Proto-Japanese; Issues and Prospects (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Proto-Japanese AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E.F.K. KOERNER (Zentr...

Author: Bjarke Frellesvig | John Whitman

256 downloads 1408 Views 4MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Proto-Japanese

AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E.F.K. KOERNER (Zentrum für Allgemeine Sprachwissenschaft, Typologie und Universalienforschung, Berlin) Series IV – CURRENT ISSUES IN LINGUISTIC THEORY Advisory Editorial Board Lyle Campbell (Salt Lake City); Sheila Embleton (Toronto) Brian D. Joseph (Columbus, Ohio); John E. Joseph (Edinburgh) Manfred Krifka (Berlin); E. Wyn Roberts (Vancouver, B.C.) Joseph C. Salmons (Madison, Wis.); Hans-Jürgen Sasse (Köln)

Volume 294

Bjarke Frellesvig and John Whitman (eds.) Proto-Japanese. Issues and Prospects

Proto-Japanese Issues and Prospects

Edited by

Bjarke Frellesvig

University of Oxford & University of Oslo

John Whitman Cornell University

JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA

4-

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences — Permanence of Paper for Printed Library Materials, ANSI Z39.48-1984.

Library of Congress Cataloging-in-Publication Data Proto-Japanese : Issues and Prospects/ edited by Bjarke Frellesvig and John Whitman. p. cm. -- (Amsterdam studies in the theory and history of linguistic science. Series IV, Current issues in linguistic theory, ISSN 0304-0763 ; v. 294) Includes bibliographical references and index. 1. Japanese language--Morphophonemics. 2. Japanese language--Variation. 3. Ryukyuan language--Phonetics. I. Frellesvig, Bjarke. II. Whitman, John. PL558.9.P76 2008 495.6'15--dc22 2007049685 ISBN 978 90 272 4809 1 (Hb; alk. paper) © 2008 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. • P.O.Box 36224 • 1020 ME Amsterdam • The Netherlands John Benjamins North America • P.O.Box 27519 • Philadelphia PA 19118-0519 • USA

CONTENTS

Acknowledgements Introduction Bjarke Frellesvig & John Whitman Abbreviations

vii 1

10

Part I: Reconstructing the basic phoneme inventory 1. Evidence for seven vowels in proto-Japanese Bjarke Frellesvig & John Whitman

15

2. Early Japanese lexical strata and the allophones of /g/ J. Marshall Unger

43

Part II: Use of dialects in reconstruction 3. Proto-Japanese and the distribution of dialects Takuichiro Onishi

57

4. The uses of Ryukyuan in understanding Japanese language history Leon A. Serafim

79

Part III: Reconstructing accent 5. On the reconstruction of the proto-accentual system of Japanese Akiko Matsumori

103

6. A reconstruction of proto-Japanese accent for disyllabic nouns: focusing on the problem of subclasses Moriyo Shimabukuro

125

7. Proto-Japanese beyond the accent system Alexander Vovin

141

vi

CONTENTS

Part IV: Reconstructing morphology and syntax 8. The source of the bigrade conjugation and stem shape in pre-Old Japanese John Whitman

159

9. On reconstruction of proto-Japanese and pre-Old Japanese verb inflection Bjarke Frellesvig

175

10. The Nominal and Adnominal forms in Old Japanese: consequences for a reconstruction of pre-Old Japanese syntax Janick Wrona

193

Master list of references

217

Index

227

Acknowledgements The editors would like to acknowledge the help we received in the preparation of this book. The Faculty of Oriental Studies, University of Oxford, provided generous financial support for copy-editing and formatting. In the course of that, we were invaluably helped by the following Oxford students who worked hard on turning the disparate manuscripts into this book: Madeleine Brook, Patrick Lavelli, Tuukka Toivonen, James Webb and Asa Yoneda. Every time we created or developed IT related problems, Greg Jennings, Ross Wackett and Matt Brock of the IT Department of Hertford College, Oxford, came to the rescue and solved them. A congenial environment for manuscript editing was provided by the Hôtel du Manoir St-Denis in Montreal. We are very grateful to Wes Jacobsen for his detailed, perceptive, and helpful comments on the entire manuscript which helped us improve it greatly. Sven Osterkamp made important corrections and suggestions to several chapters at a late stage of the editing process. We would like to thank the series editor Professor E.F.K. Koerner for his interest in the project and suggestions on the book. Finally, we owe a big thanks to Anke de Looper of John Benjamins who gently guided us safely through the production process. Bjarke Frellesvig John Whitman Oxford — Ithaca January 2008

INTRODUCTION BJARKE FRELLESVIG & JOHN WHITMAN University of Oxford & Cornell University Proto-Japanese (pJ) is the reconstructed language stage from which all later varieties of Japanese descend.1 It has been studied both as an end in itself (as the genetic code of the Japanese language) and as part of endeavors to clarify the genetic affiliation of Japanese. Overall, two complementary methodologies have been employed in the reconstruction of pJ: internal reconstruction, especially applied to Old Japanese (OJ; 8th century), the oldest attested stage of Japanese; and, the comparative method applied to Japanese varieties, including those of the Ryukyus. Apart from the reconstruction of pJ suprasegmental phonology (usually, and also in this volume, loosely referred to as a whole as ‘accent’), most earlier work has focused on internal reconstruction and has mainly incorporated dialect comparison in passing. Samuel E. Martin’s The Japanese language through time (JLTT), published in 1987, which is a landmark in the study of the history and pre-history of Japanese, can be said to represent the culmination of earlier efforts, incorporating and synthesizing most previous research and presenting much original work. It gives the first comprehensive reconstruction of a large pJ lexicon and of pJ phonology, and also includes a reconstruction of the basic inflectional and derivational morphology. The segmental phonology and the morphology are reconstructed on the basis of internal reconstruction, but accent is reconstructed on the basis of dialect comparison. The only large scale attempt to base reconstruction of pJ primarily on dialect comparison is Shirô Hattori’s Nihon sogo ni tsuite (NSNT) published as a serial in 22 installments in the journal Gekkan Gengo in 1978-1979. It incorporates large amounts of Ryukyuan dialect material and brims with brilliant insights and proposals, but is unfortunately, in addition to being difficult to get hold of, also not easily approachable. Fundamentally different publications as they are, these two are the most important contributions in the 20th century to the study of pJ; it is befitting that they were produced by the two towering figures of Japanese linguistics of that century in the West and in Japan itself. 1 In this book we use ‘Japanese’ to cover all of Japanese, including Ryukyuan, whereas some scholars prefer ‘Japonic’, reserving ‘Japanese’ to refer to what we here call ‘mainland Japanese’. Consequently, ‘proto-Japanese’ is in this book the ancestral language of Ryukyuan and mainland Japanese.

2

FRELLESVIG & WHITMAN

In August 1999 the two editors of this book, together with Ross King of the University of British Columbia, organized a Workshop on Korean-Japanese Comparative Historical Linguistics held during the XIVth International Conference on Historical Linguistics. The papers presented at that meeting were diverse and wide ranging, addressing a number of topics related to the pre-history of Japanese and Korean, but in the course of the workshop it became clear to us that significant progress had been made in the reconstruction of pJ since the publication of JLTT, and using the energy and momentum created by that meeting, we decided to organize a more focused effort on the reconstruction of pJ, with a view to the production of this book. Based on the state of the field, especially as represented in JLTT, we singled out key areas in the reconstruction of pJ within which progress had been made or promised to be immediately forthcoming and invited contributions from scholars working on those topics: segmental phonology, use of dialect evidence, accent, and finally morpho-syntax. We used a meeting of the International Society for Historical Linguistics, the XVIth conference held in Copenhagen in August 2003, to bring the group of contributors together in a Workshop on proto-Japanese where the initial drafts of papers were read and discussed. All contributors subsequently revised their papers and submitted them as chapters to this book. While the primary goal of this book is to present new research which advances our understanding of pJ, we also hope to have produced a book which gives an overview over the state-of-the-art of the field and of its main issues, making the book useful for students and researchers alike. In this Introduction we first (§1) present some basics about OJ segmental phonology and its implications for internal reconstruction of pJ segmental phonology; second, we give an overview over the structure of the book and a presentation of the individual chapters (§2). 1. OJ and pJ segmental phonology 1.1 Onset consonants 1.1.1 OJ /p/. The consonant reflected in Modern Japanese (NJ) as /h/ in wordinitial position and /w, Ø/ in internal position–in Japanese usually referred to as ハ行の子音 ha-gyô no shi’in ‘the consonant of the h-column’–goes back to OJ (and pJ) /p/. OJ /p/ changed in internal position to /-w/ around 1000, merging with OJ /-w/, and was eventually lost before other vowels than /a/ (e.g., NJ kai ‘shellfish’ < OJ kapi, kawa ‘river’ < kapa). In initial position OJ /p/ changed to /f/ sometime in the Middle Japanese (MJ) period and then to /h/ in most positions in the transition between MJ and NJ (e.g., NJ hana ‘flower’ < OJ pana, hito ‘person’ < pito). The changes in initial position, which only involved a phonemic redefinition, but no merger or loss, and thus found no orthographic expression, are more difficult to date than the merger in internal position, and earlier it was thought that this consonant had shifted to /f/ (or a bilabial fricative /∏/) as early as, or even before, the OJ period.

INTRODUCTION

3

1.1.2 Secondary mediae (/b, d, g, z/). The OJ and MJ mediae, /b, d, g, z/, which in inherited vocabulary are not found in word-initial position, were prenasalized, i.e., pronounced with a nasal onset: [mb, nd, Ng, nz]. This reflects their secondary origin as contractions of sequences of nasal and tenues (/p, t, k, s/), sometimes with preceding loss of a vowel: OJ /b/ < pre-OJ */Np/ or */NVp/ (where ‘N’ is an unspecified nasal, not to be identified with the moraic nasal of MJ). The implications for pJ are obvious: OJ /b, d, g, z/ do not reflect single pJ segments and are not reconstructed as such. 1.1.3 Basic pJ consonant inventory. OJ had the following onset consonants: /p, t, k, s; b, d, g, z; m, n; r; w, y/. On the whole they map on to a basic pJ consonant phoneme inventory in a fairly straightforward way: OJ /p, t, k, s; m, n; r/ < pJ /*p, *t, *k, *s; *m, *n; *r/, and as mentioned OJ /b, d, g, z/ are diachronically secondary. Some issues remain, however, which are addressed by Unger in chapter 2. First, whereas it is agreed that OJ /w, y/ reflect two pJ segments, the question has not been settled of whether these simply were pJ /*w, *y/, or whether they in fact were /*b, *d/. A second issue, related to the first, is whether pJ had consonants which merged or were lost into OJ: It has thus been proposed that pJ in addition to the phonemes above had /*g, *z/, reflected as OJ /Ø, Ø~s/; Unger further argues that pJ had a velar nasal phoneme /*N/, reflected as OJ /g/. 1.2

Vowels and diphthongs Since the early 20th century when Shinkichi Hashimoto published his article “Kokugo kanazukai kenkyûshi-jô no ichi-hakken: Ishizuka Tatsumaro no Kanazukai no oku no yamamichi ni tsuite” (1917), it has become widely known that OJ had syllable distinctions which disappeared in the transition to MJ, the socalled kô-rui–otsu-rui ( 甲類・乙類 ‘A-type–B-type’) syllable distinctions. Corresponding to a number of MJ syllables with the vowels /i/, /e/, and /o/, OJ had two of each, so that where MJ for example, like NJ, had the k-initial (short) syllables /ka, ki, ku, ke, ko/, OJ had what may fairly neutrally be transcribed as /ka, ki1, ki2, ku, ke1, ke2, ko1, ko2/, with OJ /ki1, ki2/ merging as MJ /ki/, OJ /ke1, ke2/ > MJ /ke/, and OJ /ko1, ko2/ > MJ /ko/. /Ci1/ ≠ /Ci2/ and /Ce1/ ≠ /Ce2/ were distinct when the onset consonant was /p, k, b, g, m/, while /Co1/ ≠ /Co2/ were distinct when the consonant was /t, k, s, d, g, z, (m,) n, r, w, y/.2 It is usually agreed that the distinction resided in the part of the syllable following the onset consonant, but other than that the phonetic and phonemic reconstruction and definition of these syllables is still debated, and that is reflected in the various main transcription systems employed to note the distinctions, see Table 1.3 Some systems overdifferentiate phonemically in order to copy the orthographic 2 /mo1/ and /mo2/ are only distinct in the songs of the Kojiki (712), but merged as /mo/ in later OJ texts. 3 There are, needless to say, yet more systems of transcription which have been used. For example, Whitman (1985) uses -uy for -i2; Unger (1993 [1977]) is almost like Frellesvig & Whitman, but aligns neutral -e with -e1 (noting both -ye), rather than with -e2.


4

distinctions (e.g., the subscript indexing notation used above, the Yale system (JLTT), or modified Mathias-Miller (Mathias 1973, Miller 1986b:198)), whereas others take more explicit positions on the phonemic identity of the distinctions (e.g., Ohno 1990 or Frellesvig & Whitman 2004). In the interest of uniformity, we adopt in this volume a single system of transcription of OJ, here that of Frellesvig & Whitman (although it is not the first choice of all contributors). Examples, with equivalents in the other systems, are given in Table 2. Translation between the systems is entirely straightforward and they may, by anyone who so wishes, be regarded as notational variants (bearing in mind that the reader has to know how the ‘neutral’ syllables are aligned). Finally, Table 3 lists the 88 distinct syllables of OJ in the Frellesvig & Whitman transcription. Syllable type

Index notation

Ohno

A B neutral

i1 i2 i

i ï i

A B neutral

e1 e2 e

e ë e

A B neutral

o1 o2 o

Modified Mathias-Miller î ï i

Yale

Frellesvig & Whitman

yi iy i

i wi i

ê ë e

ye ey e

ye e e

o ö o

ô ö o

wo o o

wo o o

u

u

u

u

u

a

a

a

a

a

Table 1: Transcription systems for OJ

Gloss

NJ

‘fire’ ‘sun’ ‘blood’

hi hi chi

‘woman’ ‘eye’ ‘hand’

me me te

‘child’ ‘this’ ‘ear (of rice)’

ko ko ho

Frellesvig & Whitman pwi pi ti

Index notation pi2 pi1 ti

piy pyi ti

mye me te

me1 me2 te

mye mey te

mê më te

me më te

kwo ko po

ko1 ko2 po

kwo ko po

kô kö po

ko kö po

Table 2: Examples

Yale

Modified Mathias-Miller pï pî ti

Ohno pï pi ti

INTRODUCTION .i wi pi bi mi ki gi ti di si zi ni ri

pwi bwi mwi kwi gwi

pye bye mye kye gye

.e we pe be me ke ge te de se ze ne re ye

.a wa pa ba ma ka ga ta da sa za na ra ya

5 .o wo po bo mo ko go to do so zo no ro yo

.u

mwo kwo gwo two dwo swo zwo nwo rwo ywo

pu bu mu ku gu tu du su zu nu ru yu

Table 3: Distinct OJ syllables 1.3

Proto-Japanese vowels All hypotheses about the vowels of pJ recognize that a number of OJ vowels and/or diphthongs are diachronically secondary and derive from contractions of pre-OJ sequences of vowels. Until recently, the most widely accepted hypothesis about the pJ vowel system held that it consisted of only four vowels, /*i, *a, *u, *ə/, reflected as OJ /i, a, u, o/, and that all other OJ vowels and diphthongs arose from contractions of combinations of those four vowels, e.g., pJ *-ai > OJ -e (= -e2), *-ui > -wi (= -i2), or *ia > -ye (= -e1). This hypothesis was challenged in the 1970s by Hattori, who argued largely on the basis of Ryukyuan evidence for a six vowel system that included the mid vowels /*e, *o/. The chapters by Serafim and Frellesvig & Whitman address the reconstruction of the pJ vowels, arguing for more than four pJ vowels. 2.

Organization of the book The chapters in this book are organized into four parts around the main subject areas of the book, although some chapters contribute to more than one area. Part I, on segmental phonology, has chapters by Frellesvig & Whitman on vowels and by Unger on consonants. The chapters in part II present and incorporate recent findings from Japanese dialect studies, addressing the use of dialect evidence in reconstruction of pJ, with chapters by Onishi on the primary split within mainland varieties and Serafim on Ryukyuan. The three chapters in part III, by Matsumori, Shimabukuro, and Vovin, address the issue of the reconstruction of pJ accent classes, focusing in particular on the status of accent classes for bimoraic nouns, but arriving at rather different reconstructions of these classes in the proto-language. Finally, in part IV, Whitman, Frellesvig, and Wrona have chapters on the reconstruction of pJ and pre-OJ morpho-syntax.

6


Part I: Reconstructing the basic phoneme inventory In chapter 1, Evidence for seven vowels in proto-Japanese, Bjarke Frellesvig and John Whitman address the reconstruction of the pJ vowels. Until recently the main view held that only four primary vowels should be reconstructed for pJ: /*i, *a, *u, *ə/ (cf. §1.2 above), and more recently proposals of six primary pJ vowels have gained some currency: /*i, *e, *a, *o, *u, *ə/. Amplifying and strengthening the case for reconstructing two mid vowels, /*e, *o/, this chapter further argues for the reconstruction of an additional vowel, for a total of seven short primary vowels; the seventh vowel is here reconstructed as high central vowel /*ˆ/, giving the following basic vowel inventory of pJ: /*i, *e, *a, *o, *u, *ˆ, *ə/. The evidence offered in support of this proposal is both internal (alternations within OJ) and comparative (between Japanese dialects, and between Japanese and Korean). In chapter 2, Early Japanese lexical strata and the allophones of /g/, J. Marshall Unger contributes to the debate about the reconstruction of the consonant inventory of pJ, proposing that it had the following onset consonants: /*p, *t, *k, *s; *b, *d, *g, *z; *m, *n, *N/, of which /*b, *d, *g, *z/ lenited and are reflected in OJ as /w, y, Ø, Ø~s/. It is generally agreed that the OJ mediae, /b, d, g, z/, are secondary, reflecting contraction of some nasal segment with a following /*p, *t, *k, *s/, respectively (cf. §1.1 above). In this chapter Unger argues that OJ /g/ additionally directly reflects pJ /*N/ and thus in contrast with the other OJ mediae has two distinct sources. Unger argues that OJ had two main phonetic realizations of the phoneme /g/: [Ng] and [N], diachronically reflecting its two sources, /*NVk/ and /*N /, respectively, and further that this variation, originally diachronically motivated, continued through MJ and eventually formed the basis for the observable dialectal and other variation in the realization of /g/ in Japanese. Part II: Use of dialects in reconstruction Takuichiro Onishi constructs in chapter 3, Proto-Japanese and the distribution of dialects, an argument for a language-level divergence between eastern and western Japanese subsequent to the breakup of pJ. Onishi primarily examines grammatical and morphological features. He examines and rejects the hypothesis that the ancestor of eastern Japanese (EJ) descends directly from the proto-language and was originally distributed all over the main islands (Mase 1992); according to this ‘donut’ model, long popular in Japanese dialect geography, western Japanese (WJ) would be the product of radiating diffusion from an innovating prestige center in the Kansai region. Onishi’s model posits proto-Ryukyuan (pR) and proto-mainland Japanese (pMJ) as primary descendants of pJ. pMJ then bifurcates into EJ and WJ, demarcated along the natural boundary of the North and Central Japan Alps. Onishi’s argument is based on the absence of evidence for marginal EJ features in far western Japan, and asymmetries in the direction of diffusion: WJ features spread west from Kansai but tend not to cross the historical EJ-WJ boundary.

INTRODUCTION

7

In chapter 4, The uses of Ryukyuan in understanding Japanese language history, Leon Serafim provides an overview of the crucial contributions of Ryukyuan to our understanding of pJ. Since Hattori (1976b) it has been thought that the Ryukyuan parent language(s) and the ancestor of Western OJ (WOJ) diverged prior to the 8th century. Serafim reviews the evidence for this provided by comparison of the vowel systems of OJ and Ryukyuan (primarily Shuri), and comments on the relevance of the Ryukyuan vowel data for reconstruction of the pJ vowel inventory. Ryukyuan provides particularly strong evidence for the existence of a primary mid front vowel *e, and confirmation for the existence of a primary mid back *o with a distribution somewhat wider than what can be recovered from OJ evidence alone. Serafim also reviews two cases in which Ryukyuan data bear on the reconstruction of pJ morphosyntax. It is well known that the Adnominal and Conclusive verb endings in Ryukyuan derive from contraction of the Infinitive form of the verb and an auxiliary cognate with Shuri wu- ‘exist’ (Hattori 1977), and it has been proposed that the verb endings for these categories have a similar source in pJ. Serafim shows, contra Thorpe (1983), that the Ryukyuan developments are relatively recent and thus completely independent from whatever might have happened in pJ. In the realm of syntax, Serafim builds on recent work with Rumiko Shinzato to show that the kakari-musubi focus pattern attested in OJ and MJ as well as Ryukyuan must be reconstructed for pJ. Part III: Reconstructing accent In chapter 5, On the reconstruction of the proto-accentual system of Japanese, Akiko Matsumori reviews the central issues involved in the reconstruction of pJ accent and builds on the proposals in her earlier work, particularly Matsumori (1993). This work rejects the widespread assumption that the 11th century Kyoto system attested in the Ruiju myôgishô dictionary is identical or close to the proto-system for all Japanese varieties. Instead, Matsumoto reconstructs a pMJ system based on comparison between the Ruiju myôgishô system and other varieties which maintain a large number of distinct accent patterns, primarily Ibukijima and other dialects located in the Inland Sea area. Matsumori’s pMJ system has the same number of accentual classes for one, two, and three mora nouns as the system attested in the Ruiju myôgishô, but posits substantially different tonal realizations for these patterns; she shows how the Ruiju myôgishô, Ibukijima, and modern Tokyo-style patterns can be derived from her proto-system in a simple way. Matsumori then takes on the problem of reconstructing a proto-system that accommodates Ryukyuan as well. It has been known since Hattori (NSNT) that Ryukyuan accentual correspondences appear to require reconstructing eight distinct accent classes for bimoraic nouns (see Shimabukuro’s discussion of this issue in chapter 6). Matsumori considers the alternative possibility that pJ had only four classes for such nouns (cf. Tokugawa 1962); the additional distinction in pMJ and the complications introduced by the Ryukyuan correspondences may then result from splits conditioned by factors

8


other than pitch, such as vowel length or initial voiced segments (cf. Shimabukuro, and Vovin in chapter 7). Chapter 6, A reconstruction of proto-Japanese accent for disyllabic nouns: focusing on the problem of subclasses, by Moriyo Shimabukuro proposes a reconstruction of the accent of bimoraic nouns. In the Ruiju myôgishô, five accentual classes of bimoraic nouns are distinguished. It has traditionally been thought that this classification should be projected back to the pJ stage and thus five classes of bimoraic nouns have also been reconstructed for pJ. Hattori (NSNT) pointed to the existence of three accentual classes of disyllabic nouns in Ryukyuan which cut across the classes in the Ruiju myôgishô. In the present chapter Shimabukuro examines the Ryukyuan evidence in more detail. Using the comparative method, he argues that eight accentual classes of bimoraic nouns should be reconstructed for pJ and that these eight classes merged in different ways in Ryukyuan and in mainland varieties. He further argues that the phonological feature of initial register (or tone) should be reconstructed for the proto-language. In chapter 7, Proto-Japanese beyond the accent system, Alexander Vovin takes a close look at accent class 2.5 (bimoraic nouns with the unique [lowfalling] pitch contour found mainly in Kansai dialects). There has been some debate over whether or not this accent class should be reconstructed for pJ, or whether it constitutes an innovation exclusive to Kansai. In this chapter Vovin argues that this class should be reconstructed for pJ and presents detailed internal evidence that it largely does reflect loss of final pJ *-m, an idea originally offered by Polivanov (1924). Vovin goes on to examine in detail the further implications for the reconstruction of the ancestors of the accent classes of present-day dialects. He concludes that initial register was not a feature of the proto-language, but a secondary development with two different sources: vowel length (as earlier proposed by Hattori (NSNT) and Martin (JLLT)) and original initial voiced segments. Part IV: Reconstructing morphology and syntax Chapter 8, The source of the bigrade conjugation and stem shape in preOld Japanese, by John Whitman. Attempts to reconstruct the verb system of preOJ have typically reconstructed a shape directly predictable from the conjugation class (Ohno 1953) or an invariant open syllable stem shape (Unger 1993 [1977], JLLT). Yet an older approach focuses on the importance of hifukukei (embedded forms), which reveal a stem shape not directly observable in the synchronic OJ conjugations. This chapter focuses on correlations between the attestation of hifukukei and conjugation class membership (upper bigrade verbs typically attest hifukukei while lower bigrade verbs typically do not), between transitivity and conjugation class, and on the semantic functions of the bigrade endings. Whitman argues in detail, on the basis of both internal and cross-linguistic evidence, that the bigrade classes originate in the grammaticalization of an ancestral form of the verb e- ‘get’ .

INTRODUCTION

9

In chapter 9, On reconstruction of proto-Japanese and pre-Old Japanese verb inflection, Bjarke Frellesvig examines the relative age of the eight distinct OJ verb classes. After first reviewing the main different previous hypotheses about the origin of the bigrade verb classes (by Ohno (1953) and Unger (1993 [1977])) and proposing some minor amendments to Whitman’s new hypothesis (chapter 8), Frellesvig argues that while it has generally been acknowledged at least since Unger (1993 [1977]) that the bigrade verbs are derived and secondary, in the sense of incorporating originally derivational matter, the further implications of this have not previously been fully explored. He presents detailed evidence to show that the bigrade verbs are not just etymologically secondary, but belong to a different, younger, chronological layer within Japanese than the quadrigrade verbs. This has profound implications for reconstruction of pJ and pre-OJ verb inflection: The forms of the secondary and younger verb classes cannot be assumed to incorporate the same morphological material as the quadrigrade (and other primary) verbs, but must have arisen through analogy. Reconstruction of early inflected forms and verb paradigms should therefore be based exclusively on the inflected forms of the primary verb classes. In chapter 10, The Nominal and Adnominal forms in Old Japanese: consequences for a reconstruction of pre-Old Japanese syntax, Janick Wrona compares the distribution and functions of the Adnominal verb form and the Nominal form in -aku in OJ. While the Adnominal is the main strategy of complementation in MJ, in OJ its distribution is more restricted, and the Nominal form in -aku instead occurs more widely in direct complements of a wide variety of predicates. Wrona argues that the Adnominal form originated as a marker for clausal modifiers of nominal heads, and developed as a marker of complementation in cases where the nominal head was dropped. More generally, Wrona’s chapter shows that, by applying modern syntactic theory to the analysis of OJ, a reconstruction of important aspects of pJ syntax is not beyond our reach.

ABBREVIATIONS Languages CJ CLMC EJ EMC EMJ EMK EOJ J K LMC LMJ LMK LOC LOJ MC MJ MK NJ OC OJ pK pKJ pJ pMJ pNR pR pSR Sr SJ SK WJ WOJ

Central Japanese Chang’an Late Middle Chinese Eastern Japanese Early Middle Chinese Early Middle Japanese Early Middle Korean Eastern Old Japanese Japanese Korean Late Middle Chinese Late Middle Japanese Late Middle Korean Late Old Chinese Late Old Japanese Middle Chinese Middle Japanese Middle Korean Modern Japanese Old Chinese Old Japanese proto-Korean proto-Korean-Japanese proto-Japanese proto-Mainland Japanese proto-Northern Ryukyuan proto-Ryukyuan proto-Southern Ryukyuan Shuri Sino-Japanese Sino-Korean Western Japanese Western Old Japanese

ABBREVIATIONS

Sources KS MYS NS RM SM

Kojiki songs/poems Man’yôshû Nihon shoki songs/poems Ruiju myôgishô (Shoku-Nihongi) Senmyô

4VH 7VH ABC ABL ACC ADN ADV ALL CAs CAf CAB CAUS COMP CONC CONCES CONJ COP CP DAT DEC DEM DMM EMM EMPH EW pattern F FOC GEN GER GIS H HON HYP INF INTR

four vowel hypothesis seven vowel hypothesis loss of A::B opposition in favor of neutral C ablative accusative adnominal form (rentaikei) adversative allative coronal assibilation coronal affrication capability causative complementizer conclusive form (shûshikei) concessive conjectural copula complementizer phrase dative declarative demonstrative deviation metrical method east masking method emphatic east and west contrastive distribution pattern falling pitch focus genitive gerund geographical information system high tone honorific hypocoristic infinitive (ren’yôkei) intransitive

Others

11

ABBREVIATIONS

12 KAK K-Irr KM KP L LB LOC M MB MPST MP MVR NEG N-Irr NML NOM PASS PERF PNPM PP PPA PROG PROHIB PROV PSF Q QG QUOT REAL R-Irr S-Irr PST SoA STAT TBU TOP TR UB UM VAf VR1 VR2

knowledge and acquisition of knowledge k-irregular (kagyô henkaku) kakari-musubi kakari particle low tone lower bigrade (shimo nidan) locative mid tone middle bigrade (naka nidan) modal past tense (kyeri) medio-passive mid-vowel raising negative n-irregular (nagyô henkaku) nominal form nominative passive perfective palatal/non-palatal merger post-positional phrase progressive palatal assimilation progressive prohibitive provisional post-sibilant fronting question quadrigade (yodan) quotative realis (izenkei) r-irregular (ragyô henkaku) s-irregular (sagyô henkaku) simple past tense (ki) state of affairs stative tone-bearing unit topic transitive upper bigrade (kami nidan) upper monograde (kami ichidan) velar affrication first vowel raising second vowel raising

PART I RECONSTRUCTING THE BASIC PHONEME INVENTORY

CHAPTER 1 EVIDENCE FOR SEVEN VOWELS IN PROTO-JAPANESE BJARKE FRELLESVIG & JOHN WHITMAN University of Oxford & Cornell University 1.

Introduction In this chapter we present evidence to support the reconstruction of seven short vowels for proto-Japanese in (1) originally proposed and preliminarily presented in Frellesvig and Whitman (2004, forthcoming). 1 The seven vowel system for Japanese is reconstructed on Japanese-internal grounds (primarily internal reconstruction and dialect comparison), but it turns out to find further support from Japanese/Korean comparative evidence. (1)

*i *e

*ˆ *ə *a

*u *o

OJ had the following eight distinct vowels and postconsonantal diphthongs, shown in (2) in the notation used in this book. These are the eight entities which earlier were thought to be unitary vowel phonemes and referred to as the ‘eight vowels’ of OJ. Since Lange (1973) and Matsumoto (1974) a diphthongal interpretation has become current. (2)

i e

u o

-wi -ye

-wo

a

1 Vowel length has been reconstructed for pJ, based mainly on interpreting low pitch in EMJ as reflecting pJ long vowels, supplemented with Ryukyuan evidence in the form of what seem to be primary long vowels. Vovin (1993a) offers additional external evidence. The precise role of this feature in changes between pJ and OJ is far from clear. Vowel length has been proposed to have been a conditioning environment for certain sound changes; for example, vowel raising only applying to short vowels (Hayata 1998), or loss of *m and *r only taking place after (some) short vowels (Whitman 1985). The hypothesis we set forth here deals only with short vowels.


16

The correspondences between the pJ vowels we reconstruct and the vowels and postconsonantal diphthongs of OJ are shown in (3). (3) Seven vowel hypothesis (7VH) pJ *i *e *a *o *u *ˆ *ə

OJ i i, -ye a u, -wo u o o

Pre-OJ *ui *î *əi *ai

OJ wi wi e e

*iˆ *iə *ia *uˆ *uə *ua

-ye -ye -ye -wo -wo -wo

Until recently, a four vowel hypothesis (4VH) has been dominant, holding that pJ had only four primary vowels (e.g., Miller 1967, Matsumoto 1974, 1984, 1995, Ohno 1977, Whitman 1985), /*i, *a, *u, *ə/, reflected as OJ /i, a, u, o/. (4)

*i

*u *ə *a

The correspondences between the pJ four vowel system and OJ are given in (5). (5) Four vowel hypothesis (4VH) pJ *i *a *u *ə

OJ i a u o

EVIDENCE FOR SEVEN VOWELS IN PROTO-JAPANESE

Pre-OJ *ui *əi *ai *ia *iə *ua *uə

17

OJ wi wi e -ye -ye -wo -wo

Comparing (3) and (5), it will be seen that the 7VH offers a more differentiated and complex account of the origin of OJ -i, -u, -o, -e, -ye, -wo, -wi than does the 4VH. Thus, the 4VH has single sources for -i, -e, -o, -u and it regards all OJ -ye and -wo as secondary, resulting from contractions of pre-OJ vowel sequences. As opposed to this, we argue (§3) that some -i, -ye, -u, -wo reflect primary mid vowels, *e and *o, which in mainland varieties in most positions rose to merge with *i and *u, but in some positions gave -ye and -wo. We also argue (§4) that OJ -o had two sources, a mid and a high central vowel, *´ and *ˆ, which in the central dialect merged to give OJ o, outside of contraction with *-i (or -y) where the outcomes were *əi > -e and *î > -wi, respectively, thus also positing two sources for OJ -e (viz. *ai, *əi). 2.

-e, -ye, -wo, -wi The proposals we present in this chapter are mainly based on interpretations of the origin of OJ -e, -ye, -wo, -wi different from those of previous scholars. While -e, -ye, -wo, -wi, which are all regarded as secondary in the 4VH, are similar in being relatively infrequent and mostly occurring in morpheme- or root-final position, we argue that this is in fact coincidental and that their distributions have different sources: -wi and -e are truly secondary, but most -ye and -wo directly reflect primary vowels. Frequency and distribution A count of the text occurrence of the orthographically distinct syllable types in the Man’yôshû shows the following distribution (Ohno 1980:151ff), (6)2. Although a frequency count in running text gives no real picture of lexical distribution, these figures do give some indication that the main primary vowels were those in the syllables Ci(1), Ca, Co(2), Cu. Likewise, the fact that whereas simple morphemes of the structure CiCi, CaCa, CoCo, CuCu are all well represented in the lexicon, CwoCwo is very rare 3 and there are no simple morphemes of the type CyeCye, CeCe, or CwiCwi, shows that Cwo, Cye, Ce, Cwi are somehow not on a par with Ci(1), Ca, Co(2), Cu.

2.1

2

We use the subscript notation here to distinguish neutral from kô or otsu syllables. Mwomwo ‘hundred’ and mwomwo ‘thigh’ look like reduplications; early OJ mwokwo (~ later OJ mukwo) ‘partner, bridegroom’ may be a compound with kwo ‘child’.

3


18 (6)

Ci1 Ci Ci2

3,160 6,103 370

Ce1 Ce Ce2

686 2,299 853

Ca

12,120

Co1 Co Co2

1,030 3,631 5,280

Cu

6,415

In terms of lexical distribution, -wi, -e, -ye, -wo are similar in being relatively infrequent and mostly occurring in morpheme- or root-final position (cf. JLTT, Matsumoto 1995). In nonfinal position, -wi, -e, -ye, -wo are all rare. In simple forms, nonfinal -wi is found only in:4 pwiwe- ‘to scrape, slice thin’, kwisi ‘shore’, kwiri ‘fog’.5 Truly nonfinal -wo is rare, although a number of instances are in a morpheme or root final syllable and not in absolute word final position (cf. §§3.1, 3.3 below); JLTT:60-2 gives a list of OJ words with -wo. Also -e and -ye are infrequent in nonfinal position, although not exceptional; some cases may be positively identified as loanwords, 6 and others may in fact be loanwords (see further §5.3 below about loanwords). 2.2

Lexical contraction Common to any account of pre-OJ phonology is the observation that OJ -wi, -e, -wo, -ye in some words result from contraction of vowel sequences,

4 The word mwina ‘all’ is most likely a compound: mwi-na ‘body-person’. Sometimes kokwida ‘this much’ is also cited as an example of nonfinal -wi, but that is at least bimorphemic: kokwi-da, cf. the following OJ forms built on koko-/kokwi- ‘this much’: kokoda, kokwida, kokodaku, kokwidaku, kokoba, kokobaku, kokwibaku. koko-/kokwi- is itself analyzable at least into ko ‘proximal’ + -kV ‘degree, quantity’, cf. i-ku- ‘how much?’ and so-kV- ‘that much’, both with a number of further derived forms. 5 -wi in kwisi and kwiri (and possibly pwiwe-, see further below) may in fact be rare examples of sporadic ‘umlaut’ under influence of i in the following syllable, going back to *kusi/kɨsi (cf. MK k√z(√) < *k√s√ ‘bank, shore’) and *kuri (cf. Hachijojima kuri ‘fog’, MK kurum ‘cloud’). Other potential examples of sporadic umlaut include seki (~ saki) ‘barrier’, neri ‘paste’, semi ‘cicada’, seri ‘parsley’. In the following bigrade verbs, the bigrade theme (cf. §2.3 below) may have induced umlaut in the penultimate syllable (in addition to having been contracted with the root final vowel): nedi- ‘twist’ (?~ yodi-), negwi- ‘appease’ (~ nagu- ~ nagwo-), nige- ‘flee’(~ EMJ nogare-), pwiwe- (see Whitman 1985:133 for a different etymology), wosipe- ‘teach’ (~ NJ osowar- ‘learn’ < OJ *wosopar-). 6 E.g., kyesa ‘priest’s robe’ (袈裟, cf. LMC *kja: a:, EMC *kaɨ aɨ; Skt. kaṣāya ‘the (yellow) robe of Buddhist clergy’), pyera ‘spatula, pallet’ (cf. MK pyət ‘moldboard’), tera ‘temple’ (cf. MK tyəl ‘temple’, 刹 EMC *thaɨt, OC *tshraat/ksraat, Skt. kśetra ‘place’). EMC and LMC reconstructions follow Pulleyblank (1991); OC follows Miyake (1997 and p.c.).


19

e.g., (7),7 where juxtapositions with a vowel initial morpheme as the second part were univerbated. (7)

a. wi < u+i wi < o+i e < o+i e < a+i

wakwiratukwo < waku+iratukwo ‘young+honored male’ opwisi Proper name < opo+isi ‘big+stone’ toneri ‘palace servant’ < tono+iri ‘palace+enter’ taketi Proper name < taka+iti ‘high+market’

b. wo < u+o wo < u+a ye < i+a ye < i+o

sitwori ‘type of cloth’ ~ situ-ori ‘native.weaving-weave’ kazwope- ‘count’ < kazu+ape- ‘number+join’ sakyeri ‘is blooming’ < saki+ari ‘bloom+is’ pyeki Proper name < pi+oki ‘sun-put’

Although these contractions have, rightly, attracted much attention it must be noted that examples of lexical contractions of this sort are in fact quite few. The only solid cases of -wi and -wo are those listed in (7) and there is a small handful of examples with -e. Only -ye is found in a sizeable number of forms, which all are fairly transparent contractions from *i-a involving grammatical morphemes, including: -yer- Stative < *-i+ar- Infinitive+be;8 -kyeku Simple.Past Nominal < *-ki+aku Simple.Past+nominalizer; -kyeku Adjectival Nominal < *-ki+aku Adjectival.Adnominal+nominalizer; -kyem- Simple.Past Conjectural < *-ki+am- Simple.Past+Conjectural; -kyer- Modal.Past < *-ki+ar- Simple.Past+be/come+be; myes- see.Honorific < *mi-as- see-Honorific: kyes- wear.Honorific < *ki-as- wear-Honorific: kyesa this morning (cf. §3.4 below) < *ki-asa this-morning. 2.3

Apophony A number of nouns have alternating shapes with variation in the final syllable; these are often referred to as ‘apophonic’ stems. One variant, known as roshutsukei ‘exposed form’, occurs in word final position, while the other, hifukukei ‘embedded form’, usually occurs in compounds or derived forms. Traditionally the hifukukei is thought to represent the original shape of the root.

7

The notation in (7) is shorthand for saying, e.g., that wakwiratukwo is thought to reflect a contraction of two pre-OJ forms which are reflected as OJ waku and iratukwo, exemplifying OJ -wi originating in a contraction of two pre-OJ vowels which are reflected as OJ u and i. 8 This is the regular morphological Stative, exemplified in (7) by sakyeri; included here are also some lexicalized forms: kyer- ‘be wearing’ < *-ki-ar- ‘wear-be’, kyer- ‘have come’ < *-ki-ar‘come-be’, ser- ‘have done’ (< *syer-) < *si-ar-.


20 (8)

roshutsukei mwi kwi tukwi

∼ mu-kapari (‘substitute’) ‘hostage’ ∼ ku-gane (‘metal’) ‘gold’ ∼ tuku-ywo (‘night’) ‘moonlit night’

kwi

∼ ko-dati (‘stand’) ‘grove’

‘Hades’

yomwi

∼ yomo-tu-kuni (‘GEN-land’) ‘id.’

e∼o

‘back’

se

∼ so-muku ‘turn’

e∼a

‘eye’ ‘saké’

me sake

∼ ma-pye (‘side, direction’) ‘front’ ∼ saka-duki (‘cup’) ‘saké cup’

wi ∼ u

‘body’ ‘yellow’ ‘moon’

wi ∼ o ‘tree’

hifukukei

One view of these cases of nominal apophony is that the roshutsukei originates from the contraction of the original root-final vowel with a morpheme of the shape *i (Yoshitake 1930), as shown in (9a). Phonologically this conforms to the contractions noted in §2.2. The function and origin of this *i is subject to some debate, but one view relates it to the subject marker i vestigially attested in OJ. A further development of this derives the apophonic nouns from consonant final shapes, with the final consonant being lost before *i (Whitman 1985), see (9b). (9) ‘body’ ‘yellow’ ‘eye’

mwi kwi me

< *muy < *kuy < *may

(a) < *mu-i < *ku-i < *ma-i

(b) < *mum-i < *kur-i < *mar-i

Yet a third view (Murayama 1962) of the apophonic nouns reconstructs a specific final consonant, *-r, which weakened to a yod (yodization) in phrase final position, (10). (10)

‘yellow’ ‘eye’

kwi me

< *kuy < *may

< *kur < *mar

On both Whitman’s and Murayama’s views, the final consonant of the reconstructed forms is deleted in composition before another consonant. The evidence for the identity of the reconstructed final consonants is in all cases comparative, from Korean. (11)

‘hostage’ ‘gold’ ‘front’

mukapari kugane mapye

i, u, non-final position ‘full MVR’ (merger of pJ pJ *i vs. *e, and *u vs. *o, respectively). On our proposal, dialects/varieties differed in the domain (word, morpheme, root) with reference to which ‘final’ was defined. (14)

*e > -ye in final position (partial MVR) i elsewhere (full MVR) *o > -wo in final position (partial MVR) u elsewhere (full MVR)

It should be noted that the effect of partial MVR, because of the merger of Ce and Cye, and Co and Cwo, respectively, was neutralized in OJ for *e after coronals and glides, for *o after labials, and eventually in EMJ after all consonants, (15). (15)

pJ

pre-OJ / C = {p, b, m, k, g}

OJ Cye >

EMJ Ce

*Ce > *Cye > / C = {t, d, s, z, n, r, w (, y)} Ce > 9

Ce

Some cases of final -wo, -ye may go back to forms of the shape *CVCV, > *CVV through consonant loss and further > Cwo, Cye through contraction; but it is in fact difficult to come up with plausible candidates for such developments motivated by anything other than the presence of final -wo, -ye. Whitman (1985) compares OJ kwo with MK koma ‘child’, seemingly providing an external motivation for the reconstruction of the -wo as originating in contraction, but Lee (1991:246) points out that the latter is a loan from Mongolian.


pJ

23

pre-OJ OJ / C = {t, d, k, g, s, z, n, r, y} Cwo >

EMJ Co

/ C = {p, b, (m, ) w}

Co

*Co > *Cwo > Co >

Finally in this section we show how the hypothesis of MVR helps us understand the correspondences between the Japanese and Korean demonstratives (§3.4). 3.1

pJ *o > OJ u, -wo Internal evidence for raising of *o may be found in variation between -u and -wo within the OJ lexicon (16) and across dialects (17). See Hayata (1998), Matsumoto (1995:79ff, 132f) for -wo~-u forms, also Thorpe (1983). (16)

*yo(-)ri *mo(-)ko

(17) *yo *no

ywo(-)ri ‘from’ ~ yuri ‘behind; from’ mwokwo ‘partner, bridegroom’ ~ mukwo (> EMJ muko)

WOJ -wo ywo ‘night’ :: nwo ‘field’ (?~ nu(ma) ‘marsh’) ::

EOJ -u yu- (yutoko ‘night bed’) nu- (nu-no ‘field-GEN’)

Note that final position is implicated in the alternations above: in (16), ywori~yuri is usually analyzed as involving a morpheme boundary between ywo~yu ‘from’ and the ending -ri, while mwokwo > mukwo ‘partner, bridegroom’ is arguably a compound involving a morpheme *mwo + kwo ‘child, man’. In each case, *o > u can be associated with loss of transparency in the morpheme boundary. The facts are similar in (17); -wo appears finally as in ywo ‘night’ and nwo ‘field’, but u appears word-internally. The forms with u are attested in EOJ material, but what is at issue here is not dialectal correspondences but different responses to the boundary in different varieties.10 The crucial point in these examples is that *o is reflected as -wo in final position in one variety, but raised fully to -u before a morpheme boundary in another. While the pattern of EOJ u for WOJ -wo reflects varying interpretations of morpheme boundaries, or varying degrees of lexicalization, a second pattern of contrast between EOJ and WOJ may reflect more systematic variation in the environment for the raising of *o.11 EOJ -wo for WOJ u is attested throughout the EOJ region. In the westernmost part of the region, this is limited to a single 10

It is instructive here to note that the WOJ attestation of ‘night bed’ as ywodoko in Nihon Shoki (NS 47) has rendaku after the partially raised -wo, showing that the compound (or word) boundary was more transparent: ywo-doko. 11 Fukuda (1965:301) observes that the pattern of EOJ u corresponding to WOJ -wo is not found in the four easternmost kuni (Shimotsuke, Kozuke, Kazusa, and Mutsu); on the other hand EOJ -wo :: WOJ u, as in (18), is found across the EOJ region. The vast majority of examples of the latter type involve the verb ending -wo, in traditional terminology, the rentaikei or Adnominal ending in consonant base verbs. See, however, Frellesvig, this volume (fn. 17).


24

morpheme, the verb ending -wo corresponding to WOJ -u. Further to the east, *o is reflected as -wo before derivational suffix boundaries in some inflecting stems. (18) *-o *su/ogo-su *popo-mu

WOJ u -u finite verb ending sugusu ‘pass it’ pupumu ‘to bud’

:: :: ::

EOJ -wo -u ~ -wo sugwosu popomu

These facts suggest that at least some EOJ varieties—those most distant from the central region—had a more restricted environment for raising *o, although the exact nature of what counted as a ‘final’ environment for the relevant EOJ varieties—perhaps before transparent derivational suffixes—is difficult to recover from the surviving data. Examples like those in (18) allow us to recover pJ *o in contexts where in WOJ it raised and merged with *u. Comparative Ryukyuan data, here due to Thorpe (1983), reveals further examples of this sort. Ryukyuan in isolation provides no basis for distinguishing vowels corresponding to OJ -wo (= -o1) and -o (= -o2), and in most varieties pR *o raises and merges with *u. But in some Amami varieties (a subset of those which distinguish reflexes of *i and *e; cf. Thorpe 1983:32), o is preserved. The examples in (19) are cases where this vowel corresponds to WOJ u. (19)

WOJ -u pR *o *siro(-)si12 sirusi ‘mark’ :: *sirosi *mo(-)ko mwokwo ~ mukwo (see (16) above) :: *moko ‘son in law’ *tuko-yo tukuywo ‘moonlit night, moon’ ::*tukoyo ‘moon’ < *tuko ‘moon’ + *yo ‘night’

As with the EOJ data, pR gives no evidence for word nonfinal pJ *o except before possible morpheme boundaries (the argument that sirus- ‘write, mark down’ originates from a morphologically complex form *sirV-s is weaker than in the case of the other two compound forms, but is not to be rejected out of hand). This raises the possibility that raising of *o was already underway at the pre-OJ stage at which pR split off. 3.2

pJ *e > OJ i, -ye Comparative Ryukyuan-Japanese evidence gives the primary impetus for reconstructing an additional pJ front vowel, reconstructed by Hattori (1976b, NSNT) and Thorpe (1983) as pJ *e. This evidence involves cases where pR *e corresponds to OJ i. Hattori and Thorpe reconstruct pR *e for words where North 12

Note further that this word was borrowed into Ainu as sirosi, giving further support to the reconstruction of pJ *o. Whether Ainu borrowed the word from a variety of Japanese which did not have MVR, or, less likely, before MVR took place, is not important here.


25

Amami dialects give a central vowel ï corresponding to i elsewhere in the Ryukyus as well as OJ. Examples of this sort are listed in (20); 13 the pR reconstructions follow Thorpe (1983). See Serafim (this volume) for further examples and discussion of this correspondence. (20)

OJ i midu ‘water’ EMJ oyobi ‘finger’ piru ‘garlic’ kizu ‘wound’ idu- ‘wh-’ pidi ‘elbow’

:: :: :: :: :: ::

pR *e *medu *UyUbe *peru *kezu *edu *pedi

(*U represents pR *u or *o)

OJ internal evidence for *e may be found in variation between -ye and -i (21). (21)

*me *e *te

mye ‘woman’ ~ -mi- (womina ‘young woman’) ye ‘placenta’ ~ i- (iro ‘of same mother’) te ~ (-)ti(-) (miti-no-naga-te ‘long way’; (mi)ti ‘way’)

Examples of this sort are not as numerous as the examples we saw in (17) showing final *o reflected as -u word-internally, but the pattern is the same. Cross-dialectal correspondences within OJ show some of the variation parallel to the correspondences between pR and OJ. These include well known cases of WOJ -i :: EOJ -e (22), where the Eastern varieties pattern like pR. Thus, for example, the adjectival adnominal ending EOJ -ke (::WOJ -ki) may be considered to reflect a pre-raising stage (cf. also Hino 2003); note there is also a single example of WOJ -kye for -ki of this auxiliary in an early poem (K 32). (22) *ke

WOJ -ki adj.Adn.

::

EOJ -ke

There are also, however, examples of WOJ -ye :: EOJ -i . These examples highlight the fact that different dialects had different criteria for full and partial raising, as we saw in the case of *o > OJ u, -wo. 13 Whitman (1985:57), working within the 4VH, proposes instead that these correspondences reflect pJ *ə, and that an umlaut-like process raised and fronted *ə before *Cson(C)i, where Cson represents a sonorant. He suggests that in the same environment, *ə was fronted to *e in pR, where it fell together with secondary *e from the sources discussed below. pJ *ə > OJ o elsewhere and the chief argument for this hypothesis is a distributional gap: OJ o is rare before Cvoii. However, Miyake (2003) cites Serafim (1999) for a number of forms that counter-exemplify Whitman’s umlaut hypothesis and thus strengthen the Hattori/Thorpe argument for pJ *e. These include OJ sima ‘island’ :: pR *sema and OJ sita ‘tongue’ :: pR *seta (Serafim’s reconstructions). These reconstructions, too, if correct, indicate that some instances of pR *e :: OJ i must reflect a pJ vowel other than *i or *ə.


26 (23) *ipe *kaper-

WOJ ipye ‘house’ kapyer- ‘return’

:: ::

EOJ ipi kapir-

We propose that pJ *e raised in the same environments as *o, merging with *i in all positions except final, where it became OJ -ye. Thus pJ *e is reflected as WOJ -ye in *me > mye ‘woman’, but nonfinally this morpheme surfaces as -mi-, as in womina ‘young woman’ (wo- ‘small’, na ‘person’) or omina ‘old woman’ (oprefix, cf. also okina ‘old man’, probably related to oi- ‘age, get old’). 3.3

The environment for mid vowel raising Several proposals have been made about the environments and conditions for MVR. Serafim (1999) and Miyake (2003) advance the view that pJ *e, *o raised in pre-OJ and merged unconditionally with *i, *u respectively, giving OJ i, u, but on this view, all instances of OJ -ye and -wo are secondary, resulting from contraction, and as we noted above (§2.4) virtually none of these instances have clear etymologies involving contraction. We therefore adopt the view that the outcome of MVR was variable and conditioned, as proposed by Hattori (1976b, NSNT) (and Hayata 1998, 2000 for *o). Hattori and Hayata propose that length was the crucial conditioning factor for raising. Under this view, short *e and *o merged with *i, *u, but long *ee and *oo gave OJ -ye and -wo.14 There are a number of problems with this view. First, there seems to be no independent evidence for the long vowel in virtually all of the relevant cases. Second, there is no explanation for why we find no CyeCV < *CeeCV or CwoCV < *CooCV. Finally, the condition seems to imply that all word final vowels were lengthened; although allophonic lengthening is plausible for monomoraic words in phrasefinal position, as in modern Kansai dialects, it is difficult to find supporting evidence for lengthening of all final vowels. Without invoking vowel lengthening or long vowels it seems likely, and simpler to assume, that raising followed a course involving an intermediate phonetic diphthongization, eventually giving a high vowel, except in certain environments where the diphthongal realization was phonemicized as a diphthong, see (24), thus distinguishing between ‘partial’ (*o, *e > -wo, -ye) and ‘full’ (*o, *e > u, i) MVR as detailed in §3. (24)

/*o/ > [wo > u] > /-wo, u/ /*e/ > [je > i] > /-ye, i/

As outlined in §3 and as noted throughout §3.1 and §3.2, we propose that ‘final position’ is the core environment for partial MVR (i.e., the environment in diphthongal realization was phonemicized as a diphthong). As we saw in §§3.1-2, 14 This is simply tantamount to proposing that the first half of a long vowel raised (*e > i, *o > u, but *ee > ie, *oo > uo), although not described in that way by Hattori and Hayata.


27

there seems to have been variation between (and possibly also within) dialects with regard to the domains with reference to which ‘final’ was defined, especially for inflecting word classes. In WOJ ‘word final’ seems to have been the main environment for partial MVR, but for example, the well-known contrast between WOJ sugus- ‘pass it’ and EOJ sugwos- ‘id.’ indicates that in the latter dialects root- or stem-final, rather than word-final, position was the environment for *o > -wo. More generally, examples of non-final partial raising are found before fairly transparent boundaries, e.g., WOJ sukwo-si ‘a little’ (attested with -wo in the Shinsen jikyô 898-901) (~ sukuna- ‘few’) < *su/oko + *-si ‘adverbial formant’, or EOJ te-kwo/te-gwo ‘baby’ (~ tigo ‘id.’, attested in EMJ) < *te ‘breast, suckle’ + *ko ‘child’. Such forms exhibit partial MVR either because the word structure remained transparent until after MVR was complete and the variety in question had morpheme-, or lexical morpheme, final as the environment for partial MVR (e.g., *te-ko > *tye-kwo > te-kwo > tekwo), or because the derivative or compound itself was only formed after the completion of MVR (e.g., *te > *tye > te, then te+kwo => tekwo). It is finally possible that some dialects had unconditional full MVR, i.e., merger of *i and *e as i, and *u and *o as u, or at least very narrowly defined environments for partial MVR. Forms such as ipi and kapir- cited in (24) above may represent such dialects.15 3.4

Mid vowel raising and demonstratives. As is well known, the Japanese and Korean systems of demonstrative/ interrogative pronouns are structurally identical, with a three way distinction among the demonstratives (25). There is also a good form fit between MK kˆ, tyə, ə- ~ ənu and EMJ ko, so, i- ~ idu-. However, the semantics do not match in two of the sets: MK mesial kˆ corresponds to EMJ proximal ko, and MK distal tyə to EMJ mesial so. In this section we show how raising of pJ mid vowels offers support for a scenario of changes which has led to this situation. (25) MK EMJ

proximal i ko

mesial kˆ so

distal tyə ka

interrogative ə- ~ ənu i- ~ idu-

Previous researchers have pointed out that there is some evidence for an earlier additional demonstrative i-, retained as the OJ personal pronoun i (second person), and also in ima ‘now’ < i ‘this’ + ma ‘interval’. If i does reflect an earlier demonstrative, its function would be proximal. This raises the possibility that OJ i descends from an earlier proximal pronoun, which was displaced by the ancestor of OJ ko.

15

Other apparent examples of final full raising may have a different origin, namely as having been extracted from non-final position in compounds after completion of full raising. A likely example is ti ‘breasts’ which in OJ mainly occurs as the first member of a compound, but which from EMJ onwards is found as an independent noun; see §5.2.


