Is Japanese related to Korean, Tungusic, Mongolic and Turkic? [thesis]

648 91 20MB

English Pages 968 Year 2005

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Is Japanese related to Korean, Tungusic, Mongolic and Turkic?  [thesis]

Table of contents :
0.Dedication......Page 1
01.7-10Contents......Page 2
02.11preface......Page 6
03.13-17Ch1Conventions......Page 7
04.18-29Ch2History......Page 12
05.30-40Ch3Interdisciplin......Page 24
06.41-51Ch4Methodology......Page 35
07.52-80Ch5Individual......Page 46
08.81-173Ch6Internal......Page 75
09.174-186Ch7Universals......Page 168
10.187-217Ch8Borrowing......Page 181
11.218-285nCh9Semantics......Page 212
12.286-377Ch10Sound......Page 280
13.378-421Ch11Core......Page 372
14.422-423Ch12Conclusion......Page 416
15.425-428Abbreviations......Page 418
16.429-452Bibliography......Page 422
17.453-480LexiconA......Page 446
18.481-484LexiconB-D......Page 474
19.485-488LexiconE......Page 478
20.489-491LexiconF-G......Page 482
21.492-548LexiconH......Page 485
22.549-569LexiconI......Page 542
23.570-673LexiconK......Page 563
24.674-712LexiconM......Page 667
25.713-747LexiconN......Page 706
26.748-768LexiconO......Page 741
27.769-774LexiconP-Q......Page 762
28.775-837LexiconS......Page 768
29.838-907LexiconT......Page 831
30.908-926LexiconU......Page 901
31.927-934LexiconW......Page 920
32.935-954LexiconY-Z......Page 928
33.955-975Morphology......Page 948

Citation preview

Voor Frits

Contents Preface ...........................................................................................................................00011 Chapter 1. Conventions................................................................................................ 00013 1.1. Transliteration conventions................................................................... 00013 1.2. Chronological division............................................................................ 00014 1.2.1. Japanese................................................................................................. 00014 1.2.2. Korean................................................................................................... 00015 1.2.3. Tungusic................................................................................................ 00016 1.2.4. Mongolic................................................................................................ 00016 1.2.5. Turkic.....................................................................................................00017 Chapter 2. The history of the question........................................................................00018 2.1. The Altaic question................................................................................. 00018 2.2. The Japanese-Altaic question................................................................ 00020 2.3. The Japanese-Korean question..............................................................00024 2.4. The present question............................................................................... 00026 Chapter 3. Interdisciplinary research......................................................................... 00030 3.1. Archaeology............................................................................................. 00030 3.2. Biology......................................................................................................00035 3.3. The Austronesian connection................................................................. 00036 3.4. A working hypothesis............................................................................ 00038 Chapter 4. Methodology............................................................................................... 00041 4.1. What is a genetic relationship?..............................................................00041 4.2. What is a genetic argument?..................................................................00042 4.2.1. A negative argument.............................................................................. 00042 4.2.2. A probabilistic argument........................................................................00047 4.3. Measuring time-depth............................................................................ 00051 Chapter 5. The individual phonological inventories.................................................. 00052 5.1. Japanese................................................................................................... 00052 5.1.1. Consonant inventory.............................................................................. 00052 5.1.2. Vowel inventory.....................................................................................00057 5.2. Korean......................................................................................................00060 5.2.1. Consonant inventory.............................................................................. 00060 5.2.2. Vowel inventory.....................................................................................00066 5.3. Tungusic................................................................................................... 00068 5.3.1. Consonant inventory.............................................................................. 00068

8

Contents 5.3.2. Vowel inventory.....................................................................................00071 5.4. Mongolic...................................................................................................00072 5.4.1. Consonant inventory.............................................................................. 00072 5.4.2. Vowel inventory.....................................................................................00074 5.5. Turkic....................................................................................................... 00075 5.5.1. Consonant inventory.............................................................................. 00075 5.5.2. Vowel inventory.....................................................................................00079

Chapter 6. Internal evidence contradicts the etymology........................................... 0081 6.1. Lexicon..................................................................................................... 0081 6.1.1. Dubious morphological analysis............................................................ 0081 6.1.1.1. Undetected morpheme boundary........................................................ 0081 6.1.1.1.1. Nouns............................................................................................... 0082 6.1.1.1.2. Verbs................................................................................................ 00114 6.1.1.1.3. Adjectives........................................................................................ 00137 6.1.1.2. Unjustified morpheme boundary........................................................ 00139 6.1.1.2.1. Nouns............................................................................................... 00140 6.1.1.2.2. Verbs................................................................................................ 00142 6.1.1.2.3. Adjectives........................................................................................ 00144 6.1.2. Dubious phonological analysis.............................................................. 00145 6.1.2.1. Problematic *y.................................................................................... 00145 6.1.2.2. Problematic *w................................................................................... 00147 6.1.2.3. m~b alternation................................................................................... 00148 6.1.2.4. p~k alternation.................................................................................... 00151 6.1.2.5. s~t alternation......................................................................................00152 6.1.2.6. Sporadic consonant elision................................................................. 00153 6.1.2.7. Vocalism............................................................................................. 00155 6.2. Morphology............................................................................................. 00157 Chapter 7. Similarities due to general properties of language................................. 00174 7.1. Nursery words......................................................................................... 00174 7.2. Sound symbolism.................................................................................... 00176 Chapter 8. Contact induced similarities: borrowing................................................. 00187 8.1. Donorword attested................................................................................ 00189 8.1.1. Nouns..................................................................................................... 00190 8.1.2. Verbs...................................................................................................... 00196 8.2. Cultural vocabulary............................................................................... 00196 8.2.1. Agricultural vocabulary........................................................................ 00198 8.2.2. Technological vocabulary.....................................................................00202 8.2.3. Medical vocabulary................................................................................00205 8.2.4. Religious vocabulary............................................................................. 00205 8.2.5. Governmental and administrative vocabulary....................................... 00207 8.2.6. Martial vocabulary................................................................................. 00208 8.2.7. Artistic vocabulary.................................................................................00209

Contents

9

8.2.8. Clothing and jewellery...........................................................................00209 8.3. Poor attestation....................................................................................... 00209 Chapter 9. Suspect semantics....................................................................................... 00218 9.1. Ghost words and misglosses................................................................... 00218 9.2. Secondary semantics...............................................................................00229 9.3. Different etymologies for one etymon................................................... 00235 9.4. One etymology for different etyma....................................................... 00248 9.5. Semantic overpermissiveness.................................................................00251 Chapter 10. Sound correspondences: Look-alike or cognate?................................. 00286 10.1. Consonant correspondences.................................................................00287 10.1.1. Initial consonant correspondences....................................................... 00287 10.1.1.1. pJ *p-.................................................................................................00287 10.1.1.2. pJ *t-................................................................................................. 00292 10.1.1.3. pJ *s-................................................................................................. 00299 10.1.1.4. pJ *k-.................................................................................................00304 10.1.1.5. pJ *m-............................................................................................... 00312 10.1.1.6. pJ *n-.................................................................................................00316 10.1.1.7. pJ *w-................................................................................................00319 10.1.1.8. pJ *y-.................................................................................................00321 10.1.2. Medial consonant correspondences..................................................... 00323 10.1.2.1. pJ *-p-............................................................................................... 00323 10.1.2.2. pJ *-t-................................................................................................ 00326 10.1.2.3. pJ *-s- and pJ *-ns-........................................................................... 00330 10.1.2.4. pJ *-k-............................................................................................... 00333 10.1.2.5. pJ *-m-.............................................................................................. 00336 10.1.2.6. pJ *-n-............................................................................................... 00338 10.1.2.7. pJ *-w-.............................................................................................. 00339 10.1.2.8. pJ *-r-................................................................................................ 00340 10.1.2.9. pJ *-y-............................................................................................... 00344 10.1.3. Medial cluster correspondences...........................................................00345 10.1.3.1. pJ *-np-............................................................................................. 00345 10.1.3.2. pJ *-nt-.............................................................................................. 00347 10.1.3.3. pJ *-nk-............................................................................................. 00349 10.2. Vowel correspondences........................................................................ 00351 10.2.1. Initial vowel correspondences..............................................................00351 10.2.1.1. pJ *a-.................................................................................................00351 10.2.1.2. pJ *i-................................................................................................. 00353 10.2.1.3. pJ *o-.................................................................................................00354 10.2.1.4. pJ *u-.................................................................................................00356 10.2.2. Medial vowel correspondences............................................................00357 10.2.2.1. pJ *-a-................................................................................................00357 10.2.2.2. pJ *-i-................................................................................................ 00363

10

Contents 10.2.2.3. pJ *-o-............................................................................................... 00365 10.2.2.4. pJ *-u-............................................................................................... 00368 10.3. The overall system of sound correspondences....................................00373 10.3.1. The consonant correspondences.......................................................... 00373 10.3.2. The vowel correspondences.................................................................00376

Chapter 11. The core evidence..................................................................................... 00378 11.1. Lexical core evidence............................................................................ 00378 11.1.1. Pronouns.............................................................................................. 00379 11.1.2. Verbs.................................................................................................... 00380 11.1.3. Adjectives and quality nouns............................................................... 00395 11.1.4. Nouns................................................................................................... 00399 11.1.4.1. Basic vocabulary...............................................................................00399 11.1.4.2. Non-basic vocabulary....................................................................... 00407 11.2. Morphological core evidence............................................................... 00412 11.2.1. Bound derivational morphemes........................................................... 00413 11.2.1.1. Verbal derivation.............................................................................. 00413 11.2.1.2. Nominal derivation .......................................................................... 00416 11.2.2. Unbound morphemes........................................................................... 00416 11.2.2.1. Case markers.....................................................................................00416 11.2.2.2. General markers................................................................................ 00420 Chapter 12. Conclusion: ‘Yes.’.................................................................................... 00422 List of abbreviations..................................................................................................... 00425 Bibliography.................................................................................................................. 00429 Etymological Index of Japanese...................................................................................00453 Lexicon............................................................................................................ 00453 Morphology..................................................................................................... 0955

11

Preface The present work is based on my doctoral dissertation which was written under supervision of Prof. Dr. F. Kortlandt and which was defended at Leiden University on October 22, 2003. Many people and organizations have supported me in one way or another to finish this book. I must acknowledge the facilities created by the resources of the Spinoza Prize awarded to Prof. Dr. F. Kortlandt by the Netherlands Organization for Scientific Research NWO. I owe gratitude to the Canon Europe Foundation for its support, creating the facilities for my postdoctoral reserach at the department of linguistics of the University of Tokyo in 2004. I would further like to thank the Japan Foundation for awarding me a Japan Foundation Fellowship for ten months, from February 2005 to December 2005, without which I could not have continued my research at the University of Tokyo. Much of this work has grown out of discussions with Prof. Dr. S. A. Starostin, Prof. Dr. R. Finch and Prof. Dr. M. H. Miyake whose visits to Leiden have been supported by the Spinoza Prize. Needless to say, this does not imply that they are in agreement with me on the ideas that I advance in this book. A special word of thank goes out to Richard James Lockett for checking the manuscript for readability and for helping me with my English. I could not have written this book without the loving support of my husband Marc and I owe him many pizzas for his help in editing the manuscript. But most of all I am grateful to the man to whom this book is dedicated. Tokyo, July 14, 2005

Martine Robbeets

Chapter 1. Conventions 1.1. Transliteration conventions The present work is an attempt to gather and evaluate the etymologies that have been proposed in the past relating Japanese to Korean and Altaic. The cover term Altaic will be used in reference to the Tungusic, Mongolic and Turkic languages. It refers to a group of North-East Asian languages that share a number of phonological, morphological and structural similarities without neccessarily presupposing that these similarities are due to common ancestorship. The etymologies proposed so far in order to relate Japanese to Korean and Altaic are listed in the etymological index of Japanese in the appendix of this book. The etymological index is keyed into the first part of this work, referring to the page numbers on which each Japanese entry along with its etymological proposal(s) is evaluated. In the index I cite lexical and morphological data that are suggested by various authors, respecting the originally chosen romanization and pitch conventions of the author in question. Usually, when citing an author or an etymology in the textual part of this work, reference is made to the covering etymological index. Otherwise additional reference is made in the footnotes. In the textual part of this book I will use the following transliteration conventions. The transliteration of Japanese expressions adopted in this work basically follows the kunreishiki , i.e. Yale system, for cited linguistic forms such as glosses and text examples and the Hepburn system for proper nouns, such as personal names, titles of books and articles, geographical names and historical periods. The essential difference is that the syllables [tsU], [Si], [tSi], [Fu], [dZi] are written as tsu, shi, chi, fu and ji in the Hepburn system and as tu, ti, hu, zi/ di in the kunreishiki system. The kunreishiki system which is phonemic in nature and largely based on the arrangement of the hiragana syllabary, is widely used in linguistics. Although the Japanese government promulgated the kunreishiki system in an attempt to unify all different systems, the Hepburn system, which is based on the system of English spelling pronunciation, is retaining its popularity in every day use. As far as the vowel distinctions attested for Old Japanese are concerned, I will use i1 versus i2 and e1 versus e 2 for the two different types of front vowels after velars (k, g) and labials (p, b, m) and o1 and o2 for the two different types of back mid vowels after velars (k, g), dentals (t, d, s, z, n), r, y and after Kojiki OJ m. The transliteration rules for transcribing Hangu#l adopted in this work follow Yale romanization for cited linguistic forms such as glosses and text examples and the classical Mc Cune Reishauer system for proper nouns, such as personal names, titles of books and articles, geographical names and historical periods. For Middle Korean the Yale romanization is modified to allow for the now obsolete vowel written o, which was probably pronounced like a somewhat unrounded [O], and for the obsolete consonants -W- (probably [B]), -G([V] or [H]), and -∆- ([z#]). Unlike -W- and -∆-, -G- was not assigned a separate grapheme in Middle Korean, but it was written with a graphic device. When a consonant final syllable is joined to a vowel initial syllable in Middle Korean, the final consonant is written as the

14

Chapter 1. Conventions

onset of the vowel-initial syllable. Graphic junctures, whereby a syllable-final -l or -∆ is restrained from becoming the onset of the tightly joined vowel-initial syllable, are represented with -G-. The Middle-Korean orthographic convention ∆ is represented as such in order to avoid taking an a priori position about its still debated phonetic value.The Korean vowel u is unrounded [ˆ]; wu and wo are rounded [u] and [o] respectively, and for modern Korean the w can be omitted for wo, and after the labials for wu, since there are no surface contrasts with the unrounded u and o. The dots in the Middle Korean words represent the distinctive pitch of the following syllable: one dot for high, two dots for rising and unmarked syllables are treated as low. The transliteration rules for transcribing Tungusic adopted in this work follow the romanization proposed by Benzing (1955) with some modifications for the voiced velar [x] that is usually transliterated as hin Manchu and Nanai, and as x in the other Tungusic languages. Contrary to this convention I use the transliteration x for all Tungusic languages. The palatal nasal is represented by n!. The mid front vowel is represented by e. The transliteration rules for transcribing Mongolic adopted in this work follow Poppe’s conventions.1 The mid front vowel is represented by e. The palatal glide is represented by j. The transliteration rules for transcribing Turkic follow Johanson (1998), except for the palatal glide which is represented by j. 1.2. Chronological division 1.2. 1. Japanese Dividing the history of Japanese into a number of time periods is of interest to historical linguists. The following five historical stages represent the periodization of the Japanese language most widely used in Japan. These stages roughly correspond to the periods of political history. Table 1 summarizes the major language stages and political periods.

1

Poppe 1954; 1955, 95-171.

1.2. Chronological division

15

Table 1. Chronological division of Japanese

Abbreviation Language stage!

Japanese term!

Political period

Date AD

OJ

Old Japanese

jo @d ai nihongo

Nara

710-794

EMJ

Early Middle Japanese

chu @ko nihongo

Heian

794-1191

LMJ

Late Middle Japanese

chu @s ei"nihongo

Kamakura/ Muromachi

1192-1602

Mod. J

Modern Japanese

kinsei nihongo

Edo

1603-1867

J

Contemporary Japanese

gendai nihongo Meiji and after

1868-

1.2 2. Korean Korean today is for all practical purposes a single language, but it has not always been that way. During the Three Kingdoms Period three languages, Koguryo#, Paekche, and Silla are spoken on the peninsula. Referring to Old Korean, it will be necessary to specify which of the three languages is meant. For historical linguistics the invention of the Korean writing system in 1443 is an important turning-point, because all our knowledge of the phonological system before that date is fragmentary and speculative, the written records of Early Middle Korean being in Chinese characters. Of course 1443 does not mark an abrupt change of the language, but rather a sudden increase of our knowledge about the language spoken at that time. It is common in Korean historical linguistics to divide the history of the language into five major periods. Table 2 summarizes the major language divisions and political periods.

16

Chapter 1. Conventions Table 2. Chronological division of Korean Abbreviation

Language stage!

Korean term!

Historical period

Date A.D.

OK

Old Korean

kodae kugo#

Three KingdomsUnited Silla

0-918

EMK!

Early Middle Korean

chungse kugo#

Koryo#Hangu#l invention

918-1443

LMK

Late Middle Korean!

chungse kugo#

Hangu#l-Imjin wars

1443-1592

Mod. K!

Modern Korean

ku #ndae kugo#

Imjin begin 20th C

1592-1900

K

Contemporary Korean

hyo#ndae kugo#

20th C and after

1900-

1.2.3. Tungusic For the historical study of Tungusic languages it is unfortunate that written records are nonexistent for most of the languages. The oldest records are written in Jurchen, dating back to the period from 1115 to 1234, when a confederacy of Tungusic tribes ruled over North China under the dynastic name Jin. However, Jurchen writing has not been completely deciphered yet and our knowledge about the Jurchen language is fragmentary and speculative. Much more extensive is the literature that is written in Manchu. From 1644 until 1911 when the Manchus were in power in China during the Qing dynasty, Manchu was the official language. Almost all Manchu written sources are official literature. As far as the other Tungusic languages, Evenki, Ewen, Solon, Negidal, Nanai, Ulcha, Orok, Oroch, and Udehe are concerned, they are only known to us in their contemporary form. 1.2.4. Mongolic The history of the Mongolic languages begins with the erection of a stele in honor of Chinggis Khan’s nephew Yissünge between 1225 and 1270. It is improbable that this inscription represents the first attempt at writing, but it is the earliest written record that is in our possession. What is known of the language before 1225 is fragmentary and highly speculative, since it comes from a small number of Mongolic words in Chinese documents and from some Mongolic loanwords in the surrounding Tungusic and Turkic languages. Prior to the time of Chinggis Khan Mongolic may be assumed to be a unity of geographically dispersed, but mutually intelligible tribal languages. With the linguistic unification under Chinggis Khan the tribal languages develop into a more homogeneous Middle Mongolian. The modern period begins around the beginning of the seventeenth century when an extensive Buddhist literature is created. The classical written language of the modern period is characterized by a high degree of standardization. By the 17th century some important phonological changes have occurred in most of the Mongolic languages. After the liberation

1.2. Chronological division

17

of Chinese domination, contemporary Mongolian develops into a modern tool for science, politics and mass communication. Table 3 summarizes the major language divisions. Table 3. Chronological division of Mongolic

Abbrev.

Spoken Written Language stage Language stage

Period

Date A.D.

MMo.

Middle Mongolian

Pre-classical Written Mongolian

Yissünge inscription -Buddhist literature

1225-17thC

Mod. Mo. Modern Mongolian

Classical Written Mongolian

Buddhist literature

17thC-19thC

Mo.

Modern Written Mongolian

Liberation from Chinese domination

20thC

Contemporary Mongolian

1.2.5. Turkic Although the Turkic written sources do not mirror the spoken varieties of a certain region in a direct way, the periodization of Turkic is largely based on the emergence and development of literary idioms in various cultural centers. The history of Turkic begins with the earliest known inscriptions on stone steles in present-day Mongolia’s Orkhon valley. In the Middle period a wide variation of literary languages emerges. The literature is linguistically heterogeneous and contains a mixture of local elements. From the sixteenth century on a limited number of prestigeous literary languages such as Chaghatay serve as a transregional medium over a vast area. In the second half of the 19th century regional written languages replace these prestige languages again, much as a result of Russian influence. Table 4 summarizes the major language divisions of Turkic. Table 4. Chronological division of Turkic Abbreviation

Language stage

Period

Date A.D.

OTk.

Old Turkic

Orkhon inscriptions -Mongol rule

8th C -13th C

MTk.

Middle Turkic

heterogenous regional written traditions

13th C -16th C

Mod. Tk.

Modern Turkic

development of transregional literary prestige languages

16th C -20th C

Co. Tk.

Contemporary Turkic

creation of new regional written languages

1850-

Chapter 2. The history of the question 2.1. The Altaic question After a long and stormy history, the Altaic question for many scholars and linguistic amateurs remains what it originally was: a question. Nevertheless, during more than two centuries of Altaic studies some important steps forward have been made. The Turkic language family is a well established family, as are both the Mongolic and Tungusic families. The possibility of an Altaic family, comprising Turkic, Mongolic and Tungusic is accepted by a fairly large group of scholars. But, as illustrated in section 2.4., this ‘core’ Altaic family is by no means accepted by everyone in the field. There seems to be a consensus that the various languages within the individual branches of Altaic are indeed genetically related, but there is still some doubt that the various branches themselves are genetically linked together. What is most controversial, however, is whether Korean and Japanese are related as members of the Altaic family. Even some supporters of the core Altaic hypothesis do not accept the inclusion of these languages into Altaic. The first classification of languages now considered Altaic is found in von Strahlenberg (1730). The Swedish adventurer Phillip Johann von Strahlenberg posits a Tatar family comprising Finno-Ugric, Turkic, Samoyed, Mongol-Manchu, Tungusic and Caucasian. In the 19th century some scholars revise his proposal, either expanding or narrowing its scope. In 1834 Rask adds Basque, Chukchi-Kamchatkan, and Eskimo to the family proposed by Strahlenberg and renames it the Scythian family. Rasmus Rask also correctly groups Manchu with Tungusic, not with Mongolic as von Strahlenberg proposes. The family is further expanded by Max Müller in 1855, who also includes Thai, Tibetan, Dravidian, and Malay. Once again the family is renamed, this time as Turanian. The term ‘Altaic’ is first used in the middle of the 19th century by Castrén, the first scholar who applies linguistic criteria to the group of languages he believes to belong to the same family. He rejects the classical typological approach and insists on lexical and morphological comparison. Castrén narrows the scope of the Altaic family to Finno-Ugric, Samoyed, Turkic, Mongolic and Tungusic. His views are not published until 10 years after his death in 1862. Another scholar active around the same time in the same field is Schott, who posits in 1849 an Altaic or Chudic-Tatar family, consisting of a Chudic or Finnic branch and a Tatar branch. Schott’s views are very similar to Castrén’s theory, with the exception that he believes that Turkic belongs to the Tatar branch with Mongolic and Tungusic and not to the Finnic branch, as in Castrén’s proposal. The languages called ‘Altaic’ by Castrén and Schott are now called ‘Ural-Altaic’. The Ural-Altaic family consists of a Uralic branch splitting in Finno-Ugric and Samoyed and a core Altaic branch. The Ural-Altaic theory was rather widely accepted in the 19th century, but after the affinity of the Uralic languages was established, the Ural-Altaic theory lost many of its supporters. Although it is commonplace in contemporary linguistic literature to reject the Ural-Altaic theory as such, the debate on possible genetic links between well-

2.1. The Altaic question

19

established families is more animated than ever before. It is in this light that attempts to establish Nostratic or Eurasiatic can be seen. As soon as the existence of an Altaic linguistic unity is posited, the course is open to counterarguments. As early as 1820 Abel-Rémusat expresses his doubts about the matter, arguing that the languages in question do not share a sufficient amount of basic vocabulary. The torch is handed on to Gerard Clauson, who writes ‘The case against the Altaic theory’ in 1956 and reformulates his doubts in numerous articles during the sixties. Just like his French predecessor, Clauson concentrates his attention on the lack of basic vocabulary in Altaic. But Clausons arguments are more centered upon mathematical assumptions of lexicostatistics or glottochronology. The Finnish linguist Gustaf John Ramstedt is usually regarded as one of the founders of contemporary Altaic linguistics. Ramstedt’s ‘Einführung in die altaische Sprachwissenschaft’ is published posthumously by Aalto in 1957 and 1952. With his profound knowledge of the Uralic and Altaic languages and his good command of Japanese and Korean, Ramstedt makes considerable contributions to the field. However, in his posthumous work many problematic forms can be found, since the author during the last months of his final illness often cites words and their meanings from memory, rarely mentioning his sources.1 The year 1960 sees the publication of Poppe’s ‘Vergleichende Grammatik der altaischen Sprachen’. Although it is meant to be the first part of a handbook for Poppe’s students, ‘Teil 1, Vergleichende Lautlehre’ is the only part to appear. The publication largely draws on ‘Altaïsch und Urtürkisch’, Poppe’s pioneering contribution from 1926. When familiarizing oneself with the Altaic literature, Nicholas Poppe’s ‘Introduction to Altaic linguistics’ (1965) is still a must. In the same year that Poppe published his ‘Vergleichende Lautlehre’ Doerfer submits his Habilitationsschrift to Göttingen. The German linguist reworks his thesis in a monumental work dealing with Turkic and Mongolic loanwords in various stages of Iranian (1963-1975). But the true nature of the four volumes on loanwords is an item-by-item refutation of the Altaic etymologies present in literature at that time, particularly those proposed by Ramstedt and Poppe. Doerfer correctly mentions the necessity of sifting the Altaic evidence in terms of borrowing, universals in linguistic structuring and semantic latitude. But, he is probably wrong in attributing the remaining similarities to chance. An attentive reader might raise the question whether Doerfer’s ‘Zufall’ is a synonym for ‘cognate’. Today, in the light of the recent developments in the Altaic field, Doerfer’s observations become increasingly relevant, both from a factual and a methodological viewpoint. Since the pioneers of Altaic linguistics are in the first place Mongolists like Poppe and Ramstedt or Turkologists like Clauson and Doerfer, Altaic studies are until the late sixties focused on Turkic, Mongolic and Tungusic. It is not until the beginning of the seventies that scholarly interest is renewed in the possibility that Japanese and Korean may also represent branches of the Altaic family.

1

Aalto 1984, 161-193.

20

Chapter 2. The history of the question

2.2. The Japanese-Altaic question Japanese can boast a wider range of languages and language families for which an attempt to establish a genetic relationship with has been made than can any other language in the world. Hypotheses have been presented, connecting Japanese to numerous languages and assigning it to many different language families: Altaic, Uralic, Ainu, Austronesian, the Papua languages, Austro-Asiatic, Sino-Tibetan, Indo-European, Sumerian, Dravidian etc.2 The least weak is the hypothesis that Japanese is an Altaic language but even here evidence is much less clear-cut than that which relates the languages of the Indo-European family. Despite significant research activity in the Japanese-Altaic field, a clear overview of the evidence suggested so far is lacking at the moment. The lack of a global state of the art leads to the dissemination of extreme positions among both linguists and language amateurs that ‘the Altaic affiliation of Japanese is premature’ or that ‘the Altaic hypothesis for Japanese is not tenable’, or that ‘Altaic is dead’ or that ‘nobody still supports Altaic today’.3 Also the fact that so many different ideas have been offered for the affinity of Japanese causes the Altaic theory to encounter much scepticism. Whereas the first line of development in the Altaic question is to narrow the family down from Ural-Altaic to Altaic, comprising Turkic, Mongolic and Tungusic, the second line of development is the expansion of core Altaic to include Korean and Japanese. A systematic comparison of Japanese and Korean to the languages making up the proto-Altaic linguistic unity is only possible after the appearance of Poppe’s Lautlehre in 1960, but the idea that Japanese is an Altaic language is by then already of some hundred years standing in the literature. As early as 1832 von Siebold suggests a genetic relationship between Japanese and Altaic in his ‘Verhandeling over de afkomst der Japanners’.4 In spite of the poor Japanese data at his disposal, the comparativist Boller tries to compare Japanese phonology and morphology with that of the Altaic languages. Although Boller’s methodology in his 1857 publication is linguistically orthodox, the weakness of his work lies in the accuracy of his materials. Around the turn of the century Japanese linguists seem to be less 2

3

The Austronesian hypothesis concerning the origins of Japanese is supported by Polivanov 1918; Matsumoto 1928; Izui 1952; Murayama 1974 a, 1974 b, 1978; Kawamoto 1977, 1978; Benedict 1990; Maher 1996; Sakiyama 1996, 1998, 2001; Itabashi 2003. Go @1980 relates Japanese to the Papua languages of New Guinea. A comparison with Austro-Asiatic languages is made by Parker 1939 who focusses on Mon-Khmer. As for Sino-Tibetan, Yasuda 1976 compares Japanese with Lepcha and Nishida 1980 concentrates on Tibeto-Birman. Aston 1874 compares Japanese to Indo-European. Yoshiwara 1991 relates Japanese to Sumerian. Ono 1980, 1982; Shiba 1980 pursue a genetic relationship between Japanese and Tamil. Janhunen 1994,237: “For instance, while Miller (1991) may well be right in presenting himself as a better linguist than Paul K. Benedict (1991), this should not serve as a reason to prefer the “Altaic” affiliation of Japanese to the “Austric” one: both are, in their current forms, equally premature.” In his summary report of the Altaic panel,held at a conference in Stanford, and consisting of Martin, Whitman, Austerlitz, Clark and Unger himself, Unger (1990, 479) states: “In short, we found Proto-Altaic, at best, a premature hypothesis and a pragmatically poor foundation on which to build a sustained research program.” and (1990, 481): “We urge researchers to focus instead on the more likely family that, at George Grace’s suggestion, we call MacroTungusic, i.e., the generally recognized Tungusic languages plus Korean, Japanese, and the now extinct languages of Korea before the Silla unification.”

2.2. The Japanese-Altaic question

21

interested than their European colleagues in the affinity of their language. In 1908 the linguist Fujioka Katsuji points out a number of typological features shared by Japanese and Ural-Altaic. His observations are accurate and striking, but the weakness of typological evidence in support of a genetic relationship will be discussed below. Pröhle (1916-17) shows a great advance on Boller accessing the Japanese data, but working within the same vast framework of Ural-Altaic as Fujioka did, few of his etymologies are still relevant. With his detailed knowledge of the languages of Central Asia, specializing in Japanology and Turkology, and his profound command of general linguistics, the Russian polyglot Evgenij Polivanov investigates the affiliation of Japanese (1918, 1924, 1927, 1960). He is also the first to become involved in comparative studies of Japanese and Ryukyu (1914). In 1924 Polivanov argues that Japanese is hybrid in origin, consisting of both southern, insular, Austronesian elements and western continental elements, held in common with Korean and Altaic. Taking into account general structural similarities between Korean and the Altaic languages, such as agglutination, vowel harmony and the general structure of the phonological system, he proposes to include Korean as a member of the Altaic family (1927). However, it is Ramstedt (1939, 1949) who first adduces systematic comparative evidence for the kinship relation of Korean and the Altaic languages. The Finnish scholar Ramstedt has been mentioned above with respect to his contributions to the Altaic question in general, but he is also active in investigating the possible links between Japanese and Altaic (1924). Poppe restricts himself to a brief reflection on the Japanese affiliation problem, mentioning the possibility of an Austronesian substratum in his foreword to Miller’s book (1971). In a multidisciplinary approach on the origin of Japanese civilization (1956), the French linguist Haguenauer evaluates the comparative linguistic materials that have been suggested with regard to Japanese and Altaic in the century separating Boller from his time. Perhaps due to the massive scope of the problems that it handles, Haguenauer’s work at times turns a blind eye to the strict application of the comparative method. Important to mention among the Japanese scholars who make contributions to the field is the linguist Murayama Shichiro@. Trained in Berlin, Murayama is familiar with the western methods of comparative linguistics and being Japanese, he has good access to the data. In a series of works beginning in the fifties (1950, 1957, 1962 a., 1963, 1966, 1975, 1976 a), he conducts morphological and lexical research relating Japanese to the Altaic languages. Although in his previous work Murayama tries to solve the problem of the affinity of Japanese purely by means of Altaic linguistics, by the end of his career (1974 a, 1976 b,1978, 1988, 1995, Murayama and O¤bayashi 1973) he subscribes to a mixed-language hypothesis. His theory is based on Polivanov’s view that Japanese is a hybrid language built on both Austronesian and Altaic elements. Although Murayama uses the term ‘substratum’, what he really has in mind is a mixed language consisting of elements originating 4

von Siebold 1832, 240: “Daar ik nogtans het voornemen heb, de etymologische overeen-komsten, niet alleen tusschen de Japansche en Mantschoesche taal, maar ook der aan Japan grenzende landen, Korea en Jezo, bij mijne onderzoekingen naar den oorsprong der Japanners te raadplegen, vermits dezen met nog eenige andere Tartaarsche volksstammen, als het ware, tusschen de beide eerstgemelden ingevoegd, de sprekendste bewijzen voor of tegen mijne stelling kunnen opleveren, zo geloof ik mijne lezers niet te vermoeijen, met hier deze idiomen naast elkander te plaatsen, waardoor niet alleen hun onderlinge naauwe verwantschap, maar waarschijnlijk ook die met de tongvallen van de nieuwe wereld blijken zal.”

22

Chapter 2. The history of the question

from two different language families, Altaic and Austronesian. It is concerned with exactly the kind of ‘mixed language’ that Meillet in 1925 declares not to have found, but which recently is regaining popularity (Munske 1986, Bakker and Mous 1994).5 Illustrating his theory with the example of Russenorsk, a so-called mixed language, Kortlandt (2000) demonstrates that the concept of mixed language is a misconception for language shift through imperfect learning. Another Japanese linguist active in the comparative field is Ozawa Shigeo. In 1968 he brings together approximately 230 lexical comparisons between Japanese and Mongolian. The study relies on a binary comparison. Each comparison is treated separately and no attempt is made to establish systematic and recurrent sound correspondences. However, a number of Ozawa’s look-alikes survive as cognates when carefully examined. In 1971 Miller puts the cat among the pigeons with his monograph ‘Japanese and the other Altaic languages’, in which he offers a set of sound correspondences, supporting the case for the Altaic origin of Japanese. It is an attempt to link Japanese and Korean cognates to the proto-Altaic forms reconstructed by Poppe (1960). Miller dedicates his further career to elaborating on the Altaic origin of Japanese (1980, 1991, 1993 c,1996 a), including phonological problems (1971 b, 1984 a, 1987, 1996 b), and morphology such as studies on case suffixes ( 1989 a, 1993 a 1993 b) and derivative verb morphology (1981, 1982, 1985 a.). Miller is experienced in other fields such as Indo-European and Sino-Tibetan and has a good access to the Japanese data. His publications read like novels, but his harsh criticism of fellow-linguists has closed many doors to further communication. And above all, it is exactly communication and cooperation among the scholars concerned which is indispensable for progress in the Altaic field. Miller’s cooperation with John Street in order to investigate Altaic elements in Old Japanese results in 1975 in a factual refinement of the proposed etymologies. In 1977 Street writes the second part of this work, but unfortunately, just like the first part, it is never published and it is still circulating as a draft version. The year 1975 also sees the publication of ‘Japanisch und Altajisch’ in which Karl Menges concentrates on Miller’s arguments. During his academic career Menges makes important contributions to several problems of Tungusic grammar (1943, 1968 a, 1968 b, 1968 c, 1978). In his attempt to relate Japanese to Altaic, Menges pays special attention to possible Tungusic cognates. A new stage in comparative Altaic studies including Japanese is represented by Sergei Starostin’s 1991 monograph on the Altaic problem and the origin of the Japanese language. Starostin contributes hundreds of lexical comparisons of Turkic, Mongolic, Tungusic, Korean and Japanese not available in the earlier literature. In cooperation with a team of scholars, among whom the Altaicist Anna Dybo and the Turkologist Oleg Mudrak, an Altaic database is made accessible via the Internet. The further accumulation of etymologies in the database results in the monumental three volumes ‘Etymological Dictionary of the Altaic Languages’ published in 2003. The dictionary presents 2800 Altaic etymologies and opens many new data for further consideration, refinement or refutation. 5

Meillet 1925, 82-83: “Dans tous les cas observés jusqu’ à présent, il y a une tradition continue d’une langue. ....Si l’on a pu arriver à faire, par la comparaison, l’ histoire de quelques langues, c’est que l’on était sûr que chaque système nouveau devait s’expliquer en partant d’un système unique.”

