Studies in Japanese and Korean Historical and Theoretical Linguistics and Beyond: Festschrift Presented to John B. Whitman 9004351132, 9789004351134

Table of contents :
‎Contents
‎Preface
‎Acknowledgments
‎List of Tables
‎List of Contributors
‎John B. Whitman Bibliography
‎Tabula Gratulatoria
‎Part 1. Documentation
‎Chapter 1. The Digital Museum Project for the Documentation of Endangered Languages: The Case of Ikema Ryukyuan (Takubo)
‎Part 2. Historical Linguistics
‎Chapter 2. Disentangling Japonic Seaweed from Koreo-Japonic Water (Antonov)
‎Chapter 3. On Feature Ranking in Japanese Onset Obstruents (Frellesvig)
‎Chapter 4. Fishy Rhymes: Sino-Korean Evidence for Earlier Korean *e (Miyake)
‎Chapter 5. A mokkan Perspective on Some Issues in Japanese Historical Phonology (Osterkamp)
‎Chapter 6. A (More) Comparative Approach to Some Japanese Etymologies (Pellard)
‎Chapter 7. The Role of Internal Reconstruction in Comparing the Accent Systems of Korean Dialects (Ramsey)
‎Chapter 8. How Many OJ Syllables are Reflected in EMJ yo? (Unger)
‎Chapter 9. On the Etymology of the Name of Mt. Fuji (Vovin)
‎Part 3. Theoretical Linguistics
‎Chapter 10. Against a VP Ellipsis Account of Russian Verb-Stranding Constructions (Bailyn)
‎Chapter 11. A New Approach to -zhe in Mandarin Chinese (Djamouri and Paul)
‎Chapter 12. Japanese Experiential -te iru (Hughes and McClure)
‎Chapter 13. DP versus NP: A Cross-Linguistic Typology? (Kornfilt)
‎Chapter 14. The Old Japanese Accusative Revisited: Realizing All the Universal Options (Miyagawa)
‎Chapter 15. Japanese Wh-Phrases as Unvalued Operators (Saito)
‎Index

Citation preview

Studies in Japanese and Korean Historical and Theoretical Linguistics and Beyond

The Languages of Asia Series Series Editor Alexander Vovin (ehess/crlao, Paris, France)

Associate Editor José Andrés Alonso de la Fuente ( Jagiellonian University in Kraków, Poland)

Editorial Board Mark Alves (Montgomery College, Rockville, md, usa) Anna Bugaeva (Tokyo University of Science, Japan) Bjarke Frellesvig (Oxford University, uk) Guillaume Jacques (cnrs/crlao, Paris, France) Juha Janhunen (University of Helsinki, Finland) Ross King (University of British Columbia, Vancouver, Canada) Marc Miyake (British Museum, London, uk) Mehmet Ölmez (Yıldız Technical University, Istanbul, Turkey) Toshiki Osada (Institute of Nature and Humanity, Kyoto, Japan) Pittawayat Pittayaporn (Chulalongkorn University, Bangkok, Thailand) Pavel Rykin (St. Petersburg Institute of Linguistic Studies, Russia) Claus Schönig (Freie Universität Berlin, Germany) Marek Stachowski ( Jagiellonian University in Kraków, Poland) Yukinori Takubo (National Institute for Japanese Language and Linguistics, Tachikawa, Japan) John Whitman (Cornell University, Ithaca, ny, usa)

volume 16 The titles published in this series are listed at brill.com/la

John B. Whitman photo: dave burbank

Studies in Japanese and Korean Historical and Theoretical Linguistics and Beyond Festschrift Presented to John B. Whitman

Edited by

William McClure Alexander Vovin

leiden | boston

Cover illustration: Clockwise from the left: (1) painting on a wooden panel in the Pulguksa temple, Kyǒngju city, Korea, photo taken in 2005; (2) stones in the stone garden of the Ryōanji temple, Kyōto city, Japan, photo taken in 2001; (3) stone structure in the Buddhist temple of Maisan mountain, Cǒlla namdo province, Korea, photo taken in 2008; (4) a pagoda in the Yakushiji temple, Nara city, Japan, photo taken in 2008. Library of Congress Cataloging-in-Publication Data Names: McClure, William, 1962- editor. | Vovin, Alexander, editor. | Whitman, John, 1954- honoree. Title: Studies in Japanese and Korean historical and theoretical linguistics and beyond : festschrift presented to John B. Whitman / edited by William McClure, Alexander Vovin. Description: Leiden ; Boston : Brill, [2018] | Series: The languages of Asia series, issn 2452-2961 ; volume 16 | Includes bibliographical references and index. Identifiers: lccn 2017041309 (print) | lccn 2017051087 (ebook) | isbn 9789004351134 (e-book) | isbn 9789004350854 (hardcover : alk. paper) Subjects: lcsh: Japanese language–History. | Korean language–History. | Japanese language–Grammar, Comparative–Korean. | Korean language–Grammar, Comparative–Japanese. Classification: lcc pl508.w45 (ebook) | lcc pl508.w45 s78 2017 (print) | ddc 495.6–dc23 lc record available at https://lccn.loc.gov/2017041309

Typeface for the Latin, Greek, and Cyrillic scripts: “Brill”. See and download: brill.com/brill-typeface. issn 2452-2961 isbn 978-90-04-35085-4 (hardback) isbn 978-90-04-35113-4 (e-book) Copyright 2018 by Koninklijke Brill nv, Leiden, The Netherlands. Koninklijke Brill nv incorporates the imprints Brill, Brill Hes & De Graaf, Brill Nijhoff, Brill Rodopi, Brill Sense and Hotei Publishing. All rights reserved. No part of this publication may be reproduced, translated, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission from the publisher. Authorization to photocopy items for internal or personal use is granted by Koninklijke Brill nv provided that the appropriate fees are paid directly to The Copyright Clearance Center, 222 Rosewood Drive, Suite 910, Danvers, ma 01923, usa. Fees are subject to change. This book is printed on acid-free paper and produced in a sustainable manner.

Contents Preface ix Acknowledgments xii List of Tables xiii List of Contributors xiv John B. Whitman Bibliography Tabula Gratulatoria xxxi

xvi

part 1 Documentation 1

The Digital Museum Project for the Documentation of Endangered Languages: The Case of Ikema Ryukyuan 3 Yukinori Takubo

part 2 Historical Linguistics 2

Disentangling Japonic Seaweed from Koreo-Japonic Water Anton Antonov

15

3

On Feature Ranking in Japanese Onset Obstruents Bjarke Frellesvig

4

Fishy Rhymes: Sino-Korean Evidence for Earlier Korean *e Marc H. Miyake

5

A mokkan Perspective on Some Issues in Japanese Historical Phonology 45 Sven Osterkamp

6

A (More) Comparative Approach to Some Japanese Etymologies Thomas Pellard

22

37

56

viii

contents

7

The Role of Internal Reconstruction in Comparing the Accent Systems of Korean Dialects 66 S. Robert Ramsey

8

How Many oj Syllables are Reflected in emj yo? J. Marshall Unger

9

On the Etymology of the Name of Mt. Fuji Alexander Vovin

76

80

part 3 Theoretical Linguistics 10

Against a vp Ellipsis Account of Russian Verb-Stranding Constructions 93 John Frederick Bailyn

11

A New Approach to -zhe in Mandarin Chinese Redouane Djamouri and Waltraud Paul

12

Japanese Experiential -te iru 124 Mamori Sugita Hughes and William McClure

13

dp versus np: A Cross-Linguistic Typology? Jaklin Kornfilt

14

The Old Japanese Accusative Revisited: Realizing All the Universal Options 159 Shigeru Miyagawa

15

Japanese Wh-Phrases as Unvalued Operators Mamoru Saito Index

191

110

138

179

Preface This volume was long in the making: it was supposed to appear in 2014, but due to unforeseen circumstances, such as the move of one of the editors from the usa to France, and different health problems, it was delayed for three years. We offer our sincere apologies to John Whitman and to all the contributors and thank everyone for their patience. The dual nature of John Whitman’s interests in both historical and theoretical linguistics has dictated the overall structure of this congratulatory volume: the editors come from these two separate disciplines and we have decided to invite scholars who, to the best of our knowledge, are the closest associates of John Whitman in historical and theoretical linguistics to contribute their papers to the volume. Since we were quite limited in volume, we offer our humble apologies for not inviting more people, who doubtlessly would also have been able to offer significant contributions. Although the reader might be under the mistaken impression that the historical part overshadows the theoretical by the number of articles, page-wise the two parts are equal, because articles in the historical linguistics tend to be much shorter than theoretical contributions. John Whitman is a rare scholar of international repute who inspired all of the contributors and many other scholars throughout his more than thirtyyear career as a scholar and teacher. He received his PhD from Harvard in 1985, and his dissertation The Phonological Basis for the Comparison of Japanese and Korean remains the starting point for comparing the historical relationship of Japanese and Korean. At the same time, his 1987 paper, ‘Configurationality parameters’ (originally written in 1982), addressed the then timely topic of whether Japanese is a syntactically configurational language or not. That is, does a scrambling language like Japanese have a flat syntax or does it have a vp? Both of these works are cited to this day. From the very beginning, we see that John’s work as a linguist has always reflected his profound knowledge of both phonology and syntax in the context of both synchronic and diachronic perspectives. The breadth of his knowledge is reflected in hundreds of publications on Japanese and Korean, and his intellectual legacy is carried on by his many, many students. John’s interest in languages began in high school where he studied French, Latin, and, during a year in Burma, Burmese. Graduating from high school, he enrolled at Harvard where he started studying philosophy as well as Russian and Japanese (the latter of which he quickly dropped). At the same time, he also discovered linguistics. He credits Alice Harris, his first ta in linguistics, with

x

preface

enabling his success in those introductory courses. An early assignment was to write a grammatical sketch of a language—John chose Kachin (Jinghpaw), spoken in northern Burma. The only extant grammar of Kachin was in Russian, so his rudimentary Russian helped there. More importantly, The Sound Pattern of English had been recently published, and, working with Alice, John tried to figure out how to describe Kachin in terms of the many new features proposed by Chomsky and Halle. A young linguist with an interest in Asian languages was born. During his undergraduate career, John took full advantage of the faculty teaching in linguistics and philosophy at Harvard. (He specifically remembers a b- that he received from W.V.O. Quine in Quine’s graduate seminar “Word and Object”.) Significantly for his future career, in his sophomore year, John was hired by Susumu Kuno to work on an nsf grant. Professor Kuno also became John’s advisor, and his main advice might be paraphrased as “if you are going to work with me, you should study Japanese.” Starting in the summer of his sophomore year, John therefore began to study Japanese again. In John’s senior year, Professor Kuno advised him to apply for a Mombushō Fellowship to continue his study of Japanese in Japan. (Professor Kuno has since been heard to say that John was the best student of Japanese he had ever met.) He recommended study at Tsukuba University, which was new, located in the middle of nowhere, and full of scholars of the Japanese language transported from Tokyo University of Education and elsewhere. John ultimately received an ma from Tsukuba, and already he was working in both the syntactic and historical domains. His first paper, ‘Scrambled over easy and sunny side up’ was published by the Chicago Linguistic Society in 1979. This paper argues that Japanese has a flat syntax with no vp. It was later contradicted by the more well-known ‘Configurationality parameters’, published in 1987. While living in Tsukuba, John got married and his first son was born. He also met Wesley Jacobsen at Tsukuba (to this day they only speak Japanese to each other), and through Wesley he was introduced to James Unger’s thesis on the internal reconstruction of Japanese. These various experiences culminated in his decision to return to Harvard to focus on the study of Japanese historical linguistics. John’s doctoral work in historical and functional Japanese and Korean was directed by Susumu Kuno. In addition, he took classes in syntax and generative grammar at mit with Noam Chomsky and his colleagues. While he had studied a great deal of Japanese, in order to learn Korean, he took his family to Korea on a Fulbright in 1982, where he studied and worked at Yonsei, Korea, and Seoul National Universities. In 1985, his dissertation committee included Susumu Kuno, Jochem Schindler, Calvert Watkins, and Samuel Martin. That is, John’s

xi

preface

committee included expertise in functional and historical linguistics, Japanese, Korean, Altaic, Indo-European, and Classics. As noted already, the dissertation itself is entitled The Phonological Basis for the Comparison of Japanese and Korean. From 1984 to 1986, John taught introductory syntax at Harvard. In 1987, he moved to Ithaca, New York where he began teaching Japanese language and linguistics and syntax at Cornell. In addition to his teaching, he has served multiple terms as Chair of Linguistics and Director of the East Asia Program. From 2010 to 2012 he was a researcher and Section Head at the National Institute of Japanese and Linguistics in Tachikawa, where he is now emeritus. As is clear from his bibliography, his body of work is remarkable for its breadth. Since graduate school, he has been publishing in historical linguistics and syntax, with a focus on Japanese and Korean. In historical linguistics, he is known for his work on comparative phonology (the topic of many of the papers in this volume) as well as the development of verbal conjugations. In syntax, he has made well-known contributions in the areas of configurationality, phrase structure and word order, relative clauses, and rightward movement. Likewise, most of his work on language variation is focused on Japanese and Korean, although there is a separate strand on historical Chinese syntax. This work on Chinese has evolved into a general interest in diachronic syntax. He has also published in language acquisition and, quite recently, glossing systems in Chinese and Japanese (comparing them to European systems). During the past thirty years, John has taught hundreds of undergraduates, help train scores of graduate students, and directed twenty-five PhD dissertations. His students work in Japanese and Korean of all historical periods. They also work in, to name a few, Greek, Indonesian, Korean, Manchu, Mandarin, Seediq, Tagalog, and Thai. John’s students are now found around the world, and a remarkable number of them have become advanced scholars of the middle generation or highly promising scholars of the younger generation. Currently, he is teaching Linguistic 101, introducing a new group of Cornell students to his self-deprecating style and dry sense of humor, and, in his spare time, he is once again studying Burmese. Life is a circle. We congratulate John Whitman on all of his outstanding activities throughout these long years, and we wish him many more years of active research and training of excellent students. William McClure and Alexander Vovin New York, usa and Poligny, France February 14, 2017

Acknowledgments Our acknowledgements go first to John Whitman himself who was a source of constant inspiration for the research of editors and all the contributors, as well as of all colleagues who signed Tabula Gratulatoria. Second, we thank all the contributors to this volume not only for taking time out of their busy schedules, but also for being extremely patient and understanding with our procrastination in preparation the book for publication. Third, we are extremely grateful to Brill’s editors, the managing editor of linguistics, Irene Van Rossum, and the production manager Maarten Frieswijk for accepting this volume into Languages of Asia Series and walking us through all the necessary steps in improving the manuscript. Finally, the last, but not the least, we thank the members of our families for being patient and supportive during the several years that finally brought this book to fruition.

List of Tables 4.1 4.2 4.3 4.4 4.5 4.6 4.7 9.1 9.2

Selected correspondences between Sino-Korean and Sino-Japanese rhymes 37 Examples of Late Middle Chinese *-jej corresponding to Sino-Korean -(y)e 38 Examples of Middle Chinese *-ɨə/*-jə corresponding to Sino-Korean -(y)ey 39 The ‘even’, ‘sacrifice’, and ‘fish’ rhymes in modern Chinese and Sino-Vietnamese 39 Denasalization in Chinese and Sinoxenic 40 The Middle Chinese 皆 ‘all’ rhyme after *k- in Sino-Korean 42 Late Middle Chinese *kjwi and *gjwi in Sino-Korean 43 Spellings of oj punzi ‘Mt. Fuji’ 83 Accentual history of Fuji 84

List of Contributors Anton Antonov Maître de conférences, inalco/crlao, Paris John Frederick Bailyn Professor, Stony Brook University Redouane Djamouri Directeur de recherche, cnrs/crlao, Paris Bjarke Frellesvig Professor, Hertford College & Faculty of Oriental Studies, Oxford University Mamori Sugita Hughes Professor, Queens College and the Graduate Center of the City University of New York William McClure Professor, Queens College and the Graduate Center of the City University of New York Jaklin Kornfilt Professor, Syracuse University Shigeru Miyagawa Professor, Massachusets Institute of Technology Marc H. Miyake Researcher, The British Museum Sven Osterkamp Professor, Ruhr Universität Bochum Waltraud Paul Directrice de recherché, cnrs/crlao. Paris Thomas Pellard Chargé de recherché, cnrs/crlao, Paris

list of contributors

xv

S. Robert Ramsey Professor, University of Maryland Mamoru Saito Professor, Nanzan University Yukinori Takubo Professor (Emeritus), Kyoto University, National Institute for Japanese Language and Linguistics J. Marshall Unger Professor (Emeritus), The Ohio State University Alexander Vovin Directeur d’études, ehess/ crlao, Paris

John B. Whitman Bibliography Books Forthcoming The Phonological Basis for the Comparison of Japanese and Korean. Accepted for publication. Ann Arbor, Michigan Monographs in Japanese Studies. The Syntax of Korean. With Donna Gerdts. Under contract. Cambridge: Cambridge University Press. Korean: A Linguistic Introduction. With Sungdae Cho. Under contract. Cambridge: Cambridge University Press.

Already Published Nichieigo hikaku sensho, volume 9: Kaku to gojun to tōgo kōzō. [Case, Word Order, and Syntactic Structure]. With Koichi Takezawa. 1998. Tokyo: Kenkyûsha.

Edited Books Ryūkyū shogo to kodai nihongo: Nichiryū sogo no saiken ni mukete [Ryūkyūan and Premodern Japanese: Toward the Reconstruction of Proto-Japanese-Ryūkyūan]. Edited with Yukinori Takubo and Tatsuya Hirako. 2016. Tokyo: Hitsuji Shobō. Grammatical Change: Origins, Nature, Outcomes. Edited with Dianne Jonas and Andrew Garrett. 2011. Oxford: Oxford University Press. Proto-Japanese. Edited with Bjarke Frellesvig. 2008. Amsterdam: John Benjamins. Studies in Korean Syntax and Semantics by Susumu Kuno. Edited with Young-Key KimRenaud. 2004. Seoul: Pajigong Press. Syntactic and Functional Explorations: A Festschrift for Susumu Kuno. Edited with Kenichi Takami and Akio Kamio. 2000. Tokyo: Kuroshio Press. Korean Syntax and Semantics: Proceedings of the lsa Institute Workshop. Edited with Chungmin Lee. 2000. Seoul: Thaehaksa. Syntactic Theory and First Language Acquisition, Volume 1: Heads, Projections, and Learnability. Edited with Barbara Lust and Magui Suñer. 1994. Hillsdale, n.j.: Lawrence Erlbaum Associates.

john b. whitman bibliography

xvii

Edited Special Issues Kugyŏl/Kunten. Edited with J.R.P. King. To appear in Sino-Platonic Papers, University of Pennsylvania. Lecture vernaculaire des textes classique Chinois/Reading classical texts in the vernacular. Special issue editor, with Franck Cinato. 2014. Histoire Epistémologie Langage 36. Paris: Laboratoire Histoire des Théories Linguistiques (Université Paris Diderot). Korean historical linguistics. Special issue editor, with Young-key Kim-Renaud. 2013. Korean Linguistics 15.2. Nominalizations in linguistic theory. Special issue editor, with Jaklin Kornfilt. 2011. Lingua 121.7, pp. 1159–1313. Special Issue: First Language Acquisition of East Asian Languages. Edited with Yasuhiro Shirai. 2000. Journal of East Asian Linguistics 9.4.

Refereed Articles and Book Chapters Forthcoming ‘A multiple correspondence analysis of the latent structure of features in linguistic typology: A statistical reanalysis of Tsunoda, Ueda and Itoh (1995a)’. With Yohei Ono, Ryozo Yoshino, and Fumi Hayashi. Revision submitted 2016. Keiryō kokugogaku (Mathematical Linguistics). ‘Topic prominence’. With Waltraud Paul. To appear, 2017. In Everaert, Martin, Henk Van Riemsdijk, Rob Goedemans, Bart Hollebrandse (eds.), Blackwell Companion to Syntax. New York: Wiley.

Already Published ‘Diachronic interpretations of word order parameter cohesion’. With Yohei Ono. 2017. In Mathieu, Eric and Robert Truswell (eds.), From Micro-change to Macro-change, pp. 43–60. Oxford: Oxford University Press. ‘The historical source of the bigrade transitivity alternations in Japanese’. With Bjarke Frellesvig. 2016. In Kageyama, Taro and Wesely Jacobsen (eds.), Transitivity and Valency Alternations: Studies on Japanese and Beyond, pp. 289–312. Berlin: Mouton de Gruyter. ‘Nichiryū sogo no on’in taikei to rentaikei, izenkei no kigen [The phonological system of proto-Japanese-Ryūkyūan and the origin of the adnominal and realis forms]’. 2016. In Takubo, Yukinori, John Whitman, and Tatsuya Hirako (eds.), Ryūkyū shogo to kodai nihongo: Nichiryū sogo no saiken ni mukete [Ryūkyūan and Premodern Japanese: Toward the Reconstruction of Proto-Japanese-Ryūkyūan], pp. 21–38. Tokyo: Hitsuji Shobō.

xviii

john b. whitman bibliography

‘Atogaki [Afterword]’. 2016. In Takubo, Yukinori, John Whitman, and Tatsuya Hirako (eds.), Ryūkyū shogo to kodai nihongo: Nichiryū sogo no saiken ni mukete [Ryūkyūan and Premodern Japanese: Toward the Reconstruction of Proto-Japanese-Ryūkyūan]. pp. 291–296. Tokyo: Hitsuji Shobō. ‘Tōhoku Aziagengo chīki no ichizuke ni mukete [Toward the establishment of Northeast Asia as a linguistic area]’. March 2016. ninjal Project Review 6, pp. 69–82. ‘Old Korean’. 2015. In Brown, Lucien and Jae Hoon Yeon (eds.), The Handbook of Korean Linguistics, pp. 421–438. London: Wiley-Blackwell. ‘Raten-go kyōten no dokuhō to butten no kundoku [The reading of Latin clerical texts and vernacular reading of Classical Chinese]’ (in Japanese). 2015. In Niikawa, Tokio (ed.), Bunmei idō toshite no bukkyō [Buddhism as Cultural Movement], pp. 105–146. Tokyo: Benseisha. ‘Iwayuru ajia-shiki kankeisetsu ni tsuite [On so-called Asian-type relative clauses]’. 2015. In Fukuda, Shin, K. Nishida, and T. Tamura (eds.), Gengo kenkyū no shiza [Perspectives of Linguistic Research], pp. 188–203. Tokyo: Kaitakusha. ‘Uncertainty in processing relative clauses across East Asian languages’. With Jiwon Yun, Zhong Chen, Tim Hunter, and John Hale. 2015. Journal of East Asian Linguistics 24(2), pp. 113–148. ‘Comparative consequences of the tongue root harmony analysis for proto-Tungusic, proto-Mongolic, and proto-Korean’. With Seongyeon Ko and Andrew Joseph. 2014. In Robbeets, Martine, and Walter Bisang (eds.), Paradigm Change in the Transeurasian Languages and Beyond, pp. 141–176. Amsterdam: John Benjamins. ‘Présentation du projet: Reading classical texts in the vernacular’. In Whitman, John and Franck Cinato (eds.) Lecture Vernaculaire des Textes Classique Chinois/Reading Classical Texts in the Vernacular. 2014. In Histoire Epistémologie Langage 36. Paris: Laboratoire Histoire des Théories Linguistiques (Université Paris Diderot). ‘Postpositions vs. prepositions in Mandarin Chinese: The articulation of disharmony’. With Redouane Djamouri and Waltraud Paul. 2013. In Biberauer, Theresa (ed.), Theoretical Approaches to Disharmonic Word Order. pp. 74–105. Oxford, Oxford University Press. ‘Syntactic change in Chinese and the argument-adjunct asymmetry’. With Redouane Djamouri and Waltraud Paul. 2013. In Cao, Guangshun, Hillary Chappell, Redouane Djamouri, and Thekla Wiebusch (eds.), Breaking Down the Barriers: Interdisciplinary studies in Chinese linguistics and Beyond], pp. 577–594. Taipei: Academia Sinica, vol. 2, [Language and Linguistics monograph series 50. ‘The relationship between Japanese and Korean’. 2012. In Tranter, David N. (ed.), The Languages of Japan and Korea, pp. 24–38. London: Routledge. ‘The formal syntax of alignment change’. With Yuko Yanagida. 2012. In Galves, Charlotte, Sonia Cyrino, Ruth Lopes, Filomena Sandalo, Juanita Avelar (eds.), Parameter Theory & Linguistic Change, pp. 177–195. Oxford: Oxford University Press.

john b. whitman bibliography

xix

‘Genitive subjects in tp nominalizations’. With Jaklin Kornfilt. 2012. In Iordachioaia, Gianina (ed.), Proceedings of JeNom 4; Working Papers of the sfb 732; Stuttgart: opus; pp. 39–72. ‘Misparsing and syntactic reanalysis’. 2012. In Kemenade, Ans van & Nynke de Haas (eds.), Historical Linguistics 2009: Selected Papers from the 19th International Conference on Historical Linguistics, pp. 69–87. Amsterdam: John Benjamins. ‘Northeast Asian linguistic ecology and the advent of rice agriculture in Korea and Japan’. 2011. Rice 4.3–4, pp. 149–158. ‘The ubiquity of the gloss’. 2011. Scripta 3, pp. 95–121. ‘Introduction’. With Dianne Jonas and Andrew Garrett. 2011. Grammatical Change: Origins, Nature, Outcomes, pp. 1–12. Oxford: Oxford University Press. ‘Introduction: Nominalizations in linguistic theory’. (With Jaklin Kornfilt). 2011. In Kornfilt, Jaklin and John Whitman (eds.), Nominalizations in linguistic theory (special issue), Lingua.121.7, pp. 1160–1163. ‘Afterword: Nominalizations in linguistic theory’. With Jaklin Kornfilt. 2011. In Kornfilt, Jaklin and John Whitman (eds.), Nominalizations in linguistic theory (special issue), Lingua.121.7, pp. 1297–1313. ‘Toward an international vocabulary for research on vernacular reading of Chinese texts (漢文訓讀 Hanwen xundu)’. With Miyoung Oh, Jinho Park, Valerio Luigi Alberizzi, Masayuki Tsukimoto, Teiji Kosukegawa, and Tomokazu Takada. 2010. Scripta 2, pp. 61–84. ‘Hitei no kōzō to rekishiteki henka—shuyōbu to hitei kyokusei hyōgen o chūsin ni—(否 定構造と歴史的変化—主要部と否定極性表現を中心に—[The structure of negation and historical change—with a focus on headedness and negative polarity expressions])’. 2010. In Katō, Yasuhiko et al (eds.), Hitei to gengo riron [Negation and Linguistic Theory], pp. 141–169. Tokyo: Kaitakusha. ‘Applicative structure and Mandarin distransitives’. With Waltraud Paul. 2010. In Huidbro, Susan et al (eds.), Argument Structure and Syntactic Relations, pp. 261–182. Amsterdam: John Benjamins. ‘The syntax of overmarking and kes in child Korean’. 2010. In Lee, Chungmin (ed.), The Handbook of Korean Psycholinguistics, pp. 221–230. New York: Cambridge University Press. ‘Alignment and word order in Old Japanese’. (With Yuko Yanagida). 2009. Journal of East Asia Asian Linguistics 18, pp. 101–144. ‘Kokugogaku vs. gengogaku: Language process theory and Tokieda’s construction of Saussure, 60 years after’. 2009. In Bourdaghs, Michael (ed.), The Linguistic Turn in Contemporary Japanese Literary Studies: Textuality, Language, Politics. Ann Arbor: University of Michigan Center for Japanese Studies Publications. ‘Ninshō to kakkaku ruikei—jōdai nihongo no daimeishi taikei no kanten kara [Person and active typology—from the standpoint of the promonimal system of Old

xx

john b. whitman bibliography

Japanese]’ With Yuko Yanagida. 2009. In Tsubomoto, Atsurō et al (eds), Uchi to soto no gengogaku [The Linguistics of “Inside” and “Outside”], pp. 175–216. Tokyo: Kaitakusha. ‘The classification of constituent order generalizations and diachronic explanation’. 2008. In Good, Jeff (ed.), Language Universals and Language Change, pp. 233–252. Oxford: Oxford University Press. ‘The complement structure of shi … de clefts’. With Waltraud Paul. 2008. In Bartos, Huba (ed.), European Studies in Chinese Linguistics. Amsterdam: John Benjamins. ‘The source of the bigrade conjugation and stem shape in pre-Old Japanese’. 2008. In Frellesvig, Bjarke and John Whitman (eds.), Proto-Japanese, pp. 159–174. Amsterdam; John Benjamins. ‘Evidence for seven vowels in proto-Japanese’. With Bjarke Frellesvig. 2008. In Frellesvig, Bjarke and John Whitman (eds.), Proto-Japanese, pp. 15–42. Amsterdam: John Benjamins. ‘Reanalysis and conservancy of structure in Chinese’. With Waltraud Paul. 2005. In Batllori, Montserrat et al (eds.), Grammaticalization and Parametric Variation, pp. 82–94. Oxford: Oxford University Press. ‘Preverbal negation in Japanese and Korean’. 2005. In Cinque, Guglielmo and Richard Kayne (eds.), Handbook of Syntactic Variation, pp. 880–902. Oxford, Oxford University Press. ‘Licensing multiple negative polarity items’. 2004. With Susumu Kuno. In Kim-Renaud, Young-Key & John Whitman (eds.), Studies in Korean Syntax and Semantics by Susumu Kuno, pp. 207–228. Seoul: Pajigong Press. ‘The vowels of proto-Japanese’. 2004. With Bjarke Frellesvig. Japanese Language and Literature 38.2, pp. 281–299. ‘The proto-Korean shape of ha- “do”’. 2003. In Lee, Sang-Oak (ed.) Pathways into Korean Language and Culture: Essays in Honor of Young-Key Kim-Renaud, pp. 463–470. Seoul: Pajigong Press. ‘Kayne 1994: p. 143, fn. 3. 2001’. In Alexandrova, Galina and Olga Arnaudova (eds.), The Minimalist Parameter, pp. 77–100. Amsterdam: John Benjamins. ‘Give as a Higher Verb: Reassessing Nakau (1973)’. 2001. In Minoru Nakau Festschrift Editorial Committee (eds.), Imi to katachi no intaafeesu [The interface of meaning and form], pp. 785–796. Tokyo: Kuroshio Press. ‘Relabeling’. 2001. In Pintzuk, Susan, George Tsoulas, & Anthony Warner, Diachronic Syntax: Models and Mechanisms, pp. 220–240. Oxford: Oxford University Press. ‘Right dislocation in English and Japanese’. 2000. In Takami, Ken-ichi, Akio Kamio, and John Whitman (eds.), Syntactic and Functional Explorations: A Festschrift for Susumu Kuno, pp. 445–470. Tokyo, Kuroshio Press. ‘A type of head in situ construction in English’. (With Atsuro Tsubomoto). 2000. Linguistic Inquiry 30.4, pp. 176–183.

john b. whitman bibliography

xxi

‘Tongsa Kulcŏl-ŏ ŭi suptŭk kwa paltal ŭi pulyŏnsoksong [Acquisition of verbal inflection and discontinuities in development]’. 2000. In Cho, S.-K. (ed.), Ingan un ono lul otohke suptuk hanun ga [How do human beings acquire language?], pp. 173–192 Seoul, Akhanet Publishers. Translated by S.-K. Cho. ‘Personal pronoun shift in Japanese: A case study in lexical change and point of view’. 1999. In Kamio, Akio and Ken-ichi Takami (eds.), Function and Structure: In Honor of Susumu Kuno, pp. 357–386. Amsterdam, John Benjamins. ‘The category of relative clauses in Japanese and Korean’. With Tamar Kaplan. 1995. Journal of East Asian Linguistics 4.1, pp. 29–58. ‘Syntactic representation and interpretive preference’. With Bradley Pritchett. 1995. In Mazuka, Reiko and Noriko Nagai (eds.), Japanese Sentence Processing, 65–76. Hillsdale, n.j.: Lawrence Erlbaum Associates. ‘The accentuation of nominal stems in proto-Korean’. 1994. In Kim-Renaud, Young-Key (ed.), Theoretical Issues in Korean Linguistics, pp. 425–440. Stanford: Center for the Study of Language and Information. ‘In defense of the Strong Continuity account of the acquisition of Verb-Second’. 1994. In Lust, Barbara, Magui Suñer, and John Whitman (eds.), Syntactic Theory and Second Language Acquisition, Volume 1: Heads, Projections, and Learnability, pp. 273–286. Hillsdale, n.j.: Lawrence Erlbaum Associates. ‘Scope and optionality: Comments on the papers on wh-movement and quantification’. 1994. In Hermon, Gabriella, Barbara Lust, and Jaklin Kornfilt (eds.) Syntactic Theory and Second Language Acquisition, Volume 2: Binding, Dependencies, and Learnability, pp. 417–426. Hillsdale, n.j., Lawrence Erlbaum Associates. ‘Verb raising and vp ellipsis’. With Kazuyo Otani. 1991. Linguistic Inquiry 22.2, pp. 345– 358. ‘A rule of medial *-r- loss in pre-Old Japanese’. 1990. In Baldi, Philip (ed.) Linguistic Change and Reconstruction Methodology, pp. 511–546. Berlin: Mouton de Gruyer. ‘Argument positions and configurationality’. 1989. In Georgopolous, Carol and Roberta Ishihara (eds.), Interdisciplinary Approaches to Language: Essays in Honor of S.Y. Kuroda, pp. 615–228. Dordrecht: Kluwer. ‘Varieties of discourse ellipsis’. 1988. In Linguistic Society of Korea (ed.), Linguistics in the Morning Calm ii: Papers from the Second Seoul International Conference on Linguistics, pp. 149–176. Seoul: Hanshin. ‘Configurationality parameters’. 1987. In Saito, Mamoru & Takashi Imai (eds.), Issues in Japanese Linguistics, pp. 351–374. Dordrecht: Foris.

xxii

john b. whitman bibliography

Papers in Refereed Conference Proceedings ‘Deconstructing clausal noun modifying constructions’. With Anna Bugaeva. 2015. In Kenstowicz, Michael, Theodore Levin, and Ryo Masuda (eds.), Japanese/Korean Linguistics 23. Stanford: Center for the Study of Language and Information. (Poster paper at: http://web.stanford.edu/group/cslipublications/cslipublications/ja-ko -contents/JK23/) ‘A Korean grammatical borrowing in Early Middle Japanese kunten texts and its relation to the syntactic alignment of earlier Korean and Japanese’. With Yuko Yanagida. 2014. In Nam, Seungho, Heejeong Ko, and Jongho Jun (eds.), Japanese/Korean Linguistics 21, pp. 121–135. Stanford: Center for the Study of Language and Information. ‘The diachronic consequences of the rtr analysis of Tungusic vowel harmony’. With Andrew Joseph. 2013. In Özge, Umut (ed.), Proceedings of the 8th Workshop on Altaic Formal Linguistics, pp. 159–176. Cambridge, Massachusetts: mit Working Papers in Linguistics. ‘The prehead relative clause problem’. 2013. In Özge, Umut (ed.), Proceedings of the 8th Workshop on Altaic Formal Linguistics, pp. 361–380. Cambridge, Massachusetts: mit Working Papers in Linguistics. ‘The genesis of indefinite pronouns in Japanese and Korean’. With Tomohide Kinuhata. 2011. In McClure, William and Marcel den Dikken (eds.), Japanese/Korean Linguistics 18, pp. 88–100. Stanford: Center for the Study of Language and Information. ‘Prenominal complementizers and the derivation of complex nps in Japanese and Korean’. With Bjarke Frellesvig. 2011. In McClure, William and Marcel den Dikken (eds.), Japanese/Korean Linguistics 18, pp. 73–87. Stanford: Center for the Study of Language and Information. ‘Subject-object asymmetries in Korean sentence comprehension’. (With Jiwon Yun and John Hale). 2010. Proceedings of the Annual Meeting of the Cognitive Science Society 2010. ‘Multiple focus clefts in Japanese and Korean’. With Sungdae Cho and Yuko Yanagida. Proceedings of the 44th Meeting of the Chicago Linguistic Society 1, pp. 61–77. University of Chicago: Chicago Linguistic Society. ‘The Japanese-Korean vowel correspondences’. (With Bjarke Frellesvig). 2008. In Endo Simon, Mutsuko and Peter Sells (eds.), Japanese/Korean Linguistics 13. pp. 15–28. Stanford, Center for the Study of Language and Information. ‘Ditransitive and applicative structure in Greek’. With Effi Georgala. 2008. In Proceedings of the 43rd Meeting of the Chicago Linguistic Society. pp. 77–91. Chicago: Chicago Linguistic Society. ‘Thematic and expletive applicatives’. With Effi Georgala and Waltraud Paul. 2007. In Chang, Charles B. and Hannah J. Haynie (eds.), Proceedings of the 26th West Coast

john b. whitman bibliography

xxiii

Conference on Formal Linguistics. pp. 181–189. Somerville, Massachusetts: Cascadilla Proceedings Project. ‘Direct movement passives in Korean and Japanese’. With Sang Doh Park. 2003. In McClure, William (ed.), Japanese/Korean Linguistics 11, pp. 307–321. Stanford: Center for the Study of Language and Information. ‘Adjunct major subjects in Korean’. 2000. In Lee, Chungmin and John Whitman (eds.), Korean Syntax and Semantics: Proceedings of the lsa Institute Workshop, pp. 1–8. Seoul: Thaehaksa. ‘Kakarimusubi from a comparative perspective’. 1997. In Sohn, Ho-min and John Haig (eds.), Japanese and Korean Linguistics 6, pp. 161–178. Stanford: Center for the Study of Language and Information. ‘Syntactic movement of overt wh-phrases in Japanese and Korean’. With Kunio Nishiyama and Eun-Young Yi. 1996. In Akatsuka, Noriko, Shoichi Iwasaki, and Susan Strauss (eds.), Japanese and Korean Linguistics 5, pp. 337–352. Stanford: Center for the Study of Language and Information. ‘The syntax of cp in early child German: The Strong Continuity Hypothesis’. With Katharina Boser, Barbara Lust, and Lynn Santelmann. 1992. Proceedings of the Northeast Linguistic Society 22, pp. 51–66. University of Massachusetts, Amherst: glsa. ‘Continuity of the principles of universal grammar in first language acquisition: The issue of functional categories’. With K.-O. Lee and Barbara Lust. 1990. Proceedings of the Northeast Linguistic Society 21, pp. 383–397. University of Massachusetts, Amherst: glsa. ‘On the acquisition of functional categories in Korean: A study of the first language acquisition of Korean relative clauses’. With K.-O. Lee and Barbara Lust. 1990. In Baek, E.-J. (ed.), Papers from the Seventh International Conference on Korean Linguistics. pp. 313–333. Toronto: University of Toronto Press. ‘Topic, modality, and ip structure’. 1988. In Kuno, Susumu et al (eds.) Proceedings of the Third Harvard Workshop on Korean Linguistics. pp. 341–356. Seoul: Hansin. ‘Korean morphological passives and causatives’. (With Sharon Hahn). 1988. In Baek, E.-J. (ed.) Papers from the Sixth International Conference on Korean Linguistics. pp. 714– 728. Toronto, University of Toronto Press. ‘Korean clusters’. 1986. In Kuno, Susumu et al. (eds.) Proceedings of the First Harvard Workshop on Korean Linguistics. pp. 280–290. Seoul: Hanshin. ‘The internal structure of np in verb-final languages’. 1981. Papers from the Seventeenth Regional Meeting of the Chicago Linguistics Society. pp. 411–418. Chicago: Chicago Linguistic Society. ‘Scrambled, over easy, or sunnyside up’. 1979. Papers from the Fifteenth Regional Meeting of the Chicago Linguistics Society. pp. 342–352. Chicago: Chicago Linguistic Society.

xxiv

john b. whitman bibliography

Other Publications ‘wals (World Atlas of Language Structures) no gengo ruikeironteki parameta no tokeironteki bunseki to sono tsujironteki kaisyaku [A statistical analysis of the typological features in wals and its diachronic interpretation]’. With Yohei Ono. 2014. Bulletin of the Edward Sapir Society of Japan 28, 1–15. Tokyo: Edward Sapir Society of Japan. Review of Lee, Ki-moon and Robert Ramsey A History of Korean. In Korean Historical Linguistics. Special issue editor, with Young-Key Kim-Renaud. 2013. Korean Linguistics 15.2. ‘Kundoku wa kanji bunkaken dake no mono ka. [Is kundoku found only in the Sinosphere?]’. Nihongogaku 2013.11 (vol. 32.13), pp. 26–41. ‘Tagengo o jigengo de yomu koto: “kundoku” no fuhensei ni tuite. [Reading a foreign language in the vernacular: the universality of “kundoku”]’. 2013. In Takada, Tomokazu and Teiji Koskegwa (eds.), Kunten shiryō no kozoka kijutu seika hōkokusho. National Institute of Japanese Linguistics Kyōdō Kenkyū Hōkoku 12.0. Nihongo shinhakken: Sekai kara mita nihongo. [Rediscovery of Japanese: Japanese from the world’s perspective]. 2013. National Institute of Japanese Linguistics ninjal Forum Series 3 (editor). ‘Gengo shigen tosite no nihongo [Japanese as a linguistic resource]’. With Satoshi Kinsui, Yasuyuki Shimizu, and Tsutomo Yada. 2011. Bungaku 12.3, pp. 2–51. Review of J. Marshall Unger (2009) The Role of Contact in the Origins of the Japanese and Korean Languages. The Journal of Asian Studies 69.4 (November 2010), pp. 1268– 1270. Review of Mineharu Nakayama (ed.) Issues in East Asian Language Acquisition. 2006. Language 82.3, pp. 662–665. Review of Mark James Hudson (1999) Ruins of Identity: Ethnogenesis in the Japanese Islands. Harvard Journal of Asiatic Studies.62.1 (June 2002). ‘Wh gimonbun to yes/no gimonbun no bimyō na sa (A subtle difference between whand yes/no questions)’. 1996. Kōkō Eigo Tenbō. Tokyo: Shogakukan. ‘Hatarakikake no “want” [Hortative “want”]’. 1996. Kōkō Eigo Tenbō. Tokyo: Shogakukan. ‘Kanbun kundoku ni yorite tsutaeraretaru (ei) gohō [Grammatical usage (of English) based on the Japanese tradition of reading Chinese]’. 1995. Kōkō Eigo Tenbō. Tokyo: Shogakukan. Review of William Poser (ed.) Japanese Syntax. Language. 1992. ‘Nonpronominal aspects of zero pronoun phenomena’. 1985. In San Diego Working Papers in Linguistics.

john b. whitman bibliography

xxv

Refereed Conference Presentations The origins of vernacular glossing in Northeast Asia. Tapping Immaterial Resources: Glossing Practices between the Far East and the Latin West, c. 600 ce. GoetheUniversität Frankfurt am Main, December 2016. Contested vernacular readings: The Satō-bon Kegon mongi yōketsu and the Tōdaiji fuju monkō. 36th Annual Conference of The Center for Medieval Studies, Fordham University. March 2016. “Kundoku” “Kunten” wa kanjibunkaken dake no mono ka: Chūsei Ōshū Raten go no chūshaku siryō to jigengo de no “yomi” (Were “kundoku” and “kunten” practices limited to the Sinoshpere? Glossed Latin materials in medieval Latin and vernacular “reading”). 2014 Fall Meeting of the Japanese Linguistics Society, Hokkaidō University, October 2014. Deconstructing clausal noun modifying constructions (with Anna Bugaeva). Japanese/ Korean Linguistics 23. Massachusetts Institute of Technology, Cambridge, ma, October 2013. The Constituent Order of proto-Sino-Tibetan (with Redouane Djamouri and Waltraud Paul). The 45th International Conference on Sino-Tibetan Languages and Linguistics (icstll45) Nanyang Technological University, Singapore. October 2012. The diachronic consequences of the rtr analysis of Tungusic vowel harmony (with Andrew Joseph). The 8th Workshop on Altaic Formal Linguistics. University of Stuttgart, May, 2012. A Korean grammatical borrowing in Early Middle Japanese kunten texts and its relation to the syntactic alignment of earlier Korean and Japanese (with Yuko Yanagida). Japanese/Korean Linguistics 21. Seoul National University, October, 2011. Processing of Noun Complement Complex nps with subject and object pro in Korean (with Nayoung Kwon, Jiwon Yun, and John Hale). cuny 2011 (the 24th Annual Conference on Human Sentence Processing). Stanford University, March 24–26, 2011. Subject-Object Asymmetries in Korean Sentence Comprehension (with Jiwon Yun and John Hale). CogSci 2010. Portland, August 11–14, 2010. The Formal Syntax of Degramaticalization (with Yuko Yanagida). DiGS xii (the 12th Conference on Diachronic Generative Syntax). Cambridge University, July 2010. The Formal Syntax of Alignment Change (with Yuko Yanagida). DiGS xi (the 11th Conference on Diachronic Generative Syntax). University of Campinas, June 2009. Kunten shiryō to Kankoku no koketsu shiryō no setten—Satō-bon Kegon mongi yōkestu o chūsin ni (Points of contact between kunten materials and Korean kugyŏl materials, with a focus on the Satō-bon Kegon mongi yōkestu. Paper read at the 100th Meeting of the Kuntengo gakkai), Kyōto, May 2009. The Genesis of Indefinite Pronouns in Japanese and Korean (with Tomohide Kinuhata). Japanese/Korean Linguistics 18, City University of New York, November 2008.

xxvi

john b. whitman bibliography

Prenominal Complementizers and the Derivation of Complex nps in Japanese and Korean (with Bjarke Frellesvig). Japanese/Korean Linguistics 18, City University of New York, November 2008. Multiple Focus Clefts in Japanese and Korean (with Sungdae Cho and Yuko Yanagida). 44rd Meeting of the Chicago Linguistic Society, May 2008. Ditransitive and Applicative Structure in Greek (with Effi Georgala). 43rd Meeting of the Chicago Linguistic Society, May 2007. Thematic and Expletive Applicatives (with Effi Georgala and Waltraud Paul). 25th West Coast Conference on Formal Linguistics, University of California, Berkeley April 2007. Double Object and Dative Constructions in Mandarin Chinese (with Waltraud Paul). European Conference on Chinese Linguistics, Budapest, January 2006. Reanalysis and Conservancy of Structure (in Chinese) (with Waltraud Paul). digs 7, Girona, Spain, June 2002. Lexical Bases for Syntactic Change (with Dianne Jonas). lsa winter meeting, January 2002 Limiting Multiple Specifiers to the Interface (with Dianne Jonas). nels 22, October 2001. Saikin no kaigai ni okeru Nihongo, Chōsengo no hikaku kenkyû no ichi sokumen. (An aspect of recent research overseas on the comparative study of Japanese and Korean). Kokugo Gakkai (Meeting of the Society for the Study of the Japanese Language), Ryûkoku University, Kyoto, Japan, May 1995. Kakarimusubi from a Comparative Perspective. Sixth Conference on Japanese and Korean Linguistics, University of Hawaii, August 1995. Syntactic Movement of Overt Wh-Phrases in Japanese and Korean (with Kunio Nishiyama and Eun-Young Yi). Fifth Conference on Japanese and Korean Linguistics, ucla, September 1994. Q-in-situ. Linguistic Society of America Annual Meeting, January 1994. The Accentuation of Nominal Stems in Proto-Korean. Eighth International Conference on Korean Linguistics, The George Washington University, August 1992. The Category of Relative Clauses in Japanese and Korean (with T. Kaplan) presented at the Southern California Conference on Japanese and Korean Linguistics, San Diego State University, August 1992. The Acquisition of infl to comp in Early Child German: The Strong Continuity Hypothesis (with K. Boser, B. Lust, and L. Santelmann) presented at nels 22, 1991. The Theoretical Significance of Auxiliary Insertion in Early Child German (with K. Boser, B. Lust, and L. Santelmann). Boston University Conference on Language Development, 1991. Nested and Layered Agreement. Twenty-Sixth Regional Meeting of the Chicago Linguistics Society, 1990.

john b. whitman bibliography

xxvii

String Vacuous infl to comp. glow, University of Leyden, 1991. String Vacuous infl to comp. Linguistic Society of America Annual Meeting, December 1989. The Relation between the Middle Korean Voiced Spirants and Aspiration Seoul International Conference on Linguistics (sicol), 1986.