28

There are internal and typological arguments for this view. First of all, it is less well known that the EMJ system is an innovation and that OJ in fact, as demonstrated in detail by Shirô Hashimoto (1966, 1982), only has a two term plus interrogative system : ko (speaker), so (non-speaker), i- ~ idu- (interrogative) (26). Most accounts posit a three term ‘proximal - mesial - distal’ system for OJ, built on ko - so - ka. However, there is no evidence within OJ of ka being a productive member of the demonstrative system. Two forms are attested in OJ: long kare is found once (MYS 18.4045); what may be taken to be short ka, as distinct from the adverb ka ‘this way’, is attested at most twice, both in dialect poems (MYS 14.3565, 20.4384). While these forms most likely represent the budding of the distal demonstrative which is so frequent in EMJ, they clearly did not form a productive integrated part of the OJ system of demonstratives. Other ka- based forms often cited are actually attested only from EMJ. Hashimoto’s study is the first to consider the OJ system on its own merits, rather than in terms of the EMJ system. He shows that the ko- versus so- system is entirely speaker based, with no primary reference to the hearer. ‘Speaker’, ko-, refers to what is within the speaker’s domain of direct sensory perception, or experience. ‘Non-speaker’, so-, refers to what is outside of the speaker’s domain of direct experience. (26)

OJ demonstratives speaker ko

nonspeaker so

interrogative i- ~ idu-

We believe on the basis of comparative Ryukuan evidence that the pJ shape of the interrogative was *e-, and we have mentioned above evidence for an earlier proximal demonstrative *i-. We propose that the OJ system evolved from the pJ system in (27). (27)

pJ

proximal *i

mesial *kˆ

distal *sˆ

interrogative *e-

pre-OJ(a)

proximal *i

mesial *kˆ

distal *sˆ

interrogative *i-

nonparticipant *sˆ

interrogative *i-

pre-OJ(b)

participant *kˆ

OJ

speaker ko

EMJ

proximal ko

nonspeaker so mesial so

distal ka

interrogative i- ~ iduinterrogative i- ~ idu-


29

What upset the pJ system, we suggest, was raising of pre-OJ *e, resulting in homonymy between pre-OJ proximal *i (< pJ *i) and interrogative *i (< pJ *e), a stage we represent as pre-OJ(a).16 We suggest that this ‘pernicious homonymy’ was resolved by proximal *i being discarded as a productive member of the system and mesial *kˆ and distal *sˆ being reanalyzed as speech event participant and nonparticipant respectively: pre-OJ(b). This was subsequently reanalyzed as speaker and nonspeaker, respectively, the system exhibited by OJ. The OJ system was later augmented to include a distal (ka), resulting in the EMJ and later system. The pJ demonstrative system we reconstruct is a good fit, phonologically and semantically, with the MK system (28): (28) MK pJ

proximal i *i

mesial kˆ *kˆ

distal tyə *sˆ

interrogative ə *e

Internal Japanese support for the reconstruction of OJ proximal ko- as reflecting an earlier mesial term is provided by a number of deictic OJ time words with initial k- (including both ko- and a variant ki-) (29), in three sets: (a) proximal, (b) ambiguous, and (c) non-proximal. The deictic difference between (a) and (c) and the ambiguity within (b) is easily understood on the hypothesis of a deictic shift in pre-OJ from mesial to proximal of the ancestor of OJ ko-, with the (b) and (c) forms reflecting the pre-shift stage. (29)

a. kyepu ‘today’, kyesa ‘this morning’; koyopi ‘tonight’ b. kozo ‘tonight; last night’17 c. kozo ‘last year’, kiso/kizo ‘last night’, kinopu ‘yesterday’

More generally, the shift of the mesial in a three term demonstrative system to replace the original proximal is attested for example in the development of the Latin demonstrative system. The Classical Latin three term system had proximal hic (haec, hoc) ‘this’ (close to speaker), mesial iste (ista, istud) ‘that’ (close to hearer), and distal ille (illa, illud) ‘that yonder’ (distant from speaker and hearer). In the transition to medieval Latin, the original mesial term iste came to replace proximal hic in its function (Väänänen 1963:128-9). 4.

Two sources for o: pJ *-ə, *-ˆ > OJ -o A main feature of the seven vowel hypothesis is that we propose that OJ o is the reflex of two pJ vowels, which we reconstruct as *ˆ and *ə, a high and a mid central vowel respectively. The basic observation underlying this proposal is 16 Note that interrogative *e-, like OJ i-, was not used in isolation, but always with following morphological material, and that *e- therefore raised fully to > i-. 17 This word is in Zdb translated as ‘tonight’, noting that it may also mean ‘last night’, whereas Ohno 1990 translates it simply as ‘last night’; these differences provide a good illustration of the ambiguity involved.


30

evident in (7a) and (8) above: OJ o takes part in two apophonic alternations, -o ~ -(w)i and -o ~ -e,18 reflecting two outcomes of contraction with *i or *y. Of these, -o ~ -e has received little serious attention, but has usually been thought to be an irregular and/or nonstandard alternative to -o ~ -(w)i. There is, however, little support for this view and in this section we argue on the contrary that -o ~ -e simply reflects a different source for OJ -o than that reflected in -o ~ -(w)i; and more generally that each of the alternations in (7a) and (8) above reflect a different pJ vowel, as shown in (30). Pre-OJ *-ˆ and *-ə merged after these contractions took place, eventually giving OJ -o. (30)

a. b. c. d.

*-u > *-ˆ > *-ə > *-a >

-u -o -o -a

~ ~ ~ ~

-(w)i -(w)i -e -e

< *-ui < *-î < *-əi < *-ai

Below, we review relevant cases from lexical contraction (§4.1), nominal apophony (§4.2), and verbal apophony (§4.3), of the two alternations o ~ e and o ~ wi in support of the 7VH. 4.1

Lexical contraction The examples in (31) seem to be the only solid cases of lexical contraction of an ancestor of OJ o with *i. They do not show -wi to be the regular and -e the irregular outcome: there is only one case of the former versus two of the latter. We interpret this to mean that OJ o had two pJ sources which gave different outcomes when contracted with *i, as shown in (32). On this proposal, pre-OJ *ˆ and *ə merged after these contractions had taken place and eventually gave OJ o, hence OJ opo-, tono, wo-. (31)

a. opwisi Proper name < opo+isi ‘big+stone’ b. toneri ‘palace servant’ < tono+iri ‘palace+enter’ wenu ‘puppy’ < wo+inu ‘small+dog’

(32)

a. *ˆpɨ- >

opo-

~

opwisi

< *ˆpˆ-isi

b. *tənə > *wə- >

tono wo-

~ ~

toneri wenu

< *tənə-iri < *wə-inu

4.2

Nominal Apophony As mentioned, apophonic alternations such as those in (8) are thought to originate in contractions of an older shape with *-i or *-y (possibly going back 18

The notation -(w)i reflects that -wi is neutralized as -i after coronals (t, d, s, z, n, r).


31

further to a consonant). OJ monosyllabic apophonic nominal stems are listed in (33). (33)

a. Cu ~ C(w)i ku- ‘fortress’ ~ kwi mu- ‘body’ ~ mwi nu ‘bead’ ~ ni tu ‘miscantus reed’ ~ ti ‘chigaya’ turi ‘hook’ ~ ti b. Co ~ C(w)i ko- ‘tree’ ~ kwi no- ‘load’ ~ ni po- ‘fire’ ~ pwi so nonspeaker demonstrative ~ si 3rd sg. personal pronoun c. Co ~ Ce19 mo ‘algae; ?seaweed’ ~ me ‘(edible) seaweed’ moyasi- ‘bud, sprout forth’ ~ me ‘bud’ so- ‘back’ ~ se yo ‘branch’ ~ ye yo- ‘good’ ~ yed. Ca ~ Ce ka(-) ‘hair’ ~ ke ka(-) ‘day(s)’ ~ ke ma- ‘eye’ ~ me sa- ‘narrow’20 ~ seta- ‘hand’ ~ te

Not all these cases of apophony involve variation between a compound and a free variant, but it is clear that also here OJ o takes part in two alternations: -o- ~ -(w)i and -o- ~ -e. Again, the distribution is not persuasive for -o ~ -(w)i being regular and -o ~ -e being irregular. Instead, we interpret this to support our proposal that OJ o has two sources, *ˆ and *ə, and that the forms above should be reconstructed and understood as (34), with *ˆ and *ə merging (eventually to give OJ o) at some point after the contractions had taken place.

19

As mo and moyasi- are not attested in the Kojiki it is not known whether they originally had mwo (mo1) or mo (mo2). We include them here in the absence of evidence that they should have had mwo. In fact, moyasi is not attested until the early 10th century in the Honzô wamyô, but is still relevant here. 20 Thanks to Sasha Vovin for pointing out to us that sa- is not phonographically attested in OJ.


32 (34)

a. ‘tree’ ‘load’ ‘fire’ ‘dem.’

*kˆ- > *nˆ- > *pˆ- > *sˆ- >

konoposo-

b. ‘seaweed ‘sprout’21 ‘back’ ‘branch’ ‘good’

*mə-> mo *mə > mo(yasi) *sə- > so *yə > yo *yə- > yo-

~ ~ ~ ~

kwi ni pwi si

< *kî < *nî < *pî < *sî

~ ~ ~ ~ ~

me me se ye ye-

< *məi < *məi < *səi < *yəi < *yəi

The picture is different if we look at polysyllabic stems: Among polysyllabic stems we only find the three alternations OJ CVCu- ~ CVC(w)i, CVCo- ~ CVC(w)i and CVCa- ~ CVCe; but not CVCo- ~ CVCe (the alternation which supports our proposal of two pJ sources for OJ o). (35)

a. CVCu- ~ CVC(w)i kamu- ~ kamwi ‘god’ kuku- ~ kukwi ‘stem’ kuru- ~ kuri ‘chestnut’ kutu- ~ kuti ‘mouth’ satu- ~ sati ‘hunting device (or its spiritual power)’ tuku- ~ tukwi ‘moon’ tuku- ~ tukwi ‘zelkova (tree)’ b. CVCo- ~ CVC(w)i yomo- ~ yomwi ‘Hades’ woto- ~ woti ‘distant place’ c. CVCa- ~ CVCe ama- ~ ame ‘rain’ ama- ~ ame ‘sweet’ kana- ~ kane ‘metal’ kaza- ~ kaze ‘wind’ muna- (not phonographically attested in OJ) ~ mune ‘ridge’ muna- ~ mune ‘breast’ mura- ~ mure (not as noun in OJ; *kəwye/kˆwye > kowe), but it is more likely that it goes back to a consonant final form: *kəwər > (yodization, cf. §2.3 above) *kəwəy > (contraction) *kəwe > kowe; *kəwər would then also, with an epenthetic echo vowel conforming to phonotactic surface constraints, be reflected in the reduplicated mimetic OJ koworo-koworo ‘rolling, rumbling noise’. 23 Note that kome- has an upper bigrade variant which is attested only once, in the compound tuma-gomwi- ‘spouse-hiding, spouse-enclosing’ (Kojiki, KS 1). Also, the relationship of the verb yor- ‘approach’ to yosor-, yose- and yos- ‘bring close’ is not clear.


34 (37)

‘hide’ ‘close’ ‘do for me’ ‘do’ ‘copula’

*kəmə-r- > *yəsə-r- > *kəsə- > *sə > *tə >

komoryosor-koso so to

~ ~ ~ ~ ~

komeyose-kosesete

< *kəmə-Y < *yəsə-Y < *kəsə-Y < *sə-Y < *tə-Y

4.4

Summary In this subsection we have presented evidence internal to OJ for two pJ sources of OJ o. Overall this evidence consists in the existence of alternations between OJ o ~ e in addition to the more widely recognized u ~ (w)i, o ~ (w)i and a ~ e. Although o ~ e has a more limited distribution, especially numerically, than the other three, it is found in the same three general contexts: lexical contraction (§4.1), nominal apophony (§4.2), and verbal apophony (§4.3). We thus suggest that these four alternations within OJ reflect contractions of four different pJ vowels with a following *i, as in (30) above, repeated as (38). (38)

*u > *ˆ > *ə > *a >

u o o a

~ ~ ~ ~

(w)i (w)i e e

< *ui < *î < *əi < *ai

5.

Supporting comparative evidence The seven vowel system we have reconstructed for pJ has the same number of vowels as the system of MK. (39) shows the vowels of LMK (15th century), with transcription following Lee (1972a) and with hangǔl and Yale Romanization (following Martin 1992) in brackets. (39)

i (이 i)

ˆ ə  a

(으 u) (어 e) ( o) (아 a)

u (우 wu) o (오 wo)

In this section we point out that many of the vowel correspondences proposed for Japanese and Korean in previous literature can be explained rather straightforwardly by the 7 vowel system we reconstruct for pJ. OJ o has been noted to correspond both to MK ɨ and ʌ. This is explained by our hypothesis that OJ o results from the merger of pJ *ɨ and *ə. MK o has been noted to correspond to OJ u and (more rarely) wo. This is explained by the hypothesis of MVR. Matters are slightly more complex in the case of our reconstructed pJ mid front vowel *e, but here too we show that the observed correspondences support the hypothesis of Japanese MVR. We propose the basic vowel correspondences in (40) (see also Frellesvig and Whitman, in press):


(40)

a. b. c. d. e. f. g. h.

MK i ˆ u yə < *e ə o  a

:: :: :: :: :: :: :: ::

OJ i o u i, -ye a, o u, -wo o a

< < < < < < <
OJ o 3. Mid-vowel raising a. pJ *e > OJ i, -ye b. pJ *o > OJ u, -wo As for the dating and ordering of these changes, we are at the moment not in a position to offer a full account, but can say the following: Within (49.1), sets (a), contraction of falling diphthongs, and (b), contraction of rising diphthongs, are very different and should be divorced temporally; as we showed above (§2.2), the (a) set is pervasive, whereas (b) is really very marginal, with only *ia > -ye being found in more than a single example. It seems likely that the (b) set arose as singular or rare phonemicizations, coincidental to MVR when, as a result of MVR, the relevant phonemic strings (/-ye, -wo/) became available in the system. We would therefore order (49.1.b.i) after (49.3.a) and (49.1.b.ii) after (49.3.b). As for (49.2), the merger of pJ *ɨ and *ə took place after (1.a); equally obviously, the backing of the result of the merger, here tentatively *ə, to give OJ o must have taken place after (49.3.b) (or at least after the first moment of that change: *o > [wo]). It is likely that the two parts of MVR were not synchronous: In WOJ a number of early forms such as mwokwo for mukwo or pom- ‘step’ for pum- suggest that raising of *o was not entirely complete at the beginning of the 8th century. Raising of *e in WOJ, on the other hand, seems to have been completed earlier, and singular forms such as -kye for -ki may be regarded as archaic literary fossils, reflecting the pre-raising stage. As mentioned above, words such as tera or potoke were probably borrowed shortly after the introduction of Buddhism into Japan, indicating that raising of *e was complete by the first half of the 6th century. If we are right about the dating of raising of *o, the type (49.1.a) contractions must have taken place before that (and should, incidentally, include *oi > -wi, cf. iswo ‘rock’ ~ isi ‘stone’), but it is not possible, strictly speaking, to determine the ordering of contraction and raising of *e. Ryukyuan, on the other hand, suggests a contrasting chronology, with pR *e retained later and in a larger range of environments than pR *o < pJ *o, and pJ *ˆ and *ə merged prior to the pR stage, accounting for differences in the formation of bigrade verbs in the two varieties. See also Serafim’s chapter in this volume on these issues. The point, however, is the same: raising of the front and back mid vowels did not necessarily occur synchronously.

40


The following more detailed ordering for OJ of the component parts of the changes above is consistent with the facts. Note that (50.4) and (50.7) have the intermediary stages *ya and *wa; here the full change is lumped together in each case, but it is very likely that the first stages, *ia > *ya and *ua > *wa took place somewhat earlier than the second and belong earlier in the chronology, especially *ia > *ya which is found in a number of grammatical morphemes. (50)

1. Contraction of falling diphthongs Pre-OJ OJ *ui wi *î wi *əi e *ai e 2. Merger of central vowels: pJ {*ˆ, *ə} > *ə 3. Mid vowel raising: pJ *e > OJ i, -ye 4. Contraction of rising diphthongs pre-OJ *ia > *ya > -ye pre-OJ *iə > -ye 5. Mid vowel raising pJ *o > OJ u, -wo 6. Backing of central vowel pre-OJ *ə > OJ o. 7. Contraction of rising diphthongs pre-OJ *ua > *wa > -wo pre-OJ *uo > -wo

6.1

Arisaka’s Law and the 7VH. The well known ‘Arisaka’s Law’ is the observation that in OJ the syllable Co (= Co2) did not occur in a root morpheme with Ca, Cwo, Cu. That is to say, there were (almost) no apparently simple words of the structure CoCa, CoCwo, CoCu or CaCo, CwoCo, CuCo. This has been taken to indicate that Japanese earlier had some form of ‘vowel harmony’ (in the form of restrictions on the shape of root morphemes rather than the kind of vowel harmony which applies to the concatenation of stem and affix in ‘vowel harmony languages’ such as Turkish,). However, Arisaka’s Law makes little sense under the 4VH: first, four


41

vowels seem few for a harmony system of any kind; (51) is the harmony system which would have held if both the 4VH and Arisaka’s Law were correct. Second, on the 4VH, OJ -wo is always a secondary vowel, < *-u+a or *-u+ə; there is thus no scope for OJ -wo to take part in Arisaka’s law, and furthermore where (-)Cwo(-) was etymologized as < *(-)Cuə(-) < *(-)CuCə(-) (rather than *(-)CuCa(-)), the source string itself violated Arisaka’s Law. Third, on the 4VH, -e (= -e2) is a secondary vowel, deriving exclusively from *-a-i; thus the majority of OJ CoCeforms are etymologized as < *CəCa(-)(C)i, with the source string violating Arisaka’s Law. On the seven vowel hypothesis, on the other hand, Arisaka’s Law makes far better sense. It provides better balanced sets of vowels, with the ‘central’ and ‘back’ sets in opposition (52), and direct OJ reflexes of the pJ vowels which conform to the observed cooccurrence relations (53). (51)

neutral *i

central

back *u

*ə *a (52)

neutral *i *e

central *ˆ *ə

back *u *o *a

(53)

neutral -i ( [ŋ] are easy to imagine, but [ŋ] > [ŋg] (partial denasalization of a nasal) is by no means unknown in languages of the world.1 Likewise, the use of ng in Portuguese documents of the 16th century for MJ /g/ might seem at first to imply that what the Portuguese heard was [ŋg], but since Portuguese lacks a velar nasal, Kamei Takashi maintained that the ng indicated [ŋ] and perhaps also the nasalization of the preceding vowel, a theory with which Kiyose concurs (Kiyose 1991:184, 195). While the Portuguese evidence casts doubt on /g/ being realized as a simple [g] even as recently as Late Middle Japanese (LMJ), it leaves us asking, was /g/ realized as [ŋ], [ŋg], or both under different conditions?

1

German Finger [fiŋə(r)], clearly derived from fangen ‘catch’ ~ fing ‘caught’, shows that English [fiŋgər] has picked up its [g] (cf. sing ~ singer).

EARLY JAPANESE LEXICAL STRATA AND THE ALLOPHONES OF /g/

Map 1: Distribution of allophones of /g/ (from Hirayama, 1960)

45

46

J. MARSHALL UNGER

There are three kinds of fairly systematic OJ textual evidence that bear on this problem. Each is related to a writing practice involving ongana, i.e., Chinese characters used transcriptively for OJ syllables based on one of their SJ readings (on) or something similar. Ongana corresponding to Middle Chinese (MC) syllables ending in stops *-p, *-t, *-k or nasals *-m, *-n were sometimes used to transcribe OJ sequences of the form CVpV, CVtV, CVkV, CVmV, or CVnV, as in the names 博多 Hakata and 難波 Naniwa. Such phonograms may be called bimoraic ongana. In the case of MC *-ŋ, the OJ sequences are of the form CVgV, as in the names 香山 Kaguyama and 相模 Sagami (Miyake 1999:456-58, Kiyose 1991:167). Note also suguroku ‘backgammon’ 双六, a transparent borrowing (cf. Mandarin shuāngliù) with /gu/ for MC *-ŋ. In other instances, ongana corresponding to MC syllables ending in *-p, *-t, *-k, *-m, or *-n, represent only an open syllable in the middle of an OJ word. When this happens, and the next syllable is represented by an ongana, the corresponding MC syllable typically begins with *p-, *t-, *k-, *m-, or *n-, respectively. E.g., 甲斐 Kai, 印南 Inami. The first ongana in such cases may be said to take truncated transcriptive values. In the case of MC *-ŋ, the next syllable shows Early Middle Chinese (EMC) *g-, *-, or *ŋ-, i.e., the initials used nondistinctively in early texts like Kojiki to transcribe OJ g- (Kiyose 1991:190-91). Finally, ongana corresponding to MC syllables ending in *-n in the middle of OJ words sometimes stand for OJ CVdV or CVrV. In the names 但馬 Tajima and 播磨 Harima, the next ongana corresponds to an MC syllable beginning with *m- or *n-, but it may correspond to a syllable beginning with EMC *g-, *-, or *ŋ-, as in the name 駿河 Suruga (Kiyose 1991:192-93), or some other segment. Note also the occasional 万 or 満 for OJ -maro in names (Sven Osterkamp, p.c.). We may call phonograms like 但 and 播 nasal-modified ongana. The way in which velars fit into the pattern of labial and dental bimoraic, truncated, and nasal-modified ongana is strong evidence that OJ /g/ had a nasal component, but does not rule out a stop component as well. Indeed, the fact that MC *-n corresponds to OJ /d/ = [ⁿd] in the name Tajima 但馬 could be construed as showing that, in the case of, say, Sagami 相模, MC *-ŋ corresponds to [ŋg] rather than [ŋ]. However, Miyake notes that the name Soga (OJ Swoga) was written in Kojiki as 宗賀 EMC *tsawŋ ah, 蘇我 EMC *so ŋaq, and 蘇賀 EMC *so ah, to which Kiyose adds a 7th-century instance 宗我 EMC *tsawŋ ŋaq. Keeping in mind that EMC *g-, *-, and *ŋ- all correspond to OJ /g/ in early documents, it is noteworthy that they also correspond to OJ /k/ 190 times, 5 times, and once, respectively, in Kojiki poetry, which contains 730 k-initial syllables (Miyake 1999:447 [Table 7.29]). That is, ongana with EMC *g- did double duty for OJ /k/ and /g/ whereas those with EMC *ŋ- (but for a single instance) did not. Kiyose (1991:188-89) also remarks on this point. All this seems to tip the balance in favor of /g/ = [ŋ]. Nevertheless, Miyake (1999:455-61) rejects [ŋ] as an allophone of OJ /g/ for two reasons unrelated to the foregoing discussion. First, whereas OJ /m, n/ are


47

transcribed exclusively with Late Old Chinese (LOC) or EMC nasal phonemes, OJ /g/ is transcribed with both LOC and EMC nasal and non-nasal phonemes. Second, go’on syllables corresponding to LOC and EMC syllables ending in *-ŋ are reflected as final /Vu/ or /Vi/, not /Vgu/ or /Vgi/, at the time the first kana transcriptions of SJ readings appear. Are these facts compelling? Miyake’s first observation is flawed because his own account of how ongana were used does not preclude the possibility that OJ /g/ may have had a nasal allophone [ŋ] (1999:77 [Table 2.7], 80 [Table 2.9], 403-405, 421-24, 45556). Miyake notes that the most common EMC initials used to transcribe OJ voiced obstruents and nasals in Kojiki were the following: EMC *b *d *j *g *g / * < LOC *g *ŋ *m *n / *

OJ b d z g g g m n

(The non-nasal EMC phonemes above are called ‘muddy’ in Chinese terminology.) On the other hand, the ongana orthography of Nihon shoki was based on the the LMC dialect of Chang’an, which had undergone some important sound changes. These are reflected in the next table, which shows the most common Chang’an LMC (CLMC) initials used to transcribe OJ voiced obstruents and nasals in the Japanese poetic passages of Nihon shoki: EMC *b *d *z *g *g / * *m *n * *ŋ

CLMC *p *t *s *k *x *mb *nd *j *ŋg

OJ p, b t, d s, z k, g k b, m d, n z, n g

Notice that there were two ways to transcribe OJ /b, d, g, z/ in the foregoing system. Of these alternatives, Miyake says, Shoki scribes preferred to use CLMC prenasalized voiced obstruents … for OJ prenasalized voiced obstruents … They also preferred to use CLMC voiceless

48

J. MARSHALL UNGER obstruents … for OJ voiceless obstruents … Only 16.7% (348/2079) of OJ voiceless obstruents … and only 20.6% (89/432) of OJ prenasalized voiced obstruents … in Shoki poetry are written with CLMC ‘muddy’ initials. Still, ‘muddy’ initials were by no means rare in phonograms for OJ voiceless and voiced obstruents, with the exception of OJ g: only 1.7% (3/176) of OJ syllables with initial g- are written with CLMC ‘muddy’ initial phonograms. (Miyake 1999:82)

This is the real reason that Miyake does not want to acknowledge [ŋ] as an allophone of OJ /g/. He believes that the strong preference for transcribing OJ /g/ with prenasalized CLMC obstruents rather than CLMC ‘muddy’ initials means that the new CLMC prenasalized voiced velar obstruent (= partially denasalized velar nasal) must have been a much better phonetic match for OJ /g/ than EMC *ŋ had been. However, only 8.0% (51/638) or fewer of the OJ /m/ in Nihon shoki poetry were transcribed with CLMC [m] (Miyake 1999:384, 403-404), yet Miyake rejects the idea that CLMC [mb] was greatly superior to EMC *m for transcribing OJ /m/.2 More than 90% of the /g/ were transcribed with a nasal and more than 90% of the /m/ were not: if we can be sure that /m/ was a nasal despite these facts, we can hardly say that /g/ could not have been nasal because of them. Turning to Miyake’s second observation, the truth is that he does not really believe that MC *-ŋ was reflected as zero in go’on. He hypothesizes “that OJ had a short-lived phoneme *ŋ only in SJ loans which was later lost: 香 EMC *həŋ > OJ Go-on *kaŋu > post-OJ Go-on kau” (1999:460). He does not want to treat this transient phoneme as an allophone of /g/ because “if one claims that MC *ŋ was borrowed as OJ g, then one must explain why the usual reflexes of OJ g are zero in SJ forms and g in native words”. I think Miyake is correct to see *kaŋu in the past of kau—indeed, in view of the truncated and modified-nasal ongana mentioned above, such a reconstruction is hard to avoid—but in view of the foregoing evidence of an [ŋ] allophone of OJ and pre-OJ /g/, the notion of a ‘transient’ phoneme strikes me as ad hoc. Why must OJ *ŋ have been short-lived and restricted to SJ vocabulary? If early stages of Japanese had some surface velar nasals in native words, it would be much easier to understand why EMC *həŋ was borrowed as OJ *kaŋu rather than as, say, kau from the beginning. Remember too that we must eventually explain all the allophones of /g/ we find in later Japanese dialects. 3.

Lenition of pre-OJ voiced obstruents As the last quotation shows, Miyake thinks that, unless there was a transient velar nasal phoneme, any pre-OJ [ŋ] would necessarily have to be allophones of /g/. This is related to Miyake’s objections to the theory, proposed by Ramsey & Unger (1972), that OJ w and y derive from pJ plain voiced *b and *d, 2

As many as 40.6% (252/620) of OJ /n/ in Nihon shoki poetry were transcribed with CLMC [n] (Miyake 1999:423 [Table 7.15]). Clearly, /m/ patterns more like /g/ than like /n/, yet Miyake treats /m/ and /n/ alike.


49

and that there were also pJ plain voiced obstruents *g, which went to zero, and *z, which went to s or zero (adopting here JLTT’s amendment to the theory). This theory is supported by some comparative evidence for pJ *b, *d,3 and pKJ *g;4 internal evidence of s ~ Ø alternations; and certain facts about dakuon obstruents suggesting that none was original in the proto-language.5 It can also explain the distribution of voiced initials in SJ syllables provided that the lenitions occurred after the establishment of go’on and before the establishment of kan’on. Some go’on and some kan’on words begin with a dakuon or /r/—native OJ words do not—but go’on voiced obstruents, which reflect phonemically voiced Chinese initials, match voiceless obstruents in the corresponding kan’on syllables. In kan’on, dakuon reflect Chinese nasals in syllables without velar nasal codas, which match nasals in the corresponding go’on syllables. Now if Japanese has always had the same two series of obstruents, then why aren’t dakuon distributed in go’on and kan’on in the same way? According to Ramsey & Unger the answer is that go’on were established when Japanese still had plain voiced obstruents. Lenition did not affect plain voiced obstruent initials in go’on words because of their special lexical status. After lenition, only prenasalized obstruents remained, forcing the use of seion initials in kan’on except when Chinese of the period coincidentally had prenasalized initial obstruents, for which prenasalized dakuon were better matches than seion. Miyake allows that OJ w and y arose from pJ *b and *d, but doubts the existence of pJ *g and *z, and insists that the lenition of *b and *d took place before the establishment of go’on, not after. His alternative answer is that, for instance, medial /b/ (more abstractly /np/) = [mb] was pressed into service to transcribe EMC [b] faute de mieux. He ascribes the substituion of, for instance, kan’on /p/ for go’on /b/ to the change of EMC [b] to LMC [p] rather than to a change in Japanese (142-47). In support of his position, he cites 5th- and 6thcentury inscriptions in which characters with LOC or EMC glides appear to stand for the antecedents of the OJ /w/ and /y/. On this point, Miyake is right: Ramsey and I should have refrained from speculating that the lenitions occurred specifically in the 7th century. But the rest of his position is problematic. Miyake dislikes the idea that plain voiced obstruents in go’on syllables would have 3

Uchima (2004) argues for a medieval development of Ryukyu [b] and [d] or [d] for main-island /w/ and /y/, respectively, based on a differences in the history of vowel raising in Miyako and Yaeyama dialects as contrasted with those of Amami and Okinawa. If he is correct, then there would be no dialect-comparative evidence supporting pJ plain voiced obstruents. At the same time, Uchima’s theory eliminates an important piece of evidence for an early separation of Ryukyu dialects from those of the main islands; it remains to be seen whether other Ryukyu specialists will accept his claims. 4 Martin (1966) includes the sound correspondences K p :: J w and K t :: J y, to which he added K k :: J Ø in Martin (1972). 5 O J voiced obstruents occur only word-medially. Dialect-comparative and some later documentary evidence suggests that they were prenasalized as well as voiced. Many can be traced back to sequences of a nasal, vowel, and voiceless obstruent, suggesting that they resulted from syncope followed by the spread of voicing from the nasal to the obstruent.

50

J. MARSHALL UNGER

resisted lenition because of the special lexical status of go’on, yet his own hypothetical OJ *ŋ, which existed only in go’on and soon vanished, seems to acknowledge that SJ borrowings were, after all, phonologically distinctive. 6 Remember also that, as mentioned in section 1 above, native mimetic Japanese morphemes evidently constituted a phonologically distinctive stratum in later centuries (Hamano 2000).7 While the inscriptional evidence does make it less likely that voiced obstruent lenition occurred after the go’on borrowings, it is possible that the go’on tradition was more complicated and began earlier than is generally supposed. If the roots of go’on took hold substantially earlier than the 6th century, there could be sufficient time between the inception of go’on and the inscriptions for the lenition to have occurred. Keep in mind too that the lenition theory is supported not just by go’on-kan’on correspondences but by other kinds of evidence. Finally, Miyake’s explanation for those correspondences depends on the idea that the smaller phonetic distance between, say, LMC [p] and LMC [ph] or [p] than between EMC [b] and EMC [ph] or [p] accounts for the abandonment of a previously utilized phonemic distinction in SJ (the descendants of EMC /b/ had not yet merged into LMC /p/ or /ph/). For all these reasons, we should not abandon the lenition theory hastily. Instead, we should revise it to allow for the possibility that lenition preceded the introduction of go’on, which I shall now do. The core idea for this revision goes back to Unger (2000b), in which the lenition *g > Ø plays an important role. Whitman (1985) had projected OJ /w, y, b, d, g, z/ one-to-one back to pJ */w, y, b, d, g, z/, a position even more extreme than Miyake’s. I wanted to show that *g > Ø was possible even under such a view. To do this, I proposed a distinct pJ *ŋ, noting that, after *g > Ø, *ŋ would have been rephonologized as OJ /g/.8 The OJ evidence reviewed so far suggests at least that 6

Some SJ borrowings of all historical periods violate word-structure constraints that apply to ordinary native words. Note also that, for instance, ni ‘two’, a go’on, should have changed to i according to Whitman’s (1985) coronal lenition hypothesis, but obviously did not although there is no evidence that it was pre-OJ *nwi rather than *ni. 7 Miyake also makes the ingenious suggestion that SJ readings such as kai for 開 were not dissyllabic (or, from a modern perspective, bimoraic) in OJ times as were kai ‘oar’ and kui ‘regrets’, noting that 開 is used in Nihon shoki to transcribe OJ ke ‘tree’ (which Miyake reconstructs as *kəy). On this account, pre-OJ *ay and *əy merge to OJ *əy in native words but remain *ay in SJ readings, which become ai in post-OJ stages. Miyake thinks that all -Vi and -Vu in post-OJ SJ readings can be so explained (1999:598-600); if so, then certain sound changes were indeed conditioned by whether a morpheme’s lexical stratum. Why could not the same have been true earlier for go’on words, which were after all the only SJ words in the language at the time of the lenition of plain voiced obstruents? 8 I gave three other reasons for setting up pJ *ŋ. First, Korean has a velar nasal. It is rare, especially in initial position in native words, and no one has yet proposed a Korean-Japanese etymology involving it, yet no one doubts it is a phoneme of Korean. Second, Whitman (1985) proposed some etymologies in which proto-Korean Japanese (pKJ) *ŋ = K n :: J k—all coincidentally in initial position. If proto-Korean-Japanese had a velar nasal, proto-Japanese had at least reflexes of one. Finally, [ŋ] is a common—perhaps the most common—allophone of /g/ in directly observable Japanese dialects.


51

[ŋ] was an allophone of OJ /g/, pace Miyake, and, as I will now explain, there is some new evidence that enables us to spell out the development of pJ *ŋ in considerable detail. 4.

Integrating *ŋ into early Japanese language history Tranter (2001) describes one of the many mokkan, or wooden slats, dating from the 7th century that were excavated at Asuka-ike in 1998. This particular mokkan appears to be part of a word list meant as a pronunciation aid. For example, 蜚 is paired with the combination 皮伊; 尸 is paired with 之. In each item, the first character’s pronunciation is evidently to be deduced from the character(s) following it. But what sort of pronunciations did the writer have in mind? Tranter argues that they were precursors of go’on. For example, 熊 is rendered by the pair 汙吾 (Tranter corrects the text’s 汗 here). Following Baxter (1992), 熊 = 汙吾 would be EMC *hjuwng = *hju ngu. This cannot be a case of fanqie since *-uwng and *-u are different, and anyway, the technique of fanqie dates from around the estimated time of the mokkan. Hence the function of 吾 in the string 汙吾 must be to indicate the coda of the syllable 熊 by means of the initial of the syllable 吾. But read in go’on, the equation 熊 = 汙吾 comes out as u = u gu.9 Evidently, whoever wrote the mokkan was thinking of something more like *uŋ(u) = *u ŋu, and hence pronunciations antecedent to go’on as they have come down to us. Furthermore, another item on the mokkan explicates 忤 by means of 懼, that is, EMC *nguH by means of *gjuH. Both 忤 and 懼 are read gu in go’on, but that cannot be why they are equated inasmuch as all the items on this mokkan presumably reflect pronunciations from the same pre-OJ stage and the equation 熊 = 汙吾 makes no sense if taken as go’on. Hence the equation 忤 = 懼 must signify something like *ŋu = *gu, and insofar as 熊 = 汙吾 implies *uŋ = *u ŋu, it suggests that go’on initial g- were realized [ŋ] at that time.10 Tranter’s mokkan thus provides direct evidence of a pre-OJ velar nasal in the 7th century, and shows that /g/ go’on initials were realized as [ŋ], not [g] or [ŋg]. Let us therefore set up a pJ *ŋ along with plain voiced obstruents *b, *d, *g, *z and suppose that their lenition began with the loss of *g.11 When this occurred, the definition of the phoneme *ŋ (which could occur initially or medially) 9

Incidentally, 熊 = 汙吾 would be yuu = u go in kan’on. The date of the mokkan is enough to make it unlikely that the writer had pronunciations like kan’on in mind, but the difference between yuu and u adds weight to that conclusion. 10 Some might argue that the mokkan item 熊 = 汙吾 was meant to indicate *uŋg(u) = *u ŋgu, i.e., that Japanese of the time could do no better than say [ŋg] when attempting to mimic Chinese [ŋ] or [g] initials. But then we would have, under Miyake’s theory of OJ phonology, OJ u < *uu < *ugu, exactly the kind of change that Miyake says never occurs. One way or another, *ŋ is unavoidable because, at the pre-go’on stage, the medial of 熊 and initial of 吾 have to be the same, yet both different from the antecedent of OJ /g/. 11 This is a reasonable surmise since traces of the obstruents other than *g can still be seen in some Ryukyu dialects (unless Uchima 2004 is correct) and in the occasional s that alternates with zero, for which JLTT reconstructs */z/, whereas evidence for *g comes either from comparisons of Korean with Japanese or internal reconstruction from Old Japanese.

52

J. MARSHALL UNGER

expanded to include medial surface [ŋg], which had developed from reductions of *NVk sequences under certain conditions.12 When SJ words began to filter into the language, this new /g/ accommodated the medial [ŋ] preserved in bimoraic ongana like 香 /kagu/ and 相 /saga/.13 From a synchronic perspective, the [ŋ] and [ŋg] allophones of /g/ would be perceptibly different but not regularly conditioned; this helps explain why later dialects leveled them out in different ways, creating the isoglosses shown in Hirayama’s map. As the weakening of plain voiced obstruents spread, original /b/ and /d/ became glides and surface [mb] < */NVp/ and [ⁿd] < */NVt/ acquired phoneme status. In the case of *z, judging from such well-known alternations as ame ‘rain’ ~ parusame ‘spring rain’, implying *zame, non-initial *z went to s. One may ask why /m/ and /n/ continued to contrast with OJ /b/ = [mb] and /d/ = [ⁿd] whereas *ŋ had joined [ŋg] as an allophone of the new /g/. One reason may have been the time it took for the spread of lenition from *g to other voiced obstruents, but perhaps more important, is that, under this theory, *g went to zero whereas *b and *d became glides that continued to contrast with the corresponding nasals, *m and *n. A greater frequency of /m/ and /n/ compared with /ŋ/ in pJ may also have played a role. It is true that some OJ and EMJ /b/ and /d/ seem to alternate with /m/ and /n/, respectively, but I doubt that any prenasalized stops developed directly from nasals. There are no n ~ d doublets comparable to the well-known m ~ b doublets,14 and b from original m are, in my opinion, all back-formations (Unger 2004). Those OJ b that alternate with m in onbin environments proceeded through a stage [mβ] as explained in Hamano (2000).15 12

This requires modifying or rejecting Whitman (1985) pKJ *ŋ = K n :: J k correspondence. On this account, the bimoraic ongana persisted due to orthographic conservatism; one need not hypothesize special phonological developments to account for them. 14 The key word is “comparable”. There are quite a few pairs like pana.re- ‘separate’ ~ pana.t-, pana.s- ‘split apart’ :: padu.re ‘outskirts’, padi < *padui ‘shame’, odo(ro)- ‘threaten, scare’ ~ oni ‘demon’ ?< *onoi, and une.mye ‘female servant representing a clan’ :: udi ‘clan’, but, in them, the vowels following the alternating consonant also vary, and the meanings of the words are not identical. These extra variations make it easier to question whether the words in a particular pair were in fact derivationally related. Pairs involving n ~ d thus appear to be older than those involving m ~ b. 15 As noted in Section 1, Hamano claims that OJ intervocalic /p/ lenited to /w/ or zero through an intermediate stage [b] > [β], not [] as often suggested with the notations /f/ or /F/. Such a development was possible because OJ /b/ was uniformly realized [mb] and thus was always distinct from /p/ by virtue of prenasalization even if /p/ varied between [p] and [b]. At the conference in Copenhagen where I presented the lecture from which this chapter is derived, one participant objected that OJ /p/ could not have been realized as [b] because CLMC syllables corresponding to ongana for OJ syllables with /p/ initials predominantly showed CLMC /p/ or /ph/ rather than /b/. This criticism is invalid because one can and should determine the number of phonemically distinct OJ syllables strictly on the basis of Japanese evidence (Lange 1973). When Nara period scribes use different ongana for an OJ syllable of a specific phonemic shape, they in effect tell us that all those characters were equally suitable for indicating the features that kept that OJ syllable phonemically distinct from all other OJ syllables. Typically, the characters in an ongana grapheme were not all used equally often in a given text or texts, but in so far as they belonged to the same 13


53

Thus, by including a velar nasal in the pJ inventory and seeing the weakening of plain voiced obstruents as a process that started with the lenition of *g, we get a ready explanation for the allophones of /g/ we find in Japanese dialects, starting with Old Japanese, and a theory that accounts for Miyake’s transient phoneme. Indeed, it is hard to imagine how to explain Tranter’s mokkan without an early *g ≠ *ŋ distinction. Furthermore, this theory does not require that go’on borrowings preceded lenition. If the lenition occurred first, then go’on voiced initials [b, d, z, ŋ] must have been re-introduced from speech of the Korean peninsula in which they were already established, at least in Chinese borrowings.16 The reason for the substitution of kan’on seion for go’on dakuon initials has to do with the differences in who borrowed the Chinese morphemes and the historical circumstances of the borrowing. If the go’on were established first—admittedly the less likely scenario—then the borrowed initials were [b, d, z, g] and the substitution is a consequence of the change in the phonemic inventory occasioned by the lenition. In either case, we should remember that go’on were, at least at first, culturally alien terms, few in number, used by members of a small elite, and for quite special purposes. As specialist shorthand, it is often said that this or that LOC, EMC, or LMC phoneme ‘transcribes’ OJ b or g, but we know perfectly well that there was no omniscient polyglot of the age who chose this or that character for a certain Japanese syllable in a conscious effort to help us, centuries later, figure out OJ phonology. The historical reality was quite the opposite. It must have taken centuries for foreign words to gain the widespread currency they have in Japanese historical documents, let alone to penetrate into the language of ordinary folk. The documentary evidence we now possess of how those words were pronounced are copies of copies of works that in most cases were set down hundreds of years after the first borrowings occurred. We must not give Chinesecharacter data more sway than they are due; in the end, internal evidence from Japanese itself is at least as important.

grapheme, their different frequencies of occurrence indicate only preferences in writing practice of a non-phonemic nature. (Indeed, we detect the collapse of kô-otsu distinctions during the OJ period by tracing the increase of cross-graphemic substitutions over time.) From examination of the net phonetic differences between the sets of Chinese syllables corresponding to the ongana in two OJ graphemes, one can infer what made the corresponding OJ symbols phonemically distinct, but if one ongana is substituted for one another more often than can be ascribed to accident, the conclusion that they stood for one and the same syllable is inescapable. By contrast, even if two characters believed to represent the same Late Old Japanese (LOJ) or EMJ syllable are found in perfect complementary distribution in a given OJ text, one cannot conclude that two phonemically distinct syllables existed in the language of that text unless the number of distinct lexical items in which each character appears is so large that scribal error and chance can be eliminated as reasons for the lack of overlapping distribution. 16 In this regard, Mabuchi’s (1999:413-30) identification of go’on 呉音 and kan’on 漢音 with the parts of Korea associated with the names Kure 呉 (Kolye) and Aya 漢 (Anla, a constituent polity of Kaya) bears careful consideration.

PART II USE OF DIALECTS IN RECONSTRUCTION

CHAPTER 3 PROTO-JAPANESE AND THE DISTRIBUTION OF DIALECTS TAKUICHIRO ONISHI The National Institute for Japanese Language 1.

Introduction This chapter has three purposes. The first is to show the existence of a contrastive distribution between east and west, a famous and classical grouping of Japanese dialects, based on two sister languages, not dialects (§§2-3). The second is to outline a method to calculate the time when the contrastive distribution formed and to do the calculation (§§4-6). The third is to show that proto-Japanese reconstructed on the basis of mainland Japanese is not a primary proto-language but a secondary proto-language (§7). The conclusion is that eastern Japanese dialects which have not been the focus of study must be respected to clarify the nature of pJ (§8). 2.

Patterns of Japanese dialect distribution It is known that various linguistic features show different patterns of distribution in Japanese dialects (Onishi 2002). Concrete examples can be seen in Figures 1-6 (see Appendix at the end of the chapter), from the GAJ (Grammar Atlas of Japanese Dialects). 2.1

Peripheral distribution Figure 1 shows the distribution of obligational expressions corresponding to V-nakereba naranai ‘must V’ in standard NJ. Though this map is complex at first glance, typical patterns of distribution can be extracted. 2.1.1 Peripheral distribution. Figure 2 is a rewritten map based on a subset of the the same data, involving dekinai or dekin (negative of dekiru ‘can do’) and with oenai or oen (negative of oeru ‘can bear’). Words of this type form a distribution like an imperfect ring. From this distribution dekinai or dekin and oenai or oen can be seen to be remnants of words from the old central area. 2.1.2 False peripheral distribution. Figure 3, like Figure 2, is a rewritten map, but features selected grammaticalized forms, where the original analytical expression

TAKUICHIRO ONISHI

58

-nakereba naranai has been shortened to unanalyzable suffixes or modal auxiliaries. Since these forms are grammaticalized, they are obviously newer, but the distribution is similar to the imperfect ring formed by the primary peripheral distribution which was defined as a distribution formed by radiation from the central area (§2.1.1). Though the distribution of the grammaticalized forms resembles the primary peripheral distribution, we call this ‘false’ peripheral distribution since the cause of their formation is different. 2.2

East and west contrastive distribution Clearly divided distributions between eastern and western Japan are called the east and west contrastive distribution pattern (hereafter referred to as EW pattern). This distribution pattern can be divided into two subtypes. 2.2.1 East and west contrastive distribution pattern A. Figure 4 is one of the very famous EW patterns, the distribution of negative suffixes. Eastern -nai and western -n are clearly contrastive and the boundary is distinct. Figure 5 is a map of resultative aspectual forms. Eastern -teiru and western -teoru also show a clear contrast. This pattern of distribution is called the east and west contrastive distribution pattern A (hereafter EW pattern A). 2.2.2 East and west contrastive distribution pattern B. Another subtype can be recognized in the EW pattern. This pattern shows a simple distribution in the eastern area, but a complex distribution in the west. Figure 6 shows the distribution of the past tense negative suffix. Almost all of the eastern area has -nakatta, but in the western area we find -zatta, -nanda and -nkatta; thus the western distribution is complex. This pattern is called the east and west contrastive distribution pattern B (hereafter EW pattern B). EW pattern A and EW pattern B differ in their surface distribution, but the boundaries of each of them are close. While the EW patterns are well-known, the process of their formation has not yet been explained, as their distribution pattern cannot be accounted for by the theory of radiation from a central area or by peripheral innovation. 3.

Causes of the formation of the EW pattern Well-known EW patterns include the following items, shown in Figures 712, which are more detailed maps, composed from language data and altitude data (see Appendix). a. b. c. d. e. f.

Negative suffixes (Figure 7) Imperative forms of monograde verbs (Figure 8) I-onbin forms of –s-stem verbs (Figure 9) Resultative aspect (Figure 10) U-onbin forms of adjectives (Figure 11) Copula forms (Figure 12)

PROTO-JAPANESE AND THE DISTRIBUTION OF DIALECTS

59

On these maps, the western limit line of the distribution of eastern Japanese (EJ) forms closely follows the line of the Hida Mountains (Northern Alps), and Kiso Mountains (Central Alps) and the western valley of the Akaishi Mountains (Southern Alps). East of this line eastern forms can be seen, and in the western area western Japanese (WJ) forms can be seen. The Northern Alps and the Central Alps therefore act as a boundary. Mase (1992:351-378) explains the formation of the EW patterns as follows: originally, EJ forms were dispersed over the entire area; subsequently, WJ forms radiated out from the western area to replace them. Mase identified the substratum of Japanese with EJ varieties, and suggested that the difference between EJ and WJ dialects was originally a difference between languages, but he was not entirely explicit on this point. Mase’s theory is not widely accepted because of his cautious stand on the matter. However, I find the theory attractive, and in the absence of better hypotheses I will pursue Mase’s line of argument below. With regard to the boundaries between east and west, the following distributions appear: a. b. c. d. e.

Expanding WJ forms along the Pacific coast: Figures 7, 8, 9 Expanding WJ forms through valleys: Figures 7, 8, 10 Expanding WJ forms along the coast of the Sea of Japan: Figure 11 Expanding EJ forms along the Pacific coast: Figure 12 Expanding EJ forms along the coast of the Sea of Japan: Figure 12

As the distributions show, there is a tendency for expansion from west to east; in addition the range of expansion from the west is wider than from the east. Another feature is that isolated EJ forms can be seen in the western area. For example in Figure7 -nai exists in Gifu in the western area, but it seems as if it is being pressured out from west. Though these facts do not contradict Mase’s theory, the radiation theory cannot explain two issues. The first is why the boundary exists. Expansion from west to east can be seen and takes the form of radiation from the west, but the expansion does not cross the boundary between east and west. The second point is that the distributions of WJ forms are too widely biased towards the west. If the EW patterns are based solely on radiation from the Kinai area, WJ forms should be equally distributed on either side of the Kinai center. But the distribution of WJ forms is broader to the west of Kinai and stops on the east boundary. With regard to these two points it has been thought that there must be some obstruction to the progress of radiation near the boundaries. High mountains may be considered one such obstruction. However, basing the case just on high mountains is impossible. It is well-known that high mountains or wide rivers obstruct the diffusion of dialectal words or forms, but the obstruction only delays the diffusion, it does not stop it. The boundaries of the EW pattern were established for the first time by the Language Research Commission (LRC 1906). Ushiyama (1953) later researched

60

TAKUICHIRO ONISHI

the boundaries and concluded that the line had not changed over 50 years. Even when comparing the GAJ and LRC data, there is not much difference. In addition, although precise lines are not found, the differences between east and west have continued from at least the Nara period, as shown by the language in the azumauta of the Man’yôshû (Mase 2003:123-141). The thesis of radiation from the central area alone cannot explain the fact that there has been little or no movement in this situation; it must be surmised that something blocked the radiation. The obstruction certainly coincides with the natural wall formed by the high mountain line, but as was pointed out above, the lack of progress of the radiation can not be explained by a geographical obstruction alone. Instead it would be natural to think that there existed a profound cultural or social division, corresponding to the natural border, and that this division functioned as a linguistic boundary. It is likely to have been the case that the linguistic boundary exceeded the level of dialect and was at the level of a language. The two divided languages were sisters, since there are clear phonetic correspondences between EJ and WJ. 4.

Formation of EW pattern B and the masking method If the boundary within the EW pattern is the border between two sister languages (not dialects), it becomes easier to explain the EW pattern B, with simple distribution in the east, but complex distribution in the west. Originally east and west were contrastive, and the discrete radiations from the western central area of Kinai took place several times, at which points the radiated forms spread to the boundary. A few of these forms may have crossed over, but in principle WJ forms stopped at the boundary. On the other hand, from Kinai to the western area, innovated forms progressed smoothly and left ripple-like distribution patterns as their trace. In Figure 6, the distribution of past tense negative suffixes in the western area shows the sequence of historical changes in Kinai, that is, first -zatta, second -nanda, last -nkatta; this sequence is also evident in the textual sources. But in the eastern area, -zatta is not observed at all, having stopped at the boundary without entering the eastern area. Figure 13 shows progressive aspect. It is known that in Kinai -oru changed to –teoru; here, too, the distribution in the eastern area does not reflect this historical sequence (Inoue 1998). On this map the -oru area on the eastern side is obviously more limited than that on the western side. This shows that even long term radiation from Kinai stopped at the boundary and was not adopted in the eastern area. In other words EW pattern B is impossible to explain on the basis of the radiation theory alone, but it becomes possible to explain the distribution by geolinguistic methods (Fig.14), by linking the distribution and radiation order, if we mask off the eastern area and limit the scope of radiation to the western area. I call this the ‘masking method’, since it is achieved by masking off one side of the boundary. I refer to the specific application to Japanese as the ‘east masking method’ (EMM).


61

5.

Time depth calculation of the formation of the EW pattern by using a deviation metrical method As the explanation of the distribution of the progressive aspect in Figure 13 shows, one of the characteristics of the distribution of the EW patterns in the west is the difference in distance between Kinai and the eastern boundary and between Kinai and the west end. Toward the west end radial diffusion progressed smoothly, but on the eastern side the diffusion stopped at the boundary barrier. It is assumed that each word had reached each area at the time of research. It is also assumed that the words would have progressed the same distance towards the east as towards the west if the boundary barrier had not existed. Thus, by dividing the difference between the distance from Kinai to the boundary and from Kinai to the west end by the speed of the diffusion, the time depth between boundary formation and the date of research can be ascertained. The model is shown in Figure 15. Yf is the date of boundary formation of the item of the map. Yc is the date of research, that is, the time at which the map was compiled. Yt is the period during which WJ forms reached the boundary and stopped there. The relationship of these three numbers is as follows. Yf = Yc-Yt Next, a is the distance between Kinai and the west end. b is the distance between Kinai and the boundary line. c is the difference between a and b. S is the speed of diffusion. Yt, the time at which a WJ form stopped at the wall, can be found as follows: a-b = c Yt = c/S Generally, it is difficult to establish S. Tokugawa (1993:391-412) calculated the figure at 0.93km/year, while Tokugawa (1996) calculated it at 0.62km/year. In the following, using these numbers as constants, the former is called tok2, and latter is called tok1. Using the model represented in Figure 15, where the time of emergence (Yo) of the form in Kinai is clear from written sources, S can be expressed as follows. S=a/(Yc-Yo) The method of calculation based on the difference in distance from the radial centre is called the ‘deviation metrical method’ (DMM). The date of formation calculated by DMM varies for each item on each map. Furthermore, since the possibility that other forms were lost cannot be excluded, the result of the calculation is the terminus ante quem, which is to say, the formation was complete by that time. By integrating each result, it is possible to estimate the actual date of

62

TAKUICHIRO ONISHI

formation. Concretely, I have applied DMM to GAJ data, for which Yc the research time, is 1980, as in Figures 16 and 17. a. In the case of progressive aspect (Figure 16), distances are measured as a=575km, b=175km, c=400km. Since Yo of -oru seems to be around 700 AD, we can calculate S for this item as 0.45km/y. Using this figure Yf results in 1091 AD. Using Tokugawa’s constants, if S is tok1, Yf results in 1335 AD, and if S is tok2, Yf results in 1550 AD. b. In the case of the negative suffixes (Figure 17), distances are measured as a=1700 km, b=175 km, c=1525 km from the map. As there is no clear date of origin for these suffixes, the actual Yo is unknown. However, if we take S to be tok1, Yf results in 480 BC and if S is tok2, Yf results in 340AD. Thus, in the case of the negative suffixes, the date of EW pattern formation is calculated as being in prehistoric times. 6.

Issues to resolve The arguments and the methodology above are new, so there are several issues to resolve. 6.1

Speed of radial diffusion Calculation by DMM depends on the speed of radial diffusion. Tokugawa’s constants were calculated on the basis of lexical items; hence there remains a problem in applying these constants to grammatical items. Moreover, there is no certainty that the speed of diffusion is fixed. The constant will need to be independently confirmed by future research. 6.2

Setting the centre of radiation After historical times, assuming Kinai as the centre of radiation is unproblematic, but in prehistoric times the centre is unidentified. For example, 480BC was calculated as a time of boundary formation in the case of the negative suffixes, but it is necessary to treat this number with caution. 6.3

Diffusion from the eastern area Arguments here are based solely on radial diffusion from Kinai as the centre in the west. On the other hand, another hypothesis could involve diffusion from the east. But, as touched upon in §3, from a geographical viewpoint it is difficult to consider diffusion from the eastern area. If there were diffusions from the east, WJ forms would be expected to remain in the eastern area, but such data has not been found. 6.4

Conquest or collision In relation to §6.3 above, we might inquire as to whether the formation of the EW pattern was due to conquest from the west or collision between east and


63

west. On present data there is no evidence of force being exerted from the east. Had that been the case, there would have been evidence of a more compressed ripple pattern based on the discrete radiation between Kinai and the EW boundary, but such examples have not been found, so more detailed explanation may be needed. Radial diffusion in dialects is based on the cultural prestige of a central area, and it is not only linguistic forms which are radiated. If there are two centers, each point in each area should receive cultural pressure from both radial centers. It would be assumed that the ripples or waves of diffusion would become compressed. In addition, if two forces collide, there should be interference near the boundary. But few examples of this exist; one is the distinction between progressive and resultative aspect using the verb iru in the village of Kaida in Nagano prefecture. Therefore, conquest from and by the west is the best explanation. 6.5

EW patterns and proto-languages My argument here set out from Mase’s geolinguistic account, explained in §3, that EJ dialects are a substratum gradually overlaid by WJ dialects; I have concluded by rejecting Mase’s account. One piece of significant evidence for Mase’s theory is that -ro, the distribution of which in Kyushu can be seen in Figure 8, shows peripheral distribution. But as Kobayashi (1996) and Hikosaka (2001) treated it as a recently innovated form, the distribution of -ro in Kyushu may be a false peripheral distribution, and therefore cannot be taken as evidence for Mase’s theory. Even if the EJ dialects form a substratum, as in Mase’s theory, at least 1,200 years must have passed since the formation of the EW pattern, as Mase himself indicates by citing the Man’yôshû azuma-uta (Mase 2003). Considering the long period of time following the formation of the EW pattern, it is impossible to reconstruct pJ solely from contemporary EJ dialects; we must instead compare EJ and WJ dialects as sister languages. 7.

Significance of the EW patterns In the preceding sections it has been confirmed that the differences between EJ and WJ dialects reflect sister languages and that the EW patterns were formed in prehistoric times. This fact is important when considering pJ. Hitherto it has been thought that the only sister languages of Japanese were the Ryukyuan language and mainland Japanese. There are many dialects in mainland Japan, and they have been treated on a par with each other. But as this chapter has shown, the eastern area and the western area were occupied by different languages. The time at which these two languages divided does not antedate the separation of Ryukyu and mainland languages. Therefore the candidate to compare with Ryukyuan language is a secondary proto-language which will be reconstructed from EJ and WJ varieties, see (1).

TAKUICHIRO ONISHI

64 (1)

proto-Japanese

proto-Ryukyuan

proto-mainland Japanese

proto-western Japanese

proto-eastern Japanese

EJ dialects have not been used extensively for, e.g., reconstruction of protolanguage, because of their similarity with standard Japanese and historically simplified features. But as has been shown, EJ dialects do have significance for historical analysis. Certainly EJ dialects are similar to standard Japanese. This is because standard Japanese was built on the Tokyo (Edo) dialect which belongs to the EJ group. If one examines EJ dialects closely, many distinctive features can be found there. It must be especially noteworthy that most of these are basic grammatical elements. For example, the fact that the present tense progressive aspectual -ita is widely observed in the eastern area. It is used in the Tohoku area extensively, but remains distributed throughout the eastern area as a fossil-like form in greeting expressions. Other examples include: the adjective conjugation element -su in the northern Tohoku area (i.e., take-su-ta ‘was high’); consonant final verb su (i.e., sa-ne ‘not to do’, su ‘to do’) in the Tohoku area (Onishi 2004); the adjective final element -ke (i.e., taka-ke ‘to be high’) and verbs with o final forms (i.e., tat-o ‘standing’) in the Hachijo islands and Akiyamago in Nagano prefecture. These Hachijo and Akiyamago examples are regarded as unchanged descendants of ancient EJ dialect forms. Since many examples such as these which cannot be explained by Kinai textual sources can be found, the historical importance of EJ dialects must be recognized. 8.