2.2. The Japanese-Altaic question

23

In the eighties and the nineties the Japanese linguist Yoshizo Itabashi publishes a series of articles on case suffixes and personal pronouns, relating Japanese to Korean and Altaic (1988, 1989, 1990, 1991, 1993). Recently (2003) he has shifted his attention to a possible hybrid character of Japanese and he now argues that proto-Japanese was shaped by mixing Austronesian and Altaic elements. Vovin (1994 a / b) argues against Benedict’s (1990) Austro-Tai hypothesis that relates proto-Japanese to proto-Austronesian. The basis for the rejection of Benedict’s proposal is the availability of more convincing Altaic etymologies within the basic vocabulary than the Austronesian etymologies presented by Benedict. The set of Altaic sound correspondences and the Altaic etymologies for Japanese that Vovin proposes follow the proposals in Starostin 1991. Alexander Vovin also contributes original lexical and morphological evidence relating Japanese to Altaic (1994 b, 1997 c, 1997 b, 1998, 2000). In coauthorship with Manaster Ramer and others Vovin writes an article on shared body part vocabulary in Japanese and Altaic (1997). In addition he adds to the problems of classification and reconstruction of Tungusic languages (1993 b, 1997 a) and Ainu (1993 e). Recently Vovin (2003 b) has completely changed his view and now he takes the position that the riddle of the origin of the Japanese language has not yet been solved. At present, he denies the availability of basic vocabulary relating Japanese to Altaic, reopening the discussion of Benedict’s arguments. An important opponent of the theory that Japanese is genetically related to Altaic is the Finnish linguist Juha Janhunen. Relying on a profound command of many languages spoken in Northeast Asia, Janhunen points out some crucial problems that confront us when positing the Altaic affiliation of Japanese (1992, 1994). He suggests that Japanese and Korean areally belong to the context of the Manchurian linguistic region. In a genetic sense, he agrees that Japanese and Ryukyu form a small family and that the Old Koguryo# language, once spoken on the Korean peninsula, is a close historical relative of Japanese. Janhunen finds himself in the company of the American linguist Unger and of the ‘Altaic panel’ further consisting of Martin, Whitman, Austerlitz, and Clark. The panel is cited in footnote 3 above, considering the case for Altaic premature and urging researchers to concentrate instead on Macro-Tungusic, which consists of Tungusic, Japanese and Korean. As for the language of the former Ryukyu kingdom, Ryukyu or Luchuan, it is commonly agreed that there is a genetic relationship with Japanese. However, in the past there has been a debate about how Ryukyu and Japanese are related. Some linguists consider Ryukyu as an independent sister language of Japanese (Chamberlain 1885; Miller 1971), but today there seems to be a broad consensus that Ryukyu is a Japanese dialect. Serafim (1985) has made considerable contributions to historical Ryukyu dialectology. As for the genetic relationship of Ainu, there is no final answer to the question available in linguistic literature today. Looking for the origins of the Ainu language scholarship in the past has pointed to Altaic and Japanese. Notable in this context is James Patrie’s 1982 monograph, which presents 140 Ainu lexical items as having Altaic origins, but it is not without problems in its phonological correspondences. Vovin (1993) proposes that Ainu is a language of the southern type with northern strata, one of which is Japanese. He contributes an extensive list of Japanese loanwords in Ainu. It seems reasonable to argue that in spite of their geographic proximity, Japanese and Ainu are not genetically related and the similarities that are found between them are to be attributed to an early loan relationship. On the basis

24

Chapter 2. The history of the question

of parallels in the basic vocabulary, Vovin puts forward the daring hypothesis that Proto-Ainu may be related to Proto-Austroasiatic, but he does not reach a satisfactory conclusion. It is also argued that Ainu is an Austronesian language by linguists like Sternberg (1933), Dempwolf (1938), Dahl (1977), and Murayama (1992), but their arguments have not won general acceptance among linguistic scholarship. In terms of genetic classification, it seems safer for the moment to categorize Ainu as a language-isolate. 2.3. The Japanese-Korean question Until the eighties research tends to be macro-level comparisons of Japanese with Altaic or Austronesian, but the last couple of decades have seen a third line of development in the question of Japanese affiliation. Recently research focusing on binary comparison between Old Japanese and Middle Korean has increased. A major problem with these micro-level studies is that they are binary, i.e. making individual paired comparisons. The weakness of comparing two, single languages instead of a group of languages lies in the quality of the resulting evidence. Correspondences that are found on the basis of binary comparison are more likely to be attributed to chance similarity. In a situation where features occur independently from each other, the probability that a feature in language A cooccurs with a feature in language B by pure coincidence is the chance that it occurs in language A multiplied by the chance of occurrence in language B. Introducing a third language C that shares the same feature, the probability that the similarity is due to sheer chance is further multiplied by the chance of occurrence of the feature in language C. Since chance probabilities always range between 0 and 1, the more we multiply, the less convincing the sheer chance explanation becomes. If the resemblances found in two single languages are not coincidental but due to a certain relationship, the introduction of more languages can still prove relevant to the further evaluation of the correspondences. If a paired comparison turns out to be completely fruitless, evidence for relating the two languages may be in only one of them, or in neither one of them, but found elsewhere within a larger stock. Macro-level comparison is more synoptic, in a way that it provides us with overview. It is like inspecting two similar objects with the naked eye or scanning them through a microscope. Focusing on some microscopic similarities is pointless without prior inspection with the naked eye since your eyes can help you to understand the totality of the objects considered. In a similar way one can go back and forth between micro-level linguistic comparisons and macro-level inspection. But, on the other hand, when comparing two languages pairwise, one never chooses the languages completely at random. When looking for potential cognates between two languages, there is a reason to suspect relatedness. In the case of Japanese and Korean this reason is geographical closeness and the archaeological record evidencing pre-historical contacts. From the viewpoint of geographical and historical proximity Japanese and Korean are more likely to be related to each other than to any other adjacent language. Also given the increasing complexity of the data when more languages become involved, binary comparison is often conducted on safer grounds than macro-level approaches that sometimes do more to obscure than to elucidate linguistic history. This is why many linguists today direct their attention to possible ties between Japanese and Korean, rather than focusing on Altaic as a whole. The suggestion of a common origin

2.3. The Japanese-Korean question

25

of the two languages is of long standing. As early as the 18th century the Confucian scholar Arai Hakuseki (1717) puzzles over the similarity between Japanese and Korean. The archaeologist Fuji Teikan (1781) concludes that the origin of Japanese was Korean, but his arguments are linguistically naive. About a century later these ideas are further developed, most notably by Aston in 1879 in ‘A Comparative Study of the Japanese and Korean Languages’, which contains an attempt to organize the superficial resemblances between approximately seventy lexical items according to regular consonant correspondences. Around the turn of the century Japanese scholars take over the helm again with Shiratori Kurakichi’s publication (1898) on the comparison of Old Japanese and Korean, attempting to apply the comparative method on approximately 200 lexical comparisons. 1910, the year of the Japanese annexation of Korea, sees the publication of ‘The Common Origin of the Japanese and Korean Languages’ in which Kanazawa Sho@saburo@ describes Korean as ‘a branch of Japanese’, a suggestion that is partly politically motivated and practically the opposite extreme of Fuji’s earlier claims. The above mentioned Finnish scholar Ramstedt adheres to a similar theory as do his Japanese colleagues Kanazawa and Shiratori as far as the genetic connections between Japanese and Korean are concerned (1926). A breakthrough takes place in 1966 when Martin offers 320 etymologies relating Japanese to Korean on the basis of regular sound correspondences. Martin is able to reconstruct forms from proto-Korean-Japanese, relying on regular sound correspondences. After approximately two centuries of speculation on the common origins of Japanese and Korean, he is the first to provide a solid foundation of systematic and recurrent correspondences. Martin compares Middle-Korean with Modern Japanese, thus missing some valuable information that can be derived from Old Japanese phonology, in particular Old Japanese vocalism. Later, however, he continues his study with a thorough reconstruction of protoJapanese (1987). In his 1985 dissertation Whitman further refines Martin’s contribution, taking into account the Old Japanese vowels of supposed diphthongal origin i2, o1, e 1 and e 2 for which he proposes a set of vowel correspondences presuming a process of medial *r and medial *m loss in the ancestral language. In this way Whitman provides us with a better understanding of the phonological assumptions underlying the sound correspondences established earlier. It is also suggested that Japanese stands in a close connection to the now extinct language of Koguryo#. Today Korean is for all practical purposes a single, unified language, but Murayama (1962 b), Yi Ki Mun (1963, 1972, 1976, 1977) and, recently, Beckwith (2000, 2004) have shown that it has not always been this way. At one time in history there were two groups of languages spoken on the Korean Peninsula, a northern Puyo# group and a southern Han group. Both languages were related, but still significantly different. According to Murayama, Yi Ki Mun and Beckwith the language of Koguryo#, which belongs to the Puyo# group, is remarkably close to Old Japanese, even closer than it is to Middle-Korean. The Silla language that belongs to the Han-group is thought to be the immediate ancestor of Middle Korean. Murayama, Yi Ki Mun and Beckwith derive lexical items from fragmentary data of the Koguryo# language for which they postulate correspondences with Old-Japanese. In order to further elucidate the relationship of Japanese and Korean to Koguryo#, a better insight is needed in the earlier stages of the Korean language. Although the data are poor and scattered, the problem is by no means insoluble. We gained a better understanding of the pre-hanku#l material and of Korean dialects through the research of Sasse (1976, 1982,

26

Chapter 2. The history of the question

1985), Lewin (1980, 1985), Ramsey (1978, 1979 a,1986, 1991, 1993), Bentley (2000) and Miyake (2000). Building on these works the task of clarifying the nature of the relationship between Japanese and Korean can go on. 2.4. The present question Is Japanese an Altaic language? The origin and classification of the Japanese language is one of the hot spots of historical linguistics today. When consulting linguistic encyclopedias and introductory texts on language classification the answers to the question are extremely diverse, ranging from an absolute ‘no’ to an agnostic attitude to an exclamatory ‘yes!’. There are hardly any bounds to the wide range of opinions on the subject. Japanese today still is a challenge to historical linguists. To illustrate, I here refer to the different points of view that one meets when crisscrossing linguistic literature on the subject. Trask 2000, 16: “Many people have tried to relate Japanese, Korean and Ainu to Altaic; at present there is growing support for the inclusion of the first two, though not for Ainu, but no consensus has been achieved.” Lee and Ramsey 2000, 5: “Like the ancestry of the Korean people themselves, the origins of the Korean language have never been definitely established.” Ibidem, 7: “The probability that Japanese belongs to the Altaic family is believed to be somewhat less than that of Korean.” Ho-min Sohn 1999, 22: “Thus, the Altaic origin of Korean and Japanese is a generally accepted hypothesis, although the hypothesis must be further refined and verified.” Lyovin 1997, 114: “The addition of Korean, Japanese, Okinawan and Ainu languages to the ‘Altaic’ family is even more controversial because even some supporters of the Altaic hypothesis do not accept the evidence adduced in support of the inclusion of these languages into Altaic.” The Concise Oxford Dictionary of Linguistics 1997, 15: “Altaic. A proposed family of languages, including Turkic, Mongolian, and Tungusic in east Siberia.” ibidem,190: “Japanese. Beyond this [Ryukyu], Japanese has no secure genetic relationship in any wider family.” Routledge Dictionary of Language and Linguistics 1996, 18: “Altaic. The inclusion of Korean, Japanese and Ainu into this group, ... is debated.” Ibidem, 245: “Japanese. Its genetic affinity is unclear; a relationship with Korean and the Altaic languages as well as with Malayo-Polynesian is often suggested.” Beekes 1995, 7: “Altaic consists of five groups: (a) The Turkish languages in West and Central Asia, such as Uzbek and Tatar. ...(b) About ten Mongolian languages. (c) The Manchu-Tungusic languages of Eastern Siberia. ... (d) Korean, and (e) Japanese. Whether or not the last two indeed belong with the others is a question still under discussion, but it seems now to be probable that they do.” Encyclopedia of Language and Linguistics 1994, 82: “Altaic languages. The Altaic language family comprises the Turkic, Mongolian, Tungusic and Korean branches.” Ibidem, 1810: “Japanese. While attempts to relate Japanese to the Altaic family have been most systematic and perhaps

2.4. The present question

27

most persuasive, views that are attracting increasing attention in recent years are: (a) those that consider Japanese to consist of an Austronesian substratum and an Altaic superstratum, and (b) those that regard Japanese as an Austronesian-Altaic hybrid or mixed language,... . ... On the other hand, more systematic correspondences have been attempted between Japanese and Korean, but even here reliable sets of cognates are extremely small in number.” Nichols 1992, 4: “When the cognates proved not to be valid, Altaic was abandoned, and the received view is now that Turkic, Mongolian, and Tungusic are unrelated (Unger 1990).” Ruhlen 1991, 130: “The affiliation of Korean with this complex [Altaic] has also gained considerable support, and Japanese appears to be gaining in popularity as a member of Altaic, though still in a minority position.” Shibatani 1990, 118: “Thus, while most people feel that Japanese and Korean are related and that these two languages are related to the Altaic languages, no conclusive evidence has been presented either for such connections or for others.” Comrie 1990, 856: “Like Korean, its geographical neighbor, Japanese has long been the target of attempts to establish a genetic relationship between it and other languages and language families. Hypotheses have been presented assigning Japanese to virtually all major language families: Altaic, Austronesian, Sino-Tibetan, Indo-European and Dravidian. The most persuasive is the Altaic theory, but even here evidence is hardly as convincing as that which relates the languages of the Indo-European family, as can been seen in ongoing speculations among both scholars and linguistic amateurs.” Lewin 1989, 114: “Zur exakteren Klärung der genetischen Zusammenhänge bleibt noch viel zu tun; allein schon die kritische Sammlung der bisher gegebenen Etymologien japanischer Wörter über den unzureichenden Versuch von Yamanaka (1976) hinaus bleibt ein dringendes Desiderat. Doch mehr als vor 150 Jahren gilt das Wort von Siebold: “De Japansche taal staat niet eenzaam meer onder de Aziatische idiomen.” (1832, 273)” Miller 1980, 46: “In other words, it is in terms of these two basic assumptions of the methodology of linguistic comparison that we are able to explain what Western scholarship has in mind when it arrives at its present consensus concerning the Altaic origins of the Japanese language.” Kodansha 1983, 19: “Japanese and the Altaic languages. Still there is little question but that Japanese has its full share of such early loans from Altaic, some of them entering the language as late as the Kofun period (ca. AD 300-710); but these forms are over and above the Japanese inherited stock of words (lexicon), word formations (morphology), and syntactic patterns that are genetically inherited from proto-Altaic. (Roy Andrew Miller)”

Given the lack of consensus as illustrated above, it is useful to identify the causes for the present disagreement in the field. In my view it is possible to determine a number of obstacles in the area of the historical linguistic comparison of Japanese with Korean and Altaic. The problems can be described as difficult accessibility, individualism, a partisan approach, and lack of overview. The historical comparison of Japanese, Korean, Tungusic, Mongolic and Turkic does not only cover a vast geographical area stretching from the Pacific Ocean in the east to the

28

Chapter 2. The history of the question

Black Sea in the west, it also requires a far-reaching linguistic knowledge. Tungusic, Mongolic and Turkic are language families that consist of many different languages and even if Japanese and Korean are both treated as single languages today, it has not always been this way in the past. Besides, studying the history of the individual languages many writing systems are involved. Apart from access to the languages themselves the historical linguist also needs profound knowledge of the comparative methodology and an insight in the mechanisms of language change in general. A considerable portion of the secondary literature in the Altaic field is written in German or in Russian. Given all these aspects of the study of Korean, Japanese and Altaic, there is a relative inaccessibility awaiting practically every historical linguist who is so bold as to approach the problem. Instead of attacking fellow historical linguists on their individual weak points in dealing with the available data, we should rather encourage them to overcome their relative inaccessibility and attempt to work out a solution by uniting forces. Uniting the intellectual forces brings us to the second problem in the Japanese-Altaic field, a problem which I have called individualism. Although communication and cooperation among Altaic linguists is indispensable for progress in the field, this is exactly what it lacks most. From the bulk of the etymological proposals in the appended index, it can be seen that over the years Japanese-Altaic etymologies have been proposed extensively. But skimming through the large amounts of data, it becomes clear that they hardly converge on any correspondence. Every Altaicist seems to create his own Altaic. Defending etymologies from feedback as if they are some untouchable property and refusing to alter ones own proposals by the slightest diacritic is not a healthy attitude to take in a dynamic intellectual process like etymological research. It goes without saying that Indo-European and other well established families have been revised continually over the years and the work goes on even today. Revising, evaluating and attempting to falsify etymologies does not invalidate a theory but rather strengthens it. Modern media, such as electronic mail and the Internet have a role to play here. The Altaic database project of the team around the Russian linguist Starostin provides a schoolbook example of modern intellectual exchange.6 Being able to consult their database electronically and privileged to attend some of the Altaic sessions in Moscow, I have been impressed by the open scholarship, that makes it free for anyone to enter the debate. Identifying causes for the lack of consensus, a third problem relates more to the ‘family’ of Altaicists than to the Altaic family. It lies in the partisan fashion in which the matter is sometimes addressed. Admitting that they do add some juice to the Altaic debate, slogans like “And then,...,it’s Miller time!” 7 and “...he [Martin] attempts to cover his tracks ... hoping that the reader will not know Japanese and hence not see through this transparent deception.”8 should better be avoided in future research because they do not match the high standards that we expect from the authorities in question. That numerous scholars ‘feel’ or ‘believe’ that Japanese is related to Korean and Altaic -or not- is rather polemic. When it concerns scholarly matters ‘feeling’ or ‘believing’ is not the issue. Positing or falsifying the Altaic hypothesis for Japanese is a matter of ‘argument’ rather than a matter of ‘belief’. 6

http:// starling.rinet.ru

7

Martin 1996 ,94, note 47.

8

Miller 1996, 204.

2.4. The present question

29

But even when approaching the problem with a serious, academic attitude, it is not always easy to get an overview on the arguments that have been made in the past. The obstacle is that the evidence is scattered around in linguistic literature. And as the materials are accumulating, we risk losing track of who-said-what-where. Looking back on decennia of ongoing discussions, it is useful to gather all the piles of data and investigate whether -when put together- they indeed form the burden of evidence sometimes referred to. It is also important to test the accuracy of such often heard complaints that Altaic lacks shared morphology or has no basic vocabulary in common. The objective of the present work is to present a state of the art for the genetic relationship between Japanese, Korean, Tungusic, Mongolic and Turkic. The approach can be described as empirical in the sense that the material, the body of evidence that authors advocated in the past increasing the Altaic family with Japanese and Korean, or interrelating the latter languages, is at the core of my study. The etymological proposals for lexical and morphological items are listed in the attached etymological index. Describing my objective as a state of the art, I do not mean to present a mere duplication of the existing discussions around this topic. In the textual part of this book, I wish to adopt an approach that is focused on the evaluation of the evidence, as opposed to the citation of evidence in the etymological index. This approach is meant to be original in its overall evaluation of the materials proposed so far. I am convinced that a holistic approach constitutes the best hope for clarifying the nature of the historical relation of Japanese to Korean and Altaic. Some solutions may have been given, but the problem is to lift them out of the bulk of etymological proposals and link them to other solutions within a clear structure. Other proposals may only be superficial solutions and prove incompatible within the overall line of reasoning. Of course it is easy to acknowledge people’s accomplishments in the Japanese comparative field, but it is a more daring task to point out the difficulties that confront us when putting the ideas together. The task before me is not simply to take the work of individual scholars in the Japanese-Altaic field as an unalterable whole and render a simple verdict of right or wrong. Rather, applying a set of severe criteria, I intend to separate the stronger etymologies from the weaker ones. Working to point out the individual inconsistencies and arguing for or against language relationship with a healthy skepticism will prove to be a much more profitable approach than abandoning the entire enterprise a priori. A criticism that can be made is that this is too broad a topic for a virgin excursion in the field. However, it requires a mind as pure and impartial as a virgin to approach this problem. The question of cognation between Japanese and Altaic is usually addressed in a partisan fashion. The polarization into a pro-Altaic and anti-Altaic front is an impediment to progress in Japanese comparative studies. My intention is to take a detached view of the matter, not joined to either side. With an impartial mind I do not wish to imply that I am uninterested or indifferent to the outcome of the problem. Rather, the question whether Japanese is Altaic or not is truly intriguing and it represents a serious challenge, but I do not have anything to gain or to lose by the exact outcome of the question. Hence, I will attempt to approach this question with proper scholarly detachment, sifting and evaluating facts rather than defending one position or the other. After all, historical linguistic scholarship is serving a common goal. That goal is to find out about what happened in the history and pre-history of languages.

Chapter 3. Interdisciplinary research Language, culture and ethnic traits do not come in package-deals: languages in particular can move across cultural and ethnic boundaries. But in setting up a working hypothesis about the spread of the Altaic languages on the Korean Peninsula and the Japanese Islands, the language shift on the Japanese Islands and the spread of the new language, we must take into account whatever help we can find. Recent advances within archaeology, physical antropology and genetics provide us with a much richer set of tools to assess the contribution that linguistics can make to unraveling such aspects of human prehistory as population migration, identity, material culture and language. This chapter will provide a brief outline of the evidence from physical and molecular biology and archaeology concerning the roots of the Japanese people, their culture, and their language and it will attempt to show that a hypothesis postulating a genetic relationship between Japanese, Korean and Tungusic is consistent with the picture that is provided by those other disciplines. As Greenberg has pointed out, it should be realized that as an initial step one must always hypothesize a certain scenario of language classification, before testing the validity of the hypothesis by 1 means of the linguistic comparative method. 3.1. Archaeology In the postglacial period after 8000 BC, when the East-Asian littoral was reshaped by the rising sea levels, the Korean Peninsula and the Japanese Islands gradually emerged and were occupied by groups subsisting on hunting, gathering and fishing. They used ceramics, which argues for a considerable degree of sedentism, but initially practised no agriculture. Since there is no clear transition from a hunting and gathering society to an agricultural society, Barnes (1993) and Choe (1990) suggests to avoid the term Neolithic to describe the postglacial societies on Japan and Korea. The archaeological periods are known by their ceramic names, the Chulmun period (6000 - 700 BC) in Korea and the Jo@mon period (10000 - 300 BC) in Japan. On the Korean Peninsula the shift from the Chulmun period to a period that can be termed the Proto-Bronze Age is manifested around 1500 BC by a change in settlement 2 pattern, the emergence of Mumun pottery and the cist-burial. In contrast to Chulmun settlements, which occur at the sea’s edge or on river banks, Mumun villages are always located in upland or inland areas. This locational change reflects a change in the subsistence 1

2

Greenberg 1987, 6-7: “Those who have realized that as an initial step one must first choose some hypothesis in order to test it by regularity of correspondence maintain, then, that the comparative method is not a method of arriving at a classification, but a method of proving a classification already hypothised.” Barnes 1993, 160.

3.1. Archaeology

31

base away from fish, shellfish and plant foods to grain production. The current estimate is also that wet-rice technology spread from the Chinese Mainland to the Peninsula after the emergence of Mumun pottery, but before the introduction of bronze. Although there are no varieties of wild rice indigenous to the Peninsula, rice grains have been discovered from a few sites. Controversy marks the literature concerning the exact route of rice expansion to Korea, but it is most plausible that wet-rice technology must have passed on to the Korean Peninsula overland from the Shanghai Delta region of the Chinese Mainland. Both longgrained and short-grained rice were present in the Shanghai Delta region, but since the latter variety is more cold- resistant it was transmitted outwards. Studies of stone tool typologies and discoveries of harvesting tools and rice grains on the northern coast of the Shandong peninsula and the northwestern Korean Peninsula argue for a northern, land-based route of wet-rice diffusion.3 Yet, only rare sites in Korea produce evidence for animal husbandry being introduced into the Mumun way of life. According to Nelson there is evidence for domesticated pigs and dogs.4 The pottery is designated Mumun, meaning undecorated, different from the decorative technique with geometrically incised comb patterns which is characteristic of much Chulmun ceramics. The occurrence of large dolmens on the Korean Peninsula has recently been viewed as an indigenous Bronze Age (700 BC - 1 AD) development, perhaps in response to the alien phenomenon of cist burial. A dolmen is a megalithic construction standing above the ground and consisting of a large capstone with various substructures. A cist can be described as a burial facility embedded in the ground, constructed of stones to form the bottom, the edges and the cover of a box in which the corpse was laid. The tradition of cist-burial in Korea is recognized to be imported by Tungusic people from Siberia. Unlike the dolmen, often found on flat alluvial plains, cist-burials are found in association with a settlement. It has been suggested that cist-culture and dolmen-culture represent different ethnic groups.5 The original locational difference in settlement between the Chulmun and the Mumun population quickly dissolved in favor of universal Mumun occupation. It is not clear whether the Chulmun population was peacefully absorbed into or brutally converted to the Mumun culture. Excavations of weaponry such as polished stone swords, projectile points and maceheads in Mumun sites argue for the latter scenario. Little is known about the ethnicity -let alone the language- of the Chulmun population, but it seems likely that the Mumun people with their Siberian roots imposed their culture not only materially but also linguistically on the Chulmun population. After the resulting language shift a substratum of the original Chulmun language may have been left in the target language. The early period of Mumun culture from 1500 to 700 BC is thus characterized by a serious restructuring of subsistence pattern and material economy, that can probably be associated with immigrations of Tungusic-language speakers. This period is termed ProtoBronze Age because bronzes were not introduced until the 8th or 7th century BC.6 Around that time the unusually shaped Liaoning type of bronze dagger intruded from the Manchurian Basin into the Korean Peninsula. This dagger derives from the Upper Xiajiadian culture, 3 4 5 6

Choe 1990, 12; Barnes 1993, 168-170; Nelson 1993, 157-58. Nelson 1993, 146. Barnes 1993, 166-167. Nelson 1993, 147-52. Barnes 1993, 161.

32

Chapter 3. Interdisciplinary research

associated with the ethnic groups Rong, Shan-Rong, and Hu. The Rong ethnic group is thought to have consisted of Tungusic-language speakers. The bronze mirror tradition of the Peninsula grew also out of the Northern bronze complex. The gradual infiltration of bronzes into the Korean Peninsula and their derivation from Manchurian and steppe precedents have suggested to many archaeologists that they arrived with another influx of Tungusiclanguage speakers. Quantifying the scale of immigration of north-eastern groups into the Korean Peninsula is most difficult, but the archaeological data seem to suggest that several migratory waves of Tungusic-language speakers took place from 1500 BC through the ensuing millennium. If elements of the Altaic material and non-material culture became established on the Peninsula during the Korean Proto-Bronze and Bronze Age, they would then have been transmitted to Japan during the spread of rice agriculture. The Korean Peninsula is viewed as the source of wet-rice technology in Japan because many Peninsular artifacts, features and technologies from the late Bronze Age society were recovered in North-Kyushu. In association with the earliest Japanese paddy fields influences from the Peninsula are evidenced by pottery, stone and wooden agricultural tools, remains of domesticated pigs, ditched settlements and megalith burials. These paddy fields can be associated with the late Jo@mon period and therefore dated to the first half of the first millennium BC.7 Rice grains have been recovered from earlier sites, dating to the late second millennium BC, but actual rice cultivation is only documented with paddy field remains. The adoption of wet-rice initiated a transitional period that resulted in the shift of the Jo@mon period to the Yayoi period (300 BC - 300 AD), which can be seen as a synthesis of Korean Bronze Age elements and north Kyushuan Jo@mon culture. The agricultural Yayoi society is named after a site near Tokyo where pottery of this period, replacing the rope patterned Jo@mon ceramics, was first excavated. Hudson has persuasively proposed a model of Japanese ethnogenesis, which is based on agricultural colonization by Yayoi farmers, who immigrated from the Korean Peninsula.8 The continental immigrants did not arrive in a single group, but intermittently over a millennium. The transition from Jo@mon to Yayoi culture as happened on the Japanese Islands is reminiscent of the absorption of the Chulmun culture into the Mumun culture, which occurred about a millennium earlier on the Korean Peninsula. Just like in the case of the Chulmun people, it remains unclear what the ethnicity of the Jo@mon people exactly was and, there is no evidence that points in the direction of an Austronesian population.The common estimate based on the archaeological picture is that the Jo@mon language was not a single language, but that several languages were spoken on the Japanese islands. It is often suggested that the Ainu people are related to Jo@mon people. With North-Kyushu acting as an incubator for the spread of Yayoi culture, the new pottery and the agricultural way of life soon expanded throughout the western Islands. The continuation of shell-collecting and hunting alongside wet-rice cultivation and the occurrence of typical Jo@mon motifs on Yayoi pottery suggests a rapid but not instantaneous change in lifeways. Since the Jo@mon population probably was relatively big in proportion to the gradually immigrating speakers of the Yayoi target language and since the language shift is supposed to have occurred at a moderate speed, it is reasonable to expect a low to moderate substratum interference from the underlying Jo@mon languages. The imperfect learning of 7 8

Barnes 1993, 168. Hudson 1999.

3.1. Archaeology

33

the target language would be expected to leave some phonological and maybe syntactical traces of the Jo@mon languages after the language shift. Since the change did not occur instantaneously, morphology and vocabulary were probably acquired quite accurately.9 So, the Jo@mon society on the western Islands probably was peacefully acculturated to Yayoi wet-rice cultivation. The situation in the northeast was different since Jo@mon societies there were committed to an efficient marine-based way of life and were reluctant to replace their fishing and shell-collecting habits by an agricultural subsistence base. The full adoption of rice culture came belatedly to the northeast and is associated with a second stage of rice diffusion in the Middle Yayoi period. After the adoption of rice culture the northeastern Yayoi culture continued to be different from the Southwestern Yayoi. Differences are found in crops, with millet and other grains playing an important role next to rice in the northeast, in pottery motifs, in the adoption of bronzes and in burial practices. The bronze implements used in the west by Yayoi people were never adopted in the northeast at all. In contrast to southwestern burial practices, which made use of jars, coffins or pits to inter an entire body, secondary jar burial was practised in the east. After removal of the flesh, only the bones of the deceased were buried in a jar, and several jars were put in the same pit. Based on these facts, it is assumed that northeastern Yayoi comprised a distinct, sociopolitically less complex subculture than southwestern Yayoi. Moreover, despite a common subsistence base, within southwestern Yayoi distinct cultures can be found on the eastern Seto and the western Seto regions. The distribution of bronze artifacts focuses on weapons in western Seto and bells in eastern Seto. This distinction between bronze repertoire echoes the different burial systems of the two areas. Eastern precincts are moated and are viewed as family burial grounds, containing burials of adults and children. In contrast to jar and cist burials of individuals in western Seto, which often contain many bronzes, grave goods are rare in the east. These differences argue for contrasting systems of organization and it seems appropriate to speak of different Yayoi cultures. The differences between the two regions are probably due to differential contact with continental societies. Western Seto had for at least a millennium shared in the culture of the Southern Korean Peninsula. In 108 BC, when Han Dynasty troops from the Chinese Mainland conquered the Northern Korean Peninsula, that region was incorporated into the Han system of military commanderies. The commanderies of Lelang and Taifang were controlling most of the northern Korean Peninsula and the south was divided in three areas known as the Samhan or the ‘Three Han’: Mahan, P’yo#nhan and Chinhan. It is commonly agreed on that at that time there were two groups of languages spoken on the Korean Peninsula: the Puyo# group in the north and the Han group in the Samhan in the South. Although Puyo# and Han languages were related they already were distinctly different at that time.10 This period of commanderies is referred to as the Proto-Three Kingdoms period (AD 1- 300), because it is the formative period for the three Korean kingdoms of Koguryo#, Paekche and Silla, which are considered to be the first Korean states. Another group of small polities, known as Kaya, never became a true state, but existed alongside the three kingdoms. The subsequent Three Kingdoms period (300-668) is characterized by monumental mounded tomb burials for the rulers. Only fragmental records are left from the languages of Paekche, Koguryo# and Silla, and the bulk 09 Thomason and Kaufman 1988, 38-45. 10 Yi, Ki-Mun. 1972, 1977; Sohn 1999, 37-44. Ramsey and Lee 2000, 273-280.

34

Chapter 3. Interdisciplinary research

of these fragments are from Silla. The information we have about the languages of Paekche and Koguryo# mainly comes from deciphering place names. It is assumed that the extinct language of Koguryo# belonged to the Puyo# group. In 313 Koguryo# was strong enough to demolish Lelang, though it did not move its capital westwards to the P’yo#ngyang river plain until 427. Pushing further southwards it came into conflict with Paekche, which was forced to retreat southwestwards, shifting its capital south of Seoul to Ungjin, modern Kongju, in 475. The lexical items preserved from Paekche toponyms are even smaller in number than the Koguryo# words. The government of Paekche is thought to have been imposed by northern invaders. However the Mahan population of the southwest was already organized in a rather complex society by that time. Even if the Paekche aristocracy spoke a Puyo# language, it seems unlikely for a language shift to have occurred given the scenario that the power was taken over abruptly by a relatively small elite, that was not bringing any revolutionary innovations along. The most plausible linguistic scenario matching the historical data would be that the Mahan population maintained their language, but borrowed rather extensively from the Puyo# language of the ruling class for reasons of prestige. This scenario seems to be confirmed by the Chinese dynastic record Zhou Shu, which mentions that the Paekche ruling class spoke a language different from the commoners. More recently a study of Paekche words preserved in the Nihon Shoki has shown that the Paekche language is closely related to the Silla language.11 By 668, the three kingdoms became united under the rule of unified Silla (668-935). Although Koguryo# was defeated by Silla, most of its territory was not incorporated into unified Silla, but became the state of Parhae, founded by people of Koguryo# stock. In the tenth century Parhae was conquered by the Khitan people. Silla language is believed to be the direct ancestor of Middle Korean and thus of the language spoken in Korea today. Although the capital was moved to former Koguryo# territory in the 10th century, Koguryo# substratum interference in the Korean language is expected to be limited by that time. The Chinese recordings of the Han and Wei dynasties, the Houhanshu and the Weizhi describe a political organization of Wa on the western Japanese Islands that was sending envoys to the Chinese court via Lelang and Taifang from the first to the third century AD. Wa consisted of many small polities, with several under the hegemonic rule of a country called Yamato. For the late Yayoi period there are indications of Wa-disturbance or increasing strife, accompanied with the construction of mound-burials. These mounded tombs gave their name to the ensuing Kofun period (300-710). The Japanese Kofun period coexisted with the Korean Three Kingdoms period (300-668) and interactions between the mounded tomb cultures of Koguryo#, Paekche and Shilla on the Korean Peninsula and Yamato in the western Japanese Islands are documented. The state of Paekche in particular, known as the most refined of the Three Kingdoms, enjoyed diplomatic, scholarly and artistic interaction with Yamato. Egami and Ledyard’s ‘Horserider Theory’ suggests that there was an abrupt conquest of Yamato by powerful horse-riding people, who directly invaded Japan from the south Korean coast and who subjugated the people of Wa.12 But archaeologists like Barnes, Hudson and Edwards have persuasively argued that there were continuous contacts between Yamato and people on the Korean peninsula during the Late Yayoi and Kofun periods. 11 Bentley 2000. 12 Egami 1964; Ledyard 1975.

3.1. Archaeology

35

Although these contacts almost certainly had a military gloss, there is no reason to believe there was a single or sudden invasion with an abrupt impact on the Japanese culture and language. Barnes rejects the Horserider Theory of Japanese state formation because archaeological evidence such as horse trappings occurs too late in the archaeological sequence.13 The present view is that the migrations causing historical discontinuity occurred much earlier, during the transition of Jo@mon to Yayoi, and that it is not necessary to caricature these migrations as violent movements of horse-riding people. The immigrants were peaceful Peninsular farmers, who can better be compared to pig-raisers than to horseriders. The cultural continuity between Yayoi and Kofun that is suggested in the BarnesEdwards-Hudson theory is likely to imply a linguistic continuity between both historical periods, although the Peninsular languages belonging to either the Puyo# or Han language group may have heavily affected the language of the dominant proto-Japanese speech community in terms of extensive borrowing.14 3.2. Biology Recently the study of physical anthropology and molecular genetics has become more refined in a way that it can cast light on the population movements that played a role in the formation of the Japanese people. The Japanese anthropologist Hanihara (1991, 1992) published a ‘dual-structure hypothesis’, showing that the Jo@mon, Ainu and Ryukyu population is genetically remote from the population of the Yayoi period and the modern main-island Japanese. Biological data likewise reveal genetic affinities between the Jo@mon-Ainu -Ryukyu group and Southeast Asians on the one hand, and between the Yayoi-modern Japanese group and Tungusic speakers of Northeast Asia on the other. Osteological evidence comes from the measurement of skulls and the analysis of skeletal remains. The Yayoi type remains have a higher average height than the Jo@mon type. The Yayoi facial skeleton is higher and narrower, giving it a more rectangular and delicate appearance than the Jo@mon face. The Jo@mon dental pattern reflects an older type, called the Sundadont pattern which is associated with Southeast Asians, while the Yayoi teeth possess the Sinodont pattern and resemble those of the modern Japanese. So it is assumed that the taller, long-faced immigrants from the continent intermarried with the shorter, round-faced indigenous Jo@mon population, producing the modern mix of Japanese physical types. There is also an increasing amount of genetic evidence suggesting the separate origins of the Ainu -Ryukyu population and the main-island population. This evidence is based on the distribution of bloodgroups and serum proteins, the distribution of carriers of significant viruses such as T-cell leukemia and Hepatitis B and the study of mitochondrial DNA. Genetic research appears to be fairly consistent with the recent model of Yayoi agricultural immigration provided by archaeology. For an accessible overview on the evidence from biological anthropology I refer to Hudson’s 13 Barnes 1988, 19; Barnes 1993, 245. 14 Unger 2001, 96. Unger uses the term adstratum, but adstratum can rather be defined as mutual influence of two languages on each other over a period of time. Since Unger refers to extensive borrowing from the peninsular branch of pJK into Japanese, it is clear that he has a unilateral language interference in mind, which I prefer to name borrowing or superstratum interference.

36

Chapter 3. Interdisciplinary research

multidisciplinary approach on the question of Japanese ethnogenesis.15 For a broader comparison of biological and linguistic classifications, suggesting a close genetic connection between the speakers of Japanese and the population in North-East Asia I refer to the work of Cavalli-Sforza and his colleagues.16 3.3. The Austronesian connection The hypothesis of a linguistic connection between Japanese and Austronesian has a long history stretching back to the works by Polivanov (1924), Matsumoto (1928) and Murayama (1974 a, 1976 b,1978 a/b, 1988, 1995, Murayama and O¤bayashi 1973) and even today linguists continue to support a Japanese-Austronesian link under various configurations. The hypothesis that Japanese is a sister language of Austronesian became popular after the publication of Benedict’s book on Japanese and Austro-Tai in 1990. This model pictures Japanese as part of the Austro-Tai family, but it concentrates on relating proto-Japanese to proto-Austronesian. In his earlier work, Kawamoto (1977, 1978 a/b, 1979, 1982, 1984, 1985) proposes a similar theory for Japanese, restricting himself to the comparison of proto-Japanese and proto-Austronesian. A second model connecting Japanese to Austronesian is the hybrid or mixed language hypothesis, which pictures Japanese as a mixture of Austronesian and other, most commonly Altaic, elements. This viewpoint is taken by Polivanov and Murayama and continues in the work of Maher (1996), Sakiyama (1996, 1998, 2001) and in the more recent work of Kawamoto (1987) and Itabashi (2003). Finally, the third model is the hypothesis that Japanese is an Altaic language on an Austronesian substratum. The underlying idea is that of a language shift. If speakers of some daughter language of Austronesian shifted to an Altaic language, then the new language has only a single ancestor, Altaic. But through learning imperfections, some features of Austronesian could be left in the language. The difference with the mixed language hypothesis is that mixed languages suppose multiple ancestory, while in the case of language shift with substratum interference, there is only a single parent language. A number of Altaic linguists such as Poppe in his foreword to Miller 1971 and Starostin, Dybo and Mudrak (2003, 82) do not exclude the possibility of an Austronesian substratum in Japanese. There is no such thing as context-free historical linguistics. The problem with setting up a hypothesis of a linguistic connection between Japanese and Austronesian is that it is incompatible with the archeological and biological facts and that it goes against our general insights in the mechanisms of language change. Language shifts have cultural causes which should be visible in the archaeological record. Hudson (1996) finds no evidence that a substantial number of Austronesian speakers reached Japan in prehistory. Claims for a small-scale Jo@mon cultivation -as opposed to the full-scale Yayoi agriculture- go back as early as 4000 BC. Since the broadleaf evergreen forests of the west are thought to have been low in productivity, a variety of cultural traits, including the cultivation of millet, gourds, lacquer trees etc. are argued to have developed from contacts with Austronesian culture along with other southern elements including myths and customs and architectural 15 Hudson 1993, 59-81. 16 Cavalli-Sforza et al. 1988.

3.3. The Austronesian connection

37

styles .17 Hudson argues that these cultural traits should not necessarily spread in complexes and that they cannot support the influx of a serious number of Austronesian speakers. Archeologically the most likely scenario following Hudson is that the Jo@mon languages were derived from the languages of the Upper Paleolithic colonization of the Japanese Islands. This is confirmed by Hanihara’s representation of the dual structure model of the population history of the Japanese Islands in which the inhabitants with the southern genes go back to paleolithic times. Whether these pre-Jo@mon languages had an Austronesian connection or not, it is highly unlikely that they would still leave recoverable traces in the Japanese language more than 15000 years later. The first general linguistic objection to the Austronesian hypothesis is the position of Ainu. Biological and archeological findings suggest that Ainu may be derived from a Jo@mon language. If the Jo@mon languages indeed were Austronesian, as suggested in the Austronesian hypothesis, then Ainu, a descendant of a Jo@mon language is probably a sister language of Austronesian. If Japanese consists of Austronesian elements or if it has an Austronesian substratum, it is reasonable to expect that Austronesian is closer to Ainu then to Japanese. But in reality, while a number of studies have investigated linguistic connections between Ainu and Austronesian, the results are extremely poor. Perhaps this line of reasoning can explain why Murayama begins to investigate the Ainu-Austronesian problem by the end of his career, when he changes his views to a mixed-language hypothesis for Japanese. Just like Sternberg (1933), Dempwolf (1938) and Dahl (1977) did before, Murayama supports an Austronesian-Ainu link (1992, 1993), but the hypothesis has not won general acceptance among linguistic scholarship. In terms of genetic classification, it seems safer for the moment to categorize Ainu as a language-isolate. A second linguistic objection to the proposed Austronesian links is the theoretical concept of mixed languages. As a linguistic model, the idea that languages descend from a single ancestor is generally accepted. Also accepted is the idea that when speakers shift from one language to another language substratum interference may occur. But a model that continues to generate controversy amongst linguists is that of mixed languages, picturing one language as a mixture of elements coming from two or more different ancestors. Do mixed languages really exist or are they nothing but language shift with heavy substratum interference? A considerable number of linguists argue that there are no mixed languages. An even when linguists accept the notion of mixed languages, they stress that the sociolinguistic conditions are very specific, so that language mixing rarely occurs. Among the conditions are, first, a multilingual setting as opposed to a bilingual setting, second, no intention of the speakers to adopt a specific target language as a new language, and third, trade or work group situations among speaker groups of more or less equal status as opposed to prestige settings.18 In reality there is little in the archeological record that supports such a model for Japanese. Given a relatively large-scale immigration of Yayoi farmers with prestigeous agricultural knowledge, we would expect an intention to adopt the language of the new immigrants along with the innovating agricultural techniques.

17 Hudson 1996;1999, 104-06 argues that southern elements in Japanese culture such as the ones mentioned in Unger 2001, 84, 94 have not necessarily diffused together. 18 Thomason and Kaufman, 1988.

38

Chapter 3. Interdisciplinary research

From the controversial model of mixed languages this brings us to the socio-linguistic context. Mixed languages can perhaps be the outcome of language contact between socially more or less equal languages. But what happens when a language meets with a prestige language? The social context will be an important determinant of the linguistic outcome of language contact. If the speakers of a native language meet with the speakers of the prestige language through various cultural contacts, we would expect language maintenance with extensive borrowing. This is probably what happened when Japanese speakers met speakers of Shilla, Paekche and Koguryo# Korean in the Kofun period. But, if there is a moderate-scale to large-scale influx of people having a prestigeous culture and knowledge, we rather expect language shift. Depending on the social conditions in which the speakers of the native language learn the prestige language more or less mistakes will be present in the shifted language. The number of mistakes made in adopting the new language is proportional to the level of substratum interference. Given the minor cultural contribution from Jo@mon to Yayoi, it is reasonable to expect that Jo@mon speakers shifted to the Yayoi language with minor to moderate linguistic contribution of the Jo@mon language. Seriously doubting that the Jo@mon languages are related to Austronesian at all, only minor substratum interference from the underlying Jo@mon languages is what we expect for Japanese. Finally, there is the lexico-statistical observation that in binary comparisons we need exponentially more evidence in order to support a genetic hypothesis. In the Austronesian hypothesis proto-Japanese is pictured as a sister-language of proto-Austronesian. In the Altaic hypothesis proto-Japanese is a descendant of proto-Altaic, related to four sisterlanguages, proto-Korean, proto-Tungusic, proto-Mongolic and proto-Turkic. The binary correspondences that are found between proto-Japanese and proto-Austronesian are more likely to be attributed to chance similarity than the multiple correspondences of Altaic. In order to rule out chance, we must find exponentially more examples of a particular correspondence in the case of the Austronesian hypothesis. In reality, the number of etymologies proposed relating Japanese to Altaic exceeds 2000. The etymological index in appendix consists of 2055 Japanese entries along with various etymologies relating the entry to Korean or Altaic. On the other hand, the overall number of etymologies proposed in literature relating Japanese to Austronesian remains below 500. This statistical fact alone makes the Altaic hypothesis a more promising theory to test than the Austronesian hypothesis for Japanese. Linguistic hypotheses must be tested with linguistic tools alone. But before testing the validity of a certain classification, one must always set up the hypothesis. In setting up a hypothesis we are entitled to take into account any help we can find. From an archaeological, biological and general linguistic perspective setting up an Altaic hypothesis for Japanese constitutes a far better hope than setting up an Austronesian theory. 3.4. A working hypothesis If we postulate a language relationship between Japanese, Korean and Tungusic on the basis of the above evidence from material culture, biology and a general insight into the mechanics of language change, then it is conceivable to represent this relationship along the general lines of the following scheme.