Invited Talks and Conference Presentations (Selected, from 2016 to 1986) Higher functional architecture in nominal projections and identifying d. Dimensions of d Workshop, University of Rochester, September 2016. Nihongo to sono ruikeironteki shinseki: Gengo ruikeironteki shinseki no chūshutsu ni tsuite (Japanese and its typological relatives: on the identification of typological relatives). 50th Anniversary Seminar of the Tokyo Institute for Linguistic Resarch, September 2016. Kanbun kundoku to Vernacular Reading: Hikaku gurōbaru teki kanten kara (Kanbun kundoku of literary Chinese and vernacular reading: From a comparative-global perspective) Nihon Daigaku Jinbun Kagaku Kenkyūsho, July 2016. Vernacular reading as a cultural linchpin in East Asia and beyond. East Asia Border Crossing Symposium, George Washington Universtiy, April 2016. Nihongo (shogengo) no keitōzu no kanōsei ni tuite: jūrai no kenkyūto shōrai no tenbō (On the possibility of a family tree for Japanese (Japonic): previous research and future prospects). Plenary talk given at jlvc 3, National Institute of Japanese Language and Linguistics, February 2016. Glossing and other traces of vernacular reading. Program in World Philology Lecture Series, Columbia University, November 2015. Kakarimusubi from a comparative perspective: A crosslinguistic overview. National Institute for Japanese Language and Linguistics, September 2015. Tōhoku Ajiano butten kundoku to Raten go kyōten no dokuhō: Chūsei Ōshū no chūshaku bunken to kunte/kugyŏl nohikaku kenkyū (A comparative study o vernacular reading of Buddhist texts in Northeast Asia and medieval Latin glossed texts). International symposium Gengo/moji no tenkai kara mita Bukkyō ryūden. Waseda University, December 2012. Typological parameters in wals (World Atlas of Linguistic Structures) and their diachronic interpretation (with Yohei Ono). 28th Annual Meeting of the Edward Sapir Society of Japan. Seishin Women’s University, October 2013. Diachronic interpretations of word order parameter cohesion. DiGS xv. University of Ottawa, August 2013. Accessing the cosmopolitan code in the Sinographic Cosmopolis. International sym-

xxviii

john b. whitman bibliography

posium Learning kanbun in Traditional Korea and Japan. Waseda University, June 2013. Topic prominence 40 years later. Workshop on word order and information structure in East Asian Languages. Harvard-Yenching Institute, Harvard University, April 2013. Comparative consequences of the tongue root harmony analysis for proto-Tungusic, proto-Mongolic, and proto-Korean (with Seongyeon Ko and Andrew Joseph). International conference on “Paradigm change in the Transeurasian languages and beyond”. University of Mainz, March 2013. Nichiryū sogono onsetsumatu shiin to rentaikei/izenkei no kigen (Coda consonants in proto-Japanese-Ryukyuan and the origin of the attributive and realis conditional ending). Workshop Ryūkyū shogo to kodai Nihongo. Kyoto University, February 2013. The historical source of the bigrade (nidan) transitivity alternations in Japanese. International symposium on valency, National Institute of Japanese Language and Linguistics. July 2012. The relative clause problem. The 8th Workshop on Altaic Formal Linguistics, University of Stuttgart. May 2012. Japanese relative clauses. Workshop on Japanese syntax, Yokohama National University, June 2012. Nihon no kunten shiryō to Kankoku no koketsu shiryō—Chūsei ōshū no chūshaku bunken no hikaku ni oite (Japanese kunten materials and Korean kugyŏl materials from the perspective of a comparison with medieval European glossed materials). Paper read at the Nikkan Gengogakusha Kaigi (Meeting of Linguists of Japan and Korea), Reitaku University, October 2010. The ubiquity of the gloss. Paper read at the Second Hunmin Jeongeum Symposium, Seoul University, October 2010. The proto-Japanese vowel system and the origins of the verbal conjugation classes. Paper read at the 4th Atami On’inron Festa, February 2009. Hangugŏ mit Ilbonŏ eui tasu ch’ojŏm punyŏlmun. Paper read at the International Symposium on Korean Linguistics, University of Tokyo, October 2008. The origins of the Japanese verb conjugations. Paper read at the Conference on East Asian Linguistics, University of Toronto. October 2006. The evolution of English complement for np to. Paper read at the Rosendal Symposium on Linguistic Theory and Language Change, University of Oslo, June 2005. Focus structures in Mandarin Chinese: shi … de vs. bare shi (with Waltraud Paul) Paper read at the Workshop on Syntactic categories and their interpretation in Chinese. The Research Institute for Linguistics (Hungarian Academy of Sciences) Budapest, October 2004. The form and function of the Rentaikei (連体形) and Izenkei (已然形) Suffixes in ProtoJapanese. Paper read at the Oxford Kobe Symposium on the History and Structure of Japanese, September 2004.

john b. whitman bibliography

xxix

The attrition of genitive subjects in Korean and Japanese. Paper read at digs ix, Yale University, June 2004. There constructions. Paper read at the Susumufest, Department of Linguistics, cuny, May 2004. Lexical change and syntactic change. Paper read at the workshop on historical change and universals, uc Berkeley, March 2003. Preverbal negation in Japanese and Korean. Paper read at the English Linguistics Society of Japan, November 2003. A seven-vowel system for proto-Japanese (with Bjarke Frellesvig). Paper read at the 16th International Conference on Historical Linguistics, August 2002. The sources of the conjugation classes in Japanese. Paper read at the 16th International Conference on Historical Linguistics, August 2002. The Japanese-Korean vowel correspondences (with Bjarke Frellesvig). Paper read at the 12th Conference on Japanese/Korean Linguistics, July 2002. Movement and control in Passives. Paper read at the 11th Conference on Japanese/ Korean Linguistics, November 2002. The sources of the main conjugation classes in Japanese. Paper read at the first eu seminar on Japanese Linguistics, Oxford University, September 2002. Verb root shape in earlier Japanese and Korean. Paper read at the Symposium on the Historical Relation of Japanese and Korean International Conference on Historical Linguistics, University of British Columbia, August, 1999. Hikakuteki kanten kara mita Nihongo no katuyô no hattatsu (The development of the verbal conjugations in Japanese as seen from a comparative perspective). Paper read at the Symposium on Japanese Language Origins, International Center for the Study of Japanese Culture, Kyoto, Japan, July 2001. A brief look at cognate functional morphology in Korean and Japanese. Paper read at the Workshop on Altaic Languages, Formal Approaches to Japanese Linguistics Conference, mit, May 2001. Diachronic reanalysis and extended vp projections in Chinese (with Waltraud Paul). Paper read at the Workshop on Historical Syntax, University of London, January 2001. Réanalyse diachronique dans le programme Minimaliste. Paper read at the Symposium on Chinese Historical Linguistics, Centre pour Recherche Linguistique sur l’ Asie Orientale, Paris, December 2000. Genitive subjects in Middle Korean. Paper read at the International Conference on Korean Linguistics, University of Hawaii, August 1999. Postnominal particles as phrasal heads. Paper read at the Harvard Conference on Korean Linguistics, July 1998. Transformational grammar is realistic: the dtc and perceptual complexity (with B. Pritchett) Paper read at the cuny Conference on Natural Languge Processing, April 1992.

xxx

john b. whitman bibliography

Right dislocation in English and Japanese. Paper read at the International Symposium on Functional Linguistics. Dokkyô University, Tokyo, Japan, December 1991. Kakarimusubi. Paper read at the University of California at San Diego Workshop on Japanese Historical Syntax, La Jolla, California, November 1991. The response from Japanese. Presented at the Consortium for Language Teaching and Learning Symposium on Advanced and Intermediate Japanese Language Instruction, Princeton University, November 1991. Scrambling and the Derivational Theory of Complexity (with B. Pritchett) presented at the Duke University Workshop on Japanese Language Processing, October 1991. First language acquisition of closed class categories. Presented at the u.s.-Korea Bilateral Conference on Cognitive Science, Seoul National University, Korea, August 1991. Quantifier float and vp adjunction. Presented at Festival for Yuki Kuroda, ucsd, April 1988. A rule of medial *-r- Loss in Pre-Old Japanese. Paper read at the Workshop on Reconstruction Methodology, lsa Summer Institute, August 1987. The identification of null nominal expressions. Presented at the Second sdf Japanese Syntax Workshop, Stanford University, May 1986. Intransitivization. Presented at the Third sdf Workshop on Japanese Syntax, Kobe, Japan, April 1987. Configurationality and the head of s. Presented at the University of Groningen Conference on Configurationality and aux, March 1986.

Reviewer for Journal of East Asian Linguistics, Journal of Korean Linguistics, Language, Linguistic Inquiry, Natural Language and Linguistic Theory, Lingua, Gengo Kenkyū, Language and Linguistics, English Linguistics, Chicago Linguistic Society, Diachronic Generative Linguistics, Japanese/Korean Linguistics, West Coast Conference on Formal Linguistics, nels, Formal Approaches to Japanese Linguistics, Workshop on Formal Altaic Linguistics, Brill, Hawaii University Press.

Tabula Gratulatoria Dorit Abusch Edith Aldridge John R. Bentley Zhong Chen Adam I. Cooper Clifford Crawford Tejaswini Deoskar Hongyuan Dong Thomas Dougherty José Andrés Alonso de la Fuente Naoki Fukui Andrew Garrett Effi Georgala Masayuki Gibson Wesley Jacobsen Guillaume Jacques

Juha Janhunen Tamotsu Kawasaki Ritsuko Kikusawa Juwon Kim Satoshi Kinsui John Hale Teruhiro Hayata Jonathan Howell Mark Hudson Hyun Kyung Hwang Andrew Joseph Tarō Kageyama Edward Kamens Young-gey Kim-Renaud Daniel Kaufman Shinjirō Kazama Alan Hyun-Oak Kim

Cornell University University of Washington Northern Illinois University Rochester Institute of Technology Northeastern University Independent scholar University of Amsterdam George Washington University University of Hawai’i, Manoa Jagiellonian University Sophia University University of California, Berkeley Nuance Communications Synfonica llc Harvard University Centre National de la Recherche Scientifique/ Centre de Recherches Linguistiques sur l’ Asie Orientale (cnrs/crlao) University of Helsinki Nagano Prefectural Museum National Institute for Ethnography Seoul National University Osaka University Cornell University Daitō Bunka University Montclair State University Northwestern Kyushu University National Institute for Japanese Language and Linguistics Cornell University National Institute for Japanese Language and Linguistics Yale University George Washington University Queens College and the Graduate Center, City University of New York (cuny) Tokyo University of Foreign Studies Southern Illinois University-Carbondale

xxxii Ross King Dongho Ko Seongyeon Ko

tabula gratulatoria

University of British Columbia Cheonbuk National University Queens College and the Graduate Center, City University of New York (cuny) John Kupchik University of Auckland Richard Larson Stony Brook University Hye-Sook Lee Rose Financial Services David Barnett Lurie Columbia University Thomasz Majtczak Jagiellonian University Gita Martohardjono Queens College and the Graduate Center, City University of New York (cuny) Yumiko Nishi University of Iowa William O’Grady University of Hawai’i at Mānoa Takuichirō Ōnishi National Institute for Japanese Language and Linguistics Toshiki Osada International Center for Japanese Studies Alain Peyraube Centre National de la Recherche Scientifique/ Centre de Recherches Linguistiques sur l’ Asie Orientale (cnrs/crlao) John Phan Rutgers University Pittayawat Pittayaporn Chulalongkorn University Laurent Sagart Centre National de la Recherche Scientifique/ Centre de Recherches Linguistiques sur l’ Asie Orientale (cnrs/crlao) James Scanlon-Canegata Yale University Leon Serafim formerly of University of Hawai’i, Manoa Peter Sells University of York Moriyo Shimabukuro University of Ryukyus David Silva Salem State University Jun’ichi Suda Senshū University Irène Tamba Ecole des hautes études en sciences sociales/ Centre de Recherches Linguistiques sur l’ Asie Orientale (ehess/crlao, Emerita) Toshirō Tsumagari Hokkaidō University Zhiguo Xie Ohio State University Jiwon Yun Stony Brook University

part 1 Documentation

∵

chapter 1

The Digital Museum Project for the Documentation of Endangered Languages: The Case of Ikema Ryukyuan* Yukinori Takubo

1

Goals of the Project

The purpose of this paper is to introduce our project to design a web-based museum for endangered languages that is easily updatable, and can serve as a basis for collaborative research on endangered languages. Our museum also aims to provide a forum where local people can exhibit their language products, thereby helping them have better access to their language and preserve the language and culture of the local community. We have been constructing a prototype of the museum using data from our field research conducted in a village called Nishihara, where a subdialect of Ikema, a dialect of Miyako Ryukyuan, is spoken.1

* Earlier versions of this paper have been presented at the 1st International Conference on Language Documentation and Conservation (the University of Hawai’i, Manoa, March 13), the Hans Rausing Endangered Languages Project (soas, University of London. London, uk, June 8, 2010), World Oral Literature Occasional Lecture Series (University of Cambridge, Cambridge, uk, October 4, 2010), Innovation: East Asian perspective (ucla, usa, January 25). I would like to thank the audiences, especially Mark Turin for detailed and insightful comments. 1 The construction of the digital museum is a cooperative project with the Contents Production Team at Kyoto University headed by Ms. Tamaki Motoki and members of the Ikema Project: Yuka Hayashi, Chigusa Kurumada, Tsuyoshi Ono, and Shoichi Iwasaki. In constructing the digital museum, we owe special thanks to my consultants: Hiroyuki Nakama, Chieko Hanashiro, Tadashi Nakama and other people of Nishihara Village. Thanks are due to the past and present members of the Contents Production Team at Kyoto University: Mikiko Takahashi, Naomi Nagai, Hiroto Ueda, Masashi Iwakura, Makoto Miyabe, and Yasuto Masumoto. The url of the digital museum is kikigengo.jp and the accompanying Facebook site is https:// www.facebook.com/kikigengojp/. See Takubo (2011) for details of our digital museum project. See Hayashi (2010) for general information about the Ikema dialect.

© koninklijke brill nv, leiden, 2018 | doi: 10.1163/9789004351134_002

4

takubo

This paper is organized as follows. In section 2, I will give an overview of Nishihara village, the Ikema dialect spoken in Nishihara, and the people and culture of the village. In section 3, I will introduce the design features of our digital museum. In section 4, I will explain how the digital museum can be extended to other endangered languages, e.g. the Jeju dialect of Korean.

2

Overview of the Language and the Community

Ryukyuan languages are the only languages that have been proven to be genetically related to Japanese. They diverged from the same ancestor, proto-Japonic, about 1500 years ago or earlier. There are five major languages within Ryukyuan: Amami, Okinawa, Miyako, Yaeyama, and Yonaguni.2 They are mutually unintelligible and unesco below regards them as different ‘languages’ rather than ‘dialects’ of Ryukyuan. Ikema is a subdialect of Miyako Ryukyuan, spoken on Ikema island, in Sarahama on Irabu island, and in Nishihara on Miyako island. Miyako Ryukyuan is an endangered language and most of the fluent speakers of Ikema are in their seventies or older. Ikema spoken in Nishihara seems to be an exception to this situation. Some people in their forties are fluent in the dialect and even some people in their thirties can understand the language.3 The preservation of the Ikema in Nishihara is probably related to the fact that speakers there still maintain various religious rituals in which they are required to speak the language.4 The speakers of the Nishihara subdialect of Ikema, however, are all bilinguals in Standard Japanese and Ikema. We have met no monolingual speakers of the language. Younger generations, that is, people in their forties or thirties, have stopped speaking Ikema to their children, and the language will most likely disappear in twenty to thirty years. By the degrees of endangerment established by unesco, Ikema is categorized between severely endangered and definitely endangered.

2 See Shimoji and Pellard (2010) for information about the Ryukyuan languages. Pellard (2012) claims that Yonaguni belongs to Macro-Yaeyaman which includes Yaeyama and Yonaguni as the two sub-families. Genealogical status of Yonaguni is varied among linguists but it is common understanding among Ryukyuan specialists that Yonaguni constitutes an independent language mutually unintelligible with other Yaeyaman dialects. 3 See Iwasaki and Ono (2013) for the language situation in Nishihara and Ikema. 4 See Hirai (2009) for the nature of religious rituals in Nishihara.

the case of ikema ryukyuan

5

(1) Degrees of endangerment: Intergenerational language transmission. – Safe: language is spoken by all generations; intergenerational transmission is uninterrupted – Vulnerable: most children speak the language, but it may be restricted to certain domains (e.g., the home) – Definitely endangered: children no longer learn the language as a mother tongue in the home – Severely endangered: language is spoken by grandparents and older generations; while the parent generation may understand it, they do not speak it to their children or among themselves – Critically endangered: the youngest speakers are grandparents and older, and they speak the language partially and infrequently – Extinct: there are no speakers left unesco Intangible Cultural Heritage—Endangered languages, http://www .unesco.org/culture/ich/index.php?pg=00139

In order to help preserve endangered languages, we have been constructing a digital museum for endangered languages. We have chosen Ikema spoken in Nishihara to be housed in the proto-type of the digital museum.

3

Design Features of the Digital Museum

Our digital museum has four areas: two open spaces for the exhibit and the library and two closed spaces for a closed library and a data storage space. a:

b:

Closed access space a1: Closed library and archives a2: Data storage space Open access space b1: The exhibit space: permanent and special exhibition room b2: Library and references

3.1 Closed Access Space The museum has a closed library and closed archives accessed only by members. This space houses past exhibits, i.e. files with transcription and other updated archival information. It stores files that cannot be exhibited because of privacy issues and other considerations. Access to this space is restricted by an id and a password. We can store (unpublished) papers here. The closed access space also stores raw data in the data storage space. This space is only accessible to a limited number of researchers. Files of raw data are

6

takubo

figure 1.1 Example 1: Indexing by location

stored here with basic metadata, e.g., who the participants were, who did the recording, the location and the time of the recording, and what the recording devices were, etc. 3.2 Open Access Space The open access space contains an exhibit space and a library for movie files. The exhibit space may be divided into permanent exhibit spaces and special exhibit spaces. The current version of the museum constructed as a proto-type only contains permanent exhibit spaces. The open access space contains indexes to pictures and video files stored in the picture and video library. The indexing can be done in various ways. Here is an example of a link by location. The red dots on the map of Nishihara village show the locations where various daily activities of the villagers such as fishing or shrimp catching take place. On the picture a video clip featuring saitui (shrimp catching) is linked to the name of a bay, that is, to the location of the activity. The visitor can click the red dot to open the file. The indexing can be done by time line. Figure 1.2 is an example of a link to a time line. A video clip named tamunu (firewood gathering) is linked to a small red triangle in the life history of the narrator. You can click on the triangle to open the video clip. Figure 1.3 is another example of a link to a time line. A video clip of a festival called MyaakuzIcI is linked to a date on a calendar.

the case of ikema ryukyuan

7

figure 1.2 Example 2: Indexing by time line (Life history)

figure 1.3 Example 3: Indexing by time line (A year in Nishihara)

3.3 Why the Metaphor of ‘Museum’? We have been using the metaphor of ‘museum’ to refer to our digital storage spaces. Why do we need exhibit spaces similar to a ‘real’ museum in addition to the data storage spaces? To keep simple digital data archives from becoming a dead storage space visited by nobody, we need exhibit spaces where each file is given a story. The parallelism of ‘real’ museums and our digital museum is quite

8

takubo

fitting. As an excavated object in a ‘real’ museum is given a story surrounding it when it is on exhibit, a video file is given its own story in our digital museum. The purpose of the exhibit space is twofold. One is to provide a forum for local people. Another one is for the development of raw data files into full-fledged contents as more meta-data is added. 3.3.1 To Provide a Forum for Local People Important events in the community have been recorded and edited for exhibits in the museum. The purpose of the exhibits in this case is for the younger generation to watch and learn, i.e. the recordings serve as manuals when they perform the events, thereby helping intergenerational transmission. The local people can also exhibit their works and performances, e.g. story books for children, musicals, or songs. Below are some of the contents that have been uploaded to the museum. MyaakuzIcI: We filmed a religious ritual called MyaakuzIcI, which is a festival held in Nishihara to pray for a rich harvest. It is held for four days around August or September on the day of Kinoe Uma, i.e., the 31st day in the sexagenary (60 day) cycle of the Chinese lunar calendar. The festival is conducted by male members of the Nanamui, a group formed for conducting communal religious rituals. We made recordings of over a period of 24 hours and edited them down into a 45-minute film with detailed explanations. The edited film is intended to serve as a manual for the younger generation to be able to carry on the ritual. Bilingual picture books for children: We have made picture story books for children based on stories written by the principal of a local nursery school. The stories are written in both Ikema and standard Japanese. Nishihara Muradate (The Founding of Nishihara village): The local people performed a musical drama depicting the migration of their ancestors from Ikema Island to Nishihara in Miyako Island about 140 years ago. This musical drama was filmed by a local broadcasting station. We have added subtitles to the film both in Ikema and standard Japanese. We have converted the contents into digital files and uploaded them to our museum. 3.3.2 Meta-data Updating through Exhibition Digital archives can easily become a dead storage space. This is because raw audio or video files without any transcriptions or translation are inaccessible to most people. In order to make raw files as accessible as possible, we have to provide such files with meta-data such as transcriptions, translations, and glosses, in addition to the basic event-related information. Processing files for exhibition in the digital museum provides the necessary incentives for researchers to

the case of ikema ryukyuan

9

figure 1.4 Metadata updating through Exhibition at the Museum. The function of the layers.

transcribe and translate files. If sufficient meta-data is provided, the files can be accessed by people in other fields, e.g. sociologists, anthropologists, or linguists not specializing in the language. Sound or movie files acquire more meta-data as they go from the archives to the exhibit space.

4

Some Special Features of the Museum

Our digital museum is quite unique among other web-based homepages for endangered languages. It is quite easy to update: researchers themselves can update the page by filling in the template pages. It makes extensive use of links, and the same video file can be linked to various pages and viewed from different angles. 4.1 Updating Updating the site will be done directly through the internet by researchers themselves by filling in the template instead of writing scripts. Up to now, it has been necessary to manually write the action scripts as in Fig. 1.5. Moving forward, we can directly edit pages. From a page that looks like Fig. 1.6, we can directly edit the explanation and upload a file and its two subtitles, the transcription of Ikema and the translation in standard Japanese. After editing, the page will look like Fig. 1.7.

10

figure 1.5 Action Scripts

figure 1.6 Wordpress

takubo

the case of ikema ryukyuan

11

figure 1.7 Results

4.2 Use of Links Our museum makes use of multiple links to the same file. One and the same file can be interpreted from various viewpoints. For example, a file named tamunu (firewood gathering) is linked to time line page ‘a day in Nishihara’ that depicts the life of an old lady when she was young, and to a location on a map of the village that depicts where the various daily activities took place. By making use of links, different stories can be associated with the same file, giving the file different and hopefully deeper interpretations. 4.3 Extendability The system, once completed, can easily be extended to make sites for other languages and cultures. We have been constructing spaces for the Kikaijima dialect of Amami and the Jeju dialect of Korean.

5

Conclusion

One of the reasons for the current endangering of various languages and dialects in the world is that the younger generation are having less and less exposure to their language. Our digital museum provides much needed exposure to the language that helps to preserve it. Even though the revitalizaton of endangered languages needs much more than a museum, our digital museum may provide a useful tool in the preservation of endangered languages.

12

takubo

References Hayashi, Yuka. 2010. Ikema. Shimoji and Pellard (eds.) An Introduction to Ryukyuan Languages, pp. 167–188. Hirai, Meari. 2009. Facing the crisis of Nanamui: The case of Nishihara in Miyako Island. Kyoto University Global coe Program ‘Reconstruction of the Intimate and Public Spheres’: Proceedings of the 1st Next Generation Global Workshop, pp. 663–674. Iwasaki, Shoichi, and Tsuyoshi Ono. (2011). Ikema Ryukyuan: Investigation past experience and the current state through life narratives. Japanese/Korean Linguistics 19. Chicago: csli, pp. 351–365. Shimoji, Michinori and Thomas Pellard (eds.) An Introduction to Ryukyuan Languages. Research Institute for Languages and Cultures of Asia and Africa, Tokyo University of Foreign Studies. Takubo, Yukinori. 2011. ‘Kikigengo dokyumentēshon no houhou toshiteno denshihakubutsukan sakusei no kokoromi: Miyakojima Nishihara chiku o chūshin toshite [The Digital Museum Project for the Documentation of Endangered Languages: The case of Ikema Ryukyuan]’. Nihongo no Kenkyu [Studies in the Japanese Language]. Vol. 7, No. 4. The Society for Japanese Linguistics, pp. 119–134.

part 2 Historical Linguistics

∵

chapter 2

Disentangling Japonic Seaweed from Koreo-Japonic Water Anton Antonov

As is undoubtedly the case with many other researchers working on Korean and Japanese with a particular interest in their history and possible relationship, Whitman’s seminal work (1985) on the comparison of these two languages has been a source of inspiration and an example of how one should go about looking for the sound laws that would eventually prove the existence of the (still) hypothetical Koreo-Japonic language family. Even though his was not the first work on this difficult subject, it still stands out as one which until recently was the most systematic and firmly grounded in the Comparative Method. Indeed, Vovin’s recent monograph on the subject (2010) has benefited from major developments in our understanding of the history of both Koreanic and Japonic, but it still harks back to Whitman’s major endeavour in this area almost thirty years ago as it presents a point-by-point rebuttal of all but six of the 352 etymologies proposed therein (1985: 209–246). Since in the end Vovin (2010: 238–240) comes close to denying the probability of ever being able to prove these two languages to be related, I think one could claim that Whitman (2012) is a major improvement on Whitman (1985) inasmuch as it tries to present a better case for Koreo-Japonic while addressing some of the criticisms in Vovin (2010). This is hardly the place to review all the etymologies Vovin (2010) rejects and to discuss the well-foundedness of his arguments. Such an enterprise could indeed only be undertaken in a similar monographic format. Neither will I attempt to evaluate the new proposals put forward by Whitman (2012). Instead, I will try to take a fresh look at some of the comparisons and propose a new etymology.

1

Water

Frellesvig and Whitman (2008: 35) reconstruct two central vowels in protoJapanese (namely *ɨ and *ə) mainly on the basis of two sets of correspondences in which Old Japanese ö (usually assumed to reflect pre-oj *ə) corre-

© koninklijke brill nv, leiden, 2018 | doi: 10.1163/9789004351134_003

16

antonov

sponds either to Middle Korean (mk) u or to mk o. Still, given that the examples adduced are phonologically extremely similar and the fact that there is at times considerable semantic latitude in their meanings (‘wh[at]’ vs. ‘conjectural adverb’ [=if]; ‘fly’ vs. ‘ride’; ‘eldest, chief (of kin)’ vs. ‘base, origin’), we may be dealing with accidental lookalikes in some cases, and with loanwords in others. Indeed, the fact that mk koWol(h) ‘county’ :: oj köpori is one of the items compared strongly supports this latter hypothesis for at least some of them. So why not consider that this correspondence is one which is only found in loanwords? If we do this, then we would have to conclude that all of the other comparanda, among which ‘fire’, ‘seaweed’, ‘snake’ and ‘boat/prow’, are loanwords, presumably from Koreanic. I think this may well be the case of the words for ‘snake’ and ‘boat’, but the case of the word for ‘fire’ is a little bit more complicated. Vovin (2011) and Pellard (2013) have independently shown that the Japonic word for ‘fire’ was *poy, and consequently the possibility of its being cognate with mk púl ‘id.’ has been de facto denied. In defense of this comparison Whitman (2012: 32) rejects the Japonic reconstruction on the ground that the philological evidence for Old Japanese pô instead of pö in this particular word is disputable, but in doing this he overlooks the fact that the Japonic reconstruction relies on data from both Eastern Old Japanese and Ryukyuan and therefore rests on firm ground. Now, the obvious conclusion would be to say that the words for ‘fire’ in Japonic and Koreanic are simply not cognate. İt ürür, kervan yürür (The dogs bark, but the caravan moves on), as the Turkish saying goes. Still, I think that we could approach the vowel mismatch problem from a different perspective. We could say that these words are indeed cognate, and that the correspondence they exhibit—mk u(l) :: pJ *o(y) with yodization—is one diagnostic of true cognates, in which case the words for ‘fire’ in Japonic and Korean would turn out to be cognate. This has the merit of not being an obvious correspondence, and certainly not one we would expect to find in loanwords. Another advantage is that we can now tentatively rescue a few more comparisons. mk tul- ‘hold, take, lift’ would now correspond regularly to oj tôr- ‘id.’, whereas its correspondence with oj tör- would be of the loanword type, thus incidentally waiving aside Vovin’s (2010: 122) qualms about this comparison. The same would be true of mk múl ‘water’ :: oj mî ‘id’, even if here the story is a bit more complex. Indeed, by applying the correspondence mk u(l) :: pJ *o(y), we would expect mk múl to correspond to pJ *mo(y) (> oj *mï), whereas the current reconstruction is pJ *me (> oj mî). This could actually be accounted for if we assume that *me is actually an ablaut variant of *mo(y)

disentangling japonic seaweed from koreo-japonic water

17

(whose final yod was lost very early on as it only appeared in compounds), just as seems to be the case of the (initial in the) Japonic words for ‘stone’ (pJ *esoy ~*osoy [> eoj osu/i]) and ‘breath’ (pJ *eki ~*oki [> eoj okî]). Ablaut variation in Japonic has been rejected by Vovin (2010: 7), but the argument against it that he adduces, the accentual register difference between the alternants, is also applicable mutatis mutandis to Indo-European (where the existence of ablaut can hardly be questioned although there does not seem to have been a register distinction in pie) and is in fact definitional of the phenomenon. Indeed, if the register distinction in Japonic is taken to reflect earlier segmental distinctions, the existence of ablaut becomes all the more probable as once again, just as in Indo-European, ablaut could be expected to occur in derivation and inflection. This, in turn, may account for the segmental differences between the ablaut forms which later evolved into a distinction in register. Thus, the examples of (what looks like) ablaut that do exist in Japonic need to be further researched. If we accept the existence of pJ *mo ‘water’ with an ablaut variant pJ *me and thus cognate with mk múl ‘id.’ one obvious question would be whether it is attested as such in Japonic. I think that the answer is yes. I would like to argue that the oj word mopî which appears in the (Heian period) Japanese reading of the name of the Water Office (主水司 mopitori no tukasa) at the Ministry of the Imperial Household (宮内省 kunaishō) under the Ritsuryō (律令) system whose members were in charge of water, rice gruel and ice supply (Omodaka and al. 1967: 747) is actually a compound consisting of the words mo ‘water’ and pî ‘ice’. Even though the second member of this compound is phonetically attested only in the compound 宇須良婢 usura-mbî ‘thin ice’ (mys 20.4478), its use as a kungana for (kō-rui) pî is well-attested. The word mopî is phonographically attested in the oj corpus with the meaning of ‘water vessel’ and (by extension?) ‘water’ (Omodaka and al. 1967: 747) but I would suggest that these are secondary meanings which developed once mopî was no longer transparent as a compound, and was reanalyzed as the name of the vessel used to fetch water (and ice) and ultimately as a word for ‘water’ as well. That this derivation is quite plausible is confirmed by the etymology of the Russian word for ‘water bucket’vedró, which continues Proto-Slavic *vědrò, which, in turn, ultimately derives from the pie root *ued-r/n- ‘water’. Vedró is thus from the same family as the (Russian) word for ‘water’, vodá (Derksen 2008: 518–519). Unfortunately, mopî seems to have left no trace whatsoever and apart from moguru ‘to dive’, which is attested only in Modern Japanese and is probably not derived from *mo ‘water’ anyway, I have been unable to find this last element in other compounds in any modern dialect or Ryukyuan language.

18 2

antonov

Seaweed

Let us consider next mk mól ‘seaweed’. This word has been compared by Whitman (1985: 237, #253) to oj mo (mö?) ~ më and although with a different theory on the vowel correspondence in Frellesvig and Whitman (2008: 35) and Whitman (2012: 30). It has been rejected by Vovin (2010: 193) as a Koreanic loanword because of irregular vowel correspondence as per Whitman (1985: 129). Once again the story may be a bit more complex. While it is true that if as suggested mk o :: oj ö (< *ə) reflects a loanword correspondence this comparison should be discarded from the list of potential cognates, it is actually not certain that the vowel of oj mo ‘seaweed’ was originally an otsu-rui one, i. e. mö. Indeed, among the several phonographic attestations of this word in Western Old Japanese texts there is one in particular which I think allows us to posit it had a kō-rui vowel (as does Bentley 2001: 246 and pace Vovin 2010: 193 who writes that ‘although the word is written as 毛 /mô/ in several cases, none of them [appears] in texts that faithfully preserve the distinction between /mô/ and /mö/’). It is its occurrence in the Norito liturgies, where it is notated with a kō-rui vowel, i. e. as mô in nt 5 and nt 6. Given that there appears to be a ratio of one-hundred percent of etymologically correct spellings of the syllable mo in these texts (Bentley 2001: 35) we can assume that this was the original vowel of the Japonic word for ‘seaweed’. Incidentally, we cannot really be sure that the vowel of the ‘alternant’ form më is an otsu-rui vowel either, since I do not find the phonographic attestation in Fudoki uncontroversial (contra Omodaka and al. 1967: 732), especially since it occurs in the name of a well 米多-井 mëta-wi ‘mëta-well’, which is given as the ‘popular’ version of the original name which appears in the emperor’s decree naming the well in Chinese as 海藻-生-井 ‘a well where seaweed grows (lit. seaweed-grow-well)’. This ‘popular’ name however is not analyzable in a straightforward manner (in particular, the identity of the element ta seems rather obscure to me). Besides, reflexes of oj më (?) do not seem to be attested in any Ryukyuan language outside of cases of obvious borrowing from mainland Japanese. Furthermore, given that the distribution of these two presumably apophonic forms, i. e. mo (which should be the bound form if from earlier *mə) and me (which should be the independent one if from earlier *məy), does not correspond to the one generally expected in the case of bound vs. independent forms, we should consider the possibility of having to do with yet another case of ablaut variants. Now two’s company, three’s a crowd. We would not want to multiply ad infinitum this type of variation especially since we are dealing with two phonologi-

disentangling japonic seaweed from koreo-japonic water

19

cally identical words, and so instead of positing that what we have here are two homophones which happen to mean respectively ‘water’ and ‘seaweed’, and have *e ~ *o ablaut variants, it might be better to assume that we are actually dealing with a single etymon and that the word for ‘seaweed’ is derived from the word for ‘water’. And indeed, the Norito (nt 5 and nt 6) preserve what may well be the original form of the word as 毛波 môpa in the collocations 奥都毛波 oki-tu môpa ‘seaweed of the offing’, 邊津毛波 pe-tu môpa ‘seaweed of the beach’, which could be analyzed as ‘water leaves’. This word seems also to be attested in nsk 4 as 茂 播, where 茂 might well be another piece of evidence in favor of mô (Vovin, p.c.), if we take pa as the word for ‘leaf’ instead of an a priori more plausible (and the one adopted by nk commentators) analysis as the topic particle. While a topic particle seems indispensable here (in which I agree with Vovin, p.c.), I would argue for haplology whereby *mopa pa > mopa. Furthermore, mopa is still present in a number of modern dialects as moba with expected rendaku in a tatpuruṣa compound (Kitahara: 2001–2002, online edition). It is also remarkably parallel to the Slavic (Russian водоросли Bielorussian водарасці, Bulgarian водорасли), Turkic (Turkish su yosunları, Tatar суүсемнәр) and possibly Estonian vetikad (from vesi ‘water’ and a denominal adjectival suffix -kas) words for ‘seaweed’ which literally translate as ‘water plants’. Given that mogusa ‘water grass’ is another ancient way of referring to seaweed (Omodaka and al. 1967: 737), this seems to be one satisfying way of accounting for all these forms. Of course, even if one does not accept the tentative analysis of the origin of oj mo ‘seaweed’ as an apocopated version of (pre-)oj môpa just presented, oj mô (from pJ *mo(y) with apocopated final yod) would not correspond regularly to mk mól according to Whitman (2012: 30)’s new vowel correspondences and would thus count as a loanword from his viewpoint, as according to him mk o comes from ok *ə and should correspond to pJ *ə, not pJ *o. But if the vowel correspondence he has put forward is taken to be diagnostic of loanwords as proposed above, pJ *mo(y) (with subsequent loss of the yod) could actually be cognate with mk mól, with no need of positing that its reflex oj mô comes from an earlier *ua diphthong and therefore appealing to Whitman’s law of medial r loss (cf. Whitman 1985: 189–208, Whitman 1990), which would derive oj mô from earlier *mura > *mua > *mwa > mô and ultimately from pJK *mol (Whitman 1985: 237), presumably by way of pre-oj *mur (Whitman 1985: 144). Therefore, if what we originally had in Japonic was a compound based on the Japonic reflex of the proto-Koreo-Japonic word for ‘water’ and the Japonic word for ‘leaf’ (i. e. mopa) as suggested above, we would have to assume that mo ‘seaweed’ (with no kō-otsu distinction in the vowel) was in fact either the

20

antonov

apocopated version of mopa and so etymologically a(n ablaut) variant of the word for ‘water’ or else that it is cognate with mk mól. In the latter case we must assume that there were in (pre-) Old Japanese (and really, in Japonic) two competing words for ‘seaweed’, a transparent compound (‘water leaves’) and a specialized word, both having cognates in Koreanic, as the word for ‘water’ in the former, and the word for ‘seaweed’ in the latter case. This last one ended up as the only survivor in later stages of the Japonic languages as it is well attested in Ryukyuan (see Pellard, this volume). But since we also have to account for the variant më (if, after all, with an otsu-rui vowel), one final possibility, which I find the most appealing, is that mo as the word for ‘seaweed’ is the result of reanalysis of môpa ‘water leaves’ as being based upon mk mól ‘seaweed’ which was in fact borrowed at a later (than the hypothetical pJK language) date from Korean as *məy which would incidentally explain its non-attestation in Ryukyuan.

3

Concluding Remarks

I have tried to argue in this short paper that loanwords, and the regular correspondences they may exhibit, are to be taken far more seriously in the study of the (still) hypothetical relationship between Japonic and Koreanic. I think that if these languages are related, the sound correspondences they may present will most probably not be trivial (but of course regular to the extent where the corresponding lexical items have been preserved in the two languages), a point I have tentatively illustrated with the words for ‘fire’, ‘water’ and ‘seaweed’, well aware of the difficulties raised by these etymologies. I thus believe that our future work on the putative relationship between these two languages will be far easier once we have identified all the obvious loanword sound correspondences and taken them out of the picture so we could concentrate on those (if there are any left) words exhibiting non-trivial sound correspondences. From this perspective, Whitman (2012) is a welcome improvement in that some of the etymologies proposed show non-trivial correspondences which take into account historical sound changes posited for both languages. But we still have plenty of work ahead.

disentangling japonic seaweed from koreo-japonic water

21

References Bentley, John R. 2001. A Descriptive Grammar of Early Old Japanese Prose. Amsterdam: Brill. Derksen, Rick 2008. Etymological Dictionary of the Slavic Inherited Lexicon. Leiden: Brill. Frellesvig, Bjarke and John Whitman 2008. Evidence for seven vowels in protoJapanese. In Proto-Japanese: Issues and prospects, ed. Bjarke Frellesvig and John Whitman, pp. 15–41. Amsterdam/Philadelphia: John Benjamins. Kitahara, Yasuo et al., ed. 2001–2002. Nihon kokugo daijiten [A comprehensive dictionary of the Japanese language]. Shogakukan, 2nd edition. Omodaka, Hisataka, and al., ed. 1967. Jidaibetsu kokugo daijiten: jōdaihen [A chronological dictionary of the Japanese language: the language of Antiquity]. Tokyo: Sanseidō. Pellard, Thomas 2013. Ryukyuan perspectives on the Proto-Japonic vowel system. In Japanese/Korean Linguistics 20, eds. Bjarke Frellesvig and Peter Sells. csli Publications. Vovin, Alexander 2010. Koreo-Japonica. A Re-evaluation of a Common Genetic Origin. Honolulu: University of Hawaii Press. Vovin, Alexander 2011. On one more source of Old Japanese i2. Journal of East Asian Linguistics 20, pp. 219–228. Whitman, John 1985. The phonological basis for the comparison of Japanese and Korean. Doctoral Dissertation, Harvard University, Cambridge ma. Whitman, John 1990. A rule of medial *-r- loss in pre-Old Japanese. In Linguistic Change and Reconstruction Methodology, ed. Philip Baldi, pp. 511–546. Berlin: Mouton de Gruyter. Whitman, John 2012. The relationship between Japanese and Korean. In The Languages of Japan and Korea, ed. Nicolas Tranter. Oxon: Routledge.

chapter 3

On Feature Ranking in Japanese Onset Obstruents Bjarke Frellesvig

Introduction The scholarship of John Whitman is of an impressive width and depth, with original and seminal contributions within many branches of linguistics, over many languages, and to linguistic theory. This is apparent from, amongst much else, the wide array of papers in this celebration of John and his scholarship. For me personally, working with John on some of the many things he does has been, and continues to be inspirational and formative for my own work and its direction. But before anything else, a lot of fun. John’s dissertation, The phonological basis for the comparison of Japanese and Korean (1985), is widely read and cited for its contribution to the comparative study of Japanese and Korean, and to a large extent it remains the baseline within this field that other scholars relate to, and argue with, in their own work on the cognation—or not—of Japanese and Korean. However, his dissertation is also an important contribution to the study of the historical phonologies of both Japanese and Korean, demonstrating that John in addition to everything else also is an accomplished phonologist. In this paper, I offer a partial phonological analysis of the system of the obstruent onset consonants of Japanese through its attested history, from Old Japanese (oj;1 mainly 700–800 ad, but also including material from before 700), until the middle of the 19th century. I will show that the mutual relationship, in terms of ranking, between the two distinctive feature categories [+/– continuant] and [+/– strident] plays an important part in the system and changes affecting it, and that a shift in ranking took place sometime in the second half of the Late Middle Japanese period.