Conclusion Based on the convergence of dialectal and topographical data, I have shown in this chapter that the EW patterns reflect differences between two sister languages: WJ and EJ. I have also outlined a method for calculating the time of boundary formation and applied it to the case at hand. On the basis of these results I have argued that we should reconstruct proto-mainland Japanese from WJ and EJ, respectively, taking full account of EJ dialect data. I have further argued that this pMJ in turn is what should be compared with pR in order to reconstruct pJ. There remain many issues to resolve. Above all in determining the speed of diffusion, geographical data, such as that available through the geographical information system (GIS) should be taken into consideration.


APPENDIX: Figures

Figure 1: Obligational expression: (ika)nakerebanaranai. GAJ4-206

65

66

TAKUICHIRO ONISHI

Figure 2: Obligational expression (ikanakerebanaranai) Peripheral distribution


Figure 3: Obligational expression (ikanakerebanaranai) Grammaticalized forms show false peripheral distribution

67

68

TAKUICHIRO ONISHI

Figure 4: Negative suffixes of verbs: (kaka)nai. GAJ2-80


Figure 5: Resultative aspectual forms: (tit)teiru. GAJ4-199

69

70

TAKUICHIRO ONISHI

Figure 6: Past tense negative suffixes: (ika)nakatta. GAJ4-151


Figure 7: Negative suffixes

Figure 8: Imperative forms of the monograde verb ‘look’

71

72

TAKUICHIRO ONISHI

Figure 9: I-onbin form of the s-stem verb ‘took out’

Figure 10: Resultative aspectual forms.


Figure 11: Adjective u-onbin form ‘(not) expensive’

Figure 12: Copula forms with the adjectival noun sizuka- ‘quiet’

73

74

TAKUICHIRO ONISHI

Figure 13: Progressive aspectual forms: (tit)teiru. GAJ4-198


Figure 14: East masking method (EMM)

75

76

TAKUICHIRO ONISHI

Figure 15: Deviation metrical method (DMM)


Figure 16: Application to progressive aspectual forms (GAJ4-198)

77

78

TAKUICHIRO ONISHI

Figure 17: Application to negative forms

CHAPTER 4 THE USES OF RYUKYUAN IN UNDERSTANDING JAPANESE LANGUAGE HISTORY1 LEON A. SERAFIM University of Hawai‘i at Mānoa 1.

Introduction This chapter seeks to give an indication of Ryukyuan (here, mostly Shuri) language history, and the ways in which the knowledge of that history illuminates what we know from work on the history and prehistory of mainland Japanese (mostly OJ and standard NJ). Then, turning things around, I will suggest how knowledge of mainland Japanese language history illuminates our understanding of what has happened in Ryukyuan. Following some general background (§§2-4), I will address issues of importance in the field of Ryukyuan historical phonology (§§5-7), morphology (§8), and syntax (§9). Finally, in the conclusion I discuss the relationship of Ryukyuan to mainland varieties of Japanese (§10). 2.

Historical background Shuri is the dialect of the former gentry and nobility of the old Kingdom of Ryukyu (also Chûzan). Shuri also refers to the old capital of this kingdom, which existed from approximately the 1300s to 1609 as a completely independent entity, and, as a vassal state of Satsuma in southern Kyushu (and indirectly, of the Tokugawa Shogunate) until 1868, with a following transitional period leading to full incorporation into the Meiji State as Okinawa Prefecture in 1879. The Ryukyus, geographically, culturally, and linguistically, are a separate entity from the rest of Japan. Geographically they are separated from Kyushu by a buffer zone of, south-to-north, the tiny Tokara Islands, and then by the substantial islands of Yaku and Tane. They arc from north(east) to (south)west, see Map 1: from well south of Kyushu, first (1) the Amami group, from Amami Oshima on the north side to Yoron in the south, then (2) the Okinawa group, constituting the single largest island, Okinawa, and its outliers, then (3) a large expanse of open 1

In addition to the presentation in Copenhagen, an earlier version of this chapter was presented at a lecture in October 2004 at Cornell University. I would like to thank the East Asia Program at Cornell for their gracious invitation and underwriting for the trip. I am also grateful to John Whitman and Bjarke Frellesvig for feedback on this paper.

80

LEON A. SERAFIM

Map 1: The Ryukyu Islands water, followed by (4) Miyako and its surrounding islands, including the outliers Tarama and Minna, halfway or more to Yaeyama; then (5) Yaeyama, consisting of the large islands of Ishigaki and Iriomote and their surrounding islands; and finally (6) Yonaguni Island (usually included in Yaeyama) within view of the northeast of Taiwan. (1-2) constitute the Northern Ryukyus, both geographically and linguistically, and (4-6) constitute the Southern Ryukyus, frequently called Sakishima in linguistic and geographical discussions. ‘Japanese’, that is, all the languages and dialects of mainland Japan and the Ryukyus, probably consists of five languages, according to the dialect-chain definition of Hockett (1958, ch. 38). Where the chain of mutual intelligibility breaks, that constitutes a language boundary. This is a useful but rough-and-ready way of ascertaining distinct languages. According to this definition, all of Japan except the Ryukyus constitutes one language, though distant dialects of it may be mutually unintelligible. As long as there is a chain of intelligibility, though, it is by definition one language. The Ryukyus consists of roughly four languages: (1) Northern Ryukyuan, with mutual unintelligibilty between far-flung dialects; (2) Miyako; (3) Yaeyama; (4) Yonaguni. Of course, nearly all speakers of all varieties of Japanese today have a

RYUKYUAN AND JAPANESE LANGUAGE HISTORY

81

good understanding of standard Japanese, and the mutual unintelligibility criterion necessarily excludes such overarching standard languages. 3.

Moras of the Shuri dialect (1) is an overview of the building blocks of words of the Shuri dialect (Sr), which constitutes the core of modern Common Okinawan, widely spoken in the Okinawa-Island region. It is a table of the moras of the dialect, in the version found in the Okinawago Jiten (Dictionary of Okinawan) of 1963, the dialect formerly spoken by members of the male gentry, with certain features taught to them in schooling, which the other dialects of Okinawan have largely lost. (1) Obsolescent Shuri moras (first half of 20th century—now used in performance) p… b… m… py… by… my… t… d… n… r… ny… c… z… č… ž… k… g… kw… gw… … h… y… hy… w… hw… ‘w… ‘y… Other

4.

…i pi bi mi

…e pe be me

ti [ti] di [di] ni ri

te de ne re

ci [tsi] zi [(d)zi] či [ti] ži [(d)i] ki gi kwi [ki] gwi [gi] i hi [i]

ce [tse] ze [(d)ze] če [te] že [(d)e] ke ge kwe [ke] gwe [ge] e [je] he

wi [i] hwi [i] ‘wi [i] ‘yi [ji] ‘N, N, Q, 

we [e] hwe [e] ‘we [e] ‘ye [je]

…a pa ba ma pya bya mya ta da na ra nya ca [tsa] za [(d)za] ča [ta] ža [(d)a] ka ga kwa [kwa] gwa [gwa] a ha ya hya [ja] wa [wa] hwa [(w)a] ‘wa [wa] ‘ya

…o po bo mo pyo byo myo to do no ro nyo co [tso] zo [(d)zo] čo [to] žo [(d)o] ko go

…u pu bu mu pyu byu myu tu [tu] du [du] nu ru nyu cu [tsu] zu [(d)zu] ču [tu] žu [(d)u] ku gu

o ho yo hyo [jo]

u hu [u] yu hyu [ju]

‘wo [wo] ‘yo

‘wu [wu] ‘yu

Vowel systems in Ryukyuan languages The Modern Sr vowel system looks like (2). It has five vowels, with i, u, a commonly appearing both long and short, but e, o rarely appearing short. When one hears a statement such as “The Okinawan dialect only has three vowels, i, u, and a”, this is the reason for it. The statement is almost but not quite true for short

LEON A. SERAFIM

82

vowels only. Examples of words with short e and short o, respectively, are gweQtai ‘a muddy mess’, and čoNdara ‘outcaste puppeteers’. (2)

i, i (e), e

a, a

u, u (o), o

Within Ryukyuan, only one dialect/language has only three vowels, and that is Yonaguni, with the system i, u, a. It has undergone repeated raisings and monophthongizations. All other dialects have at least four vowels, and some, as many as seven. The typical vowel system of Amami in the north looks something like this (Shodon, Kakeroma Island, Amami): (3)

High Mid Low

Front Palatal Front Non-Palatal Back i  u e  o a

A typical pattern in Miyako (Southern Ryukyuan) is like (4) (Hirara City): (4)

5.

High Mid Low

Front Central Unrounded Back Rounded i  u e o a

Common correspondences Below is a table of correspondences between Sr, NJ, and OJ, for vowels and consonants, the latter in mora templates, see (5). It is fairly obvious from the correspondence table that the most typical source of Sr mid vowels is earlier diphthongs, themselves typically resulting from the loss of a medial consonant, or the creation of a semivowel that figured in the monophthongization. It is also quite obvious that Sr has undergone a raising of mid to high vowels. In addition, comparison of OJ with both NJ and Sr vowels suggests large-scale vocalic mergers in Japanese vowel history. A comparison of Sr, NJ, and OJ leads to a suspicion that OJ vowels may in part be due to monophthongizations. There will indeed be diphthongs or two-mora sequences at the pJ level. The subtable of consonant (presented as mora) correspondences in (5) shows that affrication of pJ *t and *d (or *Nt) has occurred after high vowels, and that, in Sr, the syllables *ki and *gi, too, have affricated, e.g. (6). Further, there appear to be exceptions to the assumption that certain vowels in NJ correspond to affrication-inducing vowels in Sr, as with, e.g., the common pattern Sr či, či :: NJ či, ki shown in (7) versus the less common pattern Sr ti, ki :: NJ či, ki in (8). It seems clear enough that normally Sr ti, di, ki, gi correspond to NJ te, de, ke, ge, respectively, (9).


(5)

Sr

ee e

oo o

i

u

NJ

ae ai

ao au uwa

i e

u o

OJ

aCe aCi aCwi

awo aCu uCa

i, -wi e, -ye

u o, -wo

83

Sr

ci

zi

či

ži

ti

di

ki

gi

NJ

tsu

zu

či ki (i)te (i)ke

ĵi gi (i)de (i)ge

te či

de ĵi

ke ki

ge gi

OJ

tu

zu, du

ti ki, kwi (i)te (i)ke, (i)kye

zi, di gi, gwi (i)de (i)ge, (i)gye

te ti

de di

ke, kye kwi

ge, gye gwi

(6)

a. b.

kači ‘to write’ < *kaki wiiži ‘to swim’ < *oyogi

(7)

a. b.

Sr tači :: NJ tači ‘to stand’ Sr kači :: NJ kaki ‘to write’

(8)

a. b.

Sr uti :: NJ oči ‘to fall’ Sr uki :: NJ oki ‘to get up’

(9)

a. b. c. d.

Sr tii Sr udi Sr kii Sr kaagi

:: NJ te ‘hand’ :: NJ ude ‘arm’ :: NJ ke ‘fur, hair’ :: NJ kage ‘shade’

Some NJ te, de, ke, ge correspond to Sr či, ži, či, ži. The trigger for the sound change Progressive Palatal Assimilation is found in the preceding syllable: (10)

2

a. b. c. d.

Sr kači (< *kače < *kakitye < *kakite) :: NJ kaite ‘writing’ Sr wiiži (< *oyože < *oyogitye < *oyogite) :: NJ oyoide ‘swimming’ Sr iči2 (< *iče < *ikye < *ike) :: NJ ike ‘pond’ Sr hwiži (< *piže < *pigye < *pige) :: NJ hige ‘beard’

‘(Garden) pond’.

LEON A. SERAFIM

84 6.

Important sound changes in Shuri history (in approximate order) How, then, have such correspondences come into being? (12) to (20) present a set of sound changes, with one possible ordering, and the lexical items that instantiate them; first, though, (11) (next page) presents a table of the forms running through those changes to provide an overview. Glosses and full descriptions of the sound changes also follow. The First Vowel-Raising (VR1) raised many *e and *o to *i and *u, respectively, followed by monophthongization. (12a-d) exemplify raising; (e-g) are examples where raising is blocked. (12)

a. b. c. d. e. f. g.

?*sogoy ‘exceed’ > *suguy > *sugï (cf. OJ sugwi) [alternatively, > *sogey > *sugi] ?*tokoy ‘moon’ > *tukuy > *tukï (cf. OJ tukwi) [alternatively, > *tokey > *tuki] ?*kotoy ‘mouth’ > *kutuy > *kutï (cf. OJ kuti) [alternatively, > *kotey > *kuti] *asob- ‘play’ > *asub- (cf. OJ aswob-) *kezu ‘a scratch’ (because of following coronal plus high vowel; cf. OJ kizu) *medu ‘water’ (because of following coronal plus high vowel; cf. OJ midu) *moko ‘bridegroom’ (perhaps due to vowel length or compound status in pJ; cf. OJ mwokwo ~ mukwo)

Progressive Palatal Assimilation (PPA), *iC > *iCy, sometimes followed by loss of the triggering high front vowel, is shown in (13). (13)

( ) a. *pigë = [pjig] ‘whiskers’ > *pigy ë = [pjigj] (cf. OJ pige)3 b. (?*e-ka ‘?-Q’ >) *ika ‘how’ > *ikya > … > ča = [ta] (cf. OJ ika)4

3 Sr furnishes independent evidence for prehistoric prenasalization of voiced obstruents, such as: Nži- ‘leave’ < *idye- = *[idje] < *idë- < …. Cf. Nni ‘riceplant’ < *inye = *[ie] < *inë < …, with overt nasal. 4 The morpheme * e- ‘wh-’ is suggested by Sr ici =[itsi] ‘when?’, if it is not a loan, because it should come from proto-Ryukyuan (pR) *etu, presumably due to blocking of raising by the following coronal-plus-high-vowel environment. However, we are apparently required to reconstruct pR *i- in the case of pR *ido- :: OJ idu- :: NJ do- (< Eastern OJ ido-), but with ultimate pJ *e-n#to- ?*‘which one’ or ?*‘which GEN GEN’. Sr žiru ‘which one?’ = [diru] < [irregular metathesis] *žuri < *Nžuri < *idyore < *idorë < ?*edorë. See also Frellesvig & Whitman, this volume, for external links to Korean.

ciči čiči

siži

šiži

PNPM

post-Sr

ʔašib-

ʔasibkiži

miži

mizi

*mɪzɪ

*kɪzɪ

kizi

*mezɪ

*mezu

*medu

*medu

*kezɪ

*kezu

*kezu

muːku

muːku

*moːko

*moːko

h(w)iži

*hwiži

*pigi

*pigye

*pigyë

*pigë

*pigay

čaː

*ʔiča > čaː

*ikya

*ika

*eka/*ika

tuči

tuči

*tuki

*toki

*töki

ʔuki

ʔuki

*ʔukɪ

*oke

*ökë

*ököy

ʔagi

ʔagi

*ʔagɪ

*age

*agë

*agay

Either coronal consonant plus *uy (with *y) itself breaks a widely attested constraint against raising the first in a sequence of two identical type-A mid vowels, or else the actual sequence of vowels leading to Sr was different, most likely *oy > *ey = *[jej]. Perhaps relevant here are Sr kami ‘deity/-ies’ (< *kamey), but NčaN ‘ancestral deities’ = N-čaN ‘exalted deities’ (< *mi-kyami(y) = *[…mji(j)] < *mi- + *kamey). Pre-Sr *mi and *mu usually give N, while Sr mi or mu must have either shortened from a long-vowel syllable, or else be from *m plus mid vowel. Cf. OJ kamu#X, but—perhaps a match—kamo## ‘name of important shrine’ (if the river is named after the shrine or a deity, and the shrine is not simply named after the river). 2 i.e., Monophthongize.

1

*cɪči

kuči

*kuči

*cuki *cɪki

*kuti

*ʔasɪb

*asub-

*kutuy1 *kutï

*asob-

?*kotoy

*tuki

*sɪži

*sɪgi

*sugi

*tukï

*tukuy

?*tokoy

VAf

VR2

PSF

CAs/f

ABC

PPA

*sugï

M-ize2

?*sogoy

*suguy

pJ

VR1

(11)

LEON A. SERAFIM

86

Loss of the A::B (kô-rui–otsu-rui) vowel opposition in favor of neutral ‘C’ (ABC) is shown in (14). (14)

a. b. c. d. e. f. g. h. i.

*töki ‘time’ > *toki (cf. OJ toki, NJ toki) *sugï ‘exceeding’ > *sugi (cf. OJ sugwi, NJ sugi) *tukï ‘moon’ > *tuki (cf. OJ tukwi, NJ cuki) *kutï ‘mouth’ > *kuti (cf. OJ kuti, NJ kuči) *ökë ‘arising’ > *oke = *[okh], with lax mid front vowel (cf. OJ okwi, NJ oki)5 *pigye ‘whiskers’ (cf. OJ pige, NJ hige) *kezu ‘a scratch’ > *kezu (cf. OJ kizu, NJ kizu) [but *e > *i in some Amami dialects] *medu ‘water’ > *medu (cf. OJ midu, NJ mizu) *moko ‘bridegroom’ > *moko (cf. OJ mwokwo ~ mukwo, NJ muko)

Coronal Assibilation (CAs) and Coronal Affrication (CAf) (aspects of the same change) are shown in (15): (15)

a. *tuki ‘moon’ > *cuki (cf. OJ tukwi, NJ cuki) b. *medu ‘water’ > *mezu (cf. OJ midu, NJ mizu) c. *kuti ‘mouth’ > *kuči (cf. OJ kuti, NJ kuči)

Post-Sibilant Fronting of high back vowels (PSF): *u > * > * is shown in (16): (16)

a. b. c. d. e.

*sugi ‘exceeding’ > *sgi > *sgi (cf. OJ sugwi, NJ sugi = [sgji]) *cuki ‘moon’ > *cki > *cki (cf. OJ tukwi, NJ cuki [tskji]) *mezu ‘water’ > *mez > *mez (cf. OJ midu, NJ mizu) *kezu ‘a scratch’ > *kez > *kez (cf. OJ kizu, NJ kizu) *asub- ‘play’ > *asb- > *asb- (cf. OJ aswob-,6 NJ asob-)

Second Vowel Raising (VR2), raises all remaining mid vowels to high vowels:7 (17)

a. b. c. d.

*age ‘raising’ > *ag (cf. OJ age, NJ age) *oke ‘arising’ > *uk (cf. OJ okwi, NJ oki) *mez ‘water’ > *mz (cf. OJ midu, NJ mizu) *kez ‘a scratch’ > *kz (cf. OJ kizu, NJ kizu)

5 See below for a discussion of forms such as this in terms of their interaction with the FrellesvigWhitman pJ seven-vowel hypothesis. 6 OJ apparently has a constraint against raising a morpheme-final type-A mid vowel in a quadrigrade verb, giving such forms as kapyer- ‘return’, aswob- ‘play’, possibly also (?*ma#wos>) mawos- ~ maus- ‘humbly say’ or (?*wor- >) wor- ‘exist (animate)’. Apparently Ryukyuan does not have this constraint, resulting in an otherwise unexpected vowel correspondence (for ‘play’) of Sr i :: OJ -wo. Possibly—see below—the constraint against raising held in no dialect if the verb in question grammaticalized to an auxiliary, leading—of interest to us—to the form *...#wur(-) in the case of *wor- used as a progressive marker. See also Frellesvig & Whitman, this volume. 7 It is approximately the output of this sound change that constitutes the phonology of around 1500.


87

e. *toki ‘time’ > *tuki (cf. OJ toki, NJ toki) f. *pigye ‘whiskers’ > *pigi (not *pig;8 cf. OJ pige, NJ hige) g. *moko ‘bridegroom’ > muku (cf. OJ mwokwo ~ mukwo, NJ muko) Velar Affrication (VAf) before *i (but not *) or *y is shown in (18): *k, *g > č, ž / __ *i / *y. The epicenter of this change appears to be Sr; one may speculate that the change was triggered or accelerated by a similar change in Mandarin Chinese. 9 The Okinawan change is after 1500 (cf. Hagers 1997), but I do not know the exact periodization of the Chinese change. (18)

a. b c. d.

*sgi ‘exceeding’ > *sži (cf. NJ sugi) *cki ‘moon’ > *cči (cf. OJ tukwi, NJ cuki) *tuki ‘time’ > tuči (cf. NJ toki) *pigi ‘whiskers’ > … > hwiži (cf. NJ hige)

Palatal/Non-Palatal Merger in vowels (PNPM) is shown in (19). This must have happened just before the arrival of Westerners to Okinawa around 1800. (19)

a. b. c. d. e. f. g.

*sži ‘exceeding’ > siži (cf. NJ sugi) *cči ‘moon’ > ciči (cf. NJ cuki) *ag ‘raising’ > agi (cf. NJ age) *uk ‘arising’ > uki (cf. NJ oki) *mz ‘water’ > mizi (cf. NJ mizu) *kz ‘a scratch’ > kizi (cf. NJ kizu) *asb- ‘play’ > asib- (cf. OJ aswob-, NJ asob-)

Further changes to Common Okinawan include loss of the s :: š distinction (as [sa, so, su, še, ši]), and of the c :: č distinction and the z :: ž distinction (with allophony depending on dialect—Sr always picks č and ž), see (20). (20)

a. b. c. d. e.

siži ‘exceeding’ > šiži (cf. NJ sugi) ciči ‘moon’ > čiči (cf. NJ cuki) mizi ‘water’ > miži (cf. NJ mizu) kizi ‘a scratch’ > kiži (cf. NJ kizu) asib- ‘play’ > ašib- (cf. OJ aswob-, NJ asob-)

6.1

Evidence for an A::B vowel opposition in proto-Ryukyuan Finally in this section, I will mention independent phonetic evidence from within Ryukyuan which supports the reconstruction of an A::B vowel opposition for pR. Serafim (1977a) notes that Northern and Southern Ryukyuan differed markedly in the phonetics of their vowels, and that these facts can be afforded a 8 9

One may say that y +  > i. Ryukyuans were in an official tributary relationship with China from 1372.

LEON A. SERAFIM

88

ready explanation if the proto-vowel systems (pNR and pSR) have front vowels with the phonetics shown in (21). But it also seems plausible, though not necessary, to merge the pairs of vowels of proto-Ryukyuan (pR) at each height using opposite phonetic values in pNR and pSR to explain the vowel systems from which later developments ensue. Cf. here also Miyake’s (2003:217-220) reconstruction of the diphthong-derived Nara OJ type-B i (i2) as [y] or [].). (21)

pNR /i/ = [i] < /e/ = []
>

pSR /i/ = [] /e/ = [e]

(22)

pR *pigë = *[pig] (not *pïgë = *[pg]) ‘whiskers’ > NR *[pigy] > … (Okinawa): Ie-dialect tiži = [tjii] > SR (Miyako): Hirara pgi = [psgji] ‘hair’

(23)

pR *tukï = *[tuk] (not *tuki = *[tukji]) ‘moon’ > NR (Amami): Yamatohama cki = [tskji] ‘moon; month’ > SR (Miyako): Hirara tk = [tsks] ‘month’

(24)

pR *medu = *[mjedu] (not *mëdu = [mdu]) ‘water’ > NR (Amami): Yamatohama mz = [mwz] > SR (Miyako): Hirara miz = [mjidz]

(25)

pR *agë = *[ag] ‘raising’ (not *age = [agje]) > NR (Amami): Shodon age = [ag] > SR (Miyako): Hirara agi = [agji]

Vowel height vs the A::B distinction10 The problem to be raised here is the incongruity of bigrade conjugations between OJ and NJ, on the one hand, and Ryukyuan and even other mainland Japanese dialects on the other, (26).

7.

(26)

Sr 11

Pre-Sr

pJ

OJ

NJ

pJ UB

siži-

< *sugi-

< *suNkuCi-/*soNkoCi-‘exceed’

> sugwi-

> sugi

pJ MB pJ LB

ukiaki-

< *oke< *ake-

<
okwi> ake-

> oki > ake

*ököCi- ‘arise’ *akaCi- ‘open (it)’

The solution in the form of a pJ reconstruction is presented in the center, in the same way that Hattori presented it, and the movement to the modern dialects is presented in either direction from it. Critical in actually doing the reconstruction 10

See Hattori (1976b), Serafim (1977b, 1984), Unger (1993 [1977]), Frellesvig & Whitman (2004). UB (upper bigrade) is the Japanese-language kami-nidan and LB (lower bigrade) is shimo-nidan; MB (middle bigrade) is a new coinage. 11


89

have been the Shuri data, on the left (any other Ryukyuan dialect for which there are sufficient data would do), and the OJ data on the right. Even using NJ on the right would still afford a three-vowel-heights solution of some sort, so the setting up of pJ upper-, middle-, and lower bigrade conjugations is clear even without the knowledge of the existence of the A::B distinction. Velars appear throughout in the table, but the paucity of data for pJ upper bigrade requires use of a voiced velar in that instance. But cf. also the noun dyad Sr ciči :: OJ tukwi ‘moon’, the same type of correspondence as the pJ upper bigrade, but fully congruent in velar voicing.12 This subdivision into three bigrade verb subtypes is based on the Sr ki :: OJ kwi (even just the NJ ki) correspondence. Note that a successful reconstruction necessitates the use of syllables in which a trace of the height of the vowel is preserved in the preceding consonant, with lack of affrication showing i < *e,13 and affrication showing i < *i or *ye, (27). (27) {

{

Sr pJ OJ či, ži *kX, *gX kwi, gwi ki *kY kwi ki *kZ ke

} }

There appear to be three different vowel qualities involved, even though each of the daughter languages has only two categories. Sr has palatalizing and nonpalatalizing i, while OJ has high B-type -wi and mid B-type -e. One possible solution is to have a three-height set: *kï,*gï :: *kë :: *kä (Serafim 1977b), while another is to have a starting set of diphthongs, each with a back vowel of different color, likely, but not necessarily, of three heights. This is the *(N)kuCi :: *köCi :: *kaCi hypothesis (Hattori 1976b, Unger 1993 [1977]), outfitted with an assumption of a consonant loss to yield the diphthong. Supporting evidence for the diphthongal hypothesis comes from a comparison of OJ verb pairs (i.e., from internal reconstruction) with the alternations in (28a): -wi :: u; -wi :: -o; -e :: a. Note that the quadrigrade members of the dyads show a three-way difference in vowel heights. However, I should add that, in addition, EOJ has sugwosi, suggesting the hypothesis of a starting point of *suNko- or *soNko-, which breaks with the three heights observation, and goes with three qualities, instead.14 The starting point will of course look identical in (Pre-)pJ insofar as the verb roots are concerned—that is to say, there will be no allomorphy of roots, regardless of the particular form of the reconstruction, presented here in (28b) and (c). The second of the two alternatives assumes differing vowel quality in mid vowels for the two outcomes. Necessarily, the string *oCi must lead to a highvowel outcome in both mainland Japanese and Ryukyuan. Looking at the use of 12 The obvious choice would have been a verb corresponding to NJ cukiru ‘become used up’, but none presents itself. 13 Frellesvig & Whitman (2004) present a different hypothesis, discussed below. 14 The Frellesvig-Whitman (2004) hypothesis concerning the pJ verb roots also accounts for the data here—and other data—with quality distinctions as well as height distinctions.

LEON A. SERAFIM

90

Ryukyuan in obtaining reconstructions of this sort, it is no accident that Shirô Hattori wrote mostly about Ryukyuan in NSNT. (28)

a.

pJ UB; OJ UB pJ MB; OJ UB pJ LB; OJ LB

b.

c.

‘exceed1’ sugwi ‘arise1’ okwi ‘raise’ age

:: :: ::

sugusi okori agari

‘exceed2’ ‘arise2’ ‘rise’

OJ QG; pJ QG OJ QG; pJ QG OJ QG; pJ QG

pJ UB pJ MB pJ LB

*suNkuCi *ököCi *aNkaCi

:: :: ::

*suNkusi *ököri *aNkari

pJ QG pJ QG pJ QG

pJ UB pJ MB pJ LB

*soNkoCi *ököCi *aNkaCi

:: :: ::

*soNkosi *ököri *aNkari

pJ QG pJ QG pJ QG

7.1

The Frellesvig-Whitman *-lowering hypothesis Recently Frellesvig & Whitman (2004:290, hereafter F&W; see also Frellesvig and Whitman, this volume) have proposed a seven-vowel system for pJ, different in detail from that proposed by Hattori (e.g., 1976a, b, NSNT). In addition to the classic five vowels, they propose the high and mid central unrounded vowels * and *. Equivalent to the F&W *, a high central unrounded vowel, Hattori proposed *ü = [], a high central rounded vowel, and for F&W’s *, Hattori proposed the same, pJ * < pre-pJ **ä = [], essentially a rewriting. Both hypotheses propose that at least some instances (in the case of F&W, all monophthongs) of the high central vowel were lowered to a mid central vowel, merging with the extant mid central vowel: * > * > OJ o, but the diphthong *i does not lower (> pre-OJ *wi > OJ (w)i). Hattori’s hypothesis has some instances of the monophthong *ü merging with the back rounded vowel *u. He does not accept lowering before the formation of diphthongs (e.g., no *üi > *i > …), although he does not say so in so many words. F&W assume diphthong formation first, followed by lowering of non-diphthongized *. The big difference between the two views is that Hattori posits relatively few lexical items with pJ ü, although he implies a larger number, while F&W posit a relatively larger number of them, including all verbs characterized here as middle bigrade. Thus, where Hattori (NSNT 19:110) explicitly posits pJ middle bigrade *ki ‘arise (v.i.)’, F&W would posit an equivalent etymon *ki (or, in Hattori-style spelling, *üküi):15 *ki > OJ okwi. While Hattori’s corresponding transitive verb would be pJ *ks‘raise (v.t.)’,16 F&W would have *ks- (equivalent to a Hattori-style spelling of *üküs-, though of course Hattori does not posit a lexical item with a second high vowel), but with lowering of the two non-diphthongal vowels: *ks- > OJ okos-. 15

Their actual intransitive-verb example is opwi-, corresponding to opo- ‘big’ (Frellesvig & Whitman 2004:293). 16 Or possibly *üks- = [ks]; see below.


91

The rules are the same, but they apply to different etyma, since Hattori assumes that the diphthongizing vowel of middle bigrades is a mid vowel, *, and apparently never wavers from that view. (29)

Western OJ

proto-Japanese

Hattori

ökiB, ökös-

) pR *akedu ‘dragonfly’ (cf. OJ akidu) changed to *akdu, allowing aspiration, spirantization, and even loss of the velar consonant: an extreme example is Uke [nda] (< *akdu-wa, a hypocorism), from my own 1985 field notes. Apparently the situation is that, generally in the peripheral dialects, vowels neither lowered nor raised, so that (*)maru- ‘round’ < *mar- 24 (but cf. WOJ maro-, indicating lowering), and (*)ke ‘tree’ < *ky (but cf. WOJ kwi, indicating raising). On the other hand, vowels in the central dialects in effect switched heights, with vowels that are neither rounded nor front having a tendency to lower (* > *), and rounded and/or fronted vowels having a tendency to raise (*y > ?*ey > ?(*)iy; or *y > ?*we > ?(*)wi). While Ryukyuan is of the peripheral type in the sense that its vowels at first resisted raising, it eventually raised most of its non-low short vowels to high position, whether monophthongal or diphthongal in origin. Hattori (1976b:31) notes that Tottori uchiru ‘fall’ and ukiru ‘arise’ (NKD 3:630, 509), along with his previously mentioned Takeno, Kyoto ke ‘(general name for all vegetation)’, are likely non-central-dialect holdovers from pJ. My conjecture is that he sees the initial u of uchiru and ukiru as being vestiges of older (in this case, pre-pJ) **ü = *[], a vowel that has the same ‘feminine’ vowel-harmonic value as pJ *. Thus, apparently he meant to I differ with Hattori in viewing the phonetics of this form as [m(w)t], but that is a matter of phonetic detail. 24 An analysis of the type favored by Hattori would no doubt take advantage of the connection already posited to OJ moro-, and posit pre-pJ **märü- = *[mr], presumably—for the WOJ lineage—with a lowering in both syllables for the member of the doublet being considered here, and a lowering of only one for the other member of the pair: **[mr] > *[mr] > WOJ *[mor] = moro-. The peripheral dialects would have the doublet maru- < *marü- = *[mar] :: moro- < *mörö- = *[mr]. 23

LEON A. SERAFIM

94

reconstruct pre-pJ **ütäi- = **[ti] ‘fall’, and pre-pJ **ükäi- = **[ki] ‘arise’. Since he claims (Hattori 1976a:73f) that pre-pJ **ä (merely) rewrites as pJ *, and that pre-pJ **ü = **[] lowers to pJ *, which becomes OJ -o (o2) = [], then the conjectured Hattori pre-pJ reconstructions can have become pJ *ti- and *ki-, respectively, through lowering. If dialects like Tottori should turn out to furnish otherwise unexplainable u where WOJ-derived dialects have o (o2), and thereby be seen as furnishing evidence for pre-pJ **ü, then ‘pre-pJ’ will simply become an earlier level of pJ. If Ryukyuan is to provide independent corroboration, or be shown not to, then those vowels must appear after consonants that will preserve an indication of the original height (or in the reformulation suggested above, tenseness), such as, e.g., pR *tü :: *tö > Shuri ci :: tu. As far as I am aware, Hattori lowers monophthongal (*)*ü to *, but refrains from lowering diphthongal *üi to *i. To reiterate, he does not treat middle bigrades as having *üi, so in this respect his hypothesis clearly differs from F&W. 8.

Formation of verb forms Serafim (2007) has repackaged and retailored pre-existing hypotheses about the origin of the Conclusive, Adnominal, and Realis forms of both pJ and pR. In the process, it rejects a theory put forward by Thorpe (1983), that pR Conclusives and Adnominals 25 —which he correctly (cf., e.g., Hattori 1977) derives from an Infinitive form plus the stative verb (in his reconstruction) *wö— are direct descendants of an identical sequence of verb forms in pJ, as reconstructed at least in part in Ohno 1953 and 1955b. OJ bigrade verbs have the patterns in (30) and (31), deriving from pre-OJ (32) and (33). (30)

Upper Bigrade: ‘arise’ Irrealis ok –wi Infinitive ok –wi Conclusive ok u Adnominal ok uru Realis ok ure Imperative ok –wi (yo)

(31)

Lower Bigrade: ‘raise’ Irrealis ag –e Infinitive ag –e Conclusive ag u Adnominal ag uru Realis ag ure Imperative ag –e (yo)

(32)

Upper Bigrade: ‘arise’ Irrealis *ök ï Infinitive *ök ï Conclusive *ök ï #u Adnominal *ök ï # uru Realis *ök ï # urë Imperative *ök ï –(yö)

(33)

Lower Bigrade Irrealis *ag ë Infinitive *ag ë Conclusive *ag ë #u Adnominal *ag ë #uru Realis *ag ë #ure Imperative *ag ë (yö)

The above is worked out in terms of the mainland Japanese lineage, since 25

He does not specifically treat Realis forms as such, though he does treat them as a category that he calls the Imperfect Participle (1983:249ff).


95

the Ryukyuan and similar lineages would have merged their middle bigrades, such as *ököy- (> *ökë-), and their upper bigrades, such as *agay- (> *agë-), into just one lower bigrade class. Only the handful of original upper bigrades, such as *suguy- / *sogoy- (> *sugï- / *sugi-) would have been left in a separate category. I am currently considering the idea that the original verb from which the extensions descend is pJ *wor-, not the *ur- or *wr- posited in Serafim (2007). The reason is that it seems likely that to the right of the verb stem, in auxiliary or suffixal position, *wor- underwent raising to *wur-, the equivalent of OJ u (assuming that final *r can be merged successfully with the preceding material, or dropped, rather than cause a diphthongization). Further, there is an OJ form wor‘exist (animate)’, and it can easily be derived from the posited pJ root through neutralization of the assumed pJ *wo :: *wö opposition, to OJ undifferentiated wo. A sticking point: There would be no other known word in Ryukyuan (e.g., Sr wu‘exist (animate)’) where the initial mora comes from *wu, which should yield u, but here yields wu. The form pR xwû- would be required, by the rule stating that verb-root-final type-A vowels raise in Ryukyuan but not in mainland Japanese. An escape clause would be necessary. The modern Ryukyuan forms are indeed based on the grammaticalization of an identical sequence, but at a much later time. An equivalent to the Conclusive forms given above would be *oke#womu and *age#womu; see below. There already existed in earlier Ryukyuan—as we know from the Omoro sôshi of as early as the beginning of the 16th century—forms derived from the old grammaticalized forms of pJ, the equivalents of *ageru and *okeru. Actually attested forms include = agru ‘raise’ (N&H 338), = akru ‘dawn’ (N&H 338) and = urru ‘descend’ (N&H 360), a verb of the same type as the above-used *okeru example. (So-called ichidanka, or loss of the *e ~ *u alternation in the paradigm, happened earlier in Ryukyuan than in mainland Japanese.) Forms such as these were partially replaced by the newly grammaticalizing progressive forms of the type *oke#womu and *age#womu mentioned above, which in turn lost most of their progressive functions. The important point here is that grammaticalization of progressives into habituals and into presents is quite common in the world’s languages, that is to say, that the existence of a grammaticalized progressive in two related languages carries a relatively low degree of arbitrariness. To put it another way, because such a grammaticalization is highly likely to happen independently, it does not make for particularly good materials for comparative reconstruction. Lookalike forms can too easily be made up independently of each other, yielding false cognate strings. I was attracted to this topic by work with my co-author Rumiko Shinzato on early Japanese focus systems and their development. Forms such as OJ …(u)ram-u ‘probably’ and …(a)m-u ‘probably; you ought to; I think I’ll…’ bear a striking resemblance to the Okinawan forms …(y)ura and …(r)a, with the same meanings. However, the evidence points to independent development.

LEON A. SERAFIM

96 9.

The so-called kakari-musubi focus and its subtypes Shinzato and Serafim have examined the major kakari-musubi constructions, with an eye to syntactic comparison, to form hypotheses concerning the existence of specific focus constructions within pJ, and their development. We found (1) congruences: (a) Ok. du :: OJ so/ MJ zo with main verb in Adnominal form; (b) ga :: ka (wh-question particles) with main verb in the Suppositional form, which in turn is in the Adnominal form in Japanese (both, Serafim & Shinzato 2000); *su ~ *so :: koso (Serafim & Shinzato 2005), with main verb in the Realis form; (2) lack of congruence (as far as kakari-musubi goes): *ye :: ya (Shinzato & Serafim 2003); and (3) have not yet considered: (a) namo-type kakari-musubi; (b) utilization of focus with markers equivalent to NJ wa ‘(topic/contrast particle)’ or mo ‘even’. 9.1

du :: so/zo I will first look at du :: so/zo (Serafim & Shinzato 2000, but especially Shinzato & Serafim 2003). The construction in OJ and Classical Japanese basically works with a kakari particle attached to a noun or noun-like structure, and requires that the sentence end in the Adnominal, rather than the otherwise expected Conclusive, form. An example from Okinawan is as follows: (34)

tui nu du načoru bird NOM KP chirp.ADN ‘It is the bird that is chirping.’

And a Classical Japanese example: (35)

ima zo kuyasiki now KP sore(Adnom.) ‘It is now that (my heart) is sore.’

Both OJ and Okinawan have the capacity to distinguish Conclusives from Adnominals, and both utilize the kakari particle to trigger the musubi (ending) with an Adnominal. Our argument was that the constructions were not very likely to have developed independently, and that therefore they were probably both descended from an identical construction in pJ. 9.2

ga :: ka Next, consider ga :: ka. Serafim & Shinzato (2000) examine the kakarimusubi construction in Okinawan where the question particle ga functions as a kakari particle for an inferential suffix -(r)a.


(36)

97

are nama šumuči ga yudora. KP read-PROG (Inference) He now book ‘I wonder if he is reading a book now.’

The early Heian text Taketori monogatari has a very high occurrence of usage of kakari particle ka with the inferential ending -(a)m-, in the Adnominal form, such as the following: (37)

Ikade ka naka ni otori masari pa sir-amu. among inferior superior TOP know-Inference how KP ‘How would I know who is inferior or superior among (them)?’

We argued that not only did the basic pattern of ka … Adnominal form exist in pJ, but so did the special form with the inferential. This was shorn of its morphology in Okinawan, and independently developed into the inferential-type kakarimusubi. 9.3

*ye :: ya Third, there is the pair *ye :: ya. Briefly, Shinzato & Serafim (2003) find that the use of the Okinawan suffix -(y)i for yes/no questions is straightforwardly a sentence-ending type construction, and does not bear a resemblance to the OJ or Classical Japanese use in kakari-musubi. Thus we conclude that the construction developed only in the dialect leading to OJ. Here is an example of the use of -(y)i in Okinawan: (38) ari- -N čum-i? come(CONCL) Q ‘Is he also coming?’ 9.4

Realis-type KM, with Sr *su ~ *so and OJ koso Finally, Realis-type KM, with Sr *su ~ *so, and OJ koso. What Serafim & Shinzato (2005) argue is that the now-obsolete construction of *su … Realis in Okinawan, again, is genetically related to the OJ koso … Realis construction. But there is a hitch. First, we argue that OJ koso is actually a compound in origin, of *kö- ‘this’ and *so ‘one, thing’, the same *so seen earlier in this chapter (originally propounded in outline form in Thorpe 1983). The reason for the apparent discrepancy in vowel color is that Arisaka’s First Law disallows -o (-o2) and -wo (-o1) in the same word. We hypothesize that at the time of the divergence of Ryukyuan and the dialect leading to OJ, the form still had a word boundary in it: *kö#so, thus not requiring a harmonic vowel quality change. The loss of that boundary happened in the OJ lineage after the OJ-Ryukyuan divergence. Meanwhile, Ryukyuan excised the first of the two morphemes in *kö#so, leaving only *so, which, in context, sufficed to carry the construction. The modern form would have been Sr si (Common Okinawan ši) had it survived. It is embedded in a

98

LEON A. SERAFIM

very few lexicalizations in the modern language, but it is plentiful in the earliest text, the Omoro sôshi.26 10.

Conclusion I hope that the presentation in this chapter of historical topics and problems, and their elucidation of actual and possible solutions when applied both to mainland Japanese dialects of any period, and to pJ, has shown what sorts of possibilities lie in store for those who wish to delve a little more deeply into the Ryukyuan languages and dialects. To conclude I will discuss some implications of the relationship of Ryukyuan to mainland varieties of Japanese. First of all, the phonological correspondences, especially among the vowels, between Ryukyuan and OJ presented in §5 show clearly that it is not possible to derive one from the other. This means that pJ, the ancestor of later varieties of Japanese, cannot be the OJ language of the Nara period, but must predate it: Otherwise the complex vowel correspondences (obscured by the massive raisings and frontings of vowels in Ryukyuan), cannot be explained. See further Hattori (1976b) and Serafim (1977b) on this point. This raises two further questions: When and from where in Japan proper did Japanese enter the Ryukyus? as well as the logically prior question: Did Japanese enter the Ryukyus from Japan proper, or did it spread from the Ryukyus to Japan proper? On the latter, both positions have been taken, but all current evidence points to Japanese entering the Ryukyus from Japan proper. It is widely thought that Japanese language entered Japan from the Korean peninsula together with the carriers of the Yayoi culture, who also brought with them the tool kit for wet-rice agriculture (for examples see Hanihara (1991), who believes that in the thousand years between about 300 BC and the beginning of history in about 700 AD, one million people crossed the strait to Japan). Sometime after the beginning of that period, what was to become Ryukyuan must have entered the Ryukyuan archipelago. It can even be after 700, just as long as it is from a dialect that split from Central (Nara) Japanese before that date, an easy proposition. There is a possibility that Japanese language came to the Ryukyus from somewhere in northeast Kyushu or extreme western Honshu (the general area of the Kanmon Strait, dividing Honshu and Kyushu). At least, I have suggested as much in a recent paper (Serafim 2003). The reasons I give are as follows: The general opinion has been that Japanese must have entered the Ryukyus from southern (see, e.g., Uemura 1977) or western Kyushu (see, e.g., Asato & Doi 1999). Certainly, either view has geographical proximity going for it. What I noted, however, was that Ryukyuan has a nominalization marker (Common Okinawan ši HHLL > HHHL > HHHH LHHH > LLHH > LLLH > LLLL

A dictionary from 1081 which marks pitch.

THE PROTO-ACCENTUAL SYSTEM OF JAPANESE

105

This means, for example, that the tonal pattern HLLL tends to change to HHLL and subsequently to HHHL as in (1a), whereas change in the opposite direction (i.e., HHHL>HHLL>HLLL) is rare. Similarly, the pattern LHHH tends to shift to LLHH and then to LLLH as in (1b), rather than in the opposite direction. This tendency of rightward spreading of both H-tone and L-tone, which is a type of progressive assimilation observed throughout Japanese dialects, also provides support for the general observation made by tonologists in African languages that left to right tone spreading is more prevalent than right to left (Hyman and Schuh 1974). 2.2

Dissimilation at the edge of a tonal phrase The opposite of assimilatory change may be the insertion of a new tone at the left edge of a phrase, i.e., in phrase-initial positions. Thus, as illustrated in (2), if there is a sequence of L- tones as in LLL… at the left edge of an accentual phrase, the next stage is likely to be HLL…: on the other hand, the sequence of HHH…at the phrase-initial positions tends to change into LHH…: (2)

a. b.

[ LLL … > [ HLL … [ HHH … > [ LHH …

Notice that in this case, the tone inserted at the left edge of a phrase is always one of opposite value to the adjacent tone, which means that dissimilatory change tends to take place phrase-initially. 2.3

Systematicity of a tonal change and nonrecoverability of tonal distinctions The next generalization is related to the systematicity of a tonal change. When any kind of tonal change (either assimilatory or dissimilatory) takes place, it does not merely let one tonal pattern change all the way to the other without any effect on different tonal patterns in the same system. For example, if a tonal pattern HLLL of a certain tonal system changes into HHLL at a certain stage, and if the same system already has the HHLL pattern, the latter tonal pattern (HHLL) is most likely to move, by the influence of the similar rightward tonal shift of H tones, resulting in HHHL as shown in (3): (3)

a. b.

HLLL > HHLL HHLL > HHHL

likely associated with

However, it is not permissible for the tonal pattern HLLL to go all the way to HHHL and then to HHHH, bypassing all other tonal patterns in the same system. Thus, the scenario in (4) is highly unlikely. In other words, any shift of a tonal pattern generally affects other tonal patterns in the same system, usually by inducing a parallel tone shift. This means that accent change proceeds systematically, operating to conserve former

AKIKO MATSUMORI

106 (4)

Unlikely association of tonal changes a. b. c.

HLLL > HHLL > HHHL > HHHH (same as (1.a)) associated with HHLL ---- (no change) ---- HHLL HHHL ---- (no change) ---- HHHL

paradigmatic relations between tonal patterns in the same system. This tendency of systemic tonal change was one of the crucial points of the hypothesis concerning a new type of pMJ in Matsumori (1993, 1997). In addition to the paradigmatic conservation of tonal distinctions described above, other types of change can be seen as a type of non-conservation of former distinctions, namely when tonal patterns which used to function distinctively merge into one pattern. For example, if the two tonal patterns in (3) merge as a result of a rightward shift of H tone in (3.a) but not in (b), the outcome is one tonal pattern, in this case HHLL, as shown in (5): (5)

Merger of tonal patterns HLLL HHLL

HHLL

When previously different tonal patterns undergo mergers of this type, the groups of words that used to belong to the separate tonal classes are pronounced with the same tonal pattern. It is important to note here that once such a merger takes place, the old tonal distinction would then not be recovered. This ‘nonrecoverability of former distinctions’ in relation to mergers has been another important basis for Japanese accentology, working to constrain the diachronic analyses in Japanese accent. As a result of this nonrecoverability of former tonal contrasts, there is a general tendency in the history of Japanese tonal systems for the number of tonal patterns to decrease, rather than to increase, in the course of the historical development of accent systems unless there is a strong influence, such as language contact or standardization, to motivate the emergence of new tonal patterns. 3.

Examination of Shikoku dialects based on general directions of tonal change 3.1 Traditional assumption of Japanese dialectology Keeping the tendencies summarized in §2 in mind, this chapter will examine some accent systems of certain current dialects in Shikoku. The major island of Shikoku contains a large group of dialects generally classified into ‘Keihan (Kyoto-Osaka)-type’ or ‘Central-type’ accent systems. However, more


107

specifically, the dialects in and around Shikoku are known to display some distinctive patterns of tonal mergers, which differ significantly from other major Keihan-type dialects. In order to distinguish them from general Keihan-type systems, they are often classified as ‘Sanuki-type’ or ‘Manabe-type’ systems, 4 from the names of the areas in which these specific tonal systems are typically observed. Until now, the general assumption of Japanese linguists regarding the proto-system of Japanese accent was that it was similar (sometimes identical) to that of the 11th-century Kyoto Japanese recorded in the Ruiju myôgishô (RM), in which nouns display the accentual system shown in (6). (6)

The tonal system in the Ruiju myôgishô5 One-syllable words (µ ga) 1.1 H(H) 1.2 H(L) 1.3 L(L)

two-syllable words (µµ ga) 2.1 HH(H) 2.2 HL(L) 2.3 LL(L) 2.4 LH(H) 2.5 LH(L)

three-syllable words (µµµ ga) 3.1 HHH 3.2 HHL 3.3 HLL 3.4 LLL 3.5 LLH 3.6 LHH 3.7 LHL

This hypothesis, namely that the proto-accentual system of Japanese is the same or similar to the 11th-century Kyoto dialect shown in (6), is now considered the least controversial idea among dialectologists in Japan, and is more or less established and accepted. This assumption may be due to the fact that among the systems presently known, this dialect (i.e., the RM system) retains the maximal number of tonal 4

Classifications of ‘Keihan-type’, ‘Sanuki-type’, or ‘Manabe-type’ are intended to facilitate the comparison and investigation of the genealogical relationships of dialects. In Japanese accentology, following the pioneering work of Tokugawa (1962), the standard of such classification has been based on merger patterns of two-mora nouns. As shown in (6), the 11th-century Kyoto system had 5 tonal contrasts for two-mora nouns. The only known dialect that still retains all of these 5 contrasts is the dialect of Ibuki-jima, which is shown in (7) in this chapter. Other Japanese dialects have all undergone mergers of some of these 5 contrasts. For example, the present-day Kyoto system has undergone a merger of ‘2.1 / 2.2, 2.3 / 2.4 / 2.5’, while the present-day Tokyo system that of ‘2.1 / 2.2, 2.3 / 2.4, 2.5’, and so on. The so-called ‘Sanuki-type’ and ‘Manabe-type’ dialects are both known to show particularly distinctive types of mergers of two-mora words, i.e., ‘2.1, 2.3 / 2.2 / 2.4 / 2.5’, and ‘2.1, 2.5 / 2.2 / 2.3 / 2.4’, respectively. 5 In (6), µ stands for a mora, and the parenthesized H or L indicate the assumed tones of a suffix attached to these nouns. For example, a class 2.1 noun hásí (edge) with the nominative case marker ga is shown as the HH(H) tone pattern, as in hásígá, while a class 2.2 noun hásì (bridge) with the same case marker appears as HL(L) as in hásìgà.

108

AKIKO MATSUMORI

contrasts (i.e., the system in which one syllable nouns have three patterns, two syllable nouns five, three syllable nouns seven). No system with so many tonal contrasts has been reported in any other dialect in Japan (both current dialects and in the literature). Considering the fact that the number of tonal patterns generally decreases with time, as a result of mergers of tonal patterns, as the systems change into others (usually more accentual ones), some scholars have considered that this system (i.e., the RM system), retaining the maximal numbers of tonal contrasts, must be the oldest, and that this, therefore, must be the proto-system. However, for as long as written records have existed, tonal patterns have changed; thus, it is only natural to assume that the process of tonal change was active before the introduction of written records. Thus, it is likely that the 11thcentury Kyoto system of RM can be the result of mergers of numbers of tonal patterns which the previous system may have had; i.e., the tonal system of the RM itself probably also developed from a tonal system with greater numbers of tonal contrasts from which other dialects separated. 3.2

Tonal classes of ‘Kindaichi-goi’ It is well-known that, based on the tonal systems of Japanese, as well as the regular correspondences of tonal patterns observed among present Japanese dialects, a vocabulary list known as the ‘Kindaichi-goi’ was developed in the 1950s,6 which was also related to a hypothesis on the type of tonal system and the number of tonal patterns included in Kindaichi’s concept of pJ. (Throughout this chapter, I will refer to this vocabulary list simply as the ‘Kindaichi-goi’.) Being a hypothetical basis for the analysis of vocabulary belonging to each of the tonal patterns supposed to have existed in pJ, the Kindaichi-goi has been used by Japanese dialectologists as a useful tool for descriptive studies of dialects. After the Kindaichi-goi and this hypothesis were developed, much research was conducted based on this; consequently, various new types of accentual systems were reported, especially during the 1950s and 60s, making possible subsequent elaborate comparisons between a number of dialects. As a result, this growth of descriptive data using the Kindaichi-goi has also given us an opportunity to present some questions as to the validity of the general assumption that the accentual system recorded in the RM is indeed the pJ system of Japanese. 3.3

Examination of the general assumption based on Shikoku dialects As stated, it is generally accepted as a basic assumption, at least among many dialectologists in Japan, that the proto-system of the accent of Japanese dialects is that of the RM, shown in (6), which means that all the present dialects in Japan descended from the system similar to, or the same as, the 11th-century Kyoto dialect. However, Matsumori (1993) proposed that, considering the natural 6 The versions of Kindaichi-goi most commonly used among Japanese dialectologists are those of Kindaichi 1974, and Kindaichi & Wada 1980.


109

tendency of tonal change illustrated in §2 of this chapter, we might reasonably entertain doubts about the validity of this assumption. The crucial data for this argument was provided mainly by the Shikoku dialects, among which the most significant one may be the dialect of Ibuki-jima, a small island located in the Inland Sea close to Kan’onji city in Kagawa prefecture. Based on the data given by Uwano (1985), the tonal patterns of the Kindaichi-goi in this dialect are shown in (7): (7)

Present-day Ibuki-jima accent One-mora nouns (μ ga) 1.1 H(H) 1.2 H(L) 1.3 L(H)

two-mora nouns (μμ ga) 2.1 HH(H) 2.2 HH(L) 2.3 HH(M) 2.4 LL(H) 2.5 LH(L)

three-mora nouns (μμμ ga) 3.1 HHH(H) 3.2 HHH(L) (LHH(L) ) 3.3 HLL(L) 3.4 HHM(M) 3.5 HHM(M) or HLL(L) 3.6 LLL(H) 3.7 LHH(L)

Among the dialects presently spoken, this dialect is recognized to retain the maximum number of tonal distinctions. In fact, it is known as the sole dialect in present-day Japan which retains all the five distinctive categories of two-syllable nouns (i.e., class 2.1 - 2.5 in (7)) annotated in the RM. What is worthy of note is that in this dialect, Mid-tone (hereafter M-tone) is contrastive with H or L tones, as illustrated in (8). These examples suggest that, for this dialect, M-tone, in addition to H-tone or L-tone, is phonologically necessary. In fact, the Ibuki-jima dialect is the only known dialect in Japan in which M-tone is contrastive, while other dialects can all be phonologically described as using only two levels of tonal contrast: H-tone and L-tone. (8)

Tonal contrasts observed in Ibuki-jima a. Two-mora words háná ‘nose’ (HH) hána ‘flower’ (HM) tórí ‘bird’ (HH) yáma ‘mountain’ (HM)

háná gá (HHH) háná ga (HHM) tórí gá (HHH) yámá ga (HHM)

b. Three-mora words sákáná ‘fish’ (HHH) átáma ‘head’ (HHM) músúmè ‘daughter’ (HHL)

sákáná gá (HHHH) átáma ga (HHMM) músúmè gà (HHLL)

AKIKO MATSUMORI

110

For the purpose of effective comparison, let us focus on the isolated forms of three-mora nouns of this dialect and compare them with the RM system. If you simply compare the 11th-century Kyoto system given in (6) with that of presentday Ibuki-jima given in (7), you will notice some problems in the assumption that a system similar to that of 11th-century Kyoto is the proto-system, from which the present day Ibuki-jima dialect developed. (9)

Three-mora nouns in the RM (6) and present-day Ibuki-jima (7) RM 3.1 HHH 3.2 HHL 3.3 HLL 3.4 LLL 3.5 LLH 3.6 LHH 3.7 LHL

Ibuki-jima 3.1 HHH 3.2 HHL 3.3 HLL 3.4 HHM 3.5 HHM (HLL) 3.6 LHH 3.7 LHL

sample words sakana ‘fish’ musume ‘daughter’ tikara ‘strength’ atama ‘head’ inoti (life) usagi (rabbit) itigo (strawberry)

The main question is about the development of the tonal pattern HHM in the Ibuki-jima system, written in bold letters in (9), which is observed for the Kindaichi-goi’s class 3.4. and 3.5. nouns. Drawing on the natural tendency of tonal change illustrated in §2, there seems to be no way to explain why the tonal pattern HHM in Ibuki-jima developed from LLL or LLH in the 11th-century Kyoto system. Even if we assumed, for example, that a new H-tone is added at the left edge and, as a result, a completely new tonal pattern HHM developed, in such a process as shown in (10), it would be difficult to explain how this process was possible without causing drastic paradigmatic change to any of the other tonal patterns in the same system. (10)

Rejected derivation of Ibuki-jima’s HHM in class 3.4 and 3.5 nouns RM 3.4 LLL 3.5 LLH

> >

HLL HLH

present-day Ibuki-jima > HHL > HHM (atama) > HHL > HHM (inoti)

Whether or not the last change (HHL >HHM) in the process in (10) is a likely or natural change is an empirical question yet to be solved, but even if it were, we would still have trouble with the hypothesis that the present-day Ibuki-jima developed from the same system as the 11th-century Kyoto dialect.That is, we cannot explain why the other tonal patterns (e.g., HHL for class 3.2, or HLL for class 3.3) were left completely intact, while the tonal patterns for class 3.4 or 3.5 underwent the dramatic change described in (10).


