kî yö

wind-GEN see-PASS-NEG-ATTR-POSS like-INF trace PT no-ATTR life
impermanent life, which has not even a trace, like the wind that [one]
cannot see (MYS XV: 3625)
While both examples from the Nihonshoki clearly mean ‘foot’, the
example from MYS XV: 3625 means ‘trace’. The examples with WOJ atö
in phonetic spelling are much more numerous, although the lion’s share of
them are found in the Bussoku seki ka. As far as I can tell, all cases of WOJ
atö mean only ‘footprint‘. Some examples:
mî-atö tukur-u isi
HON-footprint make-ATTR stone
Stone, where [I] carve the footprint [of the Buddha] (BS 1)

könö mî-atö ya-yöröNtu pîkari-wo panat-i-iNtas-i

this HON-footprint eight-ten thousand light-ACC emit-INF-exit-INF
This footprint [of the Buddha] emits eighty thousand lights (BS 4)

atö-wo mî-tutu sinöp-am-u

footprint-ACC see(INF)-COOR praise-TENT-FIN
[I] will be looking at [Buddha’s] footprint and praise [it] (BS 6)

Therefore, if we disregard the evidence from the single example from

MYS XV: 3625, where atô appears as ‘trace’, 295 it looks as if we are
dealing with two different words here: WOJ atô ‘foot, base of the foot’ and
WOJ atö ‘footprint, trace’. I believe that ultimately these two words may

293
The traditional interpretation of this poem treats tuma as ‘edge (of clothes)’, not as
‘spouse’, cf., e.g., Tsuchihashi 1957: 188. Omodaka et al. follow the same interpretation
(JDB 1967: 32). However, such an interpretation faces two problems: first, it involves
understanding tôr- as ‘to grab’, i.e., ‘to grab the edge [of the clothes]’, which is not correct
(see the preceding footnote). Second, it leaves the prenasalized voiced /Nt/ in tumaNtôri
unexplained. I follow Bentley 1999 in interpreting tôr- as ‘to hold’. Consequently, I analyze
tumaNtôri as tuma-N-tôr-i, a contraction of *tuma n-i tôr-i (spouse DV-INF hold-INF)
‘holding as a spouse’.
294
Cited according to JDB 1967: 32.
295
It must be kept in mind that volume XV of the Man’yōshū does not have a good history
of textual transmission. Various kinds of misspellings are found in this particular volume.
 Lexical Comparisons———229

be partially related, as both are probably old compounds including OJ a

‘foot’, discussed in (323) above. Therefore, this etymology can be rejected
on phonetic, morphological, and semantic grounds.
(326) (R) MK àwól- ‘joins together, meets’ ~ OJ ap- ‘id’. 296 < PJK *ap-.
Whitman adds: “The OK (Silla) transcription 阿火 [apəl] ‘join together’
(Samkwuk saki Ch. 34) confirms the original stem final /p/. OK -l is the
attributive ending so /-ól/ in MK must be an extension of the original stem”
(1985: 244). There are two problems. First, MK àwól- (actually àGwól-)
means ‘to join together, to put together, to combine’, but it does not mean
‘to meet’, which is the basic meaning of OJ ap-. Therefore, the meaning
‘to meet’ is added to improve the semantic comparison. Nevertheless, the
semantics are not impossible, so if this were the only problem, the
comparison would be acceptable. Unfortunately, the second problem is
more serious: there is no internal Korean evidence for segmenting -ól
(more exactly -wól) either as a suffix or as a “stem extension.” Thus, we
face the problem of an unaccounted segment and the etymology should be
rejected.
Among the nine etymologies presented by Whitman for PJK *a- (318-
326), I accepted two as loans from Korean into Japanese, and one as a
possible cognate, although not perfect, and rejected the other six. It is very
strange that genetically related languages would have only one cognate
with initial *a-. This creates another potential gap, which strongly speaks
against a genetic relationship.
3.2.15 *ü-, *ö-
(327) (R) MK wùh ‘up’ ~ OJ upa-/upë < u + pa ‘place’ < PJK *u-
(Whitman 1985: 245).297 There are several serious problems with this
comparison. First, the segmentation of OJ upa- ~ upë as *u-pa is ad hoc,
because PJ *pa ‘place’ is a ghost (see [1] above), and because there is no
internal evidence for OJ *u- ‘top’ either. Second, we do not know whether
MK -h in MK wùh ‘top’ is a suffix or a part of the stem, but there is no
internal evidence for its being a suffix. Finally, MK -h- does not
correspond regularly to OJ -p-. Therefore, I reject this comparison as a
chance similarity.
(328) This number is absent from Whitman’s list.
(329) (R) MK wúlí ‘we, us, me’ ~ OJ wa, ware ‘id’. < PJK *bör ~ *bor
(Whitman 1985: 245). This comparison has several serious problems. First,
MK wúlí has only the plural meaning ‘we’. Second, OJ ware is just an
extended stem with the pronominal suffix -re of OJ wa ‘I, we’ (with plural
meaning attested very rarely), which can occur in isolation (especially in
Eastern Old Japanese), or followed by case markers (see Vovin 2005a:
219-232 for details). On the other hand, there is no internal evidence for
segmenting MK wúlí into *wú-lí. Third, the Proto-Japonic form was either
*wan[u] or *ban[u], depending on the treatment of OJ w- vs. Sakishima b-
(see 1.2.2.1). Finally, MK /wu/ does not correspond regularly to OJ /a/: the
296
The same etymology is found in Martin (1966), #120.
297
This comparison is also found in Martin (1966), #266.
230———Koreo-Japonica

suggested PJK *bör would require MK /u/ and OJ /ö/, while the suggested
PJK *bor would require MK /o/ and either OJ /u/ or /a/ according to
Whitman’s vocalic correspondences (Whitman 1985: 129). The
correspondence of MK Ø- to OJ w- is also irregular. Therefore, I reject this
etymology on the basis of its irregularity and ad hoc morphemic analysis.
(330) (R) MK wúmh ‘cellar, hole’; wùmùk / wòmòk ‘hollow, depressed’
~ OJ oku LL ‘interior’. Whitman notes: “Ryūkyūan (Shuri) uuku confirms
the OJ long vowel from medial *-m- loss. Normally this should occur only
after *u or *o, so we reconstruct PJK *umuko or *umoko” (1985: 245).
There are several problems with this etymology. First, this etymology is
found in the section illustrating PJK *ü- or *ö-. However, Whitman’s PJK
*ü requires a correspondence of MK /wu/ to OJ /u/, and his PJK *ö
requires a correspondence of MK /u/ to OJ /ö/ (Whitman 1985: 129).
Neither is present in this case. Reconstructing PJK *u- hardly helps,
because it should be reflected as MK /wo/ and OJ /u/ (Whitman 1985: 129).
Whichever way we go, we face an irregular correspondence not stipulated
by the proposed rules. Second, it is well known that OJ /ö/ and /u/ do not
combine within the same morpheme; therefore OJ oku can go back only to
pre-OJ *ôku. However, the correspondence of pre-OJ *ô to MK /wu/ is not
regular, either. Third, the loss of *-m- is a speculative proposal, which
cannot be verified on the basis of internal Japonic evidence; see also (136,
185-186, 212, and 293) above. In addition, Shuri uuku just indicates the
Proto-Japonic vowel length (Shimabukuro 2002: 203). Fourth, MK wúmh
is actually an ‘excavated cave’ or ‘underground dwelling‘, as can clearly
be seen from textual examples:

Chilce koz-ay-z wumh-ul hwuseng-i nil-osi-n-i … Cektwo anh-ay-z wumh-

ul cikum-ey pwo-zop-no-n-i …
Chilce edge-LOC-GEN cave-ACC later.saints-NOM speak-HON-
ATTR/REAL-NML … Red.island inside-LOC-GEN cave-ACC now-LOC
see-HUM-PRES-ATTR/REAL-NML
Later saints spoke about the caves at the bank of Chilce river … [we] see
[even] today the caves at Red Island (YP 5) 298

Fifth, I have doubts whether there is any internal connection between

MK wúmh ‘cave’ and MK wùmwúk- 299 ‘to be depressed’ due to the
difference in accentuation patterns and semantics. MK wòmwòk
(misspelled by Whitman as *wòmòk) is attested only as a nominal in
Middle Korean texts, since it can function as a verb only in combination
with the following ho- ‘to do’ (LCT 1987: 576; Nam 1997: 1098). Finally,
the semantic difference between MK wúmh ‘cave’ and OJ oku ‘interior’ is
quite substantial. I reject this etymology due to all these problems.
(331) (R) MK wùmúl ‘well’ < *wu + múl ‘water’ ~ OJ wi ‘id’. < PJK
*ör. Whitman cites the Old Korean form [öl] ‘well’ in reference to the
Samkwuk saki, and proposes a speculative development from OK öl > MK
298
Caves mentioned here were excavated and made into dwellings.
299
The form *wùmùk cited by Whitman is not attested.
 Lexical Comparisons———231

ul > wul “in initial position” (1985: 245), but since he does not indicate
either a character spelling of the word in Old Korean, or an exact reference
to the relevant volume (let alone leaf number) in the Samkwuk saki, it is
impossible to locate the source of his quote. In addition, this etymology
has several problems. First, there is no internal Korean evidence for the
segmentation of MK wùmúl ‘well’ into *wu (not otherwise attested) with
an unknown meaning + múl ‘water’. Second, OJ wi ‘well’ 300 can go back
to three possible forms: pre-OJ *wi, *wuy, 301 or *wöy. The form needed
for this comparison is pre-OJ *wöy, since the others will not work in this
case. However, given the three possible proto-forms, it leaves us with no
more than a one in three chance that this is correct. Even in the case of pre-
OJ *wöy we would still expect MK *u, rather than /wu-/, to preserve the
regularity of the correspondences, in accordance with Whitman’s rules. 302
Fourth, and most important, this etymology demonstrates the same
irregular correspondence as (329) above: MK Ø- : OJ w-, which cannot be
accepted. Fifth, the word for ‘well’ in any given language does not
necessarily have to be derived from the word for ‘water’, cf. English well
vs. water, German Brunnen ‘well’ vs. Wasser ‘water’, French puits ‘well’
vs. eau ‘water’, Russian kolodec ‘well’ vs. voda ‘water’, Mandarin
Chinese jing3 ‘well’ vs. shui3 ‘water’, Manchu šeri ‘well’ vs. muke ‘water’,
Chuvash pusă ‘well’ vs. šiv ‘water’, etc. Such a derivation becomes even
more suspicious in an SOV language like Korean, since the word for
‘water’ supposedly follows the alleged word for ‘well’, rather than the
reverse. Therefore, I reject this etymology.
I have rejected all four etymologies presented by Whitman for PJK *ü-,
*ö- (327, 329-331). Since the phonemes *ü- and *ö- are more uncommon
than *a- discussed above, the lack of cognates might either indicate
another gap speaking against a genetic relationship; or it might just point to
the fact that these phonemes were not present in the proto-language.

3.2.16 *u-, *o-

(332) (R) MK wòlhí ‘wild duck’ ~ OJ u F ‘cormorant’ < PJK *or
(+ [-ogi]) (Whitman 1985: 245). This etymology has two problems. The
first is that cormorants, while certainly ‘birds’, are not ‘ducks’. Comparing
one kind of bird to another is unlikely to be persuasive, even if both
happen to be water fowl. Second, there are serious phonological problems
in this comparison. MK wòlhí ‘wild duck’ goes back to PK *wòlókí, but

300
OJ *wi ‘well’ is not attested per se in phonetic writing in Old Japanese texts, but since
the kungana /wi/ spelled with the character 井 ‘well’ is well represented in various Old
Japanese sources, it leaves no doubt that the word in Old Japanese was the same as MJ wi
‘well’.
301
Some scholars may argue against pre-OJ *wu-i as a possible form, since pre-OJ *wu >
OJ u. However, it is most likely that the monophthongization *ui > ï took place first before
the merger of *î with *ï after *w-, and certainly before the disappearance of initial *w-
before /u/.
302
Initial MK /u-/ is rare. To the best of my knowledge it occurs only in ùph- ‘to recite’ and
ùstùm ‘top’ (LCT 1987: 604; Nam 1997: 1143, 1148-1149). Nevertheless, it needs to be
proven that in all other cases original PK *u- became MK /wu-/, and not some other vowel.
232———Koreo-Japonica

WOJ u ‘cormorant’ (not attested in Eastern Old Japanese or Ryukyuan)

can go back to pre-OJ *u or *um. Both pre-Old Japanese forms are
difficult to reconcile with PK *wòlókí: it is impossible to explain the loss
of *-k- in proto-Japonic, and even less so the disappearance of *-m- in
Proto-Korean, if the Proto-Japonic form was *um. Therefore, the
etymology has to be rejected.
(333) (R) MK wòlm- ‘moves’ (transitive and intransitive) ~ OJ um- H
‘gives birth to’ < PJK *orom- (Whitman 1985: 245). Besides the vague
semantics (the act of giving birth is not unique in requiring movement
from point A to point B), this etymology rests solely on the assumption of
*-r- loss in Old Japanese, which cannot be verified internally in this and
the many other cases surveyed above. Although in this case, *-r- loss
confirms to the accentual environment postulated by Whitman (H pitched
vowel, reflecting an original short vowel [Whitman 1985: 190 ff]), there
are etymologies that involve a vowel with L pitch in Old Japanese. Cf. e.g.,
(155) above. Therefore, I reject this comparison.
(334) (R) MK wòlòk- ‘rises’ ~ OJ aNkë- ‘id’. < PJK *orok- (Whitman
1985: 245). This etymology has many problems. First, MK *wòlòk- ‘to
rise’ is a ghost, as far as I can tell. There is a Middle Korean hapax
legomenon wòlwòk-hò- ‘to be full, to overflow’, attested only in the
Sincung yuhap (1576) (LCT 1987: 575; Nam 1997: 1096), which is
unlikely to be the word Whitman had in mind, or MK wòló- ‘to go up, to
climb’ (LCT 1987: 574; Nam 1997: 1097). MK wòló- has an alternate stem
wòlG-, which points to PK *wòlòk-, but *wòlòk- is not attested in Middle
Korean. Second, according to Whitman’s own vocalic correspondences,
MK /wo-/ does not correspond to OJ /a-/ (Whitman 1985: 245). Third, OJ
aNkë- is certainly not identical to MK wòló- ~ wòlG- ‘to rise‘, as it is a
transitive verb meaning ‘to raise‘, with aNkar- as the intransitive form.
Even if one accepts Martin’s reconstruction of aNkar- as PJ *anka-ra-
(Martin 1987: 674), PK *wòlòkó- ‘to go up, to climb’ and PJ *anka-ra- ‘to
rise’ still do not match due to the irregularity mentioned above and also
due to another irregular correspondence of PK *-lok- to PJ *-nk-.
Therefore, I reject this etymology.
(335) (R) MK wós ‘clothing’ ~ OJ uNsu LL ‘hairdress’ < PJK *oz
(Whitman 1985: 245). Since MK wós ‘clothes’ has a non-leniting -s- < PK
*-ns-, the etymology is phonetically plausible. However, there are two
problems with this etymology. The first is semantic: WOJ uNsu is really
not the ‘hairdress’ itself, but the various decorations, such as flower
branches, artificial flowers, or pieces of silver and gold, inserted into the
hair or headgear (JDB 1967: 115). The second problem is with its
distribution: WOJ uNsu is attested only in Western Old Japanese with the
meaning mentioned above, and in Middle Japanese as uzu with the
meaning ‘mica’, and possibly with the meaning ‘decoration on a horse
[harness]’ (JDB 1967: 115), but there are no attestations in Eastern Old
Japanese or Ryukyuan. It is remotely possible that this is a loan from
Korean into Japanese, but given that WOJ uNsu and MJ uzu indicate
‘decorations’ or material used for decorations (‘mica’), while MK wós
 Lexical Comparisons———233