1 ‘Standard’ or common Modern Japanese (MdJ, 1600 onwards) may be regarded as a direct descendant of the Early Middle Japanese (emj, 800–1200) and Late Middle Japanese (lmj, 1200–1600) of Kyoto reflected in the majority of pre-modern written sources, which was geographically displaced from Kansai to Kanto in the course of the establishment of Edo (present-day Tokyo) as the capital at the very beginning of the MdJ period, and which subsequently was subject to substratum influence, in some parts of the language to a significant degree, from the surrounding Kanto dialects and other varieties of Japanese brought to the expanding capital in the course of the Edo period (1603–1867).

© koninklijke brill nv, leiden, 2018 | doi: 10.1163/9789004351134_004

on feature ranking in japanese onset obstruents

23

The approach taken in this paper situates itself within the theoretical framework of phonological structure and language change established by Henning Andersen, who taught at Harvard when John was an undergraduate there.2 As background, I first give a brief, general overview over the main sound changes which have taken place within Japanese.3

Sound Changes in Japanese Through thirteen centuries of attested history, from Old Japanese to the present day, the Japanese language has undergone only one major structural sound change, namely the introduction of quantity sensitivity (or: the distinction between long and short, or heavy and light, syllables), structurally to be understood as the introduction of complex or branching syllable nuclei (or: the appearance of a new, post-peak, position within the syllable nucleus). This change, which seems to have taken place around 800 ad, included the addition of consonant and vowel phonemes, which could occur in the new syllable position, to the phoneme inventory of the language, often written as for example /n, q, c, i, ĩ, u, ũ, v/; see Frellesvig (2010: 185ff.) for details.4 Other than the introduction of new syllable post-peak phonemes accompanying the change in syllable structure around the beginning of the Early Middle Japanese (emj) period, and a few subsequent changes involving these phonemes (hoj: 309 f., 319 ff., 385f.), the Japanese language has undergone remarkably few segmental sound changes, both in its prehistory and through its attested history.

Vowels Between proto-Japanese (pJ), which had a seven vowel (/*i, *e, *a, *o, *u, *ɨ, *ə/) system, and oj, a number of sound changes took place, mainly as summarized in (1) (see Frellesvig and Whitman 2008 for details). These changes resulted in

2 See for example Andersen 1974 and 1980 (cited elsewhere in this paper), alongside many other publications by Andersen. 3 The exposition is brief and does not problematize. Further references and on some points reference to differing views may be found in the cited sections of Frellesvig and Whitman 2008 and Frellesvig 2010. 4 Frellesvig 2010 will in this paper be abbreviated as ‘hoj’.

24

frellesvig

the stable five vowel system, /i, e, a, o, u/, of oj which has remained unchanged since then. (1) Changes affecting pJ vowels Contraction of /vv/ diphthongs to unitary vowels:

pJ /*ei/ > oj /e/;5 /*ai/ > /e/; /*əi/ > /e/ to sequences of glide and vowel: /*ia/ > /ye/; /*iə/ > /ye/; /*ui/ > /wi/; /*oi/ > /wi/; /*ɨi/ > /wi/; /*ua/ > /wo/; /*uə/ > pre-oj /*uo/ > /wo/ Mid vowel raising /*e/ > /i, ye/; /*o/ > /u, wo/ Merger of central vowels and backing /*ɨ, *ə/ > *ə > /o/

Consonants pj had a small, simple inventory of consonants: /*p, *t, *k, *s, *m, *n, *r, *w, *y/. The main change which took place between pJ and oj was the contraction of sequences of nasal and obstruent to yield oj /b, d, g, z/ (e.g. /*np/ > /b/, /*nt/ > /d/, etc.), giving the oj consonant inventory of /p, t, k, s, b, d, g, z, m, n, r, w, y/. As is well known, the oj mediae, /b, d, g, z/ were prenasalized, that is pronounced with a nasal onglide, e.g. /d/ [nd]. It is also widely thought that the tenues, /p, t, k/, were allophonically voiced in medial position, e.g. /-t-/ [d]; it is not clear whether medial voicing also applied to /s/. Sometime in the Middle Japanese period the phoneme /f/ (which changed to /h/ probably in the second half of the 18th century) was added to the phoneme inventory through a phonemic split of /p/ into /p/ and /f/,6 but other than that also the consonant inventory

5 This paper is about the phonemes of Japanese and therefore Japanese sounds and words are written in a phonemic transcription appropriate to the period under discussion, enclosed in phoneme brackets. For oj, the phonemic analysis and interpretation adopted here is that also reflected in the transcription system often referred to as the ‘Frellesvig&Whitman’ system. See hoj: 33, 176, 414f. about the chronology of sound changes in oj, emj and lmj and timeappropriate phonemic transcription for the various phonological stages of the language. 6 The dating of this change is disputed, but as we shall see below it is overwhelmingly likely that it, viewed from a strictly phonemic point of view, took place in the second half of the lmj period.

on feature ranking in japanese onset obstruents

25

in syllable onset position remained unchanged from oj until the second half of the 19th century, when the sudden large-scale intake of loanwords from European languages resulted in the addition of more onset consonants. (2) oj and emj onset consonants: /p, t, k, b, d, g, s, z, m, n, r, w, y/ late lmj onset consonants: /p, t, k, b, d, g, f, s, z, m, n, r, w, y/ Pre-1900 MdJ onset consonants: /p, t, k, b, d, g, s, z, m, n, r, w, y, h/ However, although the inventory of onset consonant phonemes remained quite stable from oj until the mid to late 19th century, a number of sound changes took place which both have changed the phonemic shape of words and morphemes and appreciably have changed the sound texture of the language. These changes fall in a two overall groups, which may roughly be characterized as segmental changes, involving loss, change and split of segments, and (primarily) phonetic changes, i.e., changes in phonetic realization rules which only to a smaller extent resulted in phonotactic or segmental changes. The former group, segmental changes, consists largely of two partially interrelated sets: (3) a. oj /w, y/ were lost in many environments: /w/ as lost before /o, i, e/, and /y/ was lost before /e/, resulting in changes in the phonotactic distribution of these segments (hoj: 205ff., 310 f.). These changes include both the loss of the so-called kō-otsu syllable distinctions and the loss of syllable initial /w/ before /o, i, e/ and of syllable initial /y/ before /e/. b. oj /p/ lenited: Intervocalic /-p-/ changed to /-w-/ in the 10th century (and was subsequently lost before /u, o, i, e/, cf. immediately above) (hoj: 201ff.); later, remaining /p/ changed to /f/ in most environments, see (7) below. Some examples of changes in phonemic word shapes resulting from these two sets of changes are given in (4), showing the oj and the current, presentday forms in bold, and indicating when the forms reached their present-day phonemic shapes.

26

frellesvig

(4) oj

emj

/kwo.pwi/ ‘love’ > /winokwo/ ‘boar, pig’ > /wotokwo/ ‘man’ > /mapye/ ‘front’ > /yeda/ ‘branch’ > /nuye/ ‘thrush’ > /kapo/ ‘face’ > /kapa/ ‘river’ >

lmj

/kwo.pi/ > /ko.pi/ > /ko.wi/ > /koi/ /winoko/ > /inoko/ /wotoko/ > /otoko/ /mape/ > /mawe/ > /mae/ /eda/ /nue/ /kawo/ > /kao/ /kawa/

Obstruents The second group of changes (changes in phonetic realization rules) only to a smaller extent resulted in phonotactic or segmental changes, but they jointly changed the sound texture of the language. They all concern the phonetic realization rules of the obstruent onset consonants /p, t, k, b, d, g, s, z/. Also these changes fall in two sets, (a) one concerning prenasalization and voicing, and (b) another concerning the relationship between stop, fricative and affricative realizations of onset obstruents. The former consists in the loss of prenasalization of /b, d, g, z/ and of medial voicing of /p, t, k (, s)/ (this accompanied by introduction of aspiration of /p, t, k/). These two interrelated changes in phonetic realization rules are difficult to date, as they had no orthographic expression and had no visible phonemic effect, but loss of medial voicing appears to have been completed by the beginning of MdJ and loss of prenasalization was well under way by then. These two changes resulted in a significant change in the sound texture of Japanese, as shown here in the pronunciation of the word /tanabata/ ‘Vega, the Weaver’ in oj and in MdJ which segmentally has remained unchanged since oj, but sounds appreciably different:

(5)

oj

MdJ

/tanabata/ [tãnãmbada] [t(h)anabat(h)a]

on feature ranking in japanese onset obstruents

27

Stops, Fricatives and Affricates As for the second set of phonetic changes we can summarize what is known and may be reconstructed of the phonetic realization of onset obstruents as stops, fricatives, and affricates in Old and Early Middle Japanese as follows. (6) a. /s, z/ had both fricative and affricative variants, seemingly to some extent conditioned by the following vowel (hoj: 36 ff.) /s/ [ s, ts, -z, -dz ] /z/ [ nz, ndz ] b. /p/ had both occlusive and fricative variants (hoj: 37) /p/ [ p, ɸ, -b, -β ] c. /b, k, g/ most likely had both occlusive and fricative variants (hoj: 37 f., 196f.) /b/ [ mb, mβ ] /k/ [ k, x, -g, -ɣ ] /g/ [ ŋg, ŋɣ ] After emj the following changes occurred. (7) a. /p/ (remaining after /-p-/ > /-w-/ had taken place) sometime in the Middle Japanese period lost its occlusive variants in most environments and became /f/, but remained as /p/ with only occlusive realization in some environments, mainly after /n, q/ (hoj: 311 ff.), but also used in loanwords.7 This change is difficult to date, as it found no orthographic expression, but it was certainly concluded by the end of the lmj period. There is no persuasive independent evidence to show whether the phoneme /f/ phonetically was [f] or [ɸ], but as we shall see below it must have been [ɸ] at the point of change from /p/ to /f/. The split of /p/ into /p, f/ was followed by a change of /f/ to /h/ probably late in the 18th century (hoj: 386f.). Thus, for example oj /para/ ‘field’ > mj /fara/ > MdJ /hara/. b. /t, k, b, d, g/ lost their fricative variants, e.g. /k/ [ k(h) ]

7 Leaving aside here the issue of whether /p/ was retained or reinnovated in mimetics, e.g., MdJ pikapika ‘sparklingly’ (hoj: 315f.).

28

frellesvig

c. /t, d/ acquired affricative variants before high vowels /i, u/, sometime in the latter half of the lmj period (hoj: 322) /ti, tu/ [ ʧi, ʦu ] /di, du/ [ ʤi, ʣu ] d. /s, z/ lost their affricative variants /s/ [ s, ʃ ] /z/ [z] In the course, or intersection, of the latter two changes, /d/ and /z/ merged before the high vowels /i, u/ sometime early in the MdJ period (see further below), and as mentioned /p/ split into /p/ and /f/. However, other than that, the outcome of these changes was that the pronunciation of individual words and morphemes which contained the affected phonemes changed, but with no further phonemic mergers or splits, for example:

(8)

oj

MdJ

/tuma/ ‘spouse’ [tũmã] [ʦuma] /tikara/ ‘strength’ [tigara ~ tiɣara] [ʧikhara ~ ʧikara] /sama/ ‘mode, way’ [tsãmã] [sama]

The Phonemic System(s) of the Japanese Obstruents As may be seen from the summaries presented above, the Japanese onset obstruents have had phonetic realizations that include stops, fricatives and affricates at all stages of the language, but with these variants distributed differently across classes of phonemes at different stages of the language. The three consonant types stops, fricatives and affricates are distinguished in terms of the phonetic features [+/– continuant] and [+/– strident] which are defined as follows in the acoustic feature definitions from Jakobson and Halle (1968: 739–740), the last revision of the feature set emerging from Roman Jakobson’s work on the phonic, phonetic, and phonological structure of language. abrupt/continuant ([+/– continuant]); distinguishes occlusives (stops and affricates) from fricatives: “silence (at least in the frequency range above the vocal cord vibration) followed and/or preceded by a spread of energy over a wide frequency

29

on feature ranking in japanese onset obstruents

region, either as a burst or as a rapid transition of vowel formants (vs. absence of abrupt transition between sound and silence)” strident/mellow ([+/– strident]); distinguishes sibilants from non-sibilants: “presence (vs. absence) of a higher intensity noise accompanied by a characteristic amplification of the higher frequencies and weakening of the lower formants” A general classification or characterization of types of obstruents by these two phonetic parameters is shown in (9), distinguishing stops, two kinds of fricatives (strident and mellow), and affricates. The two unmarked obstruent types are stops ([-continuant] and [-strident]) and strident fricatives ([+continuant] and [+strident]) which have identical values for the two phonetic features, whereas the two marked sound types, affricates ([-continuant] and [+strident]) and mellow fricatives ([+continuant] and [-strident]) have opposite values for the two categories.

(9)

Stops

Fricatives Strident

Affricates

Mellow

[ p, b, t, d, k, g ] [ f, v, s, ʃ, z, ʒ ] [ ɸ, β, θ, ð, x, ɣ ] [ ʧ, ʦ, ʣ, ʤ ] [continuant] [strident]

– –

+ +

+ –

– +

Phonemes are not mere ‘bundles’ of distinctive features, but structured syntagms of distinctive features. Each phoneme in a language is defined in terms of a number of phonemic (distinctive) features (which are mutually ranked), i.e., values for a subset of the universal set of distinctive feature categories. Values for features which are phonemic in a particular language remain invariant in phonetic realization,8 whereas values for features which are not phonemic, that is to say, which are redundant, in a language are assigned by phonetic

8 Except where affected by neutralization rules (which are feature deleting rules) and except for distorted or extreme allegro speech.

30

frellesvig

realization rules with reference to one or more contexts and are open to variation. One important part of the phonological analysis of a language is therefore to try to determine which distinctive feature categories are phonemic in that language and which are redundant and have values assigned by phonetic realization rules (and also to try to find the contexts with reference to which values for redundant features are assigned). Applying that point of view to the analysis of the Japanese onset obstruents, whose phonetic realizations have included stops, fricatives and affricates at all stages of the language, it is easy to see that in oj it was the feature [+/– strident] which was invariant and phonemic, whereas the feature [+/– continuant] was subject to variation, see (10) which shows the feature values for these two categories against the phonetic variants of the onset obstruents in oj. The values for the feature [+/– strident] were invariant across the phonetic realization of all onset obstruents, whereas the phonetic realization exhibited variation with respect to the values for [+/– continuant]. Thus, the main opposition within the class of onset obstruents in oj was between non-sibilants ([-strident]) /p, t, k, b, d, g/ versus sibilants ([+strident]) /s, z/, as summarized in (11) which also shows that variation was possible with respect to the redundant category of [+/– continuant]. This analysis of the oj obstruent system is similar to the one I gave in hoj: 34ff. Turning now to the onset obstruents of lmj and early MdJ, we see that the situation there is the opposite of that in oj/emj. As shown in (12) and summarized in (13), the phonetic variation found in early MdJ shows that values for [+/– continuant] are invariant, and that that category therefore was phonemic, whereas values for [+/– strident] exhibit variation and that that category was redundant. Thus, by early MdJ a change in ranking and distinctiveness had taken place between the two categories [+/– strident] and [+/– continuant], such that the main opposition between classes of obstruents now was between occlusives ([-continuant]) (/p, t, k, b, d, g/) and continuants (/s, z, f/).

on feature ranking in japanese onset obstruents

oj and emj Phonemic System

(10)

/s/ /z/ /p/ /b/ /k/ /g/ /t/ /d/

[ s, -z ] [ ts, -dz ] [ nz ] [ ndz ] [ p, -b ] [ ɸ, -β ] [ mb ] [ mβ ] [ k, -g ] [ x, -ɣ ] [ ŋg ] [ ŋɣ ] [ t, -d ] [ nd ]

(11)

[ Strident ]

[ Continuant ]

Phonemic Invariant

Redundant Subject to variation

+ + + + – – – – – – – – – –

+ – + – – + – + – + – + – –

/ p, b, t, d, k, g / / s, z / strident [ continuant ]

– –/+

+ +/–

31

32

frellesvig

Early MdJ Phonemic System

(12)

/p/ /t/

[p] [t] [ʧ, ʦ ] /k/ [ k ] /b/ [ b ] /d/ [d] [ʣ, ʤ] /g/9 [g, ŋ ] /s/ [ s, ʃ ] /z/ [ z ] /f/ [ f ] or [ɸ]

(13)

[ Continuant ]

[ Strident ]

Phonemic, invariant

Redundant, subject to variation

– – – – – – – – + + + +

– – + – – – + – + + + –

/ p, b, t, d, k, g / / f / / s, z / continuant [strident]

– –/+

+ ?

+ +

Thus, Japanese has through its attested history, at least until the second half of the 19th century, had more or less the same inventory of onset obstruent phonemes (/p, t, k, b, d, g, s, z/ in oj, /p, t, k, b, d, g, s, z, f/ at the end of lmj, and /p, t, k, b, d, g, s, z, h/ in mid-19th century nj), but the phonemic systems were different at different stages.10 One significant sound change in the history 9

10

(12) includes [ŋ], a fully nasal variant realization of /g/. It is not clear when this variant established itself as the main, normative intervocalic realization of /g/ in the MdJ period, but in any case, nasal stops such as [ m, n, ŋ ] are [-continuant] (see e.g. Anderson 1976), and the presence of [ŋ] does not affect the analysis given here. This, of course, does raise the question—which I shall not attempt to address here—of in

on feature ranking in japanese onset obstruents

33

of Japanese may be identified as a shift in ranking between the two phonetic feature categories [+/– strident], which was phonemic in oj, but later became redundant and subject to variation, and [+/– continuant] which was redundant and subject to variation in oj, but later became phonemic. The dating of this shift in ranking is difficult to determine, but it formed part of the complex of changes whereby /s, z/ lost their affricative realizations, but /t, d/ acquired affricative realizations, and /p/ split into /f/ (in most contexts) and /p/ (mainly after /n, q/). The dating of the loss of affricativization of /s, z/ is very difficult to trace, but was probably the first of these phonetic changes. The assibilation of /t, d/ started in the second half of the lmj period. The dating of the split of /p/ > /p, f/ is disputed. However, looking at that change from a phonemic perspective, we see that /p/ > /p, f/ took place as an integral part of the re-ranking of [+/– strident] and [+/– continuant], such that the phonetic realization of the reflex of oj /p/, which included a fricative variant [ɸ], was reinterpreted as being phonemically [+continuant], /f/, in most environments, except after /n, q/ where it was interpreted as [-continuant], /p/.11 As long as the phonemic category was [+/– strident], reinterpretation of /p/ as /f/ was not relevant, as both [p] and [ɸ] are [-strident] and both could serve as realizations of /p/. However, once [+/– continuant] became the phonemic category, values for that category had to be established for the realizations of the reflex of oj/emj /p/: Language learners had to make decisions in the course of acquisition about the values for the realizations [p] and [ɸ] in terms of the feature [+/– continuant].12 There are two possibilities: either (a) to interpret the different observable realizations as realizations of a single phoneme (either [-continuant] /p/, or [+continuant] /f/), in which case one of the observed phonetic realizations would have to be disregarded, or (b) to interpret the different realizations as

11

12

which sense, for example, oj /p, t, k, b, d, g/ which are defined as being [-strident] but have no phonemic specification for [+/– continuant] are the same phonemes as early MdJ /p, t, k, b, d, g/ which are defined as being [-continuant] but have no phonemic specification for [+/– strident]. Needless to say, once the split of /p/ into /p, f/ had taken place, /p/ could occur outside of the environments in which it had mainly been found before the split, for example in any phonotactic position in loanwords. This description is simplistic, of course. In reality, decisions about which categories are phonemic (and which are redundant) as well as about values for phonemic features are made through a recursive series of decisions about valuation, segmentation, and ranking in the course of the acquisition of the phonological system of a language, see Andersen 1974, 1980. The points of principle remain, however.

34

frellesvig

realizations of two different phonemes with different values for [+/– continuant]: [-continuant] /p/, and [+continuant] /f/. Here, evidently, the second possibility, a phonemic split, was the outcome. The phonetic realizations and the phonemic systems and the phonemic features at the stages of oj and late lmj are shown in (14).

(14) oj phonemic system

Phonetic realization

Late lmj phonemic system

[-strident]

[p]

[-continuant]

/p/

[-strident]

[ɸ]

[+continuant]

/f/

/p/

In this way we are able to understand the split of oj/emj /p/ into late lmj/MdJ /p, f/ in the context of the re-ranking of [+/– strident] and [+/– continuant]. The affricativization of /t, d/ is also part of that change, as /t, d/ could not have affricative variants (which are [+strident]) as long as [+/– strident] was the phonemic category distinguishing the [-strident] non-sibilants (including /t, d/) from the [+strident] sibilants /s, z/. The affricativization of /t, d/ set in well into the second half of the lmj period, and accordingly the phonemic split of /p/ into /p, f/ (or, the change of /p/ > /f/ in most environments) must from a strictly phonemic point of view also be thought to date from the second half of lmj. It is also worth noting that while /f/ must have had the sound value [ɸ] at the point of re-interpretation from /p/ in the late lmj period, it is not possible to rule out the possibility that it acquired a variant, or phonetically mainly became, [f] (which is [+strident]) in the MdJ period. We have little evidence either way; the fact that /f/ later became /h/ does not tell us much, as changes of [+strident] /f/ [f] to /h/ are amply attested in languages of the world. Considering the mutual ranking and difference in distinctiveness between [+/– continuant] and [+/– strident] also gives a perspective on the outcome of the merger of /z, d/ before /i, u/. The traditional statement is that /d/ and /z/ merged before front vowels as /z/. This is reflected for example in current kana spelling, which uses じ and ず in all but a few morphophonemic spellings which retain ぢ and づ, and in both kunreishiki romanization as well as the transcription system favored by most linguists, which use zi and zu, and also uncritically adopted, for example, in my hoj: 384f. However, considering the fact that from late lmj and in MdJ [+/– continuant] became the phonemic category, we can see that the traditional interpretation of the outcome of that

on feature ranking in japanese onset obstruents

35

merger is misguided. While it does seem clear that /z/ and /d/ indeed merged as [+continuant] /z/ [z] before /u/, it seems on the other hand equally clear that /z/ and /d/ merged as [-continuant] /d/ [ʤ] before /i/, see (15).13 (15) midu ‘water’, mi-zu ‘not seeing’ > [mizu] /mizu/ kizi ‘pheasant’, kidi ‘the road to Kii’ > [khiʤi] /kidi/ Descriptions of Japanese historical phonology have not to a large extent made use of distinctive feature based analyses, one of a handful of exceptions being John’s dissertation (1985). I hope to have shown in this paper that insights and worthwhile perspectives may be gained on the historical phonology of Japanese by employing an understanding of historical phonology which not only includes distinctive features, but also takes seriously the structure of phonological systems (including the organization of distinctive features within the system) and the structure and mechanisms of phonological change.

References Andersen, Henning 1974. ‘Towards a typology of change: Bifurcating changes and binary relations’. In Historical Linguistics ii: Theory and Description in Phonology, edited by John M. Anderson and Charles Jones, pp. 17–61. Amsterdam: North-Holland Publishing Company. Andersen, Henning 1980. ‘Morphological change: Towards a typology’. In Recent Developments in Historical Morphology, edited by Jacek Fisiak, pp. 1–51. The Hague and Berlin: Mouton. Anderson, Stephen R. 1976 ‘Nasal consonants and the internal structure of segments’. Language 52: 326–345. Frellesvig, Bjarke 2010. A History of the Japanese Language. Cambridge: Cambridge University Press.

13

Since the end of the nineteenth century, the large intake of loanwords from European languages has, as mentioned earlier, resulted in the addition of a number of phonemes to the language including on most analyses unitary affricative phonemes, and on such analyses both [+/– strident] and [+/– continuant] are phonemic in the language today. That does not affect the fact that initially the outcome of the merger of /z, d/ before /i/ was /d/, and that writing kidi for the word pronounced [khiʤi] (in parallel with writing kiti for the word for ‘military base’ which is pronounced [khiʧi]) is preferable over writing kizi.

36

frellesvig

Frellesvig, Bjarke and John Whitman 2008. ‘Evidence for seven vowels in protoJapanese’. In Proto-Japanese: Issues and prospects, edited by Bjarke Frellesvig and John Whitman, pp. 15–42. Amsterdam & Philadephia: John Benjamins. hoj = Frellesvig 2010. Jakobson, Roman and Morris Halle 1968. ‘The revised version of the list of inherent features’. In Roman Jakobson: Selected Writings 1: Phonological Studies, pp. 738–742. The Hague and Berlin: Mouton. Whitman, John. 1985. The phonological basis for the comparison of Japanese and Korean. Unpublished Ph.D. dissertation. Harvard University.

chapter 4

Fishy Rhymes: Sino-Korean Evidence for Earlier Korean *e Marc H. Miyake

When I first started studying Korean in 1987, I had already been studying Japanese for years while also dabbling on and off in Chinese. Although native Korean vocabulary was alien to me, Sino-Korean was not, as it resembled SinoJapanese. While adding Sino-Korean readings by hand to entries in my copy of Nelson’s Modern Reader’s Japanese-English Character Dictionary, I began to notice patterns of correspondence (Table 4.1). table 4.1

Selected correspondences between Sino-Korean and Sino-Japanese rhymes

Middle Chinese rhyme namea

Early Middle Chinese rhymeb

Late Middle Chinese rhymeb

Tongkwuk cengwun rhyme namea

Modern Sino-Korean (Yale romanization)

Modern SinoJapanese (Go-on)

Modern SinoJapanese (Kan-on)

齊 ‘even’

*-ej

*-jej

雞 ‘chicken’

-(y)ey

-ai

-ei

祭 ‘sacrifice’

*-iejʰ

魚 ‘fish’

*-ɨə

*-ɰə / -jə

居 ‘dwell’

-(y)e

-o

-yo

a For simplicity I will only use level tone rhyme names whenever possible to represent Middle Chinese rhymes regardless of tone: e.g., 魚 ‘fish’ represents the rhymes 語 (rising) and 御 (departing) as well as 魚 (level). The departing tone rhyme 祭 ‘sacrifice’ has no counterparts with other tones, so I cannot use a level tone rhyme name to refer to it. b All Chinese reconstructions are my own unless stated otherwise.

However, the readings of various common characters do not conform to these patterns. Some Middle Chinese 齊 ‘even’ and 祭 ‘sacrifice’ rhyme characters have modern Sino-Korean readings ending in -(y)e instead of the expected -yey found in the prescriptive readings of Tongkwuk cengwun (東國正韻) (Table 4.2).

© koninklijke brill nv, leiden, 2018 | doi: 10.1163/9789004351134_005

38

miyake

table 4.2

Examplesa of Late Middle Chinese *-jej corresponding to Sino-Korean -(y)e

Sinograph

Early Middle Chinese

Late Middle Chinese

Tongkwuk cengwun

Modern Sino-Koreanb

Modern Sino-Japanesec

低

*tej

*tjej

tyèy

ce < tye

tei

底抵邸

*tejˀ

*tjéj

tyěy

黎

*lej

*ljej

lyèy

lye

rei

麗戾

*lejʰ

*ljèj

lyéy

勵

*liejʰ

西栖棲犀

*sej

*sjej

syèy

se < sye

sai ~ sei (西), sei (栖棲), sai (犀)

誓逝

*dʑiejʰ

*ɕɦ jèj

ssyéy

妻悽凄

*tsʰej

*tsʰjej

chyèy

sei che < chye

sai

a This is not an exhaustive list. Some low-frequency homophones have been omitted. b Pre-1933 reform spellings are given after “ Go-on n-, Cantonese j-)

ŋ- (> Go-on g-)

Liao Chinese ź-, Standard Mandarin r-, Sino-Korean ∅< z-

Liao Chinese ŋ-, Standard Mandarin and Sino-Korean ∅< ŋ-

*ʑ-

*(ŋ)g-

n- ~ dz- ~ l-

g-

In the sixth century, Saṃghabhara transcribed Sanskrit -ai with the 齊 ‘even’ rhyme (Pulleyblank 1984: 199) borrowed into Go-on as -ai and Sino-Vietnamese as -ây [əj]. And the early southern pronunciation of the 魚 ‘fish’ rhyme was probably something like *-ɨə, not *-jəj, judging from early Sino-Vietnamese -ưa [ɨə]. The Suiko period 魚 ‘fish’ rhyme man’yōgana 居 and 擧 transcribed Old Japanese kë [kəj] with ‘b-type e’ and suggest a falling diphthong in some southern Middle Chinese dialect. However, those syllables had velar initials unlike the problematic cases of Sino-Korean -yey examined here. The two phenomena may not be related. If the anomalous Sino-Korean reflexes of the 齊 ‘even’, 祭 ‘sacrifice’ and 魚 ‘fish’ rhymes cannot be explained in terms of Chinese, are they simply random distortions which were unworthy of inclusion in the prescriptive Tongkwuk cengwun, or do they reflect Korean-internal changes? Serafim (1999) proposed that Korean ye [jə] came from a Proto-KoreoJaponic *é whose Proto-Japonic reflex was *e. I brought this to the attention of Frellesvig and Whitman (2008) who expanded upon Serafim’s proposal.

fishy rhymes: sino-korean evidence for earlier korean *e

41

One need not believe in a common ancestor of Koreanic and Japonic to believe that Korean ye [jə] is in part from an earlier *e. Frellesvig and Whitman (2008: 38) listed Japanese loanwords borrowed from Koreanic before and after mid vowel raising. One of these words was 高句麗 Kōkuri ‘Koguryo’ whose final -i is irregular (Table 4.2) and may be from an *-e that was anomalously raised in word-final position. In any case, that -i cannot be from a *-jə. Nor is it likely to reflect the -í of Mandarin 高句麗 Gāogōulí ‘Koguryo’ since the Japanese would probably not have borrowed Kōkuri from northern Chinese centuries after the fall of Koguryo. Although there is no guarantee that the Silla dialect which borrowed SinoKorean had an *e around the same time as the source(s) of Koreanic loanwords in Old Japanese, reconstructing an *e in early Sino-Korean may shed light on otherwise inexplicable instances of ye [jə] in the above readings and others. Sino-Korean -ye [jə] for the 齊 ‘even’ and 祭 ‘sacrifice’ rhymes may go back to *-e which is a better match for Early Middle Chinese *-(i)ej and Late Middle Chinese *-jej.1 Sino-Korean -ye ~ -yey (< *-e ~ *-ej?) variation for those rhymes after acute initials may reflect allophony in their northeastern Chinese source. On the other hand, an early Sino-Korean *-ej that broke to *-yey [jəj] in the 魚 ‘fish’ rhyme morphemes 除諸豫預 is not a good match for my Late Middle Chinese reconstruction *-ɰə / -jə. In Liao Chinese postdating the borrowing of Sino-Korean, the 魚 ‘fish’ rhyme was something like *-y; it was transcribed in the Khitan small script as ⟨iú⟩ in Kane’s (2009) notation.2 Perhaps early SinoKorean *-ej was an attempt to render a northeastern Late Middle Chinese *-øy, an intermediate stage between early Tang *-jə with a mid vowel and Liao *-y with a high vowel: Pre-Tang *-ɨə > early Tang *-jə > *-jo > *-ø > *-øy > Liao *-y Cantonese has a similar front rounded diphthong -œy after alveolars (Table 4.4). The use of *e for a foreign ø has a parallel in Japanese borrowings like rentogen ‘X-ray’ from German Röntgen.

1 *-e is also a match for modern Sino-Japanese -e in 世 se with the 祭 ‘sacrifice’ rhyme. However, this Sino-Japanese -e may be a monophthongization of an earlier *-ai and hence may not necessarily reflect a Sino-Paekche source like *se. 2 ⟨iú⟩ is Kane’s transcription of Khitan small script character 289 which appears in only two contexts: in Chinese loanwords and as a dative-locative suffix after nouns with the vowel a. Character 289 may have been read as *y (i.e., as a Chinese vowel absent from native Khitan words) in the former context and as *-da in the latter context.

42

miyake

table 4.6

The Middle Chinese 皆 ‘all’ rhyme after *k- in Sino-Korean

Sinograph

Early Middle Chinese

Late Middle Chinese

Tongkwuk cengwun

Modern Sino-Korean

皆

*kɛj

*kæj

kày [kaj]

kay [kɛ] kyey [ke] (!)

階 介芥价疥 界堺戒誡屆

*kɛjʰ

*kæ̀ j

káy [kaj]

kay [kɛ] kyey [ke] (!)

It is unclear how many other instances of Sino-Korean ye go back to *e. Was, for instance, 雞 kyey (Middle Korean [kjəj], modern Korean [ke]) once read as *kej or *kjəj? It is also unclear when *e broke to [jə]: e.g., was Middle Korean tyèh ‘flute’ borrowed before breaking as *teh from a very late northeastern Middle Chinese 笛 *tɦ jeɣ? At present I prefer to reconstruct *e only to explain anomalous Sino-Korean readings, though I do not rule out the possibility that all Sino-Korean ye were once *e. Two more classes of anomalous Sino-Korean readings which may have had *e are indicated with exclamation marks in Tables 4.6 and 4.7. The normal Sino-Korean reflex of the Middle Chinese 皆 ‘all’ rhyme is -ay, but several Sino-Korean morphemes have -yey after k- (Table 4.6). If this -yey is from *-ej, it could be an archaism reflecting Early Middle Chinese *-ɛj. Another possibility is that Late Middle Chinese *-æj was borrowed both as *-aj and *-ej by Silla speakers who lacked a low front vowel *æ. The -yey (< *-wej?3) instead of the expected -yu [ju] (< *-juj?) for Late Middle Chinese *-jwi (Table 4.7) is harder to explain. Why would Silla speakers borrow *-jwi as *-wej as well as *-ju( j)? *-wej and *-ju( j) may have been competing attempts to imitate Late Middle Chinese *-jwi, a sequence that was probably impossible in Silla as well as in later, attested stages of Korean. *-wej had a palatal vowel like *-jwi but lacked a palatal medial *-j-, whereas *-ju( j) had that medial without a palatal vowel.

3 I reconstruct a *-w- since Sino-Korean normally has -w- for Middle Chinese *-w-, and it is difficult to believe that Late Middle Chinese *kjwi with *-w- would be borrowed as SinoKorean *kej without *-w-.

fishy rhymes: sino-korean evidence for earlier korean *e table 4.7

43

Late Middle Chinese *kjwi and *gjwi in Sino-Korean

Sinograph

Early Middle Chinese

Late Middle Chinese

Tongkwuk cengwun

Modern Sino-Korean

規

*kwi

*kjwi

kywùy [kjuj]

kyu [kju]

癸

*kwiˀ

*kjwí

kywǔy [kjuj]

kyey [ke] (!)

季

*kwiʰ

*kjwì

kywúy [kjuj]

葵

*gwi

*kɦjwi

kkywùy [gjuj]

揆

*gwiˀ

*kɦjwì

kkywǔy [gjuj]

悸

*gwiʰ

*kɦjwì

kkywúy [gjuj]

kyu [kju]

kyey [ke] (!)

This short study demonstrates how comparative Chinese phonology and Korean historical phonology can complement each other. I hope to do more such studies in the future.

Primary Source Tongkwuk cengwun (東國正韻), 1448ad

Secondary Sources Beijing daxue Zhongguo yuyan wenxuexi yuyanxue jiaoyanshi [Beijing University Chinese Language and Literature Linguistics Teaching and Research Office] 1960. Hanyu fangyin zihui [Collection of characters in Chinese regional pronunciation]. Beijing: Wenzi gaige. Frellesvig, Bjarke; Whitman, John B. 2008. ‘Evidence for seven vowels in proto-Japanese.’ In: Frellesvig, Bjarke; Whitman, John B. (eds.). Proto-Japanese: Issues and Prospects. Amsterdam & Philadelphia: John Benjamins, pp. 15–41. Gong, Hwang-cherng 2002. ‘12 shiji-mo Hanyu de xibei fangyin (shengmu bufen)’ [The northwestern dialect of Chinese at the end of the 12th century: initials]. In: Gong Hwang-cherng. Han-Zangyu yanjiu lunwenji [Collected Articles on Sino-Tibetan Studies]. Taipei: Academia Sinica, pp. 243–281.

44

miyake

Kane, Daniel 2009. The Kitan Language and Script. Leiden: Brill. Pulleyblank, Edwin G. 1984. Middle Chinese: A Study in Historical Phonology. Vancouver: University of British Columbia. Serafim, Leon A. 1999. ‘Reflexes of Proto-Koreo-Japonic mid vowels in Japonic and in Korean.’ Presentation at ichl xiv, Workshop on Korean-Japanese Comparative Linguistics.

chapter 5

A mokkan Perspective on Some Issues in Japanese Historical Phonology* Sven Osterkamp

The ever growing number of inscribed wooden tablets, or mokkan, excavated from numerous sites in Japan provides scholars of various disciplines with invaluable first-hand data specifically from the 7th and 8th centuries, often even dated explicitly. From the perspective of language and writing there are a good many reasons to pay due attention to this corpus: Mokkan provide us with countless examples of spontaneous and often rather casual writing, leading to characteristics not commonly observed in the traditional corpus of received texts, such as the often mentioned laxness towards the mixed use of ongana and kungana type phonograms and also towards voicing oppositions. Furthermore, the sprouts of katakana-like abbreviations can be seen here, just as examples of senmyō-gaki are found. So-called uta mokkan are among the earliest specimens of Japanese poetry preserved in a form actually dating from the Old Japanese (oj) period. Similarly, ongi mokkan grant us insights into the beginnings of the tradition of glossing Chinese texts and, by extension, the beginnings of lexicography in Japan. The corpus is likewise a goldmine for attestations of further phonograms not easily found elsewhere, especially disyllabic ones, and of lexical items not appearing in the received texts. It has also long been a major source for research on numeral classifiers. In terms of grammar, mokkan may appear less insightful at first glance, among other reasons, due to the paucity of inflected words spelled out phonographically, but there are valuable exceptions.1 Below we will have a closer look at some mokkan that are of special significance for the study of oj and Middle Chinese phonology.

* The author would like to thank Tomasz Majtczak, Gordian Schreiber and Alexander Vovin, in alphabetical order, for their helpful comments on an earlier draft of this paper. 1 To name just one example: Instead of V+Am.ë=ya=mö, as it is widely attested in the poetry of the Man’yōshū to form rhetorical questions, we find V+Am.u=ya=mö on the well-known ongi mokkan from the Kita-Ōtsu excavation site (see e.g. mr #32 for a reproduction and transcription).

© koninklijke brill nv, leiden, 2018 | doi: 10.1163/9789004351134_006

46

osterkamp

Mid Vowel Raising It is common nowadays to posit the two mid vowels *e and *o for Proto-Japonic as one source for Western Old Japanese (woj) /ê, î/ and /ô, u/ respectively. Apart from the exact conditions for partial and full raising of these mid vowels, for which different proposals have been put forward, there is another important issue to be solved, namely that of dating the change.2 Frellesvig and Whitman (2008: 38), for instance, consider tera ‘Buddhist temple’ and potoke ‘Buddha’ to be loanwords from post-raising times, which “must have been borrowed after the introduction of Buddhism into Japan in the first half of the 6th century.” This suggests that raising of *e at least was already complete at that time. However, there is no proof that the words in question were actually borrowed in forms containing *e. For instance, Vovin (2007: 76) convincingly suggests “pre-woj *taira or *tiara” as the immediate predecessor of tera, so that mid vowel raising ceases to be relevant here. Similarly, Pellard (2014: 692, 694) derives potoke from Proto-Japonic *pətək[ə|a]i, whose final syllable *k[ə|a]i (rather than *ke) is suggested to go back to the Koguryŏ or Paekche word *kɛj ‘king’. The proposed dating is also in conflict with Miyake (2003b: 126), who had previously dated the raising of *e and *o to circa 590 on different grounds. Compared to the overall number of mokkan the amount of uta mokkan is almost negligible. They are undoubtedly far from being negligible in qualitative terms, however, when it comes to the study of oj—and, as it turns out, also when it comes to the issue of mid vowel raising. The single most widely attested poem on mokkan is undoubtedly the Naniwazu 難波津 poem, which is otherwise probably best known for its prominent position in the kana preface to the Kokin waka-shū (913). It runs as follows in its oj form:3 Nanipa-ndu=ni / sak.u=ya kö=nö pana # / puyu-ŋgömöri / ima=pa parumbê=tö / sak.u=ya kö=nö pana # ‘These blossoms blooming at Naniwa port—these blossoms blooming, [as if to say]: “now [it is] spring time!”, having stayed inside during winter.’ 2 Or rather changes, as one should be cautious in taking it for granted that both processes of raising occurred simultaneously. Cf. already Frellesvig and Whitman (2008: 39) on the chronological ordering of the two. 3 Note that kö=nö pana can mean either ‘these blossoms’ or ‘blossoms on the trees’ (usually taken to refer to plum blossoms), or maybe both meanings are alluded to here. We have opted for the former interpretation here.

a mokkan perspective on japanese historical phonology

47

What interests us here is the only verb form occurring in the poem, namely the attributive form sak.u ‘to bloom’, followed by the interjectional particle =ya. It is found twice in adnominal position to the noun pana ‘blossom(s)’. Now consider the following 7th century mokkan excavated at the Ishigami site, the inscription on one side of which reads thus (see e.g. mr #26): 奈尓波ツ尓佐児矢己乃波奈[布由?]

Nanipa-ndu=ni / sak.ô=ya kö=nö pana / [puyu?] […] As the excavation site is located in Asukamura and thus not on the periphery but rather in the linguistic center of Japan at the time in question, this suggests that similar to 8th century Eastern Old Japanese and several modern Japanese dialects (see Pellard 2008 among others), the central language likewise had partially raised -ô from *o in attributive forms at one point. There is yet another mokkan featuring the Naniwazu poem that clearly has sak.ô—which comes from the Heijōkyū excavations, however, and will therefore date from the 8th century instead (see Mokkan kenkyū 23: 12). In its first instance sak.ô is apparently again written with 児 /kô/ here, but the bad state of preservation renders the decipherment difficult. The second instance involves straightforward 作古 /sakô/. Regrettably only a vague dating is possible here, but in any case we are provided with another case from the central language from the oj period. In passing we may also note a case of -ô from the Western end of Honshū. On an ongi mokkan excavated from the archaeological site situated on the Yoshida campus of Yamaguchi University in 2015 we find, e.g. 露 (lù ‘dew’) glossed as ツ由 /tuyu/, but more importantly 雨 ( yù ‘to rain’) glossed as 不路 /purô/, i.e. pur.ô ‘to rain’.4 As a consequence, we arguably have to posit two steps for the process of mid vowel raising: *o → /ô/ → /u/. The above-mentioned mokkan are witnesses of the second stage, whereas the usual woj texts generally reflect the third. Let us turn to another case, this time involving *e. According to the Kōzanji manuscript of Wamyō ruijū-shō (vii: 45b) the place name in Echizen province, Tsuruga district underlying the spelling 鹿蒜 (implying ka ‘deer’ plus firu ‘Liliaceae’) had the form 加比留 /kafiru/. In the Daitōkyū manuscript and the

4 In the press release of Yamaguchi University (2015) 路 is interpreted as rendering /ru/, thus yielding pur.u. In view of the fact that 路 was in general use to write /rô/ but not /ru/, however, this seems hardly tenable.