111

Another mystery is presented by other Shikoku dialects. In the surrounding areas near Ibuki-jima, in Shikoku’s main island, in the vicinity of the city of Kan’onji, to which Ibuki-jima belongs, the HHM tonal pattern appears also in class 3.1, in addition to class 3.4 and 3.5. Among these dialects, the dialect of Omi in the town of Takase, located about 10 km from the central area of the city of Kan’onji, is a typical example. In (11), based on Sato’s (1985) data for Omi, I rearranged the tonal patterns according to the Kindaichi-goi categorization: (11)

Present-day Omi accent (based on Sato 1985) One-mora nouns (μ ga) 1.1 H(L) 1.2 H(L) 1.3 L(L)

two-mora nouns (μμ ga) 2.1 HH(M) 2.2 HL(L) 2.3 HH(M) 2.4 LH(H) 2.5 LH(L)

three-mora nouns (μμμ ga) 3.1 HHM(M) 3.2 LHL(L) (HLL(L)) 3.3 HLL(L) 3.4 HHM(M) 3.5 HHM(M) or HLL(L) 3.6 LHH(H) 3.7 LHL(L)

Again, the comparison of three-mora nouns in this system with those of the RM system, given in (6), is shown in (12): (12)

Three-mora nouns in the RM and present-day Omi RM 3.1 HHH 3.2 HHL 3.3 HLL 3.4 LLL 3.5 LLH 3.6 LHH 3.7 LHL

Omi 3.1 HHM 3.2 LHL 3.3 HLL 3.4 HHM 3.5 HHM or HLL 3.6 LHH 3.7 LHL

sample words sakana ‘fish’ azuki ‘red bean’ tikara ‘strength’ atama ‘head namida ‘teardrop’ suzume ‘sparrow’ itigo ‘strawberry’

For classes 3.4 and 3.5, we have the same problem as with Ibuki-jima; that is, we cannot explain how the tonal pattern HHM (including an M-tone) developed from the RM’s LLL (class 3.4) or LLH (class 3.5), with no effect on other patterns. Moreover, this dialect has an additional problem: i.e., we do not know how the HHM pattern for class 3.1 could have developed from the 11thcentury Kyoto pattern HHH shown in (6). Similarly, as shown in (11), the class 2.1 nouns in the Omi system also appear with the HH(M) tonal pattern, as in háná ga (HHM) (‘nose’ + nominative marker ga), while the 11th-century Kyoto system, shown in (6), displays HH(H) for the same class 2.1 nouns, as in háná gá (HHH).

AKIKO MATSUMORI

112

Adding to these mysteries, there are other dialects spoken in fairly small islands scattered in the Inland Sea (the area north of Ibuki-jima), in which class 2.1 nouns appear with HH(L) or LH(L), and class 3.1 nouns with HHL or LHL (when ga is attached, HHL(L) or LHL(L)). Let me give an example from the Shishi-jima dialect, another dialect spoken on an island in the Inland Sea, located about 20 km northeast of Ibuki-jima: (13)

Present-day Shishi-jima accent One-mora nouns (μ ga) 1.1 H(L) 1.2 H(L) 1.3 L(L)

two-mora nouns (μμ ga) 2.1 LH(L) 2.2 HL(L) 2.3 HH(L) 2.4 LL(L) 2.5 LH(L)

three-mora nouns (μμμ ga) 3.1 LHL(L) 3.2 LHL(L) 3.3 HLL(L) 3.4 LHL(L) 3.5 LHL(L) or HLL(L) 3.6 LLL(L) 3.7 LHL(L)

Notice that we can observe the tonal patterns H(L), LH(L), and LHL(L) in class 1 nouns (i.e., the nouns belonging to class 1.1, 2.1, and 3.1 of the Kindaichi-goi) here. Thus, the words ‘nose’ (class 2.1) and ‘fish’ (class 3.1), for example, appear with LH(L) and LHL(L) tonal patterns in this dialect, as in hàná gà and sàkánà gà, respectively. Again, we have difficulty explaining how these tonal patterns developed from the patterns in the RM system, in which the class 1 nouns are H(H), HH(H), and HHH, as shown in (6). The comparison of three-mora nouns in this system with those of the RM system, (14), clearly illustrates the problem: (14)

Three-mora nouns in the RM and present-day Shishi-jima RM 3.1 HHH 3.2 HHL 3.3 HLL 3.4 LLL 3.5 LLH 3.6 LHH 3.7 LHL

Shishi-jima 3.1 LHL 3.2 LHL 3.3 HLL 3.4 LHL 3.5 LHL 3.6 LLL 3.7 LHL

sample words sakana ‘fish’ musume ‘daughter’ tikara ‘strength’ atama ‘head’ inoti ‘life’ usagi ‘rabbit’ itigo ‘strawberry’

In order to explain the change of class 3.1 from the RM's HHH to Shishi-jima’s LHL, we might have to assume the leftward shift of H-tone (i.e., HHH > HHL), which is an obvious contradiction to the general rightward trend of tonal shift delineated in §2 in this chapter.


113

3.4

Proto-accentual system of mainland Japanese dialects: going back one stage earlier than the Ruiju myôgishô system In order to solve the problems posed above, Matsumori (1993) proposed a proto-system one stage prior to the RM system, and suggested it to be the protosystem of mainland dialects. The starting point for the hypothesis is the development of tonal patterns of class 1 nouns (i.e., nouns belonging to class 1.1, 2.1, and 3.1) in Shikoku dialects. Assuming that the HHM pattern observed in Omi and certain other dialects in Shikoku (i.e., Sanuki-type dialects) originate from HLH, by a process called ‘downstep’ (HLH > HHM), I set HLH as the proto-form of class 1 nouns for pMJ. The development of class 1 nouns’ HLH proto-form to the corresponding tonal patterns of the present Shikoku dialects, as well as to the system of 11th-century Kyoto is illustrated below: (15)

Change of class 1 nouns to the present-day Shikoku dialects and the RM HHH (Ruiju myôgishô and Ibuki-jima) *HLH

HHM (Omi and other major Sanuki-type dialects) HLL > HHL > LHL (Shishi-jima)

The development HLH>HHH for class 1 nouns assumed for the RM and Ibukijima is achieved by another common type of change called plateauing (*HLH>HHH). On the other hand, Shishi-jima’s LHL can be developed by the natural rightward tone spreading of both L-tone and H-tone as in *HLH>HLL>HHL, as well as the subsequent dissimilation at the left edge of a tonal phrase (HHL>LHL). Based on this new idea about the tonal development of class 1 nouns, Matsumori (1993, 1997, 1999) proposed a system one stage earlier than the RM, the tonal system of pMJ, as shown in (16), in which the parenthesized H or L indicate the tones of regular affixes: (16)

Accent system of pMJ (based on Matsumori 1993, 1999) One-syllable words 1.1 HL(H) 1.2 L(L) 1.3 H(H)

two-syllable words 2.1 HL(H) 2.2 LL 2.3 LH 2.4 HH 2.5 HL

three-syllable words 3.1 HLH 3.2 HLL 3.3 LLL 3.4 LLH 3.5 LHL 3.6 HHH 3.7 HHL

AKIKO MATSUMORI

114

All the systems dealt with in this section can be naturally derived from (16), as is illustrated by the processes shown in (17) to (19), focusing on the three-mora nouns. (17)

The development of the RM system from pMJ pMJ 3.1 3.2 3.3 3.4 3.5 3.6 3.7

(18)

HLH HLL LLL LLH LHL HHH HHL

RM HHH HHL HLL LLL LLH LHH LHL

Development of the Ibuki-jima system pMJ 3.1 HLH 3.2 HLL 3.3 LLL 3.4 LLH 3.5 LHL 3.6 HHH 3.7 HHL

(19)

> > > > > > >

> > HHL > > > HLH > > HLH > > >

present-day Ibuki-jima HHH LHL, LHH(L) HLL HHM HHM LLH, LLL(H) LHL, LHH(L)

Development of the Omi system pMJ 3.1 HLH 3.2 HLL 3.3 LLL 3.4 LLH 3.5 LHL 3.6 HHH 3.7 HHL

> > HHL > > > HLH > > HLH > > >

present-day Omi HHM LHL HLL HHM HHM LHH LHL

A proto-system is a result of attempts to relate the forms in attested languages. It is not only the ultimate proof of genetic relationships, but a tool to understand how the related languages have evolved, and to explain why they differ and to what extent the contemporary languages resemble each other. Considering this, we obviously have to abandon the previously embraced hypothesis that the tonal system of the RM is the proto-system of Japanese


115

dialects, or even for all the mainland dialects including those in Shikoku. Although, due to its fairly abstract nature, the hypothesis proposed in (16) has not met with general acceptance, one thing we can be certain of at this stage is that we have to go beyond the written record of the RM in order to search for a better, more persuasive proto-system. 4.

Some related issues concerning the proto-mainland Japanese system Following the traditional assumption about the RM system, the accentual system of the current Tokyo dialect (i.e., Standard Japanese) is also supposed to have developed from it. Thus, the LHH pattern for class 1 nouns in the presentday Tokyo system, which is shown in (20), is generally explained to have developed from the HHH of the 11th-century Kyoto system by initial lowering, as illustrated in (21): (20)

(21)

Tonal patterns of class 1 nouns in present-day Tokyo class 2.1

LH(H)

mìtí (road), mìtí gá; mùsí (insect), mùsí gá; kàzé (wind), kàzé gá

class 3.1

LHH(H)

sàkáná (fish), sàkáná gá; kùrúmá (wheel), kùrúmá gá; sàkúrá (cherry blossom), sàkúrá gá

Standard assumption about the development of tonal patterns for class 1 nouns in Tokyo RM HHH mítí gá sákáná

>

present-day Tokyo LHH mìtí gá sàkáná

In contrast, as was shown in (16), the present analysis has proposed that these tonal patterns for class 1 nouns used to be HLH in the pMJ system, with a pitch drop from H to L tone within the tonal pattern. Therefore, this chapter holds that Tokyo dialect, as well as similar dialects around Tokyo, all developed from (16), in which class 1 nouns underwent the following process of change: (22)

My hypothesis about the development of class 1 nouns in Tokyo pMJ *HLH *mítì gá *sákàná

> HHH > mítí gá sákáná

present-day Tokyo LHH mìtí gá sàkáná

AKIKO MATSUMORI

116

This chapter contends that the data supporting this analysis have been found in certain remote areas in the Kanto district in which the Tokyo-type system is observed. For example, in an isolated village amidst the steep mountains in Yamanashi prefecture, we find the Narada dialect, which has preserved particularly distinctive tonal patterns which are completely different from any other dialects in eastern Honshu. In this dialect, the patterns for class 1 nouns such as kaze (wind), or sakura (cherry blossom) exhibit a sharp pitch drop from H to L tones, as in kázè, kázè gà (HL, HLL), or sákùrà, sákùrà gà (HLL, HLLL). Similarly, in an eastern area of Saitama prefecture, located in the suburbs of Tokyo in the same Kanto district, a unique type of accentual systems has been reported, among which the one in Hasuda city (Uwano 1977) is best-known. Throughout the dialects in the area around the city of Hasuda, class 1 nouns also exhibit a pattern of pitch drop from H to L tones, as in HL, HLL. This is illustrated most clearly by Yoshida’s (1992) description of the accent system of Kuki city in Saitama prefecture. (23)

The accent patterns of Kuki, Saitama prefecture, based on Yoshida (1992) class 2.1 class 2.2 class 2.3 class 2.4 class 2.5

HL, HHL(LHL) HL, HLL HL, HLL LH, LHL LH, LHL

kámà, kámá gà ‘oven’; mízù, mízú gà ‘water’ hátà, hátà gà ‘flag’; hásì, hásì gà ‘bridge’ hánà, hánà gà ‘flower’; ásì, ásìgà ‘leg’ kàmá, kàmá gà ‘sickle’; hàsí, hàsí gà ‘chopstick’ àmé, àmé gà ‘rain’; hàrú, hàrú gà ‘spring’

Notice here that class 2.1 nouns such as kama (oven) or mizu (water) appear with the HL, HHL tonal pattern. Both of the dialects above, i.e., the dialects spoken in Narada and in the area around Hasuda city including Kuki, which are both located in peripheral areas of the Kanto district where Tokyo-type systems are most commonly observed, exhibit a pitch drop from H to L tones in their tonal patterns for class 1 nouns. This fact comes to assume a special significance when reconsidered together with the new hypothesis about pMJ given in (16). That is to say, these pitch drops may be a retention of the pMJ tonal pattern HLH for class 1 nouns shown in (15).7 Not only do we in these remote villages or isolated peripheral areas find data with important diachronic implications, but we can also observe phenomena relevant to our analysis in some exceptional patterns of certain synchronic processes in the Tokyo dialect (i.e., Standard Japanese). I would like to propose that the older generation’s patterns of compound accentuation in Tokyo Japanese 7

In order to clarify this situation, we need to collect more data in these isolated areas in the Kanto district. Although the data in these areas are of a special significance to the diachronic study of Japanese accent, the numbers of speakers of these dialects have recently been progressively declining, mainly because of the strong influence of Standard Japanese.


117

may provide important evidence relevant to diachronic considerations. That is, some of the old patterns of accents of nouns may be retained in the second members of compounds in Tokyo dialect. In Tokyo Japanese, the most unmarked, and thus productive pattern of compound accentuation is the attraction of accent at the syllable containing the antepenultimate mora (Kubozono 1999), which is illustrated by the following examples: (24)

Compound accentuation in Tokyo toori ‘passing’ + miti ‘road’ => toori’-miti ‘road to pass, route’ kogane ‘gold’ + musi ‘insect’ => kogane’-musi ‘maybeetle’ na’ma ‘raw’ + sakana ‘fish’ => nama-za’kana ‘raw fish’ yama’ ‘mountain’ + sakura ‘cherry tree’ => yama-za’kura ‘mountain cherry tree’ isi’ ‘stone’ + atama ‘head’ => isi-a’tama ‘stubborn person’

As shown in (24), regardless of whether the second elements are two-mora nouns (miti, musi) or three-mora nouns (sakana, sakura, ata’ma), they show accent (which is realized with a pitch drop from H to L) on the antepenultimate moras. On the other hand, the compounds where the second element is longer than four (sometimes three) moras tend to retain their original accent in compounding, see (25). (25) Compound nouns where the second member is longer than four (or three) moras in Tokyo sibu ‘persimmon astringency’ + uti’wa ‘a type of fan’ => sibu-uti’wa ‘type of fan varnished with persimmon astringency’ neziri ‘twisting’ + hati’maki ‘towel worn around the head’ => neziri-hati’maki ‘twisted towel worn aroud the head’ isoppu ‘Aesop’ + monoga’tari ‘story’ => isoppu-monoga’tari ‘Aesop’s Fables’ gakusyuu ‘study’ + sankoosyo’ ‘reference book’ => gakusyuu-sankoosyo’ ‘academic reference book’ tihoo ‘regional area’ + saibansyo’ ‘court’ => tihoo-saibansyo’ ‘rural court’ tentai ‘space object’ + booenkyoo ‘telescope’ => tentai-booenkyoo ‘astronomy telescope’ However, Akinaga (2001:22) reported that in the older generation’s patterns of compounding in Tokyo, class 3.1 nouns such as sakana and tatami have final accent when used as the second element of compounds, see (26).

118 (26)

AKIKO MATSUMORI

The older generation’s pattern of compound accentuation in Tokyo ni ‘boiling’ + sakana ‘fish’ => ni-zakana’ ‘boiled fish’ isi’ ‘stone’ + tatami ‘straw matress’ => isi-datami’ ‘cobbled paving’

Simplex forms of these words (sakana and tatami, etc.) are pronounced with an unaccented pattern LHH(H), as in sàkáná gá (‘fish’ + nominative case marker ga) or tàtámí gá (‘straw mat’ + nominative case marker ga). But when appearing as the second member of compounds, the accent on the final mora shows up, as in nì-zákáná gà (boiled fish + nominative case marker ga). Considering the fact that, as shown in (25), longer words tend to retain their accent patterns in the second members of compounds, it is likely that such exceptional patterns as shown in (26) are in fact residues of old compound accentuation patterns, which are no longer retained in their simplex forms. I suggest that older patterns of accentuation may have been retained up to recently in second members of compounds in Tokyo. In other words, even in Tokyo, the class 3.1 nouns in (26) may have had a pitch drop from H to L word-finally at some stage in the past, as in sàkáná gà (LHHL), suggesting that class 1 nouns used to have pitch drops from H to L even in the Tokyo dialect.8 So far, I have tried to show that typology-based generalizations can be a good heuristic device. We have seen that a new hypothesis based on typological observations about tonal change serves to shed new light on data which have been known for some time, but which have hitherto been considered merely deviant. In addition, the hypothesis in (16) can generate new topics for future field research on Japanese dialects. 5.

Analysis of Ryukyuan accent systems I have shown that typological observations about tonal change provide us with some standards for evaluation of a particular proto-system, thereby functioning to give us some control over reconstruction of the proto-system. In this section, we will deal with Ryukyuan, and show how the typological plausibility of accentual systems generates new topics for the study of accent in this area. Ryukyuan is spoken in the southernmost part of the Japanese archipelago. The dialects spoken in this area have been considered to be of great significance for the historical study of Japanese. Unlike the Tokyo type system (which is referred to as the n+1 type), in which the number of tonal patterns increases in 8 Given such data, the next thing we might ask is whether there are any other mainland dialects in which this type of pitch drop for class 1 nouns is retained in the second member of compounds? In fact, we can speculate that this type of old compound accentuation might be observed in other peripheral areas in the Kanto district. Evidence for this would lend strong support for the hypothesis given in (16), which reconstructs HLH, instead of the traditional assumption’s HHH, for class 1 nouns. This question requires further investigation.


119

proportion to the number of tone bearing units (TBU) in each word, the number of tonal patterns in most Ryukyuan dialects do not increase according to the number of TBU, but are limited in number. Thus, they are called N-type accent systems; in fact, a number of the systems in Ryukyuan either have only 2-patterns or 3patterns regardless of the length of word. As the N-type system is also observed in a wide area in the southwestern part of Kyushu island, Ryukyuan has been said to have a close geneological bond with the Kyushu dialects (Kindaichi 1960, Tokugawa 1981. Even a recent accent dictionary published by NHK in 1998 clearly states that “Ryukyuan is a geneological subgroup of Kyushu dialects, in terms of their accent features” (NHK Hôsô-bunka Kenkyûjo 1998:135). However, the present chapter argues that such superficial similarity does not confirm the actual geneological closeness. Matsumori (2001) pointed out that in contrast to Kyushu 2-type systems such as Kagoshima or Nagasaki, proto-Ryukyuan shows a unique pattern of tonal mergers. (For details, see Appendix in Matsumori 2001.) Seen through the Kindaichi-goi, the patterns of mergers for the southwestern Kyushu dialects are illustrated in (27). On the other hand, according to Matsumori (1998, 2000a, b, 2001), the pattern of mergers for pR is as shown in (28). We can see that there is a discrepancy between the type of words belonging to each tonal pattern in Ryukyuan systems and those in Kyushu. In fact, the unique merger patterns shown in (28) suggest that the Ryukyuan dialects are totally different from any mainland system. If Ryukyuan data had been available when Haruhiko Kindaichi developed his Kindaichi-goi, the outcome might have been completely different. Thus, until new types of evidence are found for Kyushu, we have to consider, at least at the present stage, that there is no close genealogical relationship between dialects in southwestern Kyushu and the Ryukyuan dialects. In fact, it is at least equally likely that they both developed their 2-type or 3-type accent systems independently. Why such is the case is another interesting question, for which there is no answer at present. Now, if we give equal status to the Ryukyuan and mainland dialects, and insist on regularity, we may have to consider the possibility that pJ had 3 contrasts for one-syllable nouns, but as many as 8 contrasts for two-syllable nouns, and 8 for three-syllable nouns, as illustrated in (29).

AKIKO MATSUMORI

120 (27)

Patterns of merger in the southwestern Kyushu dialects One-syllable nouns

(1A) class 1.1, 1.2 (1B) class 1.3

Two-syllable nouns

(2A) class 2.1, 2.2 (2B) class 2.3, 2.4, 2.5

Three-syllable nouns

(3A) class 3.1, 3.2, 3.3 (3B) class 3.4, 3.5, 3.6, 3.7

(28)

Correspondences between Ryukyuan and Kindaichi-goi accent classes 9 Ryukyuan One-syllable nouns 1a 1b Two-syllable nouns 2a 2b 2c Three-syllable nouns 3a 3b 3c 3d?

9

Kindaichi-goi class 1.1 and 1.2 class 1.3 class 2.1 and 2.2 class 2.3, 2.4, and 2.5 class 2.3, 2.4, and 2.5 class 3.1 and 3.2 class 3.4 and 3.5 class 3.4, 3.5, 3.6, and 3.7 class 3.4, 3.5, and 3.7

The nouns belonging to class 3.3 of the Kindaichi-goi either do not exist in Ryukyuan dialects, or appear to have been adopted later due to borrowing from mainland Japanese (the group of nouns in Kindaichi-goi’s class 3.3 was originally small in number). Thus, because the existence of class 3.3 in pR is extremely doubtful, this is therefore excluded from the tables shown in (28) and (29). Furthermore, it seems that Kindaichi himself, after having proposed his original Kindaichi-goi (Kindaichi & Wada 1955), later came to have doubts about, and seems to have eventually denied or ignored its existence in pJ, considering the fact that his later version of the Kindaichi-goi (Kindaichi 1974) does not include this class.


(29)

121

Patterns of merger in mainland and Ryukyuan dialects a. One-syllable nouns Merger pattern in mainland Kindaichi-goi (1-1) (1-2) (1-3)

Proto-Japanese classes

1

/

2

/

3

(1-a) (1-b) Merger pattern in Ryukyuan b. Two-syllable nouns Merger pattern in mainland Kindaichi-goi (2-1) (2-2) (2-3) (2-4) (2-5)


1 /

2 / 3 / 3 / 4 / 4/ 5 / 5

(2-a) (2-b) (2-c) Merger pattern in Ryukyuan c. Three-syllable nouns Merger pattern in mainland Kindaichi-goi (3-1) (3-2) (3-4) (3-5) (3-6) (3-7)


1 / 2 / 4 / 4/ 5/ 5 / 6 / 7

(3-a) (3-b) (3-c) Merger pattern in Ryukyuan

AKIKO MATSUMORI

122

However, giving consideration to the typological plausibility of tonal systems, the proto-system shown in (29) is rather odd.10 If we assume that this type of system is really possible, we have to ask what kind of paradigmatic relationships each of those 8 tone classes have with other patterns in the same system, and how such a system developed. Especially noteworthy in (29) are the 8 tonal patterns for two-syllable nouns in the assumed proto-system. Here, we can recall Tokugawa’s assumption concerning pJ, which is summarized in (30). Noticing the fact that the distinction between classes 2.4 and 2.5 is only observed in a restricted subset of dialects in Japan, that is, in the central region of mainland Japan, Tokugawa (1962) argued against the assumption that this distinction was present from pJ, and suggested that it is a shared innovation in the central group. If we accept Tokugawa’s hypothesis that the distinction between 2.4 and 2.5 arose later in the process of diversification of separate dialects, we have to ask what were the conditions for the development of such new tonal classes. Now, if the same distinction between class 2.4 and 2.5 was independently found in a remote, non-contiguous area of Japan, such as Tohoku, or Kyushu or the Ryukyus, it would lend strong support for the actual existence of this distinction in pJ. However, the situation is much more complicated. It seems that both earlier central Japanese and pR developed their distinctions separately, see (31). (30)

Tokugawa's family tree based on merger patterns of two-mora nouns (based on Shibatani 1990:213) *proto-Japanese 2.1 / 2.2 / 2.3 / 2.4, 2.5

10

*western Japan 2.1 / 2.2 / 2.3 / 2.4, 2.5

*earlier central Japan 2.1 / 2.2 / 2.3 / 2.4 / 2.5

*eastern Japan 2.1 / 2.2 / 2.3 / 2.4, 2.5

Oita 2.1, 2.2 / 2.3 / 2.4, 2.5

Kyoto-Osaka 2.1 / 2.2 , 2.3 / 2.4 / 2.5

Tokyo 2.1 / 2.2, 2.3 / 2.4, 2.5

However, we have to be aware that there is also a limit to the extent to which typology constrains reconstruction; that is, the fact that a certain proto-system is typologically rare does not exclude the possibility of it being valid. In fact, any proto-system can be in a transitional stage between types. There is no assurance that all language systems are typologically plausible. Thus, we cannot say that a particular proto-system is not justifiable simply because it fails to conform to the prototypes presently considered plausible.


(31)

123

Development of class 2.4 and 2.5 *proto-Japanese 2.1 / 2.2 / 2.3 / 2.4, 2.5

*proto-Ryukyuan 2.1 / 2.2 / 2.3 / 2.4a, 2.5a/ 2.4b, 2.5b

*earlier central Japanese 2.1 / 2.2 / 2.3 / 2.4 / 2.5

In relation to (31), we have other topics to pursue: i.e., what were the conditions and reasons for the split of the 2.4, 2.5 group in pR? We might have to ask such questions as whether they arose in relation to voicing contrasts among obstruents as often observed in other Asian languages, in which the tone was initially predictable from the voicing features of consonants in the same syllables and later became contrastive; or whether they developed as a compensatory mechanism from the loss of former vowel length contrasts or by the contraction of hetero-syllabic vowels. (Cf. Vovin, this volume.) In addition to these questions, we have to ask why it is the case that Ryukyuan and earlier central Japanese developed this distinction separately? In other words, the fact that the group of words belonging to the Kindaichi-goi’s class 2.4 and 2.5 both split (but in a different manner) in two widely separated regions (central Japan and the Ryukyus) suggests that there must be some features of this group (i.e., class 2.4 and 2.5) that triggered further splits. At present, the material at our disposal is too limited to attempt further speculation on this matter. Although Matsumori (2000a) attempted to reconstruct a pR accent system, it seems that it is too early at this stage to establish the proto-system with any degree of confidence. When actually working on the reconstruction, we need to work in a bottom-up fashion, first reconstructing proto-systems for the various Ryukyuan dialect groups, and then comparing them with one another to reconstruct more distant proto-systems, thereby finally reaching the overall pR system. This will be a time-consuming process, considering the mass of accumulated changes in all the daughters of the pR system. One thing which has become clear in the past decade or so is that for the Ryukyuan dialects, there is a limit to the extent to which we can use the Kindaichi-goi, which was developed based on mainland dialects. The Ryukyuan dialects contain a number of words which are not found in mainland dialects. A large number of vocabulary items in Ryukyuan have no known cognates in any of the mainland dialects whatsoever, let alone in the Kindaichi-goi. Some of those indigenous words might have existed since pJ, and have been lost in the mainland but retained in Ryukyuan. On the other hand, some of them may be innovations in Ryukyuan which have become indistinguishable from inherited vocabulary.

124

AKIKO MATSUMORI

However, we have not reached the stage at which we are able to evaluate each word and make reasonable distinctions between them. It is necessary first to develop a word list comparable to the Kindaichi-goi, which has proven to be extremely effective in mainland dialect research. I have tried to develop a vocabulary list tailored to field research on Ryukyuan accent, based on the dialect of Okinoerabu-jima in the Amami region, which is well known for retaining the largest number of tonal distinctions in the Ryukyus, with an eye to the reconstruction of the pR accentual system (Matsumori 2000a). If a suitable word list is successfully developed, it could provide a much better database and framework for comparison than the Kindaichi-goi word lists. This in turn would enable more reliable analyses on which to base our reconstruction of pR. 6.

Conclusion We have examined previous analyses of accent in pJ from a typological point of view, proposed a hypothesis concerning pMJ, and suggested some topics for future research on the Ryukyuan accent systems. However, the main aim of this chapter is not so much to propose a particular proto-system for (any subgroup of) Japanese, as to suggest some topics for future research relating to the reconstruction of pJ. Drawing on typological findings about patterns of tonal change and accent system, this chapter has proposed some basic topics to be pursed, as well as suggesting the kinds of research that should be conducted and data to be collected in the future, relying on, but also going beyond the RM and the Kindaichi-goi.

CHAPTER 6 A RECONSTRUCTION OF PROTO-JAPANESE ACCENT FOR DISYLLABIC NOUNS FOCUSING ON THE PROBLEM OF SUBCLASSES

MORIYO SHIMABUKURO University of the Ryukyus 1.

Introduction Based on the modern Japanese languages and dialects and their historical written materials, this chapter attempts to reconstruct a proto-Japanese (pJ) accent system for disyllabic nouns. This is not the first attempt; there have been a number of insightful studies that have investigated earlier forms of Japanese accent and have reconstructed pJ accent. This chapter will first examine the hypotheses presented in previous studies, and point out advantages and disadvantages. The chapter also suggests its own hypothesis about the accentual history of Japanese disyllabic accent, focusing on problems concerning the classification of accent classes for Ryukyuan disyllabic nouns. This chapter discusses classification of accent for disyllabic nouns, examining Hattori’s and Kindaichi’s classifications by looking into Ryukyuan data (§2);1 reconstruction of proto-Ryukyuan (pR) and pJ disyllabic accent classes, evaluating four possible hypotheses (§3); and the characteristics of pR and pJ accent, positing vowel length distinctions in earlier forms of Japanese (§4). 2.

Classification of accentuation for disyllabic nouns In Japanese accentology, accentual distinctions in the languages have been analyzed by using the Ruiju myôgishô (RM) accent classes.2 This is because the 1

The following symbols are used in this chapter: O syllable o mora ⎤ phonemic pitch fall between syllables or moras ⎞ phonemic pitch fall within a syllable or mora ⎡ phonemic pitch rise between syllables or moras ¯x x is high register _x x is low register 2 The Ruiju myôgishô is a dictionary from 1081 which marks pitch accent.

MORIYO SHIMABUKURO

126

language reflected in the RM has more distinctions than any other modern Japanese language or dialect, and also because words in the RM accent classes are more or less the same throughout the languages and their dialects. (1) shows the RM accent classes for disyllables and nouns belonging to each class. (1)

RM accent classes 2.1 2.2 2.3 2.4 2.5

nose, wind, cow, bird, loins, beard, etc. bridge, stone, sound, etc. jar, flee, flower, mountain, cloud, bone, etc. boat, chopsticks, sea, shoulder, breath, etc. voice, sweat, rain, shadow, bucket, etc.

Considering the fact that the language reflected in the RM has more distinctions than any other modern Japanese language or dialect, in principle it is natural to assume that the Japanese languages had at least these five distinctions and Japanese dialects developed their accents by merging the classes, unless we can account for the development of the accentual distinctions by splitting in a ‘reasonable’ way. As shown in (2), Tokyo and Kyoto dialects have three and four accent classes respectively. One of the Tokyo accent classes corresponds to RM 2.2 and 2.3, and another one to 2.4 and 2.5. Of four, one of the Kyoto accent classes also corresponds to 2.2 and 2.3. (2)

Correspondences between accent classes for disyllabic nouns RM Tokyo Kyoto

2.1 | 2.2 | 2.3 | 2.4 | 2.5 2.1 | 2.2 + 2.3 | 2.4 + 2.5 2.1 | 2.2 + 2.3 | 2.4 | 2.5

The merger explanation is more reasonable than the split explanation because it is more natural for similar accent forms to merge than for one accent form to develop into two distinct forms by splitting the class. In the latter case, what would cause splitting would be unclear. If a split of an accent class took place independently in dialects and as a result, created two distinct classes, it would be very unlikely that the classes in those dialects would share the same words. Keeping this assumption in mind, we will discuss the development of Japanese accent in this chapter. Ryukyuan accent can be analyzed in the same way as mainland Japanese accent. However, regarding Ryukyuan disyllabic accent, there are two different classifications. One is suggested by Hattori (1959), and the other is the traditional classification; we refer to Kindaichi (1975) for the traditional analysis. Simply put, the difference between their analyses is whether or not subclasses are recognized. The following sections present Hattori’s argument and the traditional analyses in detail.

PROTO-JAPANESE ACCENT FOR DISYLLABIC NOUNS

127

2.1

Hattori’s classification According to Hattori’s observation (1959, NSNT 21), there are three accentual classes for Amami disyllabic nouns. The classes and nouns belonging to each class are listed in (3). The corresponding RM accent class is indicated in parentheses. (3)

Hattori’s classification for Amami disyllabic nouns (X) nose (2.1), wind (2.1), bridge (2.2), stone (2.2), sound (2.2), etc. (Y) jar (2.3), flee (2.3), boat (2.4), chopsticks (2.4), voice (2.5), etc. (Z) flower (2.3), island (2.3), chisel (2.4), sweat (2.5), rain (2.5), etc.

As shown, nouns belonging to class X correspond to RM classes 2.1-2. Nouns in classes Y and Z correspond to RM classes 2.3-5. It should be noted that there is no one-to-one correspondence between Amami and RM disyllabic accent classes. Hattori adds that some Okinawa dialects such as Naha, Shuri, and Oroku dialects can be classified in the same way. What is important here is that nouns corresponding to classes 2.3-5 are grouped into two separate accent classes under Hattori’s classification. That is to say, there are two subclasses of nouns in each of the accent classes 2.3, 2.4, and 2.5. Throughout this chapter, we call class Y ‘subcategory (a)’ and class Z ‘subcategory (b)’, i.e., 2.3a, 2.3b, 2.4a, 2.4b, 2.5a, and 2.5b. 2.2

Kindaichi’s classification Unlike Hattori’s classification, in traditional Japanese accentology the existence of the subclasses in Ryukyuan dialects has not been recognized. Under this classification, Ryukyuan disyllabic nouns are grouped into the following three accent classes in (4). (4)

Accent classification in traditional accentology (A) nose (2.1), wind (2.1), bridge (2.2), stone (2.2), sound (2.2), etc. (B) flower (2.3), island (2.3), jar (2.3), flee (2.3), etc. (C) boat (2.4), chopsticks (2.4), sweat (2.5), rain (2.5), tortoise (2.5), etc.

Comparing this with Hattori’s, there are similarities and differences. Similar points are; (i) there are three accent classes in both classifications and (ii) nouns in one of the classes in both classifications correspond to classes 2.1-2 (i.e., Kindaichi’s class A and Hattori’s class X). On the other hand, significant differences are seen in the other classes. Class B consists of nouns corresponding to class 2.3 only; i.e., neither 2.4 nor 2.5 nouns are in this class, and class C is made up of 2.4 and 2.5 nouns only.

MORIYO SHIMABUKURO

128 2.3

Examining the Ryukyuan data In this section the accent systems of Ryukyuan dialects will be examined to see which classification (Hattori’s or Kindaichi’s) really captures the characteristics of Ryukyuan disyllabic accent. There are five Ryukyuan dialects presented here, from different dialect groups: Kametsu (Amami), Nakijin (Okinawa), Ikema (Miyako), Sonai (Yaeyama), Yonaguni. In (5), the accent system of Kametsu disyllabic nouns is shown, as an example of Amami Ryukyuan. In Kametsu, there are three accentual distinctions for disyllables, i.e., atonic (/OO/), initial accent (/O⎤O/), and final accent (/OO⎤/). Nouns having atonic accent correspond to RM classes 2.1-2. Both initial-accent and final-accent classes correspond to RM 2.3-5 nouns. From this, it is apparent that there are subclasses among nouns corresponding to RM classes 2.3-5. (5)

Kametsu3 Phonemic OO

phonetic LH(H)

RM 2.1 2.2

O⎤O

OO⎤

HL(L)

LF ~ LH(L)

nouns hana ‘nose’, ʔusi ‘cow’, thuɪ 4 ‘bird’, khusɪ ‘loins’, sïgi ‘beard’, khadɪ ‘wind’ ʔɪsɪ ‘stone’, kabi ‘paper’, hasi ‘bridge’

2.5

hu⎤ nɪ bone’ hu⎤ nɪ ‘boat’, ha⎤ i ‘needle’, ʔu⎤ sɪ ‘mortar’, ʔɪ⎤ khɪ ‘breath’ khu⎤ i ‘voice’, kha⎤ gɨ 5 ‘shadow’

2.3 2.4 2.5

hana⎤ ‘flower’, yama⎤ ‘mountain’, khumo⎤ ‘cloud’ khata⎤ ‘shoulder’ ʔamɨ⎤ ‘rain’, ʔasɪ⎤ ‘sweat’, ʔukï⎤ ‘bucket’

2.3 2.4

As mentioned earlier, Hattori also contends that his classification applies to Okinawa dialects. With regard to accentuation of Okinawa dialects, in (6) we present Nakijin data. The data demonstrate Hattori’s point, that is, there are subclasses for nouns corresponding to RM classes 2.3-5. This distinction is not revealed by Kindaichi’s classification. 3 Bold-faced nouns such as ‘bird’, ‘stone’, ‘breath’, ‘bone’, ‘boat’, mortar’, ‘voice’, ‘shadow’, ‘mountain’, ‘rain’, and ‘sweat’ are taken from NSNT 21. They are also available on the word list in Chapter 5 of Hirayama et al. (1966), but the accent of those words is not marked. 4 The vowel /ɪ/ in the word mɪ ‘eye’ (NSNT 21) corresponds to /ï/ in Hirayama et al. (1966). With regard to disyllabic nouns, however, in the initial syllable Hattori’s vowel /ɪ/ corresponds to /i/ in Hirayama, et al., and in the final syllable Hattori’s /ɪ / corresponds to /ï/ in Hirayama et al. Hattori Hirayama et al. ʔɪsɪ ʔisï ‘stone’ ʔikï ‘breath’ ʔɪkhɪ 5 Hattori seems to phonetically distinguish [ɨ] from [ɪ] in his description. Both symbols correspond to /ï/ in Hirayama et al. (1966)


(6)

129

Nakijin Phonemic ¯Ooo

phonetic LHH(H)

RM 2.1

HH(H)

2.1

nouns ¯phana: ‘nose’, ¯husi: ‘cow’, ¯husi: ‘loins’, ¯pizi: ‘beard’, ¯hazi: ‘wind’ ¯ʔisi: ~ ⎯¯hisi: ‘stone’, ¯habi: ‘paper’, ¯phasi: ‘bridge’ ¯thui ‘bird’

LLH(H)

2.3

_phana: ‘flower’, _yama: ‘mountain’, _kumu: ‘cloud’

2.4 2.5

_hata: ‘shoulder’, _hica: ‘board’ _ʔami: ‘rain’, _hasi: ‘sweat’6

2.3 2.4

phu⎤ ni(:) ‘bone’ phu⎤ ni(:) ‘boat’, ʔu⎤mi(:) ‘sea’, pha⎤ i ‘needle’,ʔu⎤si(:) ‘mortar’, ʔi⎤ ci(:) ‘breath’ hu⎤ i ‘voice’, ha⎤ gi(:) ‘shadow’, mu⎤hu(:) ‘bridegroom’, hu⎤ khi(:) ‘bucket’

2.2

_Ooo, _ooO

O⎤O, oo⎤O

HL(L), HHL(L)

2.5

We next examine the accent systems of the other three dialects in (7), (8), and (9). First, in Ikema, there are two distinctions among disyllabic accent: final accent (/OO⎤/) and atonic (/OO/). Final-accent nouns correspond to RM 2.1-5 nouns, and atonic nouns to RM 2.3-5 nouns. There is an overlap between RM classes 2.3-5 and the two accent classes of Ikema. That is to say, there are two subclasses for the nouns. Notice that the nouns listed in (3Y) such as ‘boat’ and ‘voice’ are in one subcategory, and the nouns listed in (3Z) such as ‘flower’, ‘sweat’, and ‘rain’ are in the other subcategory. For Sonai and Yonaguni disyllabic nouns, there are three accent distinctions. Nouns corresponding to RM classes 2.1-2 belong to one class (/OO/ in Sonai and /¯OO/ in Yonaguni), and nouns corresponding to RM classes 2.3-5 overlap two classes (/O⎤O/ and /O⎡O/ in Sonai, and /_OO/ and /OO⎞/ in Yonaguni). In conclusion, as the descriptions of the Ryukyuan dialects show, Hattori’s classification is sound. The subclasses can be summarized as in (10). Nouns attested in all the relevant dialects have been added to the list based on the data introduced above.

6 According to Nakasone (1983:634), the change from /ʔa/ to [ha] occurs when followed by a voiceless obstruent. Therefore, the word hasi: should be ʔasi: phonemically. Furthermore, this devoicing rule applies to /ʔu/ as well. Therefore, the noun hu⎤khi(:) ‘bucket’ should be phonemically ʔu⎤khi(:).

MORIYO SHIMABUKURO

130 (7)

Ikema Phonemic OO⎤ 7

OO

(8)

phonetic LH(L)

RM 2.1 2.2 2.3 2.4 2.5

nouns hana⎤ ‘nose’, usï⎤ ‘cow’, tui⎤ HL ‘bird’, kusï⎤ ‘loins’, higi⎤ ‘beard’, kadi⎤ ‘wind’ isï⎤ ‘stone’, hasï⎤ ‘bridge’ hana⎤ ‘flower’, yama⎤ ‘mountain’ icya⎤ ‘board’ ami⎤ ‘rain’, asi⎤ ‘sweat’

LHL ~ LHH(L)

2.2

kabi:⎤ ‘paper’

LL ~ HH(H)

2.3 2.4 2.5

puni ‘bone’ funi ‘boat’, pai ‘needle’, usï ‘mortar’, icï ‘breath’ kui ‘voice’, kagi ‘shadow’, muku ‘bridegroom’, wu:ki ‘bucket’

phonetic LL(L)

RM 2.1

Sonai Phonemic OO

2.2 O⎤O

O⎡O

7

HL(L)

LH ~ LL(H)

nouns pana ‘nose’, usi ‘cow’, kusi ‘loins’, pini ‘beard’, kazi ‘wind’ isi ‘stone’, kabi ‘paper’, pasi ‘bridge’

2.5

pa⎤ri ~ pa⎤:ri ‘needle’, u⎤si ‘mortar’, pa⎤si ‘chopsticks’, i⎤ki ‘breath’, ku⎤i ‘voice’

2.3

pa⎡na ‘flower’, hu⎡mu ‘cloud’, i⎡nu ‘dog’, pu⎡ni ‘bone’

2.4 2.5

ka⎡ta ‘shoulder’, i⎡ta ‘board’, fu⎡ni ‘boat’ a⎡mi ‘rain’, a⎡si ‘sweat’, mu⎡ku ‘bridegroom’, u⎡gi ‘bucket’

2.4

According to Hirayama et al. (1967), nouns with a locus on the final syllable are either LH(L) or HL ~ LH(L). Depending on their internal structure, those nouns are realized as LH(L) or HL ~ LH(L) at the phonetic level. Words with LH(L) pitch have a (C)VCV structure, and ones with HL ~ LH(L) have a structure of CVi, e.g., kui ‘this’.


(9)

Yonaguni Phonemic ¯OO

phonetic LH(H)

RM 2.1 2.2

_OO

LL(L)

2.3 2.4 2.5

OO⎞

LF ~ LH(H)

2.3 2.4 2.5

HL ~ HH(H)

(10)

131

2.4 2.5

nouns ¯uci ‘cow’, ¯khuci ‘loins’,¯khadi ‘wind’, ¯ηgi ‘beard’ ¯haci ‘bridge’, ¯khabi ‘paper’ _hana ‘flower’, _dama ‘mountain’, _inu ‘dog’, _mmu ‘cloud’ _kata ‘shoulder’, _ita ‘board’ _ami ‘rain’, _asi ‘sweat’ huni⎞ ‘bone’ uci⎞ ‘mortar’, haci⎞ ‘chopsticks’, iti⎞ ‘breath’, nni⎞ ‘boat’ khaηi⎞ ‘shadow’, mugu⎞ ‘bridegroom’, ugi⎞ ‘bucket’ hai⎞ ‘needle’ khui⎞ ‘voice’

2.3-5a ‘bone’, ‘mortar’, ‘chopsticks’, ‘needle’, ‘boat’, ‘sea’, ‘breath’, ‘shadow’, ‘bucket’, ‘bridegroom’, ‘voice’, etc. 2.3-5b ‘flower’, ‘mountain’, ‘cloud’, ‘shoulder’, ‘board’, ‘rain’, ‘sweat’, etc.

3.

Reconstruction of pR and pJ disyllabic accent classes The purpose of this section is to determine whether the subclasses go back to pJ or just to pR. I argue that on the basis of accentual correspondences, the subclasses in principle have to be reconstructed for both pR and pJ, and that they cannot be explained as an innovation because there are no linguistic conditions that trigger the change. Concerning whether or not the subclasses are reconstructed, as shown in (11), there are only four possible hypotheses. In the literature, hypotheses I, II, and IV are postulated, but there are no proponents of hypothesis III. In this section, the hypotheses and the reasoning behind them will be discussed. (11)

Four hypotheses about reconstruction of subclasses

I II III IV

pR yes yes no no

pJ yes no yes no

Proponents Hattori (1959, NSNT 21, NSNT 22) Matsumori (1998) Kindaichi (1975)

MORIYO SHIMABUKURO

132 3.1

Hypothesis I On this hypothesis, it is assumed the subcategory distinction existed in both pR and pJ, but in the Ryukyuan line the distinction was kept while in the Japanese line the distinction was lost. This is a general scenario. Hattori is one of the proponents of this hypothesis. According to hypothesis I, pJ had the subcategory distinction. As the diagram in (12) shows, when pJ branched off to pR and pMJ, pR kept the distinction, but the pMJ lost it by merging the classes. This is why the subcategory distinction is seen only in Ryukyuan dialects. (12)

pJ 2.3-5a 2.3-5b / pR 2.3-5a 2.3-5b

\ pMJ 2.3-5

3.2

Hypothesis II Proponents for Hypothesis II argue that the subclasses go back to pR, but no further than that; i.e., there is no subcategory distinction in pJ. Matsumori is one of the proponents for this hypothesis. In Matsumori (1998), she reconstructs the pR accent system for disyllabic nouns as in (13). She reconstructs two subclasses for 2.3, 2.4, and 2.5. However, Matsumori contends that these subclasses may not have existed in pJ because they are found only in Ryukyuan. (13)

Matsumori’s pR accent system for disyllables Class 1 Class 2 Class 3

(corresponding to RM 2.1 and 2.2 nouns) (corresponding to RM 2.3a, 2.4a, and 2.5a) (corresponding to RM 2.3b, 2.4b, and 2.5b)

According to this claim, the subcategory distinction came into existence in pR as an innovation. This hypothesis lacks explanations for the causes of the split of the accent class in the Ryukyuan line (see, however, Matsumori, this volume, for possible explanations for the split). It should be noted that splits are not common in the history of the Japanese languages, but mergers are. (14)

pJ 2.3-5 / pR 2.3-5a 2.3-5b

\ pMJ 2.3-5


133

3.3

Hypothesis III Hypothesis III suggests a development such as that shown in (15). The subcategory distinction existed in pJ, but was lost when pJ branched off to pR and pMJ. However, when the former developed to modern Ryukyan dialects, they regained the subcategory distinctions. Furthermore, nouns were recategorized into the ‘original’ groups that pJ had. This does not seem possible because after all the 2.3-5ab classes merged into one, it is inconceivable that the nouns were grouped into exactly the original classes. There are no proponents of this hypothesis in the literature. (15)

pJ 2.3-5a 2.3-5b / \ pR pMJ 2.3-5 2.3-5 / \ Shuri-type Nakijin-type 2.3-5a 2.3-5a 2.3-5b 2.3-5b

3.4

Hypothesis IV As shown in (16), this hypothesis claims that there is no subcategory distinction in the accentual history of the Japanese languages. Kindaichi is one of the proponents for this. However, as discussed in the previous section, in Ryukyuan dialects disyllabic nouns corresponding to RM classes 2.3-5 are grouped into two subclasses. This leads to the conclusion that we need to reconstruct the subclasses in pR. Therefore, hypothesis IV is not plausible. (16)

pJ 2.3 2.4-5 / pR 2.3 2.4-5

4.

\ pMJ 2.3 2.4-5

Characteristics of pR and pJ Accent This section discusses the characteristics of the subclasses, i.e., what distinguishes the subclasses (a) and (b), and also examines whether the subclasses can go back to pJ. In the discussions above, hypotheses III and IV were rejected, because they ignore the existence of subclasses. From the discussions so far, it is clear that pR must have had the subcategory distinction for disyllabic nouns. Examining the characteristics of the subclasses from a comparative point of view,

MORIYO SHIMABUKURO

134

I argue that the subclasses must have existed in pJ as well. In the following sections, Hattori’s hypothesis (i.e., hypothesis I in (11)) and Matsumori’s hypothesis (i.e., hypothesis II in (11)) are discussed, followed by my own hypothesis. 4.1

Hattori’s Hypothesis Hattori (NSNT 21) points out that some nouns in accent classes 2.3, 2.4, and 2.5 have a long vowel in the initial syllable in some Ryukyuan dialects, and the initial-syllable vowel length of those nouns corresponds to a short vowel with accent in other dialects. In (17), the correspondences of some Shuri and Nakijin 2.3-5a nouns are shown. The correspondence sets demonstrate that vowel length in the first syllable in Shuri corresponds to an accented syllable with a short vowel in Nakijin. (17)

Subclasses 2.3-5a8

Shuri Nakijin

2.3a ‘jar’ _ka:mi ha⎤mi

2.4a ‘breath’ _ʔi:ci ʔi⎤ci

2.5a ‘shadow’ _ka:gi ha⎤gi

LLL(L) HL(L)

In a comparison of the correspondences in (17) and those in (18) for 2.35b nouns, it is apparent that there are two regular correspondence sets. (18)

Subclasses 2.3-5b9

Shuri Nakijin

2.3b ‘hole’ _ʔana _ʔana:

2.4b ‘board’ _ʔica _hica:

2.5b ‘sweat’ _ʔasi _hasi:

LL(L) LLH(H)

Based on the correspondences shown above, Hattori (NSNT 21:106) proposes his reconstruction of proto-accent forms. What Hattori claims is that pJ had two subclasses corresponding to RM classes 2.3-5 (i.e., 2.3-5a and 2.3-5b); words belonging to 2.3-5a contained a long vowel in their initial syllable. The length distinction was lost in the mainland Japanese line, while it was kept in the Ryukyuan line. Furthermore, the length existed in pR, but gave rise to an accent when the long vowel became short in some Ryukyuan dialects such as Nakijin, while in some dialects such as Shuri the length has been retained. That is to say, Hattori contends that vowel length in Shuri is a remnant of pJ vowel length and its corresponding short vowel with accent in Nakijin is secondary. The diagram in 8 9

Examples for 2.3a are from NSNT 22:100 and those for 2.4a and 2.5a from NSNT 21: 103-4. Examples for 2.3b are from NSNT 22:101 and those for 2.4b and 2.5b from NSNT 21: 105.


135

(19) roughly represents Hattori’s hypothesis about the development of the Japanese languages. (19)

Hattori’s view pJ 2.3-5a *CV:CV 2.3-5b *CVCV / \ pR pMJ 2.3-5a *CV:CV 2.3-5 *CVCV 2.3-5b *CVCV / \ Shuri Nakijin 2.3-5a CV:CV 2.3-5a CV⎤CV 2.3-5b CVCV 2.3-5b _CVCV

Kindaichi (1984) argues against Hattori’s hypothesis, his reasoning being that in many dialects, there is no clear distinction between 2.3-5a and 2.3-5b. He points out that in the Kin dialect of Ryukyuan some of the putative 2.3-5a nouns lack the long vowel that they are supposed to have according to Hattori’s claim. Furthermore, some of the putative 2.3-5b nouns have a long vowel that they are not supposed to have. Therefore, Kindaichi argues that a reconstruction of the subclasses at any level makes no sense. However, as Matsumori (1998:97) points out, there is a clear distinction between the subclasses in this dialect, not reflected by vowel length, but by accentual patterns (see below). On the matter of the relationship between vowel length and accent locus, Hattori refers to the accent systems of Ainu, and explains why his hypothesis is plausible. In Ainu dialects, there are two distinct types of accent systems: pitch accent and non-pitch accent. In the non-pitch accent systems, there is a vowel length distinction, but there is no such distinction in the dialects with pitch accent . Hattori explains that the initial-syllable high pitch in Hokkaido Ainu (pitch-accent) corresponds to vowel length in the corresponding syllable of words in Sakhalin Ainu (non-pitch accent). This is apparent in the correspondence sets shown in (20). Examples are taken from Hattori (1967:219). (20)

‘red’ Saru (Hokkaido) hure HL Raichiska (Sakhalin) hu:re

‘breathe’ hese HL he:se

‘yesterday’ numan HL nu:man

Referring to these correspondences of accent and vowel length between Hokkaido and Sakhalin Ainu, Hattori (1967:220) states: “Looking at such correspondences, it would appear that Sakhalin Ainu has retained the archaic forms”. Hattori subsequently reconstructs initial-syllable vowel length in proto-Ainu for words

MORIYO SHIMABUKURO

136

like those in (20).10 It is not clear why Hattori thinks that the vowel length in Sakhalin is more conservative than the accent in Hokkaido.11 However, postulation of length in pJ has some advantages. First, shortening of long vowels is elsewhere attested in the history of Japanese. Second, merger of long and short vowels is easier to explain than a split into long and short vowels, which would require a conditioning factor. 4.2

Matsumori’s Hypotheses As for characteristics of the subclasses in pR, Matsumori claims that in earlier Ryukyuan, 2.3-5a nouns had a long vowel in their initial syllable and their locus is one mora to the left compared with the locus of 2.3-5b nouns. As a result, she reconstructs the following accent forms for the subclasses in pR. (21)

Matsumori’s reconstruction of 2.3-5ab classes 2.3-5a pR *CV:⎤CV 2.3-5b pR *CVCV⎤

Matsumori (1998) proposes that the subcategory distinction existed in pR, but not in pJ. The diagram in (22) approximates Matsumori’s view on the development of modern Japanese accent from pJ. Regarding the controversial issue of the history of the initial-syllable vowel length in the Japanese languages, Matsumori (1998) argues that the vowel length may not have existed in pJ, but came into existence in pR, because the correspondences between the initial-syllable vowel length and an accented short vowel are seen only in Ryukyuan, not in mainland Japanese. (22)

Matsumori’s view pJ 2.3-5 *CVCV / pR 2.3-5a *CV:⎤CV 2.3-5b *CVCV⎤ / Shuri 2.3-5a _CV:CV 2.3-5b _CVCV

10

\ pMJ 2.3-5 *CVCV

\ Nakijin 2.3-5a CV⎤CV 2.3-5b _CVCV(:)

Arguing that the correspondence mentioned in (20) is not regular, Vovin (1993b) reconstructs both vowel length and pitch accent for Proto-Ainu. 11 If there are principles of universal accent change on which Hattori relies, and/or if there are languages which have developed accent from vowel length, Hattori should reveal them. Hattori’s claim that length is primary and accent derived is less than convincing.


137

Matsumori’s reasoning for excluding vowel length from the reconstruction of pJ is faulty. In principle, we must reconstruct a linguistic feature for a protolanguage unless we can account for its presence. As discussed earlier, we cannot account for why there is such a correspondence between the vowel length and accent, and also cannot explain why some Ryukyuan dialects have vowel length in particular accent classes, unless we project the feature back into pJ. 4.3

My Hypothesis Based on the discussion of pR and pJ accent above, this section will deal with the following questions: a. Did the subclasses exist in pR? b. Did the subclasses exist in pJ? c. What characteristics do the subclasses have? With regard to the first question, as this investigation has revealed, Hattori’s claim concerning the subclasses is sound. There are correspondence sets that support the assumption that the subclasses must have existed in pR. Matsumori (1998) also agrees on this point. It has also been pointed out that Kindaichi’s argument is faulty. Therefore, the answer to the first question is ‘yes, the subclasses existed in pR’. Now let us turn to the second question. The answer, again, is ‘yes, the subclasses existed in pJ’. If we did not postulate pJ forms as we do, we would have a problem in explaining the development of pR from pJ. In other words, we would have to assume that splitting of accent classes took place. As explained above, as far as Japanese is concerned, merger of accentual distinctions is more natural than splitting of a distinction. Thus, a hypothesis including mergers is well-supported and more reasonable. On the third question, one can basically agree with Hattori’s view on many points. Taking Hattori’s proposal into consideration, a detailed reconstruction of pR and pJ disyllabic accent classes and also an account for the development of pR and pMJ from pJ accent can be suggested. First, length in the initial syllable has to be reconstructed in both pR and pJ because it cannot be explained as an innovation. It is much simpler and more plausible to assume that length was passed down to Ryukyuan dialects while it was lost in the mainland Japanese line. Initial-syllable long vowels are generally seen in 2.3-5a nouns in Ryukyuan dialects, but rarely seen in classes 2.1, 2.2, 2.3b, 2.4b, and 2.5b. This suggests the feature existed in nouns belonging to classes 2.3a, 2.4a, and 2.5a. Furthermore, it has also been argued that a word-initial register system existed in pR and pJ as in modern Kansai (cf. JLTT 142ff). In other words, in both pR and pJ the initial pitches of words are distinctive, in addition to their loci, which indicate a rise or a fall in pitch. Words with initial high pitch are high-

MORIYO SHIMABUKURO

138

register words and those with initial low pitch are low-register words. As in (23), 2.1-2 are the former type, and 2.3-5ab the latter type. (23)

Proposed register system High register 2.1, 2.2

low register 2.3a, 2.4a, 2.5a 2.3b, 2.3b, 2.5b

As far as the Ryukyuan dialects are concerned, there are some characteristic patterns of accent class mergers. For example, class 2.1 is not likely to merge with class 2.5. Classes tend to merge with ones with the same register. This tendency is apparent from the Ryukyuan data shown above, as well. On the mergers of accent classes in the mainland Japanese line, two basic assumptions can be made. First, classes 2.3a, 4a, and 5a are phonologically more similar to classes 2.3b 4b, and 5b respectively than to other classes. Second, pJ classes 2.3a, 4a, and 5a had a long vowel in the initial syllable while classes 2.3b 4b, and 5b had a short vowel in the corresponding syllable. When the former group lost the length, both groups merged–the subcategory distinction became no longer distinctive. In the Ryukyuan line, 2.3a, 2.4a, and 2.5a merged as one, and 2.3b, 2.3b, and 2.5b as one in pR. This is because they (i.e., the merged classes) were phonologically very similar. The reason why classes 2.3-5a and 2.3-5b did not merge in the Ryukyuan line is due to either a phonological distinction involving a long vowel, seen in Shuri, or simply accentual differences, for example, seen in Yonaguni (e.g., /OO⎞/ (2.3-5a) vs. /_OO/ (2.3-5b)). As shown in (24), eight prosodic classes have been reconstructed for pJ disyllables, five for pMJ disyllables, and three for pR disyllables. The patterns of merger differ from dialect to dialect. The reconstructions in (24) show phonological forms for the classes in detail. On the development of pMJ accent, the following scenario is assumed. The accent forms of pJ 2.3a, 4a, and 5a are very similar to those of pJ 2.3b, 4b, and 5b; the only difference is in vowel length. Therefore, it is natural to assume that they have merged. It is suggested that this is what happened in pMJ. On the other hand, since in the Ryukyuan line the pJ vowel length was retained, classes with length (*/_CV:CV/, */_CV:⎡CV/, and */_CV:CV⎤/) merged and the ones without length (*/_CVCV/, */_CV⎡CV/, and */_CVCV⎤/) also merged as one. In the Shuri line, the subcategory distinction has been maintained as a vowel length distinction. Conversely, in the Nakijin line, the subcategory distinction was retained as an accentual distinction after the length distinction was lost.