refers just to simple ‘clothing’, chances are these two words are just
coincidentally similar. Therefore, I reject this etymology.
(336) (R) MK wòs, K och ‘lacquer’ ~ OJ urusi HHH ‘id’. < PJK *uros
(Whitman 1985: 245). This etymology has to be rejected, because there is
no internal evidence for *-r- loss in Korean, making the correspondence of
MK /wo/ to OJ /uru/ irregular. It is also inconceivable that a product such
as lacquer would have been known to the Proto-Koreo-Japonic speaking
community several thousand years ago.
(337) (C) MK :woy ‘melon’ ~ OJ uri LH ‘id’. 303 < PJK *uri (Whitman
1985: 245). This etymology can be accepted as a potential cognate, since
the correspondences are regular and the Japonic word is attested not only
in Western Old Japanese but also in Ryukyuan, including South Ryukyuan,
e.g., Taketomi ui, Kurojima urï, Hateruma u:ru, etc. (Miyara 1981: 329). If
it is indeed a cognate, the Proto-Japanese-Korean reconstruction should be
revised as *ori, since in this case OJ u- must be a product of raising *o- >
*u-.
Among the six etymologies presented by Whitman for PJK *o-, *u-
(332-337), I have rejected five and accepted one which can reflect only
PJK *o-. The absence of reliable etymologies for *u- indicates another
significant gap.
3.2.17 *e-
(338) (L) MK è(-) ‘which’ ~ OJ i- in iNtu ‘id’. < *in+tu < PJK *en-
(Whitman 1985: 245). This comparison was discussed above in 2.1.2.3. It
should be treated as a loan because neither MK ènú ‘which’ nor OJ iNtu-
‘wh-’ can be segmented as e- or i- on the basis of the internal evidence.
(339) (R) MK écúl- 304 ‘is dizzy, is disorderly’ ~ OJ oNti- ‘is frightened,
is surprised’; oNtörök- ‘surprise, frighten’; cf. also MJ odor- ‘dances,
cavorts’ < PJK *ejör- (Whitman 1985: 246). In addition to the semantic
differences, there is another problem. MK écúlèp- ~ écùl-hò- ‘be
dizzy/disorderly’ must be an onomatopoetic word by origin, as witnessed
by the Middle Korean form écùl-hò- (no independent verbal stem *écùl- is
attested) and MdK ecil-ecil or eccil-eccil ‘dizzily’. Because any
comparison involving onomatopoeia is methodologically unacceptable, I
reject this etymology.
(340) (L) MK èlí-, LMK 305 elisyek- ‘foolish, stupid’ ~ OJ oröka < orö-
+ -ka (adjectival suffix), orösöka ‘id’. 306 < PJK *er- (Whitman 1985: 246).
There are two problems with this etymology. First, the reconstruction of
PJK *er- simply evades the problem of the non-correspondence of MK -i
to OJ -ö in the second syllable, and the morphological segmentation of MK
èlí- as èl-í and WOJ orö- as or-ö cannot be justified. Second, WOJ oröka is
limited only to Central Japanese, as it does not appear in Eastern Old

303
This comparison is also found in Martin (1966), #140.
304
Only the forms écúlèp- and écùl-hò- ‘be dizzy’ as well as a causative form écùli- ‘to
make dizzy’ are attested (LCT 1987: 549; Nam 1997: 1059).
305
More precisely, Early Modern Korean.
306
This comparison is also found in Martin (1966), #227.
234———Koreo-Japonica

Japanese, and the only attestation in Ryukyuan is Shuri ʔuruka ‘stupid’

(RGJ 1976: 562), which is isolated and likely to be a loan from mainland
Japanese. As I noted in (19) above, MK -Vli may go back to PK *-VlVi,
because PK *-Vli normally > MK -Vy, as in (337) above. This, alongside
with the limited distribution in Japonic, suggests that the word is a loan
from Korean to Japonic.
(341) (R) MK èlkwúl 307 ‘face’ ~ OJ aNkî HL ‘upper jaw, gills‘, MdJ
ago HL < ‘jaw’ < PJK *erkür (Whitman 1985: 246). There are several
problems with this etymology. First, OJ aNkî ‘upper jaw, gills’ and MdJ
ago are unlikely to have an etymological connection with each other due to
insurmountable difficulties of relating OJ /î/ to MdJ /o/ in the second
syllable, as well as the differences in accentuation: Modern Japanese
(Tokyo) LH would be an accent class related to Middle Japanese HL. 308
Second, WOJ /î/ does not correspond regularly to MK /wu/. Third, MK
-lk- does not correspond regularly to WOJ -Nk-; in addition, Whitman
claimed that MK -lk- should correspond to OJ -n- (see [16] above). Finally,
the semantics of the comparison are far from ideal, especially considering
that MK èlkwùl actually means ‘shape, appearance’; the meaning ‘face’ is
not attested before Early Modern Korean (LCT 1987: 552; Nam 1997:
1064). Therefore, I reject this etymology.
(342) (R) MK èlú- ‘has sexual relations, marries’, :el.wun ‘elders’ ~ OJ
oi- 309 ‘gets older‘, oya LL ‘parent’ < PJK *ere- (Whitman 1985: 246).
MK :el.wun ‘elder, adult’ is in fact :elGwun, as indicated by the Middle
Korean ‘syllabification’, so it is unlikely to be etymologically connected to
MK èlú- ‘to marry’. The comparison of the latter to OJ oyi- ‘to get older’
and oya ‘parent’ also should be rejected, because MK -l- does not
correspond regularly to WOJ -y-; see also (17) above.
(343) (L) MK émí ‘mother’ ~ OJ omo LH ‘id’. 310 (Whitman 1985: 246).
MK émí ‘mother’ is probably a contraction from émà ‘mother’ +
diminutive suffix -í, cf. the MK honorific form émà-:nim ‘mother’, which
is also attested (LCT 1987: 545; Nam 1997: 1052). Only MK émà ‘mother’
can be somehow compared to WOJ omo ‘id’., because MK -i in émí
‘mother’ certainly does not correspond regularly to WOJ -o. However,
even in the former case regularity remains a problem, since MK -a does
not correspond regularly to WOJ -o. Of course, one can speculate about the
labialization of -a > -o after /m/ in Western Old Japanese. We should not
forget EOJ amo ‘mother’, which further complicates the correspondence
scheme. The Ryukyuan reflexes are also irregular: Nase, Koniya, Taketomi
ʔamma, Kamezu ʔama, Psara anna, Ishigaki appa, Hateruma abwa,
Kurojima abu (Hirayama 1966: 308; 1967: 251). They support EOJ a-
rather than WOJ o- in the first syllable. My tentative solution is that WOJ
omo is a loan from Korean, while EOJ amo and the Ryukyuan forms

307
The actual accentuation is èlkwùl (LCT 1987: 552; Nam 1997: 1064), not *èlkwúl.
308
The evidence for WOJ aNkî rather than aNkï is oblique at best, so the first part of this
objection may be overruled, but the difference in accentuation remains.
309
Actually, WOJ oyi- ‘to get older’ (Vovin 2009: 420-421).
310
This comparison is also found in Martin (1966), #145.
 Lexical Comparisons———235

represent a native Japonic (and unrelated) word. 311 In any case, even if this
word is someday proven to be a cognate, it has little value, as all
comparisons involving words for ‘mother’ or ‘female’ with [m] are suspect.
(344) This number is absent from Whitman’s list.
(345) (L) MK èp- ‘bears it, carries it on the back’ ~ OJ op- ‘id’. 312 <
PJK *öp- 313 (Whitman 1985: 246). There are three problems with this
etymology. First, one should expect OJ a-, not o-, because according to
Whitman’s sound laws, MK /e/ corresponds to OJ /ö/ either in final
position or in a word with another /ö/ (Whitman 1985: 129). Second, MK
èp- has a non-leniting -p-, making the Proto-Korean form *enp-, which
cannot be compared to WOJ op- (not attested in Eastern Old Japanese),
since WOJ *oNp- is needed. Third, attestations in Ryukyuan are few and
far between. I was able to find only Nakijin ʔuuruN (Nakasone 1983: 60),
Shuri ʔuuyuN (RGJ 1976: 572), and Yaeyama (without any indication of
location) offasïN (Miyara 1981: 225). The Yaeyama word must be a loan
from mainland Japanese, because for a true cognate initial u- would be
expected. The same conclusion seems to be appropriate for Shuri, since it
has restricted semantics: it refers to a horse carrying a load or to carrying a
responsibility or sin (RGJ 1976: 572). The only example provided in
Nakasone 1983: 60 also refers to a cart carrying loads. Therefore, on the
basis of the irregularity in the correspondences and the limited distribution
within Japonic I conclude that WOJ op- is a loanword from Korean.
(346) (R) MK èpí- ‘great-, respected person’ in èpèsí ‘parents‘, èpí
‘father‘, èpímòt < èpí + mòt ‘older brother’, ‘clan, family chief’ ~ OJ opo-
‘great’ < PJK *epe- or *epö-. Whitman notes that MK èwúy- ‘broad, large’
may also be related (1985: 246). This etymology has numerous problems.
First, MK èpèsí 314 ‘parents’ is a quite transparent compound consisting of
MK èpí ‘father’ and MK èzí < *èsí ‘mother’. The same is true of MK
èpímòt ‘family chief’ (which certainly is a function of a father), also a
transparent compound consisting of MK èpí ‘father’ and MK mòt ‘senior,
elder’. No ‘elder brothers’ are even remotely present in this case. Therefore,
MK èpí simply means ‘father’ and there is no need to assign the meaning
‘great-, respected person’ to it. Second, there are also MK àpí ‘father’ and
àpá-:nim ‘father (hon.)’ that point to PK *a rather than *e in the first
syllable, and this makes the comparison with OJ opo ‘big’ even less
credible due to the irregularity of the vowel correspondences. Third, both
MK àpí and èpí ‘father’ have a non-leniting -p-, which points to PK *anpa
or *enpe, incompatible with OJ opo ‘big’. Fourth, the basic meaning of OJ
opo is ‘big, large’, not just ‘great’. Finally, MK èwúy- ‘broad’ is in fact

311
The South Ryukyuan forms with -p- and -b- are problematic cognates.
312
This comparison is also found in Martin (1966), #165.
313
As a Proto-Japanese-Korean reconstruction this example is misplaced: it should be
included in Whitman’s *ü-, *ö- section. However, on the basis of Whitman’s vocalic
correspondences PJK *ö is reconstructed on the basis of the correspondence of MK /u/ to
OJ /ö/, and PJK *e on the basis of the correspondence of MK /e/ to Old Japanese ‘finally or
in a word with other /ö/, otherwise /a/’ (Whitman 1985: 129). Neither of these two
conditions seems to work here, as we would expect OJ /a/ rather than /o/.
314
This is much more frequently attested as èpèzí (LCT 1987: 546; Nam 1997: 1053).
236———Koreo-Japonica

èGwúy-, 315 but even if it were to go back to PK *epwuy-, there would still
be an irregularity in the vocalic correspondences of the second syllable.
Therefore, I reject this etymology as a chance resemblance with vague
semantics.
Among the eight etymologies presented by Whitman for PJK *e- (338-
346), I rejected five and accepted three as Korean loans into Japanese. The
total absence of reliable etymologies for *e- may indicate another
important gap.

3.2.18 *i-
(347) (L) MK -í (nominative suffix) ~ OJ -i ‘id’. < PJK *-i (Whitman
1985: 246). This etymology is discussed and treated as a loan from Korean
to Japanese in 2.1.1.1 above.
(348) (R) MK -í (deverbal nominative suffix) ~ OJ -i continuative
[ren’yōkei] suffix) < PJK *-i (Whitman 1985: 246). As indicated earlier,
Old Japanese continuative -i must be distinguished from its homophone,
Old Japanese nominalizer -i, due to the accentual differences (Martin 1987:
211). The Old Japanese nominalizer -i looks exactly like its Korean
counterpart, but there is an important morphological difference: while the
Old Japanese nominalizer -i is a true suffix that can follow verbal roots,
MK -i is in fact a bound noun that only follows attributive verbal forms.
This difference in morphology suggests different, and probably unrelated,
origins for these two markers.
(349) (R) MK :il- ‘boils it’ ~ OJ ir- ‘cooks it, roasts it’ 316 < PJK *i:r-
(Whitman 1985: 246). The main problem with this etymology is that
MK :il- actually means ‘to clean, to wash (esp. rice)’ and not ‘to cook’
(LCT 1987: 622; Nam 1997: 1175). In addition, since MK :il- < PK *ilu-,
it leaves PK *-u in comparison to OJ ir- unaccounted for. Therefore, this
etymology can be safely rejected.
(350) (R) MK ílhóy, 317 ílhì ‘wolf’ ~ OJ inu LL ‘dog’ < PJK *irko
(Whitman 1985: 246). This etymology has to be rejected, because neither
MK -lh- nor MK -i corresponds regularly to OJ -n- or -u respectively.
(351) (R) MK ìph- ‘recites or sings it’~ OJ ip- ‘says it’ 318 < PJK *ipV-
(Whitman 1985: 246). There are three problems with this etymology. First,
there is another Middle Korean form ùph- ‘to recite’ (LCT 1987: 604; Nam
1997: 1149), showing a different vowel. Second, there is EMdK ulph- ‘to
recite’ (LCT 1987: 603; Nam 1997: 1146), which agrees with MdK ulph-
‘id’. In spite of the fact that the Middle Korean forms are attested earlier,
and the Early Modern Korean and Modern Korean forms are attested later,
the Early Modern and Modern are more archaic, since there was a process
315
This may reflect either PK *ekwuy- or *epwuy-, but since the word is not attested in
dialects, it is impossible to verify the source.
316
As far I can tell, WOJ ir- refers only to ‘roasting’ and not to cooking in general (JDB
1967: 106).
317
MK ílhòy (not ílhóy) ‘wolf’ is a hapax legomenon attested only in the Kwukup II: 64
(LCT 1987: 625; Nam 1997: 1181).
318
Ōno Susumu suggested a comparison of OJ ip- ‘to say’ with Korean ip ‘mouth’ (Ōno
1957: 177).
 Lexical Comparisons———237

of -l- deletion before a labial in Middle Korean; cf. MK nèp- ~ MdK nelp-
‘to be wide’. Finally, MK -ph- indicates PK *-kup- or *-puk-. The first of
these reconstructions would rule out the comparison completely, and the
second one would make it difficult at best, as there is nothing in Japanese
that would correspond to the segment *uk in Korean. As this etymology
has a number of irregularities, it can be safely rejected.
(352) (R) MK i ‘this’ (proximal demonstrative) ~ OJ i- ‘this’ in ima
‘now, this time’ (Whitman 1985: 246). This etymology was discussed and
rejected earlier in the section on demonstrative pronouns (2.1.2.3).
Among the six etymologies presented by Whitman for PJK *i- (347-
352), I have rejected five and accepted one as a Korean loan into Japanese.
The total absence of reliable Koreo-Japonic cognates with *i- indicates
another important gap.