48

osterkamp

related 1617 print (vii: 20a) it is, however, given as 加倍留 /kaferu/. The latter appears to be closer to the original form of the name, as is not only suggested by various instances of Kaferu-yama ‘Mt. Kaferu’ appearing in classical poetry for instance (see e.g. Kokin waka-shū viii: 370, 382, xvii: 902), but more importantly also by several mokkan providing the alternate spelling 返 (implying kapêr.u ‘to return’).5 In view of the fact that the spelling 鹿蒜 found in Wamyō ruijū-shō is already attested in the eighth century,6 we see a clear correspondence of 蒜 with /pêru/ for that time.7 This /pêru/ is reminiscent of Thorpe’s (1983: 290, #85) reconstruction of *peru ‘garlic’ for proto-Ryukyuan—and we are inclined to think that this is no coincidence, but rather simply yet another case of an only partially raised mid vowel well into the 8th century.8 Needless to say mokkan are not the only sources providing new evidence for mid vowel raising. Notably modern Japanese dialects have not necessarily been taken into sufficient account so far in the reconstruction of proto-Japonic. In addition to those cases addressed by Pellard (2008) and others, such dialects feature a number of items that likewise seem to retain original mid vowels. Thus, instead of kusuri ‘medicine’ we repeatedly meet with forms having /o/ in the second syllable in Japanese dialects (cf. Hirayama 1992: 1616–1618). This probably would not mean too much if it was not for the fact that Thorpe (1983: 305, #130) again actually reconstructs *kuso’ri for proto-Ryukyuan on the basis of purely Ryukyuan data. It seems certain that a thorough search in various directions will likely yield further corroborative data for the process of mid vowel raising—and ideally also hints as to the dating of the changes involved. Mokkan are undoubtedly among the most promising candidates here.

5 See e.g. the one listed in Mokkan kenkyū (17: 20), where the name Kapêru is immediately preceded by a reference to Tsuruga district. 6 See document #505 in Dai-Nihon komonjo (iewake 18.2; see e.g. p. 240), dated to 766-10-21, writing 赤 instead of each 示 in 蒜 however. 7 One might be tempted to explain this away as a dialectal correspondence, interpreting both 鹿蒜 and 返 as rendering /kapîru/ based on pîru ‘Liliaceae’ (see Kojiki kayō 43) as well as on Eastern oj kapîr.u ‘to return’ (attested in Man’yōshū xx: 4339, from Suruga province). However, the involvement of Eastern oj seems unlikely for a place name in Tsuruga. Additionally, such an interpretation would leave the later form Kaferu-yama unexplained. 8 As we have already referred to Tsuruga above, we may also note the existence of no less than three different forms of this name in oj times, with mid vowel raising connecting the earlier two forms with each other: Tunôŋga → Tunuŋga → Turuŋga. See Osterkamp (2011: 191f., 194–196) for details.

a mokkan perspective on japanese historical phonology

49

lmc-based Phonograms One of the most characteristic changes from Early to Late Middle Chinese (emc/lmc), at least in the North, is what is referred to as the (partial) denasalization of nasal initials: [m] → [mb], [n] → [nd] and so on. This change is reflected in a variety of ways: in Chinese transcriptions of Sanskrit and other foreign words and names; in transcriptions of Northwestern Chinese in other scripts, such as Brāhmī, Tibetan, or Uighur script; in borrowings into other languages, such as the kan’on layer of Sino-Japanese (sj) and, to a lesser degree, also Sino-Korean (sk).9 It is likewise observed in Chinese languages of later times. Apart from Nihon shoki (720), which is widely known as making use of phonograms with “kan’on-like”—that is: Chang’an lmc-based—sound values, denasalization is generally only marginally reflected in man’yōgana. Consider however the following mokkan excavated at the Kannonji site in Tokushima prefecture (see e.g. mr #94): 麻殖評伎珥宍二升

‘two shō of 伎珥 meat [from] Wowe district’ More or less the only possible interpretation of 伎珥 is to read it as kizi ‘pheasant’, so that the second phonogram in question would render /zi/ (based on

9 On the surface only lmc *r- is clearly reflected as z- (later zero) in sk, while the other relevant initials are reflected as the nasals m-, n-, ng- (later zero). These nasals have been taken as proof by Vovin (1993: 254f.) that sk does not originate in the standard lmc as used in Chang’an, much like Arisaka (1936 [1944: 301f.]) already did in his classic study. In a similar fashion, Miyake (2003a: 113) “eliminated a northwestern lmc origin for sk on the basis of the different treatments of nasals” in Chang’an lmc and sk, and comparable views have been put forward by Eom (2001: 25) among others. In fact, however, the situation is different and the evidence less conclusive, if we also take the phonetic realization of Korean nasals into account—as was already pointed out by Kōno (1964–1967 [1979: 507f.]). In word-initial position the nasals /m, n/ may be realized as something variously described as [mb, nd], [mb, nd] or [b, d] up to the present day and we are fortunate enough to have materials at hand to demonstrate that this phenomenon has a long history. Thus, it is reflected in foreign transcriptions of Korean, especially Japanese ones—setting in with the Kamakura period Nichūreki 二中歴 (which in turn is based on two Heian period compilations)—but also European ones from the 17th century onwards (Witsen, Broughton, Hall etc.). Early Korean renderings of foreign words and names, again e.g. Japanese and European ones, reflect this as well. Such phonetic realizations can explain with ease how, for instance, even Chang’an lmc *[mb] may have been borrowed as /m/ into Korean.

50

osterkamp

lmc *ri`) instead of /ni/ (emc *ɲɨʰ / *ɲiʰ) here.10 In Nihon shoki we likewise find 珥 employed as a phonogram, predominantly for /ni/ in the β sections, but as expected for /zi/ in the α sections.11 Now the mokkan does not carry an explicit date, but judging from the fact that 評 is still used here for ‘district’, it is unlikely to date from any time later than the very beginning of the 8th century, when the Taihō code was implemented and 評 was replaced by 郡 to designate districts. As another mokkan from the same site is explicitly dated to 689 for instance,12 it is naturally the closing years of the 7th century that come into question here. In other words: This instance of 珥 /zi/ here predates the completion of Nihon shoki by at least about two and possibly more decades and thus appears to constitute the earliest piece of clear evidence from Japan for a reflex of denasalization in lmc. This seems especially significant in view of the fact that neither kan’on sj as such nor sk is safely attested for this time, just as the above-mentioned foreign transcriptions of Chinese set in only later. Also, Mizutani (1957) argued that a number of transcriptions in Buddhist texts dating from the mid-7th century onwards already reflect denasalization, but most examples have been refuted by Hashimoto (2007). Now the compilation of Nihon shoki was of course not an overnight accomplishment. If Mori (1999) is correct in assuming that the two on-hakase Xu Shouyan 續守言 and Sa Hongke 薩弘恪 were in charge for the α sections, denasalization is likely to date back to about the mid-7th century after all, as their arrival in Japan was already in the early 660s. Looking at the same mokkan from a somewhat different angle, we also note that the form kizi is not otherwise attested in oj. Only its supposed predecessor kigisi is (see e.g. Kojiki kayō 2; Nihon shoki kayō 96, 110; Man’yōshū xiv: 3375). It is not before the Heian period that kizi makes its first appearance in the received texts, together with another trisyllabic form, kigisu (cf. jkd 238): Early 10th century sources such as Honzō wamyō and Wamyō ruijū-shō thus have 岐之 /kizi/, the latter additionally has 岐々須 /kigisu/. We may however well imagine the existence of the contracted form kizi already around the year 700 in the colloquial language, which thanks to the rather casual nature of mokkan serving as shipping tags is recorded here. The retention of the conservative form kigisi in poetry cannot possibly come as a surprise anyway.

10 11 12

All emc and lmc reconstructions quoted are from Pulleyblank (1991). See e.g. the statistical tables accompanying Mori (1981). Note however that “explicitly dated” for the time in question means that the indication of the year is merely provided by means of the sexagenary cycle ( ji-chou 己丑 in the case of the year 689; see mr #141).

a mokkan perspective on japanese historical phonology

51

Reflexes of Chinese /-ŋ/ As one of only exceedingly few ongi mokkan and especially as a rare source giving us some glimpses of Chinese character readings in 7th century Japan, no. 1318 excavated from the Asukaike site in Asukamura is among the best known of its kind (see e.g. mr #33). Its inscription appears to be a brief glossary relating to some Chinese text—probably to a passage in the Foshuo guanding jing 佛説灌頂經 (cf. t. 1331.21.534a)—indicating the readings of less than a dozen characters by means of near-homophones and phonograms. Apart from an instance of a katakana-like abbreviation of 左 /sa/ as ナ the most interesting feature of this mokkan is undoubtedly the gloss 汙吾 /uŋgu/ provided for the character 熊 (emc *wuwŋ, also cf. sk wung 웅). Deeming this reading to be ancestral to later go’on sj /u/, both Tranter (2001: 154–156) and Unger (2008: 51–53) regard this as evidence for the earlier existence of /ŋ/ as a phoneme distinct from /ŋg/, which is lost at some time postdating the borrowing of Chinese /-ŋ/ as /-ŋu/ so that no trace of the consonant is left in later sj. Even if not in the context of this mokkan, Miyake (2003a: 193) has similarly “hypothesized that oj had a short-lived phoneme *ŋ found only in sj loans which was later lost.” Can we, however, take it for granted that sj—chiefly attested in gloss texts from the Heian period onwards—exists in a straight line of development from what we see on this mokkan? That we are dealing with a diachronic change in the way Chinese /-ŋ/ is treated in Japanese is obvious: While velar nasal codas are generally reflected as the nasal vowels /-ũ, -ĩ/ in Heian period sj, there is nothing definite at all to suggest that this was likewise the case in the 7th and 8th centuries. Quite on the contrary, the little data we have for this period suggests exactly the same as the mokkan at hand, namely that /-ŋgV/ was the regular reflex. There is evidence from this in two directions: early loanwords, though scarce in number, confirm that /-ŋ/ was borrowed as /-ŋgV/ and conversely phonograms with /-ŋ/ in Chinese are used to write oj /CVŋgV/ in numerous cases.13 The two reflexes of Chinese /-ŋ/ are therefore in chronologically complementary distribution—there is no need, however, to derive the newer reflex from the older one by means of regular phonological development.

13

For loanwords see e.g. Classical Japanese suŋ guroku ‘sugoroku (a board game)’, which is attested logographically as 雙六 (emc *ʂaɨwŋ / *ʂœːwŋ, *luwk) in the Man’yōshū (xvi: 3827, 3838). Phonograms include 當 (emc *taŋ) in Taŋgîma 當麻 or 英 (emc *ʔiajŋ) in Aŋgupî 英比 for instance. See Osterkamp (2011: chapter 4.9) for further cases.

52

osterkamp

Now Tranter (2001: 155) claims that “there is no clear evidence for other cases of phonemic /ũ/ and /ĩ/ in Japanese of any period,” but in fact it is reasonable to posit both /ũ/ and /ĩ/, which in Japanese proper were the result of syllable reduction involving the lenition of prenasalized consonants for instance. Attestations for cases such as kaũmbasi- ← kaŋ gupasi- ‘fragrant’ or fusaĩ.nde ← *pusaŋ gî.te ‘obstructs and’ are already found in gloss texts from the 9th century (Tsukishima 1969: 365, 370). In the 7th and 8th century, /ŋg/ *[ŋg] would certainly have been the best choice available to render foreign [ŋ], but with the newly established /ũ, ĩ/ competitors for /ŋg/ entered the stage. Neither /-ŋgV/ nor /-ũ, -ĩ/ was a perfect match for Chinese /-ŋ/,14 but the latter had the advantage that the monosyllabic nature of the originals could be preserved, whereas /-ŋgV/ would always yield disyllabic forms. Generally it is not as if sj was a monolithic, invariable entity. Rather it adapted to changes both in Chinese and Japanese, and was subject to some fluctuation beyond the usual sound changes that occurred in the respective periods. For the Heian period this is especially true when it comes to the reflexes of coda consonants and medial glides, which we may illustrate by the following two cases: 1)

2)

14

15

Early sources such as Shin’yaku Kegonkyō ongi shiki 新譯華嚴經音義私記 or the Shōgozō ms. of Yangjuemoluo jing 央掘魔羅經 with glosses thought to date from the very beginning of the Heian period reflect Chinese /-n/ as /-ni/, whereas later ones adopt /-n/ instead. Similarly to the appearance of /ũ, ĩ/, syllable-final /n/ became phonotactically possible in Heian times through reductions of /n/-initial syllables or assimilation of /-r(V)/ to a following nasal for instance. Needless to say the new /-n/ was more faithful to the donor language than older /-ni/. Both 紀 and 貴 (emc *kɨ’ / *ki’ and *kujʰ respectively) were used to write oj /kï/, but later go’on sj has ki vs. kwi here. Miyake (2003a: 218) thus envisaged a tripartite distinction: oj /kyi/ vs. /kiy/ (written /kî/ vs. /kï/ here) plus an additional /kwi/ only used in sj. Plausible early loanwords such as ŋ gakï15 餓鬼 ‘preta’ appear, however, with /ki/ instead of /kwi/ (as should be expected in this scenario) in Heian sources (see e.g. Wamyō As long as we assume that this is indeed what was still present in the donor language, cf. however Numoto (1986: 35f., 134f.) with reference to kan’on, or in general for instance the loss of /-ŋ/ as a segment in the rhyme groups dang 宕 and geng 梗 in Northwestern lmc, the similar loss of /-ŋ/ in the rhyme zheng 蒸 reflected in shin-kan’on etc. We assume that the phonological form of this word is identical to the sound values of the two characters when used as phonograms, i.e. /ŋga/ and /kï/.

a mokkan perspective on japanese historical phonology

53

ruijū-shō [1617 print] ii: 4b). Even more importantly Shin’yaku Kegonkyō ongi shiki indicates the reading of 脆 (emc *tsʰwiajʰ)—obviously in false analogy to 危 (emc *ŋwiə̆ / *ŋwi, later read gwi in sj)—as 奇伊 /ŋgï-i/, with 奇 (emc *giə̆ / *gi, go’on sj gi). There is thus no evidence for the opposition Miyake posits, and the little data we actually have rather suggests that Miyake’s /kwi/ (and /gwi/) in sj was identical with /kï/ (and /ŋgï/) in oj proper. The opposition of /ki/ vs. /kwi/ in post-oj sj can easily be explained by the re-introduction of /kwi/ after the merger of oj /kî/ vs. /kï/, in an attempt to stay as faithful as possible to Chinese despite the significant changes occurring in Japanese.16 Therefore, there is no need at all to derive later sj /-ũ/ from earlier /-ŋgu/, nor is the mokkan in question evidence for the existence of /ŋ/ besides /ŋg/.

References Arisaka, Hideyo 有坂秀世 1936(1944). ‘Kanji-no Chōsen’on-ni tsuite’ 漢字の朝鮮音につ いて. In: Hōgen 方言 6.4–5. [Reprinted in: Kokugo on’inshi-no kenkyū 國語音韻史の研 究. Tōkyō: Meiseidō shoten, 1944.] Eom Ik-sang 嚴翼相 2001. ‘Hanchaŭm -l unmi Chungguk pangŭm kiwŏnsŏr-ŭi munje’ 한자음 -l 운미 중국 방음 기원설의 문제. In: Chungŏ Chungmunhak 中語中文學 29: 21–35. Frellesvig, Bjarke; Whitman, John 2008. ‘Evidence for seven vowels in Proto-Japanese’. In: Frellesvig, Bjarke; Whitman, John (eds.): Proto-Japanese. Issues and Prospects. Amsterdam: John Benjamins, pp. 15–41. Hashimoto, Takako 橋本貴子 2007. ‘Darani-no onshaji-kara mita jidaku bion-no hibionka-ni tsuite’ 陀羅尼の音写字から見た次濁鼻音の非鼻音化について. In: Chūgokugogaku 中国語学 254: 124–142. 16

Note that after the loss of distinctive /ũ, ĩ/ we again find borrowings and transcriptions of foreign—chiefly Korean—[ŋ] as Japanese /-gV/. The aforementioned Nichūreki and the sources following it up to circa 1600 provide us with little data in this respect. From that time onwards, however, we have a wealth of data from glossaries of Korean such as those in Intokuki 陰徳記 or E-iri ikoku tabi-suzuri 繪入異國旅硯. The oldest extant Japanese primer of Korean, Amenomori Hōshū’s well-known Zen’ichi dōjin 全一道人 (1729), also shows the same correspondence, this time from the perspective of one well proficient in both languages. As the default realization of the syllable-final nasal after the merger of /m, n/ was [n] as evidenced by transcriptions in Korean and Latin script for instance, /g/— be it realized as [ŋg] or [ŋ] for instance—was again the best match available for a foreign velar nasal.

54

osterkamp

Hirayama, Teruo 平山輝男 et al. (eds.) 1992. Gendai Nihongo hōgen daijiten 現代日本語 方言大辞典, vol. 2. Tōkyō: Meiji shoin. jkd = Jōdaigo jiten henshū iinkai 上代語辞典編修委員会 1967. Jidai-betsu kokugo daijiten jōdai-hen 時代別国語大辞典上代編. Tōkyō: Sanseidō. Kōno, Rokurō 河野六郎 1964–1967 (1979). ‘Chōsen kanjion-no kenkyū’ 朝鮮漢字音の研 究. In: Chōsen gakuhō 朝鮮学報 31–33, 35, 41–44. [Reprinted in: Kōno Rokurō chosakushū 河野六郎著作集, vol. 2. Tōkyō: Heibonsha, 1979.] Miyake, Marc H. 2003a. Old Japanese. A phonetic reconstruction. London, New York: RoutledgeCurzon. Miyake, Marc H. 2003b. ‘Philological evidence for *e and *o in Pre-Old Japanese’. In: Diachronica 20.1: 83–137. Mizutani, Shinjō 水谷真成 1957. ‘Tōdai-ni okeru Chūgokugo gotō bion-no Denasalization shinkō katei’ 唐代における中國語語頭鼻音の Denasalization 進行過程. In: Tōyō gakuhō 東洋學報 39.4: 1–31. Mori, Hiromichi 森博達 1981. ‘Kanjion-yori mita jōdai Nihongo-no boin soshiki’ 漢字 音より観た上代日本語の母音組織. In: Kokugogaku 国語学 126: 30–43. Mori, Hiromichi 1999. Nihon shoki-no nazo-o toku: jussakusha-wa dare-ka 日本書紀の謎 を解く——述作者は誰か. Tōkyō: Chūō kōronsha. mr = Asuka shiryōkan 飛鳥資料館 (ed.) 2010. Mokkan reimei: Asuka-ni tsudou inishieno mojitachi 木簡黎明——飛鳥に集ういにしえの文字たち. Asukamura: Asuka shiryōkan. Numoto, Katsuaki 沼本克明 1986. Nihon kanjion-no rekishi 日本漢字音の歴史. Tōkyō: Tōkyōdō. Osterkamp, Sven 2011. Nicht-monosyllabische Phonogramme im Altjapanischen. Wiesbaden: Harrassowitz. Pellard, Thomas 2008. ‘Proto-Japonic *e and *o in Eastern Old Japanese’. In: Cahiers de Linguistique—Asie Orientale 37.2: 133–158. Pellard, Thomas 2014. ‘The Awakened Lord: The Name of the Buddha in East Asia’. In: Journal of American Oriental Society 134.4: 689–698. Pulleyblank, Edwin G. 1991. Lexicon of Reconstructed Pronunciation in Early Middle Chinese, Late Middle Chinese, and Early Mandarin. Vancouver: University of British Columbia Press. Thorpe, Maner Lawton 1983. Ryūkyūan language history. Doctoral dissertation, University of Southern California. Tranter, Nicolas 2001. ‘The Asukaike word list slat and Pre-Sino-Japanese phonology’. In: McAuley, Thomas. E. (ed.): Language change in East Asia. London: Curzon Press, pp. 143–160. Tsukishima, Hiroshi 築島裕 1969. Heian jidai-go shinron 平安時代語新論. Tōkyō: Tōkyō daigaku shuppankai. Unger, J. Marshall 2008. ‘Early Japanese lexical strata and the allophones of /g/’. In:

a mokkan perspective on japanese historical phonology

55

Frellesvig, Bjarke; Whitman, John (eds.): Proto-Japanese. Issues and Prospects. Amsterdam: John Benjamins, pp. 43–53. Vovin, Alexander 1993. ‘About the phonetic value of the Middle Korean grapheme ㅿ’. In: Bulletin of the School of Oriental and African Studies, University of London 56.2: 247–259. Vovin, Alexander 2007. ‘Korean loanwords in Jurchen and Manchu’. In: Althai hakpo 알타이학보 17: 73–84. Yamaguchi University 2015. ‘Kokunai goreime-to naru “ongi mokkan”-o Yamaguchi daigaku Yoshida kyanpasunai-de hakken’ 国内5例目となる 「音義木簡」 を山口大学吉 田キャンパス内で発見. Press release, 28 August 2015. http://www.yamaguchi-u.ac.jp/ weeklynews/2015/_4437.html

chapter 6

A (More) Comparative Approach to Some Japanese Etymologies Thomas Pellard

Chaque mot a son histoire! (‘each word has its own history’). Such was the battle cry of dialectologists and other partisans of the Wellentheorie against the sound laws defended by the Neogrammarians. Though few nowadays would deny the regularity of sound changes and the validity of the comparative method, it remains true that recovering the history of words is often akin to detective work, and that the above maxim seems to be valid in the domain of etymology. Studies on the genetic relationship of Japanese with other languages have usually been based on lexical evidence foremost, which obviously poses problems if indeed each word has its own history. The search for external cognates of Japanese etyma cannot be limited to the comparison of attested written forms but requires a thorough reconstruction of the internal history of the languages involved as a preliminary step. In the Japonic domain, important advances have been made by John Whitman (e.g., 1985, 1990, 2008, Frellesvig & Whitman 2004, 2008a), whose work still forms the base of most studies on both the internal and external history of Japonic. It seems now widely accepted that the examination of the 8th century Old Japanese (oj) texts needs to be supplemented by taking into account the data from the different Japanese dialects as well as the Ryukyuan and Hachijō languages (Frellesvig & Whitman 2004, 2008a, Vovin 2010: 3–7, Whitman 2012: 25, Pellard 2008, 2013). Still, few works on Japanese etymology have tapped the rich mines of such “peripheral” data. One notable exception is for instance the illuminating discussion of the history of personal pronoun shifts in Japonic (Whitman 1999). Along the same lines, I will propose a more comparative approach to Japanese etymology and give some more examples of how the Ryukyuan data, and to a lesser extent the data from Ainu borrowings, can clarify our knowledge of the (pre)history of the Japanese language.1

1 All transcriptions of modern languages have been converted to standard ipa notations. ProtoRyukyuan tone categories are noted by the letters a, b, c. In addition to the references given

© koninklijke brill nv, leiden, 2018 | doi: 10.1163/9789004351134_007

a (more) comparative approach to some japanese etymologies

1

57

Of the Same Womb

Frellesvig & Whitman (2008a) accept the reconstruction of a mid-vowel *e in Proto-Japonic (pj), as first proposed by Hattori (1976).2 Though most of the supporting data comes from Ryukyuan, they adduce several examples of ê/e ~ î/i alternations within oj which might constitute further evidence of a vowel raising process *e > î in pre-oj.3 One of such examples adduced by Frellesvig & Whitman (2008a: 25) is oj ye ‘placenta’, which they propose to relate to i in irö ‘of the same mother’ by vowel raising in non-final position. The root ( y)e is itself scarcely attested in Japanese, and it does not actually appear in the oj corpus proper, but only as a Late Middle Japanese reading tradition of the character 胞 ‘womb, afterbirth’ in the Nihon shoki. The first bona fide attestation is in the Shinsen jikyō (898– 901) in the gloss konoye 子乃兄 for the Chinese word 膜 ‘membrane’ (Tenchi i: v14), but the syllables e and ye are sometimes not distinguished in that source (Hashimoto 1950: 208), so that a doubt subsists as to whether this word was actually e or ye.4 It is probably related to the later form ena ‘afterbirth’, which is only attested after the merger of the syllables e and ye. The Ryukyuan words for ‘placenta, amnion, afterbirth’ point to a different scenario than that suggested by Frellesvig & Whitman. Northern Ryukyuan forms such as Izena ʔíjà, Iejima ʔìjá, Nakijin ʔìjáː or Shuri ʔíjá do not generally exhibit a loss of the initial vowel to ʔj- and thus point to a proto-form *eja a, with a mid-vowel. On the other hand, Southern Ryukyuan forms such as Hirara zɨza or Sawada ɭɭa exhibit a liquid and obstruentization of the first vowel, which requires us to reconstruct *ira, with a high vowel. The Northern Ryukyuan forms can be reconciled with the Southern ones if we posit for the former two phonetically motivated, though irregular, changes: progressive palatalization *ira > *ija, and then height assimilation *ija > *eja. Deriving Southern Ryukyuan *ira from *eja would on the other hand require to posit less natural changes.5 Though a development *( j)ena > pr *ira might be not completely impossible, ( y)e could also be the result of the contraction of *ira, by medial -r- loss

2 3 4 5

in the bibliography, I have used my fieldwork notes on Kamikatetsu, Yuwan, Shodon, Okazen, Yoron, Ōgami, and Yonaguni. Japanese prose texts are referred to by volume and page number of the relevant critical edition, or in the case of a fac-simile edition by the name of the manuscript followed by volume, recto/verso, and folio number. See also Pellard (2008, 2013). See Pellard (2010) for further comments. The merger of the two syllables e and ye was completed by the first half of the 10th c. A form ija is recorded in the Nippo jisho (1603–1604) for several Kyūshū dialects, and it can also be reconciled with the Ryukyuan data by positing a similar r-yodization after *i.

58

pellard

(Whitman 1985: 22–25, Whitman 1990), “Whitman’s Law”, with the usual subsequent crasis of *ia into palatal ê/e. The comparison is then even clearer with the root attested in many oj terms for siblings born of the same mother, e.g. irö-ndö ‘lesser sibling’, irö-mô ‘younger sister’, irö-ne ‘elder sibling’, irö-se ‘brother’.6 The semantic link between ‘afterbirth’ and ‘sibling’ is not surprising, and the case of Greek ἀδελφός ‘brother’ (lit. ‘of the same womb’) immediately comes to mind. The Chinese words bāo xiōng 胞兄 ‘elder brother of the same parents’ (lit. ‘womb/afterbirth-brother’) and bāo mèi 胞妹 ‘younger sister of the same parents’ (lit. ‘womb/afterbirth-sister’) are an even closer parallel. The possibility of a link between the root *ira ~ *irə and oj ye ‘elder brother’ may also be envisaged, since oj ye is a 1.2 noun, that is to say of the high register, and pr *ira is probably an a-tone word, which corresponds to the Japanese high register. It is further possible to compare pj *ira ~ *irə with the Ainu root ir ‘(be) sibling’, though the direction of the borrowing is not clear. The Ainu root, even though it does not occur alone, is found in many words denoting siblings and relatives of the same age, like in ír-utar ‘siblings and cousins’, ír-okkayo ‘brothers’, irwak ‘siblings and other relatives of similar age’ in the Saru dialect.

2

Flat Hills and Other Mysteries

Both the Kojiki and the Nihon shoki relate the myth of the god Izanagi going after his deceased wife Izanami in the land of the dead (Yomi). Besides its striking similarities with the Greek myth of Orpheus, this tale contains a linguistic puzzle. Having found his wife in the land of the dead, Izanagi breaks the taboo by looking upon her corpse and flees, terrified by what he saw. Then, he is pursued by a thousand five hundred army until he reaches a place called Yömo-tu-pîrasaka, the limit between the land of the living and that of the dead, and manages to escape by pulling a giant rock to obstruct the passage. In the Kojiki, the toponym is written 黄泉比良坂 (i: 66), with only the part pîra written phonographically, while the Nihon shoki has both a logographic notation 泉津平坂 (i: 14, 16) and a phonographic spelling yömo-tu pîra-saka 余母都比羅佐可 (i: 16). The word pîra-saka is surprising, since the semantic 6 We might also add ira-tu-mê ‘my lady’, with the usual a ~ ö alternation, as suggested by Jōdaigo jiten henshū iinkai (1967: 105). This word is actually an honorific used for strangers, but the use of kin terms as honorifics for non-relatives is cross-linguistically common (e.g. Chinese, Korean, Japanese, etc.).

a (more) comparative approach to some japanese etymologies

59

spelling of the Nihon shoki clearly means ‘flat slope/hill’, and the root pîra is indeed well attested in Japanese with the meaning ‘flat’. The same oxymoron is seen in other toponyms such as pîra-yama ‘flat mountain’ (平山, mys ix: 1715), pîra-woka ‘flat hill’ (枚岡 Norito 394, 比良乎加 Wamyō ruijushō Genna vi-v6). Obviously, all of these instances of pîra-X cannot mean ‘flat X’. A quick survey shows that this etymon (usually çiɾa7) is used in dialects all over Japan to designate the slope or side of a mountain or of a hill, a plateau, or a cliff. Turning to the Ryukyuan languages, the same meanings are attested for this etymon, and a pr form *pira a, and pj *pira 2.1/2, can be reconstructed from the following forms: Yamatohama çiɾa ‘road over a mountain pass’, Kamikatetsu çìɾá ‘slope at the top of a mountain’, Yoron pjâː, Izena ɸíɾáː, Nakijin pˀjáː, Shuri ɸíɾà, Ishigaki psɨ̀ṣ a ‘slope’, Taketomi piɕə ‘slope (of a roof)’. Nevertheless, the relationship with *pira ‘flat’ cannot be easily eluded since the two are homophonous and share the same tone category: Yamatohama çiɾa-saɾi, Yoron pjâː-saŋ, Izena ɸíɾá-háːŋ̀ , Shuri ɸíɾá-hàŋ, Ishigaki psɨ̀sá-sàːŋ, Taketomi piɕə-səŋ. The relationship between ‘hill, slope’ and ‘flat’ is not obvious, but I propose that ‘flat’ is the original meaning of *pira and that the sense ‘hill, slope’ is a secondary development through ‘mountain pass, plateau’. A mountain pass is a less steep and lower road between two slopes, hence the extension of *pira ‘flat’ to ‘flat place (along a mountain road)’. The derivation ‘flat’ > ‘plateau’ is also a natural one. The subsequent extension to ‘slope’, ‘hill’, ‘hill or mountain side’, ‘cliff’ is easily understood as these are what naturally surround a mountain pass or a plateau. A typological parallel can be found in the relationship of Bulgarian планина, Czech planina and Serbo-Croatian planìna ‘mountain’ with Latin plānus ‘plain, flat’ (Buck 1949: 28). This reconstructed meaning agrees well with the symbolic value of pîrasaka in the Japanese myth. Like a mountain pass is a pathway between two mountains or valleys, pîra-saka represents the border and pathway between the world of the living and that of the dead, and it is made impossible to freely pass through by Izanagi precisely when the two divinities put in motion the cycle of life and death for humans. Interestingly, Ainu has a word pira meaning ‘cliff’, and also a verb pirasa ‘to spread out’. The phonological and semantic match with Japonic *pira is perfect and could hardly be a coincidence. The Japonic word might also be compared to Middle Korean (mk) pìlé ~ pìléy ‘cliff, bank’, for which it might be a better match than oj pê ‘edge, bank, vicinity’ (Whitman 1985: 151). 7 Some dialects exhibit a form heɾa instead, which might be related to the toponym pêra-saka 幣羅坂 mentioned in the Kojiki (ii: 182). However, the Ryukyuan evidence clearly points to earlier *i and not *e in the first syllable.

60

pellard

This etymology also illuminates another passage of the Kojiki, where the goddess Amaterasu is depicted as ‘attaching a five-hundred-arrow quiver to [her] pîra’ (比良邇者附五百入之靫, i: 74). Commentators do not agree on the meaning of this pîra, but it most probably means here ‘flank’ and is thus related to the gloss fira 2.2 for 方 ‘side’ in the Ruiju myōgishō (Kanchiin Sō-chū: v18).8 The same meaning ‘side, direction, vicinity’ is attested in many Japanese and Ryukyuan dialects, e.g., Tottori attɕi-biɾa ‘other there’, Yoron miɕi-bjaː ‘North direction’, Yamatohama hata-çiɾa, Nakijin hàtà-ppjáː ‘one side, one direction’. Further semantic extensions include Ōita kawa-n-heɾa ‘river shore’ or Hachijō çiɾa ‘border of a hearth, hearthside’. The semantic path is here ‘flat’ > ‘surface, face’ > ‘flank, side’ > ‘direction, vicinity’ and is partly paralleled by the evolution of Latin costa ‘rib, side’ and related words into French côte ‘coast, seashore, slope’, coteau ‘small hill’ or côté ‘side, along, vicinity’. There are also deeper implications for the reconstruction of earlier Japanese: compounds of the type pîra-X, such as pîra-saka, pîra-yama, pira-woka, do not mean ‘flat X’ and are thus not of the modifier-head type ([Adj [n]]). The element pîra is a noun meaning ‘pass, side’ modified by the following noun which restricts its denotation (‘pass or side of X’, [[n] n]). This also explains how Izanagi managed to block the pîra-saka of the land of the dead: it was not a saka, a slope or hill, but a pîra, a pass. Such compounds thus have a head-modifier structure, which is unexpected since all Japonic languages are consistently head-final. Considering also that oj has a system of prefixes, which is uncommon for a verb-final language, this disharmony could suggest the possibility that pre-oj underwent some important syntactic changes. This resonates with some recent work on alignment and word order in oj (Vovin 2009: 589, 661, 1055–1056, Yanagida & Whitman 2009).

3

Revisiting Some Koreo-Japonic Comparisons

The seminal dissertation of Whitman (1985) remains the most thorough work on the genetic comparison of Korean and Japanese. It presents a corpus of more than 300 morphological and lexical comparisons supported by a set of detailed sound correspondences, amended in Frellesvig & Whitman (2008a) 8 The character 靫 usually stands for yukî, a quiver traditionally worn on the back, but since Amaterasu is said to already have a thousand-arrow quiver on her back (負千入之靫) and since the five-hundred-arrow quiver is said to be ‘attached’ (附), in contrast to the first one which is explicitly said to be ‘worn on her back’ (負), it is reasonable to assume that she wore the second quiver on her side.

a (more) comparative approach to some japanese etymologies

61

and Whitman (2012). Vovin (2010) tackles the painstaking task of reviewing every comparison proposed by Whitman (1985) and rejects most of them as being probable borrowings or chance resemblances. A reexamination of every proposed etymology is far beyond the scope of this contribution,9 but I wish to offer some comparative perspectives from a Ryukyuanist point of view. One of the main methodological principles of Vovin (2010: 6) is that in order to be reconstructed in pj, an etymon has to be attested in oj or mj and Southern Ryukyuan, or in Ryukyuan and a non-Central Japanese dialect. From this follows that any Koreo-Japonic comparison without a (Southern) Ryukyuan cognate is likely to be a loan from Korean to Japanese.10 I will adduce below Ryukyuan comparanda for several comparisons rejected by Vovin.11 It might thus be possible to retrieve some of the Koreo-Japonic comparisons proposed by Whitman (1985), provided that they do not pose other problems than their lack of attestation in Ryukyuan. I will nevertheless remain agnostic as to their interpretation. In other words, my aim is to show that the following words need to be reconstructed at the pj level, but whether they constitute cognates, borrowings or chance similarities with Korean remains to be discussed. 3.1 Morning The comparison oj asa :: mk àchóm ‘morning’ (Whitman 1985: 244, #321) is known since at least Polivanov (1891–1938), but seems to lack Ryukyuan cognates at first sight (Vovin 2010: 224). Indeed, the usual word for ‘morning’ in virtually every Ryukyuan variety is not related to asa: Shodon sɨkaːma, Okazen ɕìtɨ̀múːtɨ̀, Yoron ɕíkámà, Nakijin ɕìtìmítì, Shuri sútúmítí, Ōgami stumuti, Ishigaki sɨ̀túmúdí, Yonaguni tˀùmútî. Still, a cognate of asa can be found in many compounds, which more often than not do not exist as such in Japanese and thus cannot be loanwords: Yamatohama ʔasa-çikjaɾi ‘glow in the morning sky’, ʔasa-sɨkama ‘early morning’, ʔasa-tʰeda ‘morning sun’, Shodon ʔasaː-jihɘː ‘glow in the morning sky’, Yoron àsá-ɕù ‘morning tide’, Izena ʔàsà-ùí ‘morning rain’, ʔásá-ùtàí ‘morning fatigue’,

9 10

11

See Whitman (2012) for a response on some etymologies. I will not discuss the potential problems of this approach in the case of a family like Japonic, where there are only two main branches, but accept it as valid for the sake of the discussion. I will restrict myself to etyma well attested in Ryukyuan and will not address intriguing but isolated attestations such as watʰa-ɡanaɕi ‘sea god’ (Yamatohama), which might be added to the comparison oj wata ‘sea’ :: mk pàtáh ~ pàlól ‘id.’ (Whitman, 1985: 213, Vovin, 2010: 111).

62

pellard

ʔásá-zúɾì ‘morning meeting’, Nakijin hàsàː-ní ‘morning sleep’, hàsàː-tɕíjù ‘morning dew’, Iejima ʔásá-úkì ‘early rising’, ʔásá-ɡánnjáì ‘morning thunder’, ʔásá-dúɾì ‘morning calm’, Shuri ʔásá-ɡúmúí ‘cloudy morning’, ʔásá-dátɕí ‘early morning’, Ōgami asa-munu ‘breakfast’, Ishigaki ásá-útsɨ́ ‘during the morning’, ásá-páná ‘early morning’, ásá-bɨ̀ɾɨ̀ ‘morning defecation’, Taketomi əsə-uzumi ‘morning rising’, əsə-tui ‘morning rooster’, əsə-niŋɡəi ‘morning prayer’, Yonaguni asadusaŋ ‘morning and evening’. It is thus possible to reconstruct pr *asa b, and pj *asa 2.5b ‘morning’ from the comparison with Japanese asa 2.5. 3.2 Body The comparison of oj mu- ~ mï ‘body’ with mk mwóm ‘id.’ (Whitman, 1985: 237, #259) is well known and is one of the core etymologies of Koreo-Japonic comparative studies. It is nevertheless rejected by Vovin (2010: 194) on the sole basis of its lack of attestation in Ryukyuan. However, as already noticed by Whitman (2012: 32), a cognate is attested in Ryukyuan as a pronoun. Whitman quotes the Nakijin form ʔàɡáːmì ‘we’, to which we can add Yoron mǐː ‘self’. Such examples parallel the grammaticalization of oj mï into Early Middle Japanese wagami ‘I, he, himself’ and mi ‘I’. Even clearer, non-grammaticalized cognates are also found, with the meaning ‘edible flesh of fish and other animals, especially the muscular tissue, edible part of seashells and sea urchins, human flesh’, ‘pulp, flesh of fruits, fruit’, and by extension ‘substantial, solid ingredient in a soup’: Yamatohama miː ‘fruit, flesh’, Yuwan miː ‘flesh of seashells’, Shodon mîː ‘fruit, flesh’, Okazen míː ‘fruit, flesh’, Izena míː ‘fruit, flesh’, Shuri mîː ‘fruit, content’, Ōgami miɯ ‘meat, flesh’, Ishigaki mɨ̂ː ‘flesh, fruit, ingredient’, Taketomi miː ‘ingredient, meat, body’, Yonaguni míː ‘flesh, ingredient’. The reconstruction of pr *mi a ‘body, flesh, meat’ and its comparison with Japanese mu- ~ mï 1.1 ‘body’ is rather straightforward. Though the two are traditionally thought to be distinct words, this is clearly the same word as oj mï ‘fruit, seed, kernel, pulp’, which also belongs to class 1.1. We thus need to reconstruct only one pj word here: *mui 1.1 ‘body, flesh, meat, soft tissue underneath the skin or shell of animals and fruits’. 3.3 Origin The comparison of oj mötö ‘base, bottom, root’ with mk míth ‘id.’ (Whitman 1985: 240, #278) is argued by Vovin (2010: 202) to be a Korean loan, in spite of its existence in Ryukyuan as a classifier.12 12

I rather take the existence of this classifier as evidence for the antiquity of this etymon in Ryukyuan and Japonic.

a (more) comparative approach to some japanese etymologies

63

It is nevertheless possible to find several varieties which have an independent cognate form: Yamatohama mutu ‘origin’, Yuwan mutu ‘root of vegetables’, Okazen mùːtú ‘origin’, Yoron mútù ‘origin, classifier for plants’, Izena mùːtú ‘origin’, Nakijin mútù ‘origin’, Iejima mútú ‘origin, root, trunk’, Shuri múːtú ‘origin’, Ikema mutu ‘origin’, Ōgami mutu ‘stem’, Ishigaki mútú ‘origin’, Taketomi mútú ‘origin, trunk’, Yonaguni mùtû ‘origin’. I thus reconstruct pr *moto c as a cognate of Japanese mötö 2.3?, and pj *mətə 2.3a ‘base, root, origin’. 3.4 Darkness The comparison of oj yamï ‘darkness’ with mk cyèmúl-, cyémkúl ‘(day) comes to a close, gets dark’ (Whitman 1985: 232, #199) is rejected by Vovin (2010: 169) because of the scarce attestation of this etymon in Ryukyuan and of other irregularities. However, regular cognates of oj yamï can be found in all branches of Ryukyuan, though sometimes in compounds only: Yamatohama jami, Okazen jàmíː Nakijin kˀùɾáː-jàmì, Shuri jámí, Ōgami ffa-jam, Ishigaki jámɨ̀ ‘darkness’, Yoron jujaŋ, Izena jàmì-ɡà-jǔː, Iejima jú-jánì ‘dark night’. I thus propose to reconstruct pr *jami a, and from the comparison with Japanese yamï 2.3, pj *jam{u,o}i 2.3b ‘darkness’. 3.5 Seaweed The comparison of oj mo ~ më ‘seaweed’ with mk mól ‘edible seaweed’ (Whitman 1985: 237, #253, Vovin 2010: 193) cannot be rejected as a late loan on the basis that this word has no cognate in Ryukyuan. This etymon is actually well attested in both Northern and Southern Ryukyuan: Yamatohama muː, Yuwan muː, Shodon môː Okazen múː, Izena móː, Nakijin móː, Iejima môː, Shuri mûː, Ōgami muː, Ishigaki môː.13 We can thus reconstruct a pr form *mo a, and from this a form *m{o,ə} 1.2 ‘seaweed’ at the pj level.14

Primary Sources Keizai zasshisha (ed.). 1915. Kokushi taikei 1 Nihon shoki. Tokyo: Keizai zasshisha. http:// kindai.ndl.go.jp/info:ndljp/pid/991091.

13 14

The vowel o found in some varieties is irregular, but the tonal correspondence is perfect. See also Antonov (this volume).

64

pellard

Kichō tosho fukusei kai (ed.). 1937. Ruiju myōgishō. Tokyo: Kichō tosho fukusei kai. http://dl.ndl.go.jp/info:ndljp/pid/2586899. Kurano, Kenji & Takeda, Yūkichi (eds.). 1958. Kojiki Norito. Tokyo: Iwanami shoten. Kyōto daigaku Bungakubu Kokugogaku Kokubungaku kenkyūshitsu (ed.). 1968. Shohon shūsei Wamyō ruijushō. Kyoto: Rinsen shoten. Kyōto Teikoku daigaku Bungakubu Kokugogaku Kokubungaku kenkyūshitsu (ed.). 1944. Shinsen jikyō. Osaka: Zenkoku shobō. http://dl.ndl.go.jp/info:ndljp/pid/ 1126450. Nakanishi, Susumu (ed.). 1978–1985. Man’yōshū zen’yakuchū genbun tsuki. Tokyo: Kōdansha.