(24)

pJ 2.1 *¯CVCV 2.2 *¯CV⎤CV 2.3a *_CV:CV 2.3b *_CVCV 2.4a *_CV:⎡CV 2.4b *_CV⎡CV 2.5a *_CV:CV⎤ 2.5b *_CVCV⎤ / pR 2.1-2 *¯CVCV⎤ 2.3-5a *_CV:CV⎤ 2.3-5b *_CVCV⎤ / Shuri 2.3-5a CV:CV 2.3-5b CVCV

5.

139

\ Nakijin 2.3-5a CV⎤CV 2.3-5b _CVCV

\ pMJ 2.1 *¯CVCV 2.2 *¯CV⎤CV 2.3 *_CVCV 2.4 *_CVCV 2.5 *_CVCV⎤

Conclusion In conclusion, three major points have been argued in this chapter. First, a problem in classifying accentual distinctions was discussed. It was concluded that Hattori’s classification is correct. Second, I introduced the possible hypotheses concerning reconstruction of the subclasses and I discussed how far back in prehistory the subclasses can go. I pointed out that the subclasses have to be reconstructed in both pR and pJ, simply because there is no alternative explanation of how the subclasses came into being. I also pointed out that accent merger is more plausible than split because the former is more widely attested in the history of Japanese and is simpler to account for. Finally, I adopted the proposal that a register system existed in Japanese, and that it played a role in the development of accent classes.

CHAPTER 7 PROTO-JAPANESE BEYOND THE ACCENT SYSTEM ALEXANDER VOVIN University of Hawai‘i at Mānoa 1.

Introduction This chapter deals with two problems: the controversy over the archaic nature of the 2.5 accent class of nominals found only in Kansai Japanese (§2) and the possible correlation between low register and vowel length (§3). Although at first glance these two problems look unrelated, I will attempt to show below that they are both relevant for reconstructing certain segmental features of pJ that either have been considered controversial until now, or have passed altogether unnoticed. The traditional reconstruction of pJ accent for bimoraic nouns is presented in the following table. The following sources are used: the Ruiju myôgishô (RM, 1081; a dictionary which marks pitch accent), and also dialect data from modern Kyoto, Tokyo, and Kagoshima. Segmental forms of nouns are given in their Middle Japanese (MJ) form. Accent class and reconstruction 2.1 fasi ‘edge’ pJ *HH 2.2 fasi ‘bridge’ pJ *HL 2.3 fana ‘flower’ pJ *LL 2.4 fasi ‘chopsticks’ pJ *LH 2.5 faru ‘spring’ pJ *LF

RM HH HL LL LH LF/LH2

Kyoto HH, HH-H1 HL, HL-L HL, HL-L LH, LL-H LF, LH-L

Tokyo LH, LH-H LH, LH-L LH, LH-L HL, HL-L HL, HL-L

Kagoshima HL, LH-L HL, LH-L LH, LL-H LH, LL-H LH, LL-H

Table 1 This reconstruction, which can be traced to Kindaichi (1975), and is also adopted in JLTT, presents the pJ accent system as practically identical with the one found in the Ruiju myôgishô. It is also quite close to the modern Kyoto system. However, not everyone agrees: this system is considered controversial by some linguists (Tokugawa 1972, Ramsey 1979, 1980), who proposed that the pJ system 1 2

Pitch shapes in modern dialects are given both in isolation and with a following particle. Depending on particular manuscripts.

142

ALEXANDER VOVIN

was closer to the modern Tokyo system. Nevertheless, the detailed arguments presented by Martin in favor of the traditional system (JLTT:162-75) have been left virtually uncontested until today. Twelve years ago I also provided a detailed refutation of the Tokugawa-Ramsey hypothesis (Vovin 1995), based on internal, philological, and external evidence, therefore I will not discuss the whole controversy again here. 2.

The segmental source of accent class 2.5 However, there still remains one controversial point, which is in some sense peripheral to the controversy between traditional and Tokugawa-Ramsey reconstructions, but still deserves our attention and needs to be resolved one way or another. This point concerns the archaic nature of the accent class 2.5, which is found only in modern Kansai dialects. The philological evidence for its existence is more controversial, since not all manuscripts of the Ruiju myôgishô have the ‘eastern dot’ (tô-ten), believed to mark the distinctive falling pitch on the second mora of nouns belonging to the accent class 2.5. Naturally, supporters of the traditional reconstruction, who are ‘Kyoto-oriented’, believe in its archaic nature, while Tokugawa and Ramsey, who are ‘Tokyo-oriented’, insist that this accent class represents a local innovation in Kansai dialects. The most recent challenge to the archaic nature of the 2.5 class was presented by Unger who writes (2000a:20-21): The Russian linguist Polivanov speculated that final falling pitch is a trace of a “truncated” nasal phoneme. Martin (1987, 361 n. 1 to Chapter 4, §14) thinks he got the idea by comparing Korean achom ‘morning’ with Japanese asa id., which is 2.5 and could be taken back to an earlier *asa.m. However, Martin found that a handful of Ryukyu words with final nasals unreflected in mainislands cognates are scattered randomly across reconstructed accent classes, and Hirata and I have hunted in vain for additional Korean matches with final nasals for Japanese 2.5 nasals. Poilivanov’s explanation for class 2.5 is therefore dubious. (In any case, he suggests no link between classes 1.2 and 2.5, and apart from the falling pitch in Kyoto-type dialects, there seems to be none) ... the fact that class 2.5 is relatively small and distinctive only in one dialect group suggests that its origins should be sought in relatively recent developments within that group.

Here I assume that Unger confuses two different issues that have no connection between them whatsoever; first, the correspondence of 2.5 class nouns to nouns with final -m in Korean, and second, the correspondence of final -N, found in Hateruma, to -Ø in other Japanese dialects. The archaic nature of Hateruma -N has been persuasively refuted in the literature (JLTT:74, Oyler 1997), and it will not concern us here. The first problem, however, remains, and I will discuss it below in detail. First, the idea that nouns in accent class 2.5 have parallels in Middle Korean (MK) words with final -m indeed belongs to Polivanov (1924), who, however, gives not just one example but two. Second, Unger and Hirata are apparently unaware of the other literature on the subject where additional parallels

PROTO-JAPANESE BEYOND THE ACCENT SYSTEM

143

reflecting a correlation between Japanese nouns belonging to accent class 2.5 and MK words with final -m are provided, otherwise their hunt would probably not have been in vain. The following table summarizes these known parallels and the sources where they were first suggested: Old Japanese3

Middle Korean

source

asa 2.5 ‘morning’ turu 2.5 ‘crane’

àchóm ‘morning’ twúlwúm-í ‘crane’

Polivanov 1924:152

MJ fiyu 2.5 ‘pigweed’ paru 2.5 ‘spring’

pìlúm ‘pigweed’4 pwóm ‘spring’

Whitman 1985:202

sirwo ‘white’ adjective accent: B noun accent: 2.5 < *sirwo-m

chú- ‘dance’ (v.) ch-wú-m ‘dance’ (n.)

Vovin 1994:250 (pre-pJ nominalizer *-m for adjectives denoting colors)

kurwo ‘black’ adjective accent: B noun accent: 2.5 < *kurwo-m kura ‘dark’ adjective accent: A noun accent: 2.5 < *kura-m aka ‘red’ adjective accent: A noun accent: 2.5 < *aka-m

:twop- ‘help’ (v.) twòW-ú-m ‘help’ (n.) :wus- ‘laugh’ (v.) wùz-wú-m ‘laugh’ (n.) :wul- ‘cry’ (v.) wùl-wú-m ‘cry’ (n.) kól- ‘change” (v.) kòl-ò-m ‘change’ (n.)

awo ‘blue/green’ adjective accent: B noun accent: 2.5 < *awo-m

Table 2 The table above represents external evidence in favor of *-m in some Japanese nouns. However, it does not immediately tell us of the possible age and distribution of these words in Japanese itself. Thus, we still have to see whether Korean external evidence is applicable to pJ as a whole or is localized in Kansai 3 Accent notations are given according to EMJ as reflected in the Ruiju myôgishô, since the accentual data on OJ are not sufficient and/or are controversial. 4 The comparison of Japanese and Korean words for ‘pigweed’ was proposed originally by Martin (1966:238), but without the discussion of correlation between final -m in Korean and accent class 2.5 in Japanese.

144

ALEXANDER VOVIN

alone. Traditionally all the comparisons above, including the four lexical comparisons and the correspondence of accent class 2.5 in nouns designating colors to the nominalizer -m in MK were considered to be examples supporting the genetic relationship between Japanese and Korean. In order to confirm or disprove this position, I will offer a detailed analysis of all etymologies below. (1) OJ asa 2.5 ‘morning’ ~ MK àchóm ‘id.’. Although this comparison is usually used to support the theory of genetic relationship between Japanese and Korean, (cf. Martin 1966:236, Whitman 1985:244), there are problems with this etymology. First, as Whitman notes, it is difficult to account for the aspirated /-ch-/ in the MK form (Whitman 1985:244). MK àchóm < proto-Korean (pK) *acoKom or *aKocom, none of which could correspond to pJ *asa 2.5 < pre-pJ *asam. Second, Japanese asa ‘morning’ has a limited distribution in Japanese: it is found in Ryukyuan only in Shuri and Shitoi dialects as ʔasa (OGJ 1991:123, Hirayama 1966:284). It is significant that the word does not present itself in the southern Ryukyus, where the mainland Japanese influence was minimal. Thus, the limited attestation exclusively in Shuri and Shitoi suggests that in these two dialects it is a loanword from mainland Japanese, especially given that another word, pR *tutomuti ‘morning’ (a cognate of MJ tutomete ‘early morning’ discussed below), is attested throughout the Ryukyus, and is also present in Shuri (Hirayama 1966:284, Hirayama 1967:202, OGJ 1991:123). In addition, the word asa ‘morning’ is also attested in azuma poems, but most of the poems in which it appears5 do not have very distinctive EOJ features.6 I suspect that EOJ asa may be also a loan from WOJ. Given all this and the fact that it is impossible to connect pJ *asa < pre-pJ *asam with pK *acoKom or *aKocom, I think that OJ asa is a loan from some form of Old Korean like *acham that already had an aspirated *-ch- (and not a sequence *-Koc- or *-coK-) that would be expected to be borrowed as OJ or pJ *-s-. Even though OJ asa and MK àchóm are not genetically related, the accent pattern 2.5 is still confirmed by the Korean word as deriving from *-m, but since the Korean loanword is well attested only in WOJ, it can confirm the archaic nature of *-m only for Central Japanese (CJ),7 but not for all of Japanese. (2) OJ turu 2.5 ‘crane’ ~ MK twúlwúm-í ‘id.’. The correspondences seem to be impeccable, and there are no problems that would point to a loanword scenario, unlike the etymology for ‘morning’ discussed above. The -i in the MK form is likely to be a dimunitive suffix, cf. twulem without this suffix in the divergent Ceycwu dialect (Choy 1987:878). In addition, although the word is not attested in EOJ, it does not suffer from limited distribution in Ryukyuan: Kikaijima (Amami), Sani (Amami) tsuru; Tokunoshima (Amami) tsuruntui; 5

See examples conveniently provided in Mizushima (1984:646). The only real exception is MYS 3502, which has some typical EOJ verbal morphology. 7 ‘Central Japanese’ refers to a branch of Japanese which comprises WOJ, MJ, and the majority of modern mainland dialects, but it does not include EOJ and the modern Hachijo-jima dialect. 6


145

Yoron, Izena, Kumejima (Okinawa), Ishigaki (Yaeyama) tsuru; Hirara (Miyako) tʃuru; Shirobe (Miyako), Kawahira (Yaeyama) tsiru; Kuroshima (Yaeyama) ʃiru; Takebu (Yaeyama) tʃiru; (Nakamatsu 1987:48, 117, 165, 207, 248, 287, 326). Even if the Japanese word is a loanword from Korean, it must be a very early one, probably from the period of mutual coexistence on the Korean peninsula. Thus, this etymology fully supports the interpretation of accent class 2.5 as resulting from a lost final *-m. (3) MJ fiyu 2.5 ‘pigweed’ ~ MK pìlúm ‘id.’. First, there is a problem of attestation on the Japanese side. The word appears for the first time in the herbary Honzôwamyô (918), so it is MJ, not OJ. This leaves room for uncertainty whether its OJ form was *piyu with kô-rui /-i/ or *pwiyu with otsu-rui /-wi/. Since otsu-rui /-wi/ goes back to pJ *oi or *ui, this would make a comparison with Korean rather difficult. However, since kô-rui /-i/ appears more frequently in OJ than its otsu-rui counterpart /-wi/, there is a strong possibility that it was actually OJ *piyu. Nevertheless, there is another possible phonetic problem in this comparison, namely the correspondence of MJ -y- to MK -l-.8 In order for this comparison to work we must first assume that MK pìlúm < pK *pitum, with well-known lenition of -t- to -l-, which is possible, but again cannot be proven. Second, we also must assume that MJ fiyu < pJ *pidu. The hypothesis that OJ y < pJ *d is widely supported, but it is entirely based, I am afraid, on a mistaken theory that the Yonaguni dialect which indeed has initial d- corresponding to OJ y- (though OJ medial -y- corresponds to Yonaguni -y-, not -d-) preserves, alone among all temporal and geographical varieties of Japanese, a pJ *d which shifted to yelsewhere (JLTT:20). The full discussion of this problem is outside the scope of this chapter, but there are some good preliminary arguments against pJ *d based on Yonaguni d- (Whitman 1985:18).9 With pJ *pidu being rejected as a possible proto-form of the Japanese word, we are faced with two possibilities: either correspondence of pK *-t- to pJ *-y- (assuming that a lenition took place prior to MK) or a correspondence of pK *-l- to pJ *-y-. Neither is an enticing solution, and both require a great deal of explanation, but I am willing to accept this etymology conditionally for the time being. Still, it is not a very reliable etymology, and therefore, it cannot provide full support for the correspondence of Japanese accent class 2.5 to MK -m. (4) OJ paru 2.5 ‘spring’ ~ MK pwóm ‘id.’. Here, I believe, we have two grave phonetic problems. First, OJ -r- never corresponds to MK -Ø- except before /i/. Second, even if this was to be confirmed, there are no other instances of pJ *aCu corresponding to MK wo. Thus, I think that these two words represent a merely coincidental resemblance. 8 Martin proposed to reconstruct pJK *ř on the basis of this correspondence (Martin 1966:211), but nowadays it seems that this proposal is no longer supported. 9 I dedicate a substantial chapter to this issue in a forthcoming book (Vovin, forthcoming).

146

ALEXANDER VOVIN

(5) Japanese accent class 2.5 for nominal forms of adjectives10 designating colors ~ MK verbal nominalizer -m. Phonetically, the correspondence seems to be impeccable. However, there are two problems. First, this particular shift from an original accent class to accent class 2.5 when a word is used as a nominal is limited in Japanese to adjectives only. Second, we cannot verify that the same process was applicable to other varieties simply because accent 2.5 class is not attested in other varieties of Japanese outside MJ and the Kansai dialects of modern Central Japanese (CJ). The limited attestation within Japanese in this case (quite similar to the case of the word for ‘morning’ discussed above), I believe, points more to the borrowing scenario from some variety of Old Korean to CJ than to common genetic inheritance. This is even more likely in the light of the fact that we are dealing here with derivational and not inflectional morphology; and the former is more easily borrowed than the latter. Therefore, I believe, this correspondence confirms the archaic nature of accent class 2.5 within CJ, but not within Japanese itself. Therefore, an attempt to resolve this controversy with the help of external data leaves us with only a partial solution. It could be confirmed that accent class 2.5 has an archaic nature in CJ, so Tokugawa and Ramsey’s point of view that this accent class originated in the post-MJ era can be reliably refuted. However, it is impossible to confirm the existence of accent class 2.5 in pJ, because out of the five examples discussed above only (2) has a potential for being explained either as a common inheritance or as a very old borrowing. Nevertheless, external evidence should never take precedence in explaining internal data, otherwise one runs the very high risk of falling into a Moscow Nostratic trap, where internal etymologies are disregarded and external etymologies override the internal data. I believe that there is substantial internal Japanese evidence supporting the archaic nature of accent class 2.5 not only on the level of CJ, but on the pJ level. This evidence is based mainly on derivational morphology. Consider the following examples: a. OJ wosa 2.5 ‘elder’ cf. OJ wosame- ‘to rule’ < *wosamaCi- < *wosam-(r)a-Ci-. b. OJ toga 2.5 ‘blame, offence’ cf. OJ togame- ‘to blame, to reproach’ < *togamaCi- < *togam-(r)a-Ci-.

10 So-called adjectives in MJ and other later varieties of Japanese are in fact quality verbs, and not “adjectives” in the sense of the term assigned to adjectives of European languages. Thus, e.g., MJ kura- does not mean ‘dark’ but rather ‘to be dark’.


147

c. OJ tutwo(-ni)11 2.5 ‘early in the morning’ cf. MJ tutomete ‘id.’, believed to be a derivation from the OJ tutwome‘to strive, to make effort’, although the semantic connection remains unclear. d. OJ paya 2.5 ‘quickly’, cf. MJ fayame- ‘to quicken’. Examples (a-d) allow us to reanalyze the derivation of all deadjectival and some denominal verbs in -me- such as aratameru ‘to renew’ (< arata- ‘new’), atatameru “to warm up’ (< atataka- ‘warm’, itameru ‘to hurt’ (< ita- ‘painful’), kiyomeru ‘to purify’ (< kiyo- ‘pure’), sebameru ‘to narrow’ (< sem/ba- ‘narrow’), nagame- ‘to prolong voice when reciting poetry’ (< naga- ‘long’), etc.12 Martin analyzes these verbs as consisting of stem+suffix *-ma- + transitivity flipper *-Ci(JLTT:792). At first glance, his point of view seems to be supported by the fact that intransitive verbs corresponding to transitives in -meru all end in -maru, which Martin reconstructs as *-ma-ra- (1987:792): aratamaru ‘to be renewed’, atatamaru ‘to be warmed up’, sebamaru ‘to become narrow’, etc. However, there are three arguments against Martin’s analysis: (1) *-ma- is a suffix with an unclear function, thus it represents an extra unnecessary entity. By Ockham’s razor it is, therefore, preferable to replace it with the nominalizer *-m to which verbalizer *-(r)a- is added. (2) Martin’s analysis of intransitives in -maru as having a stem in *-ma-raproduces yet one more suffix *-ra- with an unclear function. I reanalyze the -marpart as nominalizer *-m + verbalizer *-(r)a- + passivizing suffix *-r-. Thus, we receive quite a symmetrical picture of how transitives and intransitives are derived: transitives: adjectival stem + *-m + *-(r)a- + *-Ciintransitives: adjectival stem + *-m + *-(r)a- + *-r(3) In Martin’s analysis it is unclear why his verbalizer *-ma- is added to both nouns and adjectives. This strange fact can be easily explained if *-m is treated as a final phoneme of the nominal root, but as a nominalizing suffix *-m after adjectives. One might argue against the connection between the final *-m in nouns and accent class 2.5, because not all nouns with denominal verbs in -meru or -maru belong to the 2.5 class. Thus, OJ kipa 2.3 ‘brink’, cf. OJ kipame- ‘to make go to the end’, kipamar-, kipam- ‘to reach the end’, OJ siwa 2.1 ‘wrinkle’, 11

Not listed in Zdb. Ohno (1990:884) transcribes it as tutwo with type A /-wo/, although there is only oblique evidence for that (since the word itself is not attested phonographically in OJ)—on the basis of the fact that the vowels /u/ and /o/ do not normally combine together within one root, as well as possibly on the basis of the type A vowel in the verb tutwome- ‘to strive, to make effort’. 12 Verbs like homeru ‘praises’ or hazimeru ‘begins’, of course, do not belong here.

148

ALEXANDER VOVIN

OJ siwam-13 ‘to wrinkle’ (Modern Japanese siwame-). The second case looks like a real exception, but I will try to show later that there is evidence for OJ kipa 2.3 ‘brink’ having an original 2.5 accent. There are also some bird names belonging to accent class 2.5 besides turu ‘crane’ already mentioned above: sigi ‘snipe’, OJ tadu ‘crane’, tobi ‘kite’, toki/tuki ‘ibis’, tuku/zuku ‘owl’. Cf. suffix -me in other bird names: suzume ‘sparrow’, kamome ‘seagull’, sime ‘hawfinch’, MJ tubakurame ‘swallow’. It is interesting that that none of the bird names found in accent class 2.5 has this suffix. Thus, it is possible to speculate that accent class 2.5 in these words originated as a truncation of the suffix -me. The evidence cited above directly points to *-m as a source of accent class 2.5. In addition, there is internal evidence that indicates that some kind of nasal (no direct evidence for *-m) underlies accent class 2.5: a. OJ us[w]o 2.5 ‘a breath exhaled through a narrowed mouth’, cf. OJ us[w]obuk-/us[w]omuk- ‘to exhale a breath making one’s mouth narrower’ < *uswonpuka- < ?*uswom-puk-(r)a-. b. J nasu 2.5 ‘eggplant’, cf. MJ nasubi 3.5b ‘id.’, with unclear suffixation < *nasunpi < ?*nasumpi, which can be an alternation of an earlier *nasumi.14 Finally, there are some examples that constitute additional external evidence (whether these are to be taken as cognates or as early loanwords): a. OJ pemi 2.5 ‘snake’ has *-m preserved intact due to the fact that it is followed by obsolete suffix -i (diminutive?). Martin believes that OJ pemi 2.5 < *paCimi 3.6 (JLTT:404). Cf. MK póyàm ‘snake’. b. OJ nusi 2.5 ‘master’ ~ OK *nVlim ‘lord’ (> MK :nim). The vocalism in the first syllable may be a problem; the vocalism of the first syllable in OK is unclear. c. OJ masa(-ni) 2.5 ‘truly’, possibly from ma- ‘true’ + -sa ‘?’, which can be compared with MK chom ‘truly, really’. d. OJ saru 2.5 ‘monkey’ ~ WM sarmacin ‘id.’. e. OJ usagi 3.6 ‘hare’, cf. development of 3.6 to 2.5 in OJ pemi ‘snake’ above. Cf. Koguryo *osigam ‘hare’. The last example brings us to an interesting internal issue. On the basis of Koguryo *osigam we would expect that the pJ *usagi would belong to accent 13

Zdb lists siwam- as a consonantal verb. The examples, both in logographic script in OJ and in phonetic in MJ, provide no basis for conclusion that siwam- used to be a consonant rather than a vowel verb in OJ. The earliest known attestation of this verb is in the form siwabitarikeru ‘has wrinkled’ (Konjaku monogatari 26.2) that shows that it is either a consonantal or upper-bigrade vowel verb (Ohno 1990:695). 14 I am indebted to Blaine Erickson for pointing this out to me.


149

class 3.7 (LHL-L) that correlates with accent class 2.5 (LH-L), but it belongs to accent class 3.6 (LHH-H), which correlates with 2.4 (LH-H). Furthermore, we have a number of cases where either a lack of accentual data or irregularity in accentuation could make a choice between 2.5 and 2.4 or 2.5 or 2.3 questionable, although the bulk of the evidence normally points to class 2.5 (cf., e.g., sake ‘salmon’ 2.5/2.3? or above-mentioned wosa ‘elder’ 2.5/2.4?). The same picture is typical for accent classes 3.6 and 3.7. This, in combination with the observation made above regarding the potential Koguryo cognate *osigam for pJ *usagi 3.6, suggests a hypothesis that the small classes 2.5 and 3.7a/b became small by attrition: a number of words that used to belong to those classes are now found under 2.4, 2.3, and 3.6.15 There is one additional external comparison that could further support this attrition hypothesis: OJ asa ‘hemp’ 2.3 ~ MK sám ‘id.’. The accentuation in both modern Kyoto and Shuri is irregular (JLTT:384), possibly suggesting a shift from another accentual class, although Kyoto and Shuri contradict each other. Now we can return to the issue of why OJ kipa ‘brink’ is 2.3 rather than 2.5. OJ kipami ‘limit’, derived from kipam- ‘to reach the end’, is 3.6 (although Kyoto has irregular accent); the 3-mora accent class that correlates with 2.3 is 3.4, and not 3.6, suggesting, however, that the original class of kipa may have been 2.5. Finally, I am not sure that Unger is right when he asserts that the monomoraic accent class 1.2 (H-L) correlates with accent class 2.5 for bimoraic words. I would rather follow the traditional position that it correlates with 2.2a (HH-L) and 2.2b (HL-L). The principal difference between 1.2 and 2.5, regarding the fall of pitch that Unger mentions, is that in Kyoto-type dialects class 2.5 has the fall within the last mora in isolated form, and after the last mora before the following case particle, but class 1.2 in both cases has the fall after the first mora. The isolation form for modern Kyoto is HL, the same as for accent class 2.2, due to the fact that Kyoto automatically lengthens one-mora words to two-moras. It is also necessary to keep in mind that 2.5 is a low-register class, while 1.2 is a highregister class. Thus, one should not expect to find any final *-m underlying words belonging to accent class 1.2. Therefore, I believe that there is sufficient evidence, mainly internal and to some extent external, to link the origins of accent class 2.5 in pJ with a loss of the final consonant *-m. It is true that it is not possible to demonstrate the probability of final *-m for every noun belonging to accent class 2.5, but that should come as no surprise, because not every noun in this class has verbal counterparts derived from these nouns. It is not surprising that the number of words with an external explanation is even smaller. Thus, the conclusion for the first part of this chapter is that accent class 2.5 is not a recent innovation in Kansai dialects, but that it rightfully belongs to pJ. However, within the pJ accent system it is secondary, because its specific shape (the fall of pitch within the last mora) is due to a 15

Martin noted that “a number of words in class 2.4/5 with final open vowel in Old Japanese have shifted to 2.3” (JLTT:177).

150

ALEXANDER VOVIN

simplification of segmental phonetics, namely the loss of the final consonant *-m. This should come as no surprise, because pitch accents, like tones, do not represent an original characteristic of a language; they must come from somewhere. This ‘somewhere’ usually turns out to be certain segmental features that either merged with other features, or have been lost altogether, leaving their traces in the shape of tones or pitch accent. This brings us to the second problem to be discussed in this chapter: the correlation between low register and vowel length. 3.

The segmental sources of low register Martin suggested that pJ low register probably reflects initial vowel length in pre-pJ (JLTT:249-53). I later argued that this solution is likely to be only partially plausible, because there are more words with initial low register than words with high register, and this creates a typologically unjustified situation with long vowels occurring more frequently in initial syllables than short vowels (Vovin 1995:131). I proposed to remedy this by positing two sources for words with low register: pre-pJ vowel length, as suggested by Martin, plus pre-pJ initial voiced consonants, and offered some external evidence for the origin of low register from initial voiced consonants (Vovin 1995:124-31). This solution, although typologically more elegant, suffers from three major problems: (1) Until recently it was impossible to tell on the basis of the internal Japanese evidence which low register words would reflect initial vowel length and which would reflect initial voiced consonants; (2) I was still adhering at this point to the traditional reconstruction that posits pJ *b- and *d- on the basis of the Sakishima Ryukyuan evidence. This created two different reconstructed series of voiced *b- and *d-: one on the basis of the Sakishima evidence, and the other on the basis of the low register; (3) It was unclear what to do with high register words that start with vowels or nasal sonorants *m- and *n-: surely these are also voiced. Thus, on my earlier explanation, they should be all expected to be found in low register classes, which certainly also have their share of vowel and nasal sonorant initial words. My tentative solution was to posit an initial glottal stop or *H- for high register words starting with vowels and nasal sonorants versus smooth vocalic ingress and pure nasal sonorants for low register words (Vovin 1995:131-32). As I will try to demonstrate below, this solution may still be partially true, although at that time without a solution to (1) it was mere speculation. Regarding the major problem (1) above, Hattori was the first linguist to suggest that Ryukyuan preserves vowel length in some words belonging to accent classes 2.3, 2.4, and 2.5 (NSNT 21, 22). He also added a couple of examples of words with vowel length from accent class 2.2. Hattori’s proposal was largely


151

based on correspondences within several Okinawan dialects, in particular Shuri and Onna, as vowel length does not seem to be preserved outside Okinawa. Hattori labeled 2.3-2.5 nouns with vowel length as subclass (1) and nouns in the same classes without vowel length as subclass (2) (these are labeled 2.3a-2.5a and 2.3b-2.5b below). The apparent weakness of Hattori’s proposal is that while there are a number of examples with and without vowel length for classes 2.4 and 2.5 (NSNT 21:103-05), the examples with vowel length for classes 2.3 and 2.2 in his presentations are very scarce (NSNT 22:100-01). In addition, as Martin pointed out, Hattori’s correspondences sometimes are not consistent (JLTT:252-53). A significant refinement of Hattori’s proposal has been made in a recent dissertation by Moriyo Shimabukuro, who demonstrated two important facts (2002; see also Shimabukuro, this volume). First, words with first syllable vowel length in Shuri mostly correlate with the low pitch classes 2.3-2.5, and initial vowel length in most words belonging to classes 2.1-2.2 is either found in compounds or is otherwise secondary (Shimabukuro 2002:293). Second, most important, Shimabukuro managed to find a strong internal correlation supporting a reconstruction of vowel length in Ryukyuan on the basis of two Okinawan dialects rather than just one, thus taking care of Martin’s objection that Hattori’s data are not consistent. Shimabukuro demonstrated that initial accent in the Nakijin dialect regularly corresponds to vowel length in Shuri for the subclasses 2.3a-2.5a, while there is no such correspondence in the case of absence of vowel length in Shuri (subclasses 2.3b-2.5b) (Shimabukuro 2002:203):16 Gloss & accent class mortar 2.4a breath 2.4a needle 2.4a shadow 2.5a bucket 2.5a bridegroom 2.5a flower 2.3b mountain 2.3b cloud 2.3b shoulder 2.4b board 2.4b rain 2.5b sweat 2.5b

Nakijin ʔu⎤si(:) i⎤ci(:): pha⎤i ha⎤gi(:) hu⎤khi(:) mu⎤hu(:) _phana: yama: kumu: _hata: _hica: _ʔami hasi:

Shuri _ʔu:si _ʔi:ci _ha:i _ka:gi _u:ki _mu:ku _hana _yama _kumu _kata _ʔica _ʔami _ʔasi

Table 3 Shimabukuro reconstructs vowel length for the words belonging to 2.4a2.5a in this table and in the absence of any other cogent explanation I would like 16

Fn. 1 in Shimabukuro’s chapter in this book explains the symbols used in Tables 3-6.

152

ALEXANDER VOVIN

to agree with him. What remains to be explained, though, is the absence of words belonging to 2.3a in this table; in this regard we achieved no significant improvement over Hattori. Shimabukuro also presents three exceptions to the correspondence of long vowel in Shuri to initial accent in Nakijin, where initial accent in Nakijin corresponds to short vowel in Shuri (Shimabukuro 2002:204): Gloss & accent class bone 2.3a boat 2.4a sea 2.4a

Nakijin phu⎤ni(:) phu⎤ni(:) ʔu⎤mi(:)

Shuri _huni _huni _ʔumi

Table 4 Shimabukuro argues that the vowel of the first syllable “must have been long earlier in its history, but it has been irregularly shortened” (Shimabukuro 2002:204). He further supports this point with examples of long vowels in these three words found in the Kamishiro, Onna, and Matsuda dialects (Shimabukuro 2002:204), but this is still reminiscent of Hattori’s irregular correspondences criticized by Martin. Although Shimabukuro may be ultimately right, for the sake of keeping the correspondences as tight as possible, I will omit these three words from further consideration. Therefore, although the issue of regularity within Okinawan raised by Martin has been, I believe, resolved answered in Shimabukuro’s dissertation, there remains a problem as to whether it is possible to reconstruct separate subclasses (a), involving vowel length, and (b), without vowel length for accent class 2.3. I believe that in spite of the fact that both Hattori and Shimabukuro failed to provide a decisive answer to this problem (both give just one example each for subclass 2.3a (NSNT 22:100, Shimabukuro 2002:204, 369)), their intuitive solution is in fact correct, because there are uncontroversial examples belonging to subclass 2.3a that I add in Table 5 which is otherwise based on a table found in Shimabukuro’s dissertation that presents the reconstruction of two different subclasses (a) and (b) for accent classes 2.3-2.5 (Shimabukuro 2002:36970).17 One can clearly see that words belonging to subclass (a) have vowel length in their first syllables, supporting Martin’s hypothesis about the origin of the low register from vowel length. But it can be supported only partially, since there are no traces of vowel length in the words belonging to subclass (b). However, it does not make sense that some words in pJ have initial low register, and others have

17

I took the liberty of making some changes to Shimabukuro’s pJ and pR reconstructions that, for example, reflect the current wisdom on preservation of primary pJ *o and *e, which underwent raising in Central Japanese to /u/ and /i/. Besides words belonging to 2.3a, I have also made some other additions where Shimabukuro’s original examples did not seem to be sufficient.


153

initial vowel length, with low register and vowel length later merging as low register. It is most likely that the initial low register in subclass (b) also has an Gloss tortoise 2.3a offshore 2.3a jar 2.3a bean 2.3a ball 2.3a root 2.3a flower 2.3b mountain 2.3b cloud 2.3b dog 2.3b wave 2.3b shank 2.3b breath 2.4a mortar 2.4a chopsticks 2.4a needle 2.4a board 2.4b footprint 2.4b shoulder 2.4b dregs 2.4b number 2.4b skin 2.4b bucket 2.5a bridegroom 2.5a voice 2.5a spider 2.5a shadow 2.5a sweat 2.5b rain 2.5b spring 2.5b crane 2.5b

pR *kha:me⎞ *_ooO⎞ *ʔu:ki⎞ *_ooO⎞ *kha:me⎞ *_ooO⎞ *ma:me⎞ *_ooO⎞ *ma:i⎞ *_ooO⎞ *mu:tu⎞ *_ooO⎞ *_phana *_OO⎤ *_yama *_OO⎤ *_khumu *_OO⎤ *_ʔenu *_OO⎤ *_nami *_OO⎤ *_suni *_OO⎤ *ʔi:ki⎞ *ooO⎞ *ʔu:si⎞ *ooO⎞ *pha:si⎞ *ooO⎞ *pha:ri⎞ *ooO⎞ *_ʔita⎤ *_OO⎤ *_ʔatu⎤ *_OO⎤ *_khata⎤ *_OO⎤ *_khasu⎤ *_OO⎤ *_khazu⎤ *_OO⎤ *_phada⎤ *_OO⎤ *wo:khe⎞ *ooO⎞ *mo:kho⎞ *ooO⎞ *khu:e⎞ *ooO⎞ *kho:bo⎞ *ooO⎞ *kha:ge⎞ *ooO⎞ *_ʔasi⎤ *_OO⎤ *_ʔame⎤ *_OO⎤ *_pharu⎤ *_OO⎤ *_thuru⎤ *_OO⎤

pMJ *_kamey *_OO *_oki *_OO *_kame *_OO *_mamey *_OO *_mari *_OO *_moto *_OO *_pana *_OO *_yama *_OO *_kumwo *_OO *_inu *_OO *_nami *_OO *_sune *_OO *i⎡kyi *O⎡O *u⎡su *O⎡O *pa⎡siy *O⎡O *pa⎡riy *O⎡O *i⎡ta *O⎡O *a⎡to *O⎡O *ka⎡ta *O⎡O *ka⎡su *O⎡O *ka⎡nsu *O⎡O *pa⎡nta *O⎡O *wokey⎤*OO⎤ *mukwo⎤ *OO⎤ *kowey⎤ *OO⎤ *kumwo⎤ *OO⎤ *kagey⎤ *OO⎤ *ase⎤ *OO⎤ *amey⎤ *OO⎤ *paru⎤ *OO⎤ *turum⎤ *OO⎤

pJ *_ka:may *_ooO *_o:ki *_ooO *_ka:me *_ooO *_ma:may *_ooO *_ma:ri *_ooO *_mo:to *_ooO *_pana *_OO *_yama *_OO *_kumo *_OO *_enu *_OO *_nami *_OO *_sune *_OO *ʔi:⎡ki *oo⎡O *ʔu:⎡su *oo⎡O *pa:⎡suy *oo⎡O *pa:⎡ruy *oo⎡O *i⎡ta *O⎡O *a⎡to *O⎡O *ka⎡ta *O⎡O *ka⎡su *O⎡O *ka⎡nsu *O⎡O *pa⎡nta *O⎡O *wo:kay⎤ *ooO⎤ *mo:ko⎤ *ooO⎤ *ko:way⎤ *ooO⎤ *ko:m[p]o⎤ *ooO⎤ *ka:nkay⎤ *ooO⎤ *ase⎤ *OO⎤ *amay⎤ *OO⎤ *paru⎤ *OO⎤ *turum⎤ *OO⎤

Table 5 origin in pre-pJ segmental phonology. My earlier hypothesis that the initials in some words belonging to the low register classes go back to voiced segments thus seems to be the only concrete solution based on internal reconstruction, because initial low register develops either from initial vowel length or from initial voiced segments. There are simply no other possible sources, and vowel length only takes account of words belonging to subclass (a). I therefore propose to reconstruct pJ words belonging to subclasses 2.3b-2.5b with initial voiced onsets, including

154

ALEXANDER VOVIN

voiced obstruents, nasal sonorants, and smooth vocalic ingress. All these words have short vowels in the first syllable. All words belonging to subclasses 2.3a2.5a are reconstructed with long vowels in the first syllable. Theoretically, they could have voiced onsets as well, but we simply have no evidence for that. I tentatively reconstruct voiceless onsets for words beginning with obstruents. Vowel-initial words are reconstructed with initial glottal stop *ʔ-, which also behaves as a voiceless onset. Nevertheless, I reconstruct words with initial nasal sonorants belonging to subclasses 2.3a-2.5a as having voiced onset and initial vowel length. The reason for this is combinatorial and will become clear when I deal with reconstruction of onsets of high register words below. As I mentioned above, there is a possible contradiction between this reconstruction and a traditional reconstruction of pJ that posits pJ *b- and *d- on the basis of the Sakishima b- and Yonaguni d- evidence. The discussion of why the Sakishima and Yonaguni evidence is not admissible falls outside the scope of this chapter, but I believe that we should reconstruct just pJ *w- and *y-, and I address this issue in greater detail in a forthcoming book (Vovin, forthcoming). Thus, the major problem (2), mentioned above will be taken care of, since there will not be two different series of initial voiced obstruents reconstructed for pJ. Thus, Table 5 above can be re-written as in Table 6, eliminating low initial register for pJ altogether and replacing it with voiced onsets for the subclasses 2.3b-2.5b and vowel length in first syllable for the subclasses 2.3a-2.5a.18 As the major problems (1) and (2) have been solved above, it is now possible to address problem (3): that is, what kind of onsets should be reconstructed for the words belonging to high-register classes. Since initial low register has been eliminated from the reconstruction, it does not make sense to preserve high register either. I think that high register words should all be reconstructed as words with voiceless onsets and with short vowels in their first syllables. Thus, the following reconstructions are suggested for the three different types of onsets (I use the syllables /a/, /ta/, and /na/ as examples): The reason for reconstructing nasal sonorant onsets belonging to subclasses 2.3a-2.5a as voiced with vowel length is purely combinatorial, as I mentioned above. Voiced nasals occur cross-linguistically much more frequently in the languages that have both voiced nasals and voiceless nasals (pre-glottalized nasals function as voiceless). Since I reconstruct *ʔn- and *ʔm- for words with nasal sonorant initials belonging to traditional accent classes 2.1-2.2, reconstructing preglottalized nasals for subclasses 2.3a-2.5a would produce more preglottalized nasal sonorants than voiced nasal sonorants in the proto-language, which would be typologically unexpected.19 18

Note that we cannot completely eliminate non-initial pitch: we have no segmental explanation for high pitch on the second mora of 2.4. For 2.5, as it was suggested above, the falling pitch on the second mora can be explained as the result of the loss of final *-m, but only one word in the table has a direct evidence for it. 19 It is necessary to keep in mind that a reconstruction of pre-glottalized nasals *ʔn- and *ʔm- may actually be rewritten as a reconstruction of voiceless nasals *hn- and *hm-.


Gloss tortoise 2.3a offshore 2.3a jar 2.3a bean 2.3a ball 2.3a root 2.3a flower 2.3b mountain 2.3b cloud 2.3b dog 2.3b wave 2.3b shank 2.3b breath 2.4a mortar 2.4a chopsticks 2.4a needle 2.4a board 2.4b footprint 2.4b shoulder 2.4b dregs 2.4b number 2.4b skin 2.4b bucket 2.5a bridegroom 2.5a voice 2.5a spider 2.5a shadow 2.5a sweat 2.5b rain 2.5b spring 2.5b crane 2.5b

pR *kha:me⎞ *_ooO⎞ *ʔu:ki⎞ *_ooO⎞ *kha:me⎞ *_ooO⎞ *ma:me⎞ *_ooO⎞ *ma:i⎞ *_ooO⎞ *mu:tu⎞ *_ooO⎞ *_phana *_OO⎤ *_yama *_OO⎤ *_khumu *_OO⎤ *_ʔenu *_OO⎤ *_nami *_OO⎤ *_suni *_OO⎤ *ʔi:ki⎞ *ooO⎞ *ʔu:si⎞ *ooO⎞ *pha:si⎞ *ooO⎞ *pha:ri⎞ *ooO⎞ *_ʔita⎤ *_OO⎤ *_ʔatu⎤ *_OO⎤ *_khata⎤ *_OO⎤ *_khasu⎤ *_OO⎤ *_khazu⎤ *_OO⎤ *_phada⎤ *_OO⎤ *wo:khe⎞ *ooO⎞ *mo:kho⎞ *ooO⎞ *khu:e⎞ *ooO⎞ *kho:bo⎞ *ooO⎞ *kha:ge⎞ *ooO⎞ *_ʔasi⎤ *_OO⎤ *_ʔame⎤ *_OO⎤ *_pharu⎤ *_OO⎤ *_thuru⎤ *_OO⎤

PMJ *_kamey *_OO *_oki *_OO *_kame *_OO *_mamey *_OO *_mari *_OO *_moto *_OO *_pana *_OO *_yama *_OO *_kumwo *_OO *_inu *_OO *_nami *_OO *_sune *_OO *i⎡kyi *O⎡O *u⎡su *O⎡O *pa⎡siy *O⎡O *pa⎡riy *O⎡O *i⎡ta *O⎡O *a⎡to *O⎡O *ka⎡ta *O⎡O *ka⎡su *O⎡O *ka⎡nsu *O⎡O *pa⎡nta *O⎡O *wokey⎤*OO⎤ *mukwo⎤ *OO⎤ *kowey⎤ *OO⎤ *kumwo⎤ *OO⎤ *kagey⎤ *OO⎤ *ase⎤ *OO⎤ *amey⎤ *OO⎤ *paru⎤ *OO⎤ *turum⎤ *OO⎤

pJ *ka:may *o:ki *ka:me *ma:may *ma:ri *mo:to *bana *yama *gumo *enu *nami *zune *ʔi:⎡ki *oo⎡O *ʔu:⎡su *oo⎡O *pa:⎡suy *oo⎡O *pa:⎡ruy *oo⎡O *i⎡ta *O⎡O *a⎡to *O⎡O *ga⎡ta *O⎡O *ga⎡su *O⎡O *ga⎡nsu *O⎡O *ba⎡nta *O⎡O *wo:kay⎤ *ooO⎤ *mo:ko⎤ *ooO⎤ *ko:way⎤ *ooO⎤ *ko:m[p]o⎤ *ooO⎤ *ka:nkay⎤ *ooO⎤ *ase⎤ *OO⎤ *amay⎤ *OO⎤ *baru⎤ *OO⎤ *durum⎤ *OO⎤

Table 6

Type of onset vowel aobstruent tanasal na-

2.1-2.2 *ʔa*ta*ʔna-

2.3a-2.5a *ʔa:*ta:*na:-

Table 7

155

2.3b-2.5b *a*da*na-

156 4.

ALEXANDER VOVIN

Conclusion If my reconstruction above is correct, it might have far-reaching consequences for the study of the external relationships of the Japanese language family. A contrast between smooth vocalic ingress and an initial glottal stop on one hand, and voiced and voiceless nasals on the other, has no parallels in the ‘Altaic’ world. Instead, these features are widely found throughout South-East Asia. If the Japanese proto-language indeed arrived in Japan together with the Yayoi culture, it should not be surprising to find some South-East Asian elements in Japanese, although at the present stage of our knowledge the exact nature of the connection between Japanese and any of the major language families found in South-East Asia, if any, remains at best hypothetical. The search for these possible connections is a legitimate enterprise, but it should be carried out in a cautious manner.

PART IV RECONSTRUCTING MORPHOLOGY AND SYNTAX

CHAPTER 8 THE SOURCE OF THE BIGRADE CONJUGATION AND STEM SHAPE IN PRE-OLD JAPANESE JOHN WHITMAN Cornell University 1.

Introduction One of the first things that a learner of Japanese notices is the contrast between the verb forms in (1): (1)

a. Tenpu ga tuku to 50Kbyte gurai ni naru. attachment NOM attach when 50kbyte about LOC becomes ‘When an attachment attaches, it (the message) becomes about 50K bytes.’ b. Tenpu o tukeru to 50Kbyte gurai ni naru. attachment ACC attach when 50kbyte about LOC becomes ‘When (one) attaches an attachment, it (the message) becomes about 50K bytes.’

She notices the formal similarity between the stem tuk- ‘attachINTR’ in (1a) and tuke- ‘attachTR’ in (1b). These represent the two main verb conjugations in Japanese: tuk- represents the consonant stem or godan conjugation, while tukerepresents the vowel stem or ichidan conjugation. The learner also notices the grammatical relationship between the two stems: tuk- is the intransitive and tukethe transitive form of ‘attach’. Soon she comes upon another remarkable fact: pairs like (1) are quite common, but it is not possible to predict which conjugation has which valence. In some cases, the relationship between conjugation class and transitivity is reversed: (2)

ga sakeru to kizu ga hukaku naru. a. Hihu skin NOM split when wound NOM deep becomes ‘When the skin splits, the wound gets deeper.’

JOHN WHITMAN

160

b. Hihu o saku to kizu ga hukaku naru. skin ACC split when wound NOM deep becomes ‘When one splits the skin, the wound gets deeper.’ In (2), the vowel-stem verb is transitive and the consonant-stem verb is transitive. This is the opposite of the situation in (1). This basic state of affairs—the existence of two main conjugation classes, the presence of formally similar stems in both conjugations, related by transitivity, and the unpredictability of the transitivity relation—has an old history in Japanese. It obtains in the oldest attested variety, OJ, as shown in (3) and (4): (3)

a. taka-ki mine ni kumo no tuk-u nos-u high-ADN peak to cloud GEN attach-ADN like ‘like clouds attaching themselves to a high peak’ (MYS 14.3514 b. yo no naka ni kokoro tuke-zu-te world GEN midst in heart put-not-ing ‘without attaching my heart to the affairs of the world’ (MYS 19.4162)

(4)

a. ipa sak-u rock split-CONC/ADN ‘splits rock.’ (Nihon shoki Jindai, phonetic note explaining name of god) b. (tume) sake-te nail split-ing ‘the nail splits and...’ (Nihon ryôiki 2.26)

Intransitive tuk- ‘attach’ and transitive sak- ‘split’ belong to the quadrigrade conjugation in OJ, while transitive ‘attach’ and intransitive ‘split’ belong to the bigrade class. The descendants of these four verbs belong to the corresponding conjugation classes in modern Tokyo and Kyoto Japanese as well as in varieties whose divergence from the ancestors of modern Tokyo and Kyoto antedates the 8th century, such as Shuri (OGJ). The conjugation classes have undergone important mergers and other changes, but the basic opposition between these two main conjugations remains in the standard language and other varieties. 2.

Previous reconstructions For the past half century, linguists have attempted to explain how the main conjugation classes came into being in Japanese. Let us first examine how these classes are situated within the eight traditional conjugation classes of OJ.

THE BIGRADE CONJUGATION AND STEM SHAPE IN PRE-OLD JAPANESE

(5)

161

The conjugation classes of OJ Conjugation Quadrigrade Upper Bigrade Lower Bigrade Monograde K-irregular S-irregular N-irregular R-irregular

Conclusive

Irrealis Infinitive

Adnominal

Realis

Imperative

oku (A) ‘put’ oku (B) ‘arise’ aku (B) ‘dawn’ miru (B) ‘see’ ku (B) ‘come’ su (A) ‘do’ inu (A) ‘go’ ari (B) ‘be’

oka okwi ake mi ko se ina ara

oku okuru akuru miru kuru suru inuru aru

oke okure akure mire kure sure inure are

okye okwi(yo) ake(yo) mi ko se ine are

oki okwi ake mi ki si ini ari

The OJ conjugations suggest that the Irrealis (mizenkei) base, which appears with a functionally heterogeneous group of suffixes such as the negative and passive, reveals the original shape of certain stems, in particular the irregular monosyllabic stems ‘come’ and ‘do’, because the vowels in the Irrealis base of these verbs are quite unpredictable. A second striking fact about the OJ conjugations is the distinctiveness of the bigrade (nidan) Infinitive (ren’yôkei) endings, -wi and -e. This last is the central motivation of Ohno’s (1953) reconstruction of the pJ verbal system. (6)

Ohno’s (1953) reconstruction of the pJ verbal system Conjugation Quadrigrade

Pre-OJ Upper Bigrade

Pre-OJ

Conclusive

Irrealis Infinitive

Adnominal Realis

Imperative

oku ‘put’ *oki-u

oka oki *ok-a *ok-i

oku *ok-ru

okye *oki-a

oku ‘arise’ *oki-u

okwi okwi *oko-i *oko-i

okuru okure oki(yo) *oko-uru *oko-urai *oki-yo

Lower Bigrade aku ‘dawn’ ake

oke *ok-ai

Pre-OJ

*ake-u

ake *aka-i *aka-i

akuru akure ake(yo) *aka-uru *aka-urai *ake-yo

Monograde

miru ‘see’ *mi-ru

mi *mi-i

miru *mi-ru

Pre-OJ

mi *mi-i

mire *mi-ra

mi *mi

Ohno reconstructed an opposition in stem shape parallel to the opposition between the modern conjugations. He reconstructed the ancestor of OJ quadrigrade conjugation and modern consonant stems as having consonant stems in pJ; he reconstructed the ancestor of OJ bigrade and modern vowel stems as having vowel stems. Ohno did not simply project the modern opposition back into the past, however. Hashimoto (1931/1949:174) had noticed that while the quadrigrade Infinitive ending is -i, the bigrade ending is -wi (in the upper bigrade conjugation) and -e (in the lower bigrade conjugation). Ohno exploited the parallel between the bigrade Infinitive endings and other cases where the nucleii /wi/, /e/ are derived from *u, o + i and *a + i respectively (1955:308) to reconstruct the bigrade stems as ending in *-u, *-o, or *-a. Further support for these reconstructions comes from what Arisaka (1931/1975:50) calls ‘embedded root forms’ (hifukukei): cases where the original

JOHN WHITMAN

162

root in *-u, *-o, or *-a is embedded in a derived form. (7) gives examples of this for the two verbs in (4), upper bigrade oku/okwi ‘arise’ and lower bigrade aku/ake ‘dawn’. (7)

Embedded root forms: a. okosu ‘raise, cause’ okoru ‘arise, occur’

< *oko-s-u < *oko-r-u

b. akasu ‘make dawn, make clear’ < *aka-s-u akasi ‘red’ (cf. uninflected adjective aka ‘red’) < *aka-si A final comment is in order here about Ohno’s reconstruction of the Irrealis forms. The table in (6) shows an Irrealis suffix of variable shape: *-a in the quadrigrade conjugation, *-i elsewhere. But Ohno’s actual proposal is that the Irrealis is a secondary category, made up of vowel-initial suffixes (traceable to auxiliaries) that are suffixed directly to the original stem. On this view, for example, negative Adnominal yukanu ‘doesn’t go’ derives from *yuk+an-u, where yuk- is the stem of ‘go’ and an-u is the Adnominal form of the negative auxiliary. Speakers later reanalyze such forms as yuk-a-n-u, taking the initial vowel of the auxiliary to be the Irrealis ending. There is good evidence that this reanalysis is still in progress in the 8th century, as subsequently productive endings that select the Irrealis base, such as causative -su/-se, are just beginning to emerge at this time. A unified reconstruction of the Irrealis as resulting from stem+aCV(CV)- is made explicit in Unger’s reconstruction, discussed in the next section. The Irrealis base thus derives from reanalysis of longer vowel-initial endings originally attached directly to the stem. Ohno’s reconstruction explains the shape of the bigrade Infinitive endings and the relationship between bigrade verbs and embedded root forms as in (5). But it fails to explain the relationship between the bigrade pattern and the transitivity alternations we saw in (1-4). We might note also that Ohno’s hypothesis that all bigrade verbs descend from vowel stems forces us to reconstruct pJ stems such as *tuka- for ‘attachTR’ and *saka- for ‘splitINTR’; but there are no embedded root forms that provide independent support for such reconstructions in the case of these verbs. Unger (1993[1977]) and Yoshida (1973) are responsible for the next important advance in the reconstruction of the bigrade classes, in my opinion. These linguists share the insight that the bigrade conjugations are derived, or secondary (see Frellesvig, this volume, for another version of this view). Unger provides a comprehensive reconstruction for all of the conjugations of pJ. (8) shows his reconstruction for the quadrigrade and bigrade conjugations.


(8)

163

Unger’s (1993[1977]) reconstruction of quadrigrade and lower bigrade stems:

a. b. c. d. e. f.

Quadrigrade ‘open’ Pre-OJ Conclusive*aka-u > Irrealis *aka-aX > Infinitive *aka-i > Adnominal*aku-re-u > *akuru > Realis *aku-re > Imperative *aka-Ø >

OJ aku akaX aki aku akure ake

a. b. c. d. e. f.