3.3 STATISTICAL SUMMARY OF LEXICAL COMPARISONS

AND CONCLUSION
I divided all of Whitman’s etymologies into three groups: (1) possible
cognates, (2) obvious loans from Korean into Central Japanese, and (3)
etymologies rejected for various reasons. These three groups have the
following statistics:

(1) Possible cognates 11

(2) Obvious loans 75
(3) Rejected etymologies 261
Total: 347
Let me start from the largest group, the rejected etymologies. These
were rejected for different reasons: irregularity in the phonetic
correspondences, faulty morphemic analysis, philological problems
(including ghost words), vague or unreliable semantics, and other
difficulties. More often than not more than one problem was present.
The second largest group, consisting of obvious loans, was rejected as
evidence for a genetic relationship because, for a variety of reasons, they
cannot be perceived as genetic cognates. This leaves us with only twelve
potential cognates.
Because there are only eleven potential cognates, I believe we have
only two possible explanations: (a) Japonic and Korean are indeed
genetically related, but the relationship is very old, or (b) Japonic and
Korean are not likely to be genetically related, and these twelve possible
cognates represent an older layer of loans. Let us tabulate these twelve
potential cognates and see whether they provide us with any regular
correspondences.
238———Koreo-Japonica

Chart 35:
Potential Koreo-Japonic cognates
Gloss Middle Korean Old Japanese ##
‘fire’ púl pï ~ *pö- 33
‘make fire’ tàhí- tak- 54
‘fill [with water]’ :tam- tamë-/tamar- 59
‘hold/take’ tùl- tör- 75
‘crane’ twúlwúmí turu 80
‘painful’ kwó:lwop- kuru-si- 135
‘crab’ :key kani 152
‘suffice’ cólà- tar- 183
‘clear/wash/bleach’ sèl- ‘clear/wash’ saras- ‘bleach/wash’ 225
‘field’ nwón ‘paddy’ nô ‘field’ 297
‘melon’ :woy uri 337

One can clearly see that among these etymologies ‘crab’ and ‘suffice’
include correspondences that do not appear anywhere else in the list above.
‘Clear/bleach/wash’ and ‘painful’ have other problems that were discussed
earlier. Two other words still seem problematic: it is difficult to explain the
final -í in MK tàhí- ‘to make fire’ which does not correspond to anything
in OJ tak- ‘id’. Therefore, it seems more prudent to remove these six
etymologies and rewrite the above chart so it will include only six reliable
cognates:

Chart 36:
Reliable Koreo-Japonic cognates
Gloss Middle Korean Old Japanese ##
‘fire’ púl pï ~ *pö- 33
‘fill [with water]’ :tam- tamë-/tamar- 59
‘hold/take’ tùl- tör- 75
‘crane’ twúlwúmí turu 80
‘field’ nwón ‘paddy’ nô ‘field’ 297
‘melon’ :woy uri 337

The remaining six possible cognates seem to be impeccable in the

regularity of their phonetic correspondences, so a remote genetic
relationship seems to be a possibility. But there are several problems.
(1) The existence of only six cognates suggests a very remote
relationship, which would go back many thousands of years. However, two
of these cognates, ‘paddy/field’ and ‘melon’ indicate the presence of
agriculture, which did not exist in East Asia more than 3,000 to 4,000
years ago. Even if a genetic relationship were taken as a fact, it would be
extremely difficult to pinpoint the homeland of the Koreo-Japonic proto-
language.
(2) It is unlikely that after more than a century of searching for cognates
new reliable cognates will be found in large numbers.
 Lexical Comparisons———239

(3) It is significant that the seventy-five obvious loanwords outnumber

the six reliable cognates twelve and a half times.
(4) It is very conspicuous that these cognates do not include any terms
for body parts, which are typically found as cognates in uncontroversial
language families.
Given all these considerations, plus the fact that no common
paradigmatic morphology is attested for Koreo-Japonic, it seems that the
more viable solution would be to view the ‘cognates’ as loanwords, too.
But how are these loanwords different from the seventy-five obvious
loanwords? In this case, the distribution appears once again. The
loanwords that constitute the major portion of the Koreo-Japonic
hypothesis are mostly attested in Central Japanese. Most of them represent
comparatively late loans from Old Korean into Central Japanese and were
probably borrowed between the late fourth and the late seventh centuries
AD. A few of them are possibly much older, since they are present in all
branches of Japonic, like the word for ‘island’. On the basis of phonology,
it is possible to determine that the direction of the borrowing was from
Korean to Japonic. Therefore, these seventy-five loanwords demonstrate
that Japonic was strongly influenced by Korean. This is especially
conspicuous in the case of Central Japanese, which became a Koreanized
version of Japonic. This Koreanization was strongest in Western Old
Japanese, where we can see not only a large amount of lexical borrowing
that resulted in doublets, but also significant borrowing of morphological
markers from Korean into Western Old Japanese (Vovin 2007). As in the
case of Norman influence on English, some Korean loans into Western Old
Japanese, both lexical and morphological, turned out to be short-lived,
disappearing from later stages of the language.
The status of the loanwords that were initially termed ‘possible
cognates’ is completely different. They are found throughout the Japonic
language family, and nothing in their phonology indicates that they must
be loans from Korean. In recent years, it has gradually become more
apparent that Korean has a Japonic substratum, as indicated, for example,
by the seemingly Japonic elements in pseudo-Koguryo place names. My
solution for these ‘possible cognates’ is that they also represent a Japonic
lexical substratum in Korean, although it cannot be completely ruled out
that they may be the earliest layer of Korean loanwords in Japonic,
acquired prior to migration of proto-Japonic speakers to the Japanese
archipelago.
In conclusion, first, it seems highly unlikely that any evidence presented
thus far in favor of a Koreo-Japonic genetic relationship can be accepted.
The nature of the relationship between Korean and Japonic seems to be
areal, not genetic. The present-day similarities between Korean and
Japanese are the product of a convergence that took place in approximately
the last sixteen hundred years; they are not the residue remaining after
several thousand years of divergence from some common source. Second,
it appears that the influence of Korean on Central Japanese was much
greater than the influence of Japonic on Korean. Central Japanese (but not
240———Koreo-Japonica

Japonic) was heavily influenced by Old Korean or a language closely

related to Old Korean (possibly Paekche) during the Kofun and Asuka
periods. That influence manifests itself most clearly in Western Old
Japanese and in Middle Japanese (and modern day descendants of Middle
Japanese), but less in Eastern Old Japanese, and practically not at all in
Ryukyuan. Modern Korean has been influenced by Japanese in modern
times too, albeit to a lesser extent. Third, in spite of the long history of
research on a possible genetic relationship between Korean and Japanese,
the issue remains controversial, with problematic sound correspondences
and very few supporting lexical and non-paradigmatic grammatical
parallels. Close inspection reveals that the majority of genetic comparisons
between Japanese and Korean accepted today can be treated either as loans
or as chance resemblances. This is due to philological, phonological
(regularity of correspondences), semantic, or functional problems (or as a
combination thereof). I believe that rather than attempting to reconstruct a
Koreo-Japonic proto-language, which might have never existed, we should
concentrate our efforts on unveiling the history of contacts between these
two languages, something that has been largely neglected.
It cannot be overemphasized that the previous work by Martin,
Whitman, and others, who meticulously collected different etymologies
over the years and tried to use them to prove a genetic relationship, is
based on high scholarly standards. Their work also paved the road to this
book. The field of comparative Koreo-Japonic is presently in a period of
stagnation, and I encourage the supporters of the genetic relationship
theory to come forward with refutations of my critical assessment of these
etymologies and my general conclusion. I may be wrong in my conclusion
that the relationship between Korean and Japonic is areal rather than
genetic. However, I hope at least that this book will trigger a lively and
friendly discussion and will help to reinvigorate the field. Without such
discussion, the truth cannot emerge.
REFERENCES
An Pyenghuy, and Yi Kwangho. 1990. Cwungsey kwuke munpep non [A
Middle Korean Grammar]. Seoul: Hakyensa.
Antonov, Anton. 2004. Poster presentation at the Second Oxford-Kōbe
Linguistics Seminar on the History and Structure of Japanese.
September 26-29, St. Catharine College (University of Oxford), Kōbe
Institute.
Aston, W. G. 1879. ‘A comparative study of Japanese and Korean
languages’. Journal of the Royal Asiatic Society, 11: 317-364.
Bentley, John R. 1997. MO and PO in Old Japanese. Master’s thesis,
University of Hawai‘i at Mānoa.
———. 1999. ‘The verb TORU in Old Japanese’. Journal of East Asian
Linguistics 8: 131-146.
———. 2001. A Descriptive Grammar of Early Old Japanese Prose.
Leiden: E.J. Brill.
Burykin, Alexei A. 1987. ‘Proiskhozhdenie cheredovanii nachal’nykh
soglasnykh v nivkhskom iazyke [The origin of initial consonant
alternations in the Nivx language]’. In: Iazyki Sibiri i Mongolii [The
languages of Siberia and Mongolia], ed. E. I. Ubriatova, 185-190.
Novosibirsk: Sibirskoe otdelenie Akademii nauk SSSR.
Ceycwu pangen yenkwuhoy, eds. 1995. Ceycwue sacen [A dictionary of
the Ceycwu language]. Ceycwuto.
Choy Hakkun. 1978. Hankwuk pangen sacen [Korean dialect dictionary].
Seoul: Hyenmunsa.
Doerfer, Gerhard. 1963. Türkische und mongolische Elemente in
Neupersischen. Band 1. Wiesbaden: Frank Steiner Verlag.
Frellesvig, Bjarke. 2001. ‘A Common Korean and Japanese Copula’.
Journal of East Asian Linguistics 10: 1-35.
———. 2003. ‘Pre-Old Japanese Verbal Morphology’. Paper presented at
the XVIth International Conference on Historical Linguistics.
University of Copenhagen, August 11-15.
Frellesvig, Bjarke, and John B. Whitman. 2003. ‘Evidence for Seven
Vowels in Proto-Japanese’. Paper presented at the XVIth International
Conference on Historical Linguistics. University of Copenhagen,
August 11-15.
Gruzdeva, Ekaterina. 1997. ‘Aspects of Nivkh Morphophonology: Initial
Consonant Alternation after Sonants’. Journal de la Société Finno-
Ougrienne 87: 76-96.
Hankul hakhoy, eds. 1999. Wuli mal khun sacen [A big dictionary of the
Korean language]. 2 vols. Seoul: Emunkak.
Hankwukhak yenkwu wen [Institute for Korean studies], eds. 1988.
Wenpen kwuke kwuk munhak chonglin, 6 [Comprehensive series of
facsimiles on Korean language and literature]. Seoul: Tayceykak.
242———Koreo-Japonica

Hattori Shirō. 1959. Nihongo no keitō [The origins of the Japanese

language]. Tokyo: Iwanami.
———. 1976. ‘Ryūkyū hōgen to hondo hōgen’ [Ryukyuan dialects and
Japanese dialects]. Okinawagaku no reimei. Tokyo: Okinawa bunka
kyōkai, pp. 7-55.
———. 1978-1979. ‘Nihon sogo ni tsuite’ [On Proto-Japonic]. Gekkan
gengo, 1-22.
———. 1979a. ‘Nihon sogo ni tsuite’ [On Proto-Japonic], rensai 21.
Gengo 8.11: 97-107.
———. 1979b. ‘Nihon sogo ni tsuite’ [On Proto-Japonic], rensai 22.
Gengo 8.12: 100-114.
Hayata Teruhiro. 1998. ‘Jōdai Nihongo no onsetsu kōzō to o-retsu kō-otsu
no betsu’ [The syllabic structure of Old Japanese and kō-otsu
distinctions in the o row]. Onsei kenkyū 2.1: 25-33.
Hendriks, Peter. 1992. Adverbial and Adnominal Modification in Old
Japanese: Copular particles and predication. Ph.D. diss., Yale
University
———. 1994. ‘Adverbial Modification in Old Japanese.’ Japanese/
Korean Linguistics, vol. 4, ed. Noriko Akatsuka, 120-136. Stanford
University: Center for the Study of Language and Information.
Hino Sukenari. 2003. ‘Nihon sogo no boin taikei – jōdai Azuma hōgen
shiryō ni yoru saikō [Proto-Japanese vowel system – a reconstruction
based on Eastern Old Japanese data]’. Nihongo keitōron no
genzai/Perspectives on the Origins of the Japanese Language, ed.
Vovin, Alexander and Osada Toshiki, 187-206. Kyōto: Kokusai Nihon
bunka kenkyū sentā.
Hirayama Teruo. 1966. Ryūkyū hōgen no sōgōteki kenkyū [A
comprehensive study of the Ryukyuan dialects]. Tokyo: Meiji shoin.
———. 1967. Ryūkyū Sakishima hōgen no sōgōteki kenkyū [A
comprehensive study of the Sakishima Ryukyuan dialects]. Tokyo:
Meiji shoin.
———. 1986. Amami hōgen kiso goi no kenkyū [A study of basic
vocabulary in Amami dialects]. Tokyo: Kadokawa.
———. 1989. Zenkoku akusento jiten [A dictionary of pitch accent in
Japanese dialects]. Tokyo: Tōkyōdō.
Hokama Shuzen, ed. 1995. Okinawa kogo daijiten [A big dictionary of
premodern Okinawan]. Tokyo: Kadokawa.
Hyen Phyengho et al., eds. 1995. Ceycwuto e sacen [A dictionary of the
Ceycwuto language]. Ceycwuto [no publisher].
Igarashi Jin’ichi. 1969. Jōdai kanazukai jiten [A dictionary of Nara period
orthography]. Tokyo: Shōgakukan.
Iha Fuyū. 1927 (1974). ‘Chōsenjin no hyōryūki ni arawareta jūgo seimakki
no nantō [Ryukyuan islands at the end of the fifteenth century as
described in the drifting diaries of Koreans]’. Shigaku zasshi 38.12.
Reprinted in: Iha Fuyū zenshū [Complete works of Iha Fuyū], 5: 54-99.
Tokyo: Heibonsha.
 References 243