3.4. A working hypothesis

39

Altaic Turkic

Mongolian

Macro-Tungusic

A

South-Tungusic

North-Tungusic Mumun = pJK

1500BC

Jomon languages

Mumun culture to Korea

S

Han

1000 BC

Japonic

Ainu ?

wet-rice to Kyushu

Puyo

700 BC

0 formation 3 kingdoms

300 AD

Koguryo

HanPaekche

B B B

S

600 AD

Silla

PuyoPaekche

S

mounded tombs

S

Chinhan Pyonhan

Mahan

bronze dagger to Korea

B S

protoJapanese

Ryukyu

B B

B

B

Old Korean

B

Old Japanese

Silla unification

B= borrowing = superstratum interference (a dominant language influences the native language) A= adstratum interference (mutual influence of two languages on each other over a period of time. e.g. Flemish and French in Belgium represent such a situation) S=substratum interference (influence of native language on the target language in a situation of language shift) Fig. 1 A working model for the relationship between Japanese, Korean and Tungusic on the basis of interdisciplinary evidence.

During the transition from the Chulmun period to the Bronze Age on the Korean Peninsula and from the Jo@mon to the Yayoi period on the Japanese islands, not only were there dramatic changes in subsistence technology, material culture, and the population genome, but language may also have been involved. In the period following the supposed Yayoi language shift the contacts between different people on the Korean Peninsula and the Japanese Islands were extremely involved, a situation that is expected to be reflected in the language through extensive borrowing.

40

Chapter 3. Interdisciplinary research

The level of complexity of the hypothetical tree-structure above raises a methodological question about the linear comparison of the 5 proto-languages involved in the reconstruction of proto-Altaic. Is it methodologically sound to compare proto-Japanese, proto-Korean, proto-Tungusic, proto-Mongolic and proto-Turkic, as if they were distinct linguistic unities spoken during a comparable period in time? I think that the answer to this question is ‘yes’. A proto-form is a diachronically invariable form reconstructed on the basis of synchronically and historically alternating, varying forms. We can reconstruct proto-Romance fairly accurately on the basis of contemporary Spanish and Italian and Old French alone, without intermediately reconstructing Ibero-Romance, Gallo-Romance, Italo-Dalmatian or BalkanRomance and subsequently West- and East-Romance. In the same way we do not need to know all participating members and the exact branching for Altaic a priori. It is only after having tested the Altaic hypothesis and after the establishment of sound correspondences that we can further refine our knowledge in terms of the branches and subbranches involved. This can only be done in a final stage on the basis of shared innovations. As an initial step it is possible to postulate an historical linguistic scenario based on archaeological and biological data, linguistic information from early Chinese dynastic records and on what is known about the nature of human language and linguistic change in general. The tree like linguistic scheme above is such a rough estimate for the linguistic evolution of Japanese and Korean in the prehistorical period. But these other disciplines can merely provide a working hypothesis to linguistic history. The exact timing and mechanics of language change cannot be recovered with archaeological or biological methods alone and thus, in the work ahead, we will have to rely on the methodology of historical linguistics. In order to test the validity of the above hypothesis about language shift, split, contact and death we will use the one and only tool we have at our disposal, namely the comparative method.

Chapter 4. Methodology 4.1. What is a genetic relationship? When we ask ourselves the question whether Japanese is related to Korean, Tungusic, Mongolic and Turkic or not, we want to know whether Japanese belongs to the same language family as Korean and the so-called Altaic languages. They belong to the same family if they go back to the same parent, in casu proto-Altaic. In this question ‘related’ means ‘relatable with the tools that we have at our disposal’.1 Chief among the methods we dispose of to demonstrate language relatedness is the comparative method. Languages are organized into language families. For all their current differences, some languages display similarities which make it possible to trace them back to a common source. Anybody who knows a few languages intuitively feels that some are closer to one another than are others. For example, Dutch and German are closer to one another than either is to French, while French and Italian are closer to one another than either is to Dutch. This notion of closeness can be formalized using the comparative method, which takes systematic and recurrent formal correspondences between languages as its data. The hypothesis underlying this method is that some similarities are too regular and frequent to be the result of mere chance. Once we have obtained a significant number of similarities, we can logically exclude coincidence as a an explanation of these shared features. In short, some similarities between languages are just too striking to be coincidental and must be attributed to their common ancestor. On the lexical level for instance, the verb for ‘to see’ is zien in Dutch, sehen in German and see in English. It becomes even more interesting on the morphological level, noticing that the past tense of these verbs is formed in a similar way: in Dutch hij zag, in German er sah, in English he saw. Although Dutch, German and English are now clearly different languages, we may hypothesize that at an earlier period in history they had a common ancestor in which *sehan was the verb for ‘to see’ and *sax corresponded to the past form. This common ancestor is proto-Germanic and it is not an attested language. Although we do not have attestations of the proto-language, it can be postulated since it is the most plausible hypothesis explaining the remarkable similarities among the languages involved. The convention is to mark the hypothetical proto-forms by a preceding asterix (*). Less obvious similarities can also be found between the Germanic languages and a number of other languages spoken in Europe. These languages share fewer similarities with the Germanic languages than individual Germanic languages share with each other. Therefore, we say that these languages are more remotely related, but still belong to an overall Indo-European family. We are able to reconstruct the common ancestor language protoIndo-European. In this way we obtain a family-tree model with proto-Indo-European at the top, and a number of branches such as Celtic, Germanic, Italic, Greek, Albanian, Balto-Slavic, 1

Anttila 1989, 320: ‘Related in linguistic terms means ‘relatable’.’

42

Chapter 4. Methodology

Armenian, Indo-Iranian, Anatolian and Tocharian, which are descendents of the protolanguage, but ancestors of the European languages spoken today. The tree model is the most widely used way of showing genetic relationships graphically. The tree model is not a theory or a method, it is a model, a simplified graphic representation visualizing the temporal development of languages. There are other models available in literature, such as the wave model, which focuses on the spatial arrangement of languages. One typical property of language change as represented by the family-tree model is that, as time goes by, genetically related languages tend to become more and more distant. This follows from the fact that once two languages have split off as separate languages from a common ancestor, each will innovate its own changes, probably different from the changes that take place in the other language. With the passage of time there will be a cumulative effect of increasing divergence. In the long run the divergence may be so great that it is no longer possible to tell that the two languages do in fact come from a common ancestor. It follows that when we say that languages are ‘unrelated’, we actually mean that they are ‘unrelatable within the methodology of comparative linguistics’. The reason for their unrelatability may be twofold. Either the languages did not descend from a common ancestor, or either they did descend from a common ancestor, but there is insufficient evidence left to demonstrate their common ancestory. The more divergent the innovations in both languages and the longer the passage of time the more the traces of genetic relatedness will be erased. For language families like Indo-European that have a long written tradition it is possible to make use of an earlier stage of the language, thus bridging a certain elapse of time. However, some languages in other parts of the world lack historical records that reach a serious time depth. The oldest Turkic texts, written in runic script, date from the eighth and ninth centuries. The first inscription written in Mongolian dates from the thirteenth century and the oldest text in Jurchen, belonging to the Tungusic family, dates from the fifteenth century. It is in the same period that the written history of Korean begins, although some fragments of earlier Korean, that are difficult to interpret, exist. The earliest systematic records of Japanese date back to the eight century. Hence, in case of Tungusic and Korean, where the written history is only five and a half centuries old, earlier traces of genetic relatedness in the languages may be erased. 4.2. What is a genetic argument? 4.2.1. A negative argument A genetic argument is a negative argument, what in classical logic is called a disjunctive syllogism.2 It means that our deduction, the process of reaching a conclusion about common ancestorship, works by elimination. One rules out all but one of the logically possible accounts of the similarities holding between the languages compared, so that only inheritance from a putative common ancestor remains. Thus, a genetic argument consists not only in the presentation of a set of similarities holding over the languages compared, it also consists in the demonstration that these similarities are not likely to be the result of nature, borrowing or chance. 2

Harrison 2003, 215.

4.2. Genetic argument

43

Applying this knowledge to the similarities holding between Japanese, Korean, Tungusic, Mongolic and Turkic, it is essential to set up a methodological framework for sifting the proposed evidence. Of course, not all etymological proposals that are listed in the appended index are valid ones. The sifting criteria that I intend to adopt in order to separate the stronger etymological proposals from the weaker ones are the following. First I omit etymologies in which the internal analysis of the individual proto-forms is in conflict with the external comparison. Second, I omit similarities that could be the result of nature or universal tendencies in the structuring of language. Third, I try to rule out borrowing as an explanation of the similarity sets. And finally, I set up semantic constraints for the comparison of the meanings. Indeterminate evidence As early as 1908 the Japanese linguist Fujioka Katsuji observed a number of structural characteristics shared by Japanese, Korean, Tungusic, Mongolic and Turkic. On the phonological level there are no liquids or consonant clusters in initial position and there is vowel harmony to a certain extent. On the morphological level we find suffixing morphology, using postpositions instead of prepositions and attaching particles. The interrogative is expressed with a sentence-final question particle. Japanese, Korean and the Altaic languages all lack articles and do not distinguish grammatical gender. Possesion is expressed by an existential verb ‘to be’ rather than by a possessive verb ‘to have’ and the ablative particle ‘from’ is used in comparative expressions. Syntax is characterized by an SOV pattern, with modifiers preceding heads and objects preceding verbs. Until today those shared characteristics are often taken as a starting point in discussions about the genetic affiliation of Japanese. And yet, however striking they may be, the kind of evidence that these structural similarities represent is indeterminate and hence insignificant as a genetic argument. Significant for the demonstration of a genetical relationship are those similarities in form and function that may be argued to result from something other than universal trends or by language contact and that are unlikely to be accounted for by chance. For some common features between languages it is extremely difficult -if not impossible- to decide whether they result from common ancestry, from contact, from general principles in linguistic structuring or from sheer coincidence. This is the case for structural similarity between languages. Structural similarities concern general phenomena and can be found on many levels: phonology, morphology, and syntax. A classical example of structural similarity that is induced by language contact is the Balkan linguistic area. Although Greek, Albanian, Serbo-Croatian, Bulgarian, Macedonian and Romanian all are Indo-European languages, they belong to different branches of that family. The salient traits of the Balkan linguistic area are phonological, such as the development of a phonemic central vowel [E] in relation to stress, morphological such as the syncretism of dative and genitive or syntactical such as the absence of infinitives and the use of verb forms instead.3 That some North American languages such as the languages of the Siouan family and Navaho, South American languages such as Quechua, languages spoken in Europe such as Basque, African languages such as Ijo, and languages spoken in Asia such as Burmese, Burushaski, Turkish, Japanese and Korean share a set of syntactic traits, one of which is an SOV pattern, is most probably due 3

Campbell 1998, 300-301.

44

Chapter 4. Methodology

to sheer coincidence. The languages of the world are numerous, which cannot be said of the ways to arrange words in a sentence.4 Cross-linguistically it is also perceivable that a high degree of structural similarity can occur due to universal principles of linguistic patterning. Some structural features tend to cluster. Greenberg has demonstrated that verb-final order and exclusively suffixing morphology cooccur.5 In the past structural similarities that were determined by geography or by linguistic universals have been mistaken to be indications of genetic relationship. Therefore, it is advisable to be extremely cautious with structural evidence. Structural similarity can eventually result from genetic inheritance, but it does not necessarily do so. Although Nichols has tried to sort out genetic, areal and universal determinants of linguistic structure in a monograph dealing with linguistic diversity, I will try in the work ahead to follow the traditional approach to historical linguistics, applying the classical comparative method.6 Therefore, I will separate structural evidence from lexical and morphological evidence and in my argumentation for or against genetic relationship I will not take into account any structural similarities. Borrowing Not only structural features can cross linguistic boundaries, lexical items, semantic associations and some morphological items can do so too. The most common way of borrowing, however, is when a language takes lexical items from another language and makes them part of the vocabulary. The loanwords are adapted in such a way that the foreign sounds are changed to conform to native phonology. When Japanese, which is syllabic in nature with the exception of [n] that can be followed by a consonant, for instance, borrows a word from English, the English consonant clusters are reshaped into open syllables. This is the case for the English word club which becomes kurabu in Japanese. The reasons for borrowing can be need or prestige. Since borrowings are not of any help in establishing a genetic relationship, it will be important to identify loanwords and leave them aside. Clues for detecting loanwords can come from the phonological or morphological pattern of a word, from its meaning or from our knowledge of the cultural, ecological and geographical context in which the word was introduced. The phonological pattern of native Japanese, for example is such that it does not allow an initial liquid phoneme and that it tends towards mono- and disyllabic roots. A word like raitomotiihu ‘leitmotiv’ can immediately be recognized as a loanword from German for obvious phonological reasons. Another indication that determines the direction of the borrowing as being from German into Japanese comes from morphology. The word is morphologically complex and can be derived from a verb leiten ‘to lead’ and a noun Motiv ‘motif’ in German, but it lacks this morphological analysis in Japanese. Besides German leiten has obvious Germanic cognates such as Dutch leiden and English lead, but no legitimate cognates exist for the Japanese word. A cultural clue is that Japanese writers have been influenced by the German literary tradition in the period following the Meiji restoration. If we compare the Japanese word mesu ‘surgical knife, scalpel’ to its Dutch counterpart mes ‘knife’, it is possible to determine the direction of the loan on semantic grounds, since the Japanese word is a case of semantic 4 5 6

Martin 1966, 186. Greenberg 1963. Nichols 1992.

4.2. Genetic argument

45

narrowing of the Dutch donor. Semantic loans are called calques or loan translations. In this case only the meaning of a form in the donor language is transferred to the recipient language. Japanese jido@sha ‘car’ for instance, is a literary translation of ‘automobile’ in the sense of ‘self moving vehicle’. Although cross-linguistically morphological borrowings are unlikely to occur, they do occur occasionally. The English word acknowledgment has a Germanic root but a borrowed Latinate suffix -ment. Nature and universals Some similarities between languages cannot be attributed to a common ancestor. They are the result of natural or universal tendencies in linguistic structuring. Observing that nursery terms like Eng. tits and J titi ‘breasts’, or Eng. papa and J haha, OJ papa ‘mother’ or sound symbolic words like Eng. zigzag and J gizagiza ‘notched’, or Eng. knock knock and J kon kon ‘knock knock’ are similar does not tell us anything about the common ancestor of English and Japanese. Being due to universals in language, mama-papa terms and sound symbolic words are generally regarded as poor diagnostics of genetic relatedness (Jakobson 1960). Nursery words like mama, papa, baba, dada, caca, pipi are cross-linguistically determined, with female forms showing a preference for nasals and male forms tending towards stops, but not exclusively so. Nursery forms can enter the adult language and develop to common forms meaning ‘mother’, ‘father’, ‘grandfather’, ‘grandmother’, ‘uncle’, ‘aunt’, ‘sister’, ‘brother’ etc. Such lexical items do not provide reliable support for genetical relationship and will thus be omitted from the evaluation of the evidence. Onomatopoeia are words that are coined through the imitation of sounds from nature, like Japanese chon ‘a clap’. Sometimes the association with the natural sound is so obvious for the speakers that it inhibits regular sound change to occur. When similarities between onomatopoeia are found among different languages, the underlying reason may be the fact that the languages have independently approximated the natural sound in a similar way, and not that the onomatopoeia share a common history. Even if in expressing a dog’s barking Dutch waf-waf is nearer to Japanese wan-wan than to English bow-wow, this tells us nothing about a possible kinship relationship between Dutch and Japanese. That is why it is safer to eliminate etymologies that are suspect of sound imitation from proposals of genetic relationship. Therefore, I have chosen to omit verbs expressing physical actions processing salient sounds like ‘laugh’, ‘sneeze’, ‘break’, nouns designing birds and frogs, and other words for which there is reason to suspect sound imitation, from my argumentation. Apart from attempts to imitate sounds, there are also sound symbolic words that describe the looks of a situation or one’s physical or emotional attitude towards a situation. Martin respectively uses the term phenomimes and psychomimes for these words.7 Japanese in particular tends to be very creative in developing mimetic lexemes, like gocha-gocha ‘confusion, disorder’ or muzu-muzu ‘be irritated, have an itch’. This kind of mimesis should be handled with the same caution as onomatopoeia. In omitting etymologies that are suspect to include nursery forms or onomatopoeia, I do not wish to imply that the etymology in question is totally invalid. It is of course possible that a certain lexical item indeed descends from the same ancestral form as the proposed 7

Martin 1988, 1025.

46

Chapter 4. Methodology

cognates in the other languages. The reason why I have chosen to eliminate possible nursery words and onomatopoeia is that the probability that we are dealing with a universally instead of a genetically determined similarity is too high. The approach that I wish to adopt in the following evaluation of the attached etymological proposals is such that I want to sift the evidence in a way that only strong etymologies pass the test. It has even been suggested that some grammatical and morphological similarities may be attributed to general principles in the phonological structuring. Nichols observes that pronoun root consonantism tends to be universally determined.8 There seems to be a general principle active that pushes personal pronouns within the statistic constraints of a fixed number of phonologically basic sounds. North-west Caucasian, for instance, has a very complex phonemic inventory, reaching about 100 consonants, but the pronominal system is based on only a handful of simple phonemes such as sa-, wa- and da-. By consequence it seems that comparing the pronouns, we could come across similarities that are rather due to universals than to genetic inheritance. A similar observation can be made for case marking, where it is seen for example that many languages throughout the world tend to form nasal genitives. Hence, it will be important to sift the available evidence in such a way that those similarities that do not add to the clarification of the problem of genetical relationship are eliminated. Semantic constraints Comparing lexicon and morphology we are not exclusively focusing on words or morphemes as a set of phonemes, the items also carry a meaning or fulfill a function that should be comparable. Meaning and function shift over history. Sweetser has observed a certain directionality in semantic change.9 But for the demonstration of a genetic relationship between Japanese, Korean and Altaic, the assumed time depth is such that the semantic shifts are often difficult to undo. If in some decades meanings can shift, like the English loan for ‘bar code’ in J ba@ko @do that developed the metaphorical meaning of ‘a bad comb-over’ in Japanese, then it must be clear that for relatively remote language relationships we have to take into account a certain measure of freedom for the compared meanings and functions. Of course, it is false to assume that genetically related languages always look alike in addition to meaning the same thing and being used in the same fashion.10 But the greater the semantic and functional latitude permitted in compared forms, the more likely it becomes that the established phonological correspondences are purely coincidental. Therefore, it is essential to set up certain constraints for semantic permissiveness. Counting only exact semantic equivalents would be untenable in the work ahead and hardly a handful of etymologies would stand the test. But unless an acceptable explanation of the assumed semantic changes can be provided, I will exclude semantically overpermissive etymologies from consideration. It is an art to decide whether a semantic association is acceptable or not, but semantic divergence can be tolerated to the extent that it answers to broad unidirectional tendencies like shifting from more concrete to more abstract. Besides, once a certain semantic shift has already been established elsewhere in the world, it gets more confirmatory force. 08 Nichols 1992, 261-62, 266-67. 09 Sweetser 1990. 10 Miller 1971, 17.

4.2. Genetic argument

47

Look-alike or cognate? Having sifted the non-significant similarities and working within semantic constraints, we come to the evidence that really matters for demonstrating or denying a kinship relationship between languages: systematic and recurrent similarities between phonemes, morphemes and lexemes. If the phonological, lexical and morphological similarities are due to inheritance from a form in the ancestral language, then we are dealing with true cognates and thus we adduce evidence that the languages involved are related. If the similarities cannot be traced back to common parental forms, then we deny the hypothesized genetical relationship. The method that checks whether look-alikes are really cognates is called the comparative method. The method is based on the postulation that sound change is regular. Exploiting the regularity of sound change, we are able to undo the sound changes that have occurred giving rise to the languages in question and so we can reconstruct the unattested ancestral language. Undoing the sound changes is based on establishing regular sound correspondences underlying morphological and lexical parallels. It is obvious that the effectiveness of the comparative method will depend on the regularity of sound change. By consequence there has been an ongoing discussion among linguists about how regular sound change really is. Recently, Ringe has tested the regularity of sound changes, using languages such as Old English that were attested a millennium ago.11 He finds that in the case of Old English about 94% of the words that survive in modern English exhibit regular development of all their sounds, the regularity of development in terms of tokens of sounds in the lexicon is greater than 98%. So from linguistic observation we know that, although few in number and sporadic, irregular sound change does occur. Nevertheless, this observation does not challenge the comparative method because there is no need to postulate that sound change is always regular or that regular sound change is without exception. The comparative method starts from the postulation that if sounds correspond regularly, there must be a reason why. If we can a priori rule out borrowing or nature as an explanation for the correspondences, so that only inheritance from a putative common ancestor or sheer chance remains, then sheer chance will be ruled out once that a significant number of cognates is attained. Is it possible that this line of reasoning holds for some language families such as Indo-European, but is not applicable to others such as Japanese, Korean and Altaic? The answer to that question is a simple ‘no’. From elaborate studies and handbooks that vindicate the basic principles of historical linguistics such as Baldi (1990), Hock (1991), Labov (1994), Sykes (1999), Campbell (1999), Joseph and Janda (2003), it is perceivable that one can gain more by applying the comparative method than one can lose by denying it. 4.2.2. A probabilistic argument Arguing for or against a linguistic unity is a probabilistic matter rather than a categorical one. Cleaning a language layer by layer from its borrowings and universally determined similarities may not always be as easy as in the case of obvious examples like raitomotiihu and wan-wan. Excluding borrowings in particular is a major problem in comparative historical linguistics. After having applied the above guidelines in order to separate the significant 11 Ringe 1999, 45-73.

48

Chapter 4. Methodology

etymologies from the insignificant ones, it is highly probable that a number of disguised borrowings and look-alikes will have slipped through the net. Given the fact that there are some domains in a language where borrowing is more likely to occur than in others, we can deduce a probabilistic criterion to weigh the stability of a certain etymological proposal. Morphology The utmost important source of evidence for genetic relationship is morphology. Scholars throughout linguistic history have favored shared morphemes as evidence for kinship, since they are highly resistant to internal replacement and borrowing. Morphological units are less susceptible to internal replacement because they organize into subsystems with a certain degree of internal cohesion. Morphological units also tend to resist borrowing more successfully than lexical units. Certain morphological units, however, are more prone to borrowing than others. Unbound morphemes are more readily copied than bound morphemes, since it is easier to perceive them as a distinct functional entity. Nevertheless, borrowing of derivational morphemes does occur occasionally. Above the borrowed Latinate suffix -ment in English acknowledgment has been pointed out. Morphemes for verbal derivation empirically prove to be less prone to borrowing than morphemes for nominal derivation. And, among verbal affixes, the borrowability of inflectional units is lower than that of derivational units. Extensive borrowing between languages in contact can lead to the assimilation of derivational morphological patterns, however, examples of borrowing inflectional affixes are exceedingly rare. Some Dravidian languages have been under intense pressure from Indo-Aryan languages, showing itself not only in lexical loans but also in changes to the phonology, such as oppositions of voicing and even aspiration developing in consonants, and the syntax such as development of specific kinds of finite subordinate clauses. The borrowing of so much vocabulary has led to the assimilation of derivational morphological patterns -paralleling ‘acknowledgment’but there is no borrowing of inflectional affixes. Even so, there are at least some instances where inflectional suffixes are borrowed, such as in the German plural suffix -s which is productive in borrowed words such as Auto - Autos ‘car - cars’. This suffix is considered to be a dialectal borrowing from the northern German dialects.12 Describing borrowability as a relative tendency along the above lines leads to the assumption that bound, verbal, inflectional, morphemes provide the most reliable evidence to demonstrate common ancestorship. Concluding that “In the verbal flexion, suffixes closest to the primary stem, markers of actionality and diathesis, seem relatively little susceptible to copying. It would be a strong clue to a common origin if this ‘intimate’ part of verbal morphology exhibited systematic correspondences of materially and semantically similar morphemes within congruent combinational patterns.” Johanson (1999, 8) has further refined our notion about borrowability -and hence genetic stability- and so he has defined a restricted core of verbal inflectional affixes that is highly telling in matters of genetic relatedness.

12 Hock 1991, 387.

4.2. Genetic argument

49

Lexicon Turning to the lexicon it is a crosslinguistical certitude that certain words are more indispensable for human communication and survival than others, no matter what the accidents of cultural surroundings are. This basic vocabulary includes pronouns, verbs and nouns referring to basic activities or concepts, such as I, you, we, eat, drink, sleep, sun, moon, water, eye, mouth, etc. Basic vocabulary tends to resist replacement more successfully than other, non-basic vocabulary. As a working model, it is possible to design a list of a certain number -for our convenience 100- of relatively stable words.13 The 100 words are a representative stock of the language in question, they represent a miniature-lexicon. Does this imply that we can define a concept ‘basic word’, as an absolutely immutable word regardless of linguistic community, cultural surroundings and time? No, accepting the concept of basic vocabulary, it is important to note that we are postulating the freedom of retention or loss of an individual item, but the determinism of the system of 100 words as a whole.14 One word may stay or leave, but the whole body of basic words is relatively stable. By consequence, we can say that there is ‘basic vocabulary’, but there is no such thing as ‘a basic word’. The expectation to find a certain rate of shared basic vocabulary in a test list of 100 words cannot be turned into a conditio sine qua non for language relatedness. This would be turning the argument upside down. The strength of the basic vocabulary argument lies on the contrary in its low borrowability and in its high resistance to replacement. If we find matches within the basic vocabulary, the probability that the matches are due to common ancestorship is higher than the probability that we are dealing with borrowings. A criticism that is often made against the assumption of basic vocabulary is that many of the basic items are not culture-free in a number of languages, but seen in the above light this criticism does not attack the concept of ‘basic vocabulary’, but rather the concept of ‘basic word’.15 With low borrowability and high resistance to replacement, we refer to relative tendencies. Low borrowability refers to the fact that words corresponding to the semantic items on the list are less prone to borrowing in comparison to the overall lexicon. High resistance to replacement refers to the relatively low probability that native items on the list get replaced by synonymous loanwords. However, every single item on the basic vocabulary list can get borrowed. And every individual item on the basic vocabulary list is free to stay or to leave. As an illustration of how words for semantic items present on the basic vocabulary list can get borrowed, I refer to a number of Chinese loanwords in Japanese. For 11 items on Swadesh’s 100 list, 11. one; 12. two; 16. woman; 17. man ; 27. bark; 29. flesh ; 30. blood; 33. grease; 52. heart; 53. liver; 72. sun, a Chinese loanword is attested in Japanese. Japanese has borrowed iti ‘one’; ni ‘two’; zyosei ‘woman’; dansei ‘man’; zyuhi ‘bark’; niku ‘flesh, meat’; ketueki ‘blood’; sibo@ ‘grease ,fat’; sinzo@ ‘heart’, kanzo@ ‘liver’, taiyo@ ‘the sun’ from Chinese. Except for one item, niku ‘flesh, meat’ for which we find OJ sisi ‘flesh, meat’, but 13 For an example of a 100 wordlist, we can refer to Swadesh 1955 and to chapter 11. 14 Swadesh 1955, 127: ‘...there is no such thing as an immutable vocabulary item...’. 15 Campbell 1999, 180-183. “There are serious problems with the assumption of a universal, culture-free basic vocabulary. One is that many of the items are not culture-free, but rather are borrowed for cultural reasons in a number of languages.” The other critisisms made by Campbell in this chapter do not attack the concept of basic vocabulary, but rather they attack the method of lexicostatistics when applied on basic vocabulary.

50

Chapter 4. Methodology

no counterpart in contemporary Japanese, the Chinese loanword is in use side by side with its native synonym. Contemporary Japanese hito- ‘one’, huta- ‘two’, onna ‘woman’, otoko ‘man’, kawa ‘bark’, ti‘blood’, abura ‘grease ,fat’, kokoro ‘heart’, kimo ‘liver’, hi ‘sun’ have not been replaced by the Chinese loanwords, but are still present in the language in competing distribution. But 11 Chinese borrowings on 100 in the basic vocabulary is 11 % while the proportion of Chinese loan vocabulary in the entire Japanese lexicon exceeds 60 %. This fact illustrates that semantic items belonging the basic vocabulary list are less prone to borrowing. Moreover, only a single item, 29. Flesh, has been completely replaced by a Chinese loanword. Strictly speaking, only 1% of the basic vocabulary has been replaced by borrowing. This illustrates the tendency of basic vocabulary to resist replacement more successfully. Among the basic vocabulary items, pronouns have the highest stability. Pronouns share the nominal property of deixis, but morphologically they can be derived in different ways in order to agree with their antecedents. Therefore, they can be situated on the dim interface between lexicon and morphology. Since pronouns occur in systematic groups, each pronoun being part of a larger syntacto-semantic system, it is rather unlikely for a language to borrow a single pronoun from a donor language, because this would undermine the logic of its native pronominal system. However, the resistance of pronouns to borrowing is by no means absolute. A classical example are the English pronouns they, their, them which are borrowings from Scandinavian.16 It would require extensive borrowing in order to take over the entire pronominal system of the donor language. Besides the fact that the comparison of pronominal systems as a whole can throw light on shared formational peculiarities, adds to the relative weight of pronominal evidence. For the sake of the evaluation of the material we will rank pronominal evidence as the best evidence among the stable part of the lexicon. It is also interesting to note that verbs are harder to borrow than nouns. Although we do not exactly understand why this is the case, part of the answer comes from observation of instances of borrowing. It just is an empirical tendency that verbs are more stable than nouns. If the need for borrowing a verb does arise, many languages instead borrow a nominal form of the verb and use a native verb covering all purposes, such as verbs meaning ‘to do’ or ‘to make’. Japanese for example has borrowed many sports along with their names from English, but there is a clear preference to borrow the nominalized form such as jogingu ‘jogging’, haikingu ‘hiking’ etc. and add suru or yaru ‘to do’ in order to express the corresponding verb. In the past linguists have argued that numerals should be reckoned among the more stable parts of the vocabulary as well. However, empirical findings can show that this faith in numerals is misplaced. Among the most striking examples of languages where nearly the entire numeral system has been borrowed we find Japanese and Korean, both having extensively borrowed numerals from Chinese. It is easily understood that counting is to a certain extent culturally determined, nomadic people, dealing with herds of cattle, for example, may have other counting needs than people involved in trade. The less stable part of the vocabulary belongs to the domain of cultural vocabulary. Apart from the numerals and counters, cultural words are terms for fauna and flora, agricultural, technological, religious, artistic, medical, political and martial vocabulary. Cultural vocabulary can easily 16 Hock 1991, 386.

4.2. Genetic argument

51

be borrowed or replaced. Evidence that comes from this part of the vocabulary can hardly add any strength to a hypothesis of language kinship. 4.3. Measuring time-depth The comparative method does not provide us with a tool for measuring time depth. Of course, using the comparative method, linguists are able to reconstruct linguistic changes in a relative chronology and with the help of archaeology it is sometimes possible to provide absolute dates for archaeological events with which these linguistic innovations can be correlated. However, the only purely linguistic tool for measuring time depth is glottochronology, which is often claimed to be far from reliable.17 Swadesh ’s formula in glottochronology has been reconsidered several times during the past decades and recently Starostin has sharpened the equation.18 What glottochronology really measures is the time that separates sister languages from their ancestral language. The method on which it relies to obtain its results is generally called lexicostatistics. The effectiveness of lexicostatistics depends entirely on the postulation that semantic drift over time is constant. Counting the number of shared semantic retentions between two sister languages A and B, the time that separates them is calculated. The main problem of glottochronology is not in choosing a sample of 100 words, called basic vocabulary, in function of their tendency for stability as a miniature-lexicon. The most important flaw is the assumption of a constant rate of retention (or loss) through time. The large differences of actual semantic loss over 1000 years in test languages are not very inviting, whatever the statistical standard deviation on their average may be. It should be noted, however, that even Starostin emphasizes that the appliance of the lexicostatistic method is limited to well established language families with a reliable system of phonological correspondences. He further stresses that the only method for establishing genetical relationships is the classical comparative method.19 Since in the work ahead I want to focus on the question whether it is possible or not to demonstrate that Japanese is related to Korean, Tungusic, Mongolic and Turkic, measuring time depth with the lexicostatistic method is beyond the scope of the present work.

17 Ross 1998; Campbell 1999, 177-86; Comrie 2000.

− nλt Starostin 1999: C = e− nλ 0Ct 18 Swadesh 1955: Classical equation: C = e

2

19 Starostin 1999, 4: ‘What is the method of establishing and proving ancient genetic relationships between language families? To this, evidently, there can be only one answer: the classical method of comparative historical linguistics, that is the discovery of a system of regular sound correspondences between proto-languages that is valid for the majority of lexical and morphological items, and the reconstruction of an earlier system based on those correspondences.” Starostin 1999, 13 (footnote): “I would like to emphasize that no lexicostatistical survey is possible until thorough comparative work has been done.”

Chapter 5. The individual phonological inventories Applying the comparative method on proto-Japanese, proto-Korean and the proto-languages that are supposed to make up the Altaic family, certainly does not assume that we already know everything we need to know about all the proto-languages involved. An important first step is to provide the consonant and vowel inventories of the individual proto-languages and to point out the critical proto-phonemes. These proto-phonemes are critical in the sense that their reconstruction is controversial and somewhat speculative. Every reconstruction of an individual proto-phoneme is by its very nature hypothetical, but for the reconstruction of the critical proto-phonemes conflicting hypotheses are available in the linguistic literature. Highlighting these critical proto-phonemes at an initial stage, it is important to bear in mind that hypotheses built atop other hypotheses can sometimes be like houses of cards. 5.1. Japanese 5.1.1. Consonant inventory Table 1. Consonant inventory of pJ labial

dental

*p

*t

nasals

*m

*n

liquid and glides

*w

*r

obstruents

palatal *s

velar *k

*y

5.1. Japanese

53

Table 2. Consonant inventory of OJ labial

dental

palatal

velar

voiceless obstruents

p

t

s

k

voiced obstruents

b

d

z

g

nasals

m

n

liquid and glides

w

r

y

The reconstruction of the proto-Japanese consonant inventory is relatively uncontroversial. Nevertheless, there is some disagreement on the validity of the reconstruction of voiced obstruents in proto-Japanese, on the number of voiced obstruents that must be assumed and on the reflexes of certain obstruents in Old Japanese. The present work takes the view that there was no phonological voice distinction in proto-Japanese. The assumption that proto-Japanese had voiced obstruents is hardly tenable in the case of pJ *g and *z.1 At the annual meeting of the American Oriental Society in 1972 Martin has delivered an unpublished paper arguing for the reconstruction of a voiced velar stop *g in proto-Japanese on the basis of a number of possible cognate pairs in which Korean initial k- corresponds to Japanese ø (zero).2 The only external cognates that he cites in his 1987 monograph are J am/b(i)- ‘bathe’ with K kam- ‘bathe’ and the -i- formant which reverses the transitivity of the verb base with K -ki-, -khi-, -hi-, -i-, a formant that makes lexical passives and causatives. Unger and Ramsey agree with Martin’s reconstruction because pJ *g fits into the symmetrical triple obstruent system (*p, *t, *s, *k / *b, *d, *z, *g / *np, *nt, *ns, *nk) that they reconstruct in proto-Japanese. But, however strong and numerous Martin’s unpublished cognates may be, apart from some sporadic k~ø alternations, there is no systematic internal evidence for the validity of pJ *g > ø.3 Therefore, I will reject the reconstruction of an initial pJ *g- underlying the distal doublet a- ~ ka- ‘that’ that is also reflected in are ‘that one’ ~ kare ‘he’, atira ‘over there’, and kanata ‘here and there, far in distance, a long way off’ . The reconstruction of the distal pronoun *ka- is justified for the entries a- ‘that’, ka- ‘that’ are ‘that one’, atira ‘over there’, kanata ‘here and there, far in distance, a long way off’, kare ‘he’ in the index, but I see no reason to reconstruct a still earlier *ga-. The elision of the initial k- may be due 1 2

3

For the reconstruction of voiced obstruents pJ *g and *z I am in disagreement with Martin 1987, 20, 36; Unger 1973, 33-35; Unger 2000 a, 666-67. Reference to this paper is found in Whitman 1985, 19; Unger 1973, 33; 2000, 666; Martin 1991 a, 273. Martin 1987, 20 refers to his discovery of ‘a fair number of possible cognate pairs in which Korean has initial k- where Japanese has zero’ without specification of the source in question. Whitman 1985, 19,186; Miyake 1999, 136-138.

54

Chapter 5. Phonological inventories

to sporadic phonological erosion for both the distal *ka- and the transitivity reversing *ki-. It can be remarked that velar elision is attested elsewhere in Japanese, as for example the k-elision in the adjective attributive or in tuitati ‘first day of the month, new moon’ that is derivable from tuki ‘moon’. Unger’s, Vovin’s and Martin’s proposal to reconstruct pJ *g as the consonant preceding the -i- formant which reverses the transitivity of the verb base can be dismissed on the same grounds. For the time being it remains safer to reconstruct pJ *k-, except for abiru ‘pour (water) over oneself, bathe in’ for which there is no internal indication whatsoever to reconstruct an initial velar. The evidence is even weaker for the reconstruction of pJ *z which is mainly made for the sake of phonological symmetry because there is little internal evidence and no consistent comparative evidence suggesting pJ *z > OJ ø . The only external proposal I am aware of is Whitman’s comparison of ito ‘thread’ with MK sil ‘thread’. Some internal motivation for the reconstruction of pJ *z can be found in the alternation of s with ø in doublets such as ame ‘rain’ and -same in harusame ‘spring rain’, kosame ‘shower’ etc., ao ‘blue’ and OJ sawo ‘deep blue’, ara ‘new’ and sara ‘newness, something new’, aku ‘open, be opened, become vacant’ and saku ‘bloom, open, come out’, sinu ‘die’ and OJ in- ‘go away’, ueru ‘plant’ and sueru ‘fix, install’. 4 Since there are many cases where the -s- alternant appears in intervocalic position, it is rather unnatural to suppose devoicing from *-z- > -s- in a vocalic environment. In the work ahead I will treat the examples of doublets of s with ø separately and I will argue that for the majority of these examples a better explanation can be found in morphology, the s being a relic of various morphemes. The rejection of pJ *zis relevant for the internal reconstruction of the following etyma in the index: ai ‘indigo plant’, akeru ‘open, empty, vacate’, aku ‘open, be opened, become vacant’, ame ‘rain, rainfall’, ao ‘blue’, arai ‘rough, coarse, natural, sparse, crude’, atarasii ‘new’, OJ in- ‘go away’, ine ‘rice plant’, ito ‘thread’, OJ saki1 ‘good luck, happiness’, OJ saki 1pap- ‘flourish’, sara ‘newness, something new’, and se ‘a woman’s familiar call for her husband or elder brother’, sinu ‘die, pass away’, ueru ‘plant’. Although I disagree, the assumption that proto-Japanese had voiced obstruents is stronger in the case of pJ *b and pJ *d due to some modern southern Ryukyu dialects that have band d- corresponding to OJ w- and OJ y-. As for the reconstruction of pJ *d the dialectal evidence rests entirely on a single dialect Yonaguni, which reflects initial OJ y- as d-. The dialectal evidence is rather weak because the Yonaguni d- is restricted to the initial position. Moreover, d- also appears in borrowings from Chinese such as Yo. dasai ‘vegetables’ which reflects a fortition y- > d-. Therefore, I doubt the antiquity of Yonaguni d- and I am reluctant to reconstruct pJ *d- as an ancestor for OJ y-.5 It is also suspect that there are no Yonaguni words available which make it possible to distinguish between initial pJ *yi and pJ *i which are neutralized in attested Japanese (OJ i), suggesting that the occurrence of Yo. d- depends on the presence of OJ y-. An argument that has often been misused in favor of the reconstruction of pJ *d- comes from external comparison with Korean or Altaic. First, for the purpose of the present study such an argument would be circular. Besides, 4 5

I refer to Martin 1987, 35-36 for additional examples of alternation of s with ø. Shibatani 1972; Thorpe 1983; Whitman 1985, 18-19; Miyake 1999, 141-42 favor the reconstruction of pJ *y and pJ *w instead of pJ *d and pJ *b. Although Miyake finds the evidence for pJ *b and *d rather ambiguous and although he favors pJ *y and pJ *w, he prefers to reconstruct pJ *b and *d since it represents the consensus view.