Bigrade ‘open it’ Pre-OJ Conclusive*aka-gi-u > *akagu > Irrealis *aka-gi-aX > *akagiX > Infinitive *aka-gi-i > *akagi > Adnominal*aka-gu-re-u > *akaguru > Realis *aka-gu-re > Imperative *aka-gi-Ø >

OJ aku akeX ake akuru akure akeØ(yo)

Unger proposes that the derivational source is an “enlarging clitic” *gi, “cognate with the passive/causative marker of Korean (Martin 1972), or a monosyllable of the shape *Ce” (1993:70). As we see below, the second of these proposals is quite close to Yoshida’s and the view that I advance in this chapter. Note that the final vowel in quadrigrade roots plays a rather minor role in Unger’s reconstruction. This vowel is always lost before the Conclusive, Irrealis, and Infinitive suffixes in (8a-c). Here Unger follows Yamaguchi’s (1971, 1985) proposal that in an early stage of pre-OJ derived V1-V2 sequences, V1 is lost if the root is longer than the suffix. This ensures that the hypothesized root-final vowel is lost in, e.g., ‘doesn’t open’ (*aka-azu > akazu), but retained in ‘doesn’t come’ (*ko+azu > kozu). In a later stage of pre-OJ, (i) *g is lost and (ii) V-V sequences contract (see Frellesvig & Whitman, this volume). This is the stage where the bigrade Infinitives are formed, as shown in (9): (9)

Unger’s (1993[1977]) derivation of bigrade Infinitives Hiatus loss of *g contraction a. Upper bigrade *oko-gi-i > *okogi > *oko-i > b. Lower bigrade *aka-gi-i > *akagi > *aka-i >

okwi akey

In more recent work, Unger (2001) relaxes the claim that all quadrigrade roots were originally vowel-final. He continues to posit an original vowel in roots with a final coronal consonant, to block the effects of the ‘strong palatalization’ process hypothesized by Whitman (1985). In other quadrigrade roots, a final

164

JOHN WHITMAN

vowel may or may not be present. However the other features of Unger’s earlier reconstruction are retained. Unger’s reconstructions represent a major step forward in that they account for the derived nature of the bigrade conjugations. But they still fail to account for a number of facts. First, the transitivity patterns of lower and upper bigrade stems are different. The lower bigrade conjugation includes both transitives and intransitives, as we saw in (1-4); the former are somewhat more numerous. As observed by Kida (1988:83), however, upper bigrade verbs include only intransitives (1988:88). Thus okwi- ‘arise’, tukwi- ‘get used up’, sugwi- ‘pass, exceed’, and amwi- ‘bathe’ are all intransitive. Kida notes two possible exceptions: todomwi- ‘bring to a stop’ and komwi- ‘enfold’.1 But as Kida points out (ibid), both of these verbs have lower bigrade alternants: todome- and kome-. It is the latter that survive in later Japanese, and have the wider distribution in OJ. The second problem with Unger’s reconstruction concerns the specific proposal that the ‘enlarging clitic’ is cognate with the Korean passive/causative marker, MK -Gi- < *-ki-. The Korean morpheme appears to have originally been a causative suffix, subsequently extended to a passive function, although both functions already exist in the first hangŭl texts of the 15th century. It derives true causatives, with agent causees (10a), and passives with (animate) experiencer surface subjects (b). (10)

a. Emeni ka ay hanthey mother NOM child to ‘The mother fed the child.’

pap ul mek-i-ess-ta. food ACC eat-CAUS-PAST-DEC

b. So ka holangi eykey mek-hi-ess-ta. cow NOM tiger by eat-pass-pasT-DEC ‘The cow was eaten by the tiger.’ Neither of these patterns is attested with Japanese bigrade stems. Japanese bigrade intransitives are anticausatives, not passives; therefore they disallow expression of an agent, overt or implicit. Thus (11) disallows expression of an agent: (11)

*Hihu ga gekai ni sakeru to kizu ga hukaku naru. skin NOM surgeon by split when wound NOM deep becomes ‘When the skin is split by the surgeon, the wound gets deeper.’

Likewise, bigrade transitives do not occur with agent causees; that is, they do not entail a true causative structure involving two agents and two events. OJ did have lower bigrade stems which are superficially causative-like in that they appear to 1 Bigrade todomwi is restricted to the MYS poems of Yamanoue no Okura and may reflect an orthographic idiosyncrasy of his. There are two more possible upper bigrade transitives; but todu/todi ‘shut (door)’ lacks a clear OJ phonographic attestation, while kozu/kozi ‘pull out by the roots’ is attested only in the infinitive.


165

add an additional argument, such as paku/pake ‘(cause to) strap on’, (12). (12)

usi ni koso pana napa pak-ure cow DAT EMPH nose rope strap.on-REAL ‘On the cow strap on the bridle’ (MYS 16.3886)

Even in a case like (12), however, the bigrade transitive does not result in a causative with a volitional (i.e., agentive) causee. Thus (12) means “put the bridle on the cow”, not “make the cow put on her bridle”. A third problem for the comparison with the Korean passive/causative marker is that in many cases, the root selected by the ancestor of the Japanese bigrade endings was an adjective; in these instances, the meaning of the derived Japanese bigrade verb is never causative (as is the case in Korean), but rather inchoative. Examples of this type include aku/ake ‘dawn, redden’, from aka- ‘red’ in (5-7), asu/ase ‘get shallow’, from asa- ‘shallow’; opu/opwi ‘grow’, from opo‘big’, sabu/sabwi ‘behave quietly’ from sabu- ‘lonely’, and many others. Adjectives are generally taken to have been uninflecting in pre-OJ. The formative involved in the derivation of the bigrade stems must have been able to select adjectival roots as complements, and turn them into inchoative verbs. Typological parallels become relevant here: while passives derived from inchoatives and causatives derived from passives are robustly attested across languages, inchoatives derived from passives or causatives are not. This suggests that the base meaning of the formative involved in the derivation of the bigrade conjugations was inchoative. A fourth problem that must be explained by any reconstruction of the bigrade stems is the statistical imbalance between upper and lower bigrade classes. Lower bigrade stems are much more numerous: Frellesvig (this volume) states that they make up 20% of all OJ verbs; quadrigrade stems make up 75%, but Frellesvig counts only approximately 30 upper bigrade stems. In terms of absolute numbers, Kida counts 1,295 quadrigrade stems, 452 lower bigrade stems, and 83 upper bigrade stems in OJ (1988:78-9).2 We would like to find a reconstruction that explains both the low type frequency of this conjugation and its restriction to intransitives. The facts summarized above indicate that whatever morpheme was involved in the derivational relationship linking quadrigrade and bigrade stems, it was not a simple “transitivity flipper”, nor was it a causative/passive morpheme like Korean -hi/ki/li/i-. Upper bigrade stems do not participate in the transitivity alternation at all. And for many bigrade verbs, the derived interpretation was inchoative, not transitive or intransitive. 2 Kida’s larger numbers (based on a tally of entries in Omodaka et. al, 1967) result from counting separately verbs involving the same derivational element: thus deadjectival verbs in upper bigrade -mu/-mwi are each counted separately. The point is the same, however: upper bigrade stems are much less frequent than the other two classes.

JOHN WHITMAN

166 3.

A “new” old proposal Yoshida (1973:85-6) proposes that the verb e- ‘get, obtain, be able to’ is the source of the bigrade conjugation. He refers to an earlier hint along these lines by Takeda (1957:201). E- is rather rare in OJ texts in phonographic attestations, except in its Infinitive form e. It occurs as a potential auxiliary, less frequently in its base lexical meaning ‘get’. It is classified as a lower bigrade verb; in fact its shape consists exactly of the lower bigrade endings: (13)

The conjugation of e- ‘get’ (lower bigrade) Conclusive Irrealis Infinitive Adnominal Realis Imperative

u (?B) e e uru ure e(yo)

Yoshida and Takeda both remark on the absence of OJ phonographic attestations of conclusive u. Infinitive and Irrealis e and Adnominal uru are attested, the latter only in the function of potential auxiliary. Takeda’s speculation about the relation between u/e and the bigrade conjugation is brief: “There are cases where it would appear that verbs in the lower bigrade conjugation with other finals may include this word” (1957:201). Yoshida is more explicit: “U exists as an independent word, but from a phonological and semantic standpoint it may be supposed that this element is incorporated in the rest of the lower bigrade conjugation. It is the original morpheme of the lower bigrade formation” (1973:87-8). Neither Takeda nor Yoshida spell out how this verb, grammaticalized as a suffix, might derive the functions and forms associated with the bigrade endings. Let us consider form first. Martin (1987:681 and 1996:13) proposes that e‘get’ is built on the same root a- that occurs in ar- ‘exist’, with the latter derived by an original suffix in *-r-. On this view, Infinitive (and Irrealis) e is derived by contraction with *-i: *a+i > e. Whether Martin’s etymology is correct or not, I will assume, following Frellesvig (this volume) that the stem of e-, like the other OJ monosyllabic stems in (5), ‘come’ and ‘do’, was identical to its Irrealis form, that is e-. This means that if e- is derived from *a+i > *ay > *ey, contraction occurred much earlier than the period under discussion. Let us examine this proposal concretely, contrasting e- and ku/ki ‘come’. (14)

e- ‘get’ a. Conclusive *e-u > u b. Irrealis *e-aX > eX c. Infinitive *e-i > e

ku/ki ‘come’ *ko-u > ku *ko-aX > koX *ko-i > ki

In the Conclusive and Irrealis (a-b), the outcomes are the same for both verbs. In the Infinitive (c), ‘come’ deletes the stem vowel, while ‘get’ appears to


167

delete the suffix. Relevant here is a crucial difference between the two stems: part of the stem (the onset /k/) remains after deletion of the stem vowel with ‘come’, but the stem would be completely obliterated by the same strategy with e-, as indeed occurs in the Conclusive. We can explain this pattern with a modification of Yamaguchi’s (1971, 1985) constraints on stem and affix contraction, mentioned in §2. Yamaguchi’s constraints make reference to the relative length of the stem and suffix. But another way to understand the pattern in (14) is in terms of two different dimensions of stem diminution: reduction and elimination. When stem or suffix is reduced to avoid vowel hiatus, as in the Irrealis pattern (14b), the stem is favored. But when the only alternative to hiatus is complete elimination of stem or suffix, the suffix wins out, as in the conclusive form of e- (14a). This last is reminiscent of the Realize Morpheme constraint proposed by Kurisu (2001): this constraint penalizes derivations that leave no trace in the output corresponding to a morpheme in the input.3 I will propose that Realize Morpheme can be subdivided to refer specifically to stem or affix. The paradigms in (14) can be captured by the constraints informally stated in (15), ranked in the order *Hiatus > RMaffix > RMstem > *Diphthong > Faithstem > Faithaffix. (15)

a. *Hiatus: *VV b. Realize Morpheme (stem/affix): A (stem/affix) morpheme must receive some realization in the output. c. *Diphthong: *Vy d. Faithfulness (stem/affix): Every segment of the input (stem/affix) must have a correspondent in the output.

Ranking RMaffix above RMstem ensures that the suffix surfaces when it is composed of a single segment, as in the conclusive forms (14a). 4 Realize Morpheme also applies in the case of the Infinitive suffix -i (14c), straightforwardly in the case of ‘come’; here *Diphthong rules out an Infinitive of the form *koy or *kwi. In the Infinitive of ‘get’ however, deletion of the stem vowel would violate a higher ranked constraint, RMstem, by completely eliminating the stem. Diphthongization allows this constraint to be satisfied, resulting in *e-i > *ey > OJ e. On this view, diphthongization is a ‘last resort’ strategy for maintaining material from stem and suffix.5 3

I am grateful to Abigail Cohn for bringing Kurisu’s work to my attention. The Adnominal forms of ‘come’ and ‘do’ in (5) may at first seem to require a modification of (15); taking ‘come’ as an example, the stem vowel in ko+uru appears to be deleted despite the fact that the suffix -uru is two syllables. The same problem arises with the Realis. In Whitman (2004), however, I argue that the pre-OJ shape of the Adnominal suffix was *-ur or *-or. I then suggest that assimilation of the stem vowel *ko+ur > *kuur followed by metathesis > kuru is the strategy that circumvents hiatus in the adnominal form. I also propose that the Realis base is derived from the Adnominal, an idea dating back to Ohno (1953) (6). The same result is obtained if the Adnominal is analyzed as attaching to Conclusive *-u as in Ohno and Unger’s reconstructions. 5 The same strategy is not available for conclusive -u because pre-OJ did not have labial off-glides in Vw. (15c) specifically disallows diphthongs of shape Vy; constraints on other kinds of 4

JOHN WHITMAN

168

Finally, Stem Faithfulness (15d) mandates loss of the suffix initial vowel when (15a-c) are satisfied. On this treatment, OJ verbal paradigms result from the interplay of two constraints, one a constraint banning the complete elimination of morphological material, favoring affixes over stems, and the other a constraint on reduction (partial loss) of material in the morpheme, favoring stems over affixes. The distinctive paradigm of ‘get’ results from the combination of a monosegmental stem and monosegmental affixes. Diphthongization allows retention of stem and suffix in the Infinitive; subsequent monophthongization makes the Infinitive and Irrealis identical, a distinctive characteristic of the bigrade conjugations. In the Conclusive, Adnominal, and Realis the stem is obliterated. It seems possible that the non-attestation of Conclusive u (and the rarity of Adnominal uru) in OJ noted by Takeda and Yoshida may have been related to the nontransparency of these forms, which retain no material from the stem. U remerges in the reading tradition for early Heian kundoku material, but this may be a learned restoration.6 Let us now apply Takeda and Yoshida’s proposal that the lower bigrade endings originate from an auxiliary derived from ‘get’, attached to other stems and eventually reanalyzed as the bigrade endings. Note that this proposal already captures the same insight as Unger’s reconstruction: that the bigrade conjugation is derived. There are two cases that we should consider: vowel stems such as adjectival aka- ‘red’ in (5-6), and cases where no evidence exists to support reconstruction of a stem-final vowel. Actually, Unger’s example ‘open’ in (8) is such a case: no related embedded root form exists to support reconstruction of *aka- ‘open’, and Unger’s revised (2001) reconstruction does not require that we posit one. Of course we must be cautious in such cases; absence of a related embedded root form in any given case may be just an historical accident. Nevertheless I will assume (like Ohno and Unger) that original consonant-final verb stems did exist. The most plausible assumption is that ‘get’ functioned as a fully conjugating auxiliary when it underwent univerbation with the lexical verb stem. The combination of auxiliary ‘get’ with consonant-final stems is straightforward: u/e attaches to the stem to derive the homophonous set of bigrade endings: (16) a. Conclusive b. Irrealis c. Infinitive

*ak- ‘open’ + u/e ‘get’ *ak + u > aku *ak + eX > akeX *ak + ey >*akey > ake

diphthongs were thus higher ranked. The view of diphthongization in the verbal paradigm as a last resort strategy clearly cannot apply to other instances of diphthongization in pre-OJ, as these include compounds of shape *CVCV+iCV > CVyCV (see Frellesvig, this volume). 6 For example, conclusive u occurs 26 times (3 times in combination with the modal adjective besi) in the reading tradition for the character ‘get, can’ in the early Heian kunten annotated text Konkômyô saishôô kyô (Kasuga 1985).


169

To explain the combination of ‘get’ with vowel-final stems, however, we must posit a phonological change. Unger (1993:41) argues that at an earlier stage (Unger’s proto-Japanese), vowel hiatus was resolved by contraction; that is, by deletion of one of the two vowels in sequence. At a later stage, diphthongs come into being, subsequently these are monophthongized. Without going into the details of Unger’s argument here, I adopt the view that the domain for diphthongization is expanded in pre-OJ.7 This can be made concrete by positing a re-ranking of the constraints in (15) whereby the Stem and Affix Faithfulness constrains come to outrank the constraint prohibiting diphthongs: *Hiatus > RMaffix > RMstem > Faithstem > Faithaffix > *Diphthong. This ranking now correctly predicts the paradigms for lower bigrade verbs derived from vowel stems: (17)

*aka- ‘red’ a. Conclusive *aka + u b. Irrealis *aka + eX c. Infinitive *aka +ey

+ u/e ‘get’ > > *akayX > *akeyX > *akay > *akey

aku > akeX > ake

The Conclusive form in (17a) follows directly from Realize Morpheme (15b). The Irrealis and Infinitive forms require that we assume that both *V-i and *V-e resulted in diphthongs of shape *Vy, or that mid vowel raising (Frellesvig & Whitman, this volume) preceded the formation of the the bigrade paradigms. The diachronic derivation for lower bigrade verbs from original vowel stems also applies to derive upper bigrade verbs: (18)

*opo- ‘big’ + u/e ‘get’ a. Conclusive *opo + u > b. Irrealis *opo + eX > *opoyX c. Infinitive * opo + ey > *opoy

> >

opu opwiX opwi

In (18) as well, the Conclusive form follows from Realize Morpheme. The Irrealis and Continuative forms follow from diphthongization as a strategy to satisfy both Realize Morpheme and Stem Affix Faithfulness. The derivation in (18) makes an important prediction: upper bigrade verbs are derived only from original vowel-final stems, specifically stems ending in *o (or its ancestors) or *u. Consonant stems result only in a lower bigrade pattern, as in (16). This provides an explanation for one of the observations made at the end of section 2: the relative paucity of upper bigrade verbs. It suggests an explanation for the second major observation in section 2 about upper bigrade verbs: the fact that they are limited to intransitive inchoatives. We return to this issue below. Yoshida comments only briefly that the meaning of *e- plausibly contributes to the semantics of the lower bigrade stems. His insight is correct; the 7

In Unger’s (1993) framework, the domain for diphthongization is created by voiced medial consonants loss. See Frellesvig (this volume) for arguments that diphthongization (‘contraction’) is to be treated as fundamentally different in kind from deletion.

170

JOHN WHITMAN

semantics of ‘get’ are perfect for explaining the three lexical subtypes instantiated by lower bigrade verbs. ‘Get’ > inchoative (change of state) as in *aka- + e- > ake- ‘get red’ is widely attested in the grammaticalization literature (Heine & Kuteva 2003:144-145, Anderson 1975). Heine & Kuteva observe that “this process appears to be associated primarily with contexts where ‘get’ has adjectives and related words as complements” (2003:145). As we observed in §2, this is exactly what occurs in Japanese: inchoative bigrades result when the original complement of *e- ‘get’ was an adjectival stem. Heine and Kuteva also present crosslinguistic evidence for the developments ‘get’ > passive (2003:144147) and ‘get’ > permissive causative (145-6). The examples of both types cited by Heine & Kuteva involve analytic formations where ‘get’ remains an independent word, and most appear to be true passives and causatives, with expressible agents and causees. A case more directly comparable to the Japanese situation may be the proto-Indo-European affix *-ne-, which, reanalyzed in Germanic as a suffix -nā/-nə, derives inchoative verbs from adjectives and inchoative intransitives (unaccusatives) from transitive verbs (description and examples from Prokosch 1939:156-7; see also Suzuki 1989): (19)

a. Gothic wakan, Old Norse vaka ‘wake’ :: Go. -wak-na-n, ON vak-na ‘wake up’ b. Go. fulls ‘full’ :: full-na-n ‘fill upINTR’

Although an independent lexical source for *-ne- is unclear, Anderson (1975:24) suggests a relationship with PIE *neud- ‘obtain something desired, use’. More generally, we may supposed that an analytic grammaticalization of get (such as in English) produces full passives or causatives, while more complete grammaticalization to affixal status might be expected to result in more lexicalized inchoatives, anticausatives, or intransitives. Let us now return to the semantics of the upper bigrade stems. A straightforward explanation of the limitation of the upper bigrade endings to an inchoative function is that inchoative *e- selected only adjective (or stative substantival) stems, while transitive and intransitive (anticausative) *e- selected verbs. It is perhaps premature to speculate about the diachronic relation between the inchoative and the other two functions of *e-, but a possible scenario emerges from an inspection of the OJ CV- bigrade stems. There are four CV- upper bigrade stems: pwi- ‘dry up’, pwi- ‘sneeze’, mwi- ‘turn’ wi- ‘sit, be at’. ‘Sneeze’ is an inchoative; the original stative root *pois detectable as a hidden root form in pos- ‘dry it’, but does not survive as an independent word. Unger (1993:116) suggests that pwi- ‘sneeze’ is a semantic extension of ‘dry up’. Mwi- is an achievement verb whose root *mo- is embedded in motopor- ‘go back’ < *mo+topor- ‘pass through’ (Unger 1993:28). Like the other three verbs, wi- ‘sit, be at’ is in the process of being reanalyzed as a monograde verb in the 8th century and is much more abundantly attested in its monograde form; but Kinsui (2006) argues persuasively that wi- is the nonstative


171

counterpart of stative wor- ‘sit, be at’. Kinsui adopts the analysis that wor- is derived from *wo+ar(-i) ‘exist, be at’ (see also Kinsui 1998). On the analysis of upper bigrade verbs proposed here, we would expect wi- to be derived from *wo + inchoative e-, deriving a change of state counterpart to stative wor-. This seems to fit well with the aspectual difference between wi- and wo- that Kinsui argues for. The facts are similar with monosyllabic lower bigrade CV- stems: ke‘vanish’, ne- ‘sleep’, pe- ‘elapse’, and the OJ perfective auxiliary te- and the two medio-passive auxiliaries ye- ~ re-. The first and third of these and the auxiliaries are transparent inchoatives. ‘Sleep’ is an activity predicate, and thus varies from the overall pattern in a manner similar to ‘sneeze’, suggesting an earlier semantic shift. What is most striking about the bigrade CV- stems is that all of them are intransitives, and most are inchoatives or achievements. This is particularly striking given that transitives are more numerous among lower bigrade verbs as a whole. If the causative/transitive function of *e- was available when this verb first developed as an auxiliary or higher predicate, we would expect at least some of the bigrade CV- stems to be transitive. When we investigate how transitives were formed from these roots, we find that they are formed with the transitivizing suffix *-s-: pos- ‘dry it’, nas- ‘put to bed’. The scenario suggested by the monosyllabic bigrade stems is the following: deriving an intransitive change of state (inchoative) predicate was the earliest function of *e-. This function was extended to other kinds of intransitives without clear adjectival roots, such as wi- < *wo+e-. Semantic extensions to nonachievement intransitives may have occurred along the lines of ‘dry up’ > sneeze. In this process, *e- came to attach to verbs; when the original verb root happened to be transitive, *e- had an intransitivizing function. See Gronemeyer 1999 for an argument that English passive get developed from inchoative get. As for the causative/transitive function of *e-, it is unclear whether this developed directly from the lexical meaning of ‘get’ or from one of the derived intransitive functions discussed above. What does seem clearer about transitive/causative *-e is that it was restricted by the category and/or shape of the root it combined with. Adjective roots, which uniformly end in a vowel, are always transitivized by suffixing -s-. The same is true of CV-verbal roots, as we saw above. Among longer derived transitives we find both bigrade stems and stems in -s-, but the two are in near-complementary distribution. This suggests a development somewhat similar to the semantic expansion of *e- above: *-s- was used first as a transitivizer for adjectives (or more broadly, stative substantivals). Later it was extended to vowel-final verbal stems more broadly, but until the development of the productive -ase- causative after the 8th century, there was no morphological device for attaching -s- to consonant stems. Transitivizing *e- may have developed as a suppletive device for deriving transitives from verb stems ending in a consonant. If this is correct, the distribution of derived -s- transitives and bigrade transitives may provide a hint as to earlier root structure: -s- betokens an original vowel root, -e- a consonant root. Over time, the root structure and from class

172

JOHN WHITMAN

restrictions on the two suffixes may have been loosened, but we might expect to find a semantic reflection of the original contrast. This is supported by the small number of roots that occur with both -sand -e-. Take, for example the root *kap- ‘exchange. ‘ In addition to the quadrigrade stem kap- ‘mingle, exchange, buy’ in OJ, this root displays embedded root forms of shape kapa- in transitive kapas- ‘exchange (actions), do reciprocally’ and transitive kapar- ‘change, take the place of’. But it also has a lower bigrade transitive kape- ‘change it’ Taken together, these forms suggest a CVCV- root *kapa-; the original root final vowel is lost in the quadrigrade, as in Unger’s reconstruction (8). The derived transitive in -s (*kapa-s-) and the derived bigrade transitive kapu/kape < *kapa+e both ‘increase’ the transitivity of the original root, in the sense of Jacobsen (1988), but in different ways. Kapasimplies no change of state on the part of the patient (if any), except a change of location. Kape- entails that the patient is changed. In this sense kape- shows a higher degree of transitivity, as we might expect from a transitivizer restricted to selecting verbs. The two suffixes also suggest a difference in degree of lexicalization. Kapas- typically appears in verb-verb compounds (mi+kapas‘exchange looks’, yobi+kapas ‘call one another’), while kapu/kape functions as an independent verb, suggesting that it may be the later formation.8 On the view that I have sketched above, the lower bigrade transitives result from a suppletive strategy for transitivizing consonant-final verb roots that were unable to host transitive -s. If this view is correct, presence of a lower bigrade transitive is a general diagnostic for a primary *CVC- root, except for a small number of cases where the bigrade transitive was extended to *CVCV- roots such as *kapa-, to represent a semantic distinction with the original transitive in -s. This view may also provide an explanation for the exceptional intransitives mentioned in §2, todomu/todomwi ‘bring to a stop’ and komu/komwi ‘enfold’. The first of these verbs occurs only in two songs by Yamanoue no Okura (MYS 4.605 and 4.657; see fn. 1); elsewhere only the lower bigrade variant todomu/todome occurs. The second occurs only in the first song of the Kojiki; in the corresponding song in the Nihon shoki, lower bigrade komu/kome occurs in its place. We have good evidence for an original CVCV root *komo- (viz. komor- ‘be hidden, snuggled). As the pattern of transitives derived with *e- was extended to CVCV roots, some varieties may have followed the earlier pattern for CVCV + *e- for adjectival roots in (18), while most varieties followed the established pattern for attaching *-e only to CVC- roots. 4.

Conclusion In this chapter I have addressed the longstanding problem of the historical source for the two main conjugations in Japanese. The basic hypothesis that I develop dates back to Takeda (1957). It holds that the source for the bigrade 8

As noted above, verb roots selected by both suffixes are rare, but another possible example is nagas- ‘let flow, drain’ and nagu/nage ‘throw’ from naga- ‘elongated’. Here too the bigrade derivative is higher in transitivity.


173

endings is the verb *e- ‘get’, which fused with verbal and adjectival roots to derive transitives, anticausatives and inchoatives. I have argued that this source accords well with the lexical patterning of bigrade verbs (inchoatives from adjectives, unaccusatives from transitives, transitives from intransitives). I have shown how it can be understood to derive not only lower bigrade stems in -e but upper bigrade stems in -i < wi, and suggested why the latter derivation is largely limited to inchoatives (because it derives from adjectives) and smaller in number (because this is its only source). Finally, I have suggested that the opposition between transitives in -s- and lower bigrade transitives may indicate a difference in original root shape.

CHAPTER 9 ON RECONSTRUCTION OF PROTO-JAPANESE AND PRE-OLD JAPANESE VERB INFLECTION BJARKE FRELLESVIG University of Oxford & University of Oslo 1.

Introduction As is well known, Old Japanese has eight distinct verb classes. At least since Unger (1993 [1977]) it has been generally acknowledged that the bigrade verbs are derived and secondary, in the sense of incorporating originally derivational matter, but the further implications of this have not previously been fully explored. In §§2-3, I outline some preliminaries: different types of ‘processes’ applying to sequences of vowels (§2), and OJ verb classes, pJ root shapes, and the etymology, or sources, of the bigrade verbs (§3). In the main part of the chapter, I first present data which show that the bigrade verbs belong to a different, younger, chronological layer within Japanese than the quadrigrade verbs (§4). This has important implications for reconstruction of pJ and pre-OJ verb inflection, entailing that reconstruction of early inflected forms and verb paradigms be based exclusively on the forms of the primary verb classes. The forms of the secondary verbs do not incorporate the same morphological material as the quadrigrade verbs and must have arisen through analogy (§5). 2.

Vowel deletion versus contraction It has long been observed that OJ and pre-OJ (surface) phonotax did not tolerate sequences of vowels. It is necessary to distinguish two different ways of ‘resolving’ vowel sequences, both of which are found within earlier Japanese. 2.1

Synchronic rules of vowel deletion First, there is the set of phonological rules in (1). These are productive, synchronic, generative rules, applying to underlying form in the course of speech production in order to give phonotactically acceptable surface forms. They were active in OJ and in pre-OJ as phonological rules, and they remained in MJ as morphophonemic rules. In short, when two vowels came together in the concatenation of a vowel-final (grammatical or lexical) morpheme with a vowelinitial morpheme to form one word, one of the vowels was deleted. It was

BJARKE FRELLESVIG

176

generally the first vowel which was deleted (1)b. The second vowel was only deleted when a monosyllabic vowel-final morpheme was followed by a polysyllabic vowel-initial morpheme, (1)a. Cf. Unger (1993 [1977]:41ff.) (1)

a. V1-V2 => V1: CV1-V2CV wa-ga-ipye I-GEN-house b. V1-V2 => V2 elsewhere: i. CV1-V2 ko-i come-INF

=>

CV1CV wagapye my.house

=>

CV2 ki come.INF

ii.

CVCV1-V2 ake-u dawn-CONCL

=>

CVCV2 aku dawn.CONCL

iii.

CVCV1-V2CV waga-ipye my-house

=>

CVCV2CV wagipye my.house

myesi-age deign.to.see-raise 2.2

myesagesummon

Diachronic laws of vowel contraction Second, there is the set of pre-OJ diachronic contractions which gave rise to secondary vowels and diphthongs, (2). The diachronic correspondences established by these sound changes are shown by ‘>’. See Frellesvig & Whitman (this volume) for detail about pJ and pre-OJ vowels and about these changes, but note here that all instances in OJ of -wi and -e (distinct from -ye) are secondary and reflect contractions of earlier sequences, whereas only some instances of -wo and -ye are secondary in this sense: the majority reflect raising of pJ mid-vowels pJ *o > OJ u, wo, and *e > i, ye. (2)

Pre-OJ

OJ

a. *ui, *î *ai, *əi

> >

wi e

b. *iˆ, *iə, *ia *uˆ,*uə, *ua

> >

ye wo

PROTO-JAPANESE AND PRE-OLD JAPANESE VERB INFLECTION

177

As opposed to the vowel deletion rules in (1) above, these diachronic laws of contraction are sound changes; the result of reanalysis through time in the course of language acquisition. This took place where vowels came together in a single morpheme, either through sound change (consonant loss: ..VCV.. > ..VV..), or as a result of univerbation due to a loss of constituent transparency (..V-V.. > ..VV..). 2.3

Deletion versus contraction The two ‘processes’ above are entirely different phenomena and must be distinguished sharply (as represented here by the use of different arrows, ‘=>’ and ‘>‘ respectively): The former are ‘deductive’ and the latter basically ‘abductive’ (Andersen 1973). They are often descriptively acknowledged as being distinct, but are mostly presented as having different temporal, or phonological and/or morphological, scope of application (most recently Russell 2003), not as being different in nature (except in a general remark by Hattori 1979:660ff). I shall in the following sections make reference to and make use of these two different types of ‘processes’. They have both provided important input into Japanese verbal morphology, but in different areas: the vowel deletion rules have determined the shapes of inflected forms, as in the examples above with the Infinitive and the Conclusive; the diachronic vowel contractions, on the other hand, were instrumental in the formation of the secondary (bigrade) verb stems (cf. 3.1 and 4.2). As an example which reflects both processes, consider the form myesage- ‘summon’: in (3)a, myes- < *mi-as- represents a diachronic reanalysis and lexicalization of the surface form */mias-/ (from a juxtaposition of *mi- and *as-), and age- involves a diachronic contraction, -e- < *-ai, typical of the formation of the bigrade verbs (cf. further 3.1, 4.2); in (3)b, myesi-age => myesage- is a generative process which takes place in the course of speech production,1 conforming to (1)b.iii above. (3)

myesage- ‘summon’ a. myes- ‘deign to see’ < *mias- < *mi-as- ‘see-HON’ age- ‘raise’ < *agai b. myesage- CVCV-. Until recently, this latter had become the prevalent view, adopted for example in JLTT. However, Whitman has argued persuasively, in lectures and talks since the late 1990s (e.g., Whitman 1999) and K-Irr hold only one verb each, and N-Irr only the auxiliary and two lexical verbs listed. UU verbs are not thought of as irregular in Japanese school grammar, although they clearly are irregular vis-à-vis the larger, regular vowel base classes, UB and LB. The OJ UM verbs are: i- ‘cast (metal)’, i- ‘shoot’, ki- ‘put on’, mi- ‘see, look’, mwi- ‘turn’, ni- ‘resemble’, ni- ‘boil’, wi- ‘lead, bring along’, wi- ‘sit’. Note that UM verbs have monosyllabic basic stems. 3 This idea of the bigrade verbs incorporating a formant of a shape like *-(C)i- whose function is to switch (or ‘flip’) the basic transitivity of the verb goes back, of course, further than Unger: see Miller (1967:65f and 364) for references.


179

and appearing for the first time in print in this volume, that both quadrigrade and bigrade verbs have both open and closed root ancestors, such that OJ quadrigrade verbs reflect both CVC- and CVCV- roots which at some point were restructured uniformly as CVC- (see 4.4.1), whereas bigrade verbs reflect CV-, CVC-, and CVCV- roots augmented with a formant which Whitman reconstructs as *-e(primarily an inchoative formant, secondarily a transitivizer; cf. Whitman (this volume) for details). The differences between Ohno, Unger, and Whitman may be set out as in (5). (5) OJ

pJ

Quadrigrade

CVCsak- ‘bloom’

*CVC*sak-

Bigrade lower upper

(CV)CVake- ‘dawn’ opwi- ‘grow’

*(CV)CV- + -i- (‘INF’) *aka-i *ɨpɨ-i (*ɨpɨ- > OJ opo-)

CVCyak- ‘burnTR’ tuk- ‘attachINTR’

*CVCV*yaka*tuka-

(CV)CVyake- ‘burnINTR’ tuke- ‘attachTR’ opwi- ‘grow’

*(CV)CV- + *-gi(‘transitivity flipper’) *yakai < *yaka-gi *tukai < *tuka-gi *ɨpɨi < *ɨpɨ-gi

CVC-

*CVCV- / *CVC-

ok- ‘put’

*ək-/ɨk- (*ək-, *ɨk- > ok-) *mapa-

Ohno (1953)

Unger (1993 [1977]) Quadrigrade

Bigrade lower upper Whitman (1999, this volume) Quadrigrade

map- ‘dance’ Bigrade lower upper

(CV)CVake- ‘dawn’ tuke- ‘attachTR’ okwi- ‘rise’

*(CV)CV-/*CVC- + -e (< *e-i ‘get-INF’) *aka-e *tuk-e *ˆkɨy < *ɨkɨ-e (*ɨkɨ- > oko-)

Whitman’s account of the sources of the bigrade verbs is superior to previous reconstructions. As, however, I disagree on some specific but important points of reconstruction, especially concerning the formants involved in the formation of the bigrade verbs, I give my own understanding of the sources of the bigrade

180

BJARKE FRELLESVIG

verbs here, (6).4 I adopt two main points of Whitman’s (this volume), viz. that pJ verb roots have both open ((CV)CV-) and closed (CVC-) members, and that a formant grammaticalized from an ancestral form of the OJ lexical verb e- ‘get’ played an important role in the derivation of bigrade verbs. Whitman reconstructs this formant as *-e- < *e-i ‘get-INF’, thus doing away with a separate bigrade ‘theme’ *-(C)i-, and has *-e- take part in the derivation of all bigrade verbs. It is with these two latter points I disagree. In my view, bigrade verbs diachronically comprise two main groups, (6)a and (6)b, of which only the latter involve a formant related to OJ e-. I believe that it is premature to abandon a bigrade theme of a shape something like *-(C)i-, distinct from the Infinitive formant *-i-. The function of this formant is difficult to pin down, but it had nothing to do with transitivity. It derived inchoative verbs in general, and in particular from adjective roots, but in some cases seems to have been a vacuous extension. A small group of bigrade verbs, (6)a, involves direct lexicalization of this theme. This group comprises all UB and some LB verbs; it has a high proportion of bigrade verbs derived from adjectives, but holds for example also the verb e- ‘get’ (< *a-(C)i-). The second and much larger group, (6)b, comprises most LB verbs, viz. those which take part in transitivity alternations. The verbs in this group are derived by the suffixation of a formant -e- which served to derive transitivity alternants from (ancestors of) quadrigrade verbs; whether this derivational pattern only emerged after the (ancestors of) quadrigrade verbs had uniformly been reanalyzed as CVCstems (see 4.4.1), is an open question at this point, but in any case the pattern remained productive into OJ and later. The formant -e- was, as per Whitman, grammaticalized from the (ancestor of the) lexical verb e- ‘get’, which, in my view, < *a-(C)i- and itself incorporates the bigrade theme *-(C)i-. This group of verbs may therefore be said to involve indirect lexicalization of the bigrade theme. Briefly, my reasons for positing two different sources for bigrade verbs, rather than one uniform source, are (a) the phonological implausibility of getting the -wi of the UB verbs from *-u-e or *-ˆ-e, as in Whitman’s proposal, and (b), the widely different functional relationship between source and derivative in the two groups of bigrade verbs, (6)a and (6)b, including the fact that -e- remained productive in OJ and later as a transitivity formant, indicating that these were truly two different derivational patterns. Finally, it is possible that the bigrade verbs include a third, small group of verbs which do not incorporate the bigrade theme, but arose, as shown in (6)c, through reduction of Infinitive-like forms or forms with root-final *i (cf. Whitman 1985:34 about lenition before *i), followed by vowel contraction: *CVCi > *CVi > Cwi-, Ce-; or alternatively through weakening (yodization) of a root final consonant followed by contraction: *CVC > *CVy > Cwi-, Ce- (cf. Frellesvig & Whitman, this volume, §2.3). Such verbs will be difficult to identify internally, but candidates include short (monosyllabic) bigrade verbs. 4

Even if I disagree on some points, I want to acknowledge that Whitman’s work and insights, generously shared over the past years, have been formative for my own thinking about pJ and preOJ verbal morphology.


(6)

181

a. direct lexicalization of ‘theme’, *-(C)i- (?)’inchoative’ LB ake- ‘dawn, redden’ < *aka-(C)i-; cf. aka- ‘red’ UB opwi- ‘grow’ < *ɨpɨ-(C)i-; cf. opo-(< *ɨpɨ-) ‘big’ b. indirect lexicalization of theme; suffixation of -e- ‘transitivity flipper’ < e- ‘get’ < *a-(C)iyake- ‘burnINTR’ -tar-) was used with all conjugation classes and gradually replaced the morphological Stative through (E)MJ (see Sandness 1999:16ff). The degree of morphophonological fusion is high in the formation of the morphological Stative, which etymologically arose through the diachronic process of contraction (cf. 2.2) from univerbation of a construction with a verb Infinitive followed by the existential verb ar- (e.g., sakyer- ‘be blooming’ < *saki ar- ‘bloom be’). It seems clear that the bigrade verbs were only lexicalized after the morphologization of the morphological Stative and that the periphrastic Stative arose to supplement the morphological Stative for use with the newer verb classes; the periphrastic Stative which could be used with all verb classes then gradually replaced the older morphological Stative (and in turn changed itself through MJ from a construction (-te ar-) to a univerbated auxiliary (-tar-), eventually giving rise to the NJ past tense inflectional ending -ta). (9) Honorific Passive Passive Causative Stative

Honorific Passive Passive Causative Stative

QG yak- ‘burnTR’

R-Irr ar- ‘be’

UM ki- ‘wear’

N-Irr sin- ‘die’

yakasyakayeyakareyakaseyakyer-

arayearare-

kyeskiyekisekyer-

sinayesinare-

LB ake- ‘dawn’

UB okwi- ‘rise’

S-Irr se- ‘do’

K-Irr ko- ‘come’

-

-

sesser-

kyer-

morpheme -s- (itself related to the verb se- ‘do’), was on its way into the language, but had not yet emerged as a fully inflected auxiliary by OJ. 9 Furthermore, derived continuative verbs in -(a)p- are only derived from quadrigrade verbs, not from bigrade verbs. In general, derivatives are not formed on bigrade verbs to any great extent in OJ.

184

BJARKE FRELLESVIG

4.4

The Negative Using the case of the Negative auxiliary, -(a)n-, it is possible to show in a little more detail that the bigrade verbs are secondary and younger. 4.4.1 Concatenation with verbs. First of all, the phonemic shape of the base of the Negative auxiliary is simply /-n-/ in OJ. The notation -(a)n- is morphophonemic and intended to show that this auxiliary selects the derived stem in -a of consonant base verbs (e.g., yak.a- ) *yak-anV => yakanV (*yəsə-10 =>) *yəsə-anV => yosanV

yakanV => => => =>

*yakanV *yəsanV *aranV *minV *kɨnV

> > > > >

yakanV yosanV aranV minV konV

However, by the same phonological rules, polysyllabic bigrade verbs + -anV give the forms in (12), which are not the correct, attested forms. (12) LB

UB

CVCV-anV ake-anV are-anV

=> => =>

CVCanV *akanV *aranV

okwi-anV oti-anV

=> =>

**okwanV/*okanV *otanV

This means that either the rules are wrong or something else happened. The rules are pervasive in pre-OJ and OJ word formation, and there is no reason to assume that they are incorrect or somehow did not apply. What must have happened instead is that the bigrade verbs were lexicalized after the reanalysis of the Negative outlined above, and that the (reanalyzed) Negative /-n-/ simply attached to the base of the newly lexicalized bigrade verbs. (13)

LB

ake-nV are-nV

=> =>

akenV arenV

UB

okwi-nV oti-nV

=> =>

okwinV otinV

BJARKE FRELLESVIG

186

4.4.2 Reformation of paradigm. The Negative exhibits a composite paradigm in OJ, with variation between two basic variants, -(a)n- and -(a)zu- in some forms (Conclusive, Infinitive, and Gerund), representing a transition between older and younger forms: For all three forms, the -(a)n- based variants went out of use in the transition from OJ to EMJ, yielding to the -(a)zu based variants. A new Gerund is found from EMJ. (14) Conclusive Adnominal Realis Infinitive

OJ (composite) paradigm -zu ~ -nu -nu -ne -zu ~ -ni

EMJ (suppletive) paradigm -zu -nu -ne -zu

Gerund Concessive Provisional Conditional

-zute ~ -nito -nedo -neba -zupa

-zute ~ -de -nedo -neba -zupa

Nominal

-naku

-naku

The two basic variants of the OJ Negative paradigm belong to different conjugations: -(a)n- conjugates like a quadrigrade verb (with a notable irregular Gerund, see Frellesvig 2001:12), (15);11 we may therefore hypothetically posit a regular quadrigrade Conditional and a regular Gerund, marked with ‘*’. On the other hand, -(a)zu does not conjugate like a verb: morphologically, it resembles the forms within the adjectival paradigm built on the adverbial formant -ku; syntactically, it resembles an existential verb in using the Infinitive as Conclusive. (15)

11

Conclusive Adnominal Realis Infinitive

QG -(a)n-nu -nu -ne -ni

Innovative forms -zu

Gerund Concessive Provisional Conditional

-nito (~ *-nite) -nedo -neba *-naba

-zute

Nominal

-naku

-zu

-zupa

Thus -(a)n- is the only n-base quadrigrade verb; there are no lexical n-base quadrigrade verbs.


187

The innovative formants are usually thought to be diachronically derived as shown in (16), based on augmentation of the Infinitive, -ni, with a formant *-su,12 or simple contraction of a hypothesized Gerund *-nite.13 (16)

-zu < *-ni-su -zute < *-ni-su-te -zupa < *-ni-su-pa -de < *-nite

It is one thing, however, to understand the etymology of the innovative formants. To understand the motivation for the reformation of the conjugation of the Negative is an entirely different matter. In order to do so it is revealing to compare the forms of the Negative of quadrigrade and lower and upper bigrade verbs with the corresponding forms of the -(i)n- Perfective (which selects the Infinitive of consonant base verbs, and not the derived stem in -a-), see (17).14 With quadrigrade verbs, the inflected forms of the Perfective and Negative are distinct throughout, as these two suffixes select different stems of the consonant base verbs. However, with the bigrade verbs some of the inflected forms of the Perfective are homonymous with forms of the -(a)n- Negative. Those forms are underlined in (17) and it is easy to see that it is precisely those forms which were replaced. Thus, it may be surmised that the paradigm of the Negative was reformed in order to avoid pernicious homonymy between the forms of the Negative and the Perfective with the bigrade verbs.15 This reformation may hence ultimately be seen to have been motivated by the emergence and lexicalization of the bigrade conjugations, for it is only with this event that the homonymy in question comes into being. This, again, shows that the bigrade conjugations are 12

This *-su has traditionally been identified with the Conclusive of se- ‘do’, parallel to the ‘dosupport’ of negatives in Korean (e.g., V-chi antha ‘doesn’t V; V-Nominalizer doesn’t’, antha *peru; pwi- ‘dry (intr.)’: pwi+uru => *pwiru). However, Conclusive (pe+u (by (1)b.i) => pu; pwi+u => (*pwu =>) pu) + -ru/-re gives the correct forms (pu+ru => puru, pu+ru => puru). 17 Whereas the quadrigrade verbs do not have a segmental distinction between Conclusive and Adnominal in WOJ, it is well-known that EOJ usually is said to have exhibited such a difference, with Adnominal in -wo, as in tatwo tuku ‘rising moon’ (MYS 14.3476), distinct from Conclusive in -u. However, the data are not as straightforward as they are usually presented: the number of occurrences of the distinct EOJ Adnominal form in -wo is in fact quite small (see Fukuda 1965:497ff; cf. also Hagers 2000), in contrast to the many occurrences of forms of the same shape as the WOJ Adnominal/Conclusive in -u in syntactic environments (including adnominal modification) requiring an Adnominal form. Furthermore, there are examples of the -wo form in conclusive function, especially of grammatical morphemes. Fukuda proposes (1965:402) that the -wo form reflects an older uniform Adnominal and Conclusive which was in the process of changing to -u, and not a distinct earlier Adnominal. Particularly in view of what we now know about midvowel raising (see Frellesvig & Whitman, this volume), this is a plausible proposal. However, whether one agrees with Fukuda (as I do) or accepts the conventional view that Eastern OJ reflects an original distinction in quadrigrade verbs between Conclusive and Adnominal (most recently reconstructed by Hino (2003) as *-u versus *-o, with raising of the Adnominal formant to merge eventually with the Conclusive), it is clear that the bigrade Adnominal reflects distinct morphological material (quite likely the obsolete Genitive formant -rwo- ~ -ru-, as originally proposed by Ohno (1953, 1978)).


191

from the same morphological material. They belong to different chronological layers in the language and their forms must be reconstructed differently. The forms of the secondary verb classes must have their origin in analogy in a wide sense. It seems, in addition, that the bigrade verbs have exerted analogical influence on the irregular vowel stem verbs (K-Irr, UM, and S-Irr), probably due to the common synchronic property of being vowel base verbs, and also on the formation of the Adnominal and Realis of the N-Irr verbs. (19) Base

quadrigrade yak- ‘burn (TR)’

R-Irr ar- ‘be’

N-Irr sin- ‘die’

derived stem

yaka-

ara-

sina-

Infinitive Imperative

yaki yakye

ari are

sini sine

Conclusive Adnominal Realis

yaku yaku yake

ari aru are

sinu sinuru sinure

Base

LB ake- ‘dawn’

UB okwi- ‘rise’


ake ake(yo)

okwi okwi(yo)


aku akuru akure

oku okuru okure

Base

K-Irr ko- ‘come’

UM mi- ‘see’

S-Irr se- ‘do’


ki ~ koko

mi mi(yo)

si ~ sese(yo)


ku kuru kure

miru miru mire

su suru sure

192

BJARKE FRELLESVIG

The major reconstructions which have been proposed up to now (Ohno 1953, 1978, Unger 1993 [1977], 2000b, Whitman 1985, Miller 1986a) have each been influential in various ways and have each, if to different extents, provided new perspectives and insights on the prehistory of Japanese verb inflection. However, they are all predicated upon the premise of deriving the inflected forms in all conjugation classes from the same morphological matter subjected to sound changes, and are characterized by a high degree of effort and ingenuity in devising sets of underlying forms and rules applying to them in order to arrive at the forms of verbs of all eight conjugation classes. These reconstructions must now be revisited critically, in light of the fact that the bigrade verbs belong to a younger chronological layer, as outlined above, and with a view to incorporating its implications for the reconstruction of pJ and early pre-OJ verb inflection.

CHAPTER 10 THE NOMINAL AND ADNOMINAL FORMS IN OLD JAPANESE CONSEQUENCES FOR A RECONSTRUCTION OF PRE-OLD JAPANESE SYNTAX

JANICK WRONA Kyoto University 1.

Introduction This chapter shows that a careful examination of the usages of the Nominal and Adnominal forms in Old Japanese can shed light on pre-OJ syntax, in particular the complement system. The Nominal and Adnominal forms are essential to a proper understanding of OJ syntax, as they are the two main strategies for encoding subordination. Each form has a number of usages which may be assumed to be diachronically related, though this cannot always be ascertained in the historical records. In §§2 and 3, I will briefly describe each of the usages of the Nominal and Adnominal forms and look at their diachronic development. In §4, I will compare the usages, making a distinction between contexts in which the two forms are in opposition and the ones in which they are in competition, and show that the usages of the Nominal and Adnominal forms were originally in opposition. In §5 I will deal with the consequences for a reconstruction of the pre-OJ system of complementation. 1.1

Background information and theoretical assumptions OJ refers to the language of the 8th century represented in a corpus consisting mainly of poetry, the main source being the Man’yôshû. It may be argued that poetry is an unsuitable genre for a syntactic investigation, but I believe it is justified in the present study for two reasons. Firstly, it is unlikely that the choice of complementizer is affected by genre. Secondly, it is impossible to gain a sufficiently detailed picture of the complement system using only OJ prose sources. The empirical foundation of this study is based on a comprehensive analysis of the OJ corpus, wherein all instances of the Nominal and Adnominal forms were recorded. Since OJ was often written in a combination of phonographic and logographic writing, a minimum requirement was that at least the final syllable of the predicate was phonographically written. Otherwise the

JANICK WRONA

194

example was excluded. All examples cited in this chapter meet this requirement. To show this important difference, all cited text examples are transcribed with logographic text portions in plain type and phonographically written text in italic type in this chapter. I make a distinction between pre-OJ and proto-Japanese. Pre-OJ, in the sense used here, means the period of the Japanese language between pJ and OJ. The period is characterized by a stable system of complementation after the morphologization of the Nominal form, which was a syntactic construction in pJ (see §2). What I have in mind is a purely “functional reconstruction” of the preOJ complement system based exclusively on internal evidence. This amounts to a kind of internal reconstruction, but only on the content side. It means that based on the usages of complements and the morphological forms involved in them in OJ, one can make certain careful inferences about the usages in the period preceding OJ, but following pJ. This procedure is quite different from that of reconstruction by comparison of two or more languages or dialects. 2.

The Nominal form Since Ohno (1957) it has been widely accepted that the Nominal form is based on the Adnominal form plus a morpheme *-aku possibly with the meaning ‘place’ or ‘(abstract) thing’. In (1) I have given the Conclusive, Adnominal, and Nominal form of verbs, adjectives, and auxiliaries. (1)

Conjugation Quadrigrade Lower bigrade Upper bigrade Upper monograde K-irregular S-irregular Adjective Simple past

Conclusive form kaku ‘write’ tatu ‘raise’ kwopu ‘love’ miru ‘see’ ku ‘come’ su ‘do’ takasi ‘is high’ ki

Adnominal form kaku taturu kwopuru miru kuru suru takaki si

*-aku -aku => -aku => -aku => -aku => -aku => -aku => -aku => -aku =>

Nominal form kakaku taturaku kwopuraku miraku kuraku suraku takakyeku siku/kyeku

The evidence on the content side of the reconstruction is rather weak, but from a functional and a morphophonological point of view it does seem to be the case that the Nominal form goes back to a relative construction involving the Adnominal form. The attestation of the Nominal form is limited to OJ, with a few frozen expressions surviving into Early Middle Japanese. It is also striking that the Nominal form has not been preserved in any modern Japanese dialects, though it is attested in OJ dialect poems. Thus it would appear that the Nominal form was once widespread, but disappeared completely. In (2) I have listed the various usages of the Nominal form dividing them into complement and non-complement usages.

THE NOMINAL AND ADNOMINAL FORMS IN OLD JAPANESE

(2)

Complement Subject Object

195

Non-complement Quotation opening device Quotation closing device Adverbial clause Exclamative reaction Exclamative (aku-dome) Topic Adnominal

2.1

Subject complements The Nominal form basically nominalizes a predicate or a clause (Wrona, 2005). Thus it may be the subject or the object of adjectives or verbs.1 There are six semantic classes of complement-taking predicates selecting Nominal complements as their subject: Commentative, Quantitative, Knowledge and Acquisition of Knowledge (KAK), Immediate Perception, Negative, and Desiderative (cf. Noonan 1985). I have given an example from each of the classes in (3). (3) kimi ni kwopuraku you DAT love.NML ‘that I love you’

wosi ‘is regrettable’ (Commentative) tabimanesi ‘is frequent’ (Quantitative) wasurayu ‘forget’ (KAK) miyu ‘is visible’ (Immediate Perception) apida nasi ‘is constant’ (Negative)

kimi ni kwopwimaku posi ‘is desirable’ (Desiderative) you DAT love.CONJ.NML ‘to love you’ The propositions expressed by Nominal complements are presupposed and discourse-dependent, except for the complement of desiderative predicates (see Wrona, forthcoming, for an explanation). Nominal complements selected by Negative predicates may not be complement constructions at all. Typologically, they are rare (Noonan 1985) and OJ already has a far more productive synthetic negative. Wrona (2003) argues that Nominal complements selected by Negative predicates are scope adjustments rather than complement constructions proper. It is particularly important to note that Nominal complements are selected not by the negative predicate alone, but by a complex predicate consisting of a temporal noun+nasi ‘not exist’. This means that the negation does not have scope over the Nominal complement, but only over the temporal noun. Thus these construction resemble more closely the group of complement-taking predicates called Phasal predicates (Noonan 1985).

1

I will follow the majority of non-generative work on complementation and preserve the term subject complement.

JANICK WRONA

196 2.2

Object complements Nominal complements may also be the object of transitive adjectives and verbs. In (4) I have given an example from each of the eight semantic classes of complement-taking predicates. The propositions expressed by the Nominal complements are presupposed and discourse-dependent, except for the complement of desiderative, negative predicates and possibly a few immediate perception contexts. (4)

kimi ni kwopuraku you DAT love.NML ‘that I love you’ kimi ni kwopwimaku you DAT love.CONJ.NML ‘to love you’

wosimi ‘consider regrettable’ (Commentative) siru ‘know’ (KAK) miru ‘see’ (Immediate Perception) ipu ‘say’ (Utterance) omopu ‘think’ (Propositional Attitude) todomu ‘stop’ (Phasal) porisu ‘want’ (Desiderative) subye nami ‘means is non-existent’ (Negative)

Many of the semantic classes of complement-taking predicates may select both subject and object Nominal complements as summarized in (5). The first four classes are derivationally related. (5)

Complement-taking predicate Intransitive

Transitive

Commentative

kurusi- ‘is awful’

kurusimi ‘think awful’

Immediate Perception

miye- ‘is visible’

mi- ‘see’

Desiderative

posi- ‘want’ (INTR)

poris- ‘want’ (TR)

Negative

na- ‘not exist’

nami ‘think id.’

Knowledge

wasuraye- ‘forget’

sir- ‘know’

2.3

Pseudoclefts and abstract nominalizations Another type of construction in which the Nominal complement can be said to be the subject is in specificational pseudoclefts. (6)

[poyo torite kazasi-turaku]-pa poyo take.GER adorn-PERF.NML-TOP so

titose poku to 1000-years pray.CONC COMP

FOC

‘That I picked the poyo and put it in my hair was to pray to live a 1000 years’ MYS 18.4136


197

The difference between the pseudocleft construction and the cases of complement constructions is that in the former there is no complement-taking predicate. Pseudoclefts involving the Nominal form focus adjuncts, predominantly of the category CP or PP (see Wrona (2005) for details). An adjective in the Nominal form may serve as an abstract nominalization. (7)

[asikyeku]-mo [yosikyeku]-mo mi-mu bad.NML-FOC good.NML-FOC see-CONJ ‘I would see good (things) and bad (things)’ MYS 5.904

It is important to emphasize that neither the complement in pseudocleft constructions nor the abstract nominalization in (7) constitute headless relative clauses. 2.4

Quotation-opening The Nominal form may be used as a quotation opening device, so that the utterance verb in the Nominal form precedes the quotation. Since OJ is a rigid SOV language, this may come as a surprise. It has been suggested that the construction has been borrowed from Chinese. This is possible, but it is clear that the construction serves a specific purpose in OJ. Its function is to establish the interlocutors as in (8) below. (8)

[wagimo-kwo-ga papa-ni my girl-HYP-GEN mother-DAT ‘My girl told to her mother: “….” MYS 9.1809

kataraku] tell.NML

“….” to COMP

This has the advantage of supplying the addressee with important information and with a backdrop to the actual quotation itself. 2.5

Quotation-closing The Nominal form may also close a quotation as in (9) below. The Nominal form is underlined. (9)

“… samu-bou-no opo-mapye-ni kasikwomi-mo 3-relics-GEN HON-front-LOC in awe-FOC kasikwomi-mo mawositamapaku]” to] mawosu. in awe-FOC say.NML COMP say ‘“I (the Emperor) speak in awe, in awe in front of the Three Relics ….” Thus I (a messenger) humbly speak.’ SM 12.7-8

198

JANICK WRONA

Precise conditions for the occurrence of the Nominal form in this context are impossible to formulate, but it would appear that the Nominal form closes a complex and lengthy quotation. Length is of course a relative concept, but complexity seems to have to do with the presence of multiple quotations. It is not uncommon to find quotations within quotations in the OJ prose texts and the least embedded quotation is frequently closed with a Nominal form. 2.6

Adverbial usage The Nominal form may encode an adverbial subordinate clause, most commonly resulting in a concessive relationship between protasis and apodosis as in (10) below. (10)

omopu swora yasukye-naku ni, nageku swora think.ADN feeling easy-NEG.NML COP lament.ADN feeling yasukara-nu mono wo easy-NEG.ADN but ‘my feeling (when I) think (of it) is not at ease, my lamenting feeling is not at ease’ MYS 4.534 (cf. MYS 8.1520)

2.7 Exclamative reaction The Nominal form can be used to express an exclamative reaction. This usage is formally characterized by a sentence-final Nominal form followed by the subordinating morpheme ni, presumably the same as ni as in the adverbial usage in (10). Exclamative reactions are most commonly focused reactions to an assertion made immediately prior to the exclamative reaction as in (11) below. Typically, there is a contrastive or causal relation between the two clauses involved. (11)

2.8

udimayama asakaze samusi. Tabi-nisite Mt Udima morning-wind is-cold.CONC journey-LOC koromo kasu-beki imo-mo ara-naku ni garment lend-CAB wife-FOC exist-NEG.NML ‘the morning wind on Mt Udima is cold. (But) I do not even have a wife who could lend me garments on the journey!’ MYS 1.75

Exclamative Exclamatives are similar to exclamative reactions in some respects: the Nominal clause is sentence final, and they can be exclamative reactions to a state of affairs (SoA) in a preceding clause. However, the exclamative clause is not in a contrastive or causal relation with the preceding clause, nor is it marked by ni.