Ikema Nae. 1998. Yonaguni kotoba jiten [A dictionary of the Yonaguni

dialect]. Yonaguni island: Privately printed.
IKJ = Ōno Susumu, Satake Akihiro, Maeda Kingorō. 1990[1974]. Kogo
jiten [A dictionary of Pre-modern Japanese], Tokyo: Iwanami.
Jakobson, Roman. 1971. ‘Notes on Gilyak’. Roman Jakobson: Selected
Writings. Vol.2, Word and Language. The Hague and Paris: Mouton.
JDB = Omodaka Hisataka et al., ed. 1967. Jidaibetsu kokugo daijiten:
jōdai hen [A large dictionary of the national language by periods: Old
Japanese]. Tokyo: Sanseidō.
Kanazawa Shōzaburō. 1910. Nikkan ryō kokugo kankei ron [A treatise on
the relationship of Japanese and Korean]. Tokyo: Sanseidō.
Karimata Shigehisa. 1999. ‘Onsei no men kara mita Ryūkyū no hōgen
[Ryukyuan dialects from the view point of phonetics]’. Kotoba no
kagaku, 9: 13-86.
Kawamoto Eiichirō. 1973. ‘Toyama-ken Shō-gawa ryūiki ni okeru ‘wa’ to
‘ba’ no bunpu to sono kaishaku [The distribution of /wa/ and /ba/ in the
basin of the Shō river in the Toyama prefecture]’. Kokugogaku kenkyū,
12. Reprinted in Nihon rettō hōgen sōsho 12. Hokuriku hōgen kō 2: 63-
77. Tokyo: Yumani shobō.
Kim Chwunghoy et al. 1993. Hankwuk pangen calyo cip. Vol. 8,
Kyengsang Namto phyen [Collection of materials on Korean dialects.
Vol. 8, Collection of Southern Kyengsang dialect materials]. Sengnam:
Hankwuk chengsin munhwa yenkwuwen.
———. 1995. Hankwuk pangen calyo cip. Vol. 9, Ceycwuto phyen
[Collection of materials on Korean dialects. Vol. 9, Collection of the
Ceycwuto dialect materials]. Sengnam: Hankwuk chengsin munhwa
yenkwuwen.
Kim Cwuwen [Juwon]. 1994. Moum cohwa yenkwu [A study of vowel
harmony]. Taekwu: Yengnam tayhakkyo chwulphanpu.
Kim Pyengcey. 1980. Pangen sacen. Phyengyang: Kwahak paykkwa sacen
chwulphansa.
Kim Thaykyun. 1986. Hampuk pangen sacen. Seoul: Cenghwun
chwulphansa.
Kim Wancin. 1994[1973]. Cwungsey kwuke sengco-uy yenkwu [A study of
tone marks in Middle Korean]. Kwukehak chongse 4 [Series on
National Linguistics 4]. Seoul: Thap chwulphansa.
King, Ross. 1988. ‘Towards a History of Transitivity in Korean’. Paper
presented at the Conference on the Theory and Practice of Historical
Linguistics. University of Chicago, April 27.
KKJ = Saeki Umetomo and Mabuchi Kazuo. 1969. Kogo jiten [A
dictionary of premodern Japanese]. Tokyo: Kōdansha.
Kojima Noriyuki, Kinoshita Masatoshi, and Satake Akihiro, eds. 1971-
1975. Man’yōshū 1-4. Nihon koten bungaku zenshū [A complete
collection of Japanese classical literature], Vols. 2-5. Tokyo:
Shōgakukan.
244———Koreo-Japonica

Kokuritsu kokugo kenkyūjo, eds. 1950. Hachijō no gengo chōsa

[Investigation of the Hachijō language]. Kokuritsu kokugo kenkyūjo
hōkoku. Vol. 1. Tokyo: Shūei shuppan.
Kolpakchi, Elena M. 1956. Ocherki po istorii iaponskogo iazyka. Vol. 1,
Morfologiia glagola [Sketches on the history of the Japanese language.
Vol. 1, The verb morphology]. Moscow-Leningrad: Nauka.
Kreinovich, Erukhim A. 1972. ‘O dialektakh nivkhskogo iazyka [On the
dialects of the Nivx language]’. Lingvisticheskie issledovaniia 1972.1:
157-174.
Kwen Sangno. 1994. Hankwuk cimyeng yenhyek sacen [An encyclopedic
dictionary of the histrocal change of Korean place names]. Seoul: Ihwa
munhwa chwulphansa.
Lass, Roger. 1984. Phonology. Cambridge: Cambridge University Press.
LCT = Yu Changton. 1987[1964]. Yico e sacen [A dictionary of the
Korean language of the Yi dynasty period]. Seoul: Yensey tayhakkyo
chwulphanpu.
Ledyard, Gari 1975. ‘Galloping Along with the Horseriders: Looking for
the Founders of Japan.’ The Journal of Japanese Studies, 1.2: 217-254.
Lessing, Ferdinand D., ed. 1995[1960]. Mongolian-English Dictionary.
Bloomington: Mongolia Society.
Mabuchi Kazuo. 1972. Jōdai no kotoba [Old Japanese]. Tokyo: Shibundō.
Martin, Samuel E. 1966. ‘Lexical Evidence Relating Japanese to Korean’.
Language 42.2: 185-251.
———. 1968a. A Korean-English Dictionary. Co-authored with Yang Ha
Lee and Sung-Un Chang. New Haven and London: Yale University
Press.
———. 1968b. ‘Grammatical Evidence Relating Korean to Japanese’.
Eighth Congress of Anthropological and Ethnological Sciences B-9:
405-407.
———. 1987. The Japanese Language Through Time. New Haven and
London: Yale University Press.
———. 1988[1975]. A Reference Grammar of Japanese. Tokyo and
Rutland: Tuttle.
———. 1990. ‘Morphological Clues to the Relationship of Japanese and
Korean’. In Philip Baldi, ed. Linguistic Change and Reconstruction
Methodology, 483-509. Trends in Linguistics: Studies and Monographs
45.
———. 1991. ‘Recent Research on the Relationships of Japanese and
Korean’. In Sydney M. Lamb and E. Douglas Mitchell, eds. Sprung
from Some Common Source, 269-292. Stanford: Stanford University
Press.
———. 1992. A Reference Grammar of Korean. Tokyo and Rutland:
Tuttle.
———. 1995. ‘On the Prehistory of Korean Grammar: Verb Forms’.
Korean Studies 19: 139-150.
 References 245

———. 1996. Consonant Lenition in Korean and the Macro-Altaic

Question. Center for Korean Studies Monograph 19. Honolulu: Center
for Korean Studies.
———. 1997. ‘How Did Korean Get -l for Middle Chinese Words Ending
in -t?’ Journal of East Asian Linguistics 6.3: 263-271.
———. 1998. ‘Copular structures in Korean and Japanese’. Unpublished
manuscript.
———. 2000. ‘How Have Korean Vowels Changed Through Time?’
Korean Linguistics 10: 1-59.
Matsuda Masayuki et al. 1991. Ōita ken shi. Hōgen hen [History of Oita
Prefecture. Volume on the dialects]. Ōita ken shuppansha.
Miller, Roy A. 1977. ‘The Altaic accusatives in the light of Old and
Middle Korean’. Altaica. Proceedings of the 19th annual meeting of the
Permanent International Altaistic Conference, 157-170. Helsinki:
Suomalais-Ugrilainen Seura.
Mineya Tōru. 1972. Etsunan kanji on no kenkyū [A study of Vietnamese
readings of Chinese characters]. Tokyo: Tōyō bunko.
Miyajima Tatsuo. 1971. Koten taishō goi hyō [A comparative chart of
vocabulary in classical literature]. Tokyo: Kasama shoin.
Miyake, Marc H. 1999. The Phonology of Eighth Century Japanese
Revisited: Another Reconstruction Based upon Written Sources. Ph.D.
diss., University of Hawai‘i.
———. 2003a. Old Japanese: A phonetic reconstruction. London:
RoutledgeCurzon.
———. 2003b. ‘Philological Evidence for *e and *o in Pre-Old Japanese’.
Diachronica 20.1: 81-137.
Miyara Tōsō. 1980. Yaeyama goi [The lexicon of Yaeyama]. Miyara Tōsō
zenshū. Vol. 8a. Tokyo: Daiichi shobō.
———. 1981. Yaeyama goi sakuin [The index to the lexicon of Yaeyama].
Miyara Tōsō zenshū. Vol. 8b. Tokyo: Daiichi shobō.
Mizushima Yoshiharu. 1974. Kōchū Man’yōshū Azuma uta Sakimori uta
[Annotated Azuma and Borderguards poems in the Man’yōshū]. Tokyo:
Kasama shoin.
———. 1984. Man’yōshū Azuma uta honbun kenkyū narabi ni sōsakuin
[A study of the original text of Azuma poems in the Man’yōshū with a
comprehensive index]. Tokyo: Kasama shoin.
———. 2003. Man’yōshū Sakimori uta zen chūshaku [Completely
annotated Borderguards poems in the Man’yōshū]. Tokyo: Kasama
shoin.
Mochizuki Ikuko, ed. 1974. Ruiju myōgishō. Tokyo: Kasama shoin.
Murayama Shichirō 1981. Ryūkyūgo no himitsu [The mystery of the
Ryukyuan language]. Tokyo: Chikuma shoten.
Nakamatsu, Takeo 1987. Ryūkyū hōgen jiten [A dictionary of the
Ryukyuan dialects]. Naha: Naha shuppansha.
Nakamoto Masachie, Kajuki Shin’ichi, Hashio Naokazu, Nohara
Mitsuyosi, Nakama Mitsunari. 1995. Urazoe-Kowan hōgen jiten [A
246———Koreo-Japonica

dictionary of the Urazoe-Kowan dialect]. Urazoe: Kowan jishi henshū

iinkai.
Nakanishi Susumu. 1984. Man’yōshū zen chūshaku genbun tsuki
[Man’yōshū fully annotated with the original text]. Tokyo: Kōdansha.
Nakasone Seizen. 1983. Okinawa Nakijin hōgen jiten [A dictionary of the
Nakijin dialect in Okinawa]. Tokyo: Kadokawa.
Nam Kwangwu. 1997. Kyohak koe sacen [Kyohaksa’s dictionary of pre-
modern Korean]. Seoul: Kyohaksa.
Nevskii, Nikolai A. 1978. Fol’klor ostrovov Miiako [Folklore of the
Miyako islands]. Moscow: Nauka.
Nohara Mitsuyoshi. 1986. Ryūkyū hōgen joshi no kenkyū [A study of
particles in the Ryukyuan dialects]. Tokyo: Musasino shoin.
———. 1998. Ryūkyū hōgen no joshi no kenkyū [A study of particles in
the Ryukyuan dialects]. Tokyo: Okinawa gaku kenkyūjo.
Ogura Shinpei. 1934. ‘Chōsengo to Nihongo [Korean and Japanese]’.
Kokugo kagaku kōza. Vol. 4, Kokugogaku. Reprinted in: Ogura Shinpei
hakase chōsaku shū 4: 315-377).
———. 1935. Chōsen gogaku shi [History of Korean Linguistics]. Tokyo:
Toe shoin.
Omodaka, Hisataka et al. (ed.) 1967. Jidai betsu kokugo dai jiten. Jōdai
hen [A big dictionary of Japanese by periods. Ancient period]. Tokyo:
Sanseidō.
Omodaka Hisataka. 1974. Man’yōshū chūshaku [Annotated Man’yōshū].
1-22. Tokyo: Chūōkōronsha.
Onions, C. T., ed. 1978. The Oxford Dictionary of English Etymology.
Oxford: Clarendon Press.
Ōno Susumu. 1957. Nihongo no kigen (kyūhan) [The origins of Japanese
(old edition)]. Tokyo: Iwanami.
Oshio Mutsuko. 1999. Okinawa Iejima hōgen jiten [A dictionary of the
Iejima dialect in Okinawa]. Iejima: Iejima mura kyōiku iinkai.
Pak Yongho. 1988. Ceycwu pangen yenkwu (caryo phyen) [A Study of the
Ceycwu Dialect (Data Section)]. Seoul: Korye tayhakkkyo minsok
munhwa yenkwuso chwulphanpu.
Panfilov, Vladimir Z. 1962. Grammatika nivxskogo iazyka. Chast’ 1-ia [A
grammar of the Nivx language. Part 1]. Moscow-Leningrad:
Izdatel’stvo Akademii Nauk SSSR.
———. 1968. ‘Nivkhskii iazyk [The Nivx language]’. Iazyki narodov
SSSR [The languages of the peoples of the USSR], Vol. 5, Mongol’skie,
tunguso-man’chzhurskie, i paleoaziatskie iazyki [Mongolian, Manchu-
Tungus, and Paleosiberian Languages], 408-434. Leningrad: Nauka.
Polivanov, Evgenii D. 1968[1924]. ‘K rabote o muzykal’noi aktsentuacii v
iaponskom iazyke (v sviazi s malaiskimi) [Paper on pitch accent in the
Japanese language (in connection to Malay languages)]’. Biulleten’ 1-
go Sredne-Aziatskogo gosudarstvennogo universiteta, 4: 101-108.
Tashkent. Reprinted in Polivanov E.D. Stat’i po obshchemu
iazykoznaniiu, ed. A.A. Leont’ev, 146-155. Moscow: Glavnaia
redakciia vostochnoi literatury.
 References 247