5.1. Japanese

55

anticipating what follows, I think that the reconstruction of pJ *d adds more to obscuring than to clarifying the comparative picture. An argument contra the reconstruction of pJ *y is that from the cross-linguistic viewpoint a fortition of a palatal glide to a dental stop is an unnatural sound change, but Miyake mentions a similar development from pre-Middle Vietnamese *y to Middle Vietnamese *d.6 Another objection to pJ *y is the symmetry of the reconstructed proto-Japanese sound system since the reconstruction of a phonological system with only one voiced obstruent pJ *b would be typologically bizarre. That brings us to the crux of the question. Is it really necessary to reconstruct pJ *b? Turning the argument upside down, the symmetry of the reconstructed proto-Japanese sound system is an important indication that pJ *w should be reconstructed instead of pJ *b. However, Japanese comparative dialectal evidence supports that pJ *b underlies OJ w both initially and medially. The evidence comes from South-western Ryu@kyu@ dialects, namely the Sakishima dialects, consisting of Miyako and Yaeyama, and the Yonaguni dialect. Martin and Whitman present a list of dialectal correspondences that give initial b- for OJ wand Martin also lists some cases for which there is evidence of medial Ryu@kyu@ -b- for OJ -w-.7 However, my suspicion is that, just like in the case of the fortition of the palatal glide, the Ryu@kyu@ b may be a fortition of pJ *w rather than a retention of pJ *b. This view is supported by Miyake’s observation that in the Suiko period attestations words for which OJ w or y is attested are transcribed with phonographs read with EMCh. w or y and not with EMCh. b or d.8 Hence, I will favor the reconstruction of a proto-Japanese sound system that lacks voice distinction as a working hypothesis in the following evaluation of the etymological proposals. Although the Old Japanese consonant inventory in table 1 does not contain any critical phonemes, OJ p may require special notice because it has been suggested that its articulatory definition had already become a bilabial fricative F by the time of Old Japanese.9 That is why Miller prefers the representation F instead of p. In his dissertation Miyake has persuasively argued against the lenition of pJ *p and he has demonstrated that p remained unchanged until the Heian period.10 Therefore, the Middle Japanese reflex of OJ p is represented by f in the present work. The difference between the proto-Japanese and Old Japanese consonant system listed in table 1 and in table 2 lies in the appearance of a voiced obstruent series b, d, g, z. There is good reason to argue that the voiced obstruents were prenasalized in Old Japanese.11 Old Japanese prenasalization can be described as a nasal onglide preceding a medial voiced obstruent and the prenasalized obstruents are usually transcribed as [nb], [ nd], [ng], [ nz]. The first indication that prenasalization has developed into voice distinction comes from the observation that Old Japanese voiceless obstruents develop into voiced obstruents in a prenasalized environment. From the ninth century on, a heterogeneous group of phonological changes known collectively as onbin altered the Old Japanese phonological pattern in a way that a syllable-final nasal was allowed. Japanese verb morphology shows 06 07 08 09 10

Miyake 1999, 139. Whitman 1985, 15-16; Martin 1987, 9-10,14. Miyake 1999, 144. Hashimoto 1928; Whitman 1985, 17; Martin 1987, 37; Miller 1996, 59. Miyake 1999, 396-400.

56

Chapter 5. Phonological inventories

the effects of onbin changes clearly. Contemporary Japanese gerunds such as yonde ‘read’, tonde ‘fly’ and sinde ‘die’ have developed from the Old Japanese forms yomite ‘read’, tobite ‘fly’ and sinite ‘die’. In the formation of the verb gerund -mite, -bite, and -nite all become -nde by vowel elision and in the new prenasalized environment, the obstruent -tbecomes voiced -d-. The second indication is that morphological and etymological boundaries sometimes occur on a voiced obstruent, evidencing the contraction of a morpheme with a final nasal syllable and a morpheme with an initial obstruent. Examples can be found in internal derivations such as hude ‘writing brush, painting brush, pen’ from humi ‘writing’ and te ‘hand’; OJ kabi1 ‘sprout, unhusked ear of grain’ from OJ kami1 ‘top, head, upper part’ and OJ ipi 1 ‘rice’; kabuto ‘helmet, headpiece’ from kami ‘top, head, upper part’ and huta, OJ puta ‘lid’, nazo ~ nanzo ‘riddle, why’ from nani ‘what’ and -so ‘thing’, nado ‘or something’ from nani ‘what’ and -to ‘place’, uzi ‘clan, lineage’ from umi, the deverbal noun of umu ‘give birth to, bring forth’ and ti ‘blood; line’, yagate ‘before long’ from yami, the deverbal noun of yamu ‘stop’ and the adjective stem kata- ‘hard’, yugake ‘archer’s glove’ from OJ yumi-kake2 ‘archer’s glove’ and originally from yumi ‘bow’ and the deverbal noun of kakeru ‘cover with (cloth), spread over, veil’. Third, a number of modern dialects, notably those of northern Honshu@, southern Shikoku and some Ryu@kyu@ dialects have retained prenasalization to some extent. Sendai in northeast Honshu@, for example, pronounces mado ‘window’ as [mãdo] with nasality on the preceding vowel. The Kobama dialect in the Ryu@kyu@s has kandu for standard kado ‘corner’ and pangun for hagu ‘strip off, tear off’. In many other dialects, including standard Tokyo, prenasalization has been lost. However in Tokyo dialect there is a residue of the prenasalized velar, like in the pronounciation of kagi ‘key’ as [kaNi] or the nominative case marker ga as [Na], with tendency of [ng] to absorb the nasality and to become [N]. Finally, foreign written sources suggest that the voiced obstruent series in Late Middle Japanese and Modern Japanese was still prenasalized to a certain extent. In the JapanesePortuguese dictionary of 1604 -nd- and -ng- is observed for -d- and -g- in the speech of Kyu@shu@ at that time. An 18th-century Korean glossary of Japanese, the Wago-ruikai, writes Hangu#l digraphs equivalent of -mp-, -nt- and -Nk- for the medial voiced stops -b-, -d- and -g-. And the Nihon Kan-yakugo, a 15th century Chinese language guide transcribes Japanese syllables preceding a voiced stop with a final nasal. On the basis of morphological, etymological, dialectal and textual evidence it is safe to assume that the Old Japanese obstruents OJ b, d, g, z resulted from the rephonologization of prenasalized obstruents pJ *np, pJ *nt, pJ *nk, pJ *ns. Reminiscent of how the Altaic languages do not allow for consonant clusters in initial position, Old Japanese did not permit word-initial voiced obstruents except in mimetic adverbs. From the ninth century on, as loans from Chinese began to have a major impact, the restriction was relaxed and initial voiced obstruent began to appear in borrowings and in contracted native forms.

5.1. Japanese

57

5.1.2. Vowel inventory 11 Table 3. Vowel inventory of pJ front high

central

*i

back *u

mid

*o

low

*a

Table 4. Inventory of pJ diphtongs *-a

*-i

*-o

*a(C)i

*a*i-

*i(C)a

(> OJ e1)

*u-

*u(C)a

(> OJ o1)

*o-

(> OJ e2)

*u(C)i

(> OJ i2)

*o(C)i

(> OJ i2)

*i(C)o

(> OJ e1)

*u(C)o

(> OJ o1)

OJ reflexes are given between brackets ( ).

Table 5. Vowel inventory of OJ front high

i1

mid

e1

low

central i2 e2

back u

o2

o1

a

11 The vowel inventories of OJ and pJ in tables 3 and 5, and the inventory of pJ diphtongs in table 4 are based on Miyake 1999, 153, 159, 592. The syllable inventory of OJ in table 5 is based on Unger 2000 a, 656.

58

Chapter 5. Phonological inventories

Table 6. Syllable inventory of OJ

a i1

i

i2

o

labial

a

pa, ba

ma

p i 1 , bi1

mi 1

p i 2 , bi2

mi 2

pu, bu

mu

i u

u e

zero

e1 e2 o1 o2

e

o

dental

pe 1, be1 me 1 pe 2, be2 me 2 po, bo

mo

palatal

velar

ya

ka, ga

wa

ta, da

na

ra

sa, za

wi

ti, di

ni

ri

si, zi

tu, du

nu

ru

su, zu

te, de

ne

re

se, ze

t o 1 , do1 no 1

ro 1

so 1, zo 1 y o 1

ko 1,go 1

t o 2 , do2 no 2

ro 2

so 2, zo 2 y o 2

ko 2,go 2

we

wo

ki 1, gi 1 ki 2, gi 2 ku, gu ke 1, ge 1 ke 2, ge 2

Old Japanese texts preserve an interesting phonograph-dichotomy for syllables that are phonemically identical at all later stages of the language. From the 8th century textual evidence it is clear that there was a certain syllablewise contrast, but it has taken reconstructions of the Middle Chinese syllables associated with these contrastive phonographs (Lange 1973, Miyake 1999), internal reconstruction on the basis of OJ morphophonemic alternations (Unger 1973, Martin 1987) and comparative evidence from Korean (Whitman 1985) to shift the discussion from the phonemic to the phonetic level and to find out how this contrast really worked. Today it is commonly agreed that the OJ texts reflect a system of eight vowels. I have listed the vowels in table 5, using subscript 1 for the A-type vowels (ko@rui) and subscript 2 for the B-type vowels (otsurui). This representation is purely orthographic in its reference, not implying any phonetic reality. Miyake hesitates to use the term ‘vowels’ for Old Japanese, since ‘we cannot be sure at this point if all of the so-called OJ ‘eight vowels’ were really vowels’ (sic.).12 Following Miyake’s argumentation OJ e 2 turns out to be a diphtong, which he does not consider as a single vowel phoneme. However for the present account I will use the term vowel for both monophtongs and diphtongs. So, the present day five-vowel system originated from an Old Japanese eight-vowel system, but how did the eight-vowel system come into being? The most widespread theory of contractions says that a proto-Japanese four-vowel system (table 3) developed into the OJ eight-vowel system, directly reflecting the four original vowels as OJ i1, a, o2, u, and deriving four new vowels OJ i2, e1, e2, o1 through contractions ( (table 4).13 These contractions resulted either from the loss of an intervocalic consonant or from an original sequence of two vowels. Morphotactic alternations such as free form OJ sake2 ‘rice wine’, and bound form OJ saka ‘rice wine’ and vowel contraction in compounds such as OJ sake1ri ‘is blooming’ < *saki 12 Miyake 1999, 159. 13 Whitman 1985; Martin 1987; Miyake 1999.

5.1. Japanese

59

‘blooming’ + *ari ‘be’, OJ opi2si ‘large rock’ < opo- ‘large’ + isi ‘rock’ or OJ kazo1p‘count’ < OJ kazu ‘number’ + ap- ‘join’ is the kind of internal evidence that underlies the contractions.14 However, the four-vowel hypothesis for proto-Japanese still is not entirely accepted. Unger reconstructs a five-vowel system in proto-Japanese to which *e is added in order to account for cases of OJ e in which it is unclear which underlying vowel sequence to reconstruct and for cases of o ~ a alternation in Japanese.15 Although Miller complains about what he calls ‘the muddled phonological conclusion of the Lange-Martin school’ (sic.), his view about the derivation of all Old-Japanese vowels -except for OJ o1- is compatible with the four-vowel hypothesis for proto-Japanese and the contraction hypothesis deriving the OJ vowels.16 Miller incorrectly points out that there is no reliable internal or external evidence to support the derivation of OJ o1, for which the example OJ kazo1p- ‘count’ is given above and for which both internal and external evidence can be found in Whitman 1985.17 He is right, however, in pointing out that o1 is rendered in a different way than the accusative suffix wo in Old Japanese texts, and thus the pronunciation of the vowel and the accusative suffix could not have been the same at that time. But this is not what Martin, Lange, Whitman, Unger, Miyake etc. are suggesting when using the Yale romanization wo for the vowel in question. Just like Miller’s representation o1 this notation is merely conventional. Miller also makes speculations about a Suiko six-vowel system and a pre-Suiko five-vowel system, as intermediary stages in the development from the proto-Japanese four-vowel system to the eight-vowel system of Old Japanese.18 He posits these intermediary stages on written records, but since he fails to give any further motivation for this reconstruction, it seems that the grounds for such an argument are lacking. Comparative dialectal studies reconstruct a much more complex vowel inventory of proto-Japanese, but since fewer vowels can explain the bulk of the material, we will not take these proposals into account for the purpose of the present study.19 Studies on vowel contractions and monophtongization abound in the linguistic literature, but it should be remarked that not every vowel sequence that is found in OJ texts necessarily develops by contraction, under some conditions vowel elision occurs instead of contraction. Reconciling Whitman’s theory with his own views, Unger has recently revised the rules of vowel contraction in Old Japanese in terms of vowel length.20 The rule claims that vowel 14 For more examples reference is made to Whitman 1985, 41-42; Miyake 1999, 160-61; Unger 2000 a, 657. 15 Unger 1993, 48; Miyake 1999, 157. 16 Miller 1996, 54 for the remark on the Martin-Lange school. Miller 1996, 50-59 summarizes the author’s view on the OJ vowel system. 17 Miller 1996, 59: ‘Unfortunately we have no reliable source in or outside our Japanese materials proper concerning the difference in articulation between o1 and o2.’ (sic.) 18 Miller 1967, 191. 19 Serafim 1994. 20 Unger 2000 a, 660-663 formulates rules for vowel contraction and vowel drop in Japanese. The view that initial low pitch in Japanese, as reflected in the RM as well as in Japanese dialects, is a reflex of original initial syllable vowel length was first suggested by Hattori and developed in Whitman 1985 and Martin 1987. Whitman 1985, 198-201 investigates the comparative evidence for the relation between initial low register in Japanese and vowel quantity. Vovin 1993 c provides internal and external evidence for long vowels and pitch accent in proto-Japanese.

60

Chapter 5. Phonological inventories

sequences contract only if the first vowel is short. If not, one of the vowels drops according to the length of the second vowel. The vowel length environment is based on treating the low initial register of Middle Japanese, represented by a low dot in the Ruiju Myo@gisho@ as distinctive vowel length. 5.2. Korean 5.2.1. Consonant inventory Table 7. Consonant inventory of pK labial

dental

voiceless obstruent

*p

*t

nasal

*m

*n

liquid and glide

*w(?)/ y

*l

*c

*s

velar

glottal

*k

*h

*ng

Table 8. Consonant inventory of LMK labial

dental

plain obstruent

p

t

aspirated obstruent

ph

th

fricative

[ß]

nasal

m

n

liquid and glide

w [?]

l

s

palatal

velar

glottal

c

k

h

ch

kh [z#]

[V] ng

y

A lingering problem for the comparative Altaic enterprise is the restricted distribution of voicing, not only in OJ as I have pointed out in the previous paragraphs, but also in Middle Korean. For Korean there is still no agreement on the question whether Middle Korean and proto-Korean were phonologically voiced or not. I agree with Martin that the notion of earlier distinctive voicing in Korean came about because the Middle Korean orthography provides a separate notation for the voiced fricatives.21 The obstruents /p/, /t/, /k/, /c/ in 21 Martin 1996, 48.

5.2. Korean

61

contemporary Korean are automatically voiced in voiced environments, and there is no reason to assume that voicing worked in a different way in Middle Korean. The sibilant /s/ is normally unvoiced in all environments in contemporary Korean, but there is evidence for a voiced allophone [z] in northern dialects, especially in earlier times.22 It cannot be excluded that /s/ became phonetically voiced [z] in a voiced environment in Middle Korean as well. I suspect that voicing was non-distinctive in Middle Korean, as it is in Korean today. However we find three graphemes , , ∆ in Middle Korean, that are generally believed to represent voiced fricatives in Middle Korean. Orthographically the graphemes and ∆ seem to correspond to the graphemes and plus one feature added. Martin’s lenition hypothesis suggests that the extra feature is fricativation for the stops p and k and voicing for the sibilant s, thus reading , , and ∆ as ß, V , z respectively.23 As far as the lenition from pK *p to /ß/ and from pK *k to /V/ is concerned, I agree with the proposals in Martin’s 1996 monograph, but as I have pointed out in an earlier publication (2000), the voiced fricative /z/ hypothesis for MK ∆ is not whithout problems. The first argument against the reading of MK ∆ as /z/ is found in the description of the grapheme in the Hunmin cho#ngu#m haerye as a ‘half front-tooth sound’, a label which conforms to the Early Middle Chinese palatal nasal /n!/ and not the voiced counterpart of /s/. A second problem is why the creators of the Korean alphabet would introduce an extra symbol for the voiced counterpart of [s], but not for the allophones of [p], [t], [k] and [c]. A third obstacle is a serious number of near . minimal pairs such as MK¨cis‘build’, inf. ci ∆eand MK sis- ‘wash’, inf. sise-, MK . . ko∆om ‘material’ and MK ka som ‘chest’, MK sa∆o ‘die’ and MK sa som ‘deer’ etc. 24 If /s/ indeed was voiced in a voiced environment in Middle Korean, the representation of [z] sometimes with grapheme and sometimes with ∆ grapheme would be completely random. If there was no intervocalic voicing of /s/ in Middle Korean, MK ∆ and MK would represent two different phonemes instead of allophones. This brings us to the fourth argument against the interpretation of MK ∆ as /z/: the symmetry of the Middle Korean sound system. From a cross-linguistic point of view it is very unlikely for a system that does not show distinctive phonological voicing, to make an exception for only one set of phonemes, /s/ and /z/. I was unable to find a single language with such a peculiar phonology. It should be clear that as far as MK /ß/ and /V/ are concerned, the distinctive feature is not voicing, since this occurs automatically in a voiced environment, but fricativation. Other arguments against the interpretation of MK ∆ as /z/ can be found in Korean readings of Chinese characters and in Chinese loanwords, for which we refer to Vovin’s article.25 Resting on Chinese loanwords and on comparative Altaic evidence, Ramstedt is the first to suggest that MK ∆ stood for a palatal nasal [n!], something like a weak nasalized [j], and his theory is supported through some of Starostin’s Altaic comparisons and through Vovin’s examination of Korean internal evidence.26 Although I have presented comparative evidence in favor of a palatal nasal fricative /n!z/! reading for MK ∆, inherited from pK *n!, in the past, I now tend to agree with 22 Martin 1996, 2. 23 Martin 1996, 65. The representation W and G is used for the voiced fricative phonemes ß and V. 24 Ramsey 1978, 46. 25 Vovin 1993, 254-55. 26 Ramstedt 1939, 10, 41, 64. Starostin 1992, 21. Vovin 1993, 247-59.

62

Chapter 5. Phonological inventories

Martin that most of the occurrences of MK ∆ result from the weakening of an original pK *s. However, following the above motivation, I do not agree that the result of the lenition of pK *s was the phoneme /z/ in Middle Korean. There was an extra feature involved, which may have been palatalization, thus MK ∆ could reflect the voiced palato-alveolar fricative /z# /. In the 15th century, /c/ was not palatal as it is in Korean today, but merely an alveolar affricate [ts]. The -sop- / -∆op- deferential auxiliary for example had a third representation -cop- after t, c, or ch, which can be taken for evidence of a [ts] pronunciation. Palatalization took place around the 17th century, but the process may have been underway during the Middle Korean period. By the 16th century MK ∆ disappeared from the language, a process that probably made room for the palatalization of the affricate c. There are quite a number of scribal doublets attested in which the grapheme for the affricate c alternates with the grapheme ∆. Martin reports that MK mo∆om ‘mind, spirit, heart’ is written as mocom in an early 15th century source.27 In Middle Korean the verb stem a(n)c- ‘to sit’ is also sporadically . represented as a∆- in a voiced environment. MK howo ∆a is found in contemporary Seoul as hwonca. ‘alone, by oneself’. If MK ¨ko∆ ¨eps- ‘have no .border’ is etymologically related to MK ko cang ‘end, extremity; extremely, most’, MK s ko ci > K kkaci ‘till, all the way to’, then it might also represent sporadic confusion between the affricate /c/ and the palatal fricative /z#/. There are also examples of scribal errors that confuse the grapheme in . . palatal environment with the ∆ grapheme, such as the writing of ta ∆i for ta si ‘again’ or the misspelling of the honorific -usi- as -u∆i-.28 Following Martin’s lenition hypothesis I now accept pK *s as the ancestor of MK z# because the lenition must be connected to the lenition of pK *p and pK *t in similar environments. In 1975 Ramsey proposes that the MK voiced fricatives lenited from the plain obstruents pK *p,*s, *k, but later he changes his view and now he argues that they are reflexes of an original voiced obstruent series pK *b, *z, *g. He also reconstructs pK *d on the basis of the anomalous t/l conjugation of Korean t-irregular verbs. The t- / p- / s- irregular verbs are at the core of his argument because they are similar in accentual behavior and because they have a stem final obstruent t, p, s before consonant . endings, but MK l, ß, ∆ before vowels. Examples are ¨ket-, inf. ke le‘walk’ versus ket-, inf. . . . ke te- ‘gather’; ¨kwup-, inf. kwu ße‘bake’ versus kwup-, inf. kwu pe‘bent’ along with¨cis-, . inf. ci ∆e- ‘build’ versus sis-, inf. sise- ‘wash’. Since Ramsey fails to find any phonetic motivation for the difference in behavior between the t- / p- / s- irregular verbs and the t- / p- / s- regular verbs without lenition in intervocalic position, he reconstructs voiced obstruents different from the plain obstruents underlying the irregular and the regular conjugations respectively. Martin, however, persuasively argues that Ramsey’s initial lenition hypothesis does not represent fatal flaws, explaining the voiced fricatives in the Middle Korean verb . stems as a weakening of the plain obstruents in the regular environment *CV Cu/o-. The weakening of /k/, /p/, /s/ led to the new voiced fricative phonemes, but the lenition of /t/ merged with an allophone of the existent liquid phoneme /r/. Another argument for lenition is that no convincing evidence has been found for an initial voicing contrast. There are also . . several examples of compounds like MK han-¨∆wum ‘deep breath’ < han ‘big’ + ¨swum ‘breath’, MK tay-¨ßem ‘a mighty tiger’ < tay ‘big’ + ¨pem ‘tiger’ in which the initial voiceless obstruent of the second member becomes a fricative when it occupies the medial 27 Martin 1996, 49. 28 Martin 1996, 70, 99.

5.2. Korean

63

position in compounding.29 Moreover, Miyake reports that the Early Middle Chinese phonographs, used to represent syllables of Koguryo#, Paekche, and Silla Old Korean toponyms in the Samguk Sagi and Samguk Yusa show random voicing alternations for the consonants.30 The data imply that Old Korean did not have a phonological voicing distinction, just as is the case in contemporary Korean. The main problem with the lenition hypothesis is that there still are a serious numbers of medial obstruents that cannot be explained by Martin’s theory. Ramsey and Martin offer a list with exceptions in which a plain obstruent before a minimal vowel in the regular . environment *CV Cu/o- does not lenite according to our expectations, but as Martin points out, there are no verbs among the exceptions.31 Besides most of the nouns belong to the sphere of cultural vocabulary, which increases the probability that we are dealing with a borrowing. For some of the exceptions like susung ‘teacher’ that is thought to be a corruption of Ch. syen-soyng there exists a suitable Chinese donor word and in other cases like MK sa . som ‘deer’ we cannot exclude borrowing from Tungusic. In the index OJ sisi ‘flesh, animal’ is related to the Korean word by Starostin incorrectly glossing it as ‘deer’, but the . Japanese word has a better cognate in MK solh ‘flesh’. Left with the binary comparison . between pTg *sesi-n ‘herd (of deer, wild animals)’ and MK sa som ‘deer’ and in consideration of the semantic sphere, borrowing becomes more likely. It is also probable that unlenited forms from surrounding dialects invaded in Middle Korean as might be .the case for MK . ka som ‘chest’ for which Martin mentions two early attestations of MK ka ∆om .32 Recently Vovin proposed a revision for Martin’s lenition hypothesis, in which an original consonant cluster *NC or occasionally *LT blocks the lenition, but as he admits himself ‘there is meager evidence to support the claim’ (sic.).33 First, this theory entails an enormous number of *NC clusters to be reconstructed in proto Korean, since every non-lenited obstruent must be derived from an underlying consonant cluster. Apart from some dialectal material, there is little evidence that systematically supports such clusters. Miyake (pc.) informs that he was unable to find any internal evidence for these *NC sequences in the early philological materials. Hence, I agree with Martin that the lenition of obstruents is an internal Korean development, without Altaic prehistory. Contrary to my past view, I am now reluctant to reconstruct pK *n! as the ancestor of the phoneme that is represented by MK ∆. This position will have consequences for my evaluation of the Altaic evidence proposed in the index. Instances in the index where the etymological proposals rely on the reconstruction of a palatal nasal n! in proto-Korean are ane ‘elder sister’, ani ‘elder brother’, hane ‘feather, plume, plumage’, hareru ‘swell’, hina ‘chicken, chick, doll’, imo@to ‘younger sister’, inabikari ‘lightning’, OJ in- ‘go, leave, depart’, inu ‘dog’, ira ‘thorn, irritation’, ka ‘mosquito’, kage ‘shadow’, kanemoti ‘man of wealth’, kinu ‘silk’, kisaragi ‘second month of lunar calender’, 29 Martin 1996, 31, 70. 30 Miyake 2001, 2. . 31 Ramsey 1978, 46-47; Martin 1996, 52. The only verb in the list is MK ka poyyaß- ‘be light (weight)’ which is thought to involve an etymological boundary. 32 Martin 1996, 99. 33 Vovin 2001 a, 124-25. Vovin 2001 b promesses typological and etymological evidence in a forthcoming publication.

64

Chapter 5. Phonological inventories

kogareru ‘pine for, have a burning love for’, kome ‘rice’, konomu ‘like, be fond of’, OJ ko2ru ‘freeze’, OJ koyi- ‘freeze’, OJ kozi- ‘dig up by the roots’, kumu ‘draw (water), mairu ‘come humbly, go humbly’, mamoru ‘protect, defend’, mitu ‘three’, mura ‘village’, mune ‘breast’, nazuna ‘shepherd’s purse’, naru ‘sound’, sane ‘kernel, core’, tane ‘seed’, tanosii ‘pleasant, joyful’, tono ‘lord, lordship, mansion, palace’, tinamu ‘be connected with’, uni ‘sea urchin, echinoid’, no genitive case marker. In these etymologies the suggested Korean cognate with pK *n! can be rejected. It is also speculated that Old Korean had a distinction between the liquid phonemes *r and *l. Yi Ki Mun is the first to observe the consequent use of two distinct Chinese characters and for occurrences of the liquid phoneme. 34 Therefore, he posits a distinction between *r and *l in Old Korean that merged as one liquid morpheme l in Middle Korean. Although Yi is correct in suggesting that a distinction existed in Old Korean, there is no evidence for postulating the precise nature of that phonemic distinction as *r and *l. Besides, as Miller shows, this distinction is problematic since the reading of the Chinese characters is *si without any hint of a liquid morpheme.35 Relying on internal reconstruction within Middle Korean, Ramsey finds evidence within Middle Korean -l verb stems that confirms an earlier distinction in the Korean . liquids.36 The vowel and pitch contrasts that Ramsey finds in the -l inserting (e.g. MK ni lu(l)- 'arrive' ) versus the l-doubling verb stems (e.g. hulu-/hull- 'flow') may have resulted from the loss of a distinction between two liquids. The accentual differentiation of l-extending verb stems (e.g. MK al- / a(l)- ‘know’) suggests that some l-extending verb stems behave like obstruents, while others behave like sonorants, but Ramsey does not establish the exact nature of the phonemic distinction. However, Miller’s hypothesis that OK *s# and *l merged into MK *l is more compatible with Ramsey’s findings than Yi Ki Mun’s reconstruction of OK *l and *r. The modern reinforced consonants (pp, tt, cc, kk, ss) all result from Middle Korean consonant clusters. At the time the Korean alphabet was introduced (1446) the newly recorded stage of the language, usually referred to as Late Middle Korean, was characterized by a richer inventory of consonant clusters than Korean is today. Changes in the clusters were already setting in and within a century many clusters had developed into reinforced consonants. The tensity which now distinguishes reinforced consonants from their plain counterparts (p, t, c, k, s) was still phonologically predictable in the 15th century. It was only after the obstruents which preceded them in the consonant clusters were lost, that . tensity became a distinctive feature. This is why we find MK kaska Wi for K kakkai ‘nearby’, MK ¨yetcop- for K yeccwup- ‘tell, inform’, MK . psol for K ssal ‘rice’.37 The creators of the Korean alphabet did not provide dinstinct symbols for the reinforced consonants because at that time they were not distinct phonemes yet.

34 Yi 1977 b, 95-96. 35 Miller 1979 a, 1979 c. Sasse 1981, 1989. Sasse has further refined Miller’s arguments and he has increased the examples in which later Korean language stages have a silibant reflecting an Old Korean word in which was attested. 36 Ramsey 1996. 37 Unger 1973, 8; 2000, 666; Whitman 1985, 7-13; Vance 1987, 108-109, 135, 140; Martin 1987, 20-25; Miyake 1999, 150.

5.2. Korean

65

The aspirated obstruents (ph, th, ch, kh) , however, were perceived as unit phonemes and were attributed distinct graphemes when the Korean alphabet was introduced. The aspirated obstruents contrasted with the plain obstruents (p, t, c, k, s,h), but in origin the aspirates were also complex. In a morphophonemic sense, the aspirated obstruents can be considered clusters of C + h, but a voiceless fricative *h may not be reconstructable for proto-Korean. One source for MK h is pK *s, but another source is pK *k.38 Nevertheless, it appears that the source of aspiration in Ch clusters is pK *Ck rather than pK *Cs.Whitman (1999) has presented internal evidence, restricting the development pK *s > MK h to a special vocalic environment when pK *s is followed by *-i- and the requirements for breaking of the *i are fulfilled. From Chinese donorwords corresponding to Korean loanwords . (e.g. OCh. *mrak ‘wheat’ is probably borrowed as pK *milk > MK milh ‘wheat’), phonogram readings in the Kyelim Yusa (e.g. ¨hwalq-huy for MK holk ‘earth’), elements in Paekcke placenames (e.g. tin-qak for ¨twolh ‘stone’), dialectal forms (e.g. dial. tolk for MK ¨twolh ‘stone’), and internal doublets (MK siphu- versus MK sikpu- ‘want’ ) it can be understood that velar lenition (*Ck > *Ch) has taken place in *Ck clusters at an early stage in Korean. Other evidence for the reconstruction of pK *Ck comes from Ramsey’s (1991, 230) distributional observation that obstruent clusters with k such as pk, tk or ck do not exist in Middle Korean, except for sk and psk. Since Middle Korean has no aspirated silibant (sh), the aspirated obstruents ph, th, ch, kh and the sk-clusters are in complementary distribution. From some materials preserved from the Silla period it can be understood that Old Korean had aspiration distinctions only in the dental stop *th and the affricate *ch. It is thus safe to assume that the development of aspiration is an internal Korean process and that the aspirated obstruent series developed out of the reduction of pK *Ck consonant clusters. Although some medial clusters are probably reconstructable for proto-Korean, others may be the result of an internal Korean development. Ramsey has shown that unaccented minimal vowels o and u rarely occur between voiceless obstruents in Middle Korean.39 Therefore, it seems probable that some of the Korean consonant clusters arose through minimal vowel syncope. (e.g. K pakk, MK pas ~pask ‘outside’ < *pas-ok or K aph, MK alph < *alpok with *-u/ok place suffix). Other medial clusters were morphologically generated through the insertion of the genitive marker MK s, e.g. MK twuystari ‘hind leg’ (< MK . twuy ‘behind’ +s+ tari ‘leg’) or through compounding, e.g. MK han¨∆wum ‘deep breath’ . (< han ‘big’ + ¨swum ‘breath’). Cross-linguistically clusters are relatively unstable and Korean is no exception to that observation. Cluster metathesis like in MK siphu- versus MK sikpu- ‘want’ can occur and the regressive assimilation of aspiration discussed below is a special kind of cluster metathesis. The liquid phoneme in lateral clusters sporadically drops, as can be observed in the . development from MK alph to K aph ‘front’ and MK polk- to K pak- ‘be bright’. Contrary to the distribution of proto-Japanese and Altaic clusters, aspirates and obstruent clusters do appear in initial position in Korean. Internal evidence, however, indicates that complex initials are secondary, internally generated through phonological or morphological developments. Korean forms with initial s- clusters are often derivations with the intensive 38 Ramstedt 1939, 17; Yi 1977, 83-84; Ramsey 1978, 51-52; 1991, 230-31; Martin 1996, 36-37; Whitman 1999. 39 Ramsey 1991, 221-22; 1997, 137.

66

Chapter 5. Phonological inventories

prefix pK *su- ~*so- . The initial s- in MK spolo- ‘be fast’, for example is a trace of this prefix as is indicated by MK polo- ‘early’ that occurs without the intensive suffix. In contemporary Korean both ‘early’ and ‘fast’ are represented by the reinforced adjective pparu-.40 One phonological cause of complex initials can be vowel syncope. Words with initial consonant clusters or aspirates are often accompanied by other phonological pecularities such as an exceptional pitch pattern, an unusual syllable structure or a special vocalism. Following Ramsey (1993, 438; 1997) verb stems with complex initials that are tonic and monosyllabic and have minimal vowels (MK o, u, i) are thought to be created through the . . loss of a first-syllable vowel. For MK tho- ‘ride’ and MK khu- ‘be big’, for example, there are additional phonogram interpretations in the Kyeylim Yusa that suggest the reconstruction *hoto- ‘ride’ and *hoko- ‘be big’.41Another phonological cause of initial aspirates is regressive assimilation of the aspiration or h-metathesis. It is the kind of assimilation that can be seen . in the development from MK kwoh toK kho 'nose’, MK polh to K phal ‘arm’, MK kalh to K khal ‘sword’etc. As a conclusion, I think that pK *n!, *z, *g, *d, *b, *s# and *l are critical phonemes for the Japanese-Altaic comparative enterprise. Experienced scholars have mined the internal Korean data thoroughly, without reaching a final conclusion. I believe that the main way to advance forward to a solution of the problem is to proceed to the comparative level. 5.2.2. Vowel inventory Table 9. Vowel inventory of pK front neutral

central

back

*i [i]

high

*e [e]

*u [ˆ]

*wu [u]

low

*a [a]

*o [E]

*wo [o]

following Miyake 1999 b

40 For other examples of verbs that include the Korean intensive prefix , I refer to Ramsey 1977 and 1997. 41 Yi 1991, 17-18.

5.2. Korean

67

Table 10. Vowel inventory of LMK front

central

back

high

i [i]

u [ˆ]

wu [u]

mid

e [e]

o [E]

wo [o]

low

a [a]

following Sohn 1999 (except for wo [o])

Middle Korean had a seven vowel system, in which each of the vowels required a separate grapheme. As for the reconstruction of proto-Korean vocalism there are basically two different viewpoints depending on whether or not a vowel shift has taken place at an earlier stage of the language. Yi Ki Mun believes that a vowel rotation took place after Early Middle Korean, whereas Martin reconstructs proto-Korean vocalism without any major changes.42 Martin’s reconstruction is supported by Miyake’s observations, taking SinoVietnamese into account as a third witness. Both advocates and opponents of the vowel shift agree, however, on the merger of o with other vowels and changes from complex vowels to monophtongs after Late Middle Korean. Middle Korean had vowel harmony, a phonological system in which the quality of the subsequent vowels depend on the first vowel in a word. The vowel harmony was not only at work inside an individual word, it also affected the behavior of suffixes. Vowel rotation advocates reconstruct a front-back vowel harmony in Korean, which may partly be motivated by the presence of front-back harmony in other supposed Altaic languages. But Miyake has argued that there is no need to force a front-back harmony into Korean since a simple two-way distinction high and low can be reconstructed for proto-Korean.

42 Yi 1977, 136. Martin 1998.

68

Chapter 5. Phonological inventories

5.3. Tungusic 5.3.1. Consonant inventory Table 11. Consonant inventory of pTg labial

dental

palatal

velar

voiceless stop/ affricate

*p

*t

*c#

*k

voiced stop/affricate

*b

*d

*Z#

*g

voiceless fricative

*s

nasal

*m

*n

liquid and glide

*w(?)

*r

*x *n! *l

*N

*y

following Benzing 1955

It is commonplace in the linguistic literature to use the cover term ‘Tungusic’ in reference to both the Manchu and Tungusic subgroups. The best known member of the Manchu subgroup is Manchu, an almost extinct language that was once spoken by the Qing dynasty rulers in China (1644-1911), but the subgroup also includes the Sibe dialect and the extinct Jurchen language of the Chinese Jin dynasty (1115-1234) . The Tungusic subgroup is traditionally divided into a Northern branch including Evenki, Ewen, Solon, and Negidal and a Central branch including Nanai, Ulcha, Orok, Oroch, and Udehe.43 In 1949 Cincius published an extensive study of comparative Tungusic phonology, to which Benzing added some revisions in 1955. For the present account we will follow Benzing’s consonant inventory of proto-Tungusic, adding the palatal nasal phoneme posited by Cincius. The phonemic inventories of all Tungusic languages are very similar. All languages show distinctive voicing, so voice distinction can safely be reconstructed for proto-Tungusic. Cincius reconstructs a series of palatal consonants (m!, d!, s!, n! next to Benzings c#, Z#, j), but 44 Benzing thinks that palatalization is secondary, resulting from a following i-diphtong. Therefore, he reconstructs diphtongs (*ia, iä, *io, *iö, *iu, *iü) consisting of an i-glide and a short vowel. However, the explanation of palatalized consonants along these lines, also requests positing diphtongs such as *ia@, iä¤, *io@, *iö¤, *iu@ consisting of an i-glide and a long vowel. Besides, the overwhelming majority of the examples are cases like pTg *panian ‘Schatten [shadow]’, *nio@ra- ‘vorangehen [walk in front]’, *niuNnia.ki @ ‘Gans [goose]’ in 43 Avrorin 1960, 3; Doerfer 1978; Whaley, Grenoble, Li 1999, 291. Vovin 1993 b revises the classification of the Tungusic subgroup on the basis of lexical data and suggests a division into three subgroups: Ewen; West Tungus (Evenki, Negidal, Solon) and East Tungus (Oroch, Ulchi, Nanai, Udehe and Orok). 44 Cincius 1949, 210-14, 250-51.. Poppe 1960, 157 also posits a palatal nasal phoneme in proto-Tungusic.

5.3. Tungusic

69

which the reconstruction of the i-diphtong is preceded by a nasal phoneme n.45 Benzing leaves the question whether dental n and palatal n! should be distinguished open. 46 On the basis of the awkward i-diphtong reconstruction and the numerous nasal examples, I tend to agree with Cincius as far as the reconstruction of a palatal nasal phoneme n! is concerned. However this leaves a relatively small number of case like pTg *bia@ga ‘Mond [moon]’, *gia@ ‘Gefährte [companion ]’, *sia@- ‘kauen [chew]’, and *mia@ban ‘Herz [hart]’ unexplained. It can also be remarked that the addition of a palatal nasal phoneme provides the consonant inventory with more symmetry. These observations indicate that the palatalization of Tungusic m!, d!, s!, is secondary due to a following high front vowel. The same is true for a number of cases with n!, although next to Tg. n! < pTg *n, there are cases in which Tg. n! is a reflex of an original palatal nasal pTg *n! . This is in agreement with Poppe’s view that ‘In Tg werden *n- und *n!- bis jetzt gut unterschieden, obgleich in einigen fällen dort eine Vermischung dieser beiden consonanten stattgefunden hat.’ (sic.).47Apart from the liquid phonemes, only n can occur word-finally. Whether it is due to language contact or to genetic inheritance is not the issue here, but it can be remarked that Japanese, Korean, Tungusic, Mongolic and Turkic share a common phonological feature, namely the absence of the liquids (l- / r-) in initial position. A striking exception, however, is the reconstruction of initial *l- in proto-Tungusic. It is possible to reconstruct two liquid phonemes *r and *l in proto-Tungusic, that can occur in medial and final position. Although pTg *r never occurs word initially, there are a number of words for which pTg *l can be reconstructed in initial position. Poppe has suggested, however, that pTg *l- is secondary, coming from an initial *n- more often than not followed by a labial nasal *m. 48 Benzing finds evidence for the opposite development pTg *l- > n- in all Tungusic languages, except for Nanai where l- is retained, when the initial liquid is followed by a nasal *m or * N. 49 Apart from the general structural argument about the absence of initial liquids in the North-East Asian language area, there is another reason to favor Poppe’s view. Agreeing with the reconstruction of a palatal nasal phoneme pTg *n!- in Tungusic, examples like Ma. lexe- ‘demand’, Ev. n!eke- ‘demand’, Neg. n!exe- ‘be busy, intend’ yield the reconstruction pTg *n!eke following Poppe’s view, but following Benzings view we would have to reconstruct an additional palatal liquid pTg *l’ to account for cases like this. The problem with Poppe’s view is that there are many cases in which there is no following *-m- in spite of the initial lin most of the Tungusic languages. Relevant entries in the covering index are MJ nagi ‘a kind of Chinese black pine with a sacred connotation’ with pTg *laka- ‘elm, oak’, OJ nakate ‘eel’ with pTg *laka ‘name of a fish, goby’ and pTg *lokija 'a k. of salmon', naku ‘cry, creep, wail, moan’ with pTg *ligi- ‘snore’, nao ‘further, furthermore, still, still more’ with pTg *lab-du ‘many, plenty’, nawa ‘rope, cord’ with pTg *lapV-ki ‘tiers, straps (for 45 Benzing 1955, 973-74. 46 Benzing 1955, 986: ‘Beim n muß vorerst die Frage offen bleiben, ob ursprübglich (dentales) *n und (palatales) *n! zu unterscheiden sind.’ (sic.). 47 Poppe 1960, 37. 48 Poppe 1960, 74: ‘Das anlautende l im Mandschu-Tungusischen ist sekundärer Herkunft und geht gewöhnlich auf ein anlautendes *n (meistens vor einem folgenden *m) zurück.’ 49 Benzing 1955, 993.