(12)

199

tamamatu-ga ye-pa pasiki ka mo. Kimi-ga Jewel pine-GEN branch-TOP dear.ADN FOC My lord-GEN mikoto-wo motite kaywopaku. hold.GER commute.NML words-ACC ‘The branch of the jewel pine is dear. It arrives bearing the esteemed words of my lord’ MYS 2.113

2.9

Topic The Nominal clause may be marked by the topic particle pa (> Modern Japanese wa) and have no discernible grammatical relation to the rest of the clause as in (13) below. Note also that the topic is right-dislocated. (13)

yuku pye-mo sira-zu, [wa-ga go.ADN direction-FOC know-NEG I-GEN ‘as for my loving , I do not know where to go’ MYS 11.2739

kwopuraku]-pa love.NML-TOP

2.10

Adnominal usage It is an important fact about the Nominal form that it does not generally have an adnominal function, i.e., it does not modify nominals. There are two examples in the OJ corpus where the Nominal form does modify a noun. (14)

[tat-eraku]-no taduki-mo sira-zu, [woraku]-no know-NEG sit.NML-GEN stand-STAT.NML-GEN means-FOC okuka-mo sira-zu know-NEG future-FOC ‘standing up I have no way out, sitting down I have no way out’ MYS 13.3272

There is, however, clear evidence that this is not a general property of the Nominal form. First, the Nominal form does not modify the noun directly, but through the mediation of the genitive particle no. This clearly proves that the Nominal form had a nominal function. Second, the nouns that the Nominal form modifies are peculiar. Taduki ‘means, way’ belongs to a class of nouns that are obligatorily modified. And the precise meaning of okuka is unclear. Finally, both examples are found in the same poem. This shows that the Nominal form did not have a general adnominal function in OJ. 2.11

Diachronic development of the usages of the Nominal form It is impossible to reconstruct the diachronic development of the functions of the Nominal form. Its usages are almost exclusively attested in OJ and the form has not been preserved in any dialects. Nor has a convincing Korean cognate been

JANICK WRONA

200

proposed. Thus the OJ data is the basis on which one has to extrapolate the diachronic development of the usages in OJ. The task is made all the more difficult because the Nominal form is already a fully-fledged form in OJ, so it is not possible to observe changes in usage in progress. If it is correct that the Nominal form is a univerbation of an Adnominal form followed by a (bound?) noun *aku ‘(abstract) thing’, then it must have happened before OJ, since there is no trace of its origin as a relative construction in the historically attested stages of Japanese. Given the importance of the Nominal form in OJ as an inflected predicate form, it is highly likely that it was treated as an inflected form of the predicate in pre-OJ as well. Whether the OJ usages of the Nominal form exhibit the full picture, or whether it was already in decline in OJ is difficult to assess. Considering the fact that the Nominal form disappears all but completely in Early Middle Japanese, it seems to follow that its usages and productivity must have been curtailed to some extent in OJ. A piece of evidence for this is that many examples appear to be of a quasi-frozen nature. (15)

[mi-maku] posikyedo see-CONJ.NML want.CONCES ‘Though I want to see you’ MYS 6.946

(16)

[asibwi-no tira-maku] wosi mo scatter-CONJ regrettable FOC ashibi-GEN ‘It is regrettable that the ashibi will scatter’ MYS 20.4513

The sequences mimaku posi ‘want to see’ in many of the desiderative constructions and tiramaku wosi ‘is regrettable that ... will scatter’ are particularly frequent. What and how many usages the Nominal form had in pre-OJ, we shall never know, but it is almost certain that it was a fully integrated member of the pre-OJ complement system and it remained an important form in the syntax of OJ. 3.

The Adnominal form As mentioned in §2, there is wide-spread agreement that the Nominal form is based on the Adnominal form+*-aku. Thus the Adnominal form is older than the Nominal form. I will come back to this in §4.1. In (17) below I have listed the various usages of the Adnominal form in OJ and divided them into complement and non-complement usages. (17)

Complement Subject Object

Non-complement Adjunct clause Kakari-musubi Exclamative (rentai-dome) Relative clauses Formal noun constructions


201

One important fact about the Adnominal complements is that they are rare compared with Nominal complements. 3.1

Subject complement Adjectives and verbs may take Adnominal complements as their subject.

(18)

Commentative. [kwopuru]-pa tomosi love.ADN-TOP painful ‘that I love is painful’ MYS 4.489

(19)

Knowledge and acquisition of knowledge. Asikaki-no kumatwo-ni tatite wagimo-kwo-ga swode-mo reed fence-GEN corner-LOC stand.GER my girl-HYP-GEN sleeve-FOC sipopo-ni [naki-si] so mopa-yu drench-COP cry-PST.ADN FOC think-MP ‘I come to think that my girl has cried, her sleeves being drenched standing at the corner of the reed fence’ MYS 20.4357

(20)

Immediate perception verbs. [sirwotape-no swode kapyesi-si]-pa ime-ni white-cloth-GEN sleeve turn-PST.ADN-TOP dream-LOC miye-ki ya Q visible-PST.CONC ‘was it visible in a dream that I had turned inside out my white-cloth sleeves?’ MYS 11.2812

There are not only far fewer examples of Adnominal subject complements compared with Nominal subject complement, but also just three semantic classes of complement-taking predicates. The proposition expressed by the Adnominal complement selected by Commentative and KAK predicates is presupposed and discourse-dependent. This is probably not the case with the example in (20) where the complement-taking predicate is an immediate perception verb. It is the only example in the corpus of an Adnominal complement selected by the intransitive perception verb miye- ‘is visible’. Object Adnominal complements selected by the transitive perception verb mi- ‘see’ denote direct perception (see §3.2), but (20) is slightly problematic because the matrix predicate is logographically written. If the reading tradition is correct in (20) then it is possible that (20) is also an example of a direct perception.

202

JANICK WRONA

3.2

Object complements The transitive counterpart to miye- ‘is visible’, mi- ‘see’ is more frequent, in fact it is the most frequent predicate selecting Adnominal complements. (21)

Immediate perception verbs. [kurenawi-no iro-mo uturopu] crimson-GEN colour-FOC change.ADN ‘Seeing the red colour change’ MYS 19.4160

mireba see.PROV

Adnominal complements selected by immediate perception verbs differ from other complement usages of the Adnominal form and from Nominal complements selected by immediate perception verbs in that they denote direct perceptions. Thus states of affairs (SoA) unfold as the subject is watching as in (21). A requirement for this reading is that the time of occurrence of the SoA in the subordinate clause can be construed as simultaneous with the matrix clause. The only other semantic class of predicates selecting Adnominal complements as objects is KAK verbs illustrated in (22). (22)

Knowledge and acquisition of knowledge verbs. [yamamori-no ari-kyeru] sira-ni mountain guard-GEN exist-MPST.ADN know-NEG ‘not knowing that there was a mountain guard’ MYS 3.401

3.3

Relative clauses By far the most frequent usage of the Adnominal form is to construct relative clauses. This goes for the primary strategy of relativization (the prenominal relative clause) as well as the less common circumnominal relative clause. The two types are illustrated in (23) and (24) (the relative clause is in square brackets and the head is underlined). (23)

(24)

Prenominal relative construction [[otitagitu] tagi-no kaputi]-pa waterfall-GEN stream-TOP cascade.ADN ‘though I see the stream of the waterfall cascade down’ MYS 6.909 Circumnominal relative construction [imo-ga ipi-si]-wo okite wife-GEN say-PST.ADN-ACC leave.GER ‘leaving behind my wife who protested’ MYS 20.4429

miredo see.CONCES


203

Relative constructions may also encode direct perception complements in OJ. (25)

[[suminoye-no oki-tu siranami kaze pukeba Suminoe-GEN offing-GEN white-waves wind blow.PROV kiyosuru] pama]-wo mireba approach.ADN beach-ACC see.PROV ‘when one looks at the beach where the white waves of the offing of Suminoe approach when the wind blows’ MYS 7.1158

The above example is ambiguous in the sense that the bracketed material may be interpreted as denoting a SoA or an entity taking part in one. Cross-linguistically, it is not uncommon to use relative constructions to denote direct perception. The same situation obtains in many other languages (Spanish, Italian and French (Felser 1999), Lori (Noonan 1985), Dyirbal (Dixon 1995)). When construed as encoding SoA, the relative construction belongs to the functional domain of complement clauses. 3.4

Pseudoclefts and headless relative clauses In §2.3 I mentioned that pseudoclefts, in which the subordinate predicate is in the Nominal form, focus adjuncts. When the subordinate predicate is in the Adnominal form only arguments are focused. (26) is an example of this. (26)

[[si-ga pa-no pirori-imasu] pro]-pa opo-kimi-ro-kamo leaf-GEN wide-HON.ADN (PRO)-TOP HON-lord-HYP-FOC 3sg-GEN ‘the one who is wide and calm is you my lord’ KS 57

Wrona (2005) argues that the Adnominal clause (delimited by square brackets) is a relative clause headed by a null pro. The same structure may be posited for other relative constructions where the head has been elided as in (27). (27)

so-no [[kanasiki] pro]-wo two-ni tate-me yamo DEM-GEN sweet.ADN (pro)-ACC outside-LOC stand-CONJ.REAL FOC ‘would (I) let that sweet one stand outside?’ MYS 14.3386

This should come as no surprise given the fact that the Adnominal form is used in relative construction (§3.3). 3.5

Adjunct clauses The Adnominal form is frequently used to form adjunct clauses of various kinds as in (28)-(33) below.

204

JANICK WRONA

(28)

Temporal allatives [wa-ga kuru]-made ni I-GEN come.ADN-ALL ‘until I come (home)’ MYS 20.4408

(29)

Temporal ablatives [wakare-ni-si]-ywori separate-PERF-PST.ADN-ABL ‘from when (we) had separated’ MYS 19.4220

(30)

Causal adjuncts [kimi-ga me-no kwoposiki] you-GEN eyes-GEN lovely.ADN ‘because your eyes are lovely’ NS 123

(31)

Simultaneity [wa-ga sekwo-ga koto I-GEN husband-GEN koto ‘as my husband takes the koto’ MYS 18.4135

kara ni because

toru] take.ADN

nape ni as

(32)

Ni-adjuncts [kimi-ni kwopuru]-ni kokoro-dwo-mo nasi you-DAT miss.ADN-NI heart-consoling-FOC not-exist When I miss you, there is no consoling my heart’ MYS 17.3972

(33)

Wo-adjuncts [tune naki koto-pa siru-ramu]-wo forever not-exist.ADN thing-TOP know-PCONJ.ADN-WO kokoro tukusu-na. heart exhaust-PROHIB ‘since you probably know that nothing is forever, do not exhaust your heart’ MYS 19.4216

It is noteworthy that many types of adjunct clauses involving the Adnominal form are constructed from bimorphemic grammaticalizations, (30), (31), of which at least one of the morphemes is of nominal origin. This may also be true of the case particles made (allative) and ywori (ablative) in temporal allatives and ablatives.


OJ morpheme made ywo/yu/ywori/yuri kara ni nape ni ni wo

205

proposed etymology < ma ‘interval’+nite copula, < mande (< maude<mawi-ide ‘come, go’ (Martin 1975:137)) < yuri ‘after’ (Zdb) < kara ‘character’? (Zdb)+ni copula ga/na (GEN)+pe ‘top, up’+ni copula copula Infinitive root related with wor-i ‘exist’?

Table 1: Etymology of adjunct morphemes 3.6

Kakari-musubi The kakari-musubi phenomenon is perhaps best viewed synchronically as a kind of agreement in which certain focus particles trigger a certain form of the predicate. So in (34) below the focus particle so triggers the Adnominal form naki of the adjective nasi ‘not exist’. (34)

mi-si pito-so naki see-PST.ADN person-FOC not-exist.ADN ‘people who saw it do not exist (anymore)’ MYS 3.446

Diachronically, however, kakari-musubi involves two interdependent morphosyntactic elements, the particle and the predicate form. No triggering was originally involved. On this analysis, the particular form of the predicate signals the non-declarative nature of the sentence, either a question or an exclamation while the focus particle signals the exact nature and scope of the question or exclamation. 3.7

Exclamative Given the proposed diachronic source of kakari-musubi constructions it should come as no surprise that one finds questions and exclamatives unaccompanied by focus particles as in (35)-(36). (35)

Exclamative wa-ga koromode-no puru toki-mo I-GEN sleeve-GEN dry.ADN time-FOC ‘there is no time when my sleeves are dry!’ MYS 10.1994

naki not-exist.ADN

206

JANICK WRONA

(36)

Question ide ika ni kokodaku kwopuru Oh why so much love.ADN ‘Oh, why do I love her this much?’ MYS 12.2889

3.8

Formal noun constructions, sa-derivations and gotosi Certain nouns in OJ obligatorily construct with a modifier. These are called formal nouns, exemplified in (37). (37)

Formal nouns a. wagimo-kwo-ni mise-mu-ga tame ni my girl-HYP-DAT show-CONJ.ADN-GEN sake ‘in order to show (it) to my dear girl’ MYS 19.4222 b. nara-no miyakwo-ni yukite ko-mu Nara-GEN capital-LOC go.GER come-CONJ.ADN ‘in order to go and come to the Nara capital’ MYS 5.806

tame sake

Tame ‘sake’ is a formal noun and there are two ways in which an adnominal clause may modify it: either through the mediation of the genitive particle ga (as in (37)a) or directly (as in (37)b). The latter is indistinguishable from relative clauses, but the former cannot be construed as such because of the genitive ga. (38)

Sa-derived nominals punapito-wo miru-ga tomosi-sa see.ADN-GEN enviable-SA boat-people-ACC ‘it is enviable to see the boat-people’ MYS 15.3658

(39)

Gotosi-construction puku kaze-no miye-nu-ga gotoku blow wind-GEN visible.NEG.ADN-GEN like yuku midu-no tomara-nu gotoku go water-GEN stop-NEG.ADN like ‘like the blowing wind is invisible and like the running water is ceaseless’ MYS 19.4160

In (38)-(39) we see two constructions similar to the formal noun construction. In (38) the nominal predicate tomosisa is derivationally related to the commentative adjective tomosi- ‘is enviable’. An Adnominal clause always attaches to a saderivative through the mediation of ga. The example in (39) involves the quasi-


207

free adjective gotosi ‘is like’ (cf. JLTT:820). Morpho-syntactically, Adnominal clauses are attached either directly or through the mediation of ga; both are exemplified in the example in (39). The constructions are ambiguous. They may be interpreted as a variant kind of relative clause as in the case of the formal noun construction in (37), or as subject complement clauses. These constructions are extremely important for the diachrony of the Japanese complement system. 3.9

Diachronic development of the usages of the Adnominal form It is easier to trace the functional development of the usages of the Adnominal form than the Nominal form. There is little doubt that the diachronically primary function of the Adnominal form is to construct relative clauses. The sheer number of relative clauses compared to other usages makes one suspect that this is the case. There are more than 2,000 relative clauses and that far exceeds the combined number of all other usages of the Adnominal form. Furthermore, there is no other productive way of forming relative clauses, nor does there appear to have been one, though there is some evidence that other strategies were possible. But more importantly, most of the usages of the Adnominal form can be directly derived from the relative clause usage. This is summarized in (40). Solid arrows designate a direct diachronic relationship and dotted arrows signify certain similarities between the two constructions which may have affected the development of the diachronically later construction. The schema in (40) is to be understood in such a way that the constructions at the top of the schema are temporally prior to the constructions lower in the schema. Each side of the schema should be seen as temporally independent. That is, the usages on the right hand side are diachronically independent from the ones on the left, so no claims as to the relative temporal relationship are made. The sole exception to this is the relationship between object complements and subject complements. While there is no direct diachronic relationship between the two usages (as indicated by the dotted arrow), there may be an indirect one and to all appearances subject complements developed later than object complements. Looking at the right side of the schema, the first step in the development of usages of the Adnominal form may not be a genuine diachronic development at all since null-pro headed relative constructions (§3.4) are just a subtype of relative constructions. It is likely that they developed simultaneously with other types of prenominal relative constructions. From relative constructions, the usage potential of the Adnominal form was extended to adjunct clauses. Some adjunct clauses are indistinguishable from relative clauses (kara, formal noun constructions), some involve relative clauses diachronically (nape) and still more are related to relative clauses (temporal allative/ablatives and ni-adjuncts). Temporal ablatives find relative construction counterparts, see (41).

JANICK WRONA

208 (40)

Relative clauses

Formal noun constructions (without ga) Formal noun constructions (with ga)

Null-pro headed relative clauses (incl. pseudo-clefts

Temporal allatives

Temporal ablatives

Gotosi/sa derivatives

niadjuncts

Object complements

Subject complements Time

(41)

[[…to ipi-si] toki]-ywori QUOT say-PST.ADN time-ABL ‘from the time when (they) said that…’ MYS 7.1311

It is not difficult to imagine a similar construction for the temporal allative. However, examples of this are rare. By far the majority of cases of allative made construct with predicates. There are cases of made constructing with nouns like pi ‘day’, yo ‘generation’ or other temporal nouns, but there is only one case in which made constructs with a relative construction. (42)

[[karigane-no ki-naka-mu] pi]-made wild-geese-GEN come-cry-CONJ.ADN day-ALL ‘Until the day when the wild geese come and cry’ MYS 10.2097

There could be several reasons for the scarcity of examples like (42). One is to assume that the grammaticalization of temporal allatives is more advanced than the grammaticalization of temporal ablatives. Though one has to exercise caution in arguing for a particular development based on numbers of examples, it is striking that the frequency of the four constructions discussed fits with the particular development argued for here. Thus while there are quite a few examples of temporal allatives and only one (clear) example of a relative construction, the opposite situation obtains with temporal ablatives: there are more relative constructions than adverbial constructions. This seems to indicate that it is in fact the case that the grammaticalization of temporal allatives is more


209

advanced than the grammaticalization of temporal ablatives. This is reflected in the schema in (40) above. Alternatively, it is possible that made retained its nominal character (ma ‘interval’+nite copula) while developing into a particle. On this latter interpretation, temporal allatives did not develop from a [[Adnominal form]N]-made construction, but nonetheless from a relative construction. One prominent usage of ni-adjunct clauses is as a temporal adjunct clause. There are numerous cases of relative constructions with virtually the same meaning as ones without the head noun toki ‘time’. (43)

[[pakari-si] separate-PST.ADN ‘when they separated’ MYS 2.167

toki]-ni time-COP

The fact that ni in these constructions already marks an adjunct is presumably a factor. Ni-adjuncts are usually said to have three meanings: temporal succession, causation and concession, but it should be emphasized that the semantics are not inherent to ni (‘fact’. This means that koto-complements compete with Nominal complements (and therefore also Adnominal complements) in certain contexts. Paratactic and participial complements constitute additional strategies with no unique functions (see Wrona (forthcoming) for an account of koto, paratactic and participial complements). The division of the complement system into a core and a peripheral part is intended to capture the aforementioned diachronic transition. The core system contains the members of the complement system with which all types of complementation in (pre-)OJ can be expressed. This also constitutes the diachronically primary system, i.e., the system that made up the pre-OJ complement system. As argued in §4, Adnominal complements play a minor part in OJ in the sense that they do not express anything that cannot be expressed by a member of the core system. The one complement usage of the Adnominal form that seems relatively well-established is direct perception selected by immediate perception verbs. As mentioned, relative constructions are also abundant in this usage. This fact, coupled with the strong evidence that all non-relative clause usages of the Adnominal form are derived, leads to the conclusion that at some point in pre-OJ Adnominal complements did not exist.3 The pre-OJ complement system is schematized in Table 5 below. Core system Nominal complements To-complements Relative constructions

Peripheral system (Paratactic) (Participial)

Table 5: The pre-OJ complement system. The important thing to emphasize is that the pre-OJ complement patterns are fully capable of expressing all the semantic categories necessary to the OJ complement system. The main distinction is between Nominal complements and tocomplements with relative constructions performing the role of expressing direct perception. With the emergence of Adnominal complements and the decrease of Nominal complements the complement system begins to change into the system one finds in OJ (Table 4). 6.

Conclusions In this chapter I have presented evidence that Adnominal complements are relatively recent innovations in OJ. I showed that the majority of usages of the Adnominal form can be shown to be diachronically related to relative

3

The same is true of koto-complements, but I will not address this issue here.


215

constructions. In the OJ complement system, Adnominal complements denote direct perceptions when selected by immediate perception verbs. In this context they are in synchronic variation with relative constructions. In other complement contexts they are in synchronic variation with Nominal complements. Eventually, Adnominal complements replace both Nominal complements and the role that relative constructions play in the complement system. The facts that (1) nonrelative usages of the Adnominal form can be derived from its usage in relative constructions and (2) Adnominal complements are not fully integrated members of the OJ complement system lead to the conclusion that Adnominal complements did not exist at some point in pre-OJ.

REFERENCES Akinaga, Kazue. 2001. “Akusento shûtoku hôsoku”. Shin-Meikai Nihongo akusento jiten ed. by Kazue Akinaga, 1-106. Tokyo: Sanseidô. Andersen, Henning. 1973. “Abductive and deductive change.” Language 49.76594. Anderson, Lloyd B. 1975. “Grammar-meaning universals and proto-language reconstruction, or Proto-World NOW!” Papers from the Eleventh regional meeting of the Chicago Linguistic Society ed. by Robin Grossman, James L. San & Timothy Vance, 15-36. Chicago: University of Chicago. Arisaka, Hideyo. 1975 [1931]. “Kokugo ni arawareru isshu no boin kôtai ni tsuite”. Kokugo on’in no kenkyû, 3-68. Tokyo: Sanseidô. Asato, Susumu & Naomi Doi. 1999. Okinawajin wa doko kara kita ka: "Ryûkyû = Okinawajin" no kigen to seiritsu. Naha: Bôdâ Inku. Baxter, William Hubbard. 1992. A handbook of old Chinese phonology. (= Trends in linguistics. Studies and monographs, 64.) Berlin: Mouton de Gruyter. Bloch, Bernard. 1946. “Studies in colloquial Japanese I: Inflection”. Journal of the American Oriental Society 66.97-109. Bybee, Joan, Revere Perkins & William Pagliuca. 1994. The evolution of grammar. Chicago: University of Chicago Press. Chamberlain, Basil Hall. 1895. Essay in aid of a grammar and dictionary of the Luchuan language. (= Transactions of TASJ, 23, Special Supplement.) Yokohama: Kelley and Walsh. Choy, Hakkun. 1987. Han’guk pang’ŏn sachŏn. Seoul: Myŏngmuntang. Crothers, John. 1978. “Typology and universals of vowel systems”. Universals of human language 2: Phonology ed. by Joseph Greenberg, 93-153. Stanford: Stanford University Press. Dixon, R. M. W. 1995. “Complement clauses and complement strategies”. Grammar and meaning: Essays in honour of Sir John Lyons ed. by F. R. Palmer, 175-220. Cambridge: Cambridge University Press. F&W = Frellesvig & Whitman 2004. Felser, Claudia. 1999. Verbal complement clauses: A minimalist study of perception constructions. Amsterdam: John Benjamins. Frellesvig, Bjarke. 2001. “A common Korean and Japanese copula”. Journal of East Asian Linguistics 10:1.1-35. Frellesvig, Bjarke & John Whitman. 2004. “The vowels of proto-Japanese”. Japanese Language and Literature 38.281-299. Frellesvig, Bjarke. 2007. “On the verb morphology of Old Japanese”. Frellesvig et al. 2007, 219-252. Frellesvig, Bjarke, Masayoshi Shibatani & John Charles Smith (eds.). 2007.

218

REFERENCES

Current issues in the history and structure of Japanese. Tokyo: Kurosio Publishers. Frellesvig, Bjarke & John Whitman. In press. “The Japanese-Korean vowel correspondences”. Japanese/Korean Linguistics 13. Fukuda, Yoshisuke. 1965. Nara jidai tôgoku hôgen no kenkyû. Tokyo: Kazama Shobô. GAJ = Kokuritsu Kokugo Kenkyûjo 1989-. Gronemeyer, C. 1999. “On deriving complex polysemy: the grammaticalization of get”. English Language and Linguistics 3.1–39. Hagers, Steven. 1997. “The importance of a Korean source on the Ryukyuan language”. Japan and Korea, contemporary studies: Proceedings of the Fourth Nordic symposium of Japanese and Korean studies ed. by Bjarke Frellesvig & Roy Starrs, 34-43. Aarhus: Aarhus University Press. Hagers, Steven. 2000. “The attributive and conclusive forms of Modern Japanese and Ryukyuan dialects in a historical perspective”. Studia Etymologia Cracoviensia 5.13-42. Hamano, Shoko. 2000. “Voicing of obstruents in Old Japanese: Evidence from the sound-symbolic stratum”. Journal of East Asian Linguistics 9:2.207-225. Hanihara, Kazuro. 1991. “Dual structure model for the formation of the Japanese population”. Japanese as a member of the Asian and Pacific populations ed. by Kazuro Hanihara, 244-51. Kyoto: International Research Center for Japanese Studies. Hashimoto, Shinkichi. 1917. “Kokugo kanazukai kenkyûshi-jô no ichi-hakken: Ishizuka Tatsumaro no Kanazukai no oku no yamamichi ni tsuite”. Teikoku Bungaku 23:11.120-154. Hashimoto, Shinkichi. 1949 [1931]. “Jôdai no bunken ni sonsuru tokushu no kanazukai to tôji no gohô”. Hashimoto Shinkichi hakase chosakushû 3: Moji oyobi kanazukai no kenkyû, 164-191. Tokyo: Iwanami. Hashimoto, Shirô. 1966. “Kodaigo no shiji taikei: Jôdai o chûshin ni”. Kokugo Kokubun 35:6.329-341. Hashimoto, Shirô. 1982. “Shijigo no shiteki tenkai”. Kôza Nihongogaku 2: Bunpôshi ed. by Kenji Morioka et al., 217-40. Tokyo: Meiji Shoin. Hattori, Shirô. 1959. Nihongo no keitô. Tokyo: Iwanami Shoten. Hattori, Shirô. 1967. “Ainugo no on’in kôzô to akusento”. Onsei no Kenkyû 13.207-223. Hattori, Shirô. 1976a. “Jôdai Nihongo no boin onso wa muttsu de atte yattsu de wa nai”. Gekkan Gengo 5:12.69-79. Hattori, Shirô. 1976b. “Ryûkyû hôgen to hondo hôgen”. Okinawagaku no reimei: Ifa Fuyû seitan hyakunen kinenshi ed. by Ifa Fuyû Seitan Hyakunen Kinenkai, 7-55. Tokyo: Okinawa Bunka Kyôkai. Hattori, Shirô. 1977. “Ryûkyû hôgen dôshi ‘shûshikei’ no tsûjiteki henka”. Gengo Kenkyû 72.19-28. Hattori, Shirô. 1978-9. Nihon sogo ni tsuite 1-22. Gekkan Gengo 7:1-7:3, 7:6-8:12.

REFERENCES

219

Hattori, Shirô. 1979. “Ôno Susumu ‘Dôshi katsuyôkei no kigen’ ni tsuite”. Nihon no gengogaku 4: Bunpô II ed. by Shirô Hattori, 658-663. Tokyo: Taishûkan Shoten. Hayata, Teruhiro. 1998. “Jôdai Nihongo no onsetsu kôzô to o-retsu kô-otsu no betsu”. Onsei Kenkyû 2.25-33. Hayata, Teruhiro. 2000. “The liquid and stem-final vowel alternations of verbs in Ancient Japanese”. Gengo Kenkyû 118.5-27. Heine, Bernd & Tania Kuteva. 2002. World lexicon of grammaticalization. Cambridge: Cambridge University Press. Hikosaka, Yoshinobu. 2001. “Kyûshû ni okeru katsuyôkei tôgô no moyô to sono kei’i”. Nihongo Kagaku 9.101-122. Hino, Sukenari. 2003. “Nihon sogo no boin taikei - jôdai azuma hôgen shiryô ni yoru saikô”. Vovin & Osada 2003. 187-205. Hirayama, Teruo. 1960. Zenkoku akusento jiten. Tokyo: Tôkyôdô Shuppan. Hirayama, Teruo, Ichirô Ôshima & Masachie Nakamoto. 1966. Ryûkyû hôgen no sôgôteki kenkyû. Tokyo: Meiji Shoin. Hirayama, Teruo, Ichirô Ôshima & Masachie Nakamoto. 1967. Ryûkyû Sakishima hôgen no sôgôteki kenkyû. Tokyo: Meiji Shoin. Hockett, Charles. 1958. A course in modern linguistics. New York: MacMillan. Hopper, Paul J. & Elizabeth Closs Traugott. 1993. Grammaticalization. Cambridge: Cambridge University Press. Hyman, Larry M. & Russel G. Schuh. 1974. “Universals of tone rules: evidence from West Africa”. Linguistic Inquiry 5:1.81-115. Inoue, Fumiko. 1998. Nihongo asupekuto no jittai. Tokyo: Akiyama Shoten. Jacobsen, Wesley. 1988. “Tadôsei to purototaipu ron”. Nihongogaku no shintenkai ed. by Susumu Kuno & Masayoshi Shibatani, 213-248. Tokyo: Kurosio Publishers. Jacobsen, Wesley. 1992. The transitive structure of events in Japanese. Tokyo: Kurosio Publishers. JLTT = Martin 1987. Kang, Sinhang. 1980. Kyerim yusa ‘Koryŏ pang’ŏn’ yŏn’gu. Seoul: Sŏnggyun’gwan University Press. Kasuga, Masaji. 1985. Saidaiji-bon Konkômyô saishôô kyô no koten no kokugogakuteki kenkyû. Tokyo: Benseisha. Kida, Akiyoshi. 1988. “Kodai Nihongo no saikôsei”. Nihon no kodai 14: Kotoba to moji ed. by Akiyoshi Kida, 35-110. Tokyo: Chûô Kôronsha. Kim, Chuwŏn. 1993. Moŭm chohwa ŭi yŏn’gu. Kyŏngsan: Yŏngnam University Press. Kindaichi, Haruhiko & Minoru Wada. 1955. “Kokugo akusento ruibetsu goihyô”, Kokugogaku jiten, 994-998. Tokyo: Tôkyôdô Shuppan. Kindaichi, Haruhiko. 1960. “Akusento kara mita Ryûkyû sho-hôgen no keitô”. Kindaichi 1975. 129-59. Kindaichi, Haruhiko. 1973. “Hikaku-hôgengaku to hôgen-chirigaku”. Kokugo to kokubungaku 50:6.1-19.

220

REFERENCES

Kindaichi, Haruhiko. 1974. Kokugo akusento no shiteki kôsatsu: genri to hôhô. Tokyo: Hanawa Shobô. Kindaichi, Haruhiko. 1975. Nihon no hôgen: Akusento no hensen to sono jissô. Tokyo: Kyôiku Shuppan. Kindaichi, Haruhiko. 1984. “Nihongo sogo no akusento to Ryûkyûgo”. Sophia Linguistica 17.3 -25. Kindaichi, Haruhiko & Minoru Wada. 1980. “Kokugo akusento ruibetsu goihyô”. Kokugogaku daijiten, 7-10. Tokyo: Tôkyôdô Shuppan. Kinsui, Satoshi. 2006. Nihongo sonzai hyôgen no rekishi. Tokyo: Hituzi Shobô. Kinsui, Satoshi. 1998. “‘Ari’ ‘wiru’ ‘wori’ – sonzai no hyôgen no igi”. Kokubungaku: Kaishaku to Kyôzai no Kenkyû 43:11.62-69. Kiyose, Gisaburô N. 1991. Nihongogaku to Arutaigogaku. Tokyo: Meiji Shoin. Kobayashi, Takashi. 1996. “Dôshi katsuyô ni okeru ra-gyô godanka keikô no chiriteki bunpu”. TôhokuDaigaku Bungakubu Kiyô 45.242-266. Kokuritsu Kokugo Kenkyûjo (eds.). 1989-. Grammar atlas of Japanese dialects. Tokyo: Kokuritsu Kokugo Kenkyûjo. Kokuritsu Kokugo Kenkyûjo (eds.). 1991 [1963]. Okinawago jiten. Tokyo: Ôkurashô Insatsu Kyoku. Kôno, Rokurô. 1979 [1959]. “Nihongo to Chôsengo no ni-san no ruiji”. Kôno Rokurô chosakushû 1: Chôsen gogaku ronbunshû, 557-62. Tokyo: Heibonsha. Kubozono, Haruo. 1999. Nihongo no onsei. Tokyo: Iwanami Shoten. Kurisu, Kazutaka. 2001. The Phonology of Morpheme Reduction. University of California at Santa Cruz dissertation. Lange, Roland. 1973. The phonology of eighth-century Japanese: A reconstruction based upon written records. Tokyo: Sophia University Press. Language Research Commission (Kokugo Chôsa Iinkai). 1906. Kôgohô bunpuzu. Tokyo: Kokugo Kyôkasho Kyôdô Hanbaisho. Lee, Ki-Moon. 1959. “On the breaking of *i in Korean”. Asea Yǒn’gu 2:2.131138. Lee, Ki-Moon. 1972a. Kug’ŏsa kaesŏl. Seoul: T’ap Ch’ulp’ansa. Lee, Ki-Moon. 1972b. Kug’ŏ ŭm’unsa yŏn’gu. Seoul: T’ap Ch’ulp’ansa. Lee, Ki-Moon. 1991. Kug’ŏ ŏhwisa yŏn’gu. Seoul: T’ap Ch’ulp’ansa. Mabuchi, Kazuo. 1968. Jôdai no kotoba. Tokyo: Tôbundô. Mabuchi, Kazuo. 1999. Kodai Nihongo no sugata. Tokyo: Musashino Shoin. Martin, Samuel E. 1966. “Lexical evidence relating Korean to Japanese”. Language 42:2.185-251. Martin, Samuel E. 1972. “A voiced velar stop for proto-Korean Japanese”. Lecture presented at the April Meeting of the American Oriental Society. Martin, Samuel E. 1975. A reference grammar of Japanese. Yale: Yale University Press. Martin, Samuel E. 1987. The Japanese language through time. New Haven & London: Yale University Press. Martin, Samuel E. 1992. A Reference grammar of Korean. Burlington, VT: Tuttle.

REFERENCES

221

Martin, Samuel E. 1996. Consonant lenition in Korean. Honolulu: University of Hawaii Press. Martin, Samuel E. 2000. “How have Korean vowels changed through time?”. Korean Linguistics 10.1-60. Mase, Yoshio (ed.). 1986. Ronshû Nihongo kenkyû 10: Hôgen. Tokyo: Yuseidô. Mase, Yoshio. 1992. Gengochirigaku kenkyû. Tokyo: Ôfûsha. Mase, Yoshio. 2003. Shinshû no kotoba. Nagano: Shinano Mainichi Shinbunsha. Mathias, Gerald B. 1973. “On the modification of certain proto-Korean-Japanese reconstructions”. Papers in Japanese Linguistics 2:21-47. Matsumori, Akiko. 1993. “Nihongo akusento no so-taikei saiken no kokoromi”. Gengo Kenkyû 103.37-91. Matsumori, Akiko. 1997. “Tokushima-ken Waki-machi, Mikamo-chô no akusento to hondo-sogo no akusento taikei”. Kokugogaku 189.15-28. Matsumori, Akiko. 1998. “Ryûkyû akusento no rekishiteki keisei katei: Ruibetsu goi 2-hakugo no tokui na gôryû no shikata o tegakari ni”. Gengo Kenkyû 114.85-114. Matsumori, Akiko. 1999. “Accentual reconstruction of the proto-system of mainland Japanese dialects”. Langue and parole in synchronic and diachronic perspective: Selected proceedings of the XXXIst Annual Meeting of the Societas Linguistica Europaea ed. by Christopher Beedham, 431-447. Oxford: Elsevier. Matsumori, Akiko. 2000a. “Ryûkyû akusento chôsa no tame no ruibetsu-goi no kaihatsu”. Onsei Kenkyû 4:1.61-71. Matsumori, Akiko. 2000b. “Ryûkyû no takei-akusento taikei ni tsuite no ichikôsatsu”. Kokugogaku 51:1.93-108. Matsumori, Akiko. 2001. “Historical tonology of Japanese dialects”. Proceedings of the symposium “Cross-linguistic studies of tonal phenomena: tonogenesis, Japanese accentology, and other topics” ed. by Shigeki Kaji, 93-122. Tokyo University of Foreign Studies: ILCAA. Matsumoto, Katsumi. 1974. “Kodai Nihongo boin soshiki kô: naiteki saiken no kokoromi”. Kanazawa Daigaku Hô-bungakubu Ronshû: Bungaku-hen 22.83-152. Matsumoto, Katsumi. 1984. “Gengoshi no saiken to gengofuhen”. Gengo Kenkyû 86.5-32. Matsumoto, Katsumi. 1995. Kodai Nihongo boinron: Jôdai tokushu kanazukai no saikaishaku. Tokyo: Hituzi Shobô. McCarthy, John & Alan Prince. 1995. “Faithfulness and reduplicative identity”. University of Massachusetts Occasional Papers 18: Papers in Optimality Theory ed. by J. N. Beckman, L.W. Dickey & S. Urbanczyk, 249-384. Miller, Roy Andrew. 1967. The Japanese language. Chicago: University of Chicago Press. Miller, Roy Andrew. 1980. Origins of the Japanese language. Seattle: University of Washington Press.

222

REFERENCES

Miller, Roy Andrew. 1986a. “A modest proposal on the origin of Japanese”. Origins of the Japanese language: An international collection of essays ed. by Kazuo Mabuchi, 57-103. Tokyo: Musashino Shoin. Miller, Roy Andrew. 1986b. In defense of Japanese. London: Athlone. Miyake, Marc Hideo. 1997. “Pre-Sino-Korean and pre-Sino-Japanese: reexamining an old problem from a modern perspective”. Japanese/Korean Linguistics 6.179-211. Miyake, Marc Hideo. 1999. The phonology of eighth century Japanese revisited: Another reconstruction based upon written records. University of Hawaii dissertation. Miyake, Marc Hideo. 2003. “Philological evidence for *e and *o in Pre-Old Japanese”. Diachronica 20:1.81-136. Mizushima, Yoshiharu. 1984. Man’yôshû azuma-uta honbun kenkyû oyobi sôsakuin. Tokyo: Kasama Shoin. Murayama, Shichirô. 1962. “Nihongo to Kôkurigo no sûshi – Nihongo no keitô mondai ni tsuite”. Kokugogaku 48.1-11. N&H = Nakahara & Hokama 1967. Nakahara, Zenchû & Shuzen Hokama (eds.). 1967. Omoro sôshi jiten, sôsakuin. Tokyo: Kadokawa Shoten. Nakamatsu, Takeo. 1987. Ryûkyû hôgen jiten. Naha: Naha Shuppansha. Nakasone, Seizen. 1983. Okinawa Nakijin hôgen jiten. Tokyo: Kadokawa Shoten. NHK Hôsô Bunka Kenkyûjo. 1998. Nihongo hatsuon akusento jiten. Tokyo: NHK Hôsô Bunka Kenkyûjo. NKBT = Iwanami (ed.). 1957-1969. Nihon koten bungaku taikei (100 vols. + 2 index vols.). Tokyo: Iwanami Shoten. NKD = Shôgakukan (ed.). 1972-6. Nihon kokugo daijiten. Tokyo: Shôgakukan. Noonan, Michael. 1985. “Complementation”. Language typology and syntactic description II ed. by Timothy Shopen, 42-140. Cambridge: Cambridge University Press. NSNT = Hattori 1978-9. OGJ = Kokuritsu Kokugo Kenkyûjo 1991 [1963]. Ohno, Susumu. 1953. “Nihongo no dôshi no katsuyôkei no kigen ni tsuite”. Kokugo to Kokubungaku 350.47-56. Ohno, Susumu. 1955a. “Dôshi”. Sekai gengo gaisetsu ed. by Shirô Hattori & Sanki Ichikawa, 277-279. Tokyo: Kenkyûsha. Ohno, Susumu. 1955b. “Man'yô jidai no on'in”. Man'yôshû Taisei 6.287-332. Ohno, Susumu. 1957. “Ku-gohô to iu koto”. NKBT 4.57-60. Ohno, Susumu. 1977. “On'in no hensen”. Iwanami kôza Nihongo 5: On’in, 147219. Tokyo: Iwanami Shoten. Ohno, Susumu. 1978. “Dôshi katsuyôkei no kigen”. Nihongo no bunpô o kangaeru, 179-222. Tokyo: Iwanami Shoten. Ohno, Susumu. 1980. Nihongo no sekai 1: Nihongo no seiritsu. Tokyo: Chûô Kôronsha. Ohno, Susumu et al. (eds.). 1990 [1974]. Iwanami kogo jiten. Tokyo: Iwanami Shoten.

REFERENCES

223

Omodaka, Hisataka et al. (eds.). 1967. Jidaibetsu kokugo daijiten: Jôdaihen. Tokyo: Sanseidô. Onishi, Takuichirô. 1997. “Atarashii hôgen to furui hôgen no zenkokubunpu”. Nihongogaku 18:13.97-110. Onishi, Takuichirô. 2002. Asakura Nihongo kôza 10: Zenkoku hôgen no bunpu. Tokyo: Asakura Shoten. Onishi, Takuichirô. 2004. “Dôshi ‘suru’ no Tôhoku hôgen ni okeru bunpu to kaishaku”. Kokugogaku Kenkyû 43.52-62. Oyler, Gary. 1997. The beginning of the /N/: In search of the origin of the nounfinal mora nasal in the language of Hateruma. University of Hawaii M.A. thesis. Polivanov, Evgenii D. 1968 [1924]. “K rabote o muzykal’noi aktsentuacii v iaponskom iazyke (v sviazi s malaiskimi)”. Stat'i po obshchemu iazykoznaniiu ed. by A.A. Leont'ev, 146-155. Moscow: Glavnaia redakciia vostochnoi literatury. Prokosch, Eduard. 1939. A comparative Germanic grammar. Philadelphia: Linguistic Society of America. Pulleyblank, Edwin G. 1991. Lexicon of reconstructed pronunciation in Early Middle Chinese, Late Middle Chinese, and Early Mandarin. Vancouver: UBC Press. Ramsey, S. Robert & J. Marshall Unger. 1972. “Evidence of a consonant shift in seventh-century Japanese”. Papers in Japanese Linguistics 1:2.270-295. Ramsey, S. Robert. 1979. “The Old Kyôto dialect and the historical development of Japanese accent”. Harvard Journal of Asiatic Studies 39:1.157-175. Ramsey, S. Robert. 1980. “Nihongo no akusento no rekishiteki henka”. Gekkan Gengo 9:2.64-76. Russell, Kerri. 2003. “Contraction and monophthongization in Old Japanese”. Vovin & Osada 2003, 511-38. Sandness, Karen. 1999. The evolution of the Japanese past and perfective suffixes. (= Michigan Monograph Series in Japanese Studies, 26.) University of Michigan: Ann Arbor: Center for Japanese Studies. Satô, Eisaku. 1985. “Kagawa-ken Ibuki-jima no akusento taikei o kangaeru”. Kokugogaku 140.31-44. Serafim, Leon A. 1977a. Front vocalism and palatalization in the Omoro sôshi: The philological and modern dialectal evidence. Unpublished manuscript. Serafim, Leon A. 1977b. “The relationship of pre-Japanese and proto-Japanese”. Paper presented at LSA Linguistic Institute summer Workshop on the relationship between Japanese and Korean. Honolulu, Hawaii. Serafim, Leon A. 1985. Shodon: The prehistory of a Northern Ryukyuan dialect of Japanese. Tokyo: Hompo Shoseki. Serafim, Leon A. 1999. “Reflexes of Proto-Korea-Japonic mid vowels in Japonic and Korean”. Paper presented at the Workshop on Korean-Japanese comparative-historical linguistics, XIVth International Conference on Historical Linguistics, 11 - 12 August 1999, University of British Columbia, Vancouver.

224

REFERENCES

Serafim, Leon A. & Rumiko Shinzato. 2000. “Reconstructing the proto-Japonic kakari musubi, … ka …-(a)m-wo”. Gengo Kenkyû 118.81-118. Serafim, Leon A. 2003. “When and from where did the Japonic language enter the Ryukyus? — A critical comparison of language, archaeology, and history”. Vovin & Osada 2003, 463-476. Serafim, Leon A. & Rumiko Shinzato. 2005. “On the Old Japanese kakari (focus) particle koso: Its origin and structure”. Gengo Kenkyû 127.1-49. Serafim, Leon A. 2007. “Progressive stative predicate extensions in Ryukyuan, and their relation to earlier Japonic”. Frellesvig et al. 2007, 207-218. Shibatani, Masayoshi. 1990. The languages of Japan. Cambridge: Cambridge University Press. Shimabukuro, Moriyo. 2002. A reconstruction of the accentual history of the Japanese and Ryukyuan languages. University of Hawaii dissertation. Shinzato, Rumiko & Leon A. Serafim. 2003. “Kakari musubi in comparative perspective: Old Japanese ka/ya and Okinawan -ga/-i”. Japanese/Korean Linguistics 11.189-202. Suzuki, Seiichi. 1989. The morphosyntax of detransitive suffixes -Þ- and -n- in Gothic: A synchronic and diachronic study. New York: P. Lang. Takeda, Yûkichi. 1957. Man'yôshû zenchûshaku 2: Gengo. Tokyo: Kadokawa Shoten. Thorpe, Maner L. 1983. Ryukyuan language history. University of Southern California dissertation. Tokugawa, Munemasa. 1962. “Nihongo sho-hôgen akusento no keifu shiron: ‘rui no tôgô’ to ‘chiriteki bunpu’ kara miru”. Gakushûin Daigaku Kokugo Kokubun Gakkaishi 6.1-19. Tokugawa, Munemasa. 1972. “Towards a family tree for accent in Japanese dialects”. Papers in Japanese Linguistics 1:2.301-320. Tokugawa, Munemasa. 1981. Kotoba: nishi to higashi. Tokyo: Chûô Kôronsha. Tokugawa, Munemasa. 1993. Hôgen-chirigaku no tenkai. Tokyo: Hituzi Shobô. Tokugawa, Munemasa. 1996. Go no chiriteki denpa sokudo. Gengogakurin 19951996. Tokyo: Sanseidô. Tranter, Nicolas. 2001. “The Asukaike word list slat and pre-Sino-Japanese phonology”. Language change in East Asia ed. by T.E. McAuley, 143-160. Richmond, Surrey: Curzon. Uchima, Chokuji. 2004. “Kodai Nihongo no wa-gyô shiin no [b] onka ni tsuite: Miyako, Yaeyama hôgen o chûshin ni”. Kokugogaku 55:2.32-44. Uemura, Yukio. 1977. “Ryûkyû hôgen kenkyû no gendai no kadai: Toku ni sono hikaku rekishi hôgengakuteki kenkyû ni tsuite”. Shin-Okinawa Bungaku 35.226-237. Unger, J. Marshall. 1993 [1977]. Studies in early Japanese morphophonemics. Bloomington: IULC. Unger, J. Marshall. 2000a. “Rendaku and Proto-Japanese Accent Classes”. Japanese/Korean Linguistics 9.17-30. Unger, J. Marshall. 2000b. “Reconciling comparative and internal reconstruction: The case of Old Japanese /ti ri ni/”. Language 76:3.655-681.

REFERENCES

225

Unger, J. Marshall. 2004. “Alternations of m and b in Early Middle Japanese: The deeper significance of the sound-symbolic stratum”. Japanese Language and Literature 38:2.323-337. Ushiyama, Hatsuo. 1953. “Gohôjô yori mitaru tôzai hôgen no kyôkaisen ni tsuite”. Kokugogaku 12.59-63. Uwano, Zendô. 1977. “Nihongo no akusento”. Iwanami-kôza Nihongo 5: On’in, 281-321. Tokyo: Iwanami Shoten. Uwano, Zendô. 1985. “Kagawa-ken Ibuki-jima hôgen no akusento”. Nihon Gakushi’in Kiyô 40:2.75-179. Uwano, Zendô. 1999. “Classification of Japanese accent systems”. Proceedings of the symposium “Cross-linguistic studies of tonal phenomena: tonogenesis, typology, and related topics” ed. by Shigeki Kaji, 151-178. Tokyo: ILCAA, Tokyo University of Foreign Studies. Vääsänen, Veikko. 1963. Introduction au Latin vulgaire. Paris: Librairie C. Klincksieck. Vovin, Alexander. 1993a. “Long vowels in proto-Japanese”. Journal of East Asian Linguistics 2:2.125-134. Vovin, Alexander. 1993b. A reconstruction of proto-Ainu. Leiden, New York, Köln: E.J. Brill. Vovin, Alexander. 1994. “Genetic affiliation of Japanese and methodology of linguistic comparison”. Journal de la Société Finno-Ougrienne 85.241256. Vovin, Alexander. 1995. “Origin of register in Japanese and the Altaic theory”. Japanese/Korean Linguistics 6.113-133. Vovin, Alexander & Toshiki Osada (eds.). 2003. Nihongo keitôron no genzai; Perspectives on the origins of the Japanese language. (= Nichibunken Japanese Studies Series 31). Kyoto: International Research Center for Japanese Studies. Vovin, Alexander. Forthcoming. Koreo-Japonica: A critical study in the language relationship. Honolulu: University of Hawaii Press. Wenck, Günther. 1959. Japanische Phonetik, Band IV: Erscheinungen und Probleme des japanischen Lautwandels. Wiesbaden: Otto Harrassowitz. Whitman, John. 1985. The phonological basis for the comparison of Japanese and Korean. Harvard University dissertation. Whitman, John. 1999. Primary root shape in Japanese and Korean. Paper presented at the Workshop on Korean-Japanese comparative-historical linguistics, XIVth International Conference on Historical Linguistics, 11 12 August 1999, University of British Columbia, Vancouver. Whitman, John. 2004. The form and function of the rentaikei and izenkei suffixes in proto-Japanese. Paper presented at the Second Oxford-Kobe Linguistics Seminar, on the History and Structure of Japanese, 27 September 2004. Wrona, Janick. Forthcoming. A study of Old Japanese syntax: synchronic and diachronic aspects of the complemental system. Folkestone: Global Oriental.

226

REFERENCES

Wrona, Janick. 2005. “Specificational pseudo-clefts in Old Japanese”. Folia Linguistica Historica XXVI(1/2). 139-157. Yamada, Yoshio. 1990 [1913]. Narachô bunpôshi. Tokyo: Hôbunkan. Yamaguchi, Yoshinori. 1971. “Kodai Nihongo ni okeru boin datsuraku no genshô ni tsuite”. Kokugogaku 85.1-17. Yamaguchi, Yoshinori. 1985. Kodai Nihongo bunpô no seiritsu no kenkyû. Tokyo: Yûseidô. Yoshida, Kanehiko. 1973. Jôdai jodôshi no shiteki kenkyû. Tokyo: Meiji Shoin. Yoshida, Kenji. 1992. “Saitama-ken tôbu akusento no chiriteki bunpu”. Nihon hôgen kenkyû-kai dai-55-kai kenkyû happyô-kai happyô genkô-shû, 9-18. Yoshitake, Saburô. 1930. “The history of the Japanese particle ‘i’”. Bulletin of the School of Oriental Studies 5:4.889-895. Zdb = Omodaka et al. 1967

INDEX A. *-aku, see Nominal form (a)n/*an Negative 161-162, 184-190 accent class 2.5, of nouns 109, 111-113, 116, 120, 122, 126-132, 134, 138, 139, 141, 142-150, 151-155 A/B distinctions, see kô-otsu distinctions Adnominal verb form (rentaikei), uses in OJ 7, 9, 193-215 Ainu, pitch accent in 115-118, 135-136, 150 Amami 79, 82, 86, 88, 92-93, 127-128, 144 accent system 82, 127-128, 144 apophony, nominal 19-21, 30-33, 37 in pre-OJ 21-41, 89, 91-92, 95, 163, 166, 168-169, 179-180, 184-185 verbal 21-24, 33-34, 89, 91-92, 95, 162164,166ff, 180 ar- ‘exist’, in OJ 166, 177-178, 183, 189, 191 Arisaka, Hideyo 161 Arisaka’s Law 40-41, 97 B. bigrade (nidan) conjugation, in OJ 159-173, 187-192 bigrade (nidan) verbs, lexical distribution of 17-18, 182-183 morphophonology of 182-183 reconstruction of 21, 33-36, 39, 160-172, 175-192 with auxiliaries 25, 33, 162, 166, 168, 171, 178, 183-184, 187 C. *-Ci (pJ transitivity flipper) 147, 179, 180181 comparative method 1, 8 consonant stem (godan) verbs, in J 8, 147 consonants, initial voiced, in pJ 8, 44, 48-51, 150, 153-154 onset, in OJ 2-3, 44, 153-154 prenasalized, in OJ 3, 43, 47-49

D. demonstratives, in Middle Korean 27-29 in pre-OJ 23, 27-29 dialect chain 57-60, 63-64, 80-81 E. e- ‘get’, in OJ 84, 166-167, 169-171, 173, 179-181 Eastern OJ 23-25, 27, 89, 92, 144, 190 F. four vowel system, for PJ 16-17, 82, 92 Frellesvig, Bjarke 1-9, 15-41, 90-94, 165, 175-192 G. go’on 44, 47-53 Gothic 170 H. Hamano, Shoko 43, 52 Hanihara, Kazuro 98 Hashimoto, Shinkichi 161 Hashimoto, Shirô 28 Hattori, Shirô 1, 5, 7-8, 22, 24, 26, 88, 90-94, 125-132, 134-139, 150-152 Hayata, Teruhiro 26 hifukukei (embedded root forms) 19-21, 36, 161-162, 168, 172, 178-181 Hino, Sukenari 190 I. Ibuki-jima accent system 109-114 Ikema (Miyako) accent system 128-130 inchoative verbs, in OJ 165, 169-173, 179181 infinitive (ren’yôkei), in OJ 161, 187 internal reconstruction 1-2, 51, 89, 153, 194 irrealis (mizenkei), in OJ 161-163, 166-169, 184 J. Jacobsen, Wesley 172

228 K. kakari-musubi 96-98, 199, 205, 210-211, 213 Kametsu (Amami) accent system, see Amami accent system kan’on 49-50, 51, 53 Keihan accent systems 106-107 Kindaichi, Haruhiko 104, 108, 119-120, 125-128, 131, 133, 135, 137, 142 Kindaichi goi 104, 107-112, 119-121, 123124 kô-otsu distinctions (= A/B distinctions) 3-4, 53 Korean, causative/passive in 263-165 loans into Japanese 23, 27, 34, 36-38 Kubozono, Haruo 117, 220 Kuki city accent system 116 L. Lange, Roland 15 lexical contraction of vowels, in pre-OJ 1819, 21, 30, 34, 161-166, 168, 175-178, 180, 189 lower bigrade (shimo nidan) conjugation 21, 89, 94 M. *-m-, pJ nominalizing suffix 147-148 Man’yôshû 17, 60, 63, 193, 222, 224 Martin, Samuel E. 1, 33, 49, 141-142, 147148, 150-152, 166-167 Mase, Yoshio 59-60, 63, 221 Matsumori, Akiko 5, 7, 103-124, 131-132, 134, 136-137 Matsumoto, Katsumi 7, 15, 23 me-, OJ denominal verbal suffix 147-148 suffix in birds’ names 148 mid vowel raising, in pre-OJ 37-38, 40, 169 in Ryukyuan 22-27, 81-87, 91-94 middle bigrade conjugation 88-89 Middle Korean, correspondences with class 2.5 nouns 142-149 Middle Korean, vowel correspondences with OJ 27-29, 34-38 Miller, Roy Andrew 4, 192 Miyake, Marc 18, 22-26, 37, 43, 46-51, 53, 88 Miyako 49, 80, 82, 88, 128, 145 mokkan 51, 53 monophthongization, in Ryukyuan 24, 37, 39, 82, 91, 93-94 in pre-OJ 17, 18, 22, 26, 29-30, 90-94, 169, 175-177, 189 Murayama, Shichirô 20, 36

INDEX N. Nakijin (Okinawa) accent system 128-129, 133-136, 138-139 negative paradigm, in OJ 186-188 negative suffix, in J dialects 58, 60, 62, 68, 70-71 Nominal form of verbs, uses in OJ 9, 193215 O. obligational expressions, in J dialects 57, 6567 Ohno, Susumu 4, 8, 9, 94, 147-148, 161-162, 168, 178-179, 181, 184, 192, 194 Okinawa 79-81, 87-88, 92, 95-98, 127-128, 145, 151-152 Old Norse 170 Omoro sôshi 92, 95, 98 Ongana 46-48, 52 Omi accent system 111 P. /p/, in OJ 2-3, 43 palatal assimilation, progressive, in R 83-84 Polivanov, Evgenii D. 1, 141, 143 post sibilant fronting, in R 86 progressive aspect, in J dialects 60-64, 74, 77, 95 Q. quadrigrade (yodan) conjugation, in OJ 9, 92, 160-163, 165, 172, 189-190, 194 quadrigrade (yodan) verbs, reanalysis of in pre-OJ 9, 184-187 R. Ramsey, S. Robert 48-49, 141-142, 146 register, word-initial, in pJ 8, 138-140, 150, 153-154 roshutsukei 19-21, 36 Ruiju myôgishô 104 accent correspondences 107, 126, 141142 Ryukyuan, accent systems 7, 118-121, 123129, 133-135 correspondences with OJ 7, 82, 88-89, 97-99 S. s- transitive stems, in OJ 171-172 Sandness, Karen 183 Serafim, Leon 5, 7, 22, 25-26, 37, 39, 87, 89, 92, 94-98

INDEX seven vowel system, for pJ 6, 15-16, 29, 34, 38, 41 Shinzato, Rumiko 95-97 Shishi-jima accent system 112-113 Shodon 82, 88, 92-93 Shuri 7, 79, 81, 84, 89, 91-92, 94, 99, 127, 134-137, 139-140, 144, 149, 151-152, 160 six vowel system, for pJ 92 Sonai (Yaeyama) accent system 128-129, 131, 145 T. Takeda, Yukichi 166, 168, 172 Thorpe, Maner 23-25, 94, 97 Tohoku dialect 64 Tokugawa, Munemasa 7, 61-62, 119, 122, 141-142, 146 Tokyo (Edo) dialect 44, 64, 115-118, 122, 126, 141-142, 160 tonal pattern merger 106-108, 119-122 tone insertion, in phrase-initial position 105106 tone shift, rightward 105-106, 112-113 Tranter, Nicholas 51, 53-54 U. umlaut, in pre-OJ 18, 25

229 Unger, J. Marshall 3, 5, 6, 8-9, 21, 48-50, 52, 89, 142, 149, 162-164, 168-170, 172, 178-179, 181, 189, 190, 192 upper bigrade (kami nidan) conjugation, in OJ 8, 18, 21, 33, 36, 39, 88-89, 94-95 V. verb conjugations, in OJ 8, 9 voiced obstruents, in PJ 3, 43-44, 49, 51-53 vowel contraction, in pre-OJ 5, 17, 19, 20-22, 26, 30, 31, 33, 39-40, 163, 166-167, 169 vowel deletion, in OJ and pre-OJ 166-167, 169 vowel length, in Ryukyuan 7-8 vowel stem, in pJ 8, 172-173, 178-180 W. Whitman, John 4-5, 8-9, 15-16, 20, 37, 50, 52, 90, 143-145, 163, 167, 169, 178-181, 184, 192 Y. Yaeyama 49, 80, 128, 145 Yamaguchi, Yoshinori 163, 167 -yer- Stative, in OJ 18-19 Yonaguni 80, 82, 128-130, 132, 139, 145, 154 Yoshida, Kanehiko 162-163, 166, 168-169

CURRENT ISSUES IN LINGUISTIC THEORY

E. F. K. Koerner, Editor

Zentrum für Allgemeine Sprachwissenschaft, Typologie und Universalienforschung, Berlin [email protected] Current Issues in Linguistic Theory (CILT) is a theory-oriented series which welcomes contributions from scholars who have significant proposals to make towards the advancement of our understanding of language, its structure, functioning and development. CILT has been established in order to provide a forum for the presentation and discussion of linguistic opinions of scholars who do not necessarily accept the prevailing mode of thought in linguistic science. It offers an outlet for meaningful contributions to the current linguistic debate, and furnishes the diversity of opinion which a healthy discipline must have. A complete list of titles in this series can be found on the publishers’ website, www.benjamins.com 297 Dossena, Marina, Richard Dury and Maurizio Gotti (eds.): English Historical Linguistics 2006. Volume III: Geo-Historical Variation in English. Selected papers from the fourteenth International Conference on English Historical Linguistics (ICEHL 14), Bergamo, 21-25 August 2006. Expected Forthcoming 296 Dury, Richard, Maurizio Gotti and Marina Dossena (eds.): English Historical Linguistics 2006. Volume II: Lexical and Semantic Change. Selected papers from the fourteenth International Conference on English Historical Linguistics (ICEHL 14), Bergamo, 21-25 August 2006. 2008. 295 Gotti, Maurizio, Marina Dossena and Richard Dury (eds.): English Historical Linguistics 2006. Volume I: Syntax and Morphology. Selected papers from the fourteenth International Conference on English Historical Linguistics (ICEHL 14), Bergamo, 21-25 August 2006. xiv, 256 pp. + index. Expected July 2008 294 Frellesvig, Bjarke and John Whitman (eds.): Proto-Japanese. Issues and Prospects. 2008. vii, 229 pp. 293 Detges, Ulrich and Richard Waltereit (eds.): The Paradox of Grammatical Change. Perspectives from Romance. 2008. vi, 252 pp. 292 Nicolov, Nicolas, Kalina Bontcheva, Galia Angelova and Ruslan Mitkov (eds.): Recent Advances in Natural Language Processing IV. Selected papers from RANLP 2005. 2007. xii, 307 pp. 291 Baauw, Sergio, Frank Drijkoningen and Manuela Pinto (eds.): Romance Languages and Linguistic Theory 2005. Selected papers from ‘Going Romance’, Utrecht, 8–10 December 2005. 2007. viii, 338 pp. 290 Mughazy, Mustafa A. (ed.): Perspectives on Arabic Linguistics XX. Papers from the twentieth annual symposium on Arabic linguistics, Kalamazoo, Michigan, March 2006. 2007. xii, 247 pp. 289 Benmamoun, Elabbas (ed.): Perspectives on Arabic Linguistics XIX. Papers from the nineteenth annual symposium on Arabic Linguistics, Urbana, Illinois, April 2005. 2007. xiv, 304 pp. 288 Toivonen, Ida and Diane Nelson (eds.): Saami Linguistics. 2007. viii, 321 pp. 287 Camacho, José, Nydia Flores-Ferrán, Liliana Sánchez, Viviane Déprez and María José Cabrera (eds.): Romance Linguistics 2006. Selected papers from the 36th Linguistic Symposium on Romance Languages (LSRL), New Brunswick, March-April 2006. 2007. viii, 340 pp. 286 Weijer, Jeroen van de and Erik Jan van der Torre (eds.): Voicing in Dutch. (De)voicing – phonology, phonetics, and psycholinguistics. 2007. x, 186 pp. 285 Sackmann, Robin (ed.): Explorations in Integrational Linguistics. Four essays on German, French, and Guaraní. ix, 217 pp. Expected April 2008 284 Salmons, Joseph C. and Shannon Dubenion-Smith (eds.): Historical Linguistics 2005. Selected papers from the 17th International Conference on Historical Linguistics, Madison, Wisconsin, 31 July - 5 August 2005. 2007. viii, 413 pp. 283 Lenker, Ursula and Anneli Meurman-Solin (eds.): Connectives in the History of English. 2007. viii, 318 pp. 282 Prieto, Pilar, Joan Mascaró and Maria-Josep Solé (eds.): Segmental and prosodic issues in Romance phonology. 2007. xvi, 262 pp. 281 Vermeerbergen, Myriam, Lorraine Leeson and Onno Crasborn (eds.): Simultaneity in Signed Languages. Form and function. 2007. viii, 360 pp. (incl. CD-Rom). 280 Hewson, John and Vit Bubenik: From Case to Adposition. The development of configurational syntax in Indo-European languages. 2006. xxx, 420 pp. 279 Nedergaard Thomsen, Ole (ed.): Competing Models of Linguistic Change. Evolution and beyond. 2006. vi, 344 pp.