Bibliography Buck, Carl Darling. 1949. A dictionary of selected synonyms in the principal IndoEuropean languages. Chicago: University of Chicago Press. Frellesvig, Bjarke & Whitman, John. 2004. The vowels of proto-Japanese. Japanese Language and Literature 38(2). 281–299. Frellesvig, Bjarke & Whitman, John. 2008a. Evidence for seven vowels in protoJapanese. In Frellesvig & Whitman (2008b), pp. 15–41. Frellesvig, Bjarke & Whitman, John (eds.). 2008b. Proto-Japanese: Issues and prospects. Amsterdam: John Benjamins. Hashimoto, Shinkichi. 1950. Kokugo on’in no kenkyū. Tokyo: Iwanami shoten. Hattori, Shirō. 1976. Ryūkyū hōgen to hondo hōgen. In Iha Fuyū tanjō hyakunen kinenkai (ed.). Okinawagaku no reimei, Tokyo: Okinawa bunka kyōkai, pp. 7–55. Izena-jima hōgen jiten henshū iinkai (ed.). 2004. Izena-jima hōgen jiten. Izena-son: Izena-son kyōiku iinkai. Jōdaigo jiten henshū iinkai (ed.). 1967. Jidaibetsu kokugo daijiten jōdai hen. Tokyo: Sanseidō. Kiku, Chiyo & Takahashi, Toshizō. 2005. Yoron hōgen jiten. Tokyo: Musashino shoin. Kokuritsu kokugo kenkyūjo (ed.). 1963. Okinawago jiten. Tokyo: Ōkurashō insatsukyoku. Maeara, Tōru. 2011. Taketomi hōgen jiten. Ishigaki: Nanzansha. Miyagi, Shin’yū. 2003. Ishigaki hōgen jiten. Naha: Okinawa taimusu. 2 vols. Miyanaga, Masamori. 1980–1981 (1930). Yaeyama goi (Miyanaga Masamori Zenshū 8). Tokyo: Daiichi shobō. 2 vols. Nakamura, Yukihiko, Okami, Masao & Sakakura, Atsuyoshi (eds.). 1982–1999. Kadokawa kogo daijiten. Tokyo: Kadokawa shoten. Nakasone, Seizen. 1983. Okinawa Nakijin hōgen jiten. Tokyo: Kadokawa shoten. Nevskij, Nikolai A. 2005 (1922–1928). Miyako hōgen nōto. Hirara: Okinawa-ken Hirarashi Kyōiku iinkai. 2 vols.

a (more) comparative approach to some japanese etymologies

65

Osada, Suma & Suyama, Nahoko. 1977–1980. Amami hōgen bunrui jiten. Tokyo: Kasama shoin. 2 vols. Oshio, Mutsuko. 1999. Okinawa Iejima hōgen jiten. Ie-son: Ie-son kyōiku iinkai. 2 vols. Pellard, Thomas. 2008. Proto-Japonic *e and *o in Eastern Old Japanese. Cahiers de Linguistique Asie Orientale 37(2). 133–158. Pellard, Thomas. 2010. Review of Bjarke Frellesvig and John Whitman (2008) ProtoJapanese: Issues and prospects. Cahiers de Linguistique Asie Orientale 39(1). 95–114. Pellard, Thomas. 2013. Ryukyuan perspectives on the Proto-Japonic vowel system. In Frellesvig, Bjarke & Sells, Peter (eds.), Japanese Korean Linguistics 20, Stanford: csli Publications, pp. 81–96. Shōgakkan kokugo jiten henshūbu (ed.). 2000–2002. Nihon kokugo daijiten, 2 edn. Tokyo: Shōgakkan. 15 vols. Tamura, Suzuko. 1996. Ainugo Saru hōgen jiten. Tokyo: Sōfūkan. Vovin, Alexander. 2009. A descriptive and comparative grammar of Western Old Japanese: 2, Adjectives, verbs, adverbs, conjunctions, particles, postpositions. Folkestone: Global Oriental. Vovin, Alexander. 2010. Koreo-Japonica: A re-evaluation of a common genetic origin. Honolulu: University of Hawai’i Press. Whitman, John. 1985. The phonological basis for the comparison of Japanese and Korean. Cambridge: Harvard University (Doctoral dissertation). Whitman, John. 1990. A rule of medial *-r- loss in pre-Old Japanese. In Baldi, Philip (ed.), Linguistic change and reconstruction methodology, Berlin: Mouton de Gruyter, pp. 511–546. Whitman, John. 1999. Personal pronoun shift in Japanese: A case study in lexical change and point of view. In Kamio, Akio & Takami, Ken-ichi (eds.), Function and structure, Amsterdam: John Benjamins, pp. 357–386. Whitman, John. 2008. The source of the bigrade conjugation and stem shape in pre-Old Japanese. In Frellesvig & Whitman (2008b), pp. 159–173. Whitman, John. 2012. The relationship between Japanese and Korean. In Tranter, Nicolas (ed.), The languages of Japan and Korea, New York: Routledge, pp. 24–38. Yanagida, Yuko & Whitman, John. 2009. Alignment and word order in Old Japanese. Journal of East Asian Studies 18. 101–144.

chapter 7

The Role of Internal Reconstruction in Comparing the Accent Systems of Korean Dialects S. Robert Ramsey

Around forty years ago, I wrote a description of an accenting dialect spoken in the village of Pukcheng1 in South Hamkyeng. (At the time, information about pitch distinctions found in North Korean speech had been completely absent in the literature.) With that synchronic description in hand, I then turned to the question of how Korean accent had developed historically. The task consisted of comparing the North Korean data from my research project with data from accenting dialects spoken in Kyengsang, and then comparing both to the “tones” recorded in the Middle Korean texts of the fifteenth and sixteenth centuries. The results of this study, a triangulation from those three points of reference (Middle Korean, Hamkyeng, Kyengsang), were first published in Ramsey 1974, and then in considerably more detail in Ramsey 1978. The task of reconstructing earlier stages of the Korean language depended upon that comparative foundation and was advanced from there principally through internal reconstruction. In this way, a number of us became convinced that the farther back in time we look, the fewer pitch and vowel length distinctions we find. Through this method, it was possible to see that the pitch accent (or tone) systems found in Middle Korean and the modern dialects were, for the most part, produced historically through syllable crasis and the elision of vowels through syncope and/or apocope. At the same time, those processes also produced the complex initial consonant clusters found in Middle Korean, as well as many of the heavily aspirated consonants still heard in Korean today. The proto-Korean system that emerged through this process was thus markedly different in structure from both Middle Korean and all modern dialects.

1 Throughout this paper Korean is romanized using the Yale system, except for personal names for which different spellings are preferred. Note, though, that Middle Korean spellings follow the modified Yale system found in Martin 1992 and Lee and Ramsey 2011. The phonetic values of syllable pitches are indicated (roughly) by transcribing low-pitched syllables with lowercase letters, and high-pitched syllables in CAPS.

© koninklijke brill nv, leiden, 2018 | doi: 10.1163/9789004351134_008

the role of internal reconstruction

67

Both the reasoning and the data underlying these conclusions can be found in many of my own publications (e.g., Ramsey 1975, 1978, 1986, 1991), as well as in Lee and Ramsey 2011. But the model was advanced in the work of others as well. I am thinking first of all of John Whitman’s 1994 paper in which he extended my original arguments regarding verb classes to account for certain accent classes found among Middle Korean nouns. Nevertheless, Whitman surmised that some of the accent distinctions he had examined appear to have been original and could not be accounted for through internal reconstruction. Perhaps as much as anyone, my mentor Samuel E. Martin (1992, 1996) worked to advance the paradigm and to sharpen the arguments. Fukui Rei has been a major contributor to the exploration of earlier Korean through internal reconstruction as well. In particular, Fukui’s summary of what has been done along these lines in his 2012 book on the history of Korean phonology is eminently readable and useful. Moreover, and still more recently, Ito Chiyuki (2013) is an exhaustive review and reworking of the theories. Both Fukui and Ito add new and compelling ideas about reconstructing earlier Korean, and both are fundamentally in agreement with me that proto-Korean had no pitch accent or tonal distinctions. But my purpose in writing the present essay is not to elaborate on these latter efforts at exploring the early phonological history of Korean. Rather, I wish to stress instead the importance of that first stage of reconstruction, the one arrived at through the comparative method applied to Middle Korean and the modern dialects, the work I began with all those decades ago. After all, the reconstructive work we’ve been engaged in since that time is based upon that initial foundation. From time to time a reaffirmation of the assumptions and conclusions we began with is called for, I think. Let’s summarize a few of the findings:

Pitch Accent in South Hamkyeng Anyone who listens to Pukcheng speech will discover right away that differences in pitch distinguish words. For example, the pronunciation of MAL i MANtha ‘(he) talks a lot’ differs from mal I MANtha ‘(he) has a lot of horses’ by the pitch of the first syllable. However, this distinction is not one of syllabic tone like Chinese, since the phonological difference between mal ‘speech’ and mal ‘horse’ can only be heard as a contrast with the pitch of a following syllable, in this case, an enclitic. And so, the pitch difference between these monosyllabic nouns is not apparent if they are pronounced in isolation.

68

ramsey

For words and phrases of two syllables or longer, the pitch distinction is the location of the most prominent syllable in the overall pitch pattern. This locus can be specified lexically by placing an accent mark on the syllable, and in nouns, the distribution is symmetrical, since an accent can occur on any one, or none, of a noun’s syllables. Thus, for two-syllable nouns there are three possible patterns:2 móki látwu ‘even though it’s a mosquito’ → MOki latwu melí látwu ‘even though it’s the head’ → meLI latwu poli látwu ‘even though it’s barley’ → poLI LAtwu Following an atonic noun such as poli ‘barley’, the pitch of a monosyllabic, enclitic particle is always high; e.g., poLI MAN ‘only barley’, poLI NUN ‘as for barley’. But any second enclitic particle added after that is always lowpitched: poLI MAN un ‘as for only barley’. A little experimentation with the morphology and phonology like this shows that all monosyllabic particles are tonic. Particles longer than one syllable are also tonic, with an accent on one of the particle’s syllables: poLI CHElem ‘like barley’; poLI MANKHUM ilatwu ‘even as much as barley’; poLI PWUTHE latwu ‘even starting from barley’. Thus the lexical accents of these particles are: chélem ‘like …’; mankhúm ‘as much as …’; pwuthé ‘starting from …’ As can be seen from this quick look, the Pukcheng accent system is symmetrical and relatively uncomplicated, with only one, consistent kind of suprasegmental. The system is strikingly similar in many ways to that of Tokyo Japanese. There are obvious differences, of course, especially phonetic; few Pukcheng accent patterns are flat, building instead in a crescendo to an accent locus, where the accented syllable can often be markedly higher in pitch than preceding syllables: hanccakNAtul ‘the one-sided visit (that is, a new bridegroom’s traditional first visit to his bride’s and in-laws’ home)’, and a phrase-final high pitch trails off rather than falling abruptly (as in poLI ‘Ta ‘it’s barley’). But basically, there is little that would seem radically different or strange to anyone familiar with Tokyo Japanese.

Pitch Accent in Kyengsang Kyengsang accent, on the other hand, is realized in a suprasegmental system that is considerably more complex. As is well known, Kyengsang accent comes 2 An accent cancels any on its right; all unaccented syllables to the left of the prevailing are H except for the first.

the role of internal reconstruction

69

in two main varieties, that of South Kyengsang (spoken in, say, Pusan) and that of North Kyengsang (heard, e.g., in Taykwu). Let us set aside the complexity that divides the two and describe the general system that characterizes both. First, in all Kyengsang suprasegmental systems (at least those that have been described so far) there are three different types of pitch patterns. One of these consists of patterns in which the first syllable is a reflex of what is called the ‘rising tone’ in Middle Korean; since this pattern is the one by which North Kyengsang and South Kyengsang differ from each other, we’ve chosen to set it aside for our purposes here. Of the two remaining pattern types, one consists of pitch patterns like those heard in Pukcheng. Here, in this familiar accent type, each word or phrase is characterized by a pitch pattern with a single prominent syllable, the location of which is the distinctive element. As in Pukcheng, it is the accent locus. Here are some Kyengsang examples in which nouns are combined with the particle mánkhum ‘as much as’ to form this type of pattern: méli mánkhum ‘as much as the head’ → MEli mankhum hánul mánkhum ‘as much as the sky’ → HAnul mankhum kkamákwu mánkhum ‘as much as a raven’ → kkaMAkwu mankhum kámulchi mánkhum ‘as much as a mullet’ → KAmulchi mankhum palam mánkhum ‘as much as the wind’ → paLAM MANkhum saytali mánkhum ‘as much as a ladder’ → sayTALI MANkhum What is immediately obvious about these representative nouns is the strange asymmetry in the distribution of lexical accent: In Kyengsang any syllable EXCEPT the last can be accented. A noun can also be unaccented, as was the case in Pukcheng. But in Kyengsang, a lexical accent never falls on the last syllable. And then there is the third type of pitch pattern, a totally different one consisting of a fixed tonal shape for the entire phrase. That is to say, no matter how long the word or phrase with such a pattern is, the first two syllables are always uniformly high in pitch, and every syllable that follows after that in the phrase is low. This decidedly odd pattern is associated with a morphophonemic accent occurring in front of the noun, which I’ve termed a ‘pre-accent’ (cf. Ramsey 1978: 74–78 for reasons why the pattern is the result of a pre-accent). Here, for example, are two such nouns with a lexical pre-accent: ′pholi ‘fly’; ′mwucikay ‘rainbow’. These nouns combine with enclitics to form pitch patterns as follows:

70

ramsey

PHOLI mankhum (HHLL) ‘as much as a fly’ MUCIkay mankhum (HHLLL) ‘as much as a rainbow’ MUCIkay man twu (HHLLL) ‘even a rainbow’

Comparing Hamkyeng and Kyengsang Here are some representative lexical correspondences:

Hamkyeng

Kyengsang

‘water’ ‘horse’

múl mal

′mul mal

‘mosquito’ ‘head’ ‘barley’

móki melí poli

′mokwu méli poli

‘rainbow’ ‘mullet’ ‘raven’ ‘ladder’

múcikay kamúlchi kamakwí saytali

′mucikay kámulchi kkamákwu saytali

For the most part, the correspondences are straightforward and regular. (That is, except for three-syllable nouns where Kyengsang has a pre-accent; in these cases the Hamkyeng correspondences vary widely.) In 1974, and again in 1978, I argued that the correspondences are to be explained historically by the fact that in Kyengsang, the accent locus shifted one syllable to the left. In cases where an accent had, before the change, occurred on the first syllable, the shift left moved it to the front of the word, creating a morphophonemic distinction, a ‘pre-accent’, and its associated word tone.

Middle Korean Confirmation This analysis was derived from the morphophonemics of the Kyengsang system of course, but it was also strongly supported by the data from Middle Korean. For there, in the MK suprasegmental system, the phonemic distinction in each

the role of internal reconstruction

71

word or phrase was demonstrably the locus of the first high pitch, and this locus corresponded regularly with the accent locus of the word’s South Hamkyeng reflex. It seemed clear that the historical innovation separating Hamkyeng and Kyengsang had taken place in Kyengsang. Additional support for this solution came in 1993 when Umeda Hiroyuki published his findings on the phonology of the Korean dialect spoken in Yanbian in northeastern China. Since the Koreans in those communities are largely the descendants of 19th-century immigrants from Hamkyeng, Umeda’s research confirmed my Hamkyeng research findings from twenty years earlier and added further useful data.

A Reconsideration and an Alternative Hypothesis Such has been the situation in Korean studies. And, for the most part, the historical explanation that an accent shift had taken place in Kyengsang has remained unquestioned. But in 2010, in a paper read at the International Symposium on Accent and Tone (sponsored by the National Institute for Japanese Language and Linguistics), Lee Yeonju (of Hokkaidō University) argued that the accent shift had not been in the Kyengsang dialects. Rather, she said, the historical innovation separating the Korean dialects had occurred instead in the form of a right-shift in both the Hamkyeng dialects and Middle Korean. This alternative interpretation of history was then taken up by Uwano Zendō at the end of an article he had written on Japanese accent and published in Lingua in 2012. In the article Uwano offered no linguistic evidence from Korean to support Lee’s hypothesis, nor, at that point, had he intended to. What he was suggesting instead was a theoretical position, his belief that in accenting languages there is a general tendency toward a right-shift of an accent locus, just as there is said to be a typological tendency in tone languages for a high tone to be advanced rather than retracted. (See Hyman 2003.) In other words, Uwano was talking mostly about where he would like to see future research on Korean accent go and what he expects it might show. It’s certainly a fair position to take, but the linguistic evidence does not follow those expectations. Uwano’s expectations about the direction of change were raised, in part, by two extralinguistic facts. First, as he points out, ‘a large number of people migrated from [Kyengsang] to [Hamkyeng] in the [Cosen] period.’ (That was certainly true, especially during King Secong’s reign in the middle of the 15th century, but this population movement says nothing about the direction of change; in fact, we know full well that immigrant communities can often

72

ramsey

be unusually conservative.) Second, tone marks disappeared from Korean literature at the beginning of the 17th century, and that means (at least most probably) that tone (or accent) distinctions were lost in the central dialects in less than 200 years from their first appearance in the textual corpus. And since the Kyengsang dialects retain a ‘robust accent system’ today, Uwano concludes that ‘MK is linguistically “younger” than modern [Kyengsang] dialects in terms of its accent system’ (2012). These tentative suggestions of Uwano’s are interesting to think about, but we have yet to find linguistic evidence to support them. (It goes without saying that any conclusions about Korean have to be based completely on linguistic data from Korean.) The question, as I understand it, boils down to this: When two related languages or dialects are separated by a regular structural difference, how can we know where the innovation occurred? If only two linguistic systems are in play, the answer cannot be determined by the comparative method alone. That’s because the change separating the two could have occurred in either system: one or the other. The answer is fundamentally indeterminate. But when there are three independent systems being compared, and one differs from the other two, it is usual in comparative practice to conclude that the odd one is where the innovation occurred. This ensures that the fewest hypotheses be entertained simultaneously. (Moreover, the agreement of Umeda’s Yanbian data with Hamkyeng and MK, and the contrast of all three with Kyengsang, make the disparity even greater: three to one.) Uwano might counter by saying that conclusions drawn from probability are more than outweighed by the ‘tendency’ of accent systems to undergo a right-shift. From the theoretical position he has staked out then, he could claim that it is perfectly reasonable to think that a right-shift would occur independently in Hamkyeng and in Middle Korean and, perhaps, in Yanbian as well. That’s because, according to his theoretical framework, all accent systems have a latent propensity to shift the accent locus right, so the shift could happen again, and again, and again, each time independently. (Such thinking is strongly reminiscent of Sapir’s theory of ‘Drift’.) Of course, by no means do I agree with this theoretical view of things, either about ‘Drift’ or about the ‘tendency’ toward a right shift. But I do agree that statistics alone do not produce a definitive conclusion. Rather, all they can produce is a likelihood that increases with how many independent comparisons are in play. Odds of two to one—or even three to one—still leave ample room for doubt.

the role of internal reconstruction

73

The Value of Internal Reconstruction Where, then, can a definitive answer to the original question be found? It is my belief that the best way to find a solution is by using the basic methodology of internal reconstruction. Here’s why. First of all, an old, elementary principle of internal reconstruction is that sound change in a language typically leaves behind irregularities. And then, as Robert Austerlitz once put it, ‘internal reconstruction is making those irregularities regular.’ In other words, by applying internal reconstruction to irregularities left by change the researcher is able to reconstruct what the language was like before the change. And so, in this particular case, we can see from the basic irregularities now found in the Kyengsang accent system that a change took place there. (In contrast, as we have seen, the Hamkyeng system is regular and relatively simple.) The most glaring of the Kyengsang irregularities is the coexistence of two types of suprasegmentals in the system. One is what is normally termed an accent—that is, an element whose essence is the locus of a prominent, high pitch followed by a pitch fall. The other suprasegmental pattern has a completely different phonological realization, namely, the word tone (or phrase tone) I mentioned earlier that is associated with a morphophonemic ‘preaccent’. It is important to remember that this pre-accent is an accent only in a morphophonemic sense. What the morphophonemics do is produce a pitch fall of the kind associated with an accent locus, but the locus of the accent is found in front of the first mora of the lexical item, on the last syllable of a preceding element. It is in this sense that a word can have a lexical ‘pre-accent’. We can see how such pre-accents behave, for example, in noun compounding. To illustrate, here is a typical example involving the atonic noun kasil ‘autumn’ and the pre-accented noun ′mokwu ‘mosquito’. First, note that the two nouns have completely different kinds of suprasegmentals in isolation or when they occur in a phrase with a particle such as mánkhum ‘as much as’: kasil + mánkhum → kaSIL MANkhum ‘as much as autumn’ [i.e., a typical accent pattern] ′mokwu + mánkhum → MOKWU mankhum ‘as much as a mosquito’ [an unvarying tone] Now, when the two nouns are combined to form a compound, a pitch pattern emerges with an accent that is now located in front of the pre-accented noun, on the last syllable of the atonic noun:

74

ramsey

kasil + ′mokwu → kasíl-mokwu [kaSIL-mokwu] ‘an autumn mosquito’ In other words, the morphophonemics of compound formation transform one kind of suprasegmental into another. This same type of transformation occurs in other processes as well. But nothing resembling this structural oddity is found in either MK or Hamkyeng. The point here is that this ‘pre-accent’ and the morphophonemics of the Kyengsang suprasegmental system is a structural irregularity of just the kind we expect to see in the wake of a historical change in the phonological system. Consider also what this irregularity means in more practical terms. What if the Kyengsang system had been original and the MK and Hamkyeng systems derived from it? What kinds of rules would have to be written? First, the morphophonemic ‘pre-accent’ would have to be converted into a normal, phonemic accent and moved onto the first syllable of a word, and then a blended, balanced accent system somehow created when the two types of suprasegmentals were combined. An equally obvious irregularity is the overall distribution of lexical accents. Accents do not occur on the last syllables of lexical items in Kyengsang. Why would that be the case? It seems logical to assume that they once did, but that they were moved (or lost) through historical change. This structural asymmetry adds to the ample evidence that the accent locus has moved onto the previous syllable. Making such ‘irregularities regular’ in this way is another classic application of internal reconstruction. Now let us look once again at the dialect correspondences, this time together with their MK reflexes:

Hamkyeng

Kyengsang

Middle Korean

‘water’ ‘horse’

múl mal

′mul mal

múl mol

‘mosquito’ ‘head’ ‘barley’

móki melí poli

′mokwu méli poli

mwókoy melí pwoli

‘rainbow’ ‘mullet’ ‘raven’ ‘ladder’

múcikay kamúlchi kamakwí saytali

′mucikay kámulchi kkamákwu saytali

múcikay kamwóthi kamakwóy –

the role of internal reconstruction

75

A glance at these correspondences shows that Kyengsang is the odd man out here. Statistics do not conflict with the findings of internal reconstruction but rather agree with them.

References Fukui, Rei. 2012. Kankokugo on’inshi no tankyū [‘An investigation of Korean phonological history’]. Tokyo: Sanseidō. Hyman, Larry 2002. ‘Is there a right-to-left bias in vowel harmony?’ Paper presented at the 9th International Phonology Meeting, Vienna. Itō, Chiyuki 2013. ‘Korean accent: internal reconstruction and historical development’. Korean Linguistics 15.2, pp. 129–198. Lee, Ki-Moon and S. Robert Ramsey. 2011. A History of Korean. Cambridge and New York: Cambridge University Press. Martin, Samuel E. 1992. A Reference Grammar of Korean. Rutland, Vermont and Tokyo: Charles E. Tuttle Company. Martin, Samuel E. 1996. Consonant Lenition in Korean and the Macro-Altaic question. Honolulu: University of Hawai‘i Press. Ramsey, S. Robert. 1974. ‘Hamkyeng kwa Kyengsang yang-pangen uy aykseynthu yenkwu [‘A study of the accent in the two dialects, Hamkyeng and Kyengsang’]’. Journal of Korean Linguistics 2, pp. 105–132. Ramsey, S. Robert. 1977. Velar lenition in Korean. In Festschrift Commemorating the Seventieth Birthday of Doctor Lee Sungnyeong, pp. 125–132, Seoul: Tower Press. Ramsey, S. Robert. 1978. Accent and Morphology in Korean Dialects. Seoul: Tower Press. Ramsey, S. Robert. 1986. ‘The inflecting stems of Proto-Korean’. Ehak yenkwu 22-2, pp. 183–194. Ramsey, S. Robert. 1991. ‘Proto-Korean and the origin of Korean accent’. In William Boltz and Michael Shapiro (Eds.), Studies in the Historical Phonology of Asian languages, pp. 215–238, Amsterdam: John Benjamins. Ramsey, S. Robert. 2001. ‘Tonogenesis in Korean’. In Shigeki Kaji (Ed.), Cross-linguistic Studies of Tonal Phenomena, pp. 3–17, Tokyo: ilcaa, Tokyo University of Foreign Studies. Umeda, Hiroyuki. 1993. ‘Enpen chōsengo no on’in’ [‘A phonological study of Yanbian Korean’]. In Studies of Linguistics and Cultural Contacts 6, pp. 131–145, Tokyo: ilcaa, Tokyo University of Foreign Studies. Uwano, Zendo. 2012. ‘Three types of accent kernels in Japanese’. Lingua 122, pp. 1415– 1440. Whitman, John. 1994. ‘The accentuation of nominal stems in Proto-Korean’. In YoungKey Kim-Renaud (ed.), Theoretical Issues in Korean linguistics, pp. 425–439, Stanford: csli publications.

chapter 8

How Many oj Syllables are Reflected in emj yo? J. Marshall Unger

Iwo Jima, best known for the eponymous battle fought there in World War ii, is also a linguistically noteworthy name. A study of its etymology shows us something important about Old Japanese phonology. The kanji corresponding to ⟨Iwo Jima⟩ are 硫黄島, implying ‘sulfur island’. But the Japanese word for ‘sulfur’ is ioo. Why, then, the spelling ⟨iwo⟩? The modern word ioo ‘sulfur’ is the result of regular Middle Japanese sounds changes: [iwau] > [iwɔɔ] = ⟨iuǒ⟩ (in the Portuguese Dictionary of 1603–1604) > [ioo], in Edo period vernacular by the early 1800s. Today’s ⟨iwo⟩—omission of the diacritic is just a bad habit left over from the cold-type era—thus goes back as far as the 16th century.1 But where did iwau come from? The usual emj word for ‘sulfur’ was yuwau as shown by the riddle in the dictionary Myōgoki (1275) 火つくるゆわう如何、答油黄也

‘How do you bind [ yuwau] tinder?2 With sulfur [ yuwau].’ and an entry in the 20-fascicle version of the earlier Wamyōshō (931–938), which reads 流黄本草云石流黄焚石液也〈和名由乃阿和俗云 由王〉

‘Sulfur … Japanese name yu no awa, vulgarly yuwau’.

1 The name reflected in Iwo Jima must have originated after 1670, before which the Bonin Islands, though sighted by Europeans in the 16th century, were unknown to the Japanese. The name Sulphur Island was already known to participants of the Cook’s third voyage of 1779. See Welsch 2004 for a thorough discussion of the complicated details. 2 In the late 18th-century, a sulfur match was called tukegi in the East but yuoo in the capital area. The punned verb is emj yuwau < oj yupapu, adnominal of yup-ap- ‘bind + aux’.

© koninklijke brill nv, leiden, 2018 | doi: 10.1163/9789004351134_009

how many oj syllables are reflected in emj yo?

77

This passage is also the basis for the two leading etymological theories about yuwau: (1) it arose from the phrase yu-awa ‘hot-water foam’;3 or (2) it is an exceptional development of the go’on compound ru-wau (硫黄 or 流黄).4 Sino-Korean l- before i or y has become y (as in Korean ywuhwang < lywu.hwang ‘sulfur’ 硫黄).5 Etymology (2) assumes that something similar happened in Japanese. However, the compounds 硫黄 or 流黄 would both be ruwau in the go’on reading and riukwau in the kan’on. The wau part of yuwau, if SinoJapanese, could only have been go’on, but only the kan’on of 硫 or 流 contained an i that could plausibly have triggered a Korean-like change of r to y. Moreover, riu → iu would have regularly developed into yuu, not yu. All this, plus the fact that 硫黄 and 流黄 both occur for yuwau, strongly suggests that they are both just ateji (falsely attributed characters), compounds acceptable as literary Chinese but only tenuously related in form to the Japanese word they are used to transcribe. Hence (2) is unsatisfactory. That leaves (1). Now 由王 is clearly a phonographic use of characters meant to transcribe yuwau, but the Wamyōshō says it is a vulgarism, so the question becomes, how could yuwau have arisen from yu no awa? The link is the word yuwa recorded in the 10th-century Ametsuchi no kotoba, which most authorities think was a noun meaning ‘sulfur’.6 But which came first, yuwa or yuwau? If yuwau had been the original native word, its false association with 硫黄 and 流 黄 would have strongly militated against it being truncated to yuwa. But if yuwa had been a jukujikun (a multi-character gloss) on 硫黄 or 流黄, then hypercorrect reading pronunciations of 硫黄 and 流黄 would have readily resulted in an extra final u.7 We may therefore safely posit the replacement yuwa → yuwau. Given this, iwau is not hard to explain. It is a case of yu → *yi > i in allegro speech (n.b. there is no phonemic contrast i ≠ yi). In other instances of sporadic yu ~ i alternation, there was no phonetic motivation for i to break into yu, either in the ancient yuk- ~ ik- ‘go’, yum- ~ im- ‘abstain from’, and yume ~ ime ‘dream’8

3 Or perhaps originally ‘steam foam’? That would perhaps make the idiom yu/*mizu o wakasu ‘boil water’ more understandable. 4 Go’on and kan’on are the names of the two principal strata of Sino-Japanese borrowings into Old and Early Middle Japanese. 5 Neither j r- nor k l- begin free native words. 6 See Frellesvig 2010:165–167. 7 A current American example of a similar literacy-driven phoneme addition is the growing popularity of [ˈɔftən] often despite older [ˈɔfən] often and still widespread [ˈsɔfən] soften. 8 The idea that ime is a compound of i ‘sleep’ + me ‘eye’ assumes i < *ni < *ne ‘sleep’ (Whitman 1985), which is not certain (Unger 2014). Ratté (p.c.) proposes k ca- < mk co- ‘sleep’ ~ cum(G)u: oj yu- < pKJ *jo- and mk nwuW- ‘lie down’ : oj ne- < *nwe- < *nuwe- < pKJ *nwube-.

78

unger

or in the modern kayui ~ kaii ‘itchy’.9 The replacement i → yu does occur, but only in analogical back-formations (e.g. zipu ‘ten’ > ziu > zyuu ~ zip-pon ‘10 sticks’ → zyuppon). Now if, as Wamyōshō implies, yuwa < *yu-awa, the initial vowel of awa was evidently lost when the genitive marker no was omitted. However, going back to Hashimoto Shinkichi, it is often asserted that there were two phonemically distinct oj syllables corresponding to emj yo, which Martin (1987), for instance, transcribes as yo ≠ ywo. Furthermore, no oj words contain the sequence /uwa/: parts of oj words for which we would reconstruct *Cuwa are reflected in oj Cwo.10 That is, when adjacent vowels in oj lexicalizations were eliminated by rules of crasis (sometimes called contraction), we have *Cu(w)a > Cwo for typical c.11 Therefore, if ywo, as distinct from yo, had been available in Old Japanese and had crasis applied, the lexicalization of *yu-awa should have produced ywowa, yielding emj yowa. But reflexes of yowa do not appear in any of the many dialectal forms listed under yuoo and ioo ‘sulfur’ in the nkd.12 Moreover, if 硫黄 were ateji for yowa(u), which is theoretically possible (since ateji are arbitrary by definition), the alternate iwa(u) would be hard to explain.13 We are thus forced to conclude that the crasis rule *Cu(w)a > Cwo did not apply in 9

10

11

12

13

In these words, yu > i is evidently the result of allegro speech, especially in absolute initial position. Note that /yu/ in zyu and tyu can also be affected: [ʃin̩ ʤikɯ] for [ʃin̩ ʤɯkɯ] /sinzyuku/ ‘Shinjuku’ (esp. Tōkyō), and [ʃiʤɯʦɯ] for [ʃɯʤɯʦɯ] /syuzyutu/ ‘surgery; operation’, which occasionally is heard as [ʃɯzɯʦɯ] and even [ʃiiʦɯ]. The compounds /syuzyutu-situ/ ‘operating room’ and /syuzyutu-tyuu/ ‘while operating’ are sometimes cited as tongue-twisters. As suggested by the pronunciations reflected in the European spellings jujitsu etc. of /zyuuzyutu/, even non-Japanese have a hard time pronouncing words of this kind. Unger 1975. oj suwe- ‘seat’ < *suwa-gi- ~ swo-pe- ‘accompany’ < *suwa-(a)pa-gi-; kuwe- ‘kick’ < *kuwa-gi- ~ kwo-ye- id. < *kuwa-ye-; and even uwe- ‘hunger’ < *uwa-gi- ~ wo-ye- ‘starve’ < *uwa-ye- ~ wo-ya-s- ‘weaken’ < *uwa-ya-se-. There is a theory that pJ *o raised to oj u except in final position, where it became wo. This suggests that *wo could have been an intermediate stage for non-final *o, which might have merged with some *wo produced by crasis from *u(w)a. However, because the oending syllables that alternate with *uwa mentioned in n. 10 all appear in 8th-century texts, yuwa cannot be explained as a result of this raising without claiming ad hoc that *yu-awa was lexicalized long before the other words cited in n. 10. The form yoo is uncommon and readily accounted for as an allegro articulation of ioo. The distinction -iyV ≠ -iV is not easy to hear; sometimes, a false y is inserted (e.g. mi-age > miyage ‘gift, souvenir’). Even #iyV ≠ #yV can be blurry (e.g. iya = ya ‘no!’). The isolated alternation yo- ~ i- ‘good’ occurs only before the adnominal/conclusive ending -i; all other paradigmatic forms are built on yo- only. Note also that e- is sometimes heard for this i-.

how many oj syllables are reflected in emj yo?

79

the case of c = y; the only way to eliminate adjacent vowels in strings of the form *yV1V2, even in cases of lexicalization, was evidently simple vowel deletion. Since yu is monosyllabic and awa is not, *yu-awa > yuwa was forced. Thus, a proper understanding of the Japanese word for sulfur constitutes indirect but compelling evidence for rejecting the alleged phonemic oj distinction yo ≠ ywo. Lange (1973) pointed out that, on distributional grounds alone, the existence of this distinction in 8th-century poems of the Man’yōshū was not justified. Perhaps it is time the rest of the field caught up with him.

References Frellesvig, Bjarke. 2010. A History of the Japanese Language. Cambridge: Cambridge University Press. Lange, Roland A. 1973. The Phonology of Eighth-Century Japanese: A Reconstruction based upon Written Records. Sophia University Press. Martin, Samuel E. 1987. The Japanese Language through Time. New Haven: Yale University Press. nkd. 1972–1976. Nihon kokugo daijiten. Tōkyō: Shōgakukan. 2nd ed. 2000–2002 (available on-line to subscribers). Unger, J. Marshall. 1975. On the kō-type o-ending syllables of Old Japanese. Journal of the Association of Teachers of Japanese, vol. 10 nos. 2–3 pp. 201–207. Unger, J. Marshall. 2014. Old Japanese bigrade paradigms and Korean passives and causatives. In Martine Robbeets and Walter Bisang (eds.), Paradigm Change in the Transeurasion Languages and Beyond. John Benjamins Publishing, pp. 177–196. Welsch, Bernhard. 2004. Was Marcus Island discovered by Bernardo de la Torre in 1543? Journal of Pacific History, vol. 39 no. 1 pp. 109–122. Whitman, John B. 1985. The phonological basis for the comparison of Japanese and Korean. Ph.D. dissertation, Harvard University.

chapter 9

On the Etymology of the Name of Mt. Fuji Alexander Vovin

Mt. Fuji (Fujisan, 富士山), the highest mountain in Japan (3,776 m.), straddles the border of modern Yamanashi and Shizuoka prefectures, but in Ancient Japan it belonged to Suruga province (駿河國), which corresponds to the central part of present-day Shizuoka prefecture. This majestically beautiful mountain, which is a perfectly shaped conical volcano, has become a symbol of modern Japan.

Mt. Fuji as seen from Hakone author’s photo (may 27, 2012)

Nowadays we know many things about Mt. Fuji, such as the history of its worship as the body of a Shintō deity, its historical eruptions, its vegetation and natural habitat, the history of its climbing, etc. For the purposes of this article, the most important information is that Mt. Fuji is nowadays a dormant volcano

© koninklijke brill nv, leiden, 2018 | doi: 10.1163/9789004351134_010

on the etymology of the name of mt. fuji

81

(MdJ kyūkazan, 休火山). The last historically recorded eruption occurred in 1707 ad, but before this date, it was quite active. The first historically attested eruption was on the sixth day of the seventh lunar month of the first year of Ten’ō (天應) (July 31, 781ad). It is recorded in the Shoku Nihongi that on this day Mt. Fuji rained ash, and where the ash had fallen, all leaves withered (sng 36.204).1 Between 781 and 1707ad Mt. Fuji erupted at least ten times (or, this is what historical sources tell us). So the last four hundred years have been very peaceful, but Mt. Fuji is far from being dead, and it can awake any time. What we do not know, however, is the etymology of its name, and it appears that it was already obscure in the early Heian period. This gave rise to endless folk etymologies through the subsequent ages. The first of these folk etymologies can be found in the closing lines of the late ninth or early tenth century text Taketori monogatari (竹取物語): かのたてまつるふしのくすりに又つほくして御つかひににたまはす 勅使にはつきのいはかさといふ人をめしてするかの國にあなる山の いただきにもてつくへきよし仰給みねにてすへきやうをしへさせ給 御ふみふしの薬つほならへて火をつけてもやすへきおほせ給そのよ しうけたまはりてつはものともあまたくして山へのぼりけるよりま むそのやまをふしのやまとは名つけけるそのけふりいまだ雲のなか へたちのほるとそいひつたへたたる

[The emperor] placed the immortality elixir that she gave him in a jar, and gave [it] to the imperial messenger. [He] summoned a man called Tuki-no Ifakasa as an imperial messenger. [The emperor] gave the command that [Tuki-no Ifakasa] must bring [this jar] to the summit of a mountain that, as they say, was in Suruga province. [The emperor] explained [to Tuki-no Ifakasa] what [he] must do at the summit. [Namely, the emperor] gave the command that the immortality elixir and the letter must be put on fire and burnt together. Upon hearing that command, [Tuki-no Ifakasa], leading many soldiers, ascended to the mountain. From this time on, [the people] called that mountain the Mountain of Immortality. [The people] continue to tell a story that the smoke [from that fire] is still rising to the clouds2 tm 66.15–67.5

1 The Shoku Nihongi text is cited according to Aoki Kazuo, Inaoka Kōji, Sasayama Haruo, and Shirafuji Noriyuki’s edition (1991–1998). 2 The Taketori monogatari text is cited according to Sakakura Atsuyoshi’s edition (1957: 3–78).

82

vovin

This folk etymology suggests ‘immortality’ (oj pusi, 不死), but as we will see below, the oj name of the mountain is clearly punzi, not pusi. In addition, pusi (不死) would be a placename based on Sino-Japanese, which seems to be unwarranted for Ancient Japan. Quite a host of such Sino-Japanese folk etymologies, such as ‘having no equal (lit. no second)’ (oj punzi, 不二), ‘happiness and benevolence’ (oj punzi, 福慈), and ‘not ending’ (oj punzi, 不盡), appeared over the ages, but they should not entertain us here any longer. Lack of agreement between these different proposals, their reliance on Sino-Japanese, as well as the absence of any evidence presented all point to their arbitrary and ad hoc nature. John Batchelor, the great pioneer of Ainu Studies, proposed an etymology that Mt. Fuji (Fuji no yama in his reading) represents a hybrid of Ainu huchi, fuchi, fuji ‘old woman’ and Ainu unji, fuchi, huchi ‘fire’ on the one hand and Japanese yama ‘mountain’ on the other, with an ultimate explanation that Fuji means ‘Goddess of Fire’ (1928: 57). In spite of its possible superficial attractiveness, this etymology is seriously flawed for several reasons. First, Ainu fuchi, fuji ‘old woman’ and Ainu fuchi, huchi ‘fire’ are lexical ghosts, not attested anywhere except in Batchelor’s own dictionary (1938: 171). Even there, the f- initial forms are not present. Second, Ainu h, unlike MdJ h does not go back to *p (Vovin 1993: 25–27). Third, Ainu huci ‘grandmother’ and Ainu unci ‘fire’ are different words: the first one has initial h-, the second one does not, the first one does not have -n-, and the second does. I have reconstructed before pA *gurti ‘grandmother’ and pA *unti ‘fire’ on the basis of their reflexes in various Ainu languages and dialects (Vovin 1993: 90, 151), but nowadays I would correct the first reconstruction as *hu(r)ti ‘grandmother’. Neither of these two words is phonetically compatible with oj punzi; therefore this etymology, which appears nowadays frequently cited on the internet, must be also abandoned.3 Before trying to establish the etymology of Mt. Fuji, one must trace its phonological and early philological history. Martin reconstructed the pJ form as *punti 2.3 (1987: 420). While his accentual reconstruction, to which I will return later after the philological history, might be correct, his segmental reconstruction as *punti rather than *punsi is mistaken. This is supported by the fact that all oj phonographic spellings provided below indicate only oj punzi, but never *pundi. These oj phonographic spellings are given in Table 9.1 below:

3 In addition, pA *unti ‘fire’ is a regional word; it is attested only in Sakhalin dialects and in the Soya dialect in the extreme north of Hokkaidō.

83

on the etymology of the name of mt. fuji table 9.1

Spellings of oj punzi ‘Mt. Fuji’

Character Transcripspelling tion 福慈 富士 不自 不盡 布仕 布士 布自 布時

punzi punzi punzi punzi punzi punzi punzi punzi

Textual sources

Statistics

hf 38.8, hf 40.4–54 sfib 447.1, sfib 448.5, sng 36.204 mys 14.3355 mys 3.317–320, mys 11.2695, mys 11.2697a, mys 14.3356, nsk 24.2055 mys 11.2697 mys 3.317, mys 3.321 mys 14.3358, mys 14.3358b mys 14.3357

2 3 1 8 1 2 2 1

So, there are altogether twenty cases of phonographic spellings in oj texts that are represented by eight graphic variants. The most frequent of these variants is 不盡, occurring eight times. The next frequent is 富士, occurring three times, which also happens to be used in the modern spelling. What is important is that all of these spelling variants indicate punzi, and there is no evidence for *pundi or *pusi, since all man’yōgana signs used for the second syllable: 慈, 士, 自, 盡, 仕, 士, 自, and 時 are used to write the syllable /nzi/, never the syllable /ndi/, and only 時 can be used sometimes for the syllable /si/. Thus, we can firmly establish oj punzi, and not *pundi or *pusi. Consequently, contrary to Martin (1987: 420), we should reconstruct pj *punsi, not *punti. To be more accurate, since this place name is not attested in Ryūkyūan, our reconstruction is protoJapanese (pJN), and not pJ. oj punzi ‘Mt. Fuji’ is attested in both eoj and woj. The only eoj text it appears in is mys 14.3358b, while the remaining attestations are either in woj or in SinoJapanese texts. Nevertheless, it is important that this place name appears in regional woj poems from Azuma. As I mentioned above, Martin reconstructs pJN accent of *punsi as 2.3 (1987: 420). He is probably right, but there are problems with this accentual reconstruction. First, the Tokyo accent is hl, and this reflects 2.4 (lh) or 2.5

4 The Fudoki texts are cited according to Akimoto Kichirō’s edition (1958). 5 The Nihonshoki text is cited according to Kuroita Katsumi and Matsuyama Jirō’s edition (1965–1966).