70

Chapter 5. Phonological inventories

skis)’, ne ‘sound, tone’ with pTg *leje-, neko ‘cat’ with pTg *lukV , nemu ‘silk tree’ with pTg *labikta / *lebukte ‘a k. of moss (cudbear), lichen, moss’, nigeru ‘flee, run away, escape’ with pTg *luktin- ‘to run some distance’, nigoru ‘become dirty’ with pTg *lakti- / *legdi- ‘to be burnt (of food), soot’, nikui ‘hateful, detestable’ with pTg *la@k - ‘difficulty, disorder’, niwa ‘garden, yard’ with pTg *lapta- ‘flat, level’, noboru ‘rise, ascend, climb’ with pTg *lep- / *lupu- ‘to raise hands, to jump out, to move out’, nori ‘laver, sloke’ with pTg *la@lbi-kta / *lelu-kte ‘moss’, nukaru ‘make a slip, blunder’ with pTg *largi@ ‘diligent, pleasant’, nuki ‘bar, brace, cross-beam’ with pTg *loka- / *laku- ‘hang, hanger’, nuki ‘woof, weft’ with pTg *luktu- ‘to sew in (ornaments)’, nuku ‘pierce, poke’ with pTg *lokto‘to pass through’, numa ‘swamp, bog, pond, lake’ with pTg *lebe@(n)- ‘swamp’, nuu ‘sew, stitch, embroider’ with pTg *lup- ‘prick’. Benzing’s proposal can account for these examples in which there is no following nasal in spite of the initial l- in most of the Tungusic languages. So, there is sufficient internal evidence to reconstruct inital *l- in proto-Tungusic. For the etymology of negau ‘desire, wish, beg’ I will regard the problematic Manchu initial l- in lexe- ‘demand’ as a secondary development and reconstruct pTg *n!eke on the basis of Ev. n!eke- ‘demand’ and Neg. n!exe- ‘be busy, intend’. The reconstruction of pTg is relatively clear for the above simple consonants, but in case of consonant clusters the reconstruction becomes more involved because processes of original vowel loss and assimilation can obscure the reconstructions. Proto-Tungusic is particularly rich of sonorant clusters, consisting of a sonorant (r, l, n and m before a labial stop) followed by a stop P (b/p), T (t/d) or K (k/g) such as *lp, *lb, *lt, *ld, *lk, *lg ; *rp, *rb,*rk, *rg; and *mb, *nt, *nd, *nk, *ng (or *N). Obstruent clusters of the following shape do also frequently occur. The obstruent clusters consist of a stop p, T (t/d) or K (k/g) followed by a dental (t/d) or velar stop (k/g) such as *pk, *pt, *tk, *dk, *gd, *kt. Although medial clusters are reconstructable for proto-Tungusic, some result from internal Tungusic developments. Some clusters are secondary due to the loss of a short vowel. Long vowels are relatively stable throughout Tungusic, but short vowels tend to change or to drop. The consonant sequence in Ma. nomxon ‘peaceful, friendly’ for example, does not reflect an original consonant cluster, but it is the result of vowel syncope. The original short vowel is still present in Na. nomoXon and Ev. nomoko@n and in the Mongolian loansource WMo. nomuVan ‘peaceful’. The instability of clusters can be observed in cases of cluster metathesis (e.g. Ud. tegbese- ‘protect’ versus Ev. tepke- ‘encase, cover’), assimilation (e.g. Orok nette‘spread out’ versus Ev. nepte- ‘spread out’) and the sporadic drop of liquid phonemes in lateral clusters (e.g. in Ma. dobo-ri ‘night’ corresponding to Neg. dolbon, Na. dolbo, Olc#. dolbo, Orok dolboni, Jur. dolwo, Ev. dolboni@ ). Initial consonant clusters are not reconstructable for proto-Tungusic.

5.3. Tungusic

71

5.3.2. Vowel inventory Table 12. Vowel inventory of pTg

high mid

front

central

*i

*ˆ *e

back

**ü *ö

low

*u *o

*a

following Benzing 1955

It is highly probable that proto-Tungusic had distinctive vowel length since vowel length is distinctive in the majority of the Tungusic languages spoken today. Vowel harmony was at work in proto-Tungusic, but there is a considerable degree in variation from language to language in the type of vowel harmony system. Traditionally Tungusic vowel harmony is referred to as a hard-soft harmony. The hard vowels are slightly lower and further back than the soft counterparts. The phonetic basis for the vowel harmony seems to be relative vowel height, the hard vowel being relatively lower than its soft counterpart, but not absolutely low within the whole system of vowels. It has been argued that the crucial feature in vowel harmony in Tungusic is not high-low, but RTR, the mechanism of the retraction of the tongue root.50 The RTR harmony requires the RTR value of root vowels to be aligned, and it determines the RTR value of most suffix vowels.

50 Ard 1980; Comrie 1981, 69-71; van der Hulst and Smith 1988; Li 1996; Whaley, Grenoble and Li 1999, 296.

72

Chapter 5. Phonological inventories

5.4. Mongolic 5.4.1. Consonant inventory Table 13. Consonant inventory of pMo labial

dental

palatal

velar

voiceless stop / affricate * p

*t

*c#

*k

*q

voiced stop / affricate

*d

*Z#

*g

*V

*b

*s

voiceless fricative nasal liquid and glide

*m

(*s#)

*n *l

*N *r

*y

following Poppe 1954 ; 1955, 95-171

The Mongolic languages can be divided into Mongolian proper and a number of isolated varieties of Mongolic, scattered as far west as Afghanistan. Mongolian proper consists of an Eastern subgroup with the standard Khalkha dialect, and related dialects Buriat, Chakhar, Urat, Ordor, and Kharchin and a Western subgroup being the Oirat dialects among which belongs Kalmuk. Many of the languages of Mongolian proper differ so little from each other that they could be thought of as dialects of the same language. The isolated Mongolic languages are Moghol, spoken in north-eastern Afghanistan, Dagur, and some Mongolic languages spoken in the Gansu and Qinghai region of China, such as Dongxiang, Bao’an, and Monguor. Written Mongolian, a language used by the Mongolians today for written communication, still makes use of the conventions and spellings of the earliest Mongolian written records and is remarkably archaic in quality. It is clear that proto-Mongolic had voicing distinction for its stops and affricates. An important feature at the basis of the classification of the Mongolic languages is the reflex of pMo *p-. The initial labial stop pMo *p- is preserved as f- in most Gansu-Qinghai dialects (e.g. Bao. felaN, Mgr fulaan ‘red’) and as x- in Dagur and Dongxiang (e.g. Dong. xulan, Dag. xulaan ‘red’), but it disappeared in the other dialects (e.g. WMo. ulaVan, Mogh. uloon, Kalm. ulaan, Khal. ulaan ‘red’). It is therefore safe to reconstruct pMo *p- in proto-Mongolic forms for which Starostin reconstructs pMo *h- (e.g. I reconstruct pMo *pulaVan ‘red’ instead of pMo *hulaVan ‘red’). WMo. p occurs only in foreign loanwords in initial and medial position. Poppe also reconstructs a pre-Mongolian *ß, which had 51 already converged with *V in proto-Mongolic. The reason for its reconstruction is purely external, based on comparison with Turkic p, as in WMo. qaVa- ‘close’ and Tk. qapa‘close’. 51 Poppe 1955 b, 98.

5.4. Mongolic

73

The deep velar consonant *q was a stop in proto-Mongolic, but is now pronounced as a fricative x. The deep velar consonant pMo *V is still pronounced as a stop in the modern dialects as it probably was in proto-Mongolic. The velar consonants are joined with only certain vowels: q and *V can be followed only by a, o, u and k, g can only be joined with e, ö, ü, i. WMo. b, d, g, V, s, m, n, N, l, r can occur in word final position, but the final voiced stops go back to final voiceless stops in proto-Mongolic, thus WMo. c#ig ‘direction’ < pMo *tik ‘id.’.52 Most of the contemporary Mongolic languages have a palatal sibilant s# resulting from an original pMo *s before *i (< *i or *ï). In Written Mongolian the palatal s# occurs also before vowels other than *i. Poppe attributes such occurrences of WMo. s# to the secondary origin of the following vowel and he assumes that this vowel developed from pMo *i.53 Another explanation for WMo. s# followed by a vowel other than -i- would be borrowing. Such may well be the case for WMo. s#atu ‘stair’ which is following Poppe a borrowing from Tk. s#atu ‘stair’. No palatal nasal can be reconstructed in proto-Mongolic, but Poppe remarks that it is possible to distinguish between pMo *n followed by *i and pMo *n followed by other vowels. He suggests that pMo. *ni reflects an earlier palatal nasal followed by any vowel *n!V.54 Proto-Mongolic distinguishes between the liquids *l and *r, but both liquids do not appear in word initial position. Initial l- and r- occur only in a small number of loanwords and onomatopoeia.55 Consonant clusters can occur in medial position, but there are no native Mongolic words that possess clusters word initially or word finally. Proto-Mongolic is particularly rich of sonorant clusters. Obstruent clusters are also reconstructable. The sonorant clusters frequently consist of a sonorant (r,l, n,m) followed by a stop P (b/p), T (t/d) or K (k/g) such as *lp, *lb, *lt, *ld, *lk, *lg ; *rp, *rb, *rt, *rd*rk, *rg;*nb, *nt, *nd, *nk, *ng (or *N(g)), *mb, *mt. The obstruent clusters consist of a stop P (p/b), T (t/d) or K (k/g) followed by a dental (t/d) or velar stop (k/g) or sibilant (s), such as *pk, *bk, *pt, *bt, *ps, *bs *tk, *dk, *gd, *kt. Medial clusters are sometimes formed through secondary developments such as vowel syncope (e.g. WMo. aluqan ‘hammer’ > Khal. alxan, Bur. alxa, Kalm alxE) or they can develop on etymological or morphological boundaries (e.g. WMo. ög- ‘give’ + -te- passive > WMo. ögte- ‘be given’). Consonant clusters are prone to metathesis and assimilation (e.g. pMo *todka- ~ *togta ‘stop, fasten’ that is reflected in WMo. todqa- ‘hesitate', Khal. totgor ‘fasten’, Kalm. totxE- ‘snare, trap’ versus WMo. toVta- 'to stop, establish', Khal. togta‘stop’, Kalm tokto- ‘stand firmly’) and sporadic elision of a liquid (e.g. the Mongolian doublet kelbej- ~ kebüji- ‘lean, be inclined to one side’).

52 53 54 55

Poppe 1960, 156. Poppe 1955 b, 125. Poppe 1960, 36-37, 155-56. Poppe 1955 b, 155, 160.

74

Chapter 5. Phonological inventories

5.4.2. Vowel inventory Table 14. Vowel inventory of pMo front

central

high

*i



mid

*e

low

back

*ï *ö

*u *o

*a

following Poppe 1954, 10; 1955, 25-58

The modern Mongolic languages have distinctively long vowels, though vowel length is always secondary. Long vowels can have two origins: either they arose by the loss of intervocalic consonants V, N, m that are still attested in written Mongolian or they developed from originally short vowels.56 Mongolian has front-back vowel harmony, having either front or back vowels in a word, but not both. The vowel harmony also affects the behavior of suffixes. The only exception to the system is the vowel i, which is not subject to these restrictions. Although i is pronounced as a front vowel in Mongolian, it is considered as a neutral vowel. In proto-Mongolic, however, complete vowel harmony is reconstructed.57 The vowel pMo *i had a back vocalic counterpart *ï. In written Mongolian and in the contemporary Mongolic languages pMo *ï has merged with *i, resulting in a single neutral vowel i.

56 Poppe 1955 b, 59-60. 57 Poppe 1954, 11; Poppe 1955 b, 33-36.

75 5.5. Turkic 5.5.1. Consonant inventory Table 15. Consonant inventory of pTk labial

dental

strong stop

*p

*t

*k

weak stop

*b

*d

*g

velar

*s *c#

fricative/ affricate nasal

palatal

*m

liquid glide

*n *l

*n! *r

*N

(*r2) *y

following Johanson 1998, 95

Although the Turkic languages are spread over a vast territory, they show a striking resemblance in many basic features. At an early stage the Turkic unity was divided into a western branch called Oghur and an eastern branch referred to as Common Turkic. The only modern representative of Oghur is Chuvash. The main feature at the basis of the distinction between Oghur and Common Turkic is the reflex of pTk *l2 and *r2 as l and r in Oghur and as s# and z in the other Turkic languages. Common Turkic further dissolved into Arghu, represented only by Khalay, spoken today in central Iran and into Old Turkic. Old Turkic is commonly divided into four branches: a southwestern, a northwestern, a southeastern and a northeastern branch. Southwestern Turkic or Oghuz Turkic comprises Turkish, Gagauz, Azerbaijanian, Turkmen, Khorasan, the dialects of Iran and Afghanistan. Northwestern Turkic consists of Kumyk, Karachay, Balkar, Crimean Tatar, Karaim, Tatar, Bashkir, Kazakh, modern Kirghiz, Karakalpak, Kipchak Uzbek and Noghay. Southeastern Turkic is subdivided into Uzbek, Uighur, Taranchi and several Eastern Turki dialects. From a linguistic point of view Salar can be considered a dialect of Uighur. Northeastern Turkic consists of Yakut and Dolgan, spoken in North Siberia and of different dialects of Sayan, Yenisey, Chulym and Altai Turkic spoken in South Siberia. Proto-Turkic had a fortis-lenis distinction for its stops though the actual phonetic features 58 corresponding to this distinction may differ from voiceless and voiced. Word initial *pdeveloped over a bilabial fricative into h- and finally disappeared in most of the contemporary Turkic languages. According to most Turkologists, the distribution of consonants in initial position seems to have been rather limited. It is generally assumed that initial *g- and *d58 Róna-Tas 1998, 71; Johanson 1998, 95. Róna-Tas reconstructs no weak counterpart for pTk *c# whereas Johanson mentions the possibility that a weak affricate *j# was present in proto-Turkic.

76

Chapter 5. Phonological inventories

disappeared in favor of *k- and *t- as a result of neutralization. It is thought that pTk *k and *g were realized as front or back according to the quality of the vocalic environment and later split into k versus q and g versus V. The medial weak *-b-, *-d-, *-g- developed by further weakening and fricativation. Contrary to the general assumption that initial *g- and *d- have completely neutralized in proto-Turkic, it is argued in the Altaic dictionary (2003) that the distinction between pTk *k- and *g- and between pTk *t- and *d- is reconstructable on the basis of the Oghuz reflexes.59 The internal basis for the reconstruction of pTk *d- are cases in which Turkish, Azerbaijanian and Turkmen have a d- reflex, whereas the other Turkic languages have a treflex. For example, under the entry yotu ‘four’ pTk *dö¤rt ‘four’ is reconstructed instead of pTk *tö¤rt , in spite of the voiceless initial in OTk. tört ‘four’ and Chuv. ta#vada# ‘four’, on the basis of the voiced initial in Tk. dört, Az. dört, Tkm. dö¤rt ‘four’. According to Starostin, Dybo, Mudrak (2003, 67, 70-72) *g- and *k- were neutralized to pTk *K- before back vowels, but the distinction between pTk *g- and *k- can be reconstructed before front vowels. The internal basis for this reconstruction are cases in which Turkish, Azerbaijanian and Turkmen have a g- reflex before a front vowel, whereas the other Turkic languages have a k- reflex. A relevant example can be found under kuru ‘come’ with pTk *ge ¢l‘come’ instead of pTk *kel- on the basis of OTk. kel-, Chu. kil-, Tat. kil-, Tk. gel-, Az. gäl-, Tkm. gel- ‘come’. Besides the actual voiced reflexes in Oghuz, the reconstruction of pTk *d- and pTk *g- in this position can be supported by the observation that Oghuz languages tend to be more conservative in general, by the consideration of phonological symmetry including word-initial position in proto-Turkic, and by the external comparison with Tungusic and Mongolic languages that distinguish *d- ~ *t- and *g- ~ *k- initially. On the other hand, there are a number of counter-arguments that strengthen the traditional Turkological viewpoint that the d- and g- reflexes in Turkish, Azerbaijanian and Turkmen are secondary. First, the observation that only three Turkic languages, belonging to a single branch, Oghuz, preserve a reflex of pTk *d- or *g- is in conflict with the majority-wins principle of linguistic reconstruction. Besides the oldest stages of Turkic preserve t- and kin the position where Oghuz has g-. Chuvash, the only modern representative of Oghur as well as Khalay, spoken today in central Iran and the only descendant of Arghu, preserve tand k-. Second, the preservation of d- and g- in Oghuz, shows many exceptions. For example under isi ‘stone’ the Turkish reflex tas# ‘stone’ has a voiceless initial, while Az. das# and Tkm. da@s# ‘stone’ have a voiced reflex. Under kumu ‘draw (water), dip up, pump’, we find Tk. göm- and Tkm. göm- ‘dig’ with a voiced reflex, while Azerbaijanian köm- ‘dig’ has a voiceless k-. This raises the suspicion that we are dealing with sporadic, secondary voicing. Another hint that the voicing is secondary comes from foreign loanwords in Turkic. It is suggested by Doerfer (1967, 615-616) that the Turkic words for ‘parasite, tick’, Tk. gene, Tkm. gä¤nä, Tkm. dial. kä ¤nä, Khalaj kana and Kaz. kenä are borrowed from Persian kana ‘tick’. Doerfer (1995, 259) mentions a Turkic loan dümen ‘rudder’ from Greek timoni ‘rudder’. Since the Turkic phonological system permits an imitation of these words with an initial voiceless k- and t-, the voicing of the initials in the Oghuz forms are probably secondary developments. Although relying on external comparison with Mongolic and Tungusic would be methodologically circular, the distribution of pTk *d- ~ *t- and pTk 59 Vovin 1994 a/b/c; 2003 a, 24 agrees with the reconstruction of pTk *d- in e.g. pTk *dial2 ‘stone’.

5.5. Turkic

77

*g- ~ *k- as it is proposed by the Russian school is not supported by that of Mongolic and Tungusic putative cognates with *g- and *k- leading to at least a three-fold consonant distinction in proto-Altaic. Therefore, I follow the traditional Turkological approach that we can only reconstruct the voiceless initials pTk *t- and pTk *k-. Relying on internal data for the reconstruction of individual proto-forms, all initial dental and velar stops will be rendered with pTk *t- and pTk *k-, independent of the external cognates. Proto-Turkic consonants not occurring in word initial position are *n-, *m-, * N- *l-, *r-. Proto-Turkic had a palatal nasal in medial position, but the reconstruction of pTk *n!- in initial position is still controversial. Altaicists project the initial palatal nasal back to Altaic. The initial palatal nasal in Altaic is assumed to have developed into pTk *y-. Etymologies in the index that rely on the reconstruction of pA *n!- as an ancestor for pTk *y- can be found under the etyma mama ‘as it is, according to’, mamidiro ‘a Siberian thrush’, maseru ‘be precocious, be too grown up (for one’s age)’, me ‘eye’, minami ‘south’, mizikai ‘short, brief’, momidi ‘maple tree’ and natu ‘summer’. However, Hungarian nyár ‘summer’ is assumed to be borrowed from pTk *n!ar2 at an early stage and if this is the case, it follows that the initial palatal nasal must be reconstructed in proto-Turkic. Although in this case pTk *y- might represent a secondary development from pTk n!-, the prevalent opinion 60 among Turkologists is that there are also cases in which pTk *y- is primary. Some phonemes can be called critical, if not for their relevance to the actual reconstruction of the original language, then for the longlasting debates they have caused. This is certainly true for the reconstruction of a four fold liquid distinction in proto-Turkic. Chuvash occupies a central place in this discussion since in certain cases Chuvash r and l are found instead of z and s# in other Turkic languages. The classic examples are Chu. s!e #r versus Tk. yüz ‘hundred’ and Chu. c #ul versus Tk. ta@s# ‘stone’ There are also cases such as Ch. yur and Tk. yer ‘snow’ or Chu. s!ul and Tat. yul ‘way’ in which Chuvash r and l correspond to Common Turkic r and l. This can be taken as internal evidence for a proto-Turkic liquid system in which apart from the liquids *r and *l two additional liquids *r2 and *l2 were present. The hypothesis that pTk *r2 became *r in Oghur Turkic and *z in Common Turkic is referred to as ‘zetacism’, whereas the hypothesis that pTk *l2 became *l in Oghur Turkic and *s# in Common Turkic is called ‘sigmatism’.61 The postulation of zetacism and sigmatism underlies numerous Altaic etymologies referred to in the index. The reconstruction of pTk *r2 is relevant for the etymologies of arai ‘rough, coarse’, arui wa ‘or, rather’, atarasii ‘new’, hiroi ‘wide’, hitudi ‘sheep’, horobiru ‘perish’, hurui ‘old’, ira ‘thorn’, iraka ‘roof tile’, ireru ‘to insert’, iro ‘color’, iru ‘cast’, kara ‘emptiness’, kara ‘nutshell’, kari ‘wild goose’, katana ‘sword’, kati ‘walking’, kiru ‘make a fire by rubbing sticks together’, kokoro ‘heart’, OJ koti ‘east wind’, kurai ‘dark’, kuru ‘reel, wind’, kuruma ‘vehicle, car’, mare ‘unusual’, natu ‘summer’, nureru ‘get wet’, siru ‘know’, sitataru ‘drip’, tatu ‘stand’, tura ‘face, surface’, OJ tura ‘row, line’, turai ‘bitter’, MJ tusim- ‘get black and blue spots on the skin’, tuti ‘earth’, tutu ‘pipe’, ura ‘creek, bay’, ura ‘divination’, uti ‘house’, yorodu ‘myriads’. The reconstruction of pTk *l2 is relevant for the etymologies of asi ‘foot, leg’, asobu ‘play’, OJ azu ‘cliff, precipice’,esa ‘food, bait’, hasira ‘pillar’, husa ‘tassel’, husi ‘brushwood, 60 Johanson 1998, 96-97. 61 Ramstedt 1957; Poppe 1960, 74-82; Pritsak 1964; Doerfer 1975-76, 1984; Tekin 1969, 1975,1979,1986; Miller 1999.

78

Chapter 5. Phonological inventories

fence’, husi ‘joint’, husuma ‘quilt, bedding’, isago ‘sand’, isao ‘great service’, isari- ‘fishing -’, isi ‘stone’, iso ‘rocky beach’, kasa ‘scab’, kase ‘shackles, fetters’, kasira ‘head’, kasiwa ‘oak’, kasuru ‘graze, rub’, kisaragi ‘the second month of the lunar calender’, kosi ‘hips’, kosu ‘cross’, kosu ‘filter’, kosuru ‘rub, scrub, scrape’, OJ kususi ‘strange’, masu ‘a measuring box’, musasabi ‘giant flying squirrel’, na ‘greens, vegetables’, nasi ‘pear’, osieru ‘teach’, sasu ‘pierce’, OJ sisi ‘flesh, animal’, wasureru ‘forget’, yasiro ‘shrine’. Japanese is thought to be particularly telling for the reconstruction of an original four fold liquid distinction in proto-Altaic, since it is the only branch apart from Turkic which is believed to preserve evidence for an original *l versus *l2 distinction in Altaic. Another opinion, which is generally referred to as ‘rhotacism’ and ‘lambdacism’, is that pTk *z and *s# shifted to r and l in certain phonetic environments in the ancestor of Chuvash, but remained z and s# in other Turkic languages.62 But switching back and forth between internal and external evidence, the latter opinion would require that the corresponding Mongolic, Tungusic -and if they would indeed be related, Korean and Japanese- forms all shifted z to r and s# to l independently like what happened in Oghur. From the viewpoint of economy of sound change this is an unlikely explanation. Besides there are some Oghur loanwords in Hungarian and proto-Turkic loanwords in Samoyed that do not have z and s#, but r and l. An internal consideration is that many Turkic words with word final -z, have morphologically related doublets such as OTk. köküz ‘breast’ and kökräk ‘breast’ or Tk. göz ‘eye’ and görmek ‘see’ in which -r- is kept intact before or after a consonant. It is 63 probable that the preceding or following consonant blocked the expected sound change. Internal doublets of this kind might raise the question whether pTk *-z- could have developed from a liquid r-cluster, parallel to the derivation below of pTk *-s#- as an earlier lateral cluster. But this is a Turkological question that is not within the scope of the present work, so for the time being I will accept the reconstruction pTk *r2 as suggested in the mainstream ‘zetacism’ theory. A third, intermediate viewpoint is that the twofold distinction in the r-sounds may indeed refer to pTk *r1 and *r2 but that this does not automatically imply that the same distinction holds for the laterals. The very existence of pTk *l2 has been called into question by Street by what he calls ‘the lateral-cluster fusion theory’.64 The theory is based on the derivation of Tk s# from lateral clusters pTk *lC, a sound change that Ramstedt once peripherally mentioned.65 Internal evidence for such a derivation can be found in words like Tk. bes#ük ‘cradle’ that derives from the verb bele- ‘swaddle’ and is borrowed in Hungarian as bölc#ö. The same historical explanation may account for semantically related verb stems for which a verb with stem-final -s#- alternates with stem-final -l-, like Tk. to@l- ‘be filled’ versus tos#- ‘fill, be full’ and Tk. bul-Va- ‘confuse’ versus bus#- ‘be irritated’. By reassigning pTk *l2 to early clusters of pTk *l with a following consonant (*b, *c, *Z#, *j) some puzzling phonological doublets in Turkic can be explained. Street further presents external evidence 62 Ligeti 1980; Róna-Tas 1998, 71-72; Doerfer 1963, 98-100. Doerfer used to adhere the hypothesis of ‘rhotacism’ and ‘lambdacism’ for proto-Turkic until the seventies, when he changed his view. 63 Tekin 1986, 142. 64 Street 1980. 65 Ramstedt 1957, 122-23.

5.5. Turkic

79

in the form of Mongolic or Tungusic words that have an -lb- cluster where Turkic has -s#-: Tk. qas#uq ‘spoon’ and Mo. qalbuVa ‘spoon’; Tk. qos#uV ‘poem, song’ and Mo. qolbuVa@n ‘verse, alliterative words or phrase’; Tk. es#ü- ‘cover’ and Ev. elbe- ‘cover, roof a tent’ etc. Suggesting that some of these words are early borrowings from Turkic into Mongolic or the other way around does not alter the ground of the argument that pTk *s# developed from a labial cluster. So far I am inclined to agree with Street’s lateral-cluster fusion theory, but in paragraph 8.1.2.3. I will study external comparative data that could throw a light on the problematic reconstruction of pTk *l2 . For the purpose of the present study I will agree only with the reconstruction of two r-sounds in Turkic: pTk *r and *r2 . As for the actual phonetic features of the pTk *r2 it possibly represents some retroflex or palatal liquid sometimes represented as *r¢ or *r! but what matters more than the exact pronunciation is the fact that we are dealing with a phonologically distinct set of r-sounds as indicated by the symbols *r and *r2. Except for the lateral-cluster fusion theory, the reconstruction of medial clusters in proto-Turkic is reminiscent of that in Korean, Tungusic and Mongolic in the sense that its cluster inventory consists mainly of sonorant clusters (r, l, n, j followed by an obstruent) and that obstruent clusters can also occur. Clusters are not possible in initial position and they tend to be unstable due to assimilation (e.g. Karakh. üple- ‘rob’ versus Tuva üpte-‘rob’), metathesis and consonant loss ( e.g. OTk. qapqaq ~ qapaq ‘a cover’). Although Old Turkic has final clusters, ultimately they may be the result of secondary developments such as final 66 vowel loss. 5.5.2. Vowel inventory Table 16. Vowel inventory of pTk

high mid low

front

central

*i

*ü *e

back

*ï *ö

*u *o

*a

following Johanson 1998, 89

Many Turkic languages have eight vowel phonemes, which are supposed to have occurred in proto-Turkic. But it is sometimes assumed that proto-Turkic had a ninth vowel phoneme, a short closed *eW next to the open *e. 67 This vowel has no connection with the closed eW that is found in some modern languages since this eW is an allophonic variation of e. It is still unclear, however, whether other sounds in the later languages reflect pTk *eW. Proto-Turkic

66 Johanson & Csató 1998, 73; Poppe 1960, 83. 67 Róna-Tas 1998, 70; Johanson 1998, 89.

80

Chapter 5. Phonological inventories

had distinctive vowel length, which is reflected by primary length distinctions in Yakut, Turkmen and Khalaj. Although Old Turkic used graphical conventions that display front-back harmony at the earliest recorded stages of the language, there are also disharmonic sequences in the oldest Turkic words. Some Old Turkic suffixes have an invariable non-harmonic form. The frontback harmony was still under development and further widened after Old Turkic. Although Turkic front-back harmony is most manifest in the selection of vowels to match the preceding syllable, it can occasionally also affect the consonants. The front consonants k and g occur in front syllables, whereas q and V appear in back syllables. Not all consonants, however, have clear-cut front and back variants.

Chapter 6. Internal evidence contradicts the etymology In the past it has been suggested that internal reconstruction should be undertaken first and the comparative method applied afterwards. But linguistic scholarship today agrees that comparative and internal reconstruction stand in complementary relationship to one another. Etymology is a two way traffic, going from the individual languages to the ancestral language, and back again. Nevertheless, when the comparative method is applied before internal analysis, it is important to check that the individual proto-forms that are being compared to each other are legitimate reconstructions. For every Japanese etymon in the etymological index full reference is made to Martin’s internal analysis of the word in question. It is reassuring to find a serious number of Japanese etyma for which the internal reconstruction and the comparative method lead in the same direction. Unfortunately this is not always the case. The attached etymological proposals for Japanese are apparently no exception to Meillet’s general dictum that “Tout mot n’ a pas droit à une étymologie, comme on le croirait parfois à parcourir les dictionnaires étymologiques.”. 1 In the following chapter I will sort out those Japanese etyma that are not entitled to an external etymology because an alternative internal etymology is available in linguistic literature. If the alternative internal etymology is more or equally convincing than the external proposal, the comparison in question becomes too suspect to be allowed into the core of sound etymological proposals. Regarding ‘reasonable suspicion’ as a valid criterion, I will omit etymologies that have an alternative internal analysis from further consideration. 6.1. Lexicon 6.1.1. Dubious morphological analysis 6.1.1.1. Undetected morpheme boundary Compared words are sometimes analyzed as being composed of a single morpheme, whereas internal evidence argues for a morpheme boundary. As far as Japanese etymology is concerned, undetected morpheme boundaries constitute the most by far frequent case of conflicting internal evidence. Many Japanese lexemes of one or two syllables are single morphemes, but longer words often consist of two or more elements. Although the etymological structure of longer etyma is not always clear, indications for segmentation can be found in grouping Japanese lexical items consisting of similar morphemes. The appropriate morphological analysis leads to a segmentation into two or more morphemes, which occasionally reflect an etymon already listed elsewhere in the etymological index. An argument that Starostin frequently uses against the internal analysis of Japanese etyma that is proposed by Japanese or western linguistic scholarship in the past is that the 1

Meillet 1925, 38-39.

82

Chapter 6. Internal evidence contradicts

proposed analysis is ‘a folk etymology’. However, there is something obscure about the way Starostin uses this linguistic term. He seems to imply that ‘a folk etymology’ is an incorrect internal analysis of a word that is suited to the tastes, the educational level and the fantasy of the general public. However, there is nothing ‘folk’ and nothing ‘etymological’ about ‘folk etymology’. Folk etymology is a linguistic term for iconic remodeling of lexical items.2 It means that unfamiliar shapes are replaced by familiar ones. Loanwords are often subject to this kind of reinterpretation or adaptation. A word like Latin asparagus gave for example way to sparrow grass in English. So, we say that sparrow grass is a folk etymology. If Starostin is right in analyzing cases like isago ‘sand’, karamusi ‘Chinese silk plant, ramie’, kinoko ‘mushroom’, kitaru ‘come, arrive’, kotohogu ‘congratulate’, kutuwa ‘bit’, nuki ‘bar, brace, cross-beam’ etc. as folk etymologies, it follows that the words are iconic remodellings of earlier forms. And if they cannot be dismissed on suspicion of borrowing, they must be dismissed because they are phonological reinterpretations and do not reflect regular sound correspondences. Therefore, Starostin’s ‘folk etymology’ can be taken as an alarm bell, indicating that it is safer to omit the entry in question from further consideration. 6.1.1.1.1. Nouns Covert compound or nominal derivation Regarding recently coined English words such as snailmail or server, it is perceivable that compounding and nominal derivation are extensively used in order to create new words reflecting our constantly changing world and Japanese is no exception to that observation. Below I list etyma that are compared as a single noun in the etymological index, whereas internal evidence can be found in favor of an analysis as a covert compound or a nominal derivation. For segments that are not attested as individual lexical items in Japanese, I will provide the evidence on which the reconstruction of the morpheme is based. The dividing line between compounds and derivations is rather weak since it is not always clear whether a derivational element should be regarded as a suffix or as a bound noun. As far as the following etyma are concerned, there is a derivational morpheme -*ra that appears frequently. From the Old Japanese period onwards the suffix -ra is used to mark the plural. The presently unproductive suffix or bound noun *-ra which appears in OJ iku-ra ‘how much, some amount’, nani-ra ‘what, some’, koti-ra ‘this direction/ person’, ati-ra ‘that direction/ person’ and doti-ra ‘which direction/ person’ is not a plural suffix, though it is etymologically related. Suffixless alternations or variations with another formant of these pronouns can be found in a number of compounds with OJ iku- ‘how much’, in OJ nani ‘what’, in the set of demonstratives without the suffix and with compensatory lengthening of -t-: kotti, sotti, dotti. In the pronominal system we find yet another indication that pJ *-ra ‘(defined) quantity, group’ was a formant used to derive both plural and singular countable nouns, rather than only a plural marker. Traces of this pJ *-ra as are present in the first person pronoun OJ wa-, ware ‘I, we’ < *wa -ra -(C)i, in the second person pronoun OJ na-, nare ‘you’< *na -ra -(C)i, in the third person pronoun OJ so2-, so2re ‘he, she, it, they’ < *so -ra -(C)i and in the reflexive OJ ono2-, ono 2re < *ono -ra -(C)i. Other examples of an unmarked noun in variation with a noun on -ra ~ -ro(2) are present in a number of body part terms, such as OJ kabu ‘head, top’ versus OJ kabu-ro ‘short hair worn untied by a little boy or 2

Anttila 1989, 92-93.

6.1. Lexicon

83

girl’, kasi-ra ‘head, hair of the head, leader’ versus OJ kasiko1- ‘clever’ and subera-kasi ‘a woman’s hair tied at the back of the head and hanging down’, kobu-ra ‘calf (of the leg)’, siri-kobu-ra ‘the buttock’ versus kobu ‘gnarl, lump’, koko-ro ‘heart’ versus MJ koko-ti ‘heart, feelings, mood’ and in uncountable or quality nouns having a more concrete counterpart derived by -ra such as OJ no1 ‘field’ versus OJ no1ra ‘the fields’, OJ utu ‘hollow’ versus J utu-ro ‘a hollow, cavity’, OJ kapa ‘river’ versus OJ kapa-ra ‘a dry river bed, a river beach, flat place close by a river’, saba ‘mackerel’ versus sawa-ra ‘Spanish mackerel’, *na- ‘earth, ground’ versus OJ na-ra ‘Nara, the Land’, OJ taka- ‘high, elevated’ versus OJ taka-ra ‘treasure’, toko ‘place’ versus tokoro ‘place, spot, scene, site’, yo ‘evening, night’, OJ yo 1ra, yoru ‘evening’ etc. The examples of kaburo, kokoro, tokoro and uturo, yoru seem to indicate assimilation of an original final *-ra to the preceding vowel. Maybe we can also include deverbal nouns as OJ tatara ‘foot-bellows’ versus OJ tat- ‘stand’, OJ saku-ra ‘blossom’ versus OJ sak- ‘bloom’, OJ maku-ra ‘cushion’ versus OJ mak- ‘roll’, MJ tamure and tumure ‘hill’ versus tum- ‘heap up’, MJ fukura ‘swelling’ versus OJ puk- ‘blow’. Although Martin proposes that -ra ‘quantity, group’ is a truncation of a still earlier deverbal *ara- ‘what exists’, this reconstruction does not alter the ground of the argumentation, namely that a number of nouns on -ra should be analyzed as derivations instead of a treatment as morphemes that are comparable as a whole. 3 It can also be remarked that a number of the non-initial elements of the covert compounds have a voiced initial that reflects earlier prenasalization. This inserted nasal probably represents a nasal genitive *-n-. I propose to omit following etyma from the core evidence since they are treated as single nouns in the etymological index, whereas internal evidence argues for segmentation into distinct morphemes. 1. OJ agi1 ‘you’ (informal second person pronoun) This pronoun is probably a compound of the first person pronoun OJ a, followed by a nasal infix and the suffix -ki, OJ -ki1 that is used for ‘male’. The suffix -ki ‘male’ occurs in complementary distribution with -mi ‘female’ as in OJ oki1na ‘old man’ and OJ omi1na ‘old woman’ and in names for gods and goddesses like Izanagi and Izanami or Kamurogi and Kamuromi. We also find an alternation between sumera and sumeragi ‘emperor’, leaving a trace of the original suffix for ‘male’. 2. ane ‘elder sister’; ani ‘elder brother’ Both etyma are related, perhaps in the sense that ane ‘elder sister’ can be derived from ani ‘elder sibling’ + me1 ‘female’. From the viewpoint that shorter versions ni(i) ‘elder brother’ and ne(e) ‘elder sister’ are attested in Japanese, the possibility cannot be excluded that the first morpheme ani is further derivable from OJ (w)a first person pronoun and ni ‘elder sibling’. Apart from those internal considerations the etymon must also be omitted from the core evidence because like many kinship words it can be considered a nursery term.

3

Martin 1987, 514, 805; Vovin 1997, 289. Vovin rejects Martin’s reconstruction of deverbal *-ara and reconstructs pJ *-ra in the function of plural marker.

84

Chapter 6. Internal evidence contradicts

3. arare ‘hail, hailstones’ The word seems to be a derivation of the adjectival noun araara ‘roughly’, which is itself a reduplication of the adjective ara- ‘rough’, and a nominalizing formant -(C)i . 4. ato ‘back, rear’ The word might be a covert compound from asi ‘foot’ and to ‘trace’. The latter morpheme is attested in isolation as OJ to2 ‘footprint, trace’ and in the compound OJ tomi1 ‘game tracker’. The analysis leaves room for the comparison of OJ to2 ‘footprint, trace’ with MK . ¨tuyh ‘rear’, but the derivation of MK ¨tuyh ‘rear’ . from the deverbal noun MK two li ‘turning around’ from MK ¨twol- ‘turn’ (< pK *two lo-) following Martin contradicts the latter comparison.4 5. aziro ‘wickerwork, a mat of split bamboo, a reed mat’ In Old Japanese the word is attested as ami1siro 1 with the Chinese characters for ‘net’ and for ‘white’ , which has triggered the traditional etymology am(i)-siro ‘netting stuff’. But if the first morpheme is a derivation of the verb OJ am- ‘knit’, we expect low register, while aziro is high. Martin’s derivation of aziro as a compound of asi ‘reed’ and musiro ‘straw mat’ accounts for both the semantics and the register and makes Miller’s external comparison unlikely. 6. MJ figure ‘cypress plank or stave’ The word is a compound of OJ pi1 ‘cypress’ followed by a nasal genitive infix and OJ kure ‘log as it is brought down out of the mountains’. 7. MJ fitume ‘hard and horny hoof at the extremity of the legs of horses and cows’ In consideration of tume ‘nail, claw, talon, hoof’ we must probably insert a morpheme boundary after fi-. The identity of the first morpheme remains unclear. Martin has suggested that it results from contraction of hira- ‘flat’, but another speculation might be a morpheme pJ *pi ‘horn, keratin’ which perhaps underlies in hige ‘beard’ (if from pJ *pi ‘horn’ and pJ *ka-(C)i ‘hair’), hituzi (if from *pi ‘horn’ + *tu genitive + *n genitive + pJ -si ‘ox, animal’), hizu ‘edible soft bone of salmon head (if from pJ *pi ‘horn’ and pJ *-du ‘head’). Another possibility is that the first morpheme is hi- ‘extreme’, which can be supported by hiideru 'excel' (< pJ *pi- 'extreme’ + ide- ‘emerge’). However speculative the identity of the first morpheme, it is clear that the second morpheme is tume ‘nail, claw, talon, hoof’ and the insertion of a morpheme boundary is in conflict with Starostin’s external comparison. 8. hako ‘box’ The word can possibly be derived from huta ‘lid’ and -ko ‘basket’, the second member of the compound being the same morpheme as attested in mokko ‘straw basket for carrying earth’, warigo ‘a partitioned lunch box’ etc. 9. hana ‘flower, blossom’ Perhaps the word can be derived from pJ *pa- ‘red’ and *na ‘plant’. The first morpheme is 4

Martin 1996, 38, 40.