278 Doetjes, Jenny and Paz González (eds.): Romance Languages and Linguistic Theory 2004. Selected papers from ‘Going Romance’, Leiden, 9–11 December 2004. 2006. viii, 320 pp. 277 Helasvuo, Marja-Liisa and Lyle Campbell (eds.): Grammar from the Human Perspective. Case, space and person in Finnish. 2006. x, 280 pp. 276 Montreuil, Jean-Pierre Y. (ed.): New Perspectives on Romance Linguistics. Vol. II: Phonetics, Phonology and Dialectology. Selected papers from the 35th Linguistic Symposium on Romance Languages (LSRL), Austin, Texas, February 2005. 2006. x, 213 pp. 275 Nishida, Chiyo and Jean-Pierre Y. Montreuil (eds.): New Perspectives on Romance Linguistics. Vol. I: Morphology, Syntax, Semantics, and Pragmatics. Selected papers from the 35th Linguistic Symposium on Romance Languages (LSRL), Austin, Texas, February 2005. 2006. xiv, 288 pp. 274 Gess, Randall S. and Deborah Arteaga (eds.): Historical Romance Linguistics. Retrospective and perspectives. 2006. viii, 393 pp. 273 Filppula, Markku, Juhani Klemola, Marjatta Palander and Esa Penttilä (eds.): Dialects Across Borders. Selected papers from the 11th International Conference on Methods in Dialectology (Methods XI), Joensuu, August 2002. 2005. xii, 291 pp. 272 Gess, Randall S. and Edward J. Rubin (eds.): Theoretical and Experimental Approaches to Romance Linguistics. Selected papers from the 34th Linguistic Symposium on Romance Languages (LSRL), Salt Lake City, March 2004. 2005. viii, 367 pp. 271 Branner, David Prager (ed.): The Chinese Rime Tables. Linguistic philosophy and historicalcomparative phonology. 2006. viii, 358 pp. 270 Geerts, Twan, Ivo van Ginneken and Haike Jacobs (eds.): Romance Languages and Linguistic Theory 2003. Selected papers from ‘Going Romance’ 2003, Nijmegen, 20–22 November. 2005. viii, 369 pp. 269 Hargus, Sharon and Keren Rice (eds.): Athabaskan Prosody. 2005. xii, 432 pp. 268 Cravens, Thomas D. (ed.): Variation and Reconstruction. 2006. viii, 223 pp. 267 Alhawary, Mohammad T. and Elabbas Benmamoun (eds.): Perspectives on Arabic Linguistics XVII–XVIII. Papers from the seventeenth and eighteenth annual symposia on Arabic linguistics. Volume XVII–XVIII: Alexandria, 2003 and Norman, Oklahoma 2004. 2005. xvi, 315 pp. 266 Boudelaa, Sami (ed.): Perspectives on Arabic Linguistics XVI. Papers from the sixteenth annual symposium on Arabic linguistics, Cambridge, March 2002. 2006. xii, 181 pp. 265 Cornips, Leonie and Karen P. Corrigan (eds.): Syntax and Variation. Reconciling the Biological and the Social. 2005. vi, 312 pp. 264 Dressler, Wolfgang U., Dieter Kastovsky, Oskar E. Pfeiffer and Franz Rainer (eds.): Morphology and its demarcations. Selected papers from the 11th Morphology meeting, Vienna, February 2004. With the assistance of Francesco Gardani and Markus A. Pöchtrager. 2005. xiv, 320 pp. 263 Branco, António, Tony McEnery and Ruslan Mitkov (eds.): Anaphora Processing. Linguistic, cognitive and computational modelling. 2005. x, 449 pp. 262 Vajda, Edward J. (ed.): Languages and Prehistory of Central Siberia. 2004. x, 275 pp. 261 Kay, Christian J. and Jeremy J. Smith (eds.): Categorization in the History of English. 2004. viii, 268 pp. 260 Nicolov, Nicolas, Kalina Bontcheva, Galia Angelova and Ruslan Mitkov (eds.): Recent Advances in Natural Language Processing III. Selected papers from RANLP 2003. 2004. xii, 402 pp. 259 Carr, Philip, Jacques Durand and Colin J. Ewen (eds.): Headhood, Elements, Specification and Contrastivity. Phonological papers in honour of John Anderson. 2005. xxviii, 405 pp. 258 Auger, Julie, J. Clancy Clements and Barbara Vance (eds.): Contemporary Approaches to Romance Linguistics. Selected Papers from the 33rd Linguistic Symposium on Romance Languages (LSRL), Bloomington, Indiana, April 2003. With the assistance of Rachel T. Anderson. 2004. viii, 404 pp. 257 Fortescue, Michael, Eva Skafte Jensen, Jens Erik Mogensen and Lene Schøsler (eds.): Historical Linguistics 2003. Selected papers from the 16th International Conference on Historical Linguistics, Copenhagen, 11–15 August 2003. 2005. x, 312 pp. 256 Bok-Bennema, Reineke, Bart Hollebrandse, Brigitte Kampers-Manhe and Petra Sleeman (eds.): Romance Languages and Linguistic Theory 2002. Selected papers from ‘Going Romance’, Groningen, 28–30 November 2002. 2004. viii, 273 pp. 255 Meulen, Alice ter and Werner Abraham (eds.): The Composition of Meaning. From lexeme to discourse. 2004. vi, 232 pp. 254 Baldi, Philip and Pietro U. Dini (eds.): Studies in Baltic and Indo-European Linguistics. In honor of William R. Schmalstieg. 2004. xlvi, 302 pp. 253 Caffarel, Alice, J.R. Martin and Christian M.I.M. Matthiessen (eds.): Language Typology. A functional perspective. 2004. xiv, 702 pp.

252 Kay, Christian J., Carole Hough and Irené Wotherspoon (eds.): New Perspectives on English Historical Linguistics. Selected papers from 12 ICEHL, Glasgow, 21–26 August 2002. Volume II: Lexis and Transmission. 2004. xii, 273 pp. 251 Kay, Christian J., Simon Horobin and Jeremy J. Smith (eds.): New Perspectives on English Historical Linguistics. Selected papers from 12 ICEHL, Glasgow, 21–26 August 2002. Volume I: Syntax and Morphology. 2004. x, 264 pp. 250 Jensen, John T.: Principles of Generative Phonology. An introduction. 2004. xii, 324 pp. 249 Bowern, Claire and Harold Koch (eds.): Australian Languages. Classification and the comparative method. 2004. xii, 377 pp. (incl. CD-Rom). 248 Weigand, Edda (ed.): Emotion in Dialogic Interaction. Advances in the complex. 2004. xii, 284 pp. 247 Parkinson, Dilworth B. and Samira Farwaneh (eds.): Perspectives on Arabic Linguistics XV. Papers from the Fifteenth Annual Symposium on Arabic Linguistics, Salt Lake City 2001. 2003. x, 214 pp. 246 Holisky, Dee Ann and Kevin Tuite (eds.): Current Trends in Caucasian, East European and Inner Asian Linguistics. Papers in honor of Howard I. Aronson. 2003. xxviii, 426 pp. 245 Quer, Josep, Jan Schroten, Mauro Scorretti, Petra Sleeman and Els Verheugd (eds.): Romance Languages and Linguistic Theory 2001. Selected papers from 'Going Romance', Amsterdam, 6–8 December 2001. 2003. viii, 355 pp. 244 Pérez-Leroux, Ana Teresa and Yves Roberge (eds.): Romance Linguistics. Theory and Acquisition. Selected papers from the 32nd Linguistic Symposium on Romance Languages (LSRL), Toronto, April 2002. 2003. viii, 388 pp. 243 Cuyckens, Hubert, Thomas Berg, René Dirven and Klaus-Uwe Panther (eds.): Motivation in Language. Studies in honor of Günter Radden. 2003. xxvi, 403 pp. 242 Seuren, Pieter A.M. and Gerard Kempen (eds.): Verb Constructions in German and Dutch. 2003. vi, 316 pp. 241 Lecarme, Jacqueline (ed.): Research in Afroasiatic Grammar II. Selected papers from the Fifth Conference on Afroasiatic Languages, Paris, 2000. 2003. viii, 550 pp. 240 Janse, Mark and Sijmen Tol (eds.): Language Death and Language Maintenance. Theoretical, practical and descriptive approaches. With the assistance of Vincent Hendriks. 2003. xviii, 244 pp. 239 Andersen, Henning (ed.): Language Contacts in Prehistory. Studies in Stratigraphy. Papers from the Workshop on Linguistic Stratigraphy and Prehistory at the Fifteenth International Conference on Historical Linguistics, Melbourne, 17 August 2001. 2003. viii, 292 pp. 238 Núñez-Cedeño, Rafael, Luis López and Richard Cameron (eds.): A Romance Perspective on Language Knowledge and Use. Selected papers from the 31st Linguistic Symposium on Romance Languages (LSRL), Chicago, 19–22 April 2001. 2003. xvi, 386 pp. 237 Blake, Barry J. and Kate Burridge (eds.): Historical Linguistics 2001. Selected papers from the 15th International Conference on Historical Linguistics, Melbourne, 13–17 August 2001. Editorial assistance Jo Taylor. 2003. x, 444 pp. 236 Simon-Vandenbergen, Anne-Marie, Miriam Taverniers and Louise J. Ravelli (eds.): Grammatical Metaphor. Views from systemic functional linguistics. 2003. vi, 453 pp. 235 Linn, Andrew R. and Nicola McLelland (eds.): Standardization. Studies from the Germanic languages. 2002. xii, 258 pp. 234 Weijer, Jeroen van de, Vincent J. van Heuven and Harry van der Hulst (eds.): The Phonological Spectrum. Volume II: Suprasegmental structure. 2003. x, 264 pp. 233 Weijer, Jeroen van de, Vincent J. van Heuven and Harry van der Hulst (eds.): The Phonological Spectrum. Volume I: Segmental structure. 2003. x, 308 pp. 232 Beyssade, Claire, Reineke Bok-Bennema, Frank Drijkoningen and Paola Monachesi (eds.): Romance Languages and Linguistic Theory 2000. Selected papers from ‘Going Romance’ 2000, Utrecht, 30 November–2 December. 2002. viii, 354 pp. 231 Cravens, Thomas D.: Comparative Historical Dialectology. Italo-Romance clues to Ibero-Romance sound change. 2002. xii, 163 pp. 230 Parkinson, Dilworth B. and Elabbas Benmamoun (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume XIII-XIV: Stanford, 1999 and Berkeley, California 2000. 2002. xiv, 250 pp. 229 Nevin, Bruce E. and Stephen B. Johnson (eds.): The Legacy of Zellig Harris. Language and information into the 21st century. Volume 2: Mathematics and computability of language. 2002. xx, 312 pp. 228 Nevin, Bruce E. (ed.): The Legacy of Zellig Harris. Language and information into the 21st century. Volume 1: Philosophy of science, syntax and semantics. 2002. xxxvi, 323 pp. 227 Fava, Elisabetta (ed.): Clinical Linguistics. Theory and applications in speech pathology and therapy. 2002. xxiv, 353 pp.

226 Levin, Saul: Semitic and Indo-European. Volume II: Comparative morphology, syntax and phonetics. 2002. xviii, 592 pp. 225 Shahin, Kimary N.: Postvelar Harmony. 2003. viii, 344 pp. 224 Fanego, Teresa, Belén Méndez-Naya and Elena Seoane (eds.): Sounds, Words, Texts and Change. Selected papers from 11 ICEHL, Santiago de Compostela, 7–11 September 2000. Volume 2. 2002. x, 310 pp. 223 Fanego, Teresa, Javier Pérez-Guerra and María José López-Couso (eds.): English Historical Syntax and Morphology. Selected papers from 11 ICEHL, Santiago de Compostela, 7–11 September 2000. Volume 1. 2002. x, 306 pp. 222 Herschensohn, Julia, Enrique Mallén and Karen Zagona (eds.): Features and Interfaces in Romance. Essays in honor of Heles Contreras. 2001. xiv, 302 pp. 221 D’hulst, Yves, Johan Rooryck and Jan Schroten (eds.): Romance Languages and Linguistic Theory 1999. Selected papers from ‘Going Romance’ 1999, Leiden, 9–11 December 1999. 2001. viii, 406 pp. 220 Satterfield, Teresa, Christina M. Tortora and Diana Cresti (eds.): Current Issues in Romance Languages. Selected papers from the 29th Linguistic Symposium on Romance Languages (LSRL), Ann Arbor, 8–11 April 1999. 2002. viii, 412 pp. 219 Andersen, Henning (ed.): Actualization. Linguistic Change in Progress. Papers from a workshop held at the 14th International Conference on Historical Linguistics, Vancouver, B.C., 14 August 1999. 2001. vii, 250 pp. 218 Bendjaballah, Sabrina, Wolfgang U. Dressler, Oskar E. Pfeiffer and Maria D. Voeikova (eds.): Morphology 2000. Selected papers from the 9th Morphology Meeting, Vienna, 24–28 February 2000. 2002. viii, 317 pp. 217 Wiltshire, Caroline R. and Joaquim Camps (eds.): Romance Phonology and Variation. Selected papers from the 30th Linguistic Symposium on Romance Languages, Gainesville, Florida, February 2000. 2002. xii, 238 pp. 216 Camps, Joaquim and Caroline R. Wiltshire (eds.): Romance Syntax, Semantics and L2 Acquisition. Selected papers from the 30th Linguistic Symposium on Romance Languages, Gainesville, Florida, February 2000. 2001. xii, 246 pp. 215 Brinton, Laurel J. (ed.): Historical Linguistics 1999. Selected papers from the 14th International Conference on Historical Linguistics, Vancouver, 9–13 August 1999. 2001. xii, 398 pp. 214 Weigand, Edda and Marcelo Dascal (eds.): Negotiation and Power in Dialogic Interaction. 2001. viii, 303 pp. 213 Sornicola, Rosanna, Erich Poppe and Ariel Shisha-Halevy (eds.): Stability, Variation and Change of Word-Order Patterns over Time. With the assistance of Paola Como. 2000. xxxii, 323 pp. 212 Repetti, Lori (ed.): Phonological Theory and the Dialects of Italy. 2000. x, 301 pp. 211 Elšík, Viktor and Yaron Matras (eds.): Grammatical Relations in Romani. The Noun Phrase. with a Foreword by Frans Plank (Universität Konstanz). 2000. x, 244 pp. 210 Dworkin, Steven N. and Dieter Wanner (eds.): New Approaches to Old Problems. Issues in Romance historical linguistics. 2000. xiv, 235 pp. 209 King, Ruth: The Lexical Basis of Grammatical Borrowing. A Prince Edward Island French case study. 2000. xvi, 241 pp. 208 Robinson, Orrin W.: Whose German? The ach/ich alternation and related phenomena in ‘standard’ and ‘colloquial’. 2001. xii, 178 pp. 207 Sanz, Montserrat: Events and Predication. A new approach to syntactic processing in English and Spanish. 2000. xiv, 219 pp. 206 Fawcett, Robin P.: A Theory of Syntax for Systemic Functional Linguistics. 2000. xxiv, 360 pp. 205 Dirven, René, Roslyn M. Frank and Cornelia Ilie (eds.): Language and Ideology. Volume 2: descriptive cognitive approaches. 2001. vi, 264 pp. 204 Dirven, René, Bruce Hawkins and Esra Sandikcioglu (eds.): Language and Ideology. Volume 1: theoretical cognitive approaches. 2001. vi, 301 pp. 203 Norrick, Neal R.: Conversational Narrative. Storytelling in everyday talk. 2000. xiv, 233 pp. 202 Lecarme, Jacqueline, Jean Lowenstamm and Ur Shlonsky (eds.): Research in Afroasiatic Grammar. Papers from the Third conference on Afroasiatic Languages, Sophia Antipolis, 1996. 2000. vi, 386 pp. 201 Dressler, Wolfgang U., Oskar E. Pfeiffer, Markus A. Pöchtrager and John R. Rennison (eds.): Morphological Analysis in Comparison. 2000. x, 261 pp. 200 Anttila, Raimo: Greek and Indo-European Etymology in Action. Proto-Indo-European *aǵ-. 2000. xii, 314 pp. 199 Pütz, Martin and Marjolijn H. Verspoor (eds.): Explorations in Linguistic Relativity. 2000. xvi, 369 pp.

198 Niemeier, Susanne and René Dirven (eds.): Evidence for Linguistic Relativity. 2000. xxii, 240 pp. 197 Coopmans, Peter, Martin Everaert and Jane Grimshaw (eds.): Lexical Specification and Insertion. 2000. xviii, 476 pp. 196 Hannahs, S.J. and Mike Davenport (eds.): Issues in Phonological Structure. Papers from an International Workshop. 1999. xii, 268 pp. 195 Herring, Susan C., Pieter van Reenen and Lene Schøsler (eds.): Textual Parameters in Older Languages. 2001. x, 448 pp. 194 Coleman, Julie and Christian J. Kay (eds.): Lexicology, Semantics and Lexicography. Selected papers from the Fourth G. L. Brook Symposium, Manchester, August 1998. 2000. xiv, 257 pp. 193 Klausenburger, Jurgen: Grammaticalization. Studies in Latin and Romance morphosyntax. 2000. xiv, 184 pp. 192 Alexandrova, Galina M. and Olga Arnaudova (eds.): The Minimalist Parameter. Selected papers from the Open Linguistics Forum, Ottawa, 21–23 March 1997. 2001. x, 360 pp. 191 Sihler, Andrew L.: Language History. An introduction. 2000. xvi, 298 pp. 190 Benmamoun, Elabbas (ed.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume XII: Urbana-Champaign, Illinois, 1998. 1999. viii, 204 pp. 189 Nicolov, Nicolas and Ruslan Mitkov (eds.): Recent Advances in Natural Language Processing II. Selected papers from RANLP ’97. 2000. xi, 422 pp. 188 Simmons, Richard VanNess: Chinese Dialect Classification. A comparative approach to Harngjou, Old Jintarn, and Common Northern Wu. 1999. xviii, 317 pp. 187 Franco, Jon A., Alazne Landa and Juan Martín (eds.): Grammatical Analyses in Basque and Romance Linguistics. Papers in honor of Mario Saltarelli. 1999. viii, 306 pp. 186 Mišeska Tomić, Olga and Milorad Radovanović (eds.): History and Perspectives of Language Study. Papers in honor of Ranko Bugarski. 2000. xxii, 314 pp. 185 Authier, Jean-Marc, Barbara E. Bullock and Lisa A. Reed (eds.): Formal Perspectives on Romance Linguistics. Selected papers from the 28th Linguistic Symposium on Romance Languages (LSRL XXVIII), University Park, 16–19 April 1998. 1999. xii, 334 pp. 184 Sagart, Laurent: The Roots of Old Chinese. 1999. xii, 272 pp. 183 Contini-Morava, Ellen and Yishai Tobin (eds.): Between Grammar and Lexicon. 2000. xxxii, 365 pp. 182 Kenesei, István (ed.): Crossing Boundaries. Advances in the theory of Central and Eastern European languages. 1999. viii, 302 pp. 181 Mohammad, Mohammad A.: Word Order, Agreement and Pronominalization in Standard and Palestinian Arabic. 2000. xvi, 197 pp. 180 Mereu, Lunella (ed.): Boundaries of Morphology and Syntax. 1999. viii, 314 pp. 179 Rini, Joel: Exploring the Role of Morphology in the Evolution of Spanish. 1999. xvi, 187 pp. 178 Foolen, Ad and Frederike van der Leek (eds.): Constructions in Cognitive Linguistics. Selected papers from the Fifth International Cognitive Linguistics Conference, Amsterdam, 1997. 2000. xvi, 338 pp. 177 Cuyckens, Hubert and Britta E. Zawada (eds.): Polysemy in Cognitive Linguistics. Selected papers from the International Cognitive Linguistics Conference, Amsterdam, 1997. 2001. xxviii, 296 pp. 176 Van Hoek, Karen, Andrej A. Kibrik and Leo Noordman (eds.): Discourse Studies in Cognitive Linguistics. Selected papers from the 5th International Cognitive Linguistics Conference, Amsterdam, July 1997. 1999. vi, 187 pp. 175 Gibbs, Jr., Raymond W. and Gerard J. Steen (eds.): Metaphor in Cognitive Linguistics. Selected papers from the 5th International Cognitive Linguistics Conference, Amsterdam, 1997. 1999. viii, 226 pp. 174 Hall, T. Alan and Ursula Kleinhenz (eds.): Studies on the Phonological Word. 1999. viii, 298 pp. 173 Treviño, Esthela and José Lema (eds.): Semantic Issues in Romance Syntax. 1999. viii, 309 pp. 172 Dimitrova-Vulchanova, Mila and Lars Hellan (eds.): Topics in South Slavic Syntax and Semantics. 1999. xxviii, 263 pp. 171 Weigand, Edda (ed.): Contrastive Lexical Semantics. 1998. x, 270 pp. 170 Lamb, Sydney M.: Pathways of the Brain. The neurocognitive basis of language. 1999. xii, 418 pp. 169 Ghadessy, Mohsen (ed.): Text and Context in Functional Linguistics. 1999. xviii, 340 pp. 168 Ratcliffe, Robert R.: The “Broken” Plural Problem in Arabic and Comparative Semitic. Allomorphy and analogy in non-concatenative morphology. 1998. xii, 261 pp. 167 Benmamoun, Elabbas, Mushira Eid and Niloofar Haeri (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume XI: Atlanta, Georgia, 1997. 1998. viii, 231 pp. 166 Lemmens, Maarten: Lexical Perspectives on Transitivity and Ergativity. Causative constructions in English. 1998. xii, 268 pp.

165 Bubenik, Vit: A Historical Syntax of Late Middle Indo-Aryan (Apabhraṃśa). 1998. xxiv, 265 pp. 164 Schmid, Monika S., Jennifer R. Austin and Dieter Stein (eds.): Historical Linguistics 1997. Selected papers from the 13th International Conference on Historical Linguistics, Düsseldorf, 10–17 August 1997. 1998. x, 409 pp. 163 Lockwood, David G., Peter H. Fries and James E. Copeland (eds.): Functional Approaches to Language, Culture and Cognition. Papers in honor of Sydney M. Lamb. 2000. xxxiv, 656 pp. 162 Hogg, Richard M. and Linda van Bergen (eds.): Historical Linguistics 1995. Volume 2: Germanic linguistics.. Selected papers from the 12th International Conference on Historical Linguistics, Manchester, August 1995. 1998. x, 365 pp. 161 Smith, John Charles and Delia Bentley (eds.): Historical Linguistics 1995. Volume 1: General issues and non-Germanic Languages.. Selected papers from the 12th International Conference on Historical Linguistics, Manchester, August 1995. 2000. xii, 438 pp. 160 Schwegler, Armin, Bernard Tranel and Myriam Uribe-Etxebarria (eds.): Romance Linguistics: Theoretical Perspectives. Selected papers from the 27th Linguistic Symposium on Romance Languages (LSRL XXVII), Irvine, 20–22 February, 1997. 1998. vi, 349 pp. + index. 159 Joseph, Brian D., Geoffrey C. Horrocks and Irene Philippaki-Warburton (eds.): Themes in Greek Linguistics II. 1998. x, 335 pp. 158 Sánchez-Macarro, Antonia and Ronald Carter (eds.): Linguistic Choice across Genres. Variation in spoken and written English. 1998. viii, 338 pp. 157 Lema, José and Esthela Treviño (eds.): Theoretical Analyses on Romance Languages. Selected papers from the 26th Linguistic Symposium on Romance Languages (LSRL XXVI), Mexico City, 28–30 March, 1996. 1998. viii, 380 pp. 156 Matras, Yaron, Peter Bakker and Hristo Kyuchukov (eds.): The Typology and Dialectology of Romani. 1997. xxxii, 223 pp. 155 Forget, Danielle, Paul Hirschbühler, France Martineau and María Luisa Rivero (eds.): Negation and Polarity. Syntax and semantics. Selected papers from the colloquium Negation: Syntax and Semantics. Ottawa, 11–13 May 1995. 1997. viii, 367 pp. 154 Simon-Vandenbergen, Anne-Marie, Kristin Davidse and Dirk Noël (eds.): Reconnecting Language. Morphology and Syntax in Functional Perspectives. 1997. xiii, 339 pp. 153 Eid, Mushira and Robert R. Ratcliffe (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume X: Salt Lake City, 1996. 1997. vii, 296 pp. 152 Hiraga, Masako K., Christopher Sinha and Sherman Wilcox (eds.): Cultural, Psychological and Typological Issues in Cognitive Linguistics. Selected papers of the bi-annual ICLA meeting in Albuquerque, July 1995. 1999. viii, 338 pp. 151 Liebert, Wolf-Andreas, Gisela Redeker and Linda R. Waugh (eds.): Discourse and Perspective in Cognitive Linguistics. 1997. xiv, 270 pp. 150 Verspoor, Marjolijn H., Kee Dong Lee and Eve Sweetser (eds.): Lexical and Syntactical Constructions and the Construction of Meaning. Proceedings of the Bi-annual ICLA meeting in Albuquerque, July 1995. 1997. xii, 454 pp. 149 Hall, T. Alan: The Phonology of Coronals. 1997. x, 176 pp. 148 Wolf, George and Nigel Love (eds.): Linguistics Inside Out. Roy Harris and his critics. 1997. xxviii, 344 pp. 147 Hewson, John: The Cognitive System of the French Verb. 1997. xii, 187 pp. 146 Hinskens, Frans, Roeland van Hout and W. Leo Wetzels (eds.): Variation, Change, and Phonological Theory. 1997. x, 314 pp. 145 Hewson, John and Vit Bubenik: Tense and Aspect in Indo-European Languages. Theory, typology, diachrony. 1997. xii, 403 pp. 144 Singh, Rajendra (ed.): Trubetzkoy's Orphan. Proceedings of the Montréal Roundtable on “Morphonology: contemporary responses” (Montréal, October 1994). In collaboration with Richard Desrochers. 1996. xiv, 363 pp. 143 Athanasiadou, Angeliki and René Dirven (eds.): On Conditionals Again. 1997. viii, 418 pp. 142 Salmons, Joseph C. and Brian D. Joseph (eds.): Nostratic. Sifting the Evidence. 1998. vi, 293 pp. 141 Eid, Mushira and Dilworth B. Parkinson (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume IX: Washington D.C., 1995. 1996. xiii, 249 pp. 140 Black, James R. and Virginia Motapanyane (eds.): Clitics, Pronouns and Movement. 1997. 375 pp. 139 Black, James R. and Virginia Motapanyane (eds.): Microparametric Syntax and Dialect Variation. 1996. xviii, 269 pp. 138 Sackmann, Robin and Monika Budde (eds.): Theoretical Linguistics and Grammatical Description. Papers in honour of Hans-Heinrich Lieb. 1996. x, 375 pp.

137 Lippi-Green, Rosina L. and Joseph C. Salmons (eds.): Germanic Linguistics. Syntactic and diachronic. 1996. viii, 192 pp. 136 Mitkov, Ruslan and Nicolas Nicolov (eds.): Recent Advances in Natural Language Processing. Selected Papers from RANLP ’95. 1997. xii, 472 pp. 135 Britton, Derek (ed.): English Historical Linguistics 1994. Papers from the 8th International Conference on English Historical Linguistics (8 ICEHL, Edinburgh, 19–23 September 1994). 1996. viii, 403 pp. 134 Eid, Mushira (ed.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume VIII: Amherst, Massachusetts 1994. 1996. vii, 261 pp. 133 Zagona, Karen (ed.): Grammatical Theory and Romance Languages. Selected papers from the 25th Linguistic Symposium on Romance Languages (LSRL XXV) Seattle, 2–4 March 1995. 1996. vi, 330 pp. 132 Herschensohn, Julia: Case Suspension and Binary Complement Structure in French. 1996. xi, 200 pp. 131 Hualde, José Ignacio, Joseba A. Lakarra and R.L. Trask (eds.): Towards a History of the Basque Language. 1996. 365 pp. 130 Eid, Mushira (ed.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume VII: Austin, Texas 1993. 1995. vii, 192 pp. 129 Levin, Saul: Semitic and Indo-European. Volume I: The Principal Etymologies. With observations on Afro-Asiatic. 1995. xxii, 514 pp. 128 Guy, Gregory R., Crawford Feagin, Deborah Schiffrin and John Baugh (eds.): Towards a Social Science of Language. Papers in honor of William Labov. Volume 2: Social interaction and discourse structures. 1997. xviii, 358 pp. 127 Guy, Gregory R., Crawford Feagin, Deborah Schiffrin and John Baugh (eds.): Towards a Social Science of Language. Papers in honor of William Labov. Volume 1: Variation and change in language and society. 1996. xviii, 436 pp. 126 Matras, Yaron (ed.): Romani in Contact. The history, structure and sociology of a language. 1995. xvii, 208 pp. 125 Singh, Rajendra (ed.): Towards a Critical Sociolinguistics. 1996. xiii, 342 pp. 124 Andersen, Henning (ed.): Historical Linguistics 1993. Selected papers from the 11th International Conference on Historical Linguistics, Los Angeles, 16–20 August 1993. 1995. x, 460 pp. 123 Amastae, Jon, Grant Goodall, M. Montalbetti and M. Phinney (eds.): Contemporary Research in Romance Linguistics. Papers from the XXII Linguistic Symposium on Romance Languages, El Paso/Juárez, February 22–24, 1992. 1995. viii, 381 pp. 122 Smith, John Charles and Martin Maiden (eds.): Linguistic Theory and the Romance Languages. 1995. xiii, 240 pp. 121 Hasan, Ruqaiya, Carmel Cloran and David G. Butt (eds.): Functional Descriptions. Theory in practice. 1996. xxxvi, 381 pp. 120 Stonham, John T.: Combinatorial Morphology. 1994. xii, 206 pp. 119 Lippi-Green, Rosina L.: Language Ideology and Language Change in Early Modern German. A sociolinguistic study of the consonantal system of Nuremberg. 1994. xiv, 150 pp. 118 Hasan, Ruqaiya and Peter H. Fries (eds.): On Subject and Theme. A discourse functional perspective. 1995. xii, 414 pp. 117 Philippaki-Warburton, Irene, Katerina Nicolaidis and Maria Sifianou (eds.): Themes in Greek Linguistics. Papers from the First International Conference on Greek Linguistics, Reading, September 1993. 1994. xviii, 534 pp. 116 Miller, D. Gary: Ancient Scripts and Phonological Knowledge. 1994. xvi, 139 pp. 115 Eid, Mushira, Vicente Cantarino and Keith Walters (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. VolumeVI: Columbus, Ohio 1992. 1994. viii, 238 pp. 114 Egli, Urs, Peter E. Pause, Christoph Schwarze, Arnim von Stechow and Götz Wienold (eds.): Lexical Knowledge in the Organization of Language. 1995. xiv, 367 pp. 113 Moreno Fernández, Francisco, Miguel Fuster and Juan Jose Calvo (eds.): English Historical Linguistics 1992. Papers from the 7th International Conference on English Historical Linguistics, Valencia, 22–26 September 1992. 1994. viii, 388 pp. 112 Culioli, Antoine: Cognition and Representation in Linguistic Theory. Texts selected, edited and introduced by Michel Liddle. Translated with the assistance of John T. Stonham. 1995. x, 161 pp. 111 Tobin, Yishai: Invariance, Markedness and Distinctive Feature Analysis. A contrastive study of sign systems in English and Hebrew. 1994. xxii, 406 pp. 110 Simone, Raffaele (ed.): Iconicity in Language. 1995. xii, 315 pp. 109 Pagliuca, William (ed.): Perspectives on Grammaticalization. 1994. xx, 306 pp.

108 Lieb, Hans-Heinrich: Linguistic Variables. Towards a unified theory of linguistic variation. 1993. xiv, 261 pp. 107 Marle, Jaap van (ed.): Historical Linguistics 1991. Papers from the 10th International Conference on Historical Linguistics, Amsterdam, August 12–16, 1991. 1993. xviii, 395 pp. 106 Aertsen, Henk and Robert J. Jeffers (eds.): Historical Linguistics 1989. Papers from the 9th International Conference on Historical Linguistics, New Brunswick, 14–18 August 1989. 1993. xviii, 538 pp. 105 Hualde, José Ignacio and Jon Ortiz de Urbina (eds.): Generative Studies in Basque Linguistics. 1993. vi, 334 pp. 104 Kurzová, Helena: From Indo-European to Latin. The evolution of a morphosyntactic type. 1993. xiv, 259 pp. 103 Ashby, William J., Marianne Mithun and Giorgio Perissinotto (eds.): Linguistic Perspectives on Romance Languages. Selected Papers from the XXI Linguistic Symposium on Romance Languages, Santa Barbara, February 21–24, 1991. 1993. xxii, 404 pp. 102 Davis, Philip W. (ed.): Alternative Linguistics. Descriptive and theoretical modes. 1996. vii, 325 pp. 101 Eid, Mushira and Clive Holes (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume V: Ann Arbor, Michigan 1991. 1993. viii, 347 pp. 100 Mufwene, Salikoko S. and Lioba Moshi (eds.): Topics in African Linguistics. Papers from the XXI Annual Conference on African Linguistics, University of Georgia, April 1990. 1993. x, 304 pp. 99 Jensen, John T.: English Phonology. 1993. x, 251 pp. 98 Eid, Mushira and Gregory K. Iverson (eds.): Principles and Prediction. The analysis of natural language. Papers in honor of Gerald Sanders. 1993. xix, 382 pp. 97 Brogyanyi, Bela and Reiner Lipp (eds.): Comparative-Historical Linguistics: Indo-European and Finno-Ugric. Papers in honor of Oswald Szemerényi III. 1993. xii, 566 pp. 96 Lieb, Hans-Heinrich (ed.): Prospects for a New Structuralism. 1992. vii, 275 pp. 95 Miller, D. Gary: Complex Verb Formation. 1993. xx, 381 pp. 94 Hagège, Claude: The Language Builder. An essay on the human signature in linguistic morphogenesis. 1993. xii, 283 pp. 93 Lippi-Green, Rosina L. (ed.): Recent Developments in Germanic Linguistics. 1992. xii, 163 pp. 92 Poyatos, Fernando: Paralanguage: A linguistic and interdisciplinary approach to interactive speech and sounds. 1993. xii, 478 pp. 91 Hirschbühler, Paul and E.F.K. Koerner (eds.): Romance Languages and Modern Linguistic Theory. Selected papers from the XX Linguistic Symposium on Romance Languages, University of Ottawa, April 10–14, 1990. 1992. viii, 416 pp. 90 King, Larry D.: The Semantic Structure of Spanish. Meaning and grammatical form. 1992. xii, 287 pp. 89 Burridge, Kate: Syntactic Change in Germanic. Aspects of language change in Germanic with particular reference to Middle Dutch. 1993. xii, 287 pp. 88 Shields, Jr., Kenneth: A History of Indo-European Verb Morphology. 1992. viii, 160 pp. 87 Brogyanyi, Bela and Reiner Lipp (eds.): Historical Philology: Greek, Latin, and Romance. Papers in honor of Oswald Szemerényi II. 1992. xii, 386 pp. 86 Kess, Joseph F.: Psycholinguistics. Psychology, linguistics, and the study of natural language. 1992. xiv, 360 pp. 85 Broselow, Ellen, Mushira Eid and John McCarthy (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume IV: Detroit, Michigan 1990. 1992. viii, 282 pp. 84 Davis, Garry W. and Gregory K. Iverson (eds.): Explanation in Historical Linguistics. 1992. xiv, 238 pp. 83 Fife, James and Erich Poppe (eds.): Studies in Brythonic Word Order. 1991. x, 360 pp. 82 Van Valin, Jr., Robert D. (ed.): Advances in Role and Reference Grammar. 1992. xii, 569 pp. 81 Lehmann, Winfred P. and Helen-Jo Jakusz Hewitt (eds.): Language Typology 1988. Typological Models in the Service of Reconstruction. 1991. vi, 182 pp. 80 Comrie, Bernard and Mushira Eid (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume III: Salt Lake City, Utah 1989. 1991. xii, 274 pp. 79 Antonsen, Elmer H. and Hans Henrich Hock (eds.): STAEFCRAEFT: Studies in Germanic Linguistics. Selected papers from the 1st and 2nd Symposium on Germanic Linguistics, University of Chicago, 4 April 1985, and University of Illinois at Urbana-Champaign, 3–4 Oct. 1986. 1991. viii, 217 pp. 78 Kac, Michael B.: Grammars and Grammaticality. 1992. x, 259 pp. 77 Boltz, William G. and Michael C. Shapiro (eds.): Studies in the Historical Phonology of Asian Languages. 1991. viii, 249 pp.

76 Wickens, Mark A.: Grammatical Number in English Nouns. An empirical and theoretical account. 1992. xvi, 321 pp. 75 Droste, Flip G. and John E. Joseph (eds.): Linguistic Theory and Grammatical Description. Nine Current Approaches. 1991. viii, 354 pp. 74 Laeufer, Christiane and Terrell A. Morgan (eds.): Theoretical Analyses in Romance Linguistics. Selected papers from the Linguistic Symposium on Romance Languages XIX, Ohio State University, April 21–23, 1989. 1991. viii, 515 pp. 73 Stamenov, Maxim I. (ed.): Current Advances in Semantic Theory. 1991. xi, 565 pp. 72 Eid, Mushira and John McCarthy (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume II: Salt Lake City, Utah 1988. 1990. xiv, 332 pp. 71 O’Grady, William: Categories and Case. The sentence structure of Korean. 1991. vii, 294 pp. 70 Jensen, John T.: Morphology. Word structure in generative grammar. 1990. x, 210 pp. 69 Wanner, Dieter and Douglas A. Kibbee (eds.): New Analyses in Romance Linguistics. Selected papers from the Linguistic Symposium on Romance Languages XVIII, Urbana-Champaign, April 7–9, 1988. 1991. xviii, 385 pp. 68 Ball, Martin J., James Fife, Erich Poppe and Jenny Rowland (eds.): Celtic Linguistics/ Ieithyddiaeth Geltaidd. Readings in the Brythonic Languages. Festschrift for T. Arwyn Watkins. 1990. xxiv, 470 pp. 67 Lehmann, Winfred P. (ed.): Language Typology 1987. Systematic Balance in Language. Papers from the Linguistic Typology Symposium, Berkeley, 1–3 Dec 1987. 1990. x, 212 pp. 66 Andersen, Henning and E.F.K. Koerner (eds.): Historical Linguistics 1987. Papers from the 8th International Conference on Historical Linguistics, Lille, August 30-September 4, 1987. 1990. xii, 577 pp. 65 Adamson, Sylvia M., Vivien A. Law, Nigel Vincent and Susan Wright (eds.): Papers from the 5th International Conference on English Historical Linguistics. 1990. xxi, 583 pp. 64 Brogyanyi, Bela (ed.): Prehistory, History and Historiography of Language, Speech, and Linguistic Theory. Papers in honor of Oswald Szemerényi I. 1992. x, 414 pp. 63 Eid, Mushira (ed.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume I: Salt Lake City, Utah 1987. 1990. xiii, 290 pp. 62 Frajzyngier, Zygmunt (ed.): Current Progress in Chadic Linguistics. Proceedings of the International Symposium on Chadic Linguistics, Boulder, Colorado, 1–2 May 1987. 1989. vi, 312 pp. 61 Corrigan, Roberta L., Fred R. Eckman and Michael Noonan (eds.): Linguistic Categorization. Proceedings of an International Symposium in Milwaukee, Wisconsin, April 10–11, 1987. 1989. viii, 348 pp. 60 Kirschner, Carl and Janet Ann DeCesaris (eds.): Studies in Romance Linguistics. Selected Proceedings from the XVII Linguistic Symposium on Romance Languages. 1989. xii, 496 pp. 59 Voorst, Jan van: Event Structure. 1988. x, 181 pp. 58 Arbeitman, Yoël L. (ed.): Fucus: A Semitic/Afrasian Gathering in Remembrance of Albert Ehrman. 1988. xvi, 530 pp. 57 Bubenik, Vit: Hellenistic and Roman Greece as a Sociolinguistic Area. 1989. xvi, 331 pp. 56 Hockett, Charles F.: Refurbishing our Foundations. Elementary linguistics from an advanced point of view. 1987. x, 181 pp. 55 Hall, Jr., Robert A.: Linguistics and Pseudo-Linguistics. 1987. vii, 147 pp. 54 Weidert, Alfons: Tibeto-Burman Tonology. A comparative analysis. 1987. xvii, 512 pp. 53 Sankoff, David: Diversity and Diachrony. 1986. xii, 430 pp. 52 Fasold, Ralph W. and Deborah Schiffrin (eds.): Language Change and Variation. 1989. viii, 450 pp. 51 Chatterjee, Ranjit: Aspect and Meaning in Slavic and Indic. With a foreword by Paul Friedrich. 1989. xxiii, 137 pp. 50 Rudzka-Ostyn, Brygida (ed.): Topics in Cognitive Linguistics. 1988. x, 704 pp. 49 Waugh, Linda R. and Stephen Rudy (eds.): New Vistas in Grammar: Invariance and Variation. Proceedings of the Second International Roman Jakobson Conference, New York University, Nov. 5–8, 1985. 1991. x, 540 pp. 48 Giacalone-Ramat, Anna, Onofrio Carruba and Giuliano Bernini (eds.): Papers from the 7th International Conference on Historical Linguistics. 1987. xvi, 672 pp. 47 Lehmann, Winfred P. (ed.): Language Typology 1985. Papers from the Linguistic Typology Symposium, Moscow, 9–13 Dec. 1985. 1986. viii, 200 pp. 46 Prideaux, Gary D. and William J. Baker: Strategies and Structures. The processing of relative clauses. 1987. ix, 197 pp. 45 Koopman, Willem F., Frederike van der Leek, Olga Fischer and Roger Eaton (eds.): Explanation and Linguistic Change. 1986. viii, 300 pp.

44 Jungraithmayr, Herrmann and Walter W. Mueller (eds.): Proceedings of the Fourth International Hamito-Semitic Congress. 1987. xiv, 609 pp. 43 Akamatsu, Tsutomu: The Theory of Neutralization and the Archiphoneme in Functional Phonology. 1988. xxi, 533 pp. 42 Makkai, Adam and Alan K. Melby (eds.): Linguistics and Philosophy. Festschrift for Rulon S. Wells. 1985. xviii, 472 pp. 41 Eaton, Roger, Olga Fischer, Willem F. Koopman and Frederike van der Leek (eds.): Papers from the 4th International Conference on English Historical Linguistics, Amsterdam, April 10–13, 1985. 1985. xvii, 341 pp. 40 Fries, Peter H. and Nancy M. Fries (eds.): Toward an Understanding of Language. Charles C. Fries in Perspective. 1985. xvi, 384 pp. 39 Benson, James D., Michael J. Cummings and William S. Greaves (eds.): Linguistics in a Systemic Perspective. 1988. x, 452 pp. 38 Brogyanyi, Bela and Thomas Krömmelbein (eds.): Germanic Dialects. Linguistic and Philological Investigations. 1986. ix, 693 pp. 37 Griffen, Toby D.: Aspects of Dynamic Phonology. 1985. ix, 302 pp. 36 King, Larry D. and Catherine A. Maley (eds.): Selected papers from the XIIIth Linguistic Symposium on Romance. Languages, Chapel Hill, N.C., 24–26 March 1983. 1985. x, 440 pp. 35 Collinge, N.E.: The Laws of Indo-European. 1985. xviii, 273 pp. 34 Fisiak, Jacek (ed.): Papers from the VIth International Conference on Historical Linguistics, Poznaön, 22–26 August 1983. 1985. xxiii, 622 pp. 33 Versteegh, Kees: Pidginization and Creolization. The Case of Arabic. 1984. xiii, 194 pp. 32 Copeland, James E. (ed.): New Directions in Linguistics and Semiotics. 1984. xi, 269 pp. 31 Guillaume, Gustave (1883–1960): Foundations for a Science of Language. Texts selected by Roch Valin. Translated and with an introduction by Walter Hirtle and John Hewson. 1984. xxiv, 175 pp. 30 Hall, Jr., Robert A.: Proto-Romance Morphology. Comparative Romance Grammar, vol. III. 1984. xii, 304 pp. 29 Paprotté, Wolf and René Dirven (eds.): The Ubiquity of Metaphor: Metaphor in language and thought. 1985. iii, 628 pp. 28 Bynon, James (ed.): Current Progress in Afro-Asiatic Linguistics. Papers of the Third International Hamito-Semitic Congress, London, 1978. 1984. xi, 505 pp. 27 Bomhard, Allan R.: Toward Proto-Nostratic. A New Approach to the Comparison of Proto-IndoEuropean and Proto-Afroasiatic. With a foreword by Paul J. Hopper. 1984. xi, 356 pp. 26 Baldi, Philip (ed.): Papers from the XIIth Linguistic Symposium on Romance Languages, University Park, April 1–3, 1982. 1984. xii, 611 pp. 25 Andersen, Paul Kent: Word Order Typology and Comparative Constructions. 1983. xvii, 245 pp. 24 Lehmann, Winfred P. and Yakov Malkiel (eds.): Perspectives on Historical Linguistics. Papers from a conference held at the meeting of the Language Theory Division, Modern Language Assn., San Francisco, 27–30 December 1979. 1982. xii, 379 pp. 23 Danielsen, Niels: Papers in Theoretical Linguistics. Edited by Per Baerentzen. 1992. xxii, 224 pp. 22 Untermann, Jürgen und Bela Brogyanyi (Hrsg.): Das Germanische und die Rekonstruktion der Indogermanischen Grundsprache. Akten des Freiburger Kolloquiums der Indogermanischen Gesellschaft, Freiburg, 26–27 Februar 1981.. Proceedings of the Colloquium of the Indogermanische Gesellschhaft, Freiburg, 26–27 February 1981. 1984. xvii, 237 pp. 21 Ahlqvist, Anders (ed.): Papers from the Fifth International Conference on Historical Linguistics, Galway, April 6–10 1981. 1982. xxix, 527 pp. 20 Norrick, Neal R.: Semiotic Principles in Semantic Theory. 1981. xiii, 252 pp. 19 Ramat, Paolo (ed.): Linguistic Reconstruction and Indo-European Syntax. Proceedings of the Colloquium of the 'Indogermanische Gesellschaft'. University of Pavia, 6–7 September 1979. viii, 263 pp. Expected Out of print 18 Izzo, Herbert J. (ed.): Italic and Romance. Linguistic Studies in Honor of Ernst Pulgram. 1980. xxi, 338 pp. 17 Lieb, Hans-Heinrich: Integrational Linguistics I. 1984. xxiii, 527 pp. 16 Arbeitman, Yoël L. and Allan R. Bomhard (eds.): Bono Homini Donum. Essays in Historical Linguistics, in Memory of J. Alexander Kerns. (2 volumes). 1981. xvi, 557, viii, 581 pp. 15 Anderson, John A. (ed.): Language Form and Linguistic Variation. Papers dedicated to Angus McIntosh. 1982. viii, 496 pp. 14 Closs Traugott, Elizabeth, Rebecca Labrum and Susan C. Shepherd (eds.): Papers from the Fourth International Conference on Historical Linguistics, Stanford, March 26–30 1979. 1980. x, 437 pp.

Proto-Japanese; Issues and Prospects (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Saami Linguistics (Amsterdam Studies in the Theory and History of Linguistic Science Series IV: Current Issues in Linguistics Theory)

Postvelar Harmony (Current Issues in Linguistic Theory)

Linguistics and Philosophy: Essays in Honor of Rulon S. Wells (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Grammar from the Human Perspective: Case, Space and Person in Finnish (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Loan Phonology (Current Issues in Linguistic Theory)

Anaphora Processing (Current Issues in Linguistic Theory)

Loan Phonology (Current Issues in Linguistic Theory)

Emotion in Dialogic Interaction: Advances in the Complex (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Verb Constructions in German and Dutch (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Subordination and Coordination Strategies in North Asian Languages (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Simultaneity in Signed Languages: Form and Function (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Phonetics and Phonology: Interactions and interrelations (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Morphological Analysis in Comparison (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Papers in theoretical linguistics (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

The Laws of Indo-European (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

The Paradox of Grammatical Change: Perspectives from Romance (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory, CILT 293)

Language Typology: A Functional Perspective (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Language Typology: A Functional Perspective (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

English Adjective Comparison: A Historical Perspective (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory, Volume 299)

English Adjective Comparison: A Historical Perspective (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory, Volume 299)

Typological Models in Reconstruction (Amsterdam Studies in the Theory & History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Romance Languages and Linguistic Theory 2004: Selected Papers from 'Going Romance', Leiden, 911 December 2004 (Amsterdam Studies in the Theory and History ... IV: Current Issues in Linguistic Theory)

Germanic linguistics: syntactic and diachronic (Current Issues in Linguistic Theory)

Ancient Scripts and Phonological Knowledge (Current Issues in Linguistic Theory)

The Transformational-Generative Paradigm and Modern Linguistic Theory (Current Issues in Linguistic Theory)

Connectives in the History of English (Amsterdam Studies in the Theory and History of Linguistic Science)

Ergativity: Emerging Issues (Studies in Natural Language and Linguistic Theory)

Competing Models of Linguistic Change: Evolution and Beyond (Current Issues in Linguistic Theory)

Clinical Linguistics: Theory and Applications in Speech Pathology and Therapy (Current Issues in Linguistic Theory)

Linguistic Variables: Towards a Unified Theory of Linguistic Variation (Current Issues in Linguistic Theory) issue 0304-0763

Proto-Japanese; Issues and Prospects (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Saami Linguistics (Amsterdam Studies in the Theory and History of Linguistic Science Series IV: Current Issues in Linguistics Theory)

Postvelar Harmony (Current Issues in Linguistic Theory)

Linguistics and Philosophy: Essays in Honor of Rulon S. Wells (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Grammar from the Human Perspective: Case, Space and Person in Finnish (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Loan Phonology (Current Issues in Linguistic Theory)

Anaphora Processing (Current Issues in Linguistic Theory)

Loan Phonology (Current Issues in Linguistic Theory)

Emotion in Dialogic Interaction: Advances in the Complex (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Verb Constructions in German and Dutch (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Subordination and Coordination Strategies in North Asian Languages (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Simultaneity in Signed Languages: Form and Function (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Phonetics and Phonology: Interactions and interrelations (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Morphological Analysis in Comparison (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Papers in theoretical linguistics (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

The Laws of Indo-European (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

The Paradox of Grammatical Change: Perspectives from Romance (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory, CILT 293)

Language Typology: A Functional Perspective (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Language Typology: A Functional Perspective (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

English Adjective Comparison: A Historical Perspective (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory, Volume 299)

English Adjective Comparison: A Historical Perspective (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory, Volume 299)

Typological Models in Reconstruction (Amsterdam Studies in the Theory & History of Linguistic Science, Series IV: Current Issues in Linguistic Theory)

Romance Languages and Linguistic Theory 2004: Selected Papers from 'Going Romance', Leiden, 911 December 2004 (Amsterdam Studies in the Theory and History ... IV: Current Issues in Linguistic Theory)

Germanic linguistics: syntactic and diachronic (Current Issues in Linguistic Theory)

Ancient Scripts and Phonological Knowledge (Current Issues in Linguistic Theory)

The Transformational-Generative Paradigm and Modern Linguistic Theory (Current Issues in Linguistic Theory)

Connectives in the History of English (Amsterdam Studies in the Theory and History of Linguistic Science)

Ergativity: Emerging Issues (Studies in Natural Language and Linguistic Theory)

Competing Models of Linguistic Change: Evolution and Beyond (Current Issues in Linguistic Theory)

Clinical Linguistics: Theory and Applications in Speech Pathology and Therapy (Current Issues in Linguistic Theory)

Linguistic Variables: Towards a Unified Theory of Linguistic Variation (Current Issues in Linguistic Theory) issue 0304-0763

Recommend Documents