Ramsey, S. Robert. 1977. ‘Velar lenition in Korean’. Festschrift

Commemorating the Seventieth Birthday of Doctor Lee Sung Nyong,
125-132. Seoul: Ulyu.
———. 1978. Accent and Morphology in Korean Dialects. Kwukehak
chongse 9. Seoul: Thap chwulphansa.
———. 1991. ‘Proto-Korean and the Origin of Korean Accent’. In Studies
in the historical phonology of Asian languages, ed. William G. Boltz
and Michael C. Shapiro, 215-238.Amsterdam and Philadelphia: John
Benjamins Publishing Company.
———. 1993. ‘Some Remarks on Reconstructing Earlier Korean’. Ehak
yenkwu 29.4: 433-442.
———. 2003. ‘Accent, Liquids, and the Search for a Common Origin for
Korean and Japanese’. Japanese Language and Literature 38.2: 339-
350.
Ramstedt, Gustaf J. 1952. Einführung in die Altaische Sprachwissenschaft.
Vol. 2, Formenlehre. Mémoires de la Société Finno-Ougrienne 104.2.
Helsinki: Suomalais-Ugrilainen Seura.
RGJ. 1976. = Okinawa go jiten [A dictionary of the Okinawan language].
Tokyo: Ōkurashō insatsu kyoku.
Satake Akihiro, Yamada Hideo, Kudō Rikio, Ōtani Masao, and Yamazaki
Yoshiyuki, eds. 1999-2003. Man’yōshū [Ten thousand leaves (of
words)], 1-4. Shin Nihon koten bungaku taikei [New Series on the
Classical Japanese Literature]. Vols. 1-4. Tokyo: Iwanami.
Serafim, Leon. 1985. Shodon: The Prehistory of a Northern Ryukyuan
Dialect of Japanese. Tokyo: Honpo Soseki Press.
———. 1999a. ‘Why Proto-Japonic Had at Least Six, not Four, Vowels’.
Presentation given at the University of Hawai‘i at Mānoa Linguistics
Department Tuesday Seminar.
———. 1999b. ‘Reflexes of Proto-Koreo-Japonic Mid Vowels in Japonic
and Korean’. Paper presented at the XIVth International Conference on
Historical Linguistics, Vancouver. Workshop on Korean-Japanese
Comparative Linguistics.
Serebrennikov, Boris A., and Ninel’ Z. Gadzhieva 1986. Sravnitel’no-
istoricheskaia grammatika tiurkskikh iazykov [A historical-comparative
grammar of the Turkic languages]. Moscow: Nauka.
Shimabukuro, Moriyo. 2002. A Reconstruction of the Accentual History of
the Japanese and Ryukyuan Languages. Ph.D. diss., University of
Hawai‘i at Mānoa.
Shimoji Kazuaki. 1979. Miyako guntō go jiten [A dictionary of the
language of the Miyako islands]. Naha: Privately printed.
Shimono Masaaki. 1983: ‘Toyama-ken no hōgen’. In Vol. 6: Kōza
hōgengaku. Chūbu chihō no hōgen, ed. Iitomi Kiichi, Hino Sukezumi,
Satō Kōichi, 305-335.Tokyo: Kokusho kankōkai.
Starostin, Sergei. 1991. Altaiskaia problema i proiskhozhdenie iaponskogo
iazyka [The Altaic problem and the origins of the Japanese language].
Moscow: Nauka.
248———Koreo-Japonica

Starostin, Sergei, Anna Dybo, and Oleg Mudrak. 2003. The Etymological
Dictionary of the Altaic Languages. Leiden: E.J. Brill.
Street, John. 1978. ‘Proto-Altaic *ia’. In Proceedings of the VIth
International Symposium, 51-80. Seoul, National Academy of Sciences.
September 5-9.
Takagi Ichinosuke, Gomi Tomohide, and Ōno Susumu, eds. (1957-62)
Man’yōshū [Ten thousand leaves [of words]. Vols. 1-4. Nihon koten
bungaku taikei [Series of Japanese classical literature], Vols. 4-7,
Tokyo: Iwanami.
Takahashi Shunzō. 1986. Yonaguni hōgen no goi shū [A vocabulary of the
Yonaguni dialect]. Ryūkyū no hōgen 11. Yaeyama, Yonaguni shima.
Tokyo: Hōsei daigaku Okinawa bunka kenkyūjo.
Takayanagi Mitsunaga and Takeuchi Rizō, eds. 1974. Nihon shi jiten [A
dictionary of Japanese history]. Tokyo: Kadokawa.
Thompson, Laurence C. 1987. A Vietnamese Reference Grammar. Mon-
Khmer Studies XIII-XIV. Honolulu: University of Hawai‘i Press.
Thorpe, Maner. 1983. Ryukyuan Language History. Ph.D. diss., University
of South California.
Tsuchihashi Yutaka, ed. 1957. Commentary and edition of the ‘Kojiki
kayō’, ‘Nihonshoki kayō’, ‘Shoku Nihongi kayō’, ‘Fudoki kayō’, and
‘Bussoku seki no uta’. In Kodai kayōshū, Nihon Koten Bungaku Taikei
[Series of Japanese classical literature]. Vol. 3. Tokyo: Iwanami.
Uchima Chokujin and Arakaki Kumiko. 2000. Okinawa hokubu nanbu
hōgen no kijutsuteki kenkyū [A descriptive study of Northern and
Southern Ryukyuan dialects]. Tōkyō: Kazama shobō.
Unger, J. Marshall. 1990a. ‘Japanese and What Other Altaic Languages?’
In Linguistic Change and Reconstruction Methodology, ed. Philip Baldi,
547-561. Berlin and New York: Mouton de Gruyter.
———. 1990b. ‘Summary Report of the Altaic Panel’. In Linguistic
Change and Reconstruction Methodology, ed. Philip Baldi, 479-482.
Berlin and New York: Mouton de Gruyter.
———. 1993[1977]. Studies in Early Japanese Morphophonemics.
Bloomington: Indiana University Linguistics Club.
Vovin, Alexander. 1990. ‘Iaponskie zaimstvovaniia v ainskom iazyke
[Japanese Loanwords in the Ainu Language]’. In Pis'mennye
Pamiatniki i Problemy Istorii Kul'tury Narodov Vostoka (Doklady i
Soobshcheniia XXIII Godichnoi Sessii LO IV AN SSSR), 58-66.
Moscow: Nauka.
———. 1993a. A Reconstruction of Proto-Ainu. Leiden: E.J. Brill.
———. 1993b. ‘Notes on Some Japanese-Korean Phonetic
Correspondences’. In Japanese/Korean Linguistics, ed. Soonja Choi, 3:
338-350. Stanford: Center for the Study of Language and Information,
Stanford University.
———. 1993c. ‘Japanese kisaragi “second lunar month”’. Eurasian
Yearbook 65: 116.
 References 249

———. 1993d. ‘About the Phonetic Value of the Middle Korean

Grapheme ∆’. Bulletin of the School of Oriental and African Studies,
56.2: 247-259.
———. 1994. ‘Genetic Affiliation of Japanese and Methodology of
Linguistic Comparison’, Journal de la Société Finno-Ougrienne 85:
241-256.
———. 1995. ‘Once again on the accusative marker in Old Korean’.
Diachronica 12.2: 223-236.
———. 1997a. ‘On the syntactic typology of Old Japanese language’,
Journal of East Asian Linguistics, 6:273-290.
———. 1997b. ‘Voiceless velars in Manchu’. Journal de la Société Finno-
Ougrienne, 87: 263-280.
———. 2000. ‘Pre-Hankul Materials, Koreo-Japonic, and Altaic’. Korean
Studies 24: 142-155.
———. 2001. ‘Japanese, Korean, and Tungusic: Evidence for Genetic
Relationship from Verbal Morphology’. In Altaic Affinities,
Proceedings of the 40th meeting of the Permanent International Altaistic
Conference (PIAC), Provo, Utah, 1997, ed. David B. Honey and David
C. Wright, 183-202. Indiana University Uralic and Altaic Series, Vol.
168. Bloomington: Indiana University Research Institute for Inner
Asian Studies.
———. 2003a. A Reference Grammar of Classical Japanese Prose.
London: RoutledgeCurzon.
———. 2003b. ‘Once Again on Lenition in Middle Korean’. Korean
Studies 27: 85-107.
———. 2005a. A Descriptive and Comparative Grammar of Western Old
Japanese. Vol. 1: Phonology, Script, Lexicon, and Nominals.
Folkestone, England: Global Oriental.
———. 2005b. ‘The End of the Altaic Controversy’. Central Asiatic
Journal 49.1: 71-132.
———. 2007. ‘Once again on doublets in Western Old Japanese’. Current
Issues in the History and Structure of Japanese, ed. Bjarke Frellesvig,
Masayoshi Shibatani, and Charles Smith. Tokyo: Kuroshio Publishers,
pp. 351-373.
———. 2008. ‘Proto-Japanese beyond the Accent System’. In Proto-
Japanese. Issues and Prospects, ed. Bjarke Frellesvig and John
Whitman. Amsterdam and Philadelphia: John Benjamins, pp. 141-156.
———. 2009. A Descriptive and Comparative Grammar of Western Old
Japanese. Part 2: Adjectives, Verbs, Adverbs, Conjunctions, Particles,
and Postpositions. Folkestone, England: Global Oriental.
———. 2010. (forthcoming). Reconstructing Proto-Tungusic canonical
shapes: sequences of nasals and non-nasals.
Whitman, John B. 1985. The Phonological Basis for the Comparison of
Japanese and Korean. Ph.D. diss., Harvard University.
———. 1990. ‘A Rule of Medial *-r- Loss in pre-Old Japanese. In
Linguistic Change and Reconstruction Methodology, ed. Philip Baldi,
511-545. Berlin and New York: Mouton de Gruyter.
250———Koreo-Japonica

———. 2001. ‘6 (moshikuwa 7) boinsetsu kara mita Proto-Japanese-

Ryūkyūan to pre-Middle-Korean no hikaku’. Paper presented at the
meeting of the research committee on Origins of the Japanese Language,
July 20. Kyoto: International Center for Japanese Studies.
———. 2003. ‘The Source of the Bigrade Conjugation and Stem Shape in
Pre-Old Japanese’. Paper presented at the XVIth International
Conference on Historical Linguistics, Copenhagen, August 11-15.
Yamada Yoshio. 1954 [1912]. Nara chō bunpō shi [History of Nara period
grammar]. Tokyo: Hōbunkan.
Yi Kimun. 1957. ‘Kyeylim yusa-uy il-kochal’. Ilsek Yi Huysung sensayng
songswu kinyem nonchong, ed. Yi Pyengki, 393-407. Seoul: Ilcokak.
———. 1958. ‘A Comparative Study of Manchu and Korean’. Ural-
Altaische Jahrbuch 30: 104-193.
———. 1959. ‘On the Breaking of *i in Korean’. Aseya yenkwu 2.2: 131-
137.
———. 1964. Kwuke sa kaysel [An outline of the Korean language
history]. Seoul: Thap chwulphansa.
———. 1987. Kwuke umun sa yenkwu [Historical phonology of Korean].
Kwukehak chongse 3 [Series on National Linguistics 3]. Seoul: Thap
chwulphansa. 5th rev. edition. First published in 1972.
———. 1991. Kwuke ehwi sa yenkwu [A study of the history of Korean
vocabulary]. Seoul: Tonga chwulphansa.
Yi Swungnyeng. 1961 [1997]. Cwungsey kwuke munpep [A grammar of
the Middle Korean language]. Seoul: Uylyu munhwasa.
———. 1978 [1985]. Ceycwuto pangen-uy hyengthayloncek yenkwu [An
investigation of morphology of the Ceycwu dialect]. Kwukehak yenkwu
chengse 5. Seoul: Thap chwulphansa.
Yu Changton. 1964. Yico kwuke sa yenkwu [A study of the Korean
language history during the Yi dynasty]. Seoul: Senmyeng munhwasa.
Zeitoun, Elizabeth. 2000. Bunong yu cankao yufa [A reference grammar of
the Bunun language]. Taibei: Yuanliu.
INDEX
Allium porrum 195
A :alóy See :aláy
àlph 226
a ‘foot, leg’ 226
also 123, 202
a ‘I’ 62, 63
although 204
*-a 83
-am- 87, 90
-a/e 26 ama-/amë 190
a-N-kakî 227
ámh 196
a-N-pumî 227
àmkól- 23
a-na-suwe 227
Ammophilia infesta 99
a-yupî 227
amo/omo 93, 234
abscess 129
amusement party 210
ac- 17, 21
-an- 77, 90, 227
acakala 21
aN- 66, 67, 223, 224
accomplish 166
aN-tö 223
ách 223
*anaka 31
achieve 166
-anap- 53, 90
àchóm 93, 224
anc- 17, 19, 21, 29, 81
àchyét- 224
anc-hi- 29
àchyétpù- 224
àncík 21
àcík 19, 21
àncò- 19, 20
àcò- 20
àncók 21
act of doing 167
and 202
add 144
angrily 38
adult 234
ánh 205
after 214
ani 90, 227
again 123
aNkar- 232
ago 234
aNkë- 232
àhóy 224
aNkî ‘baby’ 224, 225
ahūn 31
aNkî ‘upper jaw, gills’ 234
*aka 31
aNpura 37
akari 24
-aNs- See -an-
ákí 222, 224
àns- 20
-aku 88
aNse 66
akū 31
ʔanu 70
àlá 226
ap- 229
àlà-wùh 226
àpá-:nim 235
àláy 226
àpí 13, 235
:aláy 226
appearance 234
all 113, 194, 199
 Index 253

bracelet 155 chiru 41

braid 218 chiruci 41
branch 129 churi 41
breast 14, 34, 164 chwúm 167
breath 34 *-ci, *-ci 53
bridegroom 34, 35 çiburu 93
bridge 103, 120 cici 93
bridle 96 cïhuï 34
bright 101 ciiba 216
brings toward 212 cïkururi 34
brow 34 -cil 167
brush 183 ciley 170
buckwheat noodles 39 cïno 34
buy 133 cïnu 34
cìp 93, 170
C circles 123
ciri-bira 196
càchí- 19, 21
ciribiraa 196
càlí 161
cis 20
cànchí- 19, 21
clan, family chief 235
cang-ma 190
clean 236
cangue 130
clear 101
capture 122
clear/wash/bleach 238
carries it on the back 235
clears it, washes it off 182
carry out 166
climb 232
cart 157
close 116, 121
cás 161
closes (eyes) 138
castle 161
clothing 232
cave 141, 155, 230
cloud 153
cek 21
coat 133
cèk 163
cockscomb 106
ceremony for expelling demons 206
coils it 132
certainly 152
cólà 161
cha-wan 37
cólà- 238
chalyey 14
còlG- See còlò
charcoal 185
còlò 102, 162
chastises 156
comes out 205
cheek 104
cómk- 162
chestnut 136
còmò- ~ còm- See cómk-
chici 41
complete 188
chii 41
conceal 136
child 143, 222, 224, 226
concubine 143
chima 41
confine 194
chimney 152
254———Koreo-Japonica