6.1. Lexicon

85

supported by covert compounds such as hani ‘red clay’ (< pJ *pa- ‘red’ and *ni ‘earth’), beni ‘rouge’ and hena ‘red clay’. The latter morpheme is na ‘plant, greens, vegetables’. 10. haniwa ‘burial mound figurine, clay image’ This word is probably a compound of hani ‘red clay’ < pJ *pani ‘red clay’ and ba ‘place’ < pJ *(-n-)pa. Other compounds with hani-, OJ pani- ‘red clay’ are OJ panimono ‘clay figurine’ and OJ panima ‘horse of clay’. 11. hane ‘feather, plume, plumage’ This word is probably a compound of ha ‘feather’ and ne ‘root, base’. 12. hatake ‘field, farm, plantation, garden’ Although Miller suggests that this word is a borrowing from a still undocumented Korean original reflected in K path ‘field’, this is unlikely since there are strong indications that the word is morphologically complex and can be derived from the noun hata ‘field’ and a place suffix -ka followed by the nominalizing -(C)i. The place suffix -ka frequently occurs in Japanese compounds such as arika ‘whereabouts’, sumika ‘residence’ etc. 13. hatake ‘scabby eruption, scabs’ The attestation of fadake with a voiced -d- in Middle Japanese makes the segmentation into hada ‘skin, flesh’ and a deverbal noun kaki from kaku ‘scratch’ acceptable. Sporadic -kloss is seen elsewhere in Japanese. If the internal analysis is incorrect, the entry should still be dismissed from the core evidence since it belongs to the semantic sphere of medical vocabulary (8.2.3.). 14. hayati ‘autumn wind, wind storm’ This word is a compound of hayai ‘quick, fast, early’ and ti ‘wind, direction’, which is also attested in OJ koti ‘east wind’ below. 15. heya ‘room, chamber’ The word could be a derivation of an undefined first member with ya ‘building, house’. If the first member is the same morpheme as occurs in hedataru ‘be apart from’ and henaru ‘be apart’, then it goes back to OJ pe1- ‘apart’. 16. higurasi ‘clear-toned cicada, evening cicada’ The word can be derived from hi ‘sun’ followed by the insertion of a nasal marker and a deverbal noun of the verb kurasu ‘live, see the days through’. 17. hima ‘crack, gap; time, leisure’ Although there is evidence for the insertion of a morpheme boundary on the basis of the segment -ma ‘space’, I find only weak internal evidence to reconstruct *pi- ‘empty, separated’. It concerns a number of parallel structures, such as sukima ‘gap, crack’ derived from a deverbal noun from suku ‘become empty’ and -ma ‘space’ and hazama ‘interval’ derived from the naked verb root underlying hazeru ‘burst open, separate’ (< pJ *pansa-) and -ma ‘space’. Perhaps related to *pi- ‘empty, separated’ is OJ pe1- ‘apart’ (< pJ *piCa-) and hina

86

Chapter 6. Internal evidence contradicts

‘countryside’ as a compound of pJ *pi- ‘separated, remote’ and *na ‘ground’. If this word indeed is a covert compound of ‘empty, separated’ and ‘space’ the external comparison proposed by Starostin, Dybo and Mudrak is invalid. The insertion of a morpheme boundary supports Martin’s etymology. 18. hina ‘countryside’ Following the internal analysis of hima ‘crack, gap; time, leisure’ under the previous item, the word can be segmented into pJ *pi- ‘separated, remote’ and *na ‘ground’. 19. hisasi ‘canopy, penthouse, vizor’ The word can be derived from hi ‘sun’ followed by a deverbal noun of sasu ‘shine upon’. 20. hisio ‘salted meat’ In view of sio ‘salt’ this word may incorporate sio as the second morpheme. Although the identity of the first morpheme is unclear, it could be a nominalization of a verb underlying the transitivity pair hiru ‘get dry’ / hosu ‘dry’. 21. hirame ‘flatfish, halibut, sole’ The word can be derived from hira- ‘flat’ and -me ‘bird, animal that lives in flocks’. The suffix is mainly found in bird names like kamome ‘sea gull’, sime ‘hawfinch, grosbeak’, suzume ‘sparrow’ etc. 22. hokori ‘dust’ The word could be a derivation of an undefined first member with a deverbal noun of okoru ‘arise’. A possible candidate for the first morpheme is OJ po1 ‘ear of grain’ but this remains uncertain. Whatever the identity of this morpheme is, the internal analysis is convincing enough to cast doubt on the proposed comparison. 23. honoo ‘flame, blaze’ This word appears as OJ pono2po and is probably a compound of the adjectival noun hono-, OJ pono- ‘faint, slight’ and OJ po- ‘fire’. The former morpheme is also present in honoka ‘faint, slight’, hono-jiroi ‘dimly white’, honomeku ‘appear faintly’, hodoku ‘undo, untie, loosen, unravel’(< pJ *pono ‘slightly’ + *toka- ‘dissolve’). The latter morpheme occurs in hi, OJ pi2 ‘fire’, OJ potaru ‘firefly’, OJ pokusi ‘bonfire’, OJ pokuso ‘tinder’, OJ potopor‘get heated’, OJ poter- ‘flush, be all aglow’ etc. 24. hotaru ‘firefly, glowfly’ A morpheme boundary can be inserted between ho-, OJ po- ‘fire’ and taru. For the first morpheme I refer to the previous item. The final morpheme may be an assimilated form of OJ ter- ‘shine’, Azuma OJ to1r- ‘shine’ (< pJ *ta(C)ira-). Additional support for this derivation comes from the attestation of the complex verb hoteru, OJ poter- ‘flush, be all aglow’. 25. hude ‘writing brush, painting brush, pen’ The similarity of this word with MK pwut, SV bút and reconstructed EMCh. *prit all with

6.1. Lexicon

87

the same meaning is striking and therefore the Japanese word is sometimes suggested to be an early loan from Chinese.5 However the Japanese vowel -u- and the voiced -d- are difficult to explain as imitations of EMCh. *-i- and *-t-, since an exact copy as -i- and -tlies within the phonological possibilities of Japanese. Therefore, I prefer Martin’s internal analysis for the word as humi ‘writing’ followed by te ‘hand’. The contraction of the nasal-dental stop sequence can account for the voiced consonant. Anyhow, borrowing or covert compound, the entry should be omitted from the core evidence. 26. humoto ‘foundation, foot (of a mountain)’ This word can be derived from -hu ‘place where plants grow’ and moto ‘origin, basis’. 27. husuma ‘quilt, bedding’ This word can be derived from the attributive form of the verb husu ‘lie down’ followed by the morpheme mo ‘garment’. In Old Japanese mo is used independently to design a long skirt for women, worn from the waist till the feet and often having folds. The o(2)~a alternation that we assume for this morpheme is commonly observed in Japanese, as is the case in OJ -ko2 ~ OJ -ka ‘place’, OJ kamome2 ~ OJ kamame2 ‘sea gull’, OJ ko2ro2s- ‘kill’ ~ OJ karas- ‘kill, destroy, wither’, OJ pi1ro2- ‘broad’ ~ OJ pi1rak- ‘open’, OJ siro 2sime1s- ~ OJ sirasime1s- ‘deign to rule’ etc. 28. hutokoro ‘bosom, breast, heart’ Considering the unusual length of the word it is likely to be a compound. The members of the compound probably are huto- ‘thick’ and tokoro ‘place, spot’ accounting for the loss of one syllable -to- after the other -to- by haplology. 29. ikada ‘raft’; ika ‘kite’ The word is derived from the root uka- of uku ‘float’ preceded by an i-prefix and followed by ita ‘board’. Ika ‘kite’ may be derivable from the same root, in the sense of a ‘floater’. 30. inisie ‘ancient times, old days, long ago, formerly’ OJ inisi-pe1 supports the derivation from the past attributive form of OJ in- ‘go, leave, depart’ and OJ pe 1 ‘shore, vicinity, side, direction’ in the sense of ‘the side that is left behind, times that are gone ’. The past auxiliary -si is attached to the conjunctive form of a verb to build an attributive past ending in Old Japanese. 31. isago ‘sand’/ iso ‘rocky beach, shore, strand’ Considering the parallel semantics and the congruent high atonic register both words are derivable from the same etymon. The main obstacle however is the final vowel of OJ iso1 which is A-type (ko @-rui). Therefore, it is impossible to regard iso ~ isa ‘grind, sand’ as a classical case of o(2)~a alternation in Japanese. A possible outcome is to derive isago as isa ‘grind’ followed by a nasal genitive and ko ‘little’ and to explain iswo as a compression of isago. Vovin has suggested to relate isi ‘stone’ to the same etymon, but this is problematic since Vovin’s reconstruction isi < pJ *(d)iso 2-Ci cannot account for the vowel in OJ iso1. 5

Miyake 1997, 189, 192-93.

88

Chapter 6. Internal evidence contradicts

So, it is better to consider isi ‘stone’ a separate etymon. Starting from pJ *isa ‘grind, sand’, the comparison with the Tungusic words for ‘dirt; to poach (in dirt); bay, shore inlet’ and the Turkic words for ‘red clay, ochre’ is semantically overpermissive. 32. isari- ‘fishing -’ In contemporary Japanese the word is limited to compounds like isari-bi ‘fishing fire’, and isari-bune ‘fishing boat’. Earlier attestations indicate that the word has developed from izaru ‘fish’ with a voiced sibilant. According to Martin the verb is a reduction of a compound of OJ iwo ‘fish’ and OJ asar- ‘scavenge, forage, hunt’. Even if this suggestion includes a speculative reduction, it casts doubt on the proposed external etymology. 33. itigo ‘strawberry’ In Old Japanese the word is attested as OJ itibi 1go1, an attestation that is in conflict with Starostin’s reconstruction. The Old Japanese word can probably be derived from itibi ‘Indian mallow’ followed by a nasal genitive and OJ ko 1 ‘small, little, offspring, child’. 34. itutu ‘five’ In consideration of io, OJ ipo ‘five hundred’ that can be derived from *i and pJ *-po ‘hundred’ and of OJ iso 1 ‘fifty’ that can be derived from *i and pJ *-su(C)o‘ten’, it is clear that *i- bears the meaning ‘five’ in this word. The -tutu is of obscure origin, but maybe it is a double occurrence of the genitive and substantivizing OJ tu. The reconstruction pJ *-po ‘hundred’ is supported by mio, OJ mi-po ‘three hundred’ and perhaps also by momo ‘hundred’ if assimilation of *p to the following vowel can account for m. The reconstruction pJ *-su(C)o ‘ten’ is supported by OJ so1 < *su(C)o in OJ miso1 ‘thirty’, nanaso1 ‘seventy’, yaso ‘eighty’. 35. kabane ‘corpse, dead person’ In consideration of kara ‘nutshell, hull, skin’ and its possible derivation karada ‘body’, the analysis of this word as a compound of kara ‘nutshell, hull, skin’ and hone ‘bone’ is not unlikely. 36. OJ kabi1 ‘sprout, unhusked ear of grain’ The word is possibly a covert compound of OJ kami1 ‘top, head, upper part’ and ipi1 ‘cooked rice’. 37. kabuto ‘helmet, headpiece’ If it is not from huta ‘lid’, the precise identity of the second morpheme remains unclear, but the first morpheme must be related to kabu ‘head’. It can be remarked that the verb kaburu ‘put on, wear, take upon oneself’ can also be derived from this etymon. As far as the register and the semantics are concerned kami ‘top, head, upper part’ could also be entangled with this etymon, but the quality of the final vowel in OJ kami 1 remains unexplained. 38. kaede ‘maple tree’ Considering the peculiar shape of the leaf, it is possible to analyze this word as a compound of kaeru ‘frog’ and te ‘hand’.

6.1. Lexicon

89

39. kaiko ‘silkworm’ Given the Old Japanese phonology kapi1ko1, it is possible to derive this word from a deverbal noun from OJ kap- ‘raise (animals)’ and OJ ko1’child, egg’. 6 40. OJ kami 1ra ‘leek, scallion’ The word is a petrification of the Old Japanese word for ‘leek’ mi1ra prefixed by ka ‘fragrance, smell’. A similar construction in which ka ‘fragrance’ precedes an aromatic natural product is found in (NS) OJ ka kunomi2 ‘fragrant fruit’. 41. kamome ‘sea gull’ The word is probably a compound of kamo ‘duck’ and -me ‘bird, animal that lives in flocks’, the suffix that occurs in sime ‘hawfinch, grosbeak’, suzume ‘sparrow’, tubame ‘swallow’ etc. 42. kanata ‘here and there, far in distance, a long way off’ This word can be derived from the demonstrative pronoun of the distal plan ka- ‘that’, followed by the genitive no genitive and kata ‘direction’, if we suppose that haplology can account for the elision of one syllable. This explanation makes Miller’s effort superfluous. 43. kanemoti ‘man of wealth’ . The resemblance between MK ka ∆omyel- ‘to be rich, wealthy’ and the Japanese compound J kanemoti ‘wealthy person’ is indeed striking as Vovin remarks, but the phonological correspondences are merely coincidental. Since the use of coins is already mentioned in the Nihon Shoki and their circulation was increasing gradually in Japan since the eight century, the rich were distinguished from the poor on the basis of the kane ‘money’ they motu ‘dispose of’. 44. kape1ra ‘oar, paddle, scull’ Omodaka’s suggestion that this word is a pleonasm consisting of kai, OJ kapi1 ‘oar, paddle, scull’ and hera, OJ pe1ra ‘spatula, pallet, scoop’ makes Starostin’s external comparison 7 doubtful. 45. karada ‘body’ Whether the second morpheme is an occurrence of a suffix *-da ‘(undefined) mass’, that derives nouns without precise number distinction or whether it is the bound noun ta ‘hand, limb’ preceded by a nasal genitive, it seems that the first member of the compound is the word kara ‘nutshell, hull, skin’. Other attestations of ta in the sense of ‘limb’ or ‘hand’ can be found in OJ tanapidi ‘elbow’, OJ tanamata ‘space between the fingers’ etc. The presently unproductive collective formant *-da seems to occur in examples as OJ ke2damono 2 ‘hairy being, beast’ (< ke2 ‘hair’ + -da - + mono2 ‘being’) versus OJ ke2mono2 ‘id.’, OJ e1 ‘branch’ 6

7

This internal analysis was first suggested by Kamei 1954, 8. Karlgren 1926 has suggested that this word is an early loan from OCh. *kep, but as Miyake 1997, 201 points out, the vocalism for such an explanation is poor. Either way, the word cannot be allowed to the core evidence. Omodaka et al. 1967, 214.

90

Chapter 6. Internal evidence contradicts

in alternation with eda, OJ e1da ‘branch’, OJ ikuda ‘how much, some amount’ in alternation with a number of compounds with OJ iku- ‘how much’, in OJ nami1ta, J namida ‘tear’, J nada ‘open see’, OJ kokoda ‘a mass [of], very much’ (< *koko-n-ta), and OJ sapada ‘much, a lot’ (< *sapa-n-ta). 46. karamusi ‘Chinese silk plant, ramie’ The word is a compound of kara ‘(Han) China, Korea, foreign’ and OJ musi ‘ramie’. 47. karasi ‘mustard’ Parallel the derivation of OJ ki 1tasi ‘rock salt, dried and hardened salt’ and susi ‘sushi, vinegared fish and rice’ below, the word can be derived from the conclusive form of karai, OJ kara- ‘hot, acrid, sharp, salty, bitter’. In Old Japanese the conclusive marker for adjective stems is -si . The name for the spice can be interpreted as ‘it is hot and spicy, hot stuff’. 48. karasuki ‘Chinese plow’ The word is a compound of kara ‘(Han) China, Korea, foreign’ and suki ‘spade, plow’. 49. karei ‘flatfish, plaice, turbot, dab’ The word is attested in OJ as kara-efi . I suspect that it is a compound with J ei, OJ epi ‘ray (fish)’. The first member of the compound may be kara- ‘empty’ referring to the flat shape of the fish. 50. kasiwa ‘oak’ The confusion that has arisen around this word is mainly due to the fact that the internal analysis has been overlooked. Considering kasi ‘evergreen oak’ which is also present in the attached index, it becomes clear that the meaning ‘oak leaf’ corresponding to a the compound of kasi ‘oak’ and ha ‘leaf’ was primary. However the word has developed along two different semantic pathways. As a pars pro toto the word became used to design the entire tree instead of the leaf, but the meaning has also been extended from ‘oak leaf’ to any ‘leaf used to wrap food’, a meaning that is still perceived in some dialects, to ‘vessel’. It is therefore not necessary to treat the different meanings of the word as different etyma as Miller and Starostin do. 51. kataki ‘enemy, adversary’ In view of kata(-) ‘one (of a pair), one person, the opposite side’ and the suffix -ki < OJ -ki 1 that is used for ‘male’, the word is derivable as ‘opposed man’. The suffix -ki ‘male’ occurs in complementary distribution with -mi ‘female’ as in OJ oki 1na ‘old man’ and OJ omi1na ‘old woman’ and in names for gods and goddesses like Izanagi and Izanami or Kamurogi and Kamuromi. 52. OJ katama ‘bamboo basket’ In consideration of the variant katami the word is likely to be a compound of kata- ‘hard’ and the deverbal noun of OJ am- ‘knit’ with an assimilation of the final vowel.

6.1. Lexicon

91

53. katana ‘sword, saber, knife’ This word is probably a compound of kata(-) ‘one (of a pair), one person, the opposite side’ and OJ na ‘blade’. The second morpheme is thought to be present in nata ‘hatchet, sword’, ono ‘axe’, nomi ‘chisel’ etc.. It can be remarked that this analysis does not exclude the comparison of *na with MK nolh ‘blade’ following Martin’s proposal, but this comparison can be dismissed from the core evidence since it concerns a vocabulary item in the domain of technology (8.2.2). 54. OJ katano2wi ‘beggar, leper’ This word is thought to be a compound of kata(-) ‘the opposite side’ with the genitive no2 and wi, deverbal noun of the existential verb iru, OJ wi- ‘exist, be’. The meaning ‘being at the opposite side’ probably comes from the fact that lepers and beggars were avoided by people in their community. 55. kate ‘provisions’ This word is a contraction of karite which is attested in Old Japanese with the same meaning. Although the Old Japanese word can probably be further analyzed as a compound of karei ‘dried boiled rice’, OJ karepi 1 ‘dried boiled rice to be taken along as food for a journey’ and the deverbal noun ate ‘assignation’ from the verb ateru ‘assign, allot, allocate’, the phonological derivation from karite alone is enough evidence to disregard the proposed external comparison for this word. 56. katuo ‘bonito, skipjack’ In view of OJ iwo~ uwo ‘fish’, the word is probably a compound of the adjective kata‘hard’ and the word for ‘fish’. 57. kawara ‘riverbed’ The obvious relation to kawa ‘river’ speaks for a segmentation into two morphemes. The identity of the second morpheme could be hara ‘plain’, but an equally good reconstruction is based on *-ra ‘(defined) quantity, group’ for which I have argued above. 58. kine ‘pestle, pounder’ In Old Japanese the word is attested as both ki 1 and ki 1ne, which makes us suspect that ne is a separate morpheme, maybe the same word as ne ‘root’. 59. kinoko ‘mushroom’ This word is a compound of ki ‘tree’ followed by the genitive no and ko ‘child, offspring’ in the sense of ‘miniature tree’. 60. MJ kisa ‘grain of wood’ Although the identity of the second morpheme is somewhat obscure, the first morpheme is likely to be ki ‘wood’. The *sa could be the adjectival noun ‘little (thing)’ as found in OJ sasa ‘small, little’, which is probably a reduplication.

92

Chapter 6. Internal evidence contradicts

61. OJ ki1tasi ‘rock salt, dried and hardened salt’ Parallel to the derivation of karasi ‘mustard’ above and susi ‘sushi, vinegared fish and rice’ below this word can be derived from the conclusive form on -si of an adjective *kita‘hard’. Although the adjective root is not attested as such, we find a number of internal indications for its existence, such as the adjective kitu- ‘tight, strict, hard’ that represents a case of vowel assimilation, MJ kitafi ‘dried fish’ that is derivable from *kita- ‘hard, tight, dried’ and a suffix *-pi ‘fish’ and the verbs kitaeru ‘forge, temper, drill, train’ and OJ ki1tam- ‘punish, chastise’. 62. koi ‘carp’ In view of OJ ko 1pi ‘carp’ this fish name can be derived from ko-, OJ ko1 ‘small’ and -pi ‘fish’. The Japanese counter for fish is the suffix -bi and the final syllable -pi ~-bi (? < *-n-pi) is found in a number of fish names like OJ awabi 1 ‘abalone’, ebi ‘shrimp’, tai, OJ tapi1 ‘sea bream, red snapper’ OJ ko 1pi ‘carp’, OJ kapi1 ‘shellfish’, MJ kitafi ‘dried fish’. The derivation ‘small fish’ is further supported by the Shuri form for ‘carp’ kuu-qiyu that derives from Sr. kuu- ‘little, young’ (OJ ko 1) and Sr. qiyu ‘fish’ (OJ iwo). 63. koma ‘horse, colt, pony’ In view of OJ ko1ma ‘horse, colt, pony’ this animal name can be derived from ko-, OJ ko1 ‘small’ and uma ‘horse’. Since the first morpheme of the compound is monosyllabic vowel delition is expected. There is nothing unusual about the order of the components in this analysis as is suggested in the Altaic dictionary (2003). It can be remarked that koi ‘carp’ above displays the same order, and so does OJ ko1usi ‘calf’ (< ‘little cow’) and many other compounds with ko- ‘small’ in Japanese. 64. kome ‘rice’ Considering dialectal forms such as mee, maz, mai and a probable Chinese donorword OCh. *(h)meq for the second member of the compound, the segmentation into an undefined ko and *me ‘rice’ becomes more likely.8 The segmentation does not exclude the possibility that the second member of the compound is a borrowing from the Old Chinese donorword. In the light of extensive rice trade during the Han dynasty (206 BC-220 AD) a Chinese loan source for this word gains credibility. 65. koromo ‘clothes, garment, dress’ The first member of the compound could be either the noun koro ‘self, *body’ or the attributive form of the verb kiru ‘wear’ with the vowels assimilated to the second morpheme mo ‘garment’. In view of the structural parallel with husuma ‘quilt, bedding’ above I am inclined to favor the latter segmentation.

8

Miyake 1997, 200 attempts to explain the unsegmented form OJ ko2me 2 as an early loan from Chinese, but in view of the Ryukyu evidence segmentation seems more probable. In view of J ume ‘plum’ and uma ‘horse’ the expected imitation of OCh. *hmeq in Japanese would be *(m)ume ‘rice’, as is explained in paragraph 8.1. below. Either way, the entry can be excluded from the core evidence.

6.1. Lexicon

93

66. koso ‘the very, just, indeed’ This emphatic particle is probably a compound of the proximal demonstrative ko- ‘this’ and -so ‘fact, thing’ in the sense of 'this (very) thing’. 67. OJ koti ‘east wind’ Although the identity of the first morpheme is unclear, the second morpheme can be identified as ti ‘wind, direction’, which is also attested in hayati ‘fast wind, wind storm’. 68. kotoba ‘language, speech, word’ It probably concerns koto ‘thing, matter, fact, word’ followed by a nasal genitive and ha ‘leaf’. 69. OJ ko 2to2pi1 ‘strong sacrificial bull, bull which bears heavy loads on the back’ The animal name is probably derived from its function, namely ‘burden-bear’. The first morpheme is probably koto, OJ ko2to2 ‘thing’ and the second morpheme is a deverbal noun from ou, OJ op- ‘bear, carry on the back’. 70. koziri ‘chape, tip of a sheath, point of a scabbard’ The word can be derived from ki, ko- ‘wood, tree’ followed by a nasal genitive infix -n- and siri ‘butt, rear, end’. 71. kudamono ‘fruit’ Whatever the identity of the first member(s) of this compound (< pJ *kunta-) the final morpheme seems to be mono ‘thing’. Although the etymological structure of this word is not completely clear, a serious indication for segmentation comes from its unusual length. Four syllables is unusually long for a simplex morpheme. 72. kuriya ‘kitchen, cuisine’ In consideration of OJ kuri ‘black clay, black’ and ya ‘building, house’ this word may be a compound. The kitchen being the part of the house that was smoked black, the semantics of the compound are not improbable. 73. kurumi ‘walnut’ The word can be derived from pJ *kuru ‘chestnut’ and pJ *mu(C)i ‘fruit, nut, berry’. The first morpheme underlies in kuri ‘chestnut’ and the second is mi, OJ mi2 ‘fruit, nut, berry’. 74. OJ kutupi1ki1 ‘a loom that is bound to the feet and used by pulling the feet’ This mechanical tool is an obvious compound of kutu ‘shoe’ and the deverbal noun of hiku ‘pull’. 75. kutuwa ‘bit’ The word can be derived as a compound from kutu ‘mouth’ and wa ‘wheel, circle’. Although the distinction between A-type i1 (ko@rui) and B-type (otsurui) i2 is not preserved following dental stops in Old Japanese, it is possible to reconstruct an earlier *i2 < *u(C)i underlying OJ kuti ‘mouth’.

94

Chapter 6. Internal evidence contradicts

76. mae ‘front, forepart’ In view of the morpho-semantic analysis of sirie ‘rear, back side’ as siri ‘buttocks, hips, rear or back of a person, tail end, tag end’ and OJ pe 1 ‘side’, a parallel can be drawn deriving this word from the bound nouns ma- ‘eye’ and -he ‘side, direction’. The final morpheme is listed in the index as a free noun in Old Japanese and now frequently occurs as a bound noun in words indicating a temporal or spacial dimension such as inisie ‘ancient times, former days’, kawabe ‘river side’, OJ moto2 pe1 ‘bottom-side’, uwabe ‘outside, surface’ etc. The first morpheme ma- ‘eye’ underlies OJ me 2 ‘eye’ ( *tasi-). The most natural choice for a paradigm that serves as the basis of the analogic levelling is a class of consonant roots. Regarding the numeral imbalance between approximately 200 e-bigrades and 40 i-bigrade verbs, it is clear that the majority of the reconstructed roots will have -a as the final vowel. Whitman agrees with Unger that all i bigrade verbs reflect original vowel final roots, but they incorporate the auxiliary u ‘obtain’ < *a-i as an intransitivizer (*oko- + *a-i > *oko-i > OJ oki 2-). Whitman’s criterion for the reconstruction of vowel final roots underlying quadrigrade verbs is more restricted since it is only based on covert root forms. All quadrigrade verbs that have a transitive e-bigrade counterpart are -contrary to Unger’s view- traced back to a consonant root (e.g. OJ ap- ‘meet (intr.)’ ~ OJ ape 2- ‘join (tr.)’ < pJ *ap- in spite of the covert root awaseru ‘bring together’). Consonant roots are assumed to extend the auxiliary u ‘obtain’ to a secondary transitivizing function. Following Unger a quadrigrade verb reflects a consonant root only if there is no positive internal evidence available to derive it from a vowel final root. For external evidence for the reconstruction of a consonant root he points to possible Korean cognates (e.g. OJ kak- ‘scratch, engrave, write’ < pJ *kak- in comparison with MK kolk- ‘scratch’). Although his evidence for individual consonant final 14 Martin 1987, 667. 15 Ramsey 1978; Martin 1996; Whitman 1999; Unger 2000 b.

116

Chapter 6. Internal evidence contradicts

roots is weak, the reconstruction of consonant final roots as a class in general is necessary since it serves as a basic paradigm for analogical levelling of all vowel root classes. The problem with Whitman’s view is the awkward elision of the root internal *a of the auxiliary u ‘obtain’ when it is incorporated into the i bigrade verbs and also the fact that consonant final verbs extended the auxiliary u ‘obtain’ to a secondary transitivizing function is grammatically puzzling. I prefer to follow Unger’s reconstruction criteria because his reconstruction of vowel final roots is primarily internally motivated, representing a straightforward integration of the explanations concerning morphophonemic vowel alternations, formally and semantically interrelated quadrigrade and bigrade verb pairs and the coronal lenition hypothesis. Unger’s evidence for the reconstruction of individual consonant final roots, however, is rather weak. For reasons of methodological circularity the present work cannot rely on comparative Korean evidence for the reconstruction of consonant final roots. Therefore, I will reconstruct consonant final roots for all quadrigrade verbs, unless covert roots or bigrade counterparts bear evidence for the final vowel of the root. Complex verbs Many verbs in the covering index appear to be simplex at first glance, but internal analysis reveals that these are covert complex verbs resulting from prefixing to a verb another element. The prefixed morpheme can be a prefix, an adverb, an adjective, a noun or another verb. It can be remarked that some of the underlying simplex verbs have a voiced initial that reflects earlier prenasalization. This nasal infix probably represents a reduction of an earlier case marker ni. For the following etyma there is enough doubt about their postulated simplex origin to omit them from the body of core evidence. 1. agaku ‘paw the ground, paw the air, struggle, wriggle’ This verb is usually derived from asi ‘foot, leg’ followed by a nasal infix and the verb kaku ‘scratch, stroke, engrave, write’. If this internal analysis is correct, the loss of the second syllable of asi ‘foot, leg’ must be accounted for by contraction. 2. arawareru ‘appear, emerge’ That the original root underlying this verb probably is pJ *a(-)ra- ‘be present’ is supported by areru ‘appear, be born’ (< pJ *ara-(C)i), arawasu ‘express, show’ (< pJ *ara-pa-sa-), and aru ‘be present’ (< pJ *a(-)ra-). 3. aruku ‘walk, hike, step’ Miller’s internal analysis of this word as asi ‘foot’ and yuku ‘go’ is problematic since it cannot account for the -r- in the etymon. In view of the attestation of ari yuk- and ari ik- in Old Japanese and of the occurrence MJ arik- ‘walk’ for contemporary aruku it is safe to analyse this verb as a complex derivation of aru ‘be, exist, have’ and iku / yuku ‘go’. Not excluding a certain contamination through the influence of this synonymous verb, it seems that ayumu ‘walk, step’ represents a separate etymon. 4. asobu ‘play’, arasou ‘be at variance, discord, dispute, argue’ Since Old Japanese has the verb so1bap- ‘frolic, flirt’, which is also etymologized in the

6.1. Lexicon

117

attached index, I am suspect that the initial a- of asobu ‘play’ has to be analyzed as a prefix. In consideration of the adjective stem ara- ‘rough, wild’ and a number of verbs that are composed with the prefix ara- like arabiru ‘act wild, rampage’, aragau ‘dispute’, arakeru ‘get scattered, get rough’ etc., I suspect that arasou is a compound of pJ *ara- ‘rough, wild’ and pJ *sop- ‘play’. These verbs might conserve evidence for a sporadic development pJ *-p- > pJ *-np-, as will be discussed in paragraph 10.1.3.1. 5 atukau ‘treat, handle, use’ In the light of tukau ‘use’, this verb may contain another occurrence of the a- prefix. 6. ayakaru ‘share good luck, be similar to’ This verb is probably derived from aya ‘design’ and karu ‘borrow, have a loan, hire, rent’. Both words are listed in the covering index as separate etyma. 7. ayaturu ‘handle, manipulate, operate’ This is another derivation with the noun aya ‘design’, followed by the verb toru ‘take, get, have’. The latter verb is listed separately in the index. 8. hiideru ‘excel, surpass’ The verb is probably a compound of the verb OJ ide- ‘emerge’ and a petrified intensive prefix hi- (< pJ *pi- 'extreme’). The reconstruction of an intensive prefix *pi- is speculative as mentioned above (A.1.5), but doublets like MJ fisik- ‘crush’ versus siku ‘spread out’ and hisagu ‘crush’ versus saku ‘separate, split’ support this idea. 9. hisigu ‘crush, smash, break’ The verb is attested with a voiceless velar in MJ fisik- ‘crush’. The attestation of siku ‘spread out’ suggests that we could be dealing with the petrified intensive prefix hi- (< pJ *pi- 'extreme’) followed by the latter verb. 10. hissageru ‘carry in one’s hand, take up’ The reduplication of the -s- in the contemporary Japanese verb and in MJ fissage-, that is attested next to MJ fisage- ‘carry in one’s hand, take up’, indicates that we are dealing with a contraction of two syllables into one. The verb is likely to be the result of compounding the renyo@kei (conjunctive form) of hiku ‘pull, grasp’ with sageru ‘carry by hand’. 11. hodoku ‘undo, untie, loosen, unravel’ This verb can be segmented into the adjectival noun hono- ‘faint, slight’ and the verbal root pJ *toka- ‘dissolve’ in the sense of ‘slightly undo’. The verbal root pJ *toka- occurs in the transitivity pair tokasu ‘melt, dissolve’ and tokeru ‘get melted, get dissolved’. 12. iku ‘go’ Parallel to the use in contemporary Japanese, iku is attested in the Man’yo@shu@ as a colloquial version of yuku ‘go’. I suspect that the onset vowel i- is the petrified verbal i-prefix and that the original root for ‘go, go away’ is *ka-. Following Martin’s suggestion yuku may represent a reduced form of the verbal prefix. Martin suggests that the i- prefixes probably were

118

Chapter 6. Internal evidence contradicts

independent adverbs at an earlier period. But an equally inviting proposal can be found in Unger ‘s explanation of the i-prefixes as a reanalysis of a preceding verb root followed by a conjunctive (renyo@kei). Under the following entry the possibility will be discussed that pJ *ka- ‘go away’ underlies verbs like kayou ‘ply between, commute, frequent, go’, OJ kayo2r‘approach’, OJ kare2- ‘get apart, cease, go away’. Further evidence can be found in paragraph 9.3., where it is argued that the contrast between distal ka- ‘(going) away from me’ and proximal ko- ‘(coming) towards me, near me’ reflects the original contrasts between the verb roots *ka- ‘go away’ and *ko- ‘come’. 13. OJ in- ‘go, leave, depart’ It is interesting to start with the observation that OJ in- ‘go, leave, depart’ belongs to the n-irregular verb paradigm (na-hen) along with only two other verbs: and sin- ‘die’ and the perfect auxiliary OJ -in-. The n-irregular verb paradigm is an exception to the athematic paradigm (yodan) because it has ‘long’ attributives (rentaikei) -uru and subjunctives (meireikei) -ure in contrast to the ‘short’ attributives -u and subjunctives -e of the athematic paradigm. Whitman (1985) has argued persuasively that at some pJ stage -r- was deleted after short vowels, but retained after long vowels. The loss of the intervocalic -r- in the attributives and subjunctives of the athematic paradigm is commonly attributed to this rule, so in the case of the n-irregular verb paradigm a preceding long vowel at the pJ stage must have blocked the application of the rule. This observation is an indication that these three verbs all share the same monosyllabic verbal root *na- ‘become, go’ and that OJ in- ‘go, leave, depart’ contains the verb prefix i-. This old verb prefix is occasionally attested in Old Japanese, as in OJ i-papi1 ‘crawl’ (M 199) and OJ i-kogi 1 ‘row’ (M 4408), and is probably lexicalized in a number of verbs such as ino2r- ‘pray’, imas- ‘deign to be, stay’, itar‘reach’, MJ itas- ‘achieve, cause’ etc. If pJ *ka- can indeed be reconstructed as the root ‘go’ underlying verbs like kayou ‘ply between, commute, frequent, go’, OJ kayo2r- ‘approach’ and OJ kare2- ‘get apart, cease, go away’, then the same verbal prefix may be reflected in iku ‘go’ as is suggested under the previous entry (10). Martin suggests that the i- prefixes probably were independent adverbs at an earlier period. But an equally inviting proposal can be found in Unger ‘s explanation of the i-prefixes as a reanalysis of a preceding verb root followed by a conjunctive (renyo@kei) *-i.16 Another argument in favor of the reconstruction of a monosyllabic root *na- lies in the fact that following Unger the root vowel in monosyllabic morphemes was automatically long at the pJ stage.17 A monosyllabic stem would thus account for the retention of the intervocalic -r- in the attributive OJ inuru and subjunctive OJ inure. Evidence for the final vowel of the reconstructed root can perhaps be found in the transitivity pair OJ nar- ‘become’ and OJ nas- ‘create’. On the comparative level, this internal analysis strengthens the comparison of OJ in- ‘go away’ to MK na- ‘become, go out’ and to pTg *-na@- ‘go (to do)’.18 As far as the semantics are concerned, the Korean verb 16 Martin 1987, 94, 668; Unger 2000 a, 676. 17 Unger 2000 a, 669. 18 Benzing 1955, 1068 refers to a common derivational suffix denoting motion ‘go (to do)’ in Tungusic, reflected in Ma. -na- ~ -ne- ~ -no-, Na. -nda-, Olc#. - Nda-, Oroc#, Ud. -na-, Sol. -na@-, Neg. -na-, Ev., Lam. -na@-. Followed by this suffix, in addition to their basic semantics, verbs take on the meaning of motion, e.g. Ma. ala- ‘report’ versus alana- ‘to go to report’, Ma. fata- ‘pick fruit’ versus Ma. fatana- ‘to go to pick fruit’ etc.

6.1. Lexicon

119

provides evidence for the grammaticalization from the directional verb ‘go out’ into the change of state ‘become’. It can be remarked that English go can mark the change of state in limited expressions, such as go crazy. 19 Since the grammaticalization of ‘become’ into a copula ‘be’ is also cross-linguistically attested it cannot be excluded that the defective copula OJ n- ‘be’ is related to the same etymon.20 Being an external issue, the problem whether pJ *na- can be related to other Tungusic, Mongolic and Turkic verbs meaning ‘go’ and containing a nasal will be dealt with in chapter 10. that is concerned with sound correspondences. Starting from the root pJ *na‘become, go out’ it is necessary to analyze OJ sin- ‘die’ along similar lines. The perfect auxiliary OJ -in- will be treated in paragraph 6.2. on morphology. 14. itaru ‘arrive, reach, attain’ / itasu ‘do, cause, bring about’ This transitivity pair is derivable from the verbal root pJ *ita-, which is further segmentable into the same prefix *i- as treated under the previous word and the ultimate verbal root pJ *ta- ‘reach, come’. It can be remarked that this analysis overrules Starostin’s external comparison because he compares the intransitivizing formant as part of the Japanese verb root, but it leaves room for Martin’s etymological proposal. In paragraph 6.2. I will argue that the root pJ *ta- may well be cognate with the perfective auxiliary OJ -ite-. 15. kagamu ‘stoop, lean over, bend’/ Mod. J kugum- ‘stoop, crouch’ Although the verb is not attested in Old Japanese as Starostin suggests, the Japanese-Portuguese dictionary of 1604 contains the vowel alternant kugum- and another vowel alternation is attested in contemporary kogomu both sharing the same meaning. In consideration of pJ *kuma- ‘bend’ from which kuma ‘corner, nook, recess, bend, turn’ was derived, the verb seems to be a reduplication of this root *kuma-. The resulting voiced velar stop -g- can be accounted for as the result of the nasal-obstruent cluster *mk that arose through juxtaposition and a-elision. One could speculate that the root *kuma- ‘bend’ also underlies in kumu ‘intertwine, plait, assemble, unite’, but the meanings ‘intertwine’ and ‘bend’ are too distant to make this speculation acceptable. 16. kagu ‘smell, sniff, scent’ The verb seems to be composed of ka ‘smell, odor’ and muku ‘turn toward, face’. The elision of the vowel in the verb probably resulted in the underlying nasal-obstruent *mk . 17. kaoru ‘smell’ The verb is probably a compound of ka ‘smell, odor’ and oru ‘be, exist’ in the sense of ‘have a smell’. 18. OJ kayo2r- ‘approach’ The verb is probably a compound of pJ *ka- ‘go away’ and yoru ‘approach’ (< pJ *yora-). 19 Heine and Kuteva 2002, 156. 20 Martin 1987, 40-44; Frellesvig 1999. The defective copula OJ n- appears only in its conjunctive form ni, as a gerund nite, and in its attributive form no. For crosslinguistical attestations of the grammaticalization of ‘become’ into a copula ‘be’ I refer to Heine and Kuteva 2002, 64.

120

Chapter 6. Internal evidence contradicts

For the reconstruction pJ *ka- ‘go away’ I refer to the previous items 12. iku ‘go’ and 13. OJ in- ‘go, leave, depart’ and to paragraph 9.3. 19. kazuku ‘duck one's head in water, completely cover one's head, put on’ If we may account for the loss of one syllable by contration, the verb is probably a compound of kabu ‘head’ and tuku ‘soak, be submerged, get wet’ (< pJ *kan[pu]-tuka-). 20. kizuku ‘build, construct’ In consideration of OJ tuk- ‘build, construct’ from which the indexed etymon tukuru ‘make, create, manufacture’ is derivable, the second morpheme of this complex verb can easily be identified. From the semantic viewpoint OJ ki2 ‘fortress’ seems a good candidate for the first morpheme, but this is not without phonological problems regarding the quality of the vowel in OJ ki1duk-. A better phonological fit is found in OJ ki 1 ‘pestle, pounder’ in the sense of ‘make with a pounder’, but semantically this suggestion is rather weak. 21. kisou ‘compete, strive’ Considering OJ so1bap- ‘frolic, flirt’ and asobu ‘play’ we might be dealing with a prefixed root *sopa- ‘play’. The intensive prefix *ki could also be underlying in kiou ‘strive, compete’, OJ ki1pop- versus ou ‘pursue’ and maybe in the previous etymon OJ ki1duk- versus OJ tuk-. 22. kitaru ‘come, arrive’ This verb is probably a contraction of the conjunctive of kuru ‘come’ and itaru ‘reach’, as is indicated by the Middle Japanese attestation of ki itar-. 23. kotaeru ‘endure, answer’ This verb is derivable from koto ‘thing, matter, fact, word’ and the verb au ‘meet, agree, fit’, followed by the transitivizing formant -(C)i-. 24. kotohogu ‘congratulate’ The first morpheme is the noun koto ‘thing, matter, fact, word’ as under the previous etymon and the second morpheme is the verb attested as OJ po1k- ‘pray’, which is listed in the index. This internal analysis is not only in conflict with Starostin’s etymology for the first member of the compound koto (DB 154), it also contradicts the etymology for the second member OJ po1k- ‘pray’ (DB 278) because in the following subsection this verb will be analyzed as a derivation of pJ *pu(C)a- ‘desire’. 25. kutugaeru ‘be overturned, capsize’ This complex verb probably consists of the bound noun kutu- ‘mouth, aperture, opening’, which is attested in the etymon kutuwa ‘bit’ and underlies the etymon kuti, followed by a nasal infix and the verb kaeru ‘return, turn back’ in the sense of ‘turn back the opening’. 26. mamoru ‘protect, defend’ In Old Japanese and in some dialectal forms the verb is attested without the prefix as mor‘defend’. The identity of the prefix probably is ma- ‘eye’ which underlies OJ me2 ‘eye’ and was also reconstructed in mae ‘front, forepart’.