84 table 9.2

vovin Accentual history of Fuji6

Segmental form

Accent

Textual source of accentual marks

Date

funzi funzi funzi fuzi fuzi fuzi fuzi fuzi

ll ll ll hl hl hl hl hl

various kkwks manuscripts scs Jōben-bon of siwks cmjf hkmb gk wts modern Kyoto

late 11th c.–1333ad 1185–1190 ad 1185–1333 ad end of 17th c. 1776 ad mid-Edo period first half of the 18th c. 20th c.

(lf), but not 2.3 (ll). The Kagoshima accent is even more puzzling: hl, which can reflect only 2.1 (hh) or 2.2 (hl). The modern Kyoto accent is hl, which can reflect either 2.3 (ll) or 2.2 (hl). Since the only overlap of accent patterns is between Kyoto and Kagoshima for 2.2, one might think that we should reconstruct pj *punsi as 2.2 (hl). But such a mechanical reconstruction ignores both the history of the language and geographic realities. First we must explore the historical records of the accent for oj punzi > mj funzi > MdJ fuzi. Table 9.2 above traces this accentual history. One can see from this table that all early attestations in late Heian and Kamakura periods have ll accent, while all late attestations starting from Edo period have hl accent. This is to be expected, because it is well known that the merger between 2.2 (hl) and 2.3 (ll) accent classes as a single 2.2 (hl) class did not happen in the Kyoto dialect before the fifteenth century. Therefore, the hl accent in modern Kyoto cannot reflect the original Kyoto 2.2 (hl) class, but must be a reflex of 2.3 (ll) before it merged with 2.2 (hl). As far as we rely on the historical data and the modern Kyoto dialect alone, Martin’s reconstruction of pJN accent as 2.3 is justified. But how do we explain the discrepancies with accentuation in Tokyo and Kagoshima? The difference with Kagoshima might seem quite bothersome, because Kagoshima hl clearly points not to a low, but to a high register. This is where the geographical and historical realities should be taken into serious considera-

6 All the data in the right three columns in this table are from Akinaga, Kazue; Ueno, Kazuaki; Sakamoto, Kiyoe; Satō, Eisaku; Suzuki, Yutaka (1997: 432).

on the etymology of the name of mt. fuji

85

tion. Mt. Fuji is quite close to Tokyo, somewhat far from Kyoto, but very far from Kagoshima. How many people from the uneducated classes in Kagoshima had ever seen Mt. Fuji let alone even known of its existence prior to the twentieth century? In sum, this place name in Kagoshima is unlikely to be native, going back to pJN. On the contrary, it is more likely to be a loanword from the Kyoto dialect after the accentual shift ll > hl had occurred in the latter. Thus, the Kagoshima accent can be excluded from further consideration. As far as Tokyo is concerned, we still have the same low register, because 2.3 (ll), 2.4 (lh), and 2.5 (lf) classes are all low-initial. For the purposes of this etymology, the low register is all that really matters as the reader will see below shortly. However, once the etymology has been established, I will return again to the problem of reconstructing the pJN accent for *punsi. After all these preliminaries, I am finally ready to offer my own etymology for Mt. Fuji. As I have briefly mentioned above, this etymology was already obscure in the Heian period. It is quite possible that the same was already true in the Nara period: the complete absence of logographic spellings seems to offer at least partial circumstantial evidence for this point of view. Nevertheless, I am going to demonstrate below that this place name has a perfect Japanese etymology, and that the reason that it became obscure so early is that the language that underlies is not the woj of Yamatö, but eoj of Azuma. Let me start from the second syllable nzi of punzi. I believe that it can be explained as -nzi, a contraction of oj nusi ‘master, owner’, which also occurs in oj muranzi ‘kabane title’ < *mura-nusi ‘village-master’, oj tônzi ‘mistress of the house’ < *tô-nusi ‘gate mistress’, oj arônzi (mj arunzi) ‘master’ < *ar-ônusi ‘exist-attr-master’. There is also mj miyanzi ‘majordomo’ (not attested phonographically in oj texts) < *miya-nusi ‘palace-master’. Then what is the first syllable pu-? I believe it is eoj pu ‘fire’7 that corresponds to woj pï (pô- in compounds). eoj pu ‘fire’ is a hapax legomenon attested only in mys 20.4419, a sakîmôri poem:8 (1) 伊波呂尓波 (2) 安之布多氣騰母 (3) 須美与氣乎 (4) 都久之尓伊多里弖 (5) 古布志氣毛波母

7 Technically eoj pu ‘fire’ is a word from Munzasi province, but it is from the Tatimbana district located in the west of this province, which is barely 50km. from Mt. Fuji. 8 The reconstruction of the original text, transcription, glossing and translation are from Vovin (2013: 188), with one minor adjustment in the transcription and additional highlighting made on purpose for this contribution.

86

vovin

(1) ipa-rö-ni pa (2) asi pu tak-ë-ndömö (3) sum-î yö-kë-wo (4) Tukusi-ni itar-i-te (5) kôpusi-kë-mô pa mö (1) house-dim-loc top (2) reed fire burn-ev-conc (3) live-nml goodattr-acc (4) Tukusi-loc reach-inf-sub (5) be.longing-attr-excl top pt (2) Although [we] make a fire out of reeds (1) at [my] house, (3) the living [there] is good, so (5) [I] will be longing for it (4) when I reach Tukusi! Note that pJ *poy ‘fire’ belongs to 1.3 (l) accentual class. The well-known rule of thumb for accentuation of Japanese nominal compounds is that the first element defines high or low register, while the second element defines locus, if there is any. The low initial register of punzi and its actual accentual class further confirm the possibility that the first syllable can be explained as eoj pu ‘fire’. Thus, eoj punzi < pJN *po-nusi ‘fire master,’ is quite a fitting name for an active volcano. For the usage of nusi ‘master’ in volcano deities names, cf. the oj name of the deity Opo-ana-nusi ‘big-hole-master’, which, as J. Marshall Unger once suggested (p.c.), could be the name of a volcano deity. In conclusion, two issues need to be addressed. First, if my solution of this etymology is correct, then eoj pu ‘fire’ is no longer a hapax legomenon, and we have the second independent evidence that the eoj word for ‘fire’ was indeed pu. It is almost customary now to reconstruct the pj form of woj pï as *pǝy (Martin 1987: 405) rather than *poy. However, I know of no examples where pJ *ǝ raises to eoj u, because only pJ *o > eoj u. Recently, both Pellard (on the basis of the Ryūkyūan evidence) (2011: 10), and myself on the basis of the philological evidence in woj (pace Mabuchi 1972: 88) and eoj phonological evidence just mentioned above (Vovin 2011: 222) argued for the reconstruction of pj *poy rather than *pǝy. Another attestation of eoj pu ‘fire’ gives further credence to the eoj side of argument in favor of *poy.9 Second, the reconstruction of the pJN accent of *punsi as 2.3 (ll) may not really be as certain as Martin has presented it (1987: 420). In addition to the Tokyo accent pointing to 2.4 (lh) or 2.5 (lf), there is another problem which concerns the accentuation of -nzi < nusi ‘master, owner’ in the compounds

9 Many thanks to John Whitman, with whom I have had a long discussion about the reconstruction of pJ word for ‘fire’ about five years ago. Many of his careful arguments against *poy vs. *pǝy made me revise and rethink the present argumentation about the name of Mt. Fuji as being evidence for pJ *poy rather than *pǝy.

on the etymology of the name of mt. fuji

87

listed above. The mj accent of nusi ‘master, owner’ itself is 2.4 (lh). Consequently, we would expect -nzi to have a high pitch. And indeed, this high pitch is regularly reflected in mj muranzi hhh ‘kabane title’ < *mura-nusi ‘village master’ and mj arunzi xxh ‘master’.10 The possible explanation accounting for the minor accentual discrepancy between mj funzi ll vs. the expected *funzi lh is that woj punzi and mj funzi must be loanwords from eoj punzi due to their eoj peculiar phonetic shape with -u- rather than with -ô-: we would expect woj *pônzi and mj *fonzi, if they were really cognates. Under this scenario, a minor accentual discrepancy that affects locus but not register is not totally unexpected, as such discrepancies occur even within the same language, cf. woj isô (mj accent hh) ‘rock, rocky shore’ and woj isi (mj accent hl) ‘rock, stone’.

Abbreviations Grammatical Terms acc attr conc dim ev excl inf loc nml pt sub top

Accusative Attributive Concessive Diminutive Evidential Exclamative Infinitive Locative Nominalizer Particle Subordinative Topic

Languages eoj MdJ mj oj

10

Eastern Old Japanese Modern Japanese Middle Japanese Old Japanese

The mj accent of oj tônzi ‘mistress of the house’ is unknown.

88 pA pJ pJN woj

vovin proto-Ainu proto-Japonic proto-Japanese Western Old Japanese

Primary Sources gk cmjf hf hkmb kkwks mys nsk scs sfib siwks tm wts

Gengo kokka (言語國訛), mid Edo (1601–1868ad) period Chikamatsu Jōruri fubon (近松浄瑠璃譜本), late 17th c. Hitati Fudoki (常陸風土記), 713 ad Heike mabushi (平家正節), 1776 ad Kokin waka shū (古今和歌集), 921 ad Man’yōshū (萬葉集), between 759 and 771ad Nihonshoki (日本書紀), 720 ad Shūchūshō (袖中抄), 1185–1190 ad Suruga Fudoki Ibun (駿河風土記異聞), 12th c.? Jōben-bon (浄弁本) of Shūi waka shū (拾遺和歌集), 1185–1333ad Taketori monogatari (竹取物語), late ninth or early tenth century ad Waji taikan shō (和字大観抄), first part of 18th c.

Secondary Sources Akimoto, Kichirō (ed.) 1958. Fudoki [Gazetteers]. Nihon Koten Bungaku Taikei [Series of the Japanese Classical Literature], vol. 2. Tokyo: Iwanami Shoten. Akinaga, Kazue; Ueno, Kazuaki; Sakamoto, Kiyoe; Satō, Eisaku; Suzuki, Yutaka 1997. Nihongo akusento shi sōgō shiryō. Sakuin hen [The comprehensive materials on the history of accentual history of the Japanese language. An index]. Tokyo: Tōkyōdō. Aoki, Kazuo; Inaoka, Kōji; Sasayama, Haruo; Shirafuji, Noriyuki (eds.) 1998. Shoku Nihongi [Continued Annals of Japan]. Shin Nihon koten bungaku taikei [New Series of the Japanese Classical Literature], vol. 12–16. Tokyo: Iwanami. Batchelor, John 1928. ‘Helps to the Study of Ancient Place-Names in Japan’. Transactions of the Asiatic Society of Japan, 1928, pp. 52–102. Reprinted in: Refsing, Kirsten (ed.). Early European Writings on the Ainu Language, vol. 9. London: Routledge Curzon, 1996. Batchelor, John 1938. An Ainu-English-Japanese Dictionary. 4th edition. Tokyo: Iwanami. Kuroita, Katsumi & Matsuyama, Jirō (ed.) 1965–1966. Nihonshoki [Annals of Japan]. Shintei zōho kokushi taikei [The newly corrected and enlarged series on Japanese history], vols. 1a and 1b. Tokyo: Yoshikawa kōbunkan.

on the etymology of the name of mt. fuji

89

Pellard, Thomas 2011. ‘Ryukyuan perspectives on the Proto-Japonic vowel system.’ Proceedings of the 20th Japanese/Korean Linguistics Conference. Stanford: ssli, pp. 1–15. Sakakura, Atsuyoshi (ed.) 1957. Taketori monogatari. In: Nihon koten bungaku taikei [Series of the Japanese Classical Literature], vol. 9, pp. 3–78. Tokyo: Iwanami shoten. Vovin, Alexander 1993. A Reconstruction of Proto-Ainu. Leiden & New York: Brill. Vovin, Alexander 2011. ‘On one more source of Old Japanese i2.’ Journal of East Asian Linguistics 20, pp. 219–228. Vovin, Alexander 2013. Man’yōshū. Book 20. A New English Translation Containing the Original Text, Kana Transliteration, Romanization, Glossing, and Commentary. Folkestone/Leiden: Global Oriental/Brill.

part 3 Theoretical Linguistics

∵

chapter 10

Against a vp Ellipsis Account of Russian Verb-Stranding Constructions John Frederick Bailyn

Recent proposals that Russian verb-stranding constructions are the result of vp-ellipsis after v+v raising (Gribanova 2009, 2011) (henceforth, the “vvpe analysis”) are presented and refuted on various grounds. The case against vvpe for Russian v-stranding involves examination of Gribanova’s arguments in favor of vvpe, as well as additional arguments against it. It is demonstrated, however, that one crucial insight of the vvpe analysis must be maintained, namely that there are (at least) two distinct processes allowing objects of otherwise obligatorily transitive verbs in Russian to be omitted, one of them involving ellipsis (though crucially not vp/vP ellipsis) and the other resembling discourse-licensed Argument Drop (ad). ad is revealed to be the result of the licensing of null object pronouns in a manner familiar to Huang’s (1984) Topic Drop and distinct from Rizzi’s (1986) null-object typology. The resulting analysis of Russian verb-stranding is consistent with the existing literature on true vvpe in those languages that show it uncontroversially (Goldberg 2006).

1

Russian v-Stranding Constructions

It is quite common in Russian to find grammatical constructions containing otherwise obligatorily transitive verbs without their internal argument(s). These will be referred to as v-stranding constructions (vsc). An example from Gribanova (2009) is given in (1). (In all vsc examples, the “stranded” verb will be indicated with bold face, and the neutral symbol ☉ will be used to mark the site of the missing argument(s)). (1) a. Ty poznakomil Mašu s Petej? you introduced Masha with Petya “Did you introduce Masha to Petya?”

© koninklijke brill nv, leiden, 2018 | doi: 10.1163/9789004351134_011

94

bailyn

b. Konečno, poznakomil ☉ ☉ of course introduced “Of course, I introduced ☉ ☉.” Goldberg (2006), following a long tradition dating back to Otani & Whitman (1991), argues that various languages, such as Hebrew, Swahili, Irish and, Ndendeule (but not Korean and Japanese), derive v-stranding by raising the main verb out of the extended vp/vP domain, followed by ellipsis of the vP itself. This is the vvpe account. Goldberg’s primary argument in favor of a vvpe account of v-stranding in such languages involves ruling out the alternative of Argument Drop, an approach which is similar in spirit to that taken by Gribanova (2011) for Russian, though crucially different in certain details as we will see immediately below. Gribanova’s vvpe account of Russian v-stranding differs from Goldberg’s account of the languages mentioned above in one crucial respect: Russian does not show v→t raising in overt syntax (Bailyn 1995, 2012, Kollestinova 2007, Gribanova 2009, 2010, 2011), something that is required of all the other vvpe languages in Goldberg’s typology. Rather, Gribanova argues, Russian main verbs raise out of vP to an intermediate category between vP and tp, namely Asp(ect)P, after which the lower vP is elided, stranding the main verb. A schematic representation, from Gribanova (2011), is given in (2): (2) Gribanova’s (2011) vvpe analysis of Russian vsc TP T

AspP AsP

sp … V … v

vP tv

VP tV

DP

In this article, I argue against the vvpe analysis for Russian v-stranding and propose an alternate account for deriving the relevant cases. The article is structured as follows. First, in Section 2, I present the two primary arguments in favor of a vvpe analysis of Russian v-stranding, namely (i) that there is no good alternative (in particular, that v-stranding cannot always be the result of Argument Drop) and (ii) that certain facts of Russian verbal morphology

against a vp ellipsis account of verb-stranding constructions

95

and matching in v-stranding favor vvpe. In Section 3, I present six arguments against the vvpe account. In Section 4, I present an alternative account that captures Gribanova’s core intuition (that two distinct processes can be involved in deriving Russian v-stranding) without requiring vvpe.

2

The Case for Russian vvpe

Gribanova’s (2009, 2010, 2011) case for a vvpe analysis of Russian v-stranding consists primarily of two arguments, as follows:1 (i) The alternative to vvpe, some form of Argument Drop (ad), fails (for at least some Russian vscs), and (ii) Morphological verbal matching effects of a particular kind hold (in at least some vscs), which can be accounted for under a vvpe account, but do not follow from any account that does not involve vP ellipsis. I now outline the two arguments in detail. The first argument for vvpe put forth by Gribanova is an argument against the most plausible alternative, namely (some kind of) Argument Drop (ad). The argument runs as follows: (3) Gribanova’s argument against an ad account of v-stranding: a. There are contexts where ad is unavailable (certain syntactic islands) b. In such contexts, v-stranding is still possible c. Therefore, in such contexts (at least), v-stranding can’t be derived by ad. d. Therefore in such contexts (at least), v-stranding must be derived by vvpe To start this argument, Gribanova reviews the well-known fact that uncontroversial cases of vp/vP ellipsis require a linguistic antecedent. This is shown in the distinction between successful English vP ellipsis in (4) (with a linguistic antecedent) and unsuccessful English vP ellipsis in (5): (4) a. Nobody here will be able to sew those ripped jeans! b. Don’t worry, grandma will be able to [vp sew those ripped jeans].(vPE)

1 It should be noted that the 2009 and 2010 handout versions contain additional arguments not found in the written 2011 version. I do not address those arguments here.

96

bailyn

(5) Situation: A man with ripped jeans enters the room *Don’t worry, grandma will be able to [vp sew those ripped jeans]. (*vPE) On the other hand, Argument Drop (ad) (Hoji 1998, Saito 2007, Aoun & Li 2007, Gribanova 2010, Sigurðsson 2011), does not require a linguistic antecedent, and so the Russian equivalent of (5), which is fine, must be some form of Argument Drop: (6) Situation: A man with ripped jeans enters the room Ne volnujsja, babuška zaš’et ☉ don’t worry grandma will sew “Don’t worry, grandma will sew [them].”

(ad)

Crucially for Gribanova’s argument, ad fails inside syntactic islands. Thus, if no linguistic antecedent is present, precluding vP ellipsis, and an island is present, precluding ad, the result is ungrammatical. This is shown in (7): (7) Situation: A man with ripped jeans enters the room *Ne volnujsja, sejčas pridjet [čelovek, (*ad) kotoryj zaš’et ☉ ] don’t worry now will come [man (*vPE) who will sew “Don’t worry, [someone who will sew [them]] is coming.” example from gribanova 2010

Because (7) is unacceptable, both ellipsis and ad must fail. Ellipsis is ruled out by the lack of a linguistic antecedent, leaving only ad as a possibility. The fact that the example is bad shows that ad fails within syntactic islands.2 By contrast, if a linguistic antecedent is provided, such as (8), the same response as (7) becomes possible, implicating ellipsis, and not ad, as the only viable analysis of examples such as (8), which is identical to (7) but without a linguistic antecedent:

2 Gribanova does not provide an account for the island-sensitivity of ad, other than to say that pro “establishes an A-bar relation with a topic in the matrix clause.” Gribanova does present parallels between wh-movement constraints and ad island-sensitivity, which provide support for the generalization. I will therefore assume along with Gribanova that some such island-sensitivity restricts ad in Russian, (although it must be noted that wh-movement/ad parallels do not always hold in the manner presented by Gribanova). The distribution and nature of these restrictions is outside the scope of this article.

against a vp ellipsis account of verb-stranding constructions

97

(8) a. Menja volnujet, čto nikto ne zašil džinsy me worries that no one neg sewed jeans “It worries me that no one has sewed these jeans.” b. Ne volnujsja, sejčas pridet [čelovek, kotoryj zašetv [ tv ☉ ] don’t worry now will come [man who will sew “vvpe” “Don’t worry, [someone who will sew [them]] is coming.” = example (39) from gribanova 2011

Note that this diagnostic is the only one Gribanova provides to distinguish vvpe from ad. Therefore all examples of purported vvpe should be embedded within such islands. In what follows, I will use relative clauses, which are strongly opaque to wh extraction, to maintain the parallel as much as possible. Gribanova’s second argument involves morphological matching. Recall that Gribanova’s account involves v movement of the head of AspectP, but no higher. This movement of the verb must be motivated, since it is generally accepted that Russian main verbs do not move overtly out of vP (Bailyn 1995, Kollestinova 2007, Gribanova 2009, 2010). Therefore, Gribanova relies on syntactic accounts of Russian verbal prefixation (following Svenonius 2004) to provide evidence that this movement is exactly what is involved in successful cases of vvpe. To do this, she relies on the distinction between two kinds of verbal prefixes in Russian: Lexical prefixes (lp) and Superlexical Prefixes (sp). lps are essentially part of certain lexical items (such as English refrain, retreat),3 whereas sps are heads of AspectP, by assumption, following Svenonius 2004. Assuming an identity condition on vP ellipsis of the standard kind, in successful cases of vvpe, sps, residing outside vP, can be expected to vary, while lps, being vP-internal, must be identical. The predictions for Gribanova, therefore, are as follows: v-stranding should be possible within syntactic islands, if either (i) the verbs are identical or (ii) the verbs differ only in having distinct superlexical prefixes (sp). On the other hand, v-stranding should not be possible within syntactic islands if (i) the verbs are different or (ii) the verbs have identical stems with distinct lexical prefixes (lp) (note that (ii) essentially reduces to (i) on standard approaches to lexical prefixes). Evidence to support these predictions is given in Gribanova 2011, as follows:

3 Of course, in some cartographic accounts, these too could be syntactic heads, but they certainly fall within the standard vp/vP domain, which is all that matters for Gribanova, and therefore for the current discussion.

98

bailyn

(9) Predictions about availability of v-stranding within syntactic islands (where ad is unavailable): a. Identical verbs (10) and identical stems with distinct superlexical prefixes (11) are predicted to be ok. b. Different verbs (12) and identical stems with distinct lexical prefixes (13) are predicted to be out. (10) a. Kažetsja, čto nikto ne podnjal tu vazu seems that no one neg picked up that vase “It seems that no one picked that vase up.” b. Tot fakt, čto nikto ne podnjal ∅ menja očen’ ogorčaet that fact that no one neg picked up me very upsets “vvpe” “The fact that no one picked (it) up upsets me greatly.” (11) v-stranding inside an island with identical stems and distinct Superlexical Prefixes (Gribanova, 2011, examples (73&74)) a. Kažetsja, čto nikto ne podnjal vazu kotoraja uže ne seems that no one neg picked up vase which already neg pervyj raz padaet first time falls “It seems that no one picked up the vase that fell more than once.” b. Naoborot, uže prišel čelovek, kotoryj PEREpodnjal ∅ on-contrary already came someone who re-picked up “vvpe” “On the contrary, a person who picked (it) up again already came.” (12) Unacceptable v-stranding inside an island with different verbs (Gribanova, 2011, examples (63&64)) a. Kto-to uronil ètu vazu someone dropped this vase “Someone dropped that vase.” b. *Tot fakt, čto nikto ne podnjal ∅ menja očen’ ogorčaet the fact that no one neg picked up me very upsets *“vvpe” “The fact that no one picked (it) up upsets me greatly.”

against a vp ellipsis account of verb-stranding constructions

99

(13) Unacceptable v-stranding inside an island with identical stems and distinct Lexical Prefixes (Gribanova, 2011, example (76)) a. Nepravitel’stvennye organizacii dolžny byli … RAZdavat’ Non-governmental organizations should have … Distributed butylki vody bežencam na Gaiti bottles of water refugees in Haiti “ngos were supposed to distribute bottles of water to Haitian refugees.” b. *Nas očen’ volnujut sluxi, čto oni PROdavali ∅ us very worry rumors that they sold *“vvpe” “We are very worried by rumors that they sold *(them).” These two arguments form the core of the Gribanova analysis. Next, I present six arguments against the vvpe analysis of Russian vse constructions, still assuming, along with Gribanova, that instances of ad can be controlled for by examining cases within syntactic islands only. Gribanova’s second argument, (morphological matching) is countered in 3.5. The status of what the alternative to vvpe is within islands (her first argument) is addressed in section 4.

3

The Case Against v-Stranding as vvpe

In this section, I present six arguments against the vvpe account of vscs in Russian. The first two are theoretical arguments; the final four are empirical arguments. (14) Six arguments against vvpe for Russian i. “True” Russian vP ellipsis does not strand verbs ii. vvpe occurs in v→t raising languages only iii. Russian v-stranding constructions allow strict and sloppy identity; vP ellipsis allows only sloppy readings iv. vP adverb constituency is not required in v-stranding v. Verb matching requirements do not support vvpe vi. Subject/Object drop dependencies are unexpected in v-stranding under a vvpe account 3.1 “True” Russian vP Ellipsis Does Not Strand Verbs Kazenin (2006) demonstrates that Russian has at least two kinds of phrasal ellipsis, one of which elides the entire verbal complex (including any AspP pro-

100

bailyn

jections), but nothing in the ip/tp domain. I call this kind of ellipsis “true vP+ ellipsis.” To quote, “two types of predicate ellipsis [in Russian] are possible, one retaining a polarity marker without the aux (15) and the other one retaining a polarity marker with the aux (16).” (15) type i: “polarity ellipsis” (using da, net) (Kazenin 2006) Petja [ljubit sebja], a Vasja net ∅ Petya [loves self] but Vasja net ∅ “Peter loves himself, but Vasja doesn’t [love himself]” (sloppy) (16) type ii: “true” vP+ ellipsis—using the auxiliary budet Petya budet [pomogat’ sebe] a Kolja ne budet ∅ Petya aux [to help self] but Kolja neg aux “Petya will help himself, but Kolya won’t ∅.” (∅ = help himself (Kolya) = sloppy)) The relevant structures are shown in (17): (17) a. type i ellipsis ΕP

b. type ii ellipsis TP

da/net [+F]

budet (aux)

TP | pro [D-linkes]

vP+ | pro/∅

It is the latter case that is of interest to us here. The ellipsis site in cases such as (16) comprises the entire verbal complex, including any syntactic domains headed by prefixes (since verbs are obligatory elided). This corresponds to traditional English vP ellipsis. (The + indicates that any extended cartographic domains above vP but below tp, such as AspP in Gribanova’s account, must be included here, since the extended vP is entirely elided.) Familiar diagnostics implicating traditional vP ellipsis are given in Kazenin (2006), such as the inclusion in the ellipsis site of verbal adjuncts (18), the availability of the construction in both coordinate and subordinate structures (19), the possibility that the antecedent be in a separate sentence (20) but its obligatory nature (21), and the fact that sloppy readings can be obtained (22).

against a vp ellipsis account of verb-stranding constructions

101

(18) Obama [budet zanimat’sja posle užina], Obama [aux practice] after dinner a Biden ne budet ∅ but Biden neg aux “Obama will practice after dinner, but Biden won’t ∅”. (∅ = practice after dinner) (19) a. Ja budu pomogat’ Kolje, I aux [help Kolya] a Petja ne budet [vp ∅ ] but Petya neg aux “I will help Kolya but Petya won’t ∅.” b. Ja budu pomogat’ Kolje, I aux [help Kolya] esli Petja ne budet [vp ∅ ] if Peter neg aux “I will help Kolya if Petya won’t ∅.”

(vP ellipsis)

(vP ellipsis)

(20) a. Obama [budet xodit’ na zanjatie! Obama [aux to class] “Obama will go to class!” b. Ser’ezno? a Biden ne budet ∅ seriously but Biden neg aux “Seriously? But Biden won’t” (∅ = go to class) (vP ellipsis) (21) Context: Obama puts the square block in the round hole. Obama to Biden: *A ty ne budeš ∅ (vP ellipsis) and you neg aux “But you won’t ∅” (∅ = [(be able to) put the square block in the round hole]) (22) Petja budet [pomogat’ sebe], (vP ellipsis) Petya aux [to help self] a Kolja ne budet ∅ but Kolya neg aux “Petya will help himself, but Kolya won’t ∅ [help Kolya]”

102

bailyn

I take it as uncontroversial, then, that Russian has true vP+ ellipsis. Therefore, if Gribanova is also right, there are (at least) two kinds of vP+ ellipsis (as well as ip ellipsis in sluicing). This is a theoretical weakness of the vvpe account. (23) Gribanova’s required vP ellipsis typology for Russian: a. ip/tp ellipsis: sluicing, etc. (equal to Kazenin’s PolP ellipsis) b. vP+ ellipsis: the v elides (would include AspP) (cf. Kazenin 2006; not discussed in Gribanova 2009, 2010, 2011) c. Purported vP ellipsis: the v survives (not in Kazenin 2006) This raises various questions about the theoretical claim being made: Why should a language contain all of (23)? In particular, what allows a language to elide a verbal projection lower than the full verbal complex vP+ (as needed in (23c) for the vvpe account)? Assuming any verbal xp can be elided will not help—why then does English not derive vscs through v→v movement, followed by vp ellipsis? The burden of proof is surely on the proponent of a system like (23). This leads us to the second theoretical argument against the ellipsis account of Russian vscs. 3.2 vse Occurs in v→t Raising Languages Only If a language allows something smaller than vP+ to be elided (the complement of Asp0 for Gribanova), then why do all the languages showing vscs in Goldberg’s (2006) typology have v→t movement independently? Goldberg’s generalization that v-stranding is enabled by v-raising out of the maximal verbal domain is thus contradicted, and we would not expect any such correlation to hold. And yet it does, in all known cases other than Russian. In Hebrew, for example, v→t is independently motivated by familiar word order facts: verbs precede manner adverbs and floated quantifiers. It is well-established, as readily acknowledged by Gribanova (2009, 2011), based on (24), that there is no v→t raising in Russian (Bailyn 1995, Kallestinova 2007) (neutral context and intonation are assumed): (24) … čto Ivan často celuet (*často) Mašu. [that s-adv-v-o] … that Ivan often kisses (often) Mary “I know that Ivan often kisses Mary.” (*that [s-v-adv-o])

against a vp ellipsis account of verb-stranding constructions

103

3.3 Strict vs. Sloppy Readings True vP+ ellipsis in Russian not only allows sloppy identity readings, as shown in (25) from Kazenin (2006), it also strongly disprefers strict readings in the same context: (25) Petja budet pomogat’ sebe (vP ellipsis) Petya aux [to help self] a Kolja ne budet ∅ but Kolya neg aux “Petya will help himself, but Kolya won’t ∅.” (sloppy only) (26) Dina [kupila svoej dočke škol’nye učebniki] Dina [bought self’s daughter school textbook a Paša ne kupil ☉ ☉ but Paša neg bought “Dina bought her daughter textbooks, but Paša didn’t [buy her/his daughter textbooks].” (strict possible!) ((26) is Gribanova’s v-stranding example, reported as ambiguous) (27) a. Obama budet xvalit’ sebja Obama aux praise self a Biden ne budet ∅ but Biden neg aux “Obama will praise himself but Biden won’t ∅”. (*∅ = will praise Obama) (*strict)

(vP ellipsis)

b. Obama zaxvalil sebja, Obama (over)praised self i Biden tože zaxvalil ☉ and Biden also praised “Obama (over)praised himself and Biden also (over)praised [him]”. (☉ = (praised) Obama) (strict!) The proper generalization appears to be that in Subject-oriented anaphor binding languages such as Russian (versus English, say), only sloppy readings are available. This is to be expected on the assumption that the Subject-condition on anaphor binding results from covert movement of the anaphor (or its feature) to t (Cole & Sung 1994, Saito 2003), which, at the time the vP phase is built, is unchecked/ unsatisfied, so the only value for the anaphor in the elided vp is the bound variable, leading to the sloppy reading. Once the full tp has been

104

bailyn

built, the “strict” referent is determined, and can be accessed for pronominalization purposes, that are not phase-bound. If strict readings are available, as in Verb-stranding constructions, vP ellipsis cannot be at work, because the value of the anaphor would not have accessed its antecedent’s features within vP. This in turn constitutes an argument against vP ellipsis in Russian Verb-stranding. 3.4 Adverb Constituency The fourth argument concerns adverb constituency. In true vP+ ellipsis, the adverbial may be included in the interpretation of the elided vP as in (28a), but need not, at last marginally, as in (28b): (28) a. Ty budeš’ snimat’ Vasil’eva často, a ja ne budu ∅ you aux record Vasiliev often but I neg aux “You will record Vasilev often, but I will not ∅.” ✓ (∅ = record Vasiliev often) (vP ellipsis) b. ??Ty budeš’ snimat’ Vasil’eva často, a ja budu ∅ redko you aux record Vasiliev often but I aux rarely “You will record Vasilev often, but I will rarely ∅.” ??(∅ = record Vasiliev) (vP ellipsis) With Verb-stranding, the exact opposite holds: Including the adverbial is out in (29a), whereas excluding it is fine: (29) a. Ty snimal Vasil’eva často, a ja ne snimal ☉ you recorded Vasiliev often but I neg recorded “You recorded Vasiliev often but I didn’t record [him]” ✓ (☉ = Vasil’eva) * (☉ = [t Vasil’eva often]) (cf. (28a)) b. Ty snimal Vasil’eva často, a ja snimal ☉ redko you recorded Vasiliev often but I recorded rarely “You recorded Vasiliev often but I recoded [him] rarely” (☉ = Vasil’eva) (cf. (28b)) 3.5 Verb Matching Requirements Revisited The fifth argument is simple: The proposed requirements for morphological matching between the main verb and the stranded verb that licenses vvpe do not hold. All four of the predictions about matching have systematic counterexamples. Recall that identical verbs are predicted always to license vvpe, as are identical stems with distinct Superlexical prefixes, but we see in (30) and

against a vp ellipsis account of verb-stranding constructions

105

(31) that they do not. Conversely, entirely different verbs are predicted to be out, as are identical stems with distinct lexical prefixes, but we see in (32) and (33) that these can be fine. (30) ?Nikto ne ljubit Ivana, a menja udivljaet tot fakt, čto Nadja no one neg loves Ivan but me surprises the fact that Nadya ljubit ☉ loves “No one loves Ivan, but the fact that Nadya loves [him] surprises me.” (31) *Saša nikogda ne pela “Piano Man”, daže v karaoke poetomu Sasha never neg sang “Piano Man” even at karaoke therefore menja udivil tot fakt, čto včera zapela ☉ me surprised the fact that yesterday sang “Sasha never sang ‘Piano Man’, even at karaoke, that’s why the fact that (she) started to sing [it] surprised me.” (32) ?Kto-to skazal, čto vse nenavidjat Ivana, tak čto menja someone said that everyone hates Ivan so that me udivil fakt, čto Nadja ljubit ☉ surprised fact that Nadya loves “Someone said that everyone hates Ivan, so the fact that Nadya loves [him] surprised me.” (33) Nikto nikogda ne poet “Piano Man” daže v karaoke tak čto no one ever neg sings “Piano Man” even at karaoke so that menja udivil fakt, čto Saša včera spela ☉ me surprised fact that Sasha yesterday sang “No one ever sings ‘Piano Man’ even in karaoke, so the fact that Sasha sang [it] yesterday surprised me.” To summarize, the matching requirements on v-stranding (within islands) seem problematic for vvpe: Contrary to the specific claims in Gribanova (2011), stranded verbs can have distinct lexical prefixes and indeed the verbs do not have to be identical. In fact, the conditions do not seem to be entirely syntactic: vPs differing only in the Superlexical prefix of the stranded verb do not guarantee successful stranding, and even identical verbs can fail. Therefore it is extremely unlikely that we are dealing with vP ellipsis after raising of the v+v complex.

106

bailyn

3.6 Subject/Object Drop Dependencies Finally, there is the remarkable dependence of successful v stranding on omission of the subject, something that vP ellipsis never shows. Consider the following exchange: (34) a. Kak tebe lingvistika? how youdat linguisticsnom “How do you like linguistics?” b. (*ja) Nenavižu ☉ (*I) hate1sg [it] “I hate [it].” (lit: “Hate.”) In the response, the object is missing, as is the subject. Gribanova notices this as well, (“in many examples, … the subject in the clause with the stranded verb is absent”) though she does not observe that it must be absent. Thus all other options are unacceptable, both in main clauses (35), but also in island contexts: (35) a. ☉ Nenavižu ☉ hate1sg “I hate it.” (lit: “Hate.”)

(subject and object dropped = ok)

b. *Ja nenavižu ☉ I hate1sg “I hate it.” (*lit: “I hate.”)

(*only object dropped)

c. ?☉ Nenavižu ee hate1sg it “I hate it.” (?lit: “Hate it.”)

(?only subject dropped)

d. Ja nenavižu ee I hate1sg it “I hate it.” (lit: “I hate it.”)

(nothing dropped = ok)

(36) a. Čto Saša dumaet pro lingvistiku? What Sasha thinks about linguistics “What does Sasha think about linguistics?”

against a vp ellipsis account of verb-stranding constructions

b. Menja udivljaet tot fakt, čto (*on) nenavidit ☉ Me surprise that fact that (*he) hates3sg [it] “The fact that (he) hates [it] surprises me.”

107

(v-stranding)

Crucially, true vP ellipsis has no such restriction: (37) Ja segodnja zanimalsja lingvistikoj, a zavtra ( ja) ne budu ∅ I today studied linguistics but tomorrow (I) neg aux (vP ellipsis) “Today I studied linguistics but tomorrow I won’t ∅.” Once again, we see that v-stranding constructions behave in a manner distinct from what we would expect with vP ellipsis. In the final section I turn to a plausible alternative.

4

The “Inner Constituent Ellipsis” Alternative

Gribanova’s work has definitively established that v-stranding is possible both in the purely discourse licensed environments of Argument Drop (ad), and in complex environments, such as within a’-islands, where ad typically fails. As Gribanova argues, such examples implicate syntactic ellipsis, a conclusion that seems correct. What is at issue then, is the kind of ellipsis found in such contexts. We have seen ample evidence against the vP ellipsis analysis. Instead, I propose that v-stranding within islands results from “Inner Constituent Ellipsis” (ice), a process that can eliminate vp-internal np/dps, pps, and local adverbials of the relevant kind (see Sigurðsson 2011 for related proposals). (38) Inner Constituent Ellipsis (ice) Freely elide any vp internal constituent (dp/pp) that is both identified and v-licensed. (39) Conditions on Inner Constituent Ellipsis a. ice Identification. An antecedent dp/pp must be identical to the elided dp/pp, a relationship established by agree. b. ice Licensing. Selection between the main verb and the antecedent dp/pp must be parallel to the relationship between the stranded verb and the elided dp/pp

108

bailyn

The exact definition of parallelism still requires further research. However, it is clear just from the examples in (30)–(33) above that thematic relations, case relations, semantic field, and discourse status (especially that of contrast) are all involved. And, although the kind of prefixation involved may coincide with those factors, as in Gribanova’s examples, they need not, as seen above. The kind of prefix is not (necessarily) of central concern, as the vP ellipsis account requires. Consider the ellipsis in (41) ((40) showing ad is out): (40) Context: people looking for a book *Menja udivljaet tot fakt čto kto-to prosto vzjal ☉ me surprises the fact that someone simply took “*(The fact) that someone simply took [it] surprises me.” (cf. Gribanova 2010) (41) a. Kto-to ukral moju knigu! someone stole my book “Someone stole my book!” b. ?Menja udivljaet tot fakt čto kto-to prosto vzjal ☉ me surprises the fact that someone simply took “?(The fact) that someone simply took [it] surprises me.” (same as (40)b but licensed here by ukral~vzjal) Here we find successful island-internal Verb-stranding with morphologically unrelated but parallel verbs. Absolute identity conditions on vP ellipsis would be too strong to allow such examples. And they would be too weak to disallow (30) or (31). ice, under the proper formulation of parallelism, accounts for exactly this distribution along with the other facts reported above. As for the Subject/Object dependency, we find it both in ad and in ice, indicating that some kind of blocking process is involved in both. Sigurðsson (2011) proposes specificity hierarchy requirements on null arguments in Germanic: (42) Relative Specificity Constraint (Sigurðsson 2011: 290) The dropped object cannot be more specific than the subject. Sigurðsson (2011) argues that “the Relative Specificity Constraint is puzzling at first sight. However, it can be analyzed as a minimality violation, that is, an intervention effect.” The same would then surely be true of the agree relation that establishes the identification needed for ice. The vvpe analysis has no known way to account for such blocking effects.

against a vp ellipsis account of verb-stranding constructions

109

References Aoun, Joseph and Audrey Li 2007. ‘Ellipsis and missing objects’ ms, University of Southern California. Bailyn, John F. 1995. A Configurational Account of Russian “Free” Word Order. Cornell dissertation. Bailyn, John F. 2012. The Syntax of Russian. Cambridge University Press, Cambridge, uk. Cole, P. & Sung, L.-M. 1994. ‘Head movement and long-distance reflexives,’ Linguistic Inquiry 25, pp. 355–406. Goldberg, Lotus 2006. Verb-Stranding vp Ellipsis: A Cross-Linguistic Study. McGill University dissertation. Gribanova, Vera 2009. ‘Ellipsis and the syntax of verbs in Russian.’ Talk given at nyu. Gribanova, Vera 2010. ‘On diagnosing ellipsis and argument drop: The view from Russian.’ Talk given at mit. Gribanova, Vera 2011. ‘Verb-stranding verb phrase ellipsis and the structure of the Russian verbal complex.’ Natural Language and Linguistic Theory. Hoji, Hajime 1998. ‘Null object and sloppy identity in Japanese.’ Linguistic Inquiry, 29:1, pp. 127–152. Huang, C.-T. James 1984. ‘On the distribution and reference of empty pronouns’. Linguistic Inquiry 15:3, pp. 531–574. Johnson, Kyle 2001. ‘What vp ellipsis can do, what it can’t, but not why’. The Handbook of Contemporary Syntactic Theory Kallestinova, Elena 2007. Aspects of Word Order in Russian. University of Iowa unpublished Ph.D. dissertation. Kazenin, Konstantin 2006. Polarity in Russian and typology of predicate ellipsis. Ms. Moscow State University. Otani, Kazuyo & John Whitman 1991. ‘v-raising and vp- ellipsis.’ Linguistic Inquiry, 22:2, pp. 345–358. Rizzi, Luigi 1986. ‘Null objects in Italian and the theory of pro’. Linguistic Inquiry, 17:3, pp. 501–557. Saito, Mamoru 2003. ‘A derivational approach to the interpretation of scrambling chains’. Lingua 113, pp. 481–518. Saito, Mamoru 2007. ‘Notes on East Asian argument ellipsis’. Language Research 43, pp. 203–227. Sigurðsson, Hoskildur 2011. ‘Conditions on argument drop’. Linguistic Inquiry 42:2, pp. 267–304. Svenonius, Peter 2004. ‘Slavic prefixes inside and outside vp’. Nordlyd 32:2, pp. 205–253.

chapter 11

A New Approach to -zhe in Mandarin Chinese* Redouane Djamouri and Waltraud Paul

1

Introduction

Among the aspectual suffixes in Mandarin Chinese, the so-called durative/progressive aspect -zhe has proven especially elusive, notwithstanding the large amount of literature devoted to it. The present article adopts a new approach and claims that -zhe is not on a par with the aspect markers zài, -le and -guo. Unlike the latter, which realize a “high” aspect pertaining to the entire event, -zhe is shown to instantiate a “low” aspect (aktionsart) (cf. Travis 2010 for the dichotomy high vs low aspect). Importantly, -zhe also signals the dependent status of the verbal projection concerned. As a consequence, in contrast to the aspect markers zài, -le and -guo, for which we observe constraints when in nonmatrix contexts, the opposite holds for -zhe: it is severely constrained in matrix contexts, but occurs rather freely in non-matrix contexts. The article is organized as follows. Section 2 provides data that challenge an analysis of -zhe as a high aspect marker on a par with zài, -le and -guo. Section 3 examines cases where different heads can license the non-autonomous verbal projection ‘V-zhe’. The fact that the progressive aspectual auxiliary zài is one of them further weakens the status of -zhe as a high aspect marker. Sections 4 and 5 introduce two constructions where ‘V-zhe’ is acceptable in a matrix context, viz. locative inversion sentences, on the one hand, and sentences ascribing the event as a property to the subject, on the other. Section 6 concludes the article with a short remark on the historical evolution of ‘V-zhe’.

* This article is based on our iacl-19 presentation at Nankai University in Tianjin in June 2011. We are extremely grateful to Y.-H. Audrey Li, Mai Ziyin, Victor Junnan Pan, Qiu Yiqin, and Zhitang Yang-Drocourt for discussion and data. We also thank the two editors for their careful attention.