congeal 145, 146 dama ‘obstacle’ 40

continue 123, 162 dangasa 42
cook 236 dasai 41
cookhouse, kitchen 152 daughter-in-law 196
cooking pot 132 day 33, 164
copper 153 daytime 107
cormorant 231 declare 217
correct 190 decoration 232
cow 179 deep 161
còzó 13 deep valley with soaring cliffs on both
crab 150, 238 sides 142
crane 123, 238 delay 201
crowd 200 delicious 192
cruel 156 demonic 154
cryptomeria 179 denwa 39
cùl- 165 depart 213, 214
cultivate 131 desire 99
cùlù- ~ cùlG- 166 deters through punishment 156
cùmkéy 23 die 166
cut 105 die prematurely 166
cut off 116, 166 differ 115
cwòch- 162 dig out the roots 152
cwok- 168 direction 140, 151
cwùk- 166 dirt 110
cwúl 166 dirty ground 93
cyècáy 165 disappear 136
cyècéy See cyècáy dish, vessel 180
:cyek- ~ :cyak- 168 disposable chopsticks 39
cyèmkú- 23 distant 195
cyèmkúl- 23 divide 111
cyèmkùl- 169 dividing line, divide 159
cyèmúl- 169 division 158
does 186
D dog 30, 33, 35, 236
*dolbo 40
*daba- 40
door 147
daci 41
doro 165
daguba 41
*doro 165
daguharai 41
dosi-ka 31
dagutuci 41
*dosin-ka 31
dakhunin 41
double, increase twofold 144
dakkai 41
doubt 37, 109
dakkhan 42
drag 152
dama ‘mountain’ 40
 Index 255

dragon 201 *eki 34

drain 110 El-a 83
draw 152 elbow 101
drives it out 100 elder 234
dry and bleach by exposing to the sun :elGwun 234
183 èlí- 233
dry field 98 èlkwúl 234
dry land 200 elongates 211
du 40 èlú- 234
dubagu 41 émà 234
dudan 41 embrace, entwine 139
dudi 40 embraces as a wife 194
dudu ‘business’ 42 emcügür 22
dudu ‘recuperation’ 41 émí 234
dugout cave 230 *emo 35
dugu 41 empty 101, 110
dugunin 42 :encèy 66
dukhunin ‘six people’ 41 enchi 19, 21
dukhunin ‘six years’ 41 enchyengi 22
dunNaci 41 ènték 23
dunsa 41 enter 128
duri 42 entrance way 146
duru 40 *entu- 33, 67
dusukhu 41 *entu-ma 66
dutthu 40 *entu-re 66
duya 41 *èntúy 66
ènú 25, 66, 67, 233
E èp- 235
èpèsí 235
-e/-ye/-a 83
èpí 235
-ë-/-ï- < *V-[C]i- 81
èpímòt 235
*ʔe 65
*er- 128
*e-ka 66, 67
*esi 34
*e-ku 66
e:sté 67
ear of rice 106
esthyengi 19, 22
earth 93, 102, 123, 164, 187
e:styé 66, 67
eat 195
e:styéy 66, 67
eat/drink, to gulp down something 179
*esu 34
èchí 19, 21
*esuy 127
edge 94, 97, 140, 161
*etu 66
èhí- 22
ètúW- 14
èhù- 22
ètúy 14, 25, 66, 67
eight 219
excellent 192
eighty 219
256———Koreo-Japonica

exchanges it 132 *fji-vji 16, 109

excrement 141 flavor 192
exhausted 166 flavorful 192
exorcism ceremony 154 flax 112
expended 166 flea 107
expose (to elements) 183 flesh 177
exterior 144 float 38
extinguish 136, 137 fly (n.) 103
extremity 161 foam 39
eye 93, 191 fodder 184
eyelid 133 follow 162
*ezu 33 food goes sour 178
foolish 233
F foot 96, 226, 227, 228
footprint 227, 228
fact 94
forges it 127
fafa 93
form 129
fang 216
fort 161
far 195
four 211
fart 108
foxtail 135
fast 102
freeze 146
fat 109
fresh food used as a side dish 207
father 93, 235
fathom 100
G
fati 98
fayame- 78 -Ga- See -ke-
feet end 227 gallop 227
fellow (pejor.) 209 game 210
female 196 game animal 177
fern 39 garden 94, 192
few 168 garlic 33
fi- 108 gate 146, 147
field 105, 211, 238 -Ge- See -ke-
fierce 175 -Gé/á 15
fiiï 34 general questions 128
fill 116 general term for long grasses, thatch
fill [with water] 238 135
fill up with (water) 116 generally speaking 199
film 133 generation 216
filth 110 gets dark 169
fire 107, 238 gets older 234
firewood 39, 183 gets thin, gets emaciated 221
fish 207 -Gi- 29, 81
fishing 207 gïï-baa 216
 Index 257

gills 234 hate 224

-γïr 54 HAton 151, 187
give 34, 201 hayah- 4
gives birth to (MK nàh-) 205 head 93, 191
gives birth to (OJ um-) 232 heart 194
glitter (v.) 33 hearth 132
go 212, 214, 216 heavy 197
go and come back 212 heel 153
go around [a certain object] 198 hèlí 188
go away 213 hemp 112, 173
go back 198 heron 176
go up 232 -hi- 29, 81
goes down 208 hide 131, 136
goes to meet, receives it 190 hides 184
good 34, 50, 124, 125, 176, 192 high 114
good fortune 172 hikyayuN 33
government post 155 hiru 33
grass 126 hïru 33
grass for pasture 184 hit 37
great 105, 235 ho- 186
greens, herbs 209 ho- ~ hoy- 27
ground 164 hold 122
group 200 hold/take 238
guest 195 hòlk 102, 164, 187
-Gwo See -kwo holki 164
-Gwo- 29 hollow 155
-Gwó 15 hònàh 151, 187
-Gwu- 29 hoof 153
horizontal 213
H horn 34, 102
hoto- 11
hair 131
*hoto- 114
hairdress 232
house 93, 170
hand 115, 180
howl 108
handle 102, 170
hòWózá 19, 22
handspike 170
hoy- See ho-
hang 119
huku- 11
hang down 122
huli- 29
hangs it up 145
huli-Gwu- 29
hànólh 186
hùn-hò- 188
hard 103, 143, 156
hurry 97
hardened salt 156
husk 133, 148
has sexual relations 234
huy- 4
hat 134
258———Koreo-Japonica

húy-/hóy- 189 interrogative particle 129

-huy 59 interrogative suffix 128
-hwo- 29 interrupt 116
hwòGwàk 187 intimate 38
hwòwák 187 iNtu 33, 66, 67, 233
hyé 188 iNtuku 66
hyet 188 iNture 66
iNtusi 66, 164
I iNtuti 66, 164
inu 33, 35, 236
I 203, 229
inV- 214
í 68
ip- 236
i- 28
ipa 170
-i, -î 45, 48, 49, 53, 62, 81, 87, 90
iparo 170
í-lá 27
ipe 93
i-ta 27
ipê 93, 170
ic- 18
ìph- 236
ice 109
ipî 170
icey 18, 19, 21, 22, 24
ir- 128, 236
ícèy 19, 21
ir- <*i-tV- 90
ïdï, uduru 33
is attached 162
ika 66, 67
is born 205
ikî 34
is joined 162
iku 66, 113
is locked 162
ikuNta 66, 113
isi 34, 35, 126
ikura 66
island 183
:il- 236
isô 34, 126
ilgïnda 25
ìsólàs 14
ílhì 236
isu 34, 126
ílhòy 236
ìsúl 13, 14, 50
ilwo- 76
itchy 135
ima 237
iti 164
imo 35
itö 186
in- 213, 216
itu 66
iN 28, 33
iwo 55
in general 199
ʔiyu 55
incey 18, 19, 21, 22, 24
ïzu 55
inform 217
inisi-kata 213
J
inisipê 213
insert 95, 210 jar 121
inserts it, thrusts it into 179 jaw 234
inside 37, 205 jincüg 31
inside corner 97 joins together 229
 Index 259

joint 38 kap- 133

just 113, 116 kapa 133
kape 131
K kapë- 132
kàph- 132
ka 129
kàphól 133
*ka- 68
kapi 134
-ká 128
kapî 148
-ka- See -ke-
kàpóyép- 134
ka- 27, 28
káps 134
ka-/kë 131
kara- 131
ka-kô 129
karamë- 132
kaban 39
karasi 135
kàch 129
karasukî 131
kácí 129
kari 59, 70
kàcí 129
karimo 132
kai ‘oar’ 129
karu- 134
kai ‘that over there’ 70
kasa ‘hat’ 134
-kai 60
kasa ‘scab’ 129
kak- 137
kase 130
kak-/kakë- 145
kasi 130, 150
kakayak- 140
kasi-pesi 51
kal 130, 131
kaso 93
kál 130
kàsóm 14
:kal- 131
kát 134
kàlàp 130, 131
kata ‘shape’ 129, 140, 151
kalh 134
kata ‘shore’ 140
kálh See kál
kata- ‘one’ 151
:kalkí 131
katar- 5
:kalm- 131
kati 150
kálmwó 132
kawan 39
kám- 132
kaya 131, 135
ka:m- 132
kayu- 135
kàmá 132
-ke 72
kamaNtô 132
-ke- 84, 90
kamï 24
-ké/á 15
kanatô See kaNtô
kë- 136
kanchi 18, 20
*kə- 68
kani 150, 238
kèch 144
kaNka-/kaNkë 140
:kel- 145
kaNpî 148
kèlí 146
kaNti 129
kèmúy 147
kaNtô 146
keŋi ‘crab’ 150
kanu 70
260———Koreo-Japonica

kèntí- 23 kiwa 20
kèpcìl 133 *kiyasi- 137
kèphí 148 kiye- See kë-
kèphùl 148 kìzúm 13
kèpwúp 148 kkachi 18, 20
kès 149 kkanchi 20
KESk-a 83 kkaphwul 133
ket- See kët- kkaptayki 133
:ket 150 kkepcil 133
kët- 136 kkepteyki See kkaptayki
:key 150, 238 kkol-da 25
-key 26 kkul- 152
*kezu 40 knife 134
khal See kál kö 68, 69, 144, 151, 176, 186
khal-cip 134 kô 137, 143, 152
khú- 11 kö- 180
ki 216 kö-nö pa 180
kî 159 kô-samë 191
-kî, adjectival attributive 72 kobu 147
-kî, past final 84, 90, 158 KOcal (秋察) 140
-ki- 29, 81 kòchó- 18, 19
-kí 158 kochwo- 20
kiba 216 kofi See kofu
kíc 158, 159 kofu 141
kidi 40 kól 135
:kil- 159 kòláp- 135
kìlúy 159 kólàs 135
kimchi 20 kóláy 136
kind 129, 141 kòlí- 136
ki[N]ta 159 kòlk-/kúlk- 137
kîpa 161 kòlò 15, 137
kîpam- 161 kòlòWí 12, 15
kìph- 161 :kolW- 138
kis 20 kóm- 138
kisa 150 kòmchó- 19
kisaragi 151, 152 komchwo- 20
kîsi 155 komchwu- 20
kiso 186 kòmcòk- 23
kîsö/kîNsö 186 kömë- 132
kìsúlk 13, 14 kömï- 131
kîtamë- 156 kona 137
kîtasi 156 kònchó- 19
kitu- 156 koNkor- 145
 Index 261

koNkoye- 145 *kure 34

kònòlh/kùnùl 140 kuriya 152
köNsi- 152 kurïyuN 34
köNsö 186 kuru 157
*koobo 148 kuru-/kuri 136
*koonpo 148 kuruma 157
köpor- 146 kurumi 136
kör- 145 kururu 157
:kos 140 kuru-si- 143
kösö 152 kús 152
koso 152 kusa 141
kötö 5, 50, 149 kusa- 155
kôye- 145 kusi 144, 154, 155
kòzGáy 13 kusi mî-tama 155
kòzòlàkí 140 kusirö 93, 155
kòzòlh 140 kusô 140
kòzòm 141 kusu-nuk- 144
-ku 88, 90, 140 *kusuri 41
kù 68, 144 kususi- 155
ku-/ki 153 kutu-/kuti 155
kube- 153 *kutu- 24
kubona[me] 148 kutu-wa 51
kukî 141 *kutuy 51
kuku-/kukï 142 kùtùy 14
kukupî 142 kuubaa 147
kulan/kelan 14 *kuwa 143
kùlùh 151 kuwe- 153
kùlyèkí 59 -kuy 59, 73
kum- 138, 139 kùz- 152
kum-i-ne- 139 kwaci 41
kuma 143 *kwasi 41
kumô, kumo ‘spider’ 147 Kwiiba 216
kumô < *kumo ‘cloud’ 153 -kwo 26, 88, 90, 140
kùntày 23 -kwó 15
kuNtira 142 kwóc 144
kupapë- 144 kwocci-/kwochi- 20
kupasi 144 kwóh 141
kupî See kukupî kwòhày, kwohay 141
kupîpîsu 154 kwòkwòlí 142
kupîsu 154 kwòlày 142
*kura-i- 34 kwó:lwop- 143, 238
kuraNpë- 138 :kwom 143
kuratani 142 :kwomá 143
262———Koreo-Japonica

kwonchi- 18, 20 leech 107

kwòp- 144 leek 195
:kwop- 144 leg 96, 103, 115, 226, 227
kwòsóy 14 length 159
kwùGùl- See kwùGwùl- lengthen it 211
kwùGwùl- 157 let go 210
:kwul 152 lets soak, dyes 181
kwùlèk 152 levee 123
kwùlí 153 lever 170
kwulley 24 lies down 208
*kwultu- 25 light 107
kwùlú- 153 like, be fond of 124
kwùlwúm 153 line 166
kwùngkùl- 23 linen 112
kwúp 153 little 168
:kwup- 153 lizard-tail 195
kwupul-, *kwùpwùl- 157 loads it 185
kwús 154 location 161
kwùsù 155 loft 114
kwùsúl ‘official post’ 155 -lol ~ -lul 54
kwùsúl ‘treasure, bead’ 13, 14, 93 long 159, 208
kwút 155 long ago 213
:kwut, kwùs 51 long rain 190
kwùt- 25, 156 loosen 219
kyél 150 lord 103
kyèncí- 23 lover 143
kyèncwú- 23
kyèntúy- 23 M
kyèth 151
-m 90
kyèzúlh 151
*-m 78
ma- 190
L
ma-/më 191
lacquer 233 -ma 87, 90
ladder 120 màc- 190
lagoon 140 màchí 18, 20
large number 218 mackerel 39
last night 186 máh 193
last year 186 mah 190
late 206 majestic 105
laugh 39 mak- 194
layer 101 make 23, 34, 205
lead 206 make a short-cut 166
leather 129 make fire 113, 238
 Index 263