6.1. Lexicon

121

27. mayou ‘be at a loss, go astray’; madou ‘be puzzled, lose one’s way’ Although *you ‘waver’ is not attested as a free verb, the morpheme occurs frequently in complex verbs such as kagayou ‘glitter’, OJ mogoyop- ‘turn and twist, zigzag, wave one’s way’, samayou ‘wander’, tadayou ‘be adrift’, OJ isayo1p- ‘waver, drift’ etc. The prefixed ma- probably concerns the same bound noun ‘eye’ as under the previous morpheme, in the sense of ‘have wavering eyes’. I will consider madou ‘be puzzled, lose one’s way’ as a case of phonological d ~ y alternation. Another etymological pair that is probably in d ~ y alternation is yokeru ~ dokeru ‘avoid’. Although I lack a satisfactory explanation of this d ~ y alternation, it might be connected with dialectal borrowing from forms like Yo. madurun corresponding to standard Japanese mayou. 28. mimau ‘inquire after, visit’ The verb is derivable from the conjunctive form of miru ‘see’ followed by mau ‘humbly do’. 29. minoru ‘ripen, grow ripe’ If the first morpheme is the noun mi ‘fruit’, the second morpheme could be either noru ‘ride, be borne, be carried’ or an unattested verb that is in o-alternation with naru ‘become, be borne’. 30. modoru ‘return, turn back, revert to’; OJ motor- ‘get twisted, turn to the opposite side, run counter’; OJ moto2por- ‘move around, turn round’; OJ mi2- ‘turn’ In Old Japanese modor- occurs in voice alternation with motor- ‘get twisted, turn to the opposite side, run counter’ and besides there is also the attestation moto2por- ‘return’, so it is believed that all three etyma are derivable from the same etymon moto2por-. In its turn moto2por- can be analyzed as a complex verb consisting of pJ *ma- ~ *mo- ‘turn’ and tooru, OJ to2po2r- ‘walk along, pass by’, the latter also etymologized in the index. The reconstruction of pJ *ma- ~ *mo- ‘turn’ is based on the OJ monograde verb mi2- ‘turn’, which is to be derived as pJ *mo-(C)i or *ma-(C)i and on supposed complex verbs such as megur- / makur- ‘turn round’ (< *ma- ‘turn’ + *kur- ‘wind’) and derivations such as OJ map- ‘whirl, dance’ < *ma- ‘turn’ + *-pa- formant). The achievement formant -ma- ‘to turn, to get’ may also be related. The independent noun OJ mi2 ‘circumference, round, rotation’ (< *mo-(C)i) supports the reconstruction with the back vowel pJ *mo- ‘turn’. This analysis is not in conflict with the comparison of pJ *mo- ~ *ma- ‘turn’ with MK mulu‘retreat, withdraw’ (< pK *mulul-). 31. moosu ‘humbly speak’ In view of OJ mawos- ‘humbly speak’ this verb can probably be derived from the humble prefix OJ mawi- and the verb osu ‘control, govern’, OJ wos- ‘deign to control, deign to consume, deign to wear’ (< pJ *wo-sa- ‘cause to exist, have control over, take possession of’). 32. nasuru ‘rub, smear’ This verb is probably a compound of pJ *na ‘earth, ground’ and suru (< pJ *sura-) ‘grind, rub’.

122

Chapter 6. Internal evidence contradicts

33. nigoru ‘become dirty’ In consideration of OJ ni ‘earth, dirt’ and OJ ko2r- ‘freeze, stiffen, become hard, get absorbed’ this verb is probably a compound of this noun and this verb in the sense of ‘get absorbed by dirt’ with a nasal infix that must account for the voiced velar. Both morphemes are present as separate etyma in the index. 34. niou ‘smell’ In consideration of the meaning of OJ nipop- ‘get red; shine beautifully; be fragrant', the derivation from OJ ni ‘earth, dirt, red clay, red, beautiful’ and ou ‘bear’ cannot be excluded. 35. ogamu ‘worship, bow to’ In view of OJ wogam- ‘worship, bow to’ this verb can perhaps be derived from o, OJ wo ‘tail, tag’ and kagamu ‘stoop, lean over, bend’. This derivation requires to account for the loss of the first syllable of the verb by truncation. The semantics can be understood in the sense of the English expression ‘have one’s tail between one’s legs’ that is used when one feels humbled and ashamed. 36. OJ saburap- ‘serve’ Given the attestation of OJ samorap- this verb can be derived from a prefix *sa- ‘truely, completely’, the verb mor- ‘defend’ and a *-pa- formant. I have argued for the reconstruction of the prefix under sane ‘kernel, core’. This analysis indicates that the -b- alternant is a secondary development. 37. sakebu ‘shout, cry, exclaim’ Given the quality of the vowel of OJ sake1b-, which Starostin fails to note, this verb can be derived from the conjunctive (renyo@kei) of saku ‘cut up, separate, divide, put at a distance’ and yobu ‘call, hail’ in the sense of ‘call from a distance’. It can be remarked that both verbs are individually etymologized in the index. 38. OJ saki1pap- ‘flourish’ This verb is a secondary derivation from the verb saku ‘bloom, open, come out’, from which OJ saki1 ‘good luck, happiness’ is a deverbal noun. The derivation of both etyma from saku is not in conflict with Miller’s comparison. However, considering aku ‘open, be opened, become vacant’ with congruent register and close semantics, it seems that saku ‘bloom, open, come out’ includes yet another manifestation of the intensive prefix *sa‘truely, completely’. If so, then Miller’s proposal should be disregarded and then it is needless to account for the doublet aku ‘open, be opened, become vacant’ versus saku ‘bloom, open, come out’ by the reconstruction of pJ *z, as Unger suggests.21 39. samayou ‘wander, roam’ Under sane ‘kernel, core’ in paragraph 6.1.1.1.1. I have reconstructed an intensive prefix pJ *sa- that was probably still productive in Old Japanese. This reconstruction was supported by alternating verb pairs such as OJ samorap- ‘serve’ and OJ mor- ‘defend’, OJ sagumo1r21 Unger 1977, 100.

6.1. Lexicon

123

‘become cloudy’ versus kumoru ‘id.’, OJ sane- ‘sleep’ versus neru ‘lie down, sleep’, sasuru ‘rub’ versus suru ‘grind’, OJ sawatar- ‘go across’ versus wataru ‘cross, wade, ford, change hands, be imported’, ueru ‘plant’ versus sueru ‘fix, install’, OJ ute(2)- ‘throw away, abandon, discard’ versus suteru ‘throw away, abandon, discard’, aku ‘open, be opened, become vacant’ versus saku ‘bloom, open, come out’, etc. The doublet samayou ‘wander’ versus mayou ‘be at a loss, go astray’ is another doublet in support of this prefix. The internal analysis of samayou as a compound of the intensive prefix and mayou ‘be at a loss, go astray’ is in conflict with the proposed external etymology. 40. sidaru ‘hang down, droop’ In view of sinadareru ‘hang down, droop’(< pJ *sina-n-tara-) we are probably dealing with a reduced form. The verb is a covert compound of pJ *sina- ‘bend’ and pJ *tara- ‘hang down’. The former root underlies in sinau ‘bend’, OJ sinape2- ‘bend’ and the latter root is reflected in OJ tar- ‘hang down’, tareru ‘hang down, droop’. 41. simesu ‘show’ Following Martin this verb is a compound of the renyo@kei (conjunctive form) of suru ‘do’ and mesu ‘deign to see, summon, send for’. 42. sinu ‘die, pass away’ Above it has been suggested that this verb concerns the same monosyllabic verbal root *na‘go’ as in OJ in- ‘go away, depart’. If OJ sin- ‘die’ can be analyzed as *si and *na- ‘go away’, the first component could be a conjunctive or a nominal form of a verb ‘to die’ related to J sosu and OJ sise2- both ‘to kill’ as Martin suggests. It can be remarked that Japanese has sikabane ‘corpse’ next to kabane ‘corpse’ which probably contains the same morpheme ‘death’. Unger’s suggestion that *si might be the Sino-Japanese si ‘death’ requires the postulation of an early Chinese loanmorph as has also been suggested for teru ‘shine’ and karu ‘borrow, have a loan, hire, rent’. In paragraph 8.1.2. I explain why such an explanation is doubtful for karu and teru. The analysis that I suggest here makes it superfluous to account for the doublet sinu ‘die, pass away’ versus OJ in- ‘go away, depart’ by the reconstruction of pJ *z. 43. sitataru ‘drip’ In spite of the incongruent register the suggestion that this verb concerns a segmentation of sita ‘under, lower, bottom’ and taru ‘hang down, droop’ is not unlikely. 44. OJ so2k- ‘get distant, recede’ Although the analysis is speculative, we cannot exclude that this verb is a covert compound of pJ *so ‘back’ and pJ *ka- ‘go’. The interpretation of the vowel of se, OJ se ‘back, back side’ as a B (otsu)-variant (< *se 2) is supported by the attestation of somuku ‘go back on, run counter to’ and OJ sosisi ‘backbone flesh’. Therefore, it is safe to reconstruct pJ *so ‘back’. Under 12. I have argued for the reconstruction of pJ *ka- ‘go’ as the source for iku ‘go’.

124

Chapter 6. Internal evidence contradicts

45. soru ‘warp, be warped, curve, be curved, bend back’ Parallel to the analysis of the previous entry, this verb is probably a compound of pJ *so ‘back’ and *wora- ‘bend, fold, break’. It can be remarked that soreru ‘turn aside, deviate, diverge’ shares the same verb stem (< *so-wora-(C)i-). PJ *wora- is the source of oru ‘bend, fold, break’. 46. suteru ‘throw away, abandon, discard’; sutaru ‘go out of use, decline’ It is clear that both verbs share the same stem pJ *suta- ‘go out of use’, but this verb stem can further be analysed as a compound of the intensive verb prefix pJ *sa- and pJ *uta- ‘go out of use’. The latter verb stem is present in OJ ute(2)- ‘throw away, abandon, discard’ . The doublet OJ ute(2)- versus suteru ‘throw away, abandon, discard’ has been mentioned above as evidence for the reconstruction of the intensive verb prefix pJ *sa-. 47. suwaru ‘sit down, take a seat’ In consideration of uwaru ‘get planted’, I suspect that the verb is a compound of the intensive verb prefix pJ *sa- and pJ *uwa- ‘plant’. The verb sueru ‘fix, install’ can be derived from the same stem (< pJ *sa-uwa-(C)i-) and its counterpart ueru ‘plant’ (< pJ *uwa-) confirms that the prefix pJ *sa- is involved. 48. OJ tabasir- ‘fly about violently, scatter heavily’ In consideration of hasiru ‘run’ and the petrified verbal prefix ta-, that is not necessary related to the bound noun ta- ‘hand, limb’, there is enough internal support for the segmentation of this verb. This intensive prefix ta- is also found in tabakaru ‘plan, consider, dupe’ versus hakaru ‘calculate, consult, plan’, tadoru ‘follow, pursue’ versus toru ‘take, get, have’, tagau ‘differ from, deviate’ versus kau ‘buy, exchange, cross’, takumu ‘plan, think out, scheme’ versus kumu ‘unite, assemble’, taguru ‘pull in, reel in’ versus kuru ‘reel, wind, spin, turn’, tamesu ‘try, attempt’ versus mesu ‘deign to see’, tamotu ‘preserve, maintain’ versus motu ‘take, hold’, tanomu ‘ask, request’ versus OJ no2m- ‘pray’, taoru ‘bend, break off’ versus oru ‘bend, fold, break’, tasukeru ‘help, save’ versus sukeru ‘help’, tayoru ‘rely on, depend on, lean on’ versus yoru ‘depend on, encounter’ etc.22 There are also examples of the intensive prefix ta- in adjective stems, such as OJ tayasu- ‘easy’ versus yasui ‘easy’ and OJ taputo 1- ‘exalted, sacred’ versus hutoi, OJ puto1- ‘big, fat’. 49. tadayou ‘be adrift, float’ This verb is probably a compound of the root underlying tada ‘only, solely, precisely’ and tadasii ‘right, just, correct, straight’ (< pJ *tanta) and pJ *yop- ‘waver’ that is not attested as a free verb, but frequently occurs in complex verbs such as kagayou ‘glitter’, samayou ‘wander’,OJ isayo1p- ‘waver, drift’, OJ mogoyop- ‘turn and twist, zigzag, wave one’s way’ etc.

22 I refer to Martin 1987, 95 for other examples of the occurrence of the prefix ta- in Old Japanese.

6.1. Lexicon

125

50. tagau ‘differ from, deviate’ This verb is probably a compound of the intensive prefix pJ *ta- and the root pJ *kapa- that is the source of of kau ‘cross, intersect’. 51. takuwaeru ‘store, stock, save’ In view of kuwaeru ‘add’ there is little doubt about the final member of this complex verb. Most likely the first morpheme is the intensive prefix *ta- as explained under the previous entries. 52. tamesu ‘try, attempt’ In view of mesu ‘deign to see’, I suspect that this verb is a compound of the intensive prefix *ta- and this verb. The metaphoric extension of ‘see’ over ‘perform and see what the result is’ to ‘try’ is a common semantic development. The development is reminiscent of what happens in contemporary Japanese to miru ‘see’ in the sentence ending -te miru ‘try to’. 53. tanomu ‘ask, request’ The underlying verb is OJ no2m- ‘pray’ and the prefix is another occurrence of the intensive prefix pJ *ta-. Contrary to Starostin’s proposal this etymon seems thus to be different from tanosii ‘pleasant, joyful’. 54. tasukeru ‘help, save’ This complex verb consists of the verb sukeru ‘help’ prefixed by the intensive marker pJ *ta-. 55. tattobu ‘value, esteem, honor’ The reduplication of the medial consonant in contemporary Japanese is caused by reduction of OJ taputo 1m- ‘esteem, honor’. This verb shares the same root as the adjective OJ taputo 1‘exalted, sacred’. The adjective stem is likely to be a compound of the intensive prefix pJ *ta- and hutoi, OJ puto1- ‘big, fat’. 56. tayoru ‘rely on, depend on, lean on’ The semantically overlapping verb yoru ‘depend on, encounter’ indicates that this is a complex verb consisting of the intensive prefix pJ *ta- and yoru. 57. tigau ‘differ, vary’ This verb can be analysed as a compound of OJ ti ‘road, path, way’ and the verb kau ‘coss, intersect’ in the sense of ‘the ways are crossing’. 58. tinamu ‘be connected with’ This verb is probably compounded in a similar way as its semantic counterpart in the previous entry. If the second element of this complex verb is OJ nam- ‘line up’, the first element is probably also OJ ti ‘road, path, way’ in the sense of ‘the ways are lined up’.

126

Chapter 6. Internal evidence contradicts

59. tigiru ‘vow, swear’ This verb is probably a compound of ti ‘blood’ and kiru ‘cut’, perhaps referring to the symbolic act of mixing blood when taking an oath. 60. tikau ‘swear, vow, take an oath, give one’s word’ Parallel to the internal analysis under the previous entry, this verb is probably a compound of ti ‘blood’ and kau ‘exchange, cross, intersect’, again referring to the symbolic act of mixing blood when taking an oath. 61. ukagau ‘peep through, spy, watch, infer, visit, inquire, hear’ In spite of the incongruent register, this is probably a complex verb derived from pJ *uka‘receive, hear, listen’ and pJ * kapa- ‘cross, exchange, mutually do’. These roots are reflected in ukeru ‘receive’ and kau ‘buy, cross, exchange, mutually do’. From compounds like uketamawaru ‘humbly receive, listen, hear’ that developed from OJ uke2 followed by the humble auxiliary OJ tamafar- ‘humbly receive, be given’, it can be seen that the original meaning of pJ *uka- was ‘receive’ in a broader sense because it also meant ‘hear, listen etc.’. From this viewpoint the Tungusic words for ‘submerge in thoughts, change one's mind’ are a weak comparison to pJ *uka- ‘receive, listen, hear’, but Starostin’s Mongolic and Turkic proposals meaning ‘understand, hear’ are semantically acceptable. It are the same proposals as Miller suggests under the entry ukeru ‘receive, accept’. The cognates that Starostin proposes under ukeru ‘receive, accept’ are weaker than the ones he suggest under ukagau, .so they can be rejected. Starostin’s Korean candidate under ukeru is a misgloss for MK wu huy- ‘poke, scrape, bore’ and thus semantically overpermissive. The comparison with the Tungusic words for ‘armful, handful’ and with the Mongolic and Turkic words for ‘meet’ is semantically overpermissive. 62. uramu ‘feel bitter against, resent, regret’ If the verb shares the same root as OJ urami2- ‘feel bitter against’ it can be derived from OJ ura ‘heart, mind’ and OJ mi2- ‘turn’ in the sense of ‘turn one’s heart against’. 63. utagau ‘doubt’ This verb is explained as a segmentation of the adjectival noun OJ utu ‘hollow’, followed by a nasal infix and the verb kau ‘buy, exchange, cross’, in the sense of ‘buy as hollow’. 64. yadoru ‘lodge, take shelter’ This verb is derivable from ya ‘building, house’ and toru ‘take, get, have’ with a nasal infix. 65. yosou ‘equip, help, serve, dish up’ In view of yosoou, MJ yosofof- ‘dress, equip’ and the deverbal noun yosooi, OJ yo 2so2popi1 ‘dress, equipment’, this verb is the result of haplology of pJ *yosopop- to OJ yo2so2p‘dress, equip’. The original verb is probably a compound of yoso, OJ yo2so2 ‘outside’ and OJ opop- ‘cover’.

6.1. Lexicon

127

Derived verbs The majority of the Japanese verbs are derivations from a root and one or more derivational formants. The naked verbal root can be reflected in adverbs, adjectives, nouns or other verb stems. The formants are the intransitivizing formant -r- < pJ *-ra- along with the transitivizing formant -s- < *-sa-, the transitivity flipper -i- < *(C)i, the intensive formant -p- < *-pa-, the achievement formant -m- < *-ma- ‘turn, to get’, the telic formant -t- < -ta- ‘ reach, attain the state of’, the punctual or iterative formant -k- < *-ka-, the passive formant -y- < *-ya- and the resultative formant -n- < *-na- ‘become, be the same as’.23 It is possible that some of these formants came into being through a process of gramaticalization of independent verbal roots, so the exact boundary between complex and derived verbs is vague. For the following etyma the derivation of the verb as a verbal root followed by one or more formants is in conflict with the proposed external comparison of the unsegmented verb as a whole and I will therefore exclude this external evidence from further consideration. I will also point out those etymological suggestions that are incompatible because different etymologies are attributed to verbs that are ultimately derivable from the same verbal root. 1. amaneku ‘widely, generally’; amari ‘rest, surplus, excess’; amaru ‘be in excess, be left over’, amasu ‘excel’ all share the same root pJ *ama- ‘excess’. On the basis of masu ‘excel’ Martin and Miller consider the possibility that a morpheme boundary must be inserted after *a-. The suggestion cannot be completely excluded, but whether pJ *ama- or *ma- is to be reconstructed, the main objection to Miller’s etymology is the etymologization of -r- as a stem-internal consonant. 2. akiru ‘get satiated’ goes back to OJ ak- ‘get satiated’ and has the same root as akasu ‘satiate, tire, bore’ (< pJ *aka-). The primary meaning probably is ‘get satiated’ which makes a poor semantic fit with the Tungusic words for ‘to decrease; bad, low; weak; evil’, with the Mongolic words for ‘low, uneducated, not very good’ suggested by Starostin, or for ‘sad; tired, slow’ suggested by Finch and with the Turkic words for ‘to decrease, be insufficient’. 3. aru ‘exist, have’ There is no consensus in Japanese linguistic literature about the question whether this verb should be analyzed as a simplex root pJ *ara- or as a derivation from an original existential root pJ *a- with intransitivity polarizing -ra- suffix. The verb eru ‘get, obtain’ possibly conserves internal evidence in favor of the latter explanation. Regarding e- as the result from the diphthongization of *a- ‘exist, have’ and the transitivity flipper *-i- ( on- > o- is attested. Therefore, it is impossible to insert a morpheme boundary after OJ o-. An alternative possibility is to explain the noun as a deverbal derivation of oyobu ‘reach, attain to’ but this involves an incongruent register and somewhat problematic semantics. It seems safer to stick to the reconstruction pJ *oyonpi for the time being. 6.1.1.2.2. Verbs 1. OJ apatas- ‘belittle, underrate, make light of’ is a deadjectival derivation from the adjective stem underlying awai, OJ apa- ‘light, pale’ and, OJ apaapasi- ‘pale, light, superficial’ ( -amai negative form of the presumptive ‘probably not’ OJ -amasizi - contracted to MJ -amazi- with -amaziku for the conjunctive (renyo@kei) form and -amaziki for the attributive (rentaikei) form, which indicates that the conclusive (shu@sikei) -amazi is the result of haplology from *-amazisi < *-amasizisi. OJ -amasizi - can be explained as the above subjunctive marker -amasi followed by a negation -zi- that is in its turn a contraction of the conjunctive form of the negative auxiliary -an- and of the conjunctive form of OJ su ‘do’. The morpheme boundaries that are suggested for this form are in conflict with the proposed external etymologies that treat this form as a simplex morpheme. 6. de instrumental case marker It is generally agreed that this case marker is a development of OJ ni-te with similar uses. Explaining OJ ni-te as a collocation of two separate particles, Miller and Finch leave the 47 Martin 1987, 131 calls the intervening -a- the ‘subjunctive’.

160

Chapter 6. Internal evidence contradicts

second ‘particle’ without satisfactory explanation. If we treat nite as the te-gerund of a defective copula as is suggested by Martin, then we can account for both the instrumental semantics and the phonological shape of the form. However, the treatment of ni as a conjunctive verbal form overrules the proposed external comparisons for the instrumental case marker. 7. e- ‘can (not) ...’ The quality of the Old Japanese vowel e2- indicates that this prefix is cognate with the verb eru, OJ u ‘get, obtain’. Considering the numerous occurrences of this prefix as a potential e‘can, be able’ and the fact that as a potential it often appears in negative structures, I suspect that the use of e- ‘can (not) ...’ as a negative potentialis is secondary. The internal evidence provided by Miller to treat this prefix as a -redundant- negative is weak, and I will therefore consider the prefix as a grammaticalization of eru, OJ u ‘get, obtain’ and disregarded the proposed external cognates. 8. -g- verb formant Martin explains the occurrence of the verb formant -g- in the examples cited by Miller as 48 reductions of the formant string pJ *-na-ka-. It can also be remarked that Miller has narrowed the gloss for the verb OJ pag- ‘attach, insert, fletch, feather an arrow’ to ‘make an arrow by fitting a feather to a shaft’ in a way that a denominal derivation from OJ pa ‘feather’ looks more convincing. But in view of the verb hameru ‘insert, put on one’s fingers, entrap’ with congruent register, a deverbal derivation from a root pJ *pa- ‘attach’ looks equally plausible. If there is no evidence apart from the Japanese examples to distinguish the formant pA *-g- from its voiceless counterpart, it is extremely daring to base the entire reconstruction on a suspect Japanese morpheme consisting of only a single phoneme. Therefore, the proposed etymology will be dismissed. 9. gari ‘to (where one is)’ dative-allative case marker The dative-allative ge ‘to a person’ that is found in northeastern Japanese dialects is probably a contraction of this Old Japanese case marker, as found in the example OJ kimi gari ikeba ‘if I come to my lord’.49 In Japanese the liquid phoneme r- is never allowed word initially, but it may occur suffix initially. Japanese case markers, however, behave more as separate words than as agglutinative suffixes. The farther back one goes in the history of Japanese, the fewer case markers were attached as an unstressed element to the preceding noun. Martin presumes that most case markers had independent accentuations and independent phrasing at an earlier time.50 A morphological boundary splitting gari into three separate case markers seems doubtful, especially because we need to assume an initial r- for the last case marker. An alternative explanation is to explain the case marker as a grammaticalization of the conjunctive form of the verb root pJ *kara- underlying OJ kare2- ‘get apart, cease, go away’. The contraction of a preceding allative case marker ni with the verb stem, could

48 Martin 1987, 683, 721, 775, 791. 49 Martin 1991 b, 49-50. 50 Martin 1987, 169-70.

6.2. Morphology

161

account for the initial voicing. (< pJ *n(i) kari ‘going away towards’). I suspect that the source marker kara ‘(going away) from’ is also related to this reconstructed verb root. 10. -i conjunctive (renyo@kei) ending; deverbal nominalizer When dealing with deverbal nouns in paragraph 6.1.1.1.1. it has been remarked that the conjunctive ending attached to the verb stem is -i . However there is no consensus about the exact proto-form underlying this ending. For the purpose of the present work I use the notation pJ *-(C)i, unable to specify whether the initial consonant is zero, a palatal glide *y, a velar *k or yet another phoneme. The objection to the reconstruction of the infinitive ending as *-i is the assumption that at the earliest stages of Japanese adjacent vowels within a word were not permitted. The objection to the reconstruction of a velar is that the only motivation comes from comparison with Korean and such an approach would be methodologically circular. Since there is no consensus in linguistic literature about the individual Japanese proto-form, any external comparison build on such a debated proto-form can collapse like a house of cards. 11. -(C)i- suffix which reverses the transitivity of the verb base The same objection can be made as for the previous morpheme. Given the fact that I have argued against voice distinction in proto-Japanese in paragraph 5.1.1., it goes without saying that I reject the reconstruction of pJ *-gi, which is the consensus reconstruction proposed by Vovin, Takeuchi, Martin, and Unger. If a velar consonant can be reconstructed here at all, it would be *-k-. But, I object to the reconstruction of the transitivity flipper pJ *-ki- because the velar initial is evidenced by external evidence alone. Therefore, I omit this etymology from consideration. 12. -in-/ -ite- perfective auxiliaries In Old and Middle Japanese we find two markers of the perfective aspect: -in- and -ite-.51 Although both auxiliaries seem to have identical functions, indicating that an action is completed on the point of time from which the event is viewed, they are not mutually interchangeable. The choice between -in- or -ite- largely depends on whether the event is controlled by the subject. Aside from unidirectional motion verbs that regularly construct with -in-,-ite- is used in sequence narration as a perfective aspect when the event is controlled and lack of control is expressed by -in-. Slightly different tense and aspect systems are discernible in embedded discourse. Inside embeddings -ite- shows the characteristics of recent past tense (imperfect aspect) and -itari / -ite ari takes over its functions of perfective aspect in a controlled event.52 From the morphological index it can be seen that Whitman and Martin relate the perfective OJ -in- to the MK perfective attributive suffix MK -(o/u)n, and that Vovin increases the comparison with Tungusic and Turkic so called ‘relic past attributive’ forms.53 51 OJ -in- has two allomorphs: -in- and -n-. OJ -ite- has two allomorphs: -ite- and -te. 52 Takeuchi 1988; 1999 provides a clear overview of the function of OJ -ite- and -in-, illustrated with examples. 53 The interpretation of *-n in Tungusic as a past attributive is not without problem. Benzing 1955, 1006: “pTg *-n Art Partizip einer unvollendeten Handlung.”; 1088: “Das Präteritum”.

162

Chapter 6. Internal evidence contradicts

Frellesvig reconstructs an alternating copula pJ *tV ~ *nV, having reflexes in both perfective forms OJ -in- and -ite-. Vovin compares OJ -ite- with the retrospective marker MK -te-/-taand the OTk. past tense suffix -dï/-di . Yet, convincing internal etymologies have been proposed that make it needless to look for external morphological cognates for OJ -in- and -ite-. It is interesting to start with the observation that the perfect auxiliary OJ -in- belongs to the n-irregular verb paradigm (na-hen) along with only two other verbs: in- ‘go away, depart’ and sin- ‘die’. In paragraph 6.1.1.1.2. these verbs are analyzed as complex verbs, sharing the monosyllabic verbal root *na- ‘go’. It is commonly agreed that OJ in- ‘go away, depart’ is to be considered as the independent lexical source for the perfective auxiliary OJ -in-.54 Takeuchi has demonstrated how the directional, source orientated semantics of the verb OJ in- ‘go away, depart’ could have developed into the denotations ‘anteriority, source and involition’ of the perfect auxiliary OJ -in-.55 In paragraph 6.1.1.1.2. I have reconstructed the root pJ *ta- ‘reach, come’ underlying the transitivity pair itaru ‘arrive, reach, attain’ and itasu ‘do, cause, bring about’. Starting from the original lexical opposition between pJ *na- ‘go’ and pJ *ta- ‘reach, come’, the grammaticalization into the attested perfective system with a source-goal distinction can easily be understood. Takeuchi argues that the goal-oriented auxiliary, our pJ *ta-, was reinterpreted as a recent past inside discourse probably before Old Japanese.56 It is a common pathway in grammaticalization that directional verbs like ‘come’ and ‘go’ develop into tense or aspect auxiliaries. The way that the directional verb J kuru ‘come’ participates in the formation of a recent past in modern Japanese is suggestive of this development. The opposition between source and goal orientation inside discourse was then maintained by OJ -in- and the innovative perfect -itari / -ite ari. By consequence, it is needless to look for external cognates for both OJ -in- and OJ -ite-. Martin’s comparison of the independent lexical form J itar- ‘reach’ with toT- ‘run’ is doubtful since the Korean form semantically lacks directionality and since according to Martin’s own analysis the -r- in the Japanese form is an intransitivizing formant. A better phonological and semantic match however, may be found in the first member of the compound tatoT- ‘reach’ analyzed as ta(h)- ‘reach’ + toT- ‘run’.57 The final -h of K tah- ‘reach, get to’ is probably. not intrinsic to the stem since it can be regarded as a causative formant derived from MK ho- ‘do’.58 Although the independent verbal origin of the Japanese perfective auxiliary OJ ite- factors out the suggested external morphological cognates in the index, the . connection with the Korean retrospective auxiliary MK - te/a- as proposed by Martin and Vovin is not excluded. It is possible that Korean verb pK *ta- ‘reach’ developed independently along the same pathway into the Korean retrospective. Maybe this verb is exactly the ‘full-fledged verb’ Martin declares to be looking for as the source of the retrospective auxiliary MK -ta-. It is generally agreed that the -te (2)54 55 56 57

Hashimoto 1969, 358; Akiba-Reynolds 1982; Takeuchi 1999, 108. Takeuchi 1999, 97-108. Takeuchi 1999, 107. . K ta(h)- ‘reach’ also occurs in the first member of the compound MK ta- Gwat- ‘approach, defy’ < ta(h)- ‘reach’ + pat- ‘butt’. 58 Whitman 1999, 7: “-h in K nàh- ‘give birth to’ appears to be the oldest transitive / causative suffix.”

6.2. Morphology

163

gerund and the derived -ta modern perfect (< -t(e) a(r)-) are inherited forms from the perfect auxiliary -ite-. 59 An argument in favor of this proposal, is the independent accentuation of the gerund in Middle Japanese, which implies a juncture. 13. -k- verb formant Miller supports his reconstruction of the verb formant pJ *-k- by two examples that are unhappily chosen. The first verb is mizuku, OJ mi1duku ‘get soaked in water’ which can be derived from OJ mi1 ‘water’ followed by some nasal marker, maybe n(i) and, OJ tuk‘soak’. For siku ‘spread out’ there is no internal evidence that suggests that the underlying root is pJ *si-, rather than pJ *sik-, so there is again no trace of a verb formant pJ *-k-. However Miller’s suggestion is not entirely ungrounded because Unger and Martin have accumulated evidence for a formant pJ *-ka-, that Unger attempts to describe as a formant 60 indicating punctual or iterative action. If Unger is right that pJ *-ka- is used to express either a sudden change of state (punctual) or an action that occurs repeatedly (iterative), then pJ *ka- ‘go away’ must be considered as the verbal auxiliary responsible for this aspect. Above the reconstruction of pJ *ka- ‘go away’ has been supported by OJ kare2- ‘get apart, cease’ (< pJ *ka(-)ra-Ci-), OJ kaye- ‘get apart’ (< pJ *ka(-)ya-Ci-), kayou ‘ply between, commute, frequent, go’ (< *ka(-)ywopa-), OJ tonakar- ‘fly up, jump up’ (< pJ *tona- ‘fly’ + *ka-ra-), and iku ‘go’ (< pJ *i-ka-). Cross-linguistically the grammaticalization of verbs for ‘go’ into markers of a change of state (e.g. Eng. I went mad) or into durative aspect markers (e.g. Negerhollands lo(o) < Dutch lopen ‘go, run’ is used as a durative auxiliary) is commonly observed.61 However, it is not possible to pinpoint the exact meaning of the formant in question. Therefore, we do not have a well defined functional load to match with the meaning of the proposed verb formants in the other languages. Besides it can be remarked that, if the proposed cognate pK *-k- is a legitimate verb formant at all, its meaning remains unclear. 14. -ki attributive marker for adjective stems / -si conclusive marker for adjective stems The Old Japanese adjectival paradigm displays a k- / s- consonantism, in which -si is used to build the conclusive form, -ku for the nominalizer and conjunctive form, and -ki for the attributive form of adjective stems. Parallel to these forms, there is a secondary paradigm with the attributive -karu and the conclusive -kari, that is build on contractions of the nominalizer -ku and the existential verb OJ ari . A similar paradigm is found in some Ryu@kyu@ dialects that combine the nominalizer -sa with the existential verb aru in order to create -saru for the conclusive form and in some dialects -sa-based paradigms coexist with -ku-based paradigms. This complementation of -sa-based paradigms and -ku-based paradigms in some Ryu@kyu@ dialects is suggestive of the way in which the conclusive marker OJ -si and the attributive marker OJ -ki participate in the adjectival paradigm. In this light, Martin’s proposal that the conclusive marker OJ -si and the attributive marker OJ -ki are respectively 59 Martin 1987, 121, 244; Takeuchi 1999, 108. 60 Unger 1977, 132 ; Martin 1987, 672, 796. Unger mentions 54 verb stems with the formant *-kain his study. Martin finds 32 verb stems with the formant *-ka- in his list of verbs, and refers to Unger for an attempt to find a meaning for *-ka-. 61 Heine and Kuteva 2002, 157-58.

164

Chapter 6. Internal evidence contradicts

truncations of the nominalizers -sa and -ku followed by the existential verb OJ ari, gains credibility. This proposal does not require to trace the complex nature of the Ryu@kyu@ morphology directly back to proto-Japanese, but it takes the Ryu@kyu@ morphology as an illustration of how two different adjectival paradigms can coexist. Unger and Tomita’s study (1983) confirms Martin’s view that the Japanese adjective inflection is a relatively young phenomenon. It can be assumed that originally in proto-Japanese the adjective was just the stem. Adjectives did not undergo inflection in proto-Japanese, they were just a subclass of nouns expressing quality and for semantic reasons they could take certain endings, like -sa and -ku that other nouns could not. In Old Japanese the uninflected stem of the adjective is still used to modify nouns directly in attributive constructions and the attributive use of -ki is relatively uncommon. There are examples in Old Japanese that show that the adjectival endings -si and -ki were competing for both the attributive and the predicative use, rather than being in complementary distribution.62 The competing use of OJ -si and OJ -ki further supports Martin’s internal etymology and casts doubt on the external comparisons for both adjective markers. If there was no adjective inflection in proto-Japanese, the adjective morphemes simply can not be inherited from Altaic. Contrary to Miller’s remark that “Martin’s formulation is justified in terms of internal reconstruction within Japanese, but it remains essentially a descriptive, not a historical statement, and it cannot be maintained against historic evidence from the comparative method when that evidence clearly contradicts it.”, Martin’s internal etymology still is a historical statement and a historical reconstruction cannot be wrong and right at the same time.63 If anything this remark illustrates Miller’s overaccentuation of comparative evidence. 15. -ki1 past auxiliary in the conclusive form / -si past auxiliary in the attributive form. The auxiliaries -ki1 / -si are attached to the conjunctive form of a verb to build a conclusive and attributive past ending in Old Japanese, following exactly the opposite pattern than in the above -ki attributive and -si conclusive markers for adjective stems. Just as in case of the adjective endings there is a certain competition by -ki1 for both the attributive and the conclusive form, as for example in the nominalization OJ -ke1ku < *-ki aku or in the relic omoi-ki ya ‘did one think?’ before the question marker ya. The past auxiliary -ki1 functioned also as conjunctive and combined with a following ari existential auxiliary to produce the past evidential -ke1ri. There is no consensus as to whether the perfect auxiliaries share the same origin as the adjective endings or whether the perfect auxiliaries must be derived as the conjunctive from two different verbal roots. The first view is presented by Martin (1997) and Whitman, the latter view by Martin (1995), Quinn and O¤no. The verbal source is thought to be the conjunctive form ki1 from OJ ko- ‘come’ for the first auxiliary and the conjunctive form si from OJ se- ‘do’, or perhaps a reduction of the conjunctive form s(ar)i from OJ sar- ‘leave, move away from’ for the second. The directional verb explanation based on ‘leave’ and ‘come’ looks particularly attractive in the light of the internal analysis of the perfective pair -in-/ -it- for which source and goal oriented directional verbs are also thought to have grammaticalized into tense and aspect distinctions. It is also interesting to note that the conclusive past form from OJ ko- ‘come’ is not *kiki, but kosi or kisi instead. 62 Martin 1967 a, 260; 1987, 807. Takeuchi 1999,177-180. Unger and Tomita 1983. 63 Miller 1971, 24.

6.2. Morphology

165

A certain awareness of the verbal source of -ki1 may have prevented speakers of Old Japanese attaching the ending to OJ ko- ‘come’. If we trace the past auxiliary OJ -ki 1 back to a goal oriented directional verb, we reject the proposed external cognates although there might be some connection. Elsewhere in Altaic a directional verb could have developed independently along similar lines into a past suffix. It can be remarked that the proposed . . ke/ kacan be traced back to the source oriented directional verb cognate perfective MK . MK ka- ‘go’. 16. *-m hypothetical nominalizer reconstructed for words in accent class 2.5 The asterix in the form of the entry rings an alarm bell. Internal Japanese data may preserve evidence for some hypothetical morpheme, but they do not tell us that the nature of that morpheme was pJ *-m, nor do they tell us that its function was nominalizer. Besides, the internal evidence comes from suprasegmental features and Ramsey thinks the accent class in question is secondary.64 Therefore, I take the liberty to call this a ghost morpheme. 17. -mi1 gerund in the construction noun phrase-wo adjective-mi1 In Old Japanese constructions like apu yo2si-wo nami 1 (M 213) ‘meet way-(acc.) not be (neg. aux.) / because there is no way to meet’ and pi1to2 me2 wo opomi1 ‘person eye-(acc.) be many (adj.) / because there are many eyes’ (M 207) occur frequently. The presence of the accusative case marker wo in this construction calls for a verb and therefore argues for a verbal interpretation of the -mi1 gerund. As early as 1929 Pierson drew attention on the 65 conjunctive (renyo@kei) of OJ mi1- ‘see’ as a plausible candidate for the verb in question. The conjunctive form of a verb ‘see’ would account for the causal semantics ‘because, and so’ of the construction in terms of ‘seeing (noun phrase) as (adjective); considering (noun phrase) as (adjective)’, like the English expression ‘in view of’. Martin proposes that OJ -mi1 in the construction noun phrase-wo adjective-mi 1 is the conjunctive (renyo@kei) form of a quadrigrade verb that underlies the achievement formant -m- that derives intransitive verbs from adjectives and nouns like the deadjectival verbs itamu ‘get hurt’, kuramu ‘get dark’, karomu ‘get light’, nurumu ‘get tepid’, yasumu ‘get at ease’ and the denominal verbs haramu ‘get pregnant (from hara ‘belly’), marumu ‘become round’ (from maru ‘cirkle’) and siwamu ‘get wrinkled’ (from siba ‘wrinkle’). However, as Martin points out himself, there are arguments against identifying OJ -mi1 as the conjunctive (renyo@kei) of the verb formant in question. First, there is the fact that although the noun phrase-wo adjective -mi 1 construction is widely attested and highly productive in OJ, since adjective -mi 1 could probably be made from any adjective, the OJ verb formant -m- is limited in attestation and was no longer productive in OJ. Second, the formant -m- derives verbs from both adjectives and nouns, while the construction noun phrase-wo adjective-mi1 is restricted to adjectives and adjectival nouns only. Third, although both forms are probably verbal in origin, the verb formant -m- is clearly intransitive, while OJ -mi1 appears only in constructions with the accusative case suffix -wo. Therefore, it is probable that two different verbs underlie verb formant -m- on one hand and OJ -mi1 on the other, the former an intransitive verb meaning ‘get, turn’ and the latter an transitive verb meaning ‘consider, deem’. For the intransitive 64 Ramsey 1979, 1980. 65 Pierson 1929, 86-87.