© koninklijke brill nv, leiden, 2018 | doi: 10.1163/9789004351134_012

a new approach to -zhe in mandarin chinese

2

111

Cases Challenging an Analysis of -zhe as High Aspect

In this section, we show that -zhe cannot be analysed as high aspect. Interestingly, much relevant data can be found in the literature, but arguably they have not been taken at face value. The examples provided in the literature for -zhe as an alleged high durative/progressive aspect marker typically involve sentences with ‘V-zhe’ in an adjunct clause preceding the matrix verb: (1) 他吹着口哨下楼梯。 Tā [vP[adj.clause pro chuī-zhe kǒushào] xià lóutī]1 3sg blow-zhe whistle descend stair ‘He walked down the stairs whistling.’ chen chung-yu 1986: 2; (10a), bracketing added

(2) 他看着书等火车。 Tā [vP[adj.clause pro kàn-zhe shū] děng huǒchē] 3sg read-zhe book wait train ‘He waits for the train while reading.’ First, there is no restriction on the aspect of the matrix verb (V2) in the adjunct structure ‘S [vP[adj.clause pro V1-zhe (O)] V2P]’. This would be difficult to explain if -zhe itself were a high aspect marker on a par with zài, -le and -guo. (3) 他抱着书走进了教室。 Tā bào-zhe shū zoǔjìn-le jiàoshì 3sg hold-zhe book enter-perf classroom ‘He entered the classroom holding a book.’ chen zhong 2009: 86

(4) 他闭过眼睛躺过一会儿。 Tā bì-zhe yǎnjīng tǎng-guo yī huìr 3sg close-zhe eye lie-exp one while ‘He had lain down for a while with his eyes closed.’

1 The following abbreviations are used for glossing the examples: cl classifier; exp experiental aspect (-guo); neg negation; perf perfective aspect (-le); pl plural (e.g. 3pl = 3rd person plural); progr progressive aspect (zài); sg singular; sub subordinator.

112

djamouri and paul

(5) [他]皱着眉头想了一想。 [Tā] zhòu-zhe méitóu xiǎng-le yī xiǎng 3sg knit-zhe eye.brow think-perf one think ‘Knitting his brow he thought a moment.’ jaxontov 1988: 125, (47)

(6) 老王一定会看着你笑。 Lǎo Wáng yídìng huì kàn-zhe nǐ xiào Lǎo Wáng certainly will see-zhe 2sg laugh ‘Lǎo Wáng will certainly laugh when seeing you.’ The matrix verb in (3) and (4) is suffixed by the perfective aspect -le and the experiential aspect -guo, respectively. (5) is a case of verb repetition ‘V-le yī V’ associated with the meaning of ‘doing something for a short while’. (6) finally illustrates the future auxiliary huì selecting the vP headed by xiào ‘smile’ including the adjunct clause with ‘V-zhe’. Second, in this structure -zhe cannot be replaced by -le or -guo: (7) 他站着/*了/*过吃饭。 Tā [vP[adj.clause pro zhàn-zhe/ *-le/ *-guo] chī fàn 3sg stand-zhe/ -perf/ -exp eat food ‘He eats while standing.’ (8) 他关着/*了/*过窗户睡觉。 Tā [vP[adj.clause guān-zhe/ *-le/ *-guo chuānghu] shuì jiào] 3sg shut-zhe/ -perf/ -exp window sleep sleep ‘He sleeps with the window closed.’ While (7) with -le or -guo is simply rejected, (8) becomes acceptable with -le when parsed as two successive events: ‘After he had closed the window, he went to sleep’. Third, if -zhe were a high aspect marker, its presence on both the matrix verb and the verb in the adjunct clause would be unexpected, given that the simultaneous presence of high aspect markers on all verbs within a single sentence is prohibited (cf. (11), (12)). (9) 他闭着眼睛躺着。 Tā [vP[adj.clause pro bì-zhe yǎnjīng] [matrix vP tǎng-zhe]] 3sg close-zhe eye lie-zhe ‘He lies with his eyes closed.’ jaxontov 1988: 125

a new approach to -zhe in mandarin chinese

113

(10) 他耷拉着头坐着。 Tā dāla-zhe tóu zuò-zhe 3sg lower-zhe head sit-zhe ‘He is sitting with his head lowered.’ jaxontov 1988: 125

(11) *他闭过眼睛躺过一会儿。 *Tā [vP[adj.clause pro bì-guo yǎnjīng [matrix vP tǎng-guo yī huìr]] 3sg close-exp eye lie-exp one while (12) *他耷拉了头蹲了。 *Tā [vP[adj.clause pro dāla-le tóu [matrix vP dūn-le ]] 3sg lower-perf head sit-perf While the first verb plus -zhe in (9) and (10) is contained in an adjunct clause and represents the canonical case of -zhe, the presence of -zhe on the second, i.e. the matrix verb, seems at first sight unexpected in our approach, which basically considers -zhe as signaling the dependent nature of the verbal projection. This case is discussed in section 5 below.

3

Structures Licensing ‘V-zhe’ as Dependent Predicate

Under more careful scrutiny, many of the cases where ‘V-zhe’ at first sight occurs in a matrix context turn out to involve non-matrix contexts.2 Again, much relevant data have been cited in the literature. First, ‘V-zhe’ is selected as complement by the auxiliary zài indicating progressive aspect: (13) 他*(在)看着电视。 Tā [AspP[Asp° *(zài)] kàn=-zhe diànshì] 3sg progr watch-zhe tv ‘He is watching tv.’

2 It is important to note that the matrix vs. non-matrix dichotomy with respect to -zhe is very sharp for Northern Mandarin speakers, whereas this does not seem to be the case for e.g. Taiwanese Mandarin speakers. Accordingly, the present study is based on the judgements of the first group.

114

djamouri and paul

(14) 他受了伤却仍*(在)坚持着比赛。 Tā shòu-le shāng què rēng *(zài) jiānchí-zhe bǐsài 3sg get-perf wound but still progr persist-zhe competition ‘He got hurt, but nevertheless continued the competition.’ chen zhong 2009: 81

The combination of zài with ‘V-zhe’ in (13) and (14) is well-known from the literature (cf. Chen Chung-yu 1978: 83, Paris 1988: 165, among others). However, not much attention has been paid to the obligatory presence of zài in sentences such as (13) and (14). Nor has one seen the contradiction arising from the alleged high aspectual status of -zhe and its occurring in the complement selected by another high aspect, i.e. zài. In fact, it is not -zhe that induces the progressive aspect reading in (13) and (14), but zài, as evidenced by the possibility for zài to select a verbal projection without -zhe:3 (15) 他在看电视/洗衣服。 Tā zài kàn diànshì / xǐ yīfu 3sg progr watch tv / wash clothing ‘He is watching tv/doing the laundry.’ The status of zài as a high aspectual head explains why zài cannot take as its complement another high AspP headed by e.g. -le or -guo. In other words, the unacceptability of (16) is not exclusively due to the incompatibility between the inherent semantics of -le and -guo and that of zài: (16) *他受了伤却仍在坚持{了/过}比赛。 *Tā shòu-le shāng què rēng [AspP zài [AspP jiānchí {-le /-guo} 3sg get-perf wound but still progr uphold -perf /-exp bǐsài] competition By contrast, -zhe as a low (imperfective) aspect is perfectly acceptable in the complement selected by zài, thus lending further support to the fundamental difference between -zhe and the aspectual suffixes zài, -le and -guo. 3 An analysis of the semantic differences between ‘V-zhe’ and a ‘bare’ VP as complement of zài is beyond the scope of this article. Note as a first observation, though, that ‘V-zhe’ seems to be incompatible with a habitual interpretation: (i) Tā zài liàn(*-zhe) pǎo 3sg progr practice-zhe run ‘He is practicing running.’

a new approach to -zhe in mandarin chinese

115

Second, ‘V-zhe’ can also occur in the complement of a modal auxiliary or in the complement selected by the copula shì: (17) a. 我会想着你。 Wǒ [AuxP huì [ xiǎng -zhe nǐ]] 1sg will think -zhe 2sg ‘I will be thinking of you.’ b. 成功往往是靠着毅力。 Chénggōng wǎngwǎng [vP shì [ kào-zhe yìlì]] success often be depend-zhe perseverance ‘Success often depends on perseverance.’ By contrast, an AspP headed by -le or -guo is excluded as complement for alethic modals such as huì ‘will’ and néng ‘can’: (18) 下个星期，他会/能去(*过/*了)故宫。 Xià ge xīngqi tā huì/néng [ qù (* -guo/* -le) gùgōng] next cl week 3sg will/may go -exp/ -perf Imperial.Palace (‘By next week, he will/may have gone to the Imperial Palace.’) (19) 下个星期，他会/能(*在)等你。 Xià ge xīngqi tā huì/néng [(*zài) děng nǐ ] next cl week 3sg will/may progr wait 2sg (‘Next week, he will/may be waiting for you.’) An AspP headed by -le or -guo is, however, acceptable as complement of epistemic modals such as yīnggāi ‘must (be the case), should be’: (20) 他现在应该到了机场。 Tā xiànzài yīnggāi [ dào -le jīchǎng] 3sg now must arrive -perf airport ‘He must have arrived at the airport by now.’ (21) 这件事他应该提过。 Zhèi jiàn shì, tā yīnggāi tí-guo this cl matter 3sg must mention-exp ‘This matter, he must have mentioned it.’

116

djamouri and paul

These data confirm the fundamental difference between the high aspectual suffixes zài, -le and -guo, on the one hand, and -zhe, on the other. The high vs. low aspect distinction also explains why -zhe can be selected as complement by huì ‘will’, whereas this is excluded for AspPs headed by zài, -le or -guo. Third, ‘V-zhe’ is also acceptable as predicate in a complement clause selected by a higher verb (cf. (22)) as well as in a secondary predicate (23):4 (22) 我刚才看见小李打着游戏机。 Wǒ gāngcái kànjiàn [Xiǎo Lǐ dǎ -zhe yóuxìjī] 1sg just see Xiao Li play -zhe playstation ‘I saw Xiao Li playing with his playstation just a moment ago.’ (23) 这里有很多人站着。 Zhèli [matrix-vP yǒu hěn duō rén [zhàn -zhe]] here have very much people stand -zhe ‘There are a lot of people standing here.’ Last, but not least, the head licensing ‘V-zhe’ can also be a functional category, i.e. a complementizer, realized as a sentence-final particle in Chinese and selecting the tp containing ‘V-zhe’ as its complement. (For a detailed analysis of the split cp in Mandarin, cf. Paul (2009)): (24) 外边下着雨*(呢)！ [cp[tp Wàibiān xià-zhe yǔ ] *(ne)] outside fall-zhe rain comp ‘Outside the rain is falling!’ lü shuxiang et al. 2000: 666

(25) 门开着*(来着)。 [cp [tp Mén kāi -zhe] *(láizhe)] door open -zhe comp ‘(But) the door was open just a moment ago.’ zhu dexi 1982: 209

4 For extensive evidence showing that the second VP in existential constructions of the type illustrated in (23) has to be analysed as a secondary predicate, cf. James C.-T. Huang (1987).

a new approach to -zhe in mandarin chinese

117

(26) 我们已經舒舒服服地坐着*(呢)。 [cp[tp Wǒmen yǐjīng shūshufúfude zuò-zhe] *(ne)] 2pl already comfortably sit-zhe comp ‘We are already comfortably seated.’ To summarize, this section has discussed the case where -zhe as low aspect is unacceptable as matrix predicate and needs to be licensed by a higher (lexical or functional) head. It thus contrasts with zài, -le and -guo which are higher aspectual heads themselves and accordingly severely constrained in embedded contexts.

4

Locative Inversion: ‘PlaceP V-zhe NP’

The locative inversion construction is one case where ‘V-zhe’ is acceptable in a matrix context. This is due to the semantico-syntactic properties of this construction, which are shown to be compatible with the unbounded nature of the event expressed by ‘V-zhe’. The defining semantic characteristic of the locative inversion construction as a subtype of existential sentences is to be a thetic proposition in the sense of Kuroda (1972), i.e. the event and the event participants are introduced and asserted simultaneously. In the syntax of English and Chinese, this is reflected by the fact that a PlaceP obligatorily occupies the subject position, that the agent is absent and the theme argument realized within the verbal projection, i.e. in postverbal position. (27) 墙上挂着一幅油画。 [Qiáng shàng]/*∅ guà-zhe yī fú yóuhuà wall on hang-zhe one cl oil.painting ‘*(On the wall) hangs an oil painting.’ Importantly in Chinese besides -zhe, the perfective aspect suffix -le is likewise acceptable in the case of transitive positional verbs: (28) 墙上挂了一幅油画。 Qiáng shàng guà-le yī fú yóuhuà wall on hang-perf one cl oil.painting ‘On the wall has been hung an oil painting.’

118

djamouri and paul

This observation has led to the erroneous assumption of an “alternation” between -zhe and -le in the locative inversion, which in turn is based on the incorrect analysis of -zhe as a high aspect marker. We will not discuss the various ramifications of the idea that there is a free “alternation” between -zhe and -le here, but concentrate on demonstrating the fundamental differences between the two and thereby further elucidate the characteristics associated with -zhe. First, in the locative inversion with ‘V-zhe’, the agent role is lacking, in contrast to ‘V-le’. As a result, (29) and (30) are rejected with -le instead of -zhe, because this would require the implausible interpretation that the moon is in the sky and the persimmons are in the tree due to some prior human action:5 (29) 树梢上挂着/*了一轮明月 Shùshāo shàng guà -zhe /*-le yī lùn míng yuè tree.top on hang -zhe /-perf one cl bright moon ‘Above the tree top hangs a bright moon.’ hu wenze 1995: 106, (34)

(30) 细细的枝条上挂着/*了绿色的柿子。 Xìxì de zhītiáo shang guà -zhe /*-le lǜsè de shìzi thin sub branch on hang -zhe /-perf green sub persimmon ‘Green persimmons hung from thin branches.’ (Jaxontov’s translation) ‘On thin branches hang green persimmons.’ (our translation) jaxontov 1988: 132, (78)

As a consequence, there is no corresponding structure (31) with an explicit agent for ‘V-zhe’, while there is one in the case of ‘V-le’ (cf. (32)): (31) *墙上他挂着一幅油画。 *Qiáng shàng tā guà-zhe yī fú yóuhuà wall on 3sg hang-zhe one cl oil.painting (32) 墙上他挂了一幅油画。 Qiáng shàng tā guà-le yī fú yóuhuà wall on 3sg hang-perf one cl oil.painting ‘On the wall he has hung an oil painting.’

5 This is the reason why, unlike (29), (30) may be accepted under a reading where the persimmons have been put into the tree, as an element of decoration, for example.

a new approach to -zhe in mandarin chinese

119

Furthermore, in contrast to -le, -zhe signals the absence of an end point limiting the event, as witnessed by its incompatibility with temporal adverbs such as zǎojiù ‘long since’, yòu ‘again’ etc. (33) 墙上(*已经/*早就/*又)挂着一幅油画。 Qiáng shàng (*yǐjīng/ *zǎojiù /*yòu) guà -zhe yī fú wall on already/ long.since /again hang -zhe one cl yóuhuà oil.painting ‘On the wall hangs an oil painting.’ (34) 墙上(已经/早就/又)挂了一幅油画。 Qiáng shàng (yǐjīng /zǎojiù /yòu) guà-le yī fú wall on already /long.since /again hang-perf one cl yóuhuà oil.painting ‘On the wall has (already/again/long since) been hung an oil painting.’ By contrast, adverbs such as bìngpái ‘side by side’ or yīzhí ‘always’ which do not imply temporal end points are compatible with ‘V-zhe’: (35) 墙上並排/一直挂着几幅油画。 Qiáng shàng bìngpái/ yīzhí guà-zhe jǐ fú yóuhuà wall on side.by.side/ always hang-zhe several cl oil.painting ‘On the wall (always) hang several oil paintings (side by side).’ Last, but not least, the alleged alternation between -zhe and -le in the locative inversion is restricted to transitive positional verbs and does not hold for intransitive positional verbs such as zuò ‘sit’, tǎng ‘lie’ where only -zhe is possible (cf. C.-T. James Huang 1987: 228 and references therein): (36) 沙发上坐着/*坐了两个小孩儿。 Shāfa shàng zuò-zhe/* zuo-le liang ge xiǎoháir sofa on sit-zhe/ sit-perf two cl child ‘On the sofa sit two children.’ This confirms the fundamental difference between -le and -zhe; in other words, the observed “alternation” is nothing but a surface phenomenon. To summarize, ‘V-zhe’ is acceptable as matrix predicate in locative inversion, because this construction is a type of thetic proposition, realized here as an

120

djamouri and paul

unbounded existential construction. Furthermore, in the locative inversion with -zhe, no agent role is present, neither overt nor covert. As to be discussed in the following section, the unbounded nature of the event and the lack of an agent likewise allow ‘V-zhe’ as a matrix predicate of descriptive sentences.

5

‘S V-zhe (O)’ Providing a Description (hic et nunc or gnomic)

We note as a first approximation that -zhe has the effect of ascribing the event as a property to the subject, the latter thus being more of a theme than an agent. Unlike an event, a property is not intrinsically defined in terms of boundedness. This gives rise to two interpretations for ‘S V-zhe (O)’ sentences: a hic et nunc interpretation and a gnomic interpretation. 5.1 Hic et nunc Interpretation A description is provided (rather than an activity being reported) and presented as holding at the utterance time: (37) 他戴着一頂红帽子。 Tā dài-zhe yī dǐng hóng màozi 3sg wear-zhe one cl red hat ‘He wears a red hat.’ (38) 人们跳着，唱着。 Rénmen tiào-zhe, chàng-zhe people dance-zhe sing-zhe ‘People dance and sing.’ As already pointed out by Teng Shou-hsin (1973: 21, footnote 6), sentence (37) cannot be interpreted as ‘He is putting on a red hat’; this meaning must be rendered using zài as in (39): (39) 他在戴一頂红帽子。 Tā zài dài yī dǐng hóng màozi 3sg progr wear one cl red hat ‘He is putting on a red hat.’ More examples of the same type as (37) are given below, which also illustrate the crucial role adverbs may play in strengthening the interpretation in terms of a property ascribed rather than an activity reported:

a new approach to -zhe in mandarin chinese

121

(40) 他*(高高兴兴地)唱着歌。 Tā *(gāogāoxìngxìngde) chàng-zhe gē 3sg happily sing-zhe song ‘He sings happily.’ (41) 他不停地在本子上记着什么。 Tā bùtíngde zài běnzi shàng jì-zhe shénme 3sg constantly at notebook on note-zhe anything ‘He is constantly making notes in his notebook.’ 5.2 Gnomic Interpretation Given a context or adverbs incompatible with a hic et nunc interpretation, ‘V-zhe’ in a matrix context has the effect of ascribing an event as a generally valid property to the subject (gnomic interpretation). (42) 我永远等着你。 Wǒ yǒngyuǎn děng-zhe nǐ 3sg forever wait-zhe 2sg ‘I’ll wait for you forever.’ (43) 他通年雇着三个长工。 Tā tōngnián gù-zhe sān ge chánggōng 3sg all.year hire-zhe three cl workman ‘He keeps three hired hands (farm labourers) the year round.’ jaxontov 1988: 128

(44) 要不然這麼著得了。 Yàobùrán zhème-zhe dé le otherwise be.so-zhe be.enough sfp ‘Otherwise, let it be like this and that’s it.’ To summarize, the hic et nunc or the gnomic interpretations of ‘V-zhe’ represent the default interpretations obtained from the fundamentally unbounded nature of the event signaled by -zhe.

6

Conclusion

We have argued that -zhe is a low (imperfective) aspect marker and that moreover it signals the dependent status of the verbal projection concerned. This

122

djamouri and paul

explains why a significant part of the data provided to illustrate the use of -zhe involves embedded contexts (adjunct clauses, complements to a higher head etc.). This fact has not really been considered significant in previous studies, nor has there been any attempt to explicitly incorporate it into the analysis of -zhe. Upon reflection, if -zhe were really on a par with the high aspect markers zai, -le and -guo, its prominent presence in non-matrix contexts would in fact appear rather incongruous, given the well-known observation that in Chinese aspect markers are often optional in matrix contexts and severely constrained in non-matrix contexts. When in a matrix clause, ‘V-zhe’ has the effect of ascribing the event as a property to the subject. This entails a present, hic et nunc, or a gnomic interpretation, depending on the temporal and spatial anchoring of the utterance (in presentia vs in absentia). A property interpretation is also available in the locative inversion where ‘V-zhe’ presents an agentless predicate. Last, but not least, from a diachronic point of view it is interesting to note that ‘V-zhe’ in the locative inversion is attested later than ‘V-zhe’ in embedded contexts (cf. Djamouri 2010, 2011). In other words, -zhe started out as a ‘structural marker’ (结构助词 jiégòu zhùcí) of dependency and only subsequently gave rise to the unbounded interpretation observed in locative inversion and in ‘S V-zhe O’ sentences.

References Chen, Chung-yu 1986. ‘Constraints on the ‘v1-zhe … v2’ structure’. Journal of the Chinese Language Teachers’ Association 21, pp. 1: 1–20. Chen Chung-yu 1978. ‘Aspectual features of the verb and the relative positions of the locatives.’ Journal of Chinese Linguistics 6, 1, pp. 76–103. Chen, Zhong 2009. ‘Zhe yu zhèng, zài de tihuan tiaojian ji qi liju [The substitution condition of zhe for zhèng, zài and its motivation]’. Yuyan jiaoxue yu yanjiu 2009, nr. 3, pp. 81–88. Djamouri, Redouane 2010. ‘“著”重新分析的過程 [The reanalysis of -zhe]’. Paper presented at the Seventh International Symposium on Ancient Chinese Grammar (isacg-7), Roscoff, France, September 17–18, 2010. Djamouri, Redouane 2011. ‘“著”在元朝之前的歷時發展 [The historical evolution of zhe before the Yuan]’. Paper presented at the 19th Annual Conference of the International Association of Chinese Linguistics (iacl–19), Nankai University, Tianjin, June 11–13, 2011. Hu, Wenze 1995. ‘Verbal semantics of presentative sentences’. Yuyan Yanjiu 1995, nr. 2, pp. 100–112.

a new approach to -zhe in mandarin chinese

123

Huang, C.-T. James 1987. ‘Existential sentences in Chinese and (in)definiteness’. In: Reuland, Eric J. and Alice G.B. ter Meulen (eds.). The Representation of (In)definiteness. Cambridge, Mass.: mit Press, pp. 226–253. Jaxontov, Sergei J. 1988. ‘Resultative in Chinese’. In: Nedjalkov, Vladimir P. (ed.). Typology of Resultative Constructions. Amsterdam & Philadelphia: John Benjamins, pp. 113– 133. Kuroda, S.-Y. 1972. ‘The categorical and the thetic judgement.’ Foundations of Language 9, pp. 153–185. Lü, Shuxiang et al. (eds.) 2000. Xiandai hanyu babaici [800 words of Modern Chinese]. Beijing: Shangwu yinshuguan. Paris, Marie-Claude 1988. ‘L’expression de la durée en mandarin’. In: Tersis, Nicole and Alain Kihm (eds.). Temps et aspects. Actes du colloque cnrs Paris, 24–25 Octobre 1985. Paris: Peeters, pp. 163–177. Paul, Waltraud 2009. ‘Consistent disharmony: Sentence-final particles in Chinese.’ Unpublished ms., crlao, Paris. http://crlao.ehess.fr/index.php?177. Paul, Waltraud 2014. ‘Why particles are not particular: Sentence-final particles as heads of a split cp’. Studia Linguistica 68.1: 77–115. Teng, Shou-hsin 1973. ‘Negation and aspects in Chinese’. Journal of Chinese Linguistics 1, 1, pp. 14–37. Travis, Lisa deMena 2010. Inner Aspect. The Articulation of vp. Dordrecht: Springer. Zhu, Dexi 1982. Yufa jiangyi [On grammar]. Beijing: Shangwu yinshuguan.

chapter 12

Japanese Experiential -te iru Mamori Sugita Hughes and William McClure

1

Introduction

The Japanese -te iru construction is known to have at least four different readings, as illustrated by the examples in (1). (1) -te iru a. Progressive (sinkoo) Mari-ga ima oyoi-de iru ‘Mari is swimming now.’ b. Resultative (kekka) Mari-ga ima igirisu-ni it-te iru ‘Mari (went to England and) is in England now.’ c. Experiential (keeken) Mari-wa ima-made-ni sankai kono kawa-de oyoi-de iru ‘Up to now, Mari has swum in this river three times.’ d. Habitual (syuukan) Kono goro, Mari-wa maiasa oyoi-de iru ‘These days, Mari is swimming every morning.’ The progressive/resultative dichotomy in (1a&b) is the most studied phenomenon, and analyses and discussion can be found in, for example, Kindaichi (1976), Jacobsen (1992), McClure (1995), Ogihara (1998, 1999), Shirai (2000), and Kusumoto (2003), to name just a few. Less studied are the experiential and habitual readings in (1c&d). Experiential readings are addressed in, amongst others, Fujii (1966), Soga (1983), Machida (1989), and Ogihara (1999). Similarly for habitual readings which are mentioned briefly in Fujii (1966), Kindaichi (1976), Yoshikawa (1976), Teramura (1984), and Shirai (2000). These discussions are typically descriptive, and no formal account of either reading appears in the literature. The goal of this paper is to offer a fuller accounting of experiential te iru. As turns out, we believe that this analysis gives insight into habitual -te iru as well.

© koninklijke brill nv, leiden, 2018 | doi: 10.1163/9789004351134_013

japanese experiential -te iru

2

125

Experiential -te iru

Examples of experiential -te iru are given in (2).1 Contrasting the examples in (2) and (3), it is clear that experiential -te iru is not sensitive to the same aspectual distinctions that characterize the distribution of progressive/resultative -te iru (Vendler 1967, Kindaichi 1976). Oyogu ‘swim’ in the (a)-examples is an activity, while iku ‘go’ in the (b)-examples is an achievement. The -te iru reading in (3a) is progressive, while in (3b) it is resultative. (2) a. Mari-wa ima-made-ni sankai kono kawa-de oyoi-de iru.2 ‘Up to now, Mari has swum in this river three times.’ b. Mari-wa ima-made-ni sankai igirisu-ni it-te iru. ‘Up to now, Mari has been to England three times.’ c. Mari-wa kyonen kono kawa-de oyoi-de iru. ‘Mari swam in this river last year.’ (3) a. Mari-ga ima kawa-de oyoi-de iru. ‘Mari is swimming in the river now.’ b. Mari-ga ima igirisu-ni it-te iru. ‘Mari (has gone to England and) is in England now.’ 1 The adverbials ima-made-ni sankai ‘three times up to now’ and kyonen ‘last year’ create a context that accommodates the experiential meaning, but adverbials are just one (relatively efficient) linguistic device that can perform this function. They are not the only such device. Consider the following discourse: Tanaka-san-wa gaikoku-ni itta-koto-ga arimasu ka? Ee, arimasu yo. Yooroppa-ni nandomo. Soo desu ka. Zyaa, furansu-ni-mo doitu-ni-mo … Soo desu ne. Furansu-ni-wa it-te imasu ga, doitu-ni-wa mada desu. Igirisu-ni-mo it-te inai no desu ga, kotosi ikoo ka to omotte iru n desu. Tanaka, have you ever gone abroad? Yes, I have. (I’ve gone) to Europe, many times. Really. So, (you’ve gone) to France and to Germany … Well, I have gone to France (itte imasu) but, not yet to Germany. I have also not gone to England (itte inai) but, I’m thinking of going this year. Here, the entire dialog is about travels abroad. In this context, the various -te iru forms are unambiguously experiential. In this paper, we use adverbials to create clear contexts; they are not the source of the experiential meaning. 2 The use of -ga in an experiential -te iru sentence forces the subject to receive an exhaustive reading. To maintain naturalness, such a sentence would therefore require additional context. As such, the subject in the examples here is marked with -wa and the interpretation is that of a normal topic (with no additional context required). We return to this fact in section 3.3.

126

hughes and mcclure

Before moving forward, we should also consider the class of accomplishment predicates. Beginning with Vendler’s (1967) discussion of aspectual classes in English, most theories of accomplishments consider them to be a hybrid class, with properties of both activities and achievements. In contrast, in his (independent) analysis of aspectual classes in Japanese, Kindaichi (1976) does not posit a class of accomplishments for Japanese (and the classification of a predicate such as keeki-o tukuru ‘make a cake’ is not addressed by Kindaichi). Does the experiential reading offer any insight into this question? Consider the example in (4a). (4) a. Mari-ga keeki-o tukut-te iru. b. Mari-ga ima keeki-o tukut-te iru. c. Mari-wa ima-made-ni sankai keeki-o tukut-te iru. This sentence has two possible readings. One is clearly progressive, but what is the second? Is it resultative or experiential? Consider the examples in (3b) and (4b). (3b) is an achievement with resultative -te iru. The addition of ima ‘now’ serves to emphasize the outcome state, i.e. the fact that Mari is in England at this very moment. In contrast, the addition of ima in (4b) emphasizes the progressive reading; (4b) is unambiguously about the on-going making of a cake. Note that while there is a possible outcome state, i.e. the existence of a cake, (4b) does not entail the existence of that cake.3 Turning finally to (4c), the addition of ima-made-ni sankai ‘three times up to now’ results in a reading that is unambiguously experiential. At some point in the past, Mari has made three cakes. These facts suggest that Kindaichi was correct not to posit a class of accomplishments for Japanese. The addition of a direct object to an activity verb does not change the aspectual properties of the activity. In particular, the -te iru construction of an activity, with or without a direct object, may be progressive or experiential. It is, however, never resultative.4 Having reviewed the interaction of -te iru with the traditional aspectual classes,5 we must still address the question of what an experiential reading

3 In fact, there is no reading under which (4b) entails the current existence of a cake. Even the experiential reading entails existence at some point in time, but not necessarily at the moment of speech. 4 This also suggests that keeki-o tukuru might be more accurately translated as ‘do some cakemaking’ rather than as ‘make a cake’. 5 We have actually ignored two further classes posited by Kindaichi. Statives such as iru ‘need’ are incompatible with the -te iru construction, much as the progressive in English is incompatible with statives (Vendler 1967). Kindaichi also argues for a class of predicates that

japanese experiential -te iru

127

actually is. Soga (1983) argues that the experiential reading focuses attention on a span of time, while the resultative and progressive readings focus on a point of time. In other words, the former is concerned with the state of affairs prevailing up to the moment of speech, while the latter is concerned with the state of affairs at the moment of speech. However, it is not exactly clear what the state of affairs prevailing up to the moment of speech means. Moreover, it must be noted that the experiential sentences in (2) are in the non-past. While experiential sentences seem to be about the past, they are realized grammatically in the non-past. This apparent contradiction is laid bare in example (2c), which includes kyonen ‘last year’, an adverbial making explicit reference to the past. While the sentence is in the non-past, its interpretation is (unambiguously) experiential.6 The intuition we will formalize is that, while an experienced event must have already occurred, an experiential sentence is actually about the fact of having had the experience. Experiential sentences are therefore complex in that they include reference to an event, but they are themselves presented as facts. This fact, that is, the fact of having had an experience, is a property of an individual. Moreover, it is a permanent property, one that cannot be undone. We claim that experiential sentences are actually individual level predicates (ilps) (Milsark 1974, Carlson 1977, Kratzer 1995). Evidence for this claim is given in section 3.

3

Experiential -te iru is an ilp

We discuss three grammatical properties that contrast experiential -te iru and progressive/resultative -te iru. With respect to all three properties, experiential -te iru patterns with individual level predicates.

necessarily occur in the -te iru construction (and do not, for example, occur in the simple past or simple non-past). Moriyama (1988) and others (McClure 1995, Ogihara 1998) argue that these predicates are in fact all achievements. As such, they are not given a separate treatment in our discussion. 6 As indicated by the translation, the mismatch in (2c) is not allowed in English. However, Heidi Harley (p.c.) has stated that in German, as in Japanese, an experiential sentence can be marked non-past even with the presence of a past modifier. An example is: Ich bin letztes Jahr in diesem Fluss geschwommen I am last year in this river swum ‘I swam in this river last year.’ (Thanks are due to Tamara Evans for the example and for further, clarifying discussion.)

128

hughes and mcclure

3.1 Temporal Modifiers Kratzer (1995) claims that ilps cannot be modified by temporal adverbs. As ilps are permanent properties, they are generally odd with any indication of temporariness. In the specifics of her analysis, she accounts for this fact by claiming that ilps lack an event argument. Japanese examples are found in (5). (5) a. #Mari-wa kyonen se-ga takakatta. ‘Mari was tall last year.’ b. #Mari-wa isyuukan dake amerikazin datta. ‘Mari was an American for just one week.’ Unless Mari is a character in Wonderland and her height or citizenship changes from moment to moment, the examples in (5) are pretty odd. Turning to experiential sentences, while it is not the case that such sentences may not contain a temporal modifier, it is the case that the modification is always of the event experienced, not of the fact of having had the experience. (6) a. Mari-wa kyonen kono kawa-de oyoi-de iru. [= (2c)] ‘Mari swam in this river last year.’ b. Mari-wa ima-made-ni itiziteki-ni kono kawa-de oyoi-de iru. ‘Mary has temporarily swum in this river up to now.’ As already discussed, there is no agreement between the adverbial in (6a) and the non-past form of the matrix predicate. What was experienced in (6a) is swimming-in-a-river-last-year. The fact of that experience continues to obtain. Likewise, in (6b), what was experienced was swimming-in-a-river-temporarily. The fact of that experience continues to obtain. In both examples in (6), and as will become clear below, we will say that the temporal adverb has scope over the experienced event without having scope over the matrix predicate. This will mean that experiential -te iru is really an individual level construction, not a simple individual level predicate like the examples in (5). 3.2 Tokoro da Tokoro da follows a clause to signify immediacy of action.7 It is very much focused on a change of some sort. As shown in (7b&c), it is therefore incompatible with ilps.

7 It actually has at least one other function—forcing a counterfactual interpretation when used in the consequent of conditions (cf. Takubo 2008), but that is not relevant here.

japanese experiential -te iru

129

(7) a. Mari-ga hon-o yomu tokoro da. ‘Mari is just about to read a book.’ b. *Mari-ga se-ga takai tokoro da. Attempted meaning: ‘Mari is tall right at this moment.’ c. *Hon-ga aru tokoro da. Attempted meaning: ‘There is a book right at this moment.’ Again, unless Mari is a character in Wonderland, (7b) is ungrammatical, and (7c) is also strange. As we see in (8), experiential sentences are also weird with tokoro da. (8) a. *Mari-wa ima-made-ni kono kawa-de oyoi-de iru tokoro da Attempted meaning: ‘It is just now the case that Mari has swum in this river up to now.’ b. *Mari-wa ima-made-ni igirisu-ni it-te iru tokoro da. Attempted meaning: ‘It is just now the case that Mari has been to England up to now.’ While it is possible to conceptualize the moment at which an experience begins to obtain, for example, just as soon as it has occurred, the examples in (8) cannot be used to refer to that moment. 3.3 -ga Interpretations The third parallel between an ilp and experiential -te iru concerns the interpretation of particle -ga. This observation is also found in Ogihara (1999). Kuno (1973) makes the well-known claim that particle -ga has two interpretations, a default ‘descriptive’ reading and a marked ‘exhaustive’ reading. While the marked reading can be licensed in a number of different ways, Kuroda (2005 and references therein) makes the specific observation that the exhaustive -ga reading is forced in a sentence with an ilp (p. 25). Contrast (9a&b): (9) a. John-ga Tokyo-ni itta. ‘John went to Tokyo.’ (descriptive -ga) b. John-ga se-ga takai. ‘It is John who is tall.’ (exhaustive -ga) Importantly, the exhaustive reading in (9b) obtains without any special intonation or conversation background. The effect is that it is actually impossible for -ga to have a descriptive reading with an ilp predicate.8 8 This is the case at least within matrix clauses. As is well-known, in embedded clauses, the

130

hughes and mcclure

Turning to experiential sentences, we find that the ilp pattern obtains, i.e. experiential sentences with ga-marked subjects are always exhaustive.9 (10) a. Mari-ga ima-made-ni kono kawa-de oyoi-de iru. ‘It is Mari who has swum in this river up to now.’ (exhaustive -ga) b. Mari-ga ima-made-ni igirisu-ni it-te iru. ‘It is Mari who has been to England up to now.’ (exhaustive -ga) In contrast, the exhaustive reading does not obtain in progressive/resultative te iru sentences (at least not without special intonation or context). Consider (11a&b) with the adverbial ima. (11) a. Mari-ga ima oyoi-de iru. ‘Mari is swimming now.’ (descriptive -ga) b. Mari-ga ima igirisu-ni it-te iru. ‘Mari is in England now.’ (descriptive -ga) Here the interpretations are the usual progressive in (11a) and resultative in (11b).

4

Some Syntax

Having provided evidence that experiential -te iru is an individual level predicate, certain expectations come into play with regards to its syntax. There is, of course, a lengthy literature on the structural differences between slp and ilp predicates, but we focus here on the work of Diesing (1992) and Kratzer (1995). 4.1 Raising versus Control In the literature on Japanese, it has been argued that some v-te-v compounds are raising structures while others are control.10 Raising verbs include kakeru ‘about to’ and sugiru ‘over do’; examples of control verbs include wasureru ‘forget’ and -te oku ‘do in advance’. Sugita (2009) applies two types of diagnostics to determine the complementation structures of the -te iru construction: idiom

9 10

contrast between descriptive and exhaustive -ga no longer holds (and -ga in embedded clauses is always descriptive). It is for this reason that in our discussion of the experiential, we have used examples with wa-marked subjects, cf. footnote 2. For example, see Inoue (1976, 1989), Shibatani (1978), Kuno (1983), Kageyama (1993), Nishigauchi (1993), or Matsumoto (1996).

japanese experiential -te iru

131

chunk insertion and passivization (pp. 189–201). The results of the passivization test are reviewed here. The test indicates that progressive/resultative -te iru is a raising structure while experiential -te iru is a control structure. 4.2 Passivization A raising verb shows a voice transparency in that the passive and active counterparts are synonymous. This is not the case with control verbs (Davies & Dubinsky 2004). Contrast the examples in (12a&b). (12) a. Raising The doctor seems to have examined Mary. Mary seems to have been examined by the doctor. b. Control The doctor forgot to examine Mary. Mary forgot to be examined by the doctor. While the examples in (12a) differ in voice, they are about the same event, where the doctor examines Mary. In contrast, forget is a control verb, taking different external arguments in the active voice and passive voice. As a result, the sentences in (12b) differ significantly. In the first, it is the doctor who forgets, and in the second it is Mary. The same kind of difference can be observed in Japanese. The compounds in (13) include the raising predicate kakeru ‘about to’, while the compounds in (14) include the control predicate wasureru ‘forget’. (13) Raising a. Active Isya-ga Mari-o sinsatusi kaketa. ‘The doctor was about to examine Mari.’ b. Passive Mari-ga isya-ni (yotte) sinsatusare kaketa. ‘Mari was about to be examined by the doctor.’ (14) Control a. Active Isya-ga Mari-o sinsatusi wasureta. ‘The doctor forgot to examine Mari.’ b. Passive Mari-ga isya-ni (yotte) sinsatusare wasureta. ‘Mari forgot to be examined by the doctor.’

132

hughes and mcclure

As in English, the examples in (13) differ in voice but are about the same event. In contrast, the examples in (14) are about different events, one where the doctor forgets and one where Mari forgets. The examples in (15) and (16) show that the active and passive versions of progressive/resultative -te iru preserve the underlying meaning, while the examples in (17) show that experiential -te iru does not. Specifically, the examples in (17) and (18) have different truth conditions. (15) Progressive -te iru a. Active Mari-ga ima hon-o yon-de iru. ‘Mari is reading a book now.’ b. Passive Hon-ga ima Mari-ni (yotte) yomare-te iru. ‘A book is being read by Mari now.’ (16) Resultative -te iru a. Active Mari-ga ima hon-o kari-te iru. ‘Mari (borrowed a book and) has a book now.’ b. Passive Hon-ga ima Mari-ni (yotte) karirare-te iru. ‘A book has been checked-out by Mari now.’ (17) Experiential -te iru a. Active Mari-wa ima-made-ni sankai hon-o yon-de iru. ‘Up to now, Mari has read a/the book three times.’ a’ Passive Hon-wa ima-made-ni sankai Mari-ni (yotte) yomare-te iru. ‘The book has been read by Mari three times up to now.’ b. Active Mari-wa ima-made-ni sankai hon-o kari-te iru. ‘Up to now, Mari has read a/the book three times.’ b’ Passive Hon-wa ima-made-ni sankai Mari-ni (yotte) karirare-te iru. ‘The book has been borrowed by Mari three times up to now.’ The pairs in (17a&a’) and (17b&b’) differ in meaning in that hon-o may be interpreted as definite or indefinite in the active examples, while it must be

japanese experiential -te iru

133

definite in the passive examples. That is, in the active sentences Mary may have read or borrowed the same book repeatedly, or she may have read or borrowed different books. In the passive sentences, it must be the same book that is read or borrowed three times. No such difference in interpretation is found in the examples in (15) and (16). The conclusion is that progressive/resultative -te iru has a raising structure while experiential -te iru is a control structure. 4.3 slp/ilp, raising/control With respect to the syntactic structures of slp and ilp predicates, both Diesing (1992) and Kratzer (1995) start with the observation that (in German at least), the subject of an slp is base-generated in Spec vp while the subject of an ilp is base-generated in SpecIP, that is, at a higher location in the syntax. In addition, Diesing explicitly associates an slp with a raising infl. Under her analysis, an slp subject is base-generated in SpecVP. It raises to SpecIP for case (but it may move back down to Spec vp). In contrast, an ilp subject is base generated in SpecIP where it controls a pro in SpecVP. This distribution is captured in Diesing’s well-known Mapping Hypothesis. (18) Mapping Hypothesis (Diesing 1992:10) Material from vp is mapped into the nuclear scope. Material from ip is mapped into the restrictive clause. Kratzer accounts for the difference in slp and ilp subject position by arguing that only slps include an event argument in their representations. The event argument is external, while the remaining (nominal) arguments, including the subject, are internal. The analysis in this paper assumes an event argument, but it will differ from Kratzer in the possible location of that argument. In particular, we will propose that an experiential -te iru construction includes both an individual variable and an event variable, but, in contrast to an slp, in experiential -te iru the individual has scope over the event.

5

An Analysis

We now present a proposal to account for the fact that experiential -te iru constructions have an ilp interpretation and a control structure. In addition, our proposal will account for the fact that experiential sentences are about the fact of having had an experience although they make reference to the experienced event as well. Recall that this dichotomy is captured most obviously in an example like (19), repeated from above.