makes it (into) 174 mèngGéy 23

makes it disappear 136 *mentu 33
:mal- 191 metal 180
málh 191 metal bell 180
màlí/mèlí 93, 191 *mezu 33
mallet 121 mï 201
mana 191 mï- 198
mane 131 -mî 84, 90, 91
mane- 93 mica 232
manh- 93 midwife 39
màntòlàmí 23 :mil- 201
maNtu 23 :milmúl 201
many 93, 218 mìlù 201
mara 191 mïna 35, 194
mara-/mare- 195 mina 35
marapîtö 195 mîNtu 33, 75, 146, 197
mark 34, 35 mîra 196
market 164 mìs 202
marö 62 miscanthus flower 124
marries 234 miscanthus reed 124
más 192 mît- 201
master 210 *mita ‘earth’ 93, 124
mat- 190 mita ‘with’ 202
-mata 93 míth 93, 202
match 190 mîti 200
matches (with) 138 mituti 201
material 141 mizï 33
màth 192 mïzï 33
mati (麻帝) 93, 191, 192 mizya 93
mat∫i 192 mo ~ më 193
matter 94 moho 34
*mayo 34 *moko 34, 35
mayô See *mayo mól 193
*mayU See *mayo molan 14
mê 196 moldboard of plow 106
*me, më, me 93 mole 123
meat 177 mòlk- 5, 101
mec- 17 mòlò 102, 119
mèchwú- 19, 20 mòncí- 23
meet 229 month 119
mèk- 194 moon 119
:mel- 195 mor- 198
melon 38, 233, 238 more 121
264———Koreo-Japonica

morning 93, 224 mùsúm 66, 200

morö 35, 199 musuNp- 50
mortar 38, 187 mùsús 66
mosi 199 muta 202
mother 93, 234 muti, buti 38
mötö, moto 93, 202 mutto 38
mountain 114, 143, 194 mutu 38, 202
mouth 155 mutuNp- 50
move 23, 29, 37, 103, 232 muutu 93
mòy- 194 mwòlwó 143, 194
moye-/moyï- 198 mwóm 194
mòyn- 194 mwomcwo 22
mòyngkól- 23 mwomzwo 22
mòzòlh 13 mwòncyè 23
mòzòm 13, 194 mwòngkóy- 23
mu-/mï 194 mwòtón 14, 15
mud 110, 165 :mwoy 143
mud-dauber See Ammophilia infesta :mwoyh 194
muddy 165 mwúl 200
muho 34 mwùngkúy- 23
múkèp- 197 mwùs[-] 50
mukô 34, 35 mwùth 200
múl 5, 197 -mye 90
mùl- 198 :myel 195
mùlk- 5 myènól 196
mùlù-/mùlùl- 198 myènólí 196
mulus 199 mysterious 154, 155
muna 101
muna-/mune ‘chest’ 34, 194 N
muna-/mune ‘ridge’ 101
n- 90
*munay 34
-n-, negative 77, See -an-
*mune 34
-n-, perfective 86, 90
munï 34
*n-anu- 67
mura-/mure 200
na ‘I’ 62, 63, 65, 203
mure 143, 194
na ‘inside’ 206
mus- 200
na ‘name’ 208
mus 50
na ‘side-dish’ 207, 211
mushy 165
na ‘thou’ 62, 63, 66, 203
musk- 50
naa 65
mustard plant, rape 135
naa, naN 65
mùsú 66, 67, 199
ná 203
musub- 50
ná- 205
mùsúk 66, 200
nà 62
 Index 265

naa 209 nàs 51

-na 204 nàsk- ~ nàksk- 207
-ná 204 nat- ‘to appear’ 29
-naa 60 nata 51
nác 164 nat-hwo- 29
nacay 164 *nau 66, 67, 223
nada 93 :nayh 58, 103, 207
na-dare 93 nàzí 13
nagi 206 n:da: 196
nàh- 205 ne 93, 202, 209
naka 205 ne- 208
-nakai 60, 61 né 203
naked 194 nè 54, 62, 203, 211
náks 207 :ne-, :nek- See :neyh
nàlhwó- 206 nèchwúl, nechwul 18, 20
NAli 58, 103, 207 needle 96
nam 209 negative imperative 191
naṃ 65 nèhùy 62
namari 206 nelp- 25, 211, 237
namcin 22 nəlbəsə 25
nàmcòkì 23 *nelpu- 25
name 208 nenchwul 18, 20
na-mîNta 93 nèncùsì 23
*namu 65 nengwul 18, 20
namudi 65 nèp- 25, 211, 237
*namur 206 :neyh 211
nàmwò ~ nàmk- 206 new year 186
*nan (儺) 206 -ngey 59
nànhwó- 23 ni 93
nani 66, 67, 223 ni- 214
naNka- 208 -ni 62
naNkï- 206 ní 216
naNkoya 206 ní- 216
nànkyés 23 nicuhe 31
naNsö 66 night 34
naNtö 66 nìk- 25
naNtuk- 207 *nilko/u- 25
náp 206 nìlù- ~ nìlò- ~ nìl- ~ nìlG- 217
nàpóy 12, 14 nìlú- ~ nìlúl- 217
nàpwóy 12 :nim 58, 103
nar- 205 nincuhe 31
narrow path between two peaks or cliffs niNsi 34
141 nipa 94
266———Koreo-Japonica

nipa-tu tori 49 nwòh- 210

nìpúl 12 nwòlós 210
nïrəsə 25 nwóm 209
nita 93 nwón 211, 238
-Nka 59, 62 nwú 66, 67
-Nkai 60, 61 NWUli 58, 103, 216
-Nkari 59, 60, 61 nwun 93
Nni 34 nwun-kkaphwul 133
nô 211, 238 :nwuy 58, 103, 216
-no- 26 nyé- 212
-nö 50, 54, 62, 94, 206, 217 nyèh- 212
nòlí- 208 nyèh- ~ nèh- 210
noli- 29 nyèkh 213
noli-Gwo- 29 nyelim 58, 103
nòlk- 5, 208 :nyey 213, 214, 216
nóm 208 Nzi 33
nòmòlh 209
-nón, -nún, -ón, -ún, -n 217 O
nònhwó- 23
*ʔo- 68
nöNpë- 211
-o 26, 34
nöNpï- 211
-o/uli 26
noNsi 34
-o/ulq 26, 82
*noonsi See noNsi
-o/um 78
*noozi See noNsi
-o/umye 26, 84
nör- 217
-o/un 26, 86
nor- 214
-o/uni 26
nose 141
-o/usi- 26, 88
nothing but 194
oak 131
nöti 214
oar 129
now 237
oarsman 129
-Ns- See -an-
och 233
-Ntamï 116
offasïN 235
*ntoro 165
offering to gods 143
-Ntu 75
offerings 210
nùc- 25, 214
oi- 234
nuk- 210
oki 34
*nulcu- 25
oku 230
núlì- 214
-ol ~ -ul 54
nùlk- 5, 208
old 208
nusa 210
old woman 196
nusi 210
olden times 213
nuusi 210
*omi 93
nwòh 210
omina 197
 Index 267

omo 93, 234 pàntwóy 23

omo- 197 papa 93
omuna 197 par- 100
one 110, 187 para ‘belly’ 103
only 113, 116 para ‘field’ 105
*ontaro 35 param- 103, 104
op- 235 parap- 100
opïsi 35 parent(s) 234, 235
opö, opo 93 pari 96
opo- 235 part of the shore line covered by high
ordure 140 tide 140
oröka 233 paru ‘field’ 105
-ös- See -as- paru ‘needle’ 96
osi 34, 127 paru ‘spring’ 104
osïgam 24 paru-samë 191
osu 34, 127 pàs 97
other 208 pase- 98
outer appearance 144 pasi 97
*oware- 38 pasir- 97
-oy/uy 54, 58 pàsk See pàs
oya 234 past 213
paste 100
P pasuy, *pasuy 34
pata 98
pá 94
patah 93
-pa 94
pàtáh 13, 111
pàch- 97
patakë 11, 98
paddy field 211
páth 11
pagi 96
pàth 98
pàhí- 105
pati 98
painful 143, 238
pay 172
pak- 95
pay [back] 132
pàk- 95
paya 78
pàkwòní 96
paya- 102
pákwùlGéy, pawkwulley 96
paying respect 143
pál 96
payment to the porter 172
pàlól 13, 111
pê ‘layer’ 101
pani 101
pê ‘side’ 109
paNk- 105
pë 104
paNkî 96
peel 105
pànól 96
peep into 38
pàntók- 23
pèhí- See pàhí-
*pantoli 99
pel 105
pàntwó 23
268———Koreo-Japonica

:pel 99 pîNti 101, 110

pëmî 104 piro 100
pèmkúl- 23 pîru 33, 107
pèmpúy- 23 pîsamë 109
pèngGúl- 23 pïsuma 33
pèngkúl- 23 pit 155
penis 191 pitapi 106
person 209 pîtö 110
*peru 33 pivot door 157
:pes, :pet 51 piwi 163
pèsús 13, 14 pîyu 109
pètúl 14, 15 place 94, 117, 147, 161
pey 112 plain 105
phàchyò 22 plain, unadorned 187
phalah- 4 play 210
phallus 191 pleasantly odorous 155
phélí 105 plentiful 188
phól 180 plot of land for agricultural purposes
phólì 103 192
phwulu- 4 plow 131
pí 15, 109, 163 pluck berries 198
pî ‘ice’ 108 po 106
pî ‘shuttle’ 109 po- 107
pî ‘sun’ 107 pól 100, 101
pî- See pï- pólá- 99
pï- 108 pòlh 101
pï ~ *pö- 35, 107, 238 pòlí- 100
*pi 35, 107 pólì- 111
pích 107 pòlk- 5, 101, 102
pick 198 POLK-kuy 73
pick up 122 pòlò- 100, 103
piece of split bamboo 174 pòlp- 103
pierces it, sticks in (a cane) 125 pone 34
*pigi- 16, 109 popo 104
pigweed 109 por- 99
piï 33 posi ‘desire’ 99
pïkaï 33 posi ‘star’ 107
pîkar- 33 pòsó- ~ pòzó- 50
pìlé 109 post 191
pìléy See pìlé potato 35
pìlóm 109 pötökë 11
pìlús 110 pottery vessel manufactured chiefly by
pine 179 Korean potters 180
 Index 269

pounds it (with mortar) 125 pwùthyè 11

powder 137 pwùtúl 14, 15
póy 103, 104 :pwuy- 110
pòy- 103, 104 pyé 106
póyàm 104 pyéch 106
poye- 108 :pyel 107
precious stone 155 pyèlwók 107
pregnant 103, 104 pyèsúl 13, 14
prepare, make offering to the gods 189 pyèth 107
pretty 144
price 134 Q
prosperity 172
quail 37
prow of a boat 104
quiet 206
psí- 200
quma 143
pskí- 179
Qv 37
ptóy 110
Qvi(i) 37
ptwúy 124
pu 35
R
púl 5, 107, 238
pùlk- 5, 101, 102 -r 54
pull 152 -ra 56, 58, 62, 79
pùll- See pùlù- rain 163, 190
pulls it out 210 rainbow 34
pùlù- 108 rainy season 190
pum- 103 raise 122, 232
puna-/pune 104 -rang 56
push 201 rare 122, 195
pusi 38 reach 217
put inside 210, 212 rear, rear end 124
put off 201 reason 223
put on the surface 210 recites or sings it 236
put or hang something down in order to red 101
mark off the boundary 216 reed 135
putô 38, 105, 109 remote 195
pwól 104 resemble 214
pwóm 104 resides 184
pwòntwòkí 23 restrict 194
pwòtólàp-/pwùtúlèp- 105 retreat 198
pwóy 112 reward 132
pwùk 108 rice plant 106
pwul 134 rich, fertile, fat 145
:pwul- 108 rich, thick 145
:pwut-/pwul- 109 ridge 101, 119
270———Koreo-Japonica

ripens and gets soft 198 sand 102, 187

rise 232 sangki 20
river 103, 207 saNkî 176
road 120, 146, 200 *sanpa 31
road fork 200 *sapa 31
roasts it, toasts it 153 sar- 183
rock 34, 109, 126 saraasuN 183
roll 157 saras- 182, 183, 238
root 93, 202, 209 sarasïN 183
rope 166, 210 sarasu, sarirü 183
rose 39 sarasyuN, sariyuN 183
rots, ferments 181 sare- 183
rough 175 sasa 174
rower 129 sàsól 174
rudder 129 sàsóm 14
run 97, 98 saucer 121
runs dry 110 sawa 31, 39
rush 114, 135 sawak- 175
sàWí 13
S :say 176
saykwup- 178
-s 49, 109
scabbard 133
sa 93, 173
scarce 122
-sa 90, 91, 176
scrap 137
saanap- 175
scratch 137
saba 31, 39
scrotum 134
sabi 24
se- 187
sabu/sapuy 24
:se-, :sek-, :sey- See :seyh
sacred tree 206
*sə- 68
sàGwónàW- 175
sea 93, 111
saka-/sakë 181
seat 161
saka-sama n-i 174
seaweed 193
saka-sima n-i See saka-sama n-i
second lunar month 151
sakî 172
sedge 184
sakî- 181
sèk- 181
sakîkusa 180
sèl- 182
sàks 172
:sel 186
sál 93, 173
sèl- ‘clear/wash’ 238
sal-Gi- 29
sell 37
sàm 173
sèngkúy- 23
:sam- 174
sèp 183
sama 174
seph 183
samcil 22
set 101
sanba 39
 Index 271