166

Chapter 6. Internal evidence contradicts

verbal origin of the achievement formant I have suggested the root pJ *ma- ~ *mo- ‘turn’ which is reflected in the OJ monograde verb mi2- ‘turn’ above. The interpretation of OJ -mi1 in the construction noun phrase-wo adjective-mi1 as the conjunctive form of OJ mi1- ‘see’ is acceptable and accounts for the transitivity of construction. Even if there is no consensus on the identity of the verb underlying OJ -mi1 in this construction, there is agreement on the fact that its source is verbal and that we are dealing with yet another case of grammaticalization whereby an autonomous verb acquired a grammatical status. Additional support for the observation that the origin of OJ -mi1 in this construction is not a mere gerund, that derives nominal forms from (quality) verbal stems is the fact that, contrary to the gerund for adjective stems -ku, the relics of OJ -mi1 can be found in contemporary Japanese attached to nouns expressing a quality as well: sinken-mi ‘ earnestness’.66 Establishing cognates it is needless to look for an external morphological cognate, the attention should perhaps be directed to the verb miru ‘see’ in the lexical index. It can also be noticed that Vovin provides the same Korean cognate for both the presumptive -am- and OJ -mi1 in the construction noun phrase-wo adjective-mi1 . His etymology for the . . Japanese presumptive -am- concerns the Korean assumptive K -(u)m a < MK - wu/o m a ‘I will, let me’ that can be derived from the nominalizer K -(u)m and the postsubstantive K 67 a. For OJ -mi1 Vovin proposes the Korean connective -mye. But in this way he overlooks the internal derivation of the Korean connective -mye as the nominalizing -m followed by the copula infinitive i-e68. 18. ni dative, locative, allative and mutative case marker, time marker ‘at’. Contrary to Frellesvig’s suggestion, Martin mentions that the internal analysis of ni as the defective copula infinitive is only valid for the mutative, antithesis, purpose, and manner uses of this marker, but it cannot be applied to the case marking functions.69 Although there is no conflicting internal analysis available for the case marker ni, it can be understood that the morpheme boundary inserted by Murayama between *n and *i as well as the one inserted by Miller between *n and *ri is illegitimate from the internal point of view. This leaves us with the external comparison proposed by Finch with the Korean topic marker, the Tungusic partitive, the Mongolic genitive and the Old Turkic instrumental, all of which are functionally overpermissive. 19. -p- intensive verb formant Martin and Unger agree with the reconstruction of a verb formant -p- (< pJ *-pa-), but there is no internal evidence for a clear-cut distinction between two different -p- verb formants in Japanese, as Miller suggests. Unger finds that all the verbs that share a -p- formant share the characteristic that they involve a high degree of intensity.70 But this common denominator

66 67 68 69 70

Martin 1987, 804. Martin 1992, 888. Martin 1992, 892. Martin 1991 b, 41. Unger 1977, 138 ; Martin 1987, 672, 796.

6.2. Morphology

167

is rather vague and a poor functional match to the Tungusic causative or reflexive suffix. Therefore, I will disregard the proposed etymologies. 20. pe1 allative case marker It is generally assumed that the allative case marker gramaticalized from the once independent noun OJ pe1 ‘shore, vicinity, side, direction’, which -as can be seen in the examples under mae ‘front, forepart’ - now occurs mainly as a bound noun in words indicating a temporal or spacial dimension. In our search for cognates we should direct our attention to independent lexical items in the supposedly related languages. 21. -r- / -s- transitivity polarizing formants A number of Japanese verbs polarize in transitivity pairs like OJ itar-/ MJ itas-‘reach / achieve’ and OJ nar- / nas- ‘become / create’ in which both intransitive and transitive are to be derived from some non-attested basic form. Other pairs of Japanese verbs that occur in the lexical index and that may be cited in relation to transitivity polarization are amaru / amasu ‘be in excess / increase’, hiru / hosu ‘get dry / dry’, horobiru / horobosu ‘be ruined / ruin’, huru / husu ‘fall down / lie down’, hueru /huyasu ‘increase (intr.) / increase (tr.)’, karu / kasu ‘borrow / lend’, kiru / OJ ke1s- ‘wear / clothe’, koeru / koyasu ‘get fat / fatten’, miru / miseru ‘see/ show’, mureru / musu ‘be steemed / steam’, naoru/ naosu ‘recover / repair’, niru / niseru, OJ no2s- ‘resemble / imitate’, noru / noseru ‘ride/ load’, OJ opi2- / OJ opos- ‘grow up / grow’, oru / osu ‘exist/ govern’, otoru / otosu‘be inferior to / drop’, taru / tasu ‘suffice / add’, tooru / toosu ‘pass by / let (someone) pass’, yoru / yoseru ‘approach / bring near’. It seems that the phonologically overlapping part of those pairs can be reconstructed as the original root to which a transitivizing pJ *-sa- or an intransitivizing pJ *-ra- was suffixed. Takeuchi remarks that expressing whether a predicate is transitive or not by derivational means is reminiscent of Ainu and some Austronesian languages which have incorporating morphology, coding grammatical relations by means of affixes in the predicate.71 Transitivity marking as a structural feature might be a trace of some Jo@mon substratum in Japanese. But this is beyond the scope of the problem we have to deal with at present. Miller, however, believes that these transitivity pairs cannot possibly be analyzed as resulting from the addition of suffixal formants or enlargements of the roots. He finds that they are completely anomalous as far as Japanese morphology is concerned and believes them to reflect an original *-l- / -l2- transitivity dichtomy in Altaic. Miller establishes this etymology mainly on the basis of OTk. material and he is particularly struck by the lexical and morphological parallels between the Japanese pair taru / tasu ‘suffice / add’ and OTk. tol- / tos #- ‘be full / fill’. The fact that there is shared idiosyncratic morphology that cannot be explained by internal derivation within Japanese or Old Turkic themselves leads Miller to the conclusion that ‘these forms alone would probably be sufficient to demonstrate the genetic relationship of Japanese to Old Turkic’.72 Unfortunately, the reality is none too bright. It is sometimes argued that the existential verb OJ ar- ‘be’ < pJ *ara- is the lexical form from which the endoactive stem formants, i.e. intransitivizing -r-, intransitivizing/ 71 Takeuchi 1999, 94. 72 Miller 1971, 135. sic.

168

Chapter 6. Internal evidence contradicts

passive -ra- and passive -rare- have developed.73 However, Martin also suggests that the existential verb aru ‘be, exist, have’ can be derived from the existential root pJ *a- and a verbal suffix pJ *-ra-. There is no functional or formal reason to assume that this pJ *-ra- is a different formant than the *-ra- underlying the intransitivizing -r-. So explaining the pJ *-ra- that builds the existential verb as a derivation from the existential verb is circular. If -following the argumentation under aru ‘be, exist, have’ above- pJ *a- is a legitimate reconstruction for the existential verb, then pJ *-ra- is the original intransitivizing formant and pJ *a-ra- is a secondary derivation. This does not exclude the possibility, however, that the transitivizing *-sa- from which the exoactive stem formants, i.e. transitivizing -s-, transitive/ causative/ exalted -sa- and causative -sase- have developed, is derived from the naked causal root pJ *so-~ si(C)a- that is reflected in OJ so2- ~se- ‘do, make’ following 74 Martin’s suggestion. The a-vocalism of pJ *-sa- could be explained through vowel alternation or through reduction of the original form in analogy with the intransitivizing pJ *-ra- that preceded the transitivizing suffix. Development from a causal verb into a causative is a pathway that is typical in . grammaticalization processes around the world. The fact that Korean has a -h- (< ? MK ho- ‘do’) suffixed to the verb stem, where Japanese suffixes -s(< pJ *so- ‘do’) is probably due to this universal tendency, while both verbs are directly cognate. Although the origin of pJ *-sa- is an independent verb, it may be possible to provide extra-Japanese etymologies for pJ *-ra--, but Martin’s proposal to relate the . prospective modifier K -ul(q), MK ul(q) ~ ol(q) is rather weak. While pJ *-ra- is a formant that expresses that the verb root, which is of neutral transitivity, is used in its intransitive interpretation, the Korean form is a prospective modifier that derives nominal and adnominal verb forms. This comparison is functionally overpermissive. As for the striking similarities with Old Turkic remarked by Miller, although the morphological part, i.e. the transitivity dichtomy correspondence, is most probably due to coincidence, the lexical correspondences of the verbs reflecting transitivity pairing may still hold. Besides, it is possible that a similar process of grammaticalization was independently active in Old Turkic, without supposing a genetic connection to the Japanese process. Although Starostin agrees with Miller as far as the reconstruction of a four fold Altaic liquid system is concerned, he does not follow Miller in his etymology for the polarized transitivity formants. The problem with Starostin’s etymologies for Japanese verbs that may be cited in relation to transitivity polarization is that he often disregards the internal Japanese derivation with the -r- / -s- transitivity polarizing formants . 22. -ra collective formant Above it has been argued that the plural marker OJ -ra is etymologically related to the presently unproductive suffix *-ra in OJ iku-ra ‘how much, some amount’, nani-ra ‘what, some’, koti-ra ‘this direction/ person’, ati-ra ‘that direction/ person’ and doti-ra ‘which direction/ person’ etc. and to the *-ra in the pronominal system. The semantic source that unites these different uses of pJ *-ra, deriving both plural and singular countable nouns, rather than only a plural marker, is that of ‘(defined) quantity, group’. Therefore, I have chosen to label -ra as a collective formant. The interpretation of -ra as a directive case 73 Whitman and Martin in index; Takeuchi 1999, 94. 74 For the reconstruction of the proto-form for ‘do, make’, I refer to paragraph 10.1.1.3. below.

6.2. Morphology

169

suffix as suggested by Menges, Miller and Finch is based on the semantics of the preceding lexeme in some isolated cases like koti-ra ‘this direction/ person’, ati-ra ‘that direction/ person’ and doti-ra ‘which direction/ person’. Above it has also been remarked that a naked verb root could serve as a nominal basis and in this light it is easily understood how naked verb suffixes like the intransitivizing -r< pJ *-ra- could serve as nominal suffixes. Transmitted to the noun the intransitivizing semantics ‘exist, be’ gave rise to the collective formant ‘what exists, what is, person, thing, amount’. Although I do not accept Martin’s derivation *-ra < *ara-, I do agree with his internal analysis that both the collective formant -ra and the intransitivizing -r- must be derived from the same original morpheme in proto-Japanese. From this viewpoint Miller’s comparison with the Tungusic nomen usus pTg *-ra- should be further examined in what follows. 23. -ri counter for persons, adverbializer People are usually counted by -nin in Japanese, but for hitori ‘one person’ and hutari ‘two persons’ the counter -ri is used instead. If this counter is complex like Martin suggests by deriving it as a deverbal form from the existential verb, or, following the above argumentation, as a deverbal noun formed by adding the infinitive ending -i < *-(C)i to the verb formant pJ *-ra-, then it could ultimately be the same combination of morphemes that is reflected in the adverbializer -ri. The adverbializer is present in adverbs such as assari ‘simple, plain’ (asa-asa ‘simply, lightly’), sukkari ‘completely, totally’, surari ‘smoothly’ (sura-sura ‘without a hitch, smoothly’), etc. Although a complex derivation of this formant overrules the proposed external cognates, it is interesting to observe Miller’s match with the nominal stem formant pTg *-ri@. Benzing segments the Tungusic nominal stem formant as the predicative stem formant pTg *-ra followed by a nominalizer *-gi@ which is reminiscent of the Japanese 75 internal analysis. 24. *-ro hypothetical adverbial suffix I am not aware of the existence of an adverbial suffix -ro in Old Japanese. Since Finch does not provide any examples of attestations of this morpheme, I will consider this item as a ghost morpheme. 25. *-ru hypothetical directive suffix Since the reconstruction of this suffix is based on one single doublet yo and yoru ‘evening, night’ it consists too weak a basis for further comparison. 26. OJ si ‘wind, direction’ This word is not listed in the morphological index since it is attested as a lexical item in Old Japanese. Apart from set expressions such as OJ sipuk- ‘the wind blows’, this word does not occur as a free morpheme in Old Japanese, but as a bound noun it still occurs in a number of compounds such as arasi ‘stormy wind’ (with ara- ‘rough’), sitomi ‘latticed shutters’ (with *tomi as a deverbal noun from the root pJ *toma- ‘stop’), higasi ‘east’, etc.

75 Benzing 1955, 1076-77.

170

Chapter 6. Internal evidence contradicts

Miller and Menges erroneously interpret this word as a directive case marker and overlook the free lexical source of the morpheme in question. 27. OJ sip- ‘lose the function of a sensory organ’ This verb is not listed in the morphological index since it is attested as a lexical item in Old Japanese. There seems to be no convincing internal evidence to insert a morpheme boundary in this word, let apart to analyze it as a collocation of two derivational suffixes. 28. -t- telic verb formant In the light of the above analysis of the perfective OJ ite- it is likely that this verb formant ultimately reflects the same auxiliary pJ *ta- ‘reach, attain’. This suggestion is reminiscent of the discussion under the verb formant -k-. 29. -tati plural formant Although the formal and semantic parallel with K -tolh is striking at first sight, there is an equally plausible internal etymology for the plural formant that casts doubt on the Korean cognate. Martin suggests that it may be a deverbal noun from tatu ‘stand, rise, depart, run high’, in the sense of ‘a stand (of)’. Because this conflicting internal etymology is equally plausible, we will omit the proposed comparison from our core evidence for security reasons. 30. OJ ti ‘road, path, way’ This word is not listed in the morphological index since it is attested as a lexical item in Old Japanese. Miller’s label ‘directive suffix’ is suspect since the word is attested as a free noun meaning ‘road, path, way’ in Old Japanese. Above we have also remarked that there are a substantial number of compounds having ti as a bound noun such as timata ‘crossroads’, tigai ‘difference, divergency’, doti- ‘where, which way’, tinamu ‘be connected with’, etc. There is no internal motivation to treat this lexical item as the result of lexicalization of an original case marker. 31. to 1. subjective essive case marker ‘[thinking] it to be’; 2. quotation marker; 3. inclusive conjoiner ‘and’; 4. reciprocal and comitative case marker ‘with’; 5. conjunctionalization marker ‘if, when’. The way in which this marker has been etymologized as a single morpheme, tracing back its different uses to one original copula source, illustrates the risk of squeezing homonymous morphemes under one umbrella. Due to the syllabic nature of Japanese a serious number of homonymous forms can be found in the lexicon as well as in the grammar. It is highly probable, however, that some of the homonymous forms are of accidental similarity and homonymy at one stage of the language may have been polysemy at an earlier stage. Below I argue for the treatment of the different uses of to as morphologically converged reflexes of at least three different roots. to 1. subjective essive case marker ‘[thinking] it to be’; 2. quotation marker For the use of to as a subjective essive case marker and quotation marker there are two possible internal cognates. One candidate is an old deictic *to- ‘that’ of which relics are preserved in tonikaku < to-ni kaku-ni ‘anyhow, somehow or other, all things considered’,

6.2. Morphology

171

tonimo-kakunimo ‘id.’, OJ to sama kau sama ‘all directions’. The second candidate is to ‘even, also’ with a more widespread voiced variant do ‘even, indeed’. The fact that the Ryukyu reflex du is a different form from the comitative R tu confirms the postulation that we are dealing with different but homonymous roots to in standard Japanese. It could be argued that this emphatic use of to is a secondary development from the deictic use, but strong Korean cognates for both the Japanese deictic and J to ‘even, also’ being the Korean . distal deictic K ce, MK tye and K to ‘even, also’ respectively indicate that they represent two different roots and that the subjective essive and quotation marker to should be traced back to either one or the other, but not to both. Either way, it can be excluded that this use of to, like all other uses, has diverged from a single, obsolete copula. to 3. inclusive conjoiner ‘and’; 4. reciprocal and comitative case marker ‘with’; Since the development of a comitative ‘with’ into a noun phrase conjoining ‘and’ is crosslinguistically observed (e.g. French avec ‘with’ > French based creole of Haitian ak ‘and’), it seems safe to assume that the conjoining of nouns by to ‘and’ is a special case of the use as a comitative case marker ‘with’.76 The Ryukyu reflex of this marker is tu . Martin draws attention to a likely cognate of the Altaic comitative case marker, but he does not specify which Altaic forms exactly he has in mind. In spite of Whitman’s proposal K te ‘and, more’, Martin thinks that a Korean cognate for the comitative is lacking. to 5. conjunctionalisation marker ‘if, when’. It is clear that the conjunctionalisation marker is a grammaticalization of the noun OJ to/ to1 ‘place, time’ which is listed in the lexical index and occurred as an independent noun in Old Japanese. One could think of OCh. *d++ ‘time’ as a possible loan source for OJ to/ to1 ‘place, time’, but this is in conflict with one of the most frequently cross-linguistically attested metaphors. In nearly every language of the word locative concepts 77 are extended to refer to temporal concepts and not the other way around. So I assume that the meaning ‘time’ is a secondary development from the original meaning ‘place’ that further grammaticalized into the conjunctionalization marker. Apart from the presence of the same initial syllable in toki 'time' and toko ‘place’ there is no internal motivation to accept Martin and Miller’s idea that this noun may be a truncation of toki and toko, both of which enjoy Korean cognates. The fact that following Whitman’s proposal to ‘place, time’ enjoys a plausible Korean cognate as well demonstrates the weakness of this argument. 32. tu genitive case marker Based on the observation that OJ tu was productive in Old Japanese, often expressing a temporal or locative relationship between two nouns, Miller and Murayama label the marker ‘locative genitive’. The marker is obsolete in contemporary Japanese, but it infrequently appears lexicalized in compounds. In Old Japanese, however, we find numerous examples, such as OJ asipara-no naka-tu kuni ‘middle land of reed fields’ (K 1, 124-129), ama-tu kami2 ‘heavenly god’ (M 904), soko-tu ipane ‘stone fundament in the bottom’ (K 1, 124-129), oki2-tu mopa ‘seaweed from offshore’ (N 1, 430-33), pe 1-tu mopa ‘seaweed along the shore’ 76 Heine and Kuteva 2002, 80-82. 77 Heine and Kuteva 2002, 205-06.

172

Chapter 6. Internal evidence contradicts

(N 1, 430-33), ama-tu norito ‘heavenly rules’ (N 1, 430-33), to2ki1-tu kaze ‘wind blowing at time of high tide’ (M 220) etc. It is exact that the majority of the constructions with OJ tu involve a locative or temporal relationship, but there are also examples such as asa-tu-ki ‘chives’ < ‘mild onion’ that indicate that OJ tu originally had a wider use as a genitive, not necessary involving a temporal or spacial relationship. Martin remarks that the non-adnominal -tu forms of the numerals may be the same genitive marker. Therefore, I will disregard attempts to connect this particle to the following ablative case marker and in paragraph 11.2.2.1. I will only take the comparison to the Korean genitive marker into consideration. 33. -u conclusive ending The external comparison that is proposed by Vovin and Miller is based on the hypothesis that pJ *-wu underlies the conclusive ending -u. However, *u and *wu are completely neutralized in Old Japanese and there is no internal evidence that favors the reconstruction of pJ *-wu instead of pJ *-u for this ending. The reconstruction of pJ *-wu on the basis of external evidence alone is methodologically circular, so I will disregard the proposed etymologies. 34. wa topic marker Martin is probably right in his suggestion that the topic marker is a grammaticalization of pJ *pa ‘place’. OJ ba ‘place’ can probably be derived from pJ *-n-pa. The voiced initial in the Japanese word can be accounted for as the result of analogy with compounds having the word for ‘place’ as the second member. Since the word for ‘place’ occurs more frequently in compounds than freely, as in J tatiba ‘standpoint’, J hiroba ‘market place’ etc., the original pJ *pa ‘place’ in isolation was probably replaced by its voiced alternant that is commonly heard in compounds. Internal evidence for the semantic extension from ‘place’ to ‘situation’ is found in J baai ‘case, situation’ that can be derived from ba ‘place’ and ai the deverbal noun of the verb ‘meet’. There is a good cognate for the Japanese word in Korean pa ‘place’ that will be analyzed in chapter 10. It is interesting to note that the same . grammaticalization as is assumed in Japanese has taken place in Korean: MK -n pa ‘situaton’. Benzing finds evidence for the reconstruction of pTg *-pa in several adverbs of place.78 Examples of adverbs of place are Na. boalpa ‘outside’, Ud. bu’afa ‘outside’, Na. dujlpE, dujpE ‘through / in the wood, through/ in the taiga’, Olc#. du¢jlpE ‘id.’, wajlpa, wajpa ‘along the riverside, along the shore’ and Ud. N’eœfa ‘id.’. At the first sight the Tungusic place suffix makes a nice cognate in the etymology for ‘place’. However, it can be remarked that the adverbial formant in question is to be reconstructed as *-lpa, with a liquid reflex that is preserved in Nanai and Olc#a, but lost in Udehe. 35. OJ -e 1 / -yo imperative ending The imperative marker is OJ -e 1 for consonant verbs and -yo for vowel and irregular verbs.79 Following Martin the imperative ending is build on the conjunctive form of the verb 78 Benzing 1955, 1028-29. 79 The imperative ending must be kept distinct from the hortative J -(y)o ‘let’s do it’ which is a reduction of the hortative ending -yoo following vowel stems. It goes back to the presumptive ending OJ -amu, for which the nasal consonant eroded between between the Heian period and the Edo period. For the hortative ending reference can be made to Martin 1987, 129-30.

6.2. Morphology

173

(renyo@kei -(C)i). This internal analysis explains the initial palatal glide of the -yo allomorph as a trace of the conjunctive ending and contradicts Miller’s speculative comparison with the Tungusic imperative suffixes. 36. yo1, yu, yori, yuri ablative case marker It is generally agreed on that OJ yo1, yu, and yo 1ri, yuri all reflect the same source, but there is considerable disagreement as to which form is secondary and which is primary. Following Martin OJ yo1ri is the primary form for the ablative case marker and OJ yo 1 is a reduction of that form. The way contemporary Japanese yori is abbreviated to yo in the construction yo ka (< yori ka ‘rather than ...’) and the contractions in the Ryu@kyu@ reflexes are independent, later developments, but they are suggestive of this phonological reduction. However, if we accept that OJ yo 1 is a dialectal borrowing from Azuma and that OJ yu displays the original vowel, then it follows that yuri is the primary form.80 This is in conflict with Martin’s speculation that the ablative case marker is a grammaticalization from the deverbal noun of yoru ‘select’, a proposal that is also in conflict with the quality of the vowel which is A in the ablative case marker yo 1ri and B in OJ yo 2r- ‘select’. The same phonological objection holds for an attempt to relate the ablative yo 1ri to OJ yo2r- ‘approach, gather,. rely on, be based on’, reminiscent of the connection between the Korean ablative MK pu the and the verb MK puth- ‘stick to, adhere; rely on; from’. By consequence I suspect that yuri < pJ *yuri is the original form of the ablative case marker, but the problematic vowel prevents the derivation of an original verb stem. This analysis is in conflict with Miller and Murayama’s claim that the ablative case marker OJ yu < pJ *du stands in original voice alternation with the so called ‘adnominal dative’ or ‘locative genitive’ OJ tu < pJ *tu. Supposing an original voice alternation between the so called ‘adnominal dative’ OJ tu and the ablative marker, Miller and Murayama claim that OJ yu < pJ *du is the primary form and that OJ yo 1ri is a secondary form due to the collocation of two different suffixes, i.e. the ablative pJ *du and the directive pJ *ri. However, above the connection with OJ tu and the reconstruction of pJ *d- has been rejected. Besides, if pJ *yu is the original ablative, we need to explain the identity of the second morpheme. It is suggested that this *ri is a relic of an old directive marker, but I am unable to find a secure internal motivation for this reconstruction within Japanese. Third, if OJ yo1 and yu reflect the original ablative pJ *yu and yo1ri and yuri reflect a collocation of two case markers pJ *yu-ri, we would expect a certain discrepancy in function between yo 1 / yu on one hand and yo 1ri / yuri on the other hand, but this is not the case. Therefore, I reject the idea that OJ yo 1 / yu and yori / yuri stand in morphophonemic alternation. Instead I follow Martin’s proposal, accepting that OJ yuri is primary and that OJ yo1 / yu are purely phonological reductions of this form. By consequence all external comparisons that are based on the insertion of a morpheme boundary between *yu and *ri will be disregarded.

80 Other examples of Cu ~ Co1 variation in Japanese that are thought to be the result of dialectal borrowing from Azuma are OJ nogo1p- ~ OJ nugup- ‘wipe’, OJ sugus-, Azuma OJ sugo1s-, J sugosu ‘exceed, let pass’, OJ suko1-si ‘little’~ OJ suku-na ‘scant’, OJ yuki1 ‘snow’, Azuma yo1ki 1, Kyo. yoki ‘snow’, Kyo. moromati ~ J muromati (Martin 1987, 69)

Chapter 7. Similarities due to general properties of language Not all resemblances between words across different languages are due to inheritance from a common ancestor. A set of correspondences involving nursery forms, onomatopoeia, sound symbolic words and morphological universals is likely to reflect non-inherited, ‘false’ sound correspondences. Nevertheless, it is indeed possible that such forms were present in the ancestor language and have shared a common history, but the probability that their resemblances are coincidental is too high to be admitted to the core evidence. In this chapter I will sift those entries from the covering etymological index for which the similarities are with a relatively high probability due to general properties of language. The etyma in the lexical index will be investigated on their probability of being nursery forms, onomatopoeia or sound symbolic words. For the etyma in the morphological index it must be checked whether their similarity can be due to universal principles in phonological structuring of morphemes. 7.1. Nursery words In the early sixties Roman Jakobson suggested that nursery words like mama, papa, baba, dada, caca, pipi etc. should be avoided as evidence for potential genetic relationships. The first sucking activities of a child are accompanied by nasal sounds and labial or dental stops, and become associated with the desire of food or the need of the parent. However entering the adult language and developing to common forms meaning ‘mother’, ‘father’, ‘grandfather’, ‘grandmother’, ‘uncle’, ‘aunt’, ‘sister’, ‘brother’, the straightforward sound symbolic shape of nursery forms fades to a certain extent and they are no longer recognized as such. Although there is a possibility that mama-papa forms are due to descent from a common source and have shared a common history, the probability that their resemblances are due to universals is too high to admit them to the core evidence.1 I have chosen to omit the following words from the core evidence because their phonological structure and their semantics may be the result of the mama-papa factor. 1. ane ‘elder sister’ Whether this word is to be derived from OJ a ‘I’ and ne(e) ‘older sister’ or else from ani ‘older child of one’s parents’ and me-, OJ me 1 ‘female, she-’ is irrelevant for this purpose since the nasal phonology and the meaning make the word suspect of being a nursery term.

1

I refer to Ruhlen 1994, 122-124 for the viewpoint that mama-papa words can be due to a common origin. This viewpoint is not in conflict with the approach I intend to adopt here.

7.1. Nursery words

175

2. ani ‘elder brother’ The male equivalent of the preceding word can be excluded from the core-etymology on the same grounds. 3. haha ‘mother’ The derivation from pJ *papa makes it clear that we are dealing with a mama-papa word. 4. OJ imo ‘male’s sister, beloved girl’, OJ imose ‘brothers and sisters’, imo@to ‘younger sister’ The shape of OJ imo is reminiscent of the following words me- ‘female, she-’ and omo ‘mother’. The labial nasal in combination with the meaning makes me suspect that this is a nursery form. 5. koku ‘break wind, make water, tell, excrete’; koku ‘flay, strip off (skin, leaves), thresh’ OJ ko1k- leads to the reconstruction pJ *kuaka- ‘break wind, make water, excrete’. In view of the universal nursery term caca, kaka ‘waste matter from the body’ it is safer to omit this etymon from consideration. Considering the quality of the vowel and the register of koku, OJ ko 1k- ‘flay, strip off (skin, leaves), thresh’ we are probably dealing with the same etymon. The semantics can be reconciled on the basis of the common denominator ‘remove from the body’. 6. me- ‘female, she-’; OJ womi 1na ‘woman’; onna ‘woman’ All these etyma share the formant pJ *mi ‘female, mother’ wich is directly reflected in a bound noun used to design female gods OJ -mi1 ‘god mother’ like the goddesses Izanami and Kamuromi in contrast to the male counterparts Izanagi and Kamurogi. The Azuma word corresponding to OJ me 1 ‘female, she-’ is also mi1. The female prefix me- occurred as an independent noun me 1 ‘female, she-’ in Old Japanese and regarding the quality of the vowel this word is believed to be a reduction of OJ -mi1na ‘woman’ that is present in OJ womi1na ‘old woman’. OJ womi 1na ‘woman’ is probably a compound of an unidentified prefix wo-, the same bound noun -mi1 and the na ‘name, person’ that also appears as a bound noun in mina, minna ‘all, everybody, everything’ and OJ oki1na ‘old man’. Contemporary Japanese onna ‘woman’ is a phonological reduction from OJ womi1na ‘woman’. Since the development from the word from ‘mother’ to a female formant is a common development in languages across the word (e.g. Ewe nO ‘mother’ > -nO ‘female’) and since the word has a simple labial nasal shape, the chance that we are dealing with a nursery term is too high to admit these etyma into the core evidence.2 Under the entry me- ‘female, she-’ Starostin remarks a possible connection with the following word, omo ‘mother’, which can probably be explained in the light of the phonological patterning of nursery vocabulary. 7. omo ‘mother’ In the Nihon Shoki OJ amo occurs as a synonym for omo ‘mother’. The labial nasal and the meaning turn the word into a likely candidate for a nursery form, perhaps connected to the previous word. 2

Heine and Kuteva 2002, 213.

176

Chapter 7. General properties of language

8. okaasan ‘mother’ The contemporary Japanese word okaasan comes from OJ kaka ‘mother’ whereby the medial velar has lenited. The combination of the shape and meaning in Old Japanese make it suspect of being a nursery word. 9. ozi ‘uncle’ In Ryu@kyu@ dialect the word for ‘uncle’ is buzja, pointing to pJ *wonti with initial *w-.3 If OJ wozi is to be regarded as an unidentified prefix wo- (< pJ *bo-) - possibly the same prefix as in OJ womi 1na ‘woman’ and R bui, OJ wopi ‘nephew’- followed by a nasal genitive, then the final bound noun ti is probably the same element as is reduplicated in titi ‘father’. Both meaning and form make this word suspect of being a nursery word. 10. titi ‘milk, the breasts’ Expressing the most basic desire of children and having a simple reduplicated phonological shape, this word is likely to be a nursery term. 11. titi ‘father’ Both the reduplicated shape and the meaning indicate that this word is a nursery form. 12. otoosan ‘father’ Parallel to okaasan ‘mother’, the contemporary Japanese word otoosan comes from OJ toto ‘father’ with elision of the medial consonant. The combination of the shape and meaning in Old Japanese make this entry suspect of being a nursery word. 7.2. Sound symbolism Whether it concerns real imitations of sounds or expressive words describing the looks of a situation or one’s physical or emotional attitude towards a situation, for similarities between sound symbolic words, it is always difficult to decide whether one is dealing with true cognates or accidental similarities. Therefore, sound symbolic words constitute a poor basis for establishing sound correspondences and must be omitted from the core evidence. The criterion that decides whether an entry is suspect of being sound symbolic relies on the phonological and semantic characteristics of the word in question. The phonological characteristics must be such that it is possible to interpret the shape of the word as iconic for its meaning. If the word in question is a verb, the semantics are ‘to imitate a real sound’ or ‘to perform a physical action processing a sound’ or ‘to affect the physical or emotional senses’. If the word in question is a noun, the meaning is ‘the production of a sound’ or ‘a being or instrument that produces a typical sound’ or ‘an intense physical or emotional attitude’. For both semantic and phonological reasons I suspect that the following words may be sound symbolic.

3

Whitman 1985, 16.

7.2. Sound symbolism

177

1. aegu ‘pant, gasp for breath, puff, breath heavily’ It is agreed that aegu, OJ ape 2k- is a compound of apa and iku, OJ ik- ‘breathe’. However, there is no ground to etymologize pJ *apa- as a separate verb also meaning ‘breathe’. It is more likely that we are dealing with a sound symbolic expression imitating the sound of gasping for breath. 2. akubi ‘yawn’ This word is a nominalization of MJ akub- ‘open and close the mouth with snaps, yawn’ which is probably an overlenition of pJ *paku-np-. Other possible examples of overlenition are given in paragraph 6.1.2.6. with abureru and OJ papur- ‘overflow’ , eri ‘neckband, collar, neck, lapel’ and heri ‘edge, fringe, border’ etc. The reconstruction of the root pJ *paku- is further supported by mimetical expressions like paku-paku ‘with snaps, taking big bites, biting sound of fishes’ and pakutto , pakuri ‘with ones mouth agape, openmouthed’ or pakuri taberu ‘snap up’, pakutsuku ‘munch, take quick bites of’, etc. 3. bokeru ‘senile’ Since voiced initials were not found in native words, the initial b- in this verb indicates that it is marked in some special way. The verb is probably a phonological reduction of boyakeru ‘become fuzzy’, the naked root of which is also present in the mimetic expression boya-boya ‘absentminded, fuzzy’. 4. utu ~ butu ‘beat, hit’ Again the initial b- in this verb is marked. The verb could be borrowing, a phonological reduction or an onomatopoetic word. Evidence that we are dealing with a mimetic fortition comes from the verb utu ‘beat, hit’. If utu reflects the original, unmarked root, the voiced initial in butu ‘beat, hit’ adds an intensive and sound expressive connotation. 5. damaru ‘become silent, stop speaking, say nothing’; domoru ‘stammer, stutter’ Just like for the previous items, the voiced initial d- in these verbs marks them as special, in a sense that they are likely to be borrowings, phonological reductions or a mimetic verbs. The similarity in shape with the English gloss ‘stammer’ gives an indication that the form is iconic for the meaning. Besides, the similarity in both form and meaning with the Korean onomatopoetic form tem tem ‘silent, taciturn’ seems rather due to sound symbolism than to inheritance from a common ancestral form. 6. doyomeku ‘resound, reverberate, stir’ Old Japanese also has to2yo 2m- ‘resound’ with initial t- for this verb. The voiced initial dprobably adds intensity to the sound expressive verb. The mimetic pair pJ *toyo- ~ *doyocan be considered as iconic for ‘reverberation’. In Old Japanese we also find the onomatopoetic expression to2do2 that is used to express the sound of knocking on a door or the trampling sound of horse hoofs. Another indication for the fact that *toyo ~ *doyo- is a sound symbolic pair comes from the auxiliary -meku that is often found following mimetic expressions, such as hatameku ‘flutter, flap’, kirameku ‘glitter, sparkle’, ugomeku ‘wriggle’, OJ sosomek- ‘fidget, rush; whisper’, OJ sabame1k- ‘murmur, buzz, clamor’, sasameku ‘whisper, murmur’, sazameku ‘make an uproar’, OJ sosomek- ‘fidget, move nervously,

178

Chapter 7. General properties of language

rush’, sosomeku ‘whisper’, tutumek-‘murmer’, wameku ‘scream, shriek’, zawameku ‘be noisy, rustle’. 7. emu ‘smile, beam’; wameku ‘cry, scream, shriek’; warau ‘laugh’ Considering the exclamation of lament OJ we (e.g. M 4, 486 ware pa sabusi we. I (topic) lonely (exclamation). ‘Alas, I’m lonely!’) which is likely to be an onomatopoetic expression, I suspect that there was a sound symbolic root *we perhaps in alternation with *wa for emotional exclamations of joy or lament, like the exclamations hihi, haha, hehe, waa etc. across many languages of the world. The prefix OJ we- in the verbs OJ wewarap-'laugh out loud' and OJ weturakas- ‘laugh behind one’s back’ supports this reconstruction. Deverbal derivations from the sound symbolic roots *we and *wa with the intransitivizing formant pJ *-ra- are found in warau ‘laugh’ (< pJ *wa-ra-pa-), in OJ weraku ‘laugh with joy’ (< pJ *we-ra-ka-) and probably also in the reduplicated deverbal onomatopoetic expression OJ wera wera ‘beaming with delight, beamingly’ (< *we-ra). I suspect that OJ wemu ‘smile’ is a deverbal derivation with the formant pJ *-ma- (< *we-ma-). The verb wameku ‘scream, scream, shriek’ that is also etymologized in the index is internally connected because it is composed of a sound symbolic wa and the auxiliary -meku that is commonly found in sound symbolic expressions. For reasons of iconicity I do not undo the sound changes in a regular way, which would lead to the expected reconstructions pJ *wa(C)i for *we. That phonological peculiarities such as the vowel *e in proto-Japanese are admitted in iconic expressions is illustrated by the occurrence of initial r- in contemporary Japanese expressions such as ran-ran ‘glaring, flaming’, rin-rin ‘jingling, tinkling’, rero-rero ‘speaking thickly, slurring one’s words’ etc. Approximations of natural sounds are not always strong enough to inhibit sound changes (e.g. OJ wem- > J emu). The sound of sheep . in Classical Greek [bœ ], for instance, is read as [vi] in the modern reading. 4 8. hataku ‘dust, beat, flap, clap, slap down’ Considering the formal and semantic overlap of this verb with the mimetic verb hatameku ‘flutter, flap’, I suspect that the verb is derived from the mimetic root pJ *pata which is iconic for flapping and clapping sounds. It can be remarked that sounds made by striking something crosslinguistically tend to have labial stop initials and dental stop medials like the English iconic expression pat for a sound made by striking something with a flat object or the Dutch iconic expression pats for a slapping sound. 9. hikaru ‘shine, be bright, glitter’ This verb is probably derived from an iconic root pJ *pika ‘sparkling, glittering, shining’ which is reduplicated in the contemporary mimetic expression J pika-pika ‘sparkling, glittering, shining’. That the sound change of the initial pJ *p- is inhibited in the adverb, but not in the derived verb is because the awareness of the iconicity is stronger in the case of the reduplicated adverb than it is in case of the derived verb. This observation is reminiscent of Eng. pipe, which is the regular outcome of the onomatopoetic verb MEng. pi@pen ‘make the sound of little bird chicks’, although the onomatopoetic replacement of inherited pipe, Eng. peep is the more common variant. 5 4

Anttila 1989, 86.

7.2. Sound symbolism

179

10. haeru ‘shine, glitter, excel in brilliancy’ This verb is probably derived from an iconic root pJ *paya ‘glitter, prickle, tingle’. This root is also reflected in a number of adjectives that express stimuli to the senses, such as syoppai ‘salty’, OJ sipopayu- ‘salty’, mabayui ‘dazzling, glaring’, kosobayui ‘ticklish’ etc. 11. hiwa ‘siskin, field sparrow’ In consideration of OJ pipa I suspect that the name of the bird is sound symbolic for the sound it produces. 12. hibari ‘skylark’ In consideration of OJ pi1bari (< pJ *pinpari) I suspect that the name of the bird is sound symbolic for the sound it produces. The fact that this bird name is phonologically close to the previous bird name supports its sound symbolic nature. 13. hoeru ‘howl, utter a cry, bark’ In consideration of OJ poye- with the same meaning, there is again some similarity in shape with the English glosses ‘howl’ and ‘bark’ which are also onomatopoetic in origin. Both shape and meaning of this word make me suspect that the phonological similarity with MK pulu- could be a matter of independent approximation of the same natural sound rather than a matter of inheritance from the same source. 14. hototogisu ‘little cuckoo’ It can be remarked that Japanese also has hotodori with the same meaning, so I suspect that the morpheme hoto stands for the sound produced by the bird. 15. hukuroo ‘owl’ In Old Japanese pukuropu alternates with pukuroku. This alternation along with the peculiar phonological shape and the unusual length of the word indicates that the word is mimetic for the sound produced by the bird. 16. huru ‘(rain, snow) fall, come down, come blowing’; hiru ‘break wind, evacuate, eject’, huku ‘blow, put forth’; hukureru ‘swell, expand, get big, be inflated’; OJ pukupukusi ‘lungs’; aoru ‘blow, fan, incite’, hareru ‘swell, become swollen’ In consideration of the initial labial stop p- that characterized these verbs in Old Japanese, the parallel with the English gloss ‘blow’ and the sound symbolic poeh that is iconic for a sigh is striking. Swadesh illustrates his objection to sound imitative entries in the basic vocabulary list, such as ‘blow’, ‘breathe’, ‘laugh’, ‘cry’ etc., with words for ‘blow’ that tend to have labial consonants along with rounded vowels. He mentions Spanish soplar which has developed from Latin sub-flare. 6 The Latin verb flare is of similar shape as English blow, Dutch blazen etc, but also as MK pwul- 'blow', Ma. fulgije, MMo. hüli’e < pMo *pülige@- and the Japanese verbs in question. This raises the suspicion that the phonological similarity of the verbs for ‘blow’ across different languages can be attributed to an independent 5 6

Hock 1991, 50; Campbell 1998, 320-21. Swadesh 1955, 126.

180

Chapter 7. General properties of language

approximation of the sound and not necessarily to a common history. Therefore, I eliminate the verbal entries huru (< pJ *pu-ra-), hiru (< pJ *pu-(C)i-), and huku (< pJ *pu(-)ku-), aoru ‘blow, fan, incite’ < OJ apur- (< pJ *a-pu-ra-) from the core evidence, along with the body part term OJ pukupukusi ‘lungs’ that seems to incorporate a reduplication of the naked root of huku. I further suspect that hukureru ‘swell, expand, get big, be inflated’ (< pJ *puku-ra-(C)i) shares the same root as huku ‘blow, put forth’ and must also be eliminated on ground of sound symbolism. Although hareru ‘swell, become swollen’ lacks the back vowel -u- that would make it suspect of being sound symbolic, the Tungusic and Mongolic words for ‘swelling’ to which the verb is compared, pTg *pul- and pMo *bul- seem to reflect expressive roots. 17. inanaku ‘neigh’ Although Starostin agrees with Martin’s internal analysis of this word as a prefixed manifestation of the verb naku ‘cry, creep, wail, moan’, he does not object the possible sound symbolic nature of the prefix ‘ina-’. The similarity with the English gloss ‘neigh’, Middle Dutch hinnen, Latin hinnire all with the same meaning, makes the parallel with pTg *iNi- ‘to neigh’, pMo *inc #a- ‘to neigh’, and pTk *ˆNˆ ra- suspicious. It is highly probable that the prefix in question is an onomatopoetic expression and therefore it should be eliminated from consideration. 18. kagayaku ‘shine, sparkle, gleam, twinkle’ It is generally agreed that kaga- is to be analyzed as a separate morpheme, but there is no consensus on the identity of the word. Whitman’s proposal to relate this morpheme to kage ‘shadow’ seems semantically odd at the first sight, since ‘shadow’ rather has the opposite meaning, being an area where the shining rays of light are cut off. However in Old Japanese kage2 meant not only ‘shadow’ but also ‘reflection, radiance, light’. The meaning ‘reflection, light’ is also found in kagami, OJ kagami 2 ‘mirror’ (