134

hughes and mcclure

(19) Mari-wa kyonen kono kawa-de oyoi-de iru. [= (2c)] ‘Mari swam in this river last year.’ Although the sentence includes kyonen, an adverbial making explicit reference to the past, the matrix predicate oyoi-de iru is in the non-past. The sentence is about the (currently relevant) fact of Mari having had an experience; the event experienced was a swimming-in-this-river-last-year. We propose (20) as the syntactic and semantic representations of experiential -te iru. (20) Experiential -te iru a. [ip Mari(x)i [vp proi oyoi-de iru]] b. ∃x[ip Mari’(x) ∧ past ∃e[vp swim’(e) ∧ Subject(e, x) ∧ R(e)]] (20a) is a control structure with the subject mapped into the Spec of ip; experiential -te iru is an ilp. (20b) maps the semantic interpretations into syntax.11 Note that (20b) differs from Kratzer’s claim that individual level predicates do not include an event argument in their representations. (20b) does include an event variable, but, crucially, it is mapped into the nuclear scope. Moreover, this event is itself under a past tense operator. The experienced event precedes the speech time of the experiential sentence itself. Note that kyonen in (19) corresponds exactly to the position of the tense operator. This configuration allows us to capture the intuition that experiential -te iru is about the fact of having had an experience although it includes reference to the experienced event as well. It says that Mari is (currently) the subject of a past event of swimming, that is, Mari has swum at some point in the past. This is experiential.12

11

12

r in (20b) is the semantic translation of iru. Semantically, r is based on Carlson’s (1977) notion of stage. A stage is a temporal slice of an individual. It is the ‘part’ of an individual that is actually realized in space during a given temporal interval. Formally, Carlson defines a stage in terms of a predicate r (‘realizes’). r is a “two-place, asymmetric, irreflexive, transitive relation [that] holds between stages and individuals … A formula like R(a, b) means that a is a stage of b” (p. 76). Translating this idea into event semantics, we define r as a one-place predicate defined over the set of events that are realized at a particular place and time. We leave open the metaphysical question of whether an experience obtains as soon as an event obtains. Is the experience of swimming realized as soon as the event of swimming begins? Or is the experience realized only once the swimming has stopped?

japanese experiential -te iru

6

135

A Postscript on Habitual

While the literature on habitual -te iru is complex and well beyond the scope of this paper,13 it can be shown that habitual -te iru is also an ilp.14 The tokoro da and exhaustive -ga interpretation tests are applied to habitual sentences in examples (21) through (23). (21) a. *Mari-wa mainiti seesyo-o yon-de iru tokoro da. Attempted meaning: ‘Mary reads the bible every day right at this moment.’ b. *Mari-wa maiasa hasi-o watate-te iru tokoro da. Attempted meaning: ‘Mary crosses a bridge every morning right at this moment.’ (22) a. Mari-wa mainiti seesyo-o yon-de iru. ‘Mari reads the bible every day.’ b. Mari-ga mainiti seesyo-o yon-de iru. ‘It is Mari who reads the bible every day.’ (exhaustive -ga) (23) a. Mari-wa maiasa hasi-o watat-te iru. ‘Mari crosses a bridge every morning.’ b. Mari-ga maiasa hasi-o watat-te iru. ‘It is Mari who crosses a bridge every morning.’ (exhaustive -ga) We therefore believe that habitual -te iru is a kind of experiential sentence. That is, the difference between experiential -te iru and habitual -te iru is one of degree. Habits must have been repeatedly exemplified in the past, and typically this repetition is on-going (even if no particular event is actually occurring at the moment of speech). If one has a habit, one has experiences, although it is not the case that all experiences must have such characteristics. As such, we conclude that the distinction between experiential and habitual

13

14

Krifka et al (1995) argue that habituals are a special kind of characterizing sentence (or a generic sentence). For Japanese, it is argued that habituals are a special kind of iterative sentence (cf. Fujii 1996, Kindiaichi 1976, Takahashi 1976, Yoshikawa 1976). Yoshikawa (1976), Teramura (1984) and Shirai (2000) then argue that iterative sentences are themselves a kind of progressives. Teramura (1984) also discusses the differences between simple present sentences that make habitual statements and habitual -te iru. It is therefore not a special kind of progressive (see footnote 13).

136

hughes and mcclure

-te iru is pragmatic and not semantic. Habitual readings are particular kinds of experiential readings. While such a proposal contradicts much of the existing literature,15 we believe that the facts in (21) through (23) should be the starting point for an analysis of habitual -te iru.

References Carlson, Gregory 1977. Reference to Kinds in English. PhD dissertation, University of Massachusetts, Amherst. Davies, William D. & Stanley Dubinsky 2004. The Grammar of Raising and Control: A Course in Syntactic Argumentation. Blackwell Publishers, Oxford. Diesing, Molly 1992. Indefinites. mit Press, Cambridge. Fujii, Takashi 1966. ‘Dōshi + te iru no imi [The meaning of verb + te iru]’. In Kokugo Kenkyūshitsu, 5. Tokyo University, Tokyo. Inoue, Kazuko 1976. Henkee Bunpoo to Nihongo [Transformational Grammar and Japanese]. Taishukan, Tokyo. Inoue, Kazuko 1989. ‘Shugo no imi-yakuwari to kaku-hairetsu [The semantic roles of subject and case alignment]’. In Kuno, Susumu & Masayoshi Shibatani (eds.), Nihongogaku no Shintenkai [New Developments in Japanese Linguistics]. Kuroshio, Tokyo, pp. 79–101. Jacobsen, Wesley 1992. The Transitive Structure of Events in Japanese. Kuroshio, Tokyo. Kageyama, Tarō 1993. Bunpoo to Gokeesee [Grammar and Word Formation]. Hitsuji Shobō, Tokyo. Kindaichi, Haruhiko 1976. Nihongo Dōshi no Asupekuto [Aspect in Japanese Verbs]. Mugi Shobō, Tokyo. Kratzer, Angelika 1995. ‘Stage-level and individual-level predicates’. In Gregory N. Carlson & Francis Jeffry Pelletier (eds.), The Generic Book. University of Chicago Press, Chicago, pp. 125–175. Krifka, Manfred et al 1995. ‘Genericity: an introduction’. In Gregory N. Carlson & Francis Jeffry Pelletier (eds.), The Generic Book. University of Chicago Press, Chicago, pp. 1– 124. Kuno, Susumu 1973. The Structure of the Japanese Language. mit Press, Cambridge. Kuno, Susumu 1983. Shin Nihon Bunpō Kenkyū [New Studies in Japanese Grammar]. Taishūkan, Tokyo. Kuroda, S.-Y. 2005. ‘Focusing on the matter of topic: a study of wa and ga in Japanese’. Journal of East Asian Linguistics 14, pp. 1–58.

15

See footnote 13.

japanese experiential -te iru

137

Kusumoto, Kiyomi 2003. ‘The semantics of -teiru in Japanese’. Japanese/Korean Linguistics 11. csli, Palo Alto. Machida, Ken 1989. Nihongo ni Jisei to Asupekuto [ Japanese Tense and Aspect]. Aruku, Tokyo. Matsumoto, Yo 1996. Complex Predicates in Japanese: A Syntactic and Semantic Study of the Notion “Word.” csli, Palo Alto. McClure, William 1995. Syntactic Projections of the Semantics of Aspect. Hitsuji Shobō, Tokyo. Milsark, Gregory 1974. Existential Sentences in English. PhD dissertation, mit. Moriyama, Takurō 1988. Nihongo Dōshi Jutsugobun no Kenkyū [Investigations of Japanese Verbal Predicates]. Meiji Shoin, Tokyo. Nishigauchi, Taisuke 1993. ‘Long distance passive’. In Nobuko Hasegawa (ed.), Japanese Syntax in Comparative Grammar. Kuroshio, Tokyo, pp. 79–114. Ogihara, Toshiyuki 1998. ‘The ambiguity of the -te iru form in Japanese’. Journal of East Asian Linguistics 7, pp. 87–120. Ogihara, Toshiyuki 1999. ‘Tense and aspect’. In Natsuko Tsujimura (ed.), The Handbook of Japanese Linguistics. Blackwell, Oxford, pp. 326–348. Shibatani, Masayoshi 1978. Nihongo no Bunseki [Analysis of Japanese]. Taishukan, Tokyo. Shirai, Yasuhiro 2000. ‘The semantics of the Japanese imperfective -teiru: an integrative approach’. Journal of Pragmatics, 32, pp. 327–361. Soga, Matsuo 1983. Tense and Aspect in Modern Colloquial Japanese. University of British Columbia Press, Vancouver. Sugita, Mamori 2009. Japanese -te iru and -te aru: The Aspectual Implications of the Stage-level and Individual-level Distinction. PhD dissertation, cuny. Takahashi, Tarō 1976. ‘Sugata to mokuromi [Aspect and purpose]’. In Haruhiko Kindaichi (ed.), Nihongo Dōshi no Asupekuto [Aspect in Japanese Verbs]. Mugi Shobō, Tokyo. Takubo, Yukinori 2008. ‘Tense and aspect in Japanese: the case of tokoro-da as a reference point marker’. Journal of the Association of Japanology in East Asia, 25, pp. 5–20. Teramura, Hideo 1984. Nihongo no Shintakusu to Imi [The Syntax and Semantics of Japanese. Kuroshio, Tokyo. Vendler, Zeno 1967. Linguistics in Philosophy. Cornell University Press, Ithaca. Yoshikawa, Taketoki 1976. ‘Gendai nihongo dōshi no asupekuto no kenkyū [Study of aspect in modern Japanese verbs]’. In Haruhiko Kindaichi (ed.), Nihongo Dōshi no Asupekuto [Aspect in Japanese Verbs]. Mugi Shobō, Tokyo, pp. 155–327.

chapter 13

dp versus np: A Cross-Linguistic Typology? Jaklin Kornfilt

1

Introduction

In a series of studies, Bošković (e.g. 2008, 2012, 2013) proposes a linguistic typology based on a posited dichotomy between languages whose “traditional” nps are actually dps and languages where the relevant projection does not go beyond the level of np. (This would challenge, among others, the proposal in Abney (1987), according to which all languages have dps.) One immediate clue for the relevant type of a language in this respect would be whether it has articles or not. More interestingly, Bošković proposes additional properties which a language would or would not exhibit, depending on whether it is an “np-” or a “dp-” language (e.g. np-languages disallow clause-mate npi licensing under Neg-Raising (nr), and dp-languages allow it; only dp-languages allow the majority superlative reading; inverse scope is unavailable in np-languages). In a related study, Bošković & Şener (2014) claim that Turkish is an np-language, and that it therefore exhibits the properties which Bošković’s system would ascribe to it. They further posit a structure of the np from which the relevant properties of Turkish would follow. In this paper, I shall challenge both some of the details proposed for the Turkish np and the posited correlation between the np/dp “typology” and the properties which are claimed to be found in “dp-” versus “np-” languages, and illustrate my criticism via examples mainly from Turkish, but also from German and English.

2

np/dp Generalizations

In this section, I discuss some of the generalizations proposed in work by Bošković (2012, 2013) and by Bošković & Şener (2014) with respect to properties of “np-” versus “dp-languages”. I start with an observation concerning npi licensing:

© koninklijke brill nv, leiden, 2018 | doi: 10.1163/9789004351134_014

dp versus np: a cross-linguistic typology?

139

2.1

Article-less (=np-) Languages Disallow Clause-Mate npi Licensing under nr; Article (=dp-) Languages Allow It Bošković & Şener (2014:103), based on prior work by Bošković, claim that languages that don’t have articles (which they view as “np-languages”) disallow clause-mate npi licensing under Neg-Raising (nr), while “dp-languages” (i.e. languages with articles) allow such licensing. Let us start by looking at some of their examples in English: (1) John hasn’t/*has visited her in at least two years. (2) *John doesn’t claim [that Mary has visited her [npi in at least two years]] (3) John doesn’t believe [that Mary has visited her [npi in at least two years]] cf. bošković & şener (2014:103), examples (1)–(3)

(1) is fine with the negated auxiliary, while it is ill-formed when the auxiliary is not negated. The explanation proposed is as follows: The temporal expression “in at least two years” is a negative polarity item (npi). The npi requires a clausemate Neg. This explanation is also proposed for the other two examples: (2) is illformed, given that the Neg element is not a clause-mate of the npi. In (3), Neg is a clause-mate of the embedded npi, before Neg undergoes nr. In (2), nr would not have been possible, because the matrix verb claim is not an nr-verb, while believe in (3) is an nr-verb. Turkish: Bošković & Şener (2014) claim that Turkish clause-mate npis cannot be licensed long-distance, even under typical raising verbs like san- ‘think/ believe’. They further claim that Turkish thus patterns with np languages in this respect: (4) a. Pelin Mete-yi en az iki yıl-dır ziyaret Pelin (nom) Mete-acc at least two year-for visit et-me-di/*et-ti. do-neg-past/do-past ‘Pelin hasn’t/*has visited Mete in at least two years.’ (5) a. Mete [Pelin-∅/-i (*en az iki yıl-dır) Timbuktu-ya git-ti] Mete Pelin-nom/acc at least two year-for Timbuktu-dat go-past san-mı-yor. think-neg-pres.prog ‘Mete doesn’t think Pelin went to Timbuktu in at least two years.’ cf. bošković & şener (2014:104) examples (4), (5)

140

kornfilt

(4a) establishes the familiar licensing condition for Turkish npis by a Neg element. (5a) establishes (according to b&ş) that 1. the clause-mate condition on npilicensing by Neg is valid for Turkish, as well, and that 2. npis cannot be licensed long-distance, even when the matrix verb is a typical raising verb and should thus allow nr (in those languages that have nr, according to b&ş). However, the situation is more complex. The sequence set in roman does not need to be licensed by Neg, if the past tense of the predicate is changed, for example into the present progressive: (4) b. Pelin Mete-yi en az iki yıl-dır ziyaret ed-iyor Pelin Mete-acc at least two year-for visit do-pres.prog. ‘Pelin has been visiting Mete for at least two years.’ Note that here, the English translation is fine, too, but only when the preposition in is replaced by for;1 without that replacement, the sentence is bad, irrespective of the tense/aspect of the predicate: (4) c. *Pelin has been visiting/visited Mete in at least two years. Thus, while the sequence “in at least two years” does appear to be a genuine npi in English, the sequence “en az iki yıldır” in Turkish may not be a genuine npi (in the sense of being able to be licensed exclusively by a Neg element), thus corresponding to “for at least two years,” rather than to “in at least two years”. The ungrammaticality of (4a) and (5a) may be due to reasons of incompatibility of the temporal expression with the aspect of the clause. As a matter of fact, the English translation of (4a) with for instead of in appears to exhibit some sensitivity with respect to aspect, as well: In the presence of the Neg element (i.e. in the version with hasn’t), the “npi” with for is synonymous with the npi with in; however, in the version of (4a) without the Neg, i.e. in the version with has, the English translation with for, i.e. “Pelin has visited Mete for at least two years,” loses its meaning of “dropping by within a period of time of at least two years,” and can mean only “stay for a duration 1 As Bill McClure points out, in (4a), both in and for are fine. The difference between (4b) and (4a) is that in (4b), there is no Neg element which would license the npi “in at least two years”, while such a Neg licenser is present in (4a). As suggested in the text, “for at least two years” is not a genuine npi and thus does not require a Neg licenser; where Neg is missing, as in (4b), it can also be licensed by a compatible aspectual feature—here, by the progressive, as explained in the text.

dp versus np: a cross-linguistic typology?

141

of at least two years.”2 Since this alternative meaning of ‘visit x for a duration of time’ is not available to the Turkish verb ziyaret et- (which has only the ‘drop by’ meaning), the Turkish examples are ill-formed in the perfective clauses in (4). The availability of npis may vary from language to language; the German sequences in mindestens zwei Jahren ‘in at least two years’ and seit mindestens zwei Jahren ‘since two years’ should be investigated in this regard and probably would turn out to be somewhat different from their English counterparts. I leave this to future research and turn to other German expressions which are relevant to the issue of nominal phrases. In this context, let’s look at the idiom-like einen Finger rühren ‘to move a finger’ = ‘to lift a finger’. It exhibits npi-like behavior and has properties with respect to local versus non-local licensing by Neg as sensitive to the matrix predicate (of the sort illustrated above in (1)–(3) with respect to English): (5) b. Niemand hat für Marie ein-en Finger gerührt. nobody has for Mary a-acc finger moved ‘Nobody has lifted (even) a finger for Mary.’ Without a local Neg element, the idiomatic reading is lost: (6)

Jemand hat für Marie ein-en Finger gerührt. someone has for Mary a-acc finger moved ‘Someone has lifted a finger for Mary.’

Only the literary reading is possible in this example. (7) *Hans behauptet nicht, dass Marie für ihn ein-en Finger rühren Hans claims not that Mary for him a-acc finger move würde. would ‘Hans doesn’t claim that Mary would lift a finger for him.’ Here, the Neg element is not local with respect to the idiomatic sequence, and the result is ill-formed. 2 I am indebted to Tom McKay for this native judgment and for pointing this different meaning out to me. Bill McClure reports that for him, the English translation of (4a) in its version without Neg can have both readings; thus, it appears that for speakers like him, “in at least two years” is not an npi.

142

kornfilt

(8) Hans glaubt nicht, dass Marie für ihn ein-en Finger rühren Hans believes not that Mary for him a-acc finger move würde. would ‘Hans doesn’t believe that Mary would lift a finger for him.’ In this well-formed example, although the Neg element is not local to the idiomatic sequence, the idiomatic reading is preserved. Just as in the relevant English triplet (1) through (3), the reason for this can be ascribed to a special property of the matrix verb glauben ‘believe’, which would allow for nr and thus make an analysis possible which allows construal of the Neg element within the embedded clause. In contrast, the matrix verb behaupten ‘claim’ does not have the relevant property, and thus the matrix Neg element cannot be construed in this way. We can draw here the conclusion that German lines up perfectly with English, as expected by Bošković’s and Bošković & Şener’s approach, once examples are observed which are not (clear-cut) npi expressions. Some clear-cut npis in German such as je ‘ever’ and mehr ‘(any) longer’ also behave in ways similar to English npis and the idiomatic sequence einen Finger rühren: (9) niemand hat sie je/mehr besucht. nobody has her ever/any longer visited ‘Nobody has ever visited her’/‘Nobody has visited her any longer’ (10) Hans behauptet nicht, dass ihn Marie *je/*mehr besuchen Hans claims not that him Marie ever/any longer visit würde. would ‘??/*Hans doesn’t claim that Marie would ever/any longer visit him.’3

3 Bill McClure reports that (2), as well as the English translations of (7) and (10)—i.e. examples with the matrix verb claim, and all three in the present tense—are not ill-formed for him, and that all three improve further so as to sound completely fine for him in the past tense. This strongly suggests that for some speakers, the verb claim allows nr, just like the verb believe. I don’t have an explanation, at this point in time, why the past tense should lead to an improvement and leave this issue to future research.

dp versus np: a cross-linguistic typology?

143

(11) Hans glaubt nicht, dass ihn Marie je/?mehr besuchen Hans believes not that him Marie ever/any longer visit würde. would ‘Hans doesn’t believe that Marie would ever/any longer visit him.’ (9) shows the need of je and mehr to be licensed by a Neg element. (10) shows that this need has to be satisfied locally, i.e. by a clause-mate Neg element. (11) shows that locality can be extended to the matrix domain, when the matrix verb has the property of allowing nr, i.e. of allowing construal of the matrix Neg with the embedded npi. The English translations show that the corresponding examples in English are very similar. The generalization that seems to emerge is that, while German and English are remarkably similar, up to the verbs which do or do not allow nr, one has to acknowledge the fact that different npis behave differently with respect to licensing, especially with respect to “long-distance” licensing, even within one and the same language. Note also that both of these languages are clearly dplanguages, and they both have articles, i.e. what Bošković would acknowledge as a genuine d. Let us now turn to Turkish, i.e. a language characterized by Bošković and by Bošković and Şener as an np-language. (We shall address the issue of articles later.) We saw that the long-distance licensing of the npi-like sequence en az iki yıldır ‘at least in two years’ is somewhat different from its counterpart in English, in not being able to be licensed by a non-local Neg element, even in the presence of a raising verb such as believe. However, with respect to an idiom-like sequence such as küçük parmağını (bile) oynat ‘lift (even) one’s small finger’, Turkish lines up with the corresponding German examples: (12) Mete, Osman için küçük parmağ-ın-ı bile oynat-ma-dı. Mete Osman for little finger-3.sg-acc even move-neg-past ‘Mete didn’t lift even his little finger for Osman.’ (13) *Mete, Osman için küçük parmağ -ın -ı bile oynat-tı. Mete Osman for little finger -3.sg -acc even move-past ‘Mete lifted even his little finger for Osman.’ Without a licensing Neg element, the expression loses its idiomatic reading. (The asterisk on (13) is to be understood as representing ill-formedness under

144

kornfilt

the idiomatic reading; under the literal reading of moving one’s finger, the example is fine.) Similar local licensing is found in embedded clauses, too: (14) Osman [Mete-nin kendisi için küçük parmağ-ın-ı bile Osman Mete-gen him for little finger-3.sg-acc even oynat-ma-yacağ-ın-]-ı bil-iyor. move-neg-fut.nom-3.sg-acc know-pres.prog ‘Osman knows that Mete won’t even lift his little finger for him.’ (15) Osman [Mete-nin kendisi için küçük parmağ-ın-ı bile Osman Mete-gen him for little finger-3.sg-acc even oynat-ma-yacağ-ın]-ı düşün-üyor. move-neg-fut.nom-3.sg-acc think-pres.prog ‘Osman thinks that Mete won’t even lift his little finger for him.’ (16) Osman [Mete-nin kendisi için küçük parmağ-ın-ı bile Osman Mete-gen him for little finger-3.sg-acc even oynat-ma-yacağ-ın]-ı iddia et-ti. move-neg-fut.nom-3.sg-acc claim do-past ‘Osman claimed that Mete won’t even lift his little finger for him.’ However, when Neg is not positioned within the embedded clause, but on the matrix predicate, the nature of that predicate (and of the embedded clause) matters; some of the examples are bad, but some are quite fine: (17) *Osman, [Mete-nin kendisi için küçük parmağ-ın-ı bile Osman Mete-gen him for little finger-3.sg-acc even oynat-acağ-ın]-ı bil-mi-yor. move-fut.nom-3.sg-acc know-neg-pres.prog *‘Osman doesn’t know that Mete will even lift his little finger for him.’ Under the idiomatic reading (of the npi), the Turkish example is ill-formed, as is its English counterpart. (Under the literal reading, the example is fine.) (18) Osman, [Mete-nin kendisi için küçük parmağ-ın-ı bile Osman Mete-gen him for little finger-3.sg-acc even oynat-acağ-ın]-ı düşün-mü-yor. move-fut.nom-3.sg-acc think-neg-pres.prog ‘Osman doesn’t think that Mete will lift even his little finger for him.’

dp versus np: a cross-linguistic typology?

145

My informants, as well as myself, find (18) quite acceptable under the idiomatic reading, but not (19). (19) *Osman, [Mete-nin kendisi için küçük parmağ-ın-ı bile Osman Mete-gen him for little finger-3.sg-acc even oynat-acağ-ın]-ı iddia et-me-di. move-fut.nom-3.sg-acc claim do-neg-past ‘Osman didn’t claim that Mete would lift even his little finger for him.’ It looks like the matrix verb düşün- ‘think’ is a Neg-raising verb, making it possible for the idiomatic npi item in the embedded clause to get locally licensed by the Neg on the matrix predicate. Other matrix predicates such as claim and know do not allow this. Thus, we have a similar alignment overall to German as well as English. It should further be noted that düşün- ‘think’ is not an exception in Turkish with respect to allowing “Neg-raising,” where local licensing of npi items in embedded clauses is concerned. For example, iste- ‘want’ exhibits similar behavior, as do some other matrix verbs, which take “subjunctive”-type nominalized clauses rather than “indicative”-type of clauses. 2.2 “Only Article Languages Allow the Majority Superlative Reading” Bošković & Şener (2014:107) discuss the ambiguity of examples such as (20) in English: (20) At the party, most people drank beer. The ambiguity consists of the following two readings: (20) a: Majority reading: ‘More than half the people drank beer’ (20) b: ?Plurality reading: ‘More people drank beer than any other beverage (though it could be less than half the people)’ The authors further report that in Slovenian only the plurality reading is available: (21) Največ ljudi pije pivo. most people drink beer ‘Most people drink beer.’ (Bošković & Şener (2014:108), example (15))4 4 Bošković & Şener (2014) refer to Živanović (2008) as the source of these data. No page or example numbers are given.

146

kornfilt

The authors specify the two potentially possible readings as follows: a:

‘More people drink beer than drink any other beverage.’

This is the Plurality reading (pr), i.e. beer was the more popular drink. In contrast, the Majority reading (mr) is not available for this Slovenian example: b:

‘*More than half the people drink beer.’

Note further that Slovenian is an article-less language and would thus be characterized as an np-language in Bošković’s typology of nominal phrases. Bošković & Şener further claim that the German equivalent is ambiguous: (22) Die meisten Leute tranken Bier. the most people drank beer Adapted from bošković & şener (2014:108), example (16), where the tense is the present tense.

Whether ambiguous or not, it is clear that both the English and the German examples do have the majority reading. Bošković’s and Bošković & Şener’s claim is that np-languages do not have the majority superlative reading for elements corresponding to most, and thus allow only the plurality reading for them. Slovenian would be characterized as an np-language, based on the fact that it lacks articles, and the further fact that the relevant examples lack the majority reading would be expected. Bošković & Şener further claim that Turkish patterns with Slovenian: (23) İnsan-lar en çok bira iç-ti. person-pl(nom) most very beer drink-past ‘People drank beer the most.’ Adapted from bošković & şener (2014: 108), example (17)5

As explained by Bošković & Şener, also in reference to Gajewski (2011), a pr can have a mr as an instance of inference in context, but not as a genuine, direct reading: (23) “may be interpreted indirectly as counting other objects

5 Here, Bošković & Şener (2014) refer to Gajewski (2011) as the source of data and discussion. No page or example numbers are given.

dp versus np: a cross-linguistic typology?

147

with certain background assumptions; i.e. one might get a mr as an inference. However, Gajewski shows that the Majority Reading is unavailable in [23] under the scenario in [24], although it enforces it …” (Bošković & Şener 2014: 108). In other words, the authors deny that (23) has a genuine Majority Reading. To strengthen this denial, they offer the following scenario, under which the Majority Reading should be reinforced; they further state that (23) (i.e. their 17) cannot be truthfully uttered in the context given by that scenario: (24) Bošković & Şener’s (18) Suppose people at a dinner were allowed more than one beverage. 60 % of the people had a beer. 75% of the people had a glass of wine. (2014:108) I agree with Bošković & Şener’s judgment of (23) in the context of (24); in other words, I agree that the Majority Reading is not available here. However, I disagree with the claim that this has anything to do with the status of the Turkish nominal phrase as an np rather than dp. First of all, let me point out that there are instances where superlative expressions in “dp-languages”, too, have only the plurality reading; e.g. German: (25) Auf Partys trinken die Leute meist Bier. at parties drink the people most(ly) beer ‘At parties, people drink mostly beer.’ As mentioned above, Bošković & Şener point out, in reference to Gajewski, that the pr can have the mr as an inference, and this is possible in (25)—but only as an indirect interpretation and not directly in its semantics, i.e. just like in “np-languages.” If Bošković & Şener (2014) and Živanović (2008) had looked at examples such as (25) rather than (22) for German, they probably would have made the opposite observation and generalization about article languages, and German would have looked just like Turkish with respect to example (23). In advocacy of Bošković & Şener and Živanović, one could say that (25) is irrelevant, since the superlative element is not within the “Traditional np,” as they call the relevant nominal phrase. The claim about the majority superlative reading being possible only in dp-languages may be limited only to those elements within the “Traditional Noun Phrase.” However, if so, then the Turkish example in (23) should be irrelevant, too, and for the same reason. In order to see whether a language’s status as a dp- or np-language correlates with the availability of the majority superlative

148

kornfilt

reading for a superlative element within a “Traditional Noun Phrase”, we have to examine examples of the relevant structure; e.g. for Turkish: (26) İnsan-lar-ın çoğ-u bira iç-ti. person-pl-gen most-3.prs. beer drink-past ‘Most of the people drank beer.’ For all practical purposes, (26) is a close translation of the German (22), and unambiguously has the Majority Reading. Therefore, at least in this respect, the interpretation of the “majority superlative reading” is not specific to dplanguages (or “article-languages”), contrary to Bošković’s and Bošković & Şener’s claims—unless Turkish is a dp-language. In other words, perhaps what is faulty on the authors’ part is only the categorization of Turkish as an np- (i.e. article-less) language, while their approach could still be correct in general. We shall return to this issue. 2.3

“Article-less Languages Disallow Transitive Nominals with Two Lexical Genitives” Bošković & Şener (2014:104) state that this generalization concerns “the availability of structures where both the external (not simply a possessor, but a true external argument) and the internal argument of a noun are genitive, with the genitive realized via a clitic/suffix or a dummy preposition. Such cases are disallowed in article-less languages (which don’t otherwise allow multiplication of the same case like Japanese). The same holds for Turkish.” (Bošković & Şener 2014).6 The example offered to illustrate this claim for Turkish is as follows: (27) *Osmanlı-lar-ın İstanbul-un feth -i Ottoman-pl-gen Istanbul-gen conquest-3.prs. ‘Ottomans’ conquest of Istanbul’ bošković & şener (2014:104), example (6a)

It is very unclear that the ungrammaticality of such examples has anything to do with a language’s being a dp- or an np-language. At least with respect to Turkish, the issue seems to be that: 1. there is no “dummy p” which realizes the genitive; the genitive suffix is the only means to realize this Case; 2. Specifiers and modifiers are all pre-head, and doubling of genitives in the same direction appears to be disallowed; in German, even when there are two genitives, one

6 I will return to the question of characterizing Turkish as an article-less language shortly.

dp versus np: a cross-linguistic typology?

149

would be before the head, and one after the head (e.g. Mehmet des Zweiten Eroberung Istanbuls ‘Istanbul’s conquest by Mehmet the Second = Mehmet the Second’s conquest of Istanbul’). Note that in “dp-languages” such as German and English, too, it is not possible to realize two genitives in a row, both of which are expressed by a genitive suffix (rather than one of them being realized as a p): (28) *The Ottomans’ Istanbul’s conquest In addition to the well-formed translation of (27) above, involving a “dummy p,” namely of, English can also have a construction with an agentive phrase: (29) Istanbul’s conquest by the Ottomans This type of construction is well-formed in Turkish, as well: (30) İstanbul-un Osmanlı-lar tarafından feth-i Istanbul-gen Ottoman-pl by conquest-3.sg. ‘Istanbul’s conquest by the Ottomans’ Interestingly, there is yet another construction in Turkish to express transitive nominals, involving one genitive and one accusative: (31) Osmanlı-lar-ın İstanbul-u feth-i Ottoman-pl-gen Istanbul-acc conquest-3.prs. ‘The Ottomans’ “conquering” Istanbul’ A number of verbal nouns borrowed mostly from Arabic retain their ability to license accusative on their complements, i.e. on their internal argument; cf. Keskin (2009), among others. An additional possibility (i.e. in addition to a cross-linguistic ban against two genitives in a row) for explaining the ill-formedness of Turkish examples such as (27) may be the fact that this alternative structure is available. At any rate, it is clear that the proposed generalization is incorrect. 2.4

Only Article-less (i.e. np-) Languages May Allow Scrambling (Bošković & Şener (2014:104)) The scrambling referred to in this generalization is the long-distance type as illustrated in, e.g., Japanese. Thus, the well-known phenomenon of scrambling in German is said to be irrelevant in this context. However, while Standard Ger-

150

kornfilt

man indeed does not allow long-distance scrambling, many German dialects, especially southern ones, do allow it. Interestingly, speakers of such dialects allow such movements even in Standard German; the following examples in Standard German were volunteered by Josef Bayer, who is a native speaker of Bavarian: (32) [solche Bücher]i glaube ich nicht, [dass ein Bauer ei freiwillig such books believe I not that a farmer voluntarily lesen würde] read would ‘Such books, I don’t believe that a farmer would read voluntarily.’ (33) [Die “Kritik der reinen Vernunft”]i bezweifle ich, [dass mein the critique (of) the pure reason doubt I that my Nachbar sich ei kaufen würde] neighbor self buy would ‘The Critique of Pure Reason, I doubt that my neighbor would buy for himself.’ Bayer evaluates these examples as “totally normal.”7 Facts such as these are rather damaging to the typological claim that “dp-languages” (or “article languages”) do not allow long-distance scrambling/topicalization. 2.5

“Inverse Scope is Unavailable in Article-less Languages (in Some Examples)” (Bošković & Şener (2014:109)) One of the claimed generalizations is the one expressed in this subsection’s title. The following example illustrates “inverse scope”: (34) Someone loves everyone. Examples such as this one are ambiguous in English. Here, “inverse scope” refers to the interpretation under which the internal argument takes scope over the external argument. The relevant claim in Bošković’s typology is that in examples with unmarked word order in the language under investigation, inverse scope (i.e. phrase-

7 Other native speakers of German reported similar reactions at presentations of this material at the University of Cologne and at the Max Planck Institute of Evolutionary Anthropology in Leipzig.

dp versus np: a cross-linguistic typology?

151

structurally lower expressions taking scope over phrase-structurally higher expressions) is possible only in dp-languages; np-languages (i.e. “article-less languages”) do not allow this, according to this claim. According to Bošković & Şener, Turkish conforms to “the general behavior of np languages”: (35) iki öğrenci her sandalye -yi kır -mış two student (nom) every chair -acc break evidential past ‘Two students (are said to) have broken every chair.’ bošković & şener (2014: 109), example (20); glosses and translation are mine.

The object cannot take scope over the subject. I do agree with this judgment. But in English, too, inverse scope with such examples is marginal: (36) Two students broke every chair. (37) Two students broke each chair. According to native speakers of (American) English whom I have consulted,8 examples with “each” are slightly better than those with “every”, but are still marginal in these examples. I am indebted to Tim McKinnon, one of my native speaker consultants, for pointing out that the contrast between “each” and “every” is even clearer in other examples which are similar to the ones just discussed: (38) Two diplomats represented every country. (39) Two diplomats represented each country. According to McKinnon, the inverse reading is quite fine for (39), but not for (38). In all of these English examples, the “regular” reading, with the subject taking scope over the object, is well-formed. So, at least for some examples, English patterns very similarly to Turkish. Other examples, such as in examples corresponding to (34) where inverse scope is possible in English but not in Turkish can be explained in other ways, without appeal to an np/dp dichotomy. Such an example is (40): 8 I thank Peter Cole and Tim McKinnon for their native judgments concerning the examples in this section and for discussing these facts with me.

152

kornfilt

(40) Birileri herkes-i sev -er someone everybody-acc love -pres ‘Someone loves everyone.’ Inverse scope in a word-order free language such as Turkish is typically obtained via scrambling; it is possible that this way of expressing inverse scope overtly in the syntax may block the availability of the (lf-based, i.e. abstract) scope inversion reading for the unmarked word order. 2.6

“Radical Pro-drop is Possible Only in Article-less Languages” (Bošković & Şener (2014:105)) 2.6.1 “Radical” Pro-drop Bošković’s term of “radical pro-drop” is defined as “… the productive prodrop of subjects and objects in the absence of rich verbal agreement” (cf. Bošković & Şener 2014: 105). “Spanish-type pro-drop”, i.e. pro-drop licensed by rich morphology, would be a different type of “pro-drop”. Bošković & Şener claim that Turkish has, in addition to the “Spanish-type pro-drop”, also radical pro-drop of objects, given that Turkish lacks object agreement morphology. The authors further claim that Turkish patterns with a number of “np-languages” in this respect. But is “object-drop” (in Turkish) pro-drop at all? In other words, is the empty element that arises for non-subjects a genuine pro? There are good reasons to believe that the empty category we have is not pro, but rather an a’-bound empty variable. Such elements are found not only in “np-languages” of the type mentioned by Bošković & Şener, but also in Germanic languages, which are dp-languages in the proposed typology. For example, such an empty variable would be found in connection with the phenomenon often referred to as “topic drop”; one recent in-depth discussion of this phenomenon in Germanic can be found in Sigurðsson (2011), to which I shall return. 2.6.2 Brief Discussion of the Nature of “Dropped” Objects in Turkish In some prior literature, the claim does exist that the empty category representing a “dropped” object is a pro, i.e. has pronominal properties; this is claimed most specifically in Öztürk (2005): (41) Maryi ec*i/j sev -iyor Mary love -pres.prog ‘Mary loves him/her/it/*herself.’

dp versus np: a cross-linguistic typology?

153

Öztürk (2005) claims that if condition b is violated (and we see the ill effects of that violation), the empty category is pro. However, the ill-formedness of the relevant reading could also be due to a condition c violation, if the silent object is a phonologically empty variable, bound by an abstract topicalization operator. There is evidence for the latter analysis, if additional examples are considered where condition b is irrelevant, and condition c explains the ill-formedness: (42) Maryi [Alij -nin ec*i/*j/k sev -diğ-in]-i söyle -di Mary Ali -gen love -Fact.nom.-3.sg-acc say -past ‘Maryi said that Alij loves her*i/himself*j /himk/herk/itk.’ The silent object in the embedded clause in (42) cannot be co-indexed with ‘Mary’, even though this would not give rise to a condition b violation, as ‘Mary’ would not be a local binder of the silent object, because it is not its clause-mate. However, condition c would indeed be violated if the silent object is a variable, given that elements sensitive to condition c must be free everywhere, not just in their local binding domain. Uncontroversial instances of pro, as in a silent embedded subject (i.e. “Spanish-type pro”), are well-formed in similar contexts: (43) Maryi [proi/j Ali -yi sev -diğ -in] -i söyle -di Mary Ali -acc love -Fact.nom -3.sg -acc say -past ‘Maryi said that shei/j loves Ali.’ The silent embedded subject is preferably interpreted as co-indexed with ‘Mary’ (but can also be anteceded by a discourse referent). Since the silent object clearly has different syntactic properties than the silent subject, I would claim here that the silent object is not an instance of pro at all, but is a silent, operator- (topic-) variable. Therefore, instances of (discourse-) elided objects in Turkish are not instances of pro, and hence not instances of radical prodrop (in the narrow sense of Bošković & Şener). In other words, I advocate a “traditional analysis” of null elements, whereby “Spanish-type” pro is indeed an empty pronominal category, while, in Turkish at least (and perhaps other morphologically rich languages), an empty category which is not licensed and identified by (rich) morphology is not pronominal and can only be a variable bound by a topic operator. But before leaving this issue, let us play devil’s advocate: Perhaps these instances of (topic-bound) silent variables are a hallmark of np-languages; isn’t it possible that Bošković & Şener’s diagnostic is correct, as is their typology, but that the phenomenon was simply mislabeled?

154

kornfilt

However, if Sigurðsson (2011) is correct, then at least some Germanic languages (i.e. dp-languages) do have “topic drop” as well. I thus return, as promised, to Sigurðsson’s discussion of argument drop in Germanic. A recent analysis proposed in Sigurðsson (2011) claims that instances of “dropped” topics in Germanic have a’-properties different from “Spanish-type” instances of pro, which are licensed by rich morphology and which lack a’properties. For our purposes, what is important is that dp-languages do have silent topics that are not instances of “Spanish-type” pro (and are at least in this respect similar to “dropped” non-subjects in languages such as Chinese, Japanese etc.); thus, Sigurðsson (2011) offers the following example from Swedish: (44) [Null topic operator]i Kan’ja ei inte veta. can I not know ‘That, I cannot know.’ cf. sigurðsson 2011: 291, example (56a); indications of the “dropped” object and of the silent topic are mine

Thus, these facts constitute a serious challenge to Bošković’s and Bošković & Şener’s typology.

3

The (Un)importance of Articles in the np/dp Typology

We saw earlier that in Bošković’s framework, lacking articles is a clear diagnostic for classifying a given language as an “np-language.” But we do have to ask, at this juncture, whether article-less languages are clear-cut np-languages. Why shouldn’t other determiners (e.g. demonstratives) qualify as d, as well? Yet another question is: What if a language has one type of article but not another? Turkish has demonstratives; as just mentioned, those are not accepted as a genuine d by Bošković. Furthermore, Turkish has an indefinite article. However, Bošković & Şener (2014) and Öztürk (2005) analyze that morpheme (bir) as the numeral one in all its occurrences, and not as the indefinite article a. It is therefore of some interest to address the properties of this morpheme more closely. Öztürk’s arguments against analyzing bir as an indefinite article are as follows: Crisma (1999) and Longobardi (2001) claim that if a language has only one article, it will be a definite rather than an indefinite article. Since Turkish has no definite article, Öztürk claims that bir cannot be an indefinite determiner.

dp versus np: a cross-linguistic typology?

155

But this reasoning is circular. Turkish is dismissed as a language having an indefinite, but no definite, article, based on cross-linguistic statistics. But in the very same statistics on which Öztürk bases her argument that bir is not an indefinite article, Turkish could not appear as a counterexample, given the author’s classification of bir. Thus, statistics of this nature cannot be taken seriously as the basis of arguments dismissing this morpheme as an indefinite determiner.9 In addition, there are good reasons for characterizing bir as an indefinite article, because its syntactic properties as an article differ clearly from those when it is a numeral; such distinctions are to be found in any good grammar of Turkish and include differences in placement with respect to adjectives; e.g. (45) Bir/beş yaşlı kadın one/five old woman ‘One old woman/five old women’ (46) Yaşlı bir kadın old a woman ‘An old woman’ These are the unmarked orders of the respective examples: numerals precede adjectival modifiers, while the indefinite article follows them. It is possible to override these unmarked correlations with intonation; e.g. in (45), deaccenting the numeral makes it likelier for it to be interpreted as the indefinite article, and in (46), stressing the article makes it likelier for it to be interpreted as the numeral. Nonetheless, the unmarked orders and intonation contours yield the meanings as stated above.10

4

Conclusions and Further Questions

The research agenda proposed in Bošković and Bošković & Şener’s studies has had the positive effect of triggering cross-linguistic studies, attempting to con9

10

I am indebted to Sasha Vovin for pointing out that Thai (Siamese) exhibits a situation similar to the one in Turkish: It also has ‘one’ as an indefinite article, but has no definite articles. The unmarked order within the Turkish dp/AgrP is as follows: Possessor > relative clause > demonstrative > numeral/quantifier > adjective (phrase) > indefinite article > head noun.

156

kornfilt

firm or disconfirm the detailed typological claims made. For example, Marušič & Žaucer (2010) discuss a group of Slovenian dialects, which they refer to under the cover term Gorica Slovenian, with respect to one of the claims made in this typological enterprise that was not addressed in the present paper, given that the languages addressed here do not exhibit the relevant property, namely clitic doubling. It is claimed in Bošković’s studies that only dp-languages can exhibit clitic doubling; in other words, np-languages with clitics never exhibit clitic doubling, according to this typological claim. However, Gorica Slovenian has no definite article and thus would count as an np-language in the relevant typology (this, in itself, is problematic, in light of the discussion in the present paper of Turkish as a language without a definite article but with an indefinite article). Gorica Slovenian does have clitic doubling (as opposed to Standard Slovenian, which does not). Marušič & Žaucer point out, however, that clitic doubling in Gorica Slovenian is limited to personal pronouns. After pointing out the problem for Bošković’s typology, the authors attempt to accommodate their finding within that typology, by weakening its relevant claim, so as to exclude clitic doubling of personal pronouns from the generalization in question. The present paper, while also owing its existence to Bošković’s and Bošković & Şener’s interesting and refreshingly falsifiable typology, is of a different nature. By addressing more than just one or two of the proposed typological correlations at a time, and, even more importantly, by pointing out problems not only with respect to “np-languages”, but also with respect to “dp-languages”, I hope to have demonstrated that a typology of this kind and with these claims cannot be sustained in general. With respect to Turkish in particular, probably the best move is to view it as an article-/dp-language. But this move does not make the np-dp typology much more attractive, given some of the problems posed to the posited correlations by clear-cut dp-languages such as German and English, and not only by the supposed np-nature of Turkish. In addition, the relevant properties of languages and even of individual constructions and of syntactic-semantic phenomena such as relative scope appear to vary with different lexical items, thus challenging generalizations with respect to such phenomena. I propose therefore to give up the kind of typological endeavor suggested by Bošković and Bošković & Şener, and to examine each language individually as to whether it does or does not have dps, independently from whether it does or does not have articles.

dp versus np: a cross-linguistic typology?

157

Acknowledgments My first duty of gratitude goes to John Whitman himself. He has been the most wonderful colleague, co-author and friend, over more decades than I dare say explicitly. I further thank Josef Bayer for German examples and judgments, Peter Cole, Gerry Greenberg, Bill McClure, Tom McKay, and Tim McKinnon for English judgments, Akgül Baylav, Demir Dinç, Alp Otman and Ayşe Yazgan for Turkish judgments, and Josef Bayer, Peter Cole, Martin Haspelmath, Klaus von Heusinger, Tim McKinnon, Volker Struckmeier, and Florian Schaefer for discussion. I further thank the editors of this volume for their invitation to participate in this venture, as well as for their helpful comments. The usual disclaimers apply.

Abbreviations 3.prs. acc dat dp Fact.nom fut.nom gen mr neg nom np npi nr pl pr pres pres.prog sg

Third person Accusative Dative Determiner phrase Factive nominalization Future nominalization Genitive Majority reading Negative (element) Nominative Noun phrase Negative polarity item Neg(ative) raising Plural Plurality reading Present (tense) Present progressive Singular

References Abney, Steven 1987. The English noun phrase in its sentential aspect. Unpublished Ph.D. dissertation, mit.

158

kornfilt

Bošković, Željko 2008. ‘What will you have, dp or np?’ Proceedings of nels 37, pp. 101– 114. Bošković, Željko 2012. ‘On nps and Clauses’. In: Discourse and Grammar: From Sentence Types to Lexical Categories; Günther Grewendorf and Thomas Ede Zimmermann (eds.), pp. 179–242. Bošković, Željko 2013. ‘Phases beyond clauses’. In: The Nominal Structure in