:seyh, :se-, :sek-, :sey- 180, 211 *siru 41

shade, shadow 140 sirusi 34, 35, 41
shallow cup/plate 121 sïrusï See sirusi
shape 129, 234 sisi 177
share 158 :sit-/sìlú- 185
sheath 133, 134 sita 188
shell 133, 148 situation 174
shell(fish) 148 siwa 39
shin 96, 180 sizii See sïg’i
shines, reflects 140 skewer 144
shove 201 skí- 179
shuttle 108 skim 133
-si 72 skin 133
-si- 90 skwòlí 103
siba 39 sky 186
side 109, 213 sleep embracing [each other] 139
sïg’i 180 sleeps 208
s’ig’i See sïg’i sleeve 177
sïgï See sïg’i slow 206
sign 34, 35, 41 small rain 191
sign 隱 /on/ 217 smartweed 114
siji See sïg’i smelly 155
sikep- 178 Smilax 195
sikhwu- 178 snake 104, 201
sikö tu okîna 49 snare 121
sim- 181 snow 34, 93
sima 41, 183 sö 68, 69, 122
simë- 181, 182 sö-/se 188
sing 108 -sò 176
singkeW- 29 soak in 162
sïni See sinii soaks, permeates 181
sïnï See sinii soba 39
sinii 180 soft 105
sink 120 soften 219
sinks, is immersed 162 sólh 177
siNpa 126 söm- 181
siNtum- 120 sömë- 181, 182
:sil 186 sómóy ~ sómáy 177
sira 184 son 222
sira- 189 son’s wife 196
sirô- 184, 189 sönapar- 189
sirö 167 sönapë- 189
*sirosi See sirusi soncoy 20
272———Koreo-Japonica

sòncóy 23 su 102, 187

song 37 su- 187
söNtar- 189 *su 41
sôNte 178 substitute 168
sôra 186 succeed 123
sort 129 suck 179
sours, tastes acid 178 suffice 161, 238
sowa 39 suk- 124, 179
sòy- 178 sukôsi 168
speak 217 sum- 184
spider 147, 148 sùmúy- 181
spit 167 sun 107
split 111 suna 102, 187
spól- 179 sunago 102, 187
spòlòl- 102 sunairü 189
spread 100 sunairuN 189
spring 104 sune, *sune 180
spring rain 191 suNka-/suNkë 184
sprouts, buds 198 suNkï 179
square basket for fruits 152 sunshade, hat 134
stake 144, 191 sunshine 107
stalk 142 suNsu 180
stamps feet, treads noisily 153 sup- 179
stands side by side 138 support 122
star 107 susu 41, 185
steams it 200 sute- 185
stem 142 *sutomete 93
sticks it (into), points it, pierces it 125 suwe 180
sticks it in place 210 suyur- 178
stirrup 227 swan 141
stone 34, 35, 126 sweeps it away 100
strange 155 swell 109
straw 39, 126 swelling 129
stream 103 *swo 187
street 146 swòkwòm 29
stretch it 211 swól 179
stretches 211 swólá 180
string 166 swón 180
strings to tie hakama under knees 227 swoncwo 22
stuff 141 swonzwo 22
stump 151 sword 134
stupid 233 swóy 180
stwó 123 swuch 185
 Index 273

swúm- 184 *tanto- 25

swùsk 185 tar- 161, 238
:swuy- 178 tara 114
:syem 183 tare 66
:syey- 184, 189 *taro ‘who’ 66
sywó 179 taru, *taro ‘barrel’ 35
tàsós 13, 14
T tàt- 25, 116
tat- 116
ta ‘because’ 118
-tati 119
ta ‘rice field’ 59
tawako 39
*ta ‘who’ 66, 67
táy 117
ta- 11
táylwòng 14
-ta 11, 26
tè 121
-ta- See -te-
te 115
:ta 113
-te, gerund 76, 91
tabako 39
-te, nominal suffix 161
tàhí- 113, 238
-te- 11, 26, 85, 86, 90
tak- 113, 238
tech 121
taka- 114
teci- 18, 20
taka tu sima 49
telephone 39
takara 93
tell 217
take 122
ten 218
takë 117
ten thousand 218
take to outermost limit 161
tes 121
tal 114
tètí-, teti- 17, 18, 20
:tal 114
that 122, 144
tàlàk 114
thick 105
tàlG- See tàló-
thick, fat 38
tàlí 96, 103
thicken 145
tàlì 115
thin 219
tàló- 115
thing 149
:tam- 116, 238
this 144, 237
tamar- 116
thó- 11, 40, 114
tamë 118
thou 203
tamë- 116
thread 186
tamë-/tamar- 238
three 180
:tamón, :tamóyn 116
three-grass 180
tàmpwóy 23
throng 200
tan-da 25
throws it away 100
taNkap- 115
throws it away, discards it 185
tànkì- 23
thwoki- 29
taNta 113
ti 120, 124
taNte 114
274———Koreo-Japonica

:ti- 127 topic marker 217

-ti 90 tôr- 122
tide 201 tör- 122, 238
tie 194 tösi 41
ties it together 218 -toto See -tutu
tìh- 125 tough 156
tìk- 125 töwo tu pîtö 49
ti-mata 200 trace 227, 228
timchoy 20 trap 39, 121
time 163 treads (on it) 103
tiNkap- 116 treasure 93
típh 126 tree 180, 206
tir- 122 troublesome 143
tisay 20 true 190
titi 93 t∫aban 37, 38
title for the clan leader 155 tsïdu 167
tó 118 tsïtsï 167
tô ‘door’ 147 tu 167
tô ‘place’ 147 -tu 49, 50, 51, 52, 53, 54, 62, 74, 140,
tô, tö, te 117 201
tö 73, 74, 75, 76, 90 tu-N-pana 124
-tö 121 tub/muri 93
to ~ tote 76 tuci 41
tobacco 39 tufa 167
tobosi- 122 tuk- 125, 162
*toho- 114 tuka 162
tökî 163 tukî 121
tól 119 tukï, tuku- 119
tól- 119 tukï- 166
TOlal-i 119 *tukor- See tukur-
-tólh 15, 119 tukur- 34
tòlì 103, 120 túl- 128
tóm- 120 tùl- 122, 238
töneri 35 TUl-a 83
tòngkóy- 23, 29 -túlh 15, 119
tongue 188 tùlí- 122
töNka 78 tùlúh 59
töNkamë- 78 túmpwùkí 23
toNkë- 166 túmúl- 122
töNti- 121 tuna 166
too 123 tuNk- 123
tooth 216 tunô 34, 80
top 38, 229 *tuntuk- 163
 Index 275

tu[N]tuk- 162 -u 34, 82, 90

tu[N]tukë- 162 uhu- 31
tura 166 ui 38
turf 126 ukabü 38
turn around 198 ukagoo 38
turns 123 ukaNkapu 38
turns sour 178 ukaNpu 38
turtle 148 *uku- 31
turu ‘crane’ 123, 238 ulph- 236
turu, *turu ‘bowstring, string of a *-um 90
musical instrument’ 166, 167 um- 232
-tusi See -tutu uma- 192
tuti 93, 121, 123, 124 umî, umi 93
tutô(-ni) 78 umo 193
tutômë- ‘to strive’ 78 underground dwelling 230
tutômëte 78, 93 unit of length, inch 159
-tutu 123 uNsu 232
tutui 121 uNtumoru 37
tutupe 167 uNtura 37
túy 25 up 229
tuyu 50 upa-/upë 38, 229
twines it together 218 ùph- 236
twist 218 upper jaw 234
:twochóy ~ :twochwóy 121 upper story 114
twòk 121 ur- 37
:twolh 126 ura 37
twu- 28 uri 38, 233, 238
twùlG- 123 *-urô 82
twùlù- See twùlG- uru 37
twúlwúmí 123, 238 -uru 82
*twupul 25 urusi 233
twute- 93 usaNkî 24
twùté, twùtí-, twùtú- See twùtén used up 166
twùtén, twùtúlk 123 usi 179
twùtícwúy, twutecwuy 123 usirö 124
twúy 124 usu 38, 187
:twuyh 124 usụ 38
tyé 54, 68, 122 uta 37
:tywoh- 124 utaNkapu 37
uti 37
U utu 37
uturu 37
u ‘cormorant’ 231
uuku 230
u ‘hare’ 37
276———Koreo-Japonica

uuṛï 38 warabi 39
ʔuuruN 235 warau 39
*Uwa 38 ware 62, 229
*uwabe 38 ware- 111
ʔuuyuN 235 waribasi 39
uzu 232 warui 39
wash 236
wash [out] (by waves) 183
V wasp 98
wata 93, 111
valley 142
water 5, 33, 116, 139, 197, 198, 201,
value 134
207, 218
variety 129, 141
water dragon 201
vdzdza 37
wat∫aku ‘mischief’ 38
vdzmarü 37
wave 150
vegetable(s) 207, 209
way of being 174
vicinity 213
we 229
vts 37, 38
weak 219
vu 37
wear 95
vụda 38
weaves 218
vụQtsü 37
well 230
vụra 37
wenu 35
vụsï 38
wh- 33, 67, 233
vụta 37
wh- questions 128, 129
vutagoo 37
whale 142
vụtsị̈ 37, 38
what 199, 223
vụtsụ 37
which 199, 233
vụttī 38
whip 38
white 184, 189
W
why 223
wa 27, 60, 62, 63, 65, 229 wi 230
waa 38 wife 196
waabi 38 wild 175
waari- 38 wild duck 231
ʔwacaku 38 wild, uncultivated field 211
wages 172 wild walnut 136
waist 188 winds it 132
wait 190 wine 181
walnut 136 winter 151
wana 39 with 202
wang 37 wo ‘hemp’ 112, 174
war- 111 wo ‘small’ 35
wara 39 -wo 55, 62
 Index 277

wo- 27, 28 *yacai 41

wolf 236 yak- 40
wòlG- See wòló- yakan 42
wòlhí 231 yam 193
wòlm- 232 yama 40
wolm- 29 yamï 169
wolm-ki- 29 yaNte 52, 171
wòló- 232 yasai 41
woman 196 yase- 221
womîna 196 ye 170
womuna 197 ye- 34
wòmwòk 230 year of age 186
*won- 27 yebinda 221
wood texture pattern 150 yəbinda 221
world 103, 216 yèc- 21
wós 232 yècò- 21
wòs 233 yel 218
wosa 78 :yel- 218
wosamë- 78 yèléh 218
wosaNki, *wosanki 24 yélh 25, 218, 221
wotö- 51 yellow 153
wotö-kô 51 yèlp- 219
wotö-mê 51 yèmkúl- 23
:woy 233, 238 yènc- 21
wrinkle 39 yèncò- 19, 21
wùh 229 yèns- 21
wúlí 62, 229 yeNta 52, 171
wùmchì- 23 yès- 21
wùmchú- 23 yèsús 14
wùmcùkí- 23 yètólp 219
wúmh 230 yètón 219
wùmúl 230 yètúlp 14, 25, 219, 221
wùmwúk- 230 yètún 14, 25, 219, 221
yèwúy 221
X yö 216
yö- ‘four’ 211
xatun 151, 187
yö- ‘good’ 34, 176
-yo 55
Y
yoki 34, 93
ya ‘arrow’ 93 yökö-sama n-i 174
ya ‘eight’ 219 yökö-sima n-i See yökö-sama n-i
-ya 93 yomi 169
yabunda 221 yömö- 169
278———Koreo-Japonica

yömö-nö kuni 169 ʔyu 55

yoogasa 42 -yu 55
yör- 212, 219 yuï 34
yora 218 yukî 34, 93
yorai 218 yuru 34
yöröNtu 218 yuru- 219
yôru 34, 40 ʔyuu 55
yöse- 212
yotolp See yètólp Z
young woman 196
zigzag cut pieces of cloth or paper 210
yôwa- 219
-zoW- 26
yu 40
ABOUT THE AUTHOR

Alexander Vovin has published some hundred articles on Japanese,

Korean, Ainu, and Tungusic, as well as on other languages of East
and Inner Asia. Among his major works are A Reconstruction of
Proto-Ainu (1993); A Reference Grammar of Classical Japanese
Prose (2003); Nihongo Keitōron no Genzai/Perspectives on the
Origins of the Japanese Language, a bilingual Japanese-English
volume co-edited with Osada Toshiki (2003); The Critical Edition
and the Translation of the Man’yōshū, Book 15 (2009), other books
to follow; and A Descriptive and Comparative Grammar of Western
Old Japanese, Part 1(2005) and Part 2 (2009).
 { Continued from front flap } HAWAI‘I STUDIES ON KOREA
LANGUAGE AND LINGUISTICS

Of related interest
comparisons are not due to pure chance,
they can almost always be explained as

VOVIN
borrowings from Korean into a central The Role of Contact in the Origins
group of Japanese dialects from roughly of the Japanese and Korean Languages The Japonic (Japanese and Ryukyuan)

KOREO-JAPONICA
between the third and eighth centuries J. Marshall Unger portmanteau language family and the
2008, 224 pages
A.D. The remaining group of lexical (but Korean language have long been consid-
cloth: ISBN 978-0-8248-3279-7
not morphological) comparisons that ered isolates on the fringe of northeast
cannot be explained in this way is, he Despite decades of research on the reconstruction of proto-Korean- A R E - E VA L U AT I O N O F A C O M M O N G E N E T I C O R I G I N Asia. Although in the last fifty years many
Japanese (pKJ), some scholars still reject a genetic relationship. This
argues, too small to serve as proof of even specialists in Japonic and Korean histori-
study addresses their doubts in a new way, interpreting comparative
a distant genetic relationship. cal linguistics have voiced their support
linguistic data within a context of material and cultural evidence,
In this volume, a leading historical much of which has come to light only in recent years. for a genetic relationship between the two,
linguist presents a significant challenge to this concept has not been endorsed by
a view widely held by Japonic and Korean general historical linguists, and no signifi-

KOREO-JAPONICA
historical linguistics on the relationship Also in Hawai‘i Studies on Korea cant attempts have been made to advance
between the two language families and beyond the status quo. Alexander Vovin, a
offers material support for the skepti- Questioning Minds longtime advocate of the genetic relation-
cism long espoused by general histori- Short Stories by Modern Korean Women Writers ship view, engaged in a reanalysis of the
cal linguists on the matter. His findings Yung-Hee Kim known data in the hope of finding evi-
will both challenge and illuminate issues 2010, 248 pages, illus. dence in support of this position. In the
of interest to all linguists working with cloth: ISBN 978-0-8248-3395-4 process of his work, however, he became
language contact and typology as well as paper: ISBN 978-0-8248-3409-8 convinced that the multiple similarities
those concerned with the prehistory and Available for the first time in English, the ten short stories by modern between Japonic and Korean are the result
early history of East Asia. Korean women collected here touch in one way or another on issues of several centuries of contact and do not
related to gender and kinship politics. All of the protagonists are descend from a hypothetical common
women who face personal crises or defining moments in their lives as
Alexander Vovin is professor of East ancestor.
gender-marked beings in a Confucian, patriarchal Korean society.
Asian languages at the University of In Koreo-Japonica, Vovin carefully
Hawai‘i. reviews recent advances in the recon-
struction of both language families. His
jacket art: (left) Lanterns at the back of Nigatsudō Temple, Nara, Japan;
detailed analysis of most of the morpho-
(right) a Paekche royal tomb, Gongju, Korea. (Photos by author) logical and lexical comparisons offered
jacket design: Julie Matsuo-Chun so far shows that whenever the proposed

{ Continued on back flap }

ISBN 978-0-8248-3278-0 ALEXANDER VOVIN
UNIVERSITY 90000
of HAWAI‘I
PRESS 9 780824 832780
HONOLULU, HAWAI‘I 96822-1888 www.uhpress.hawaii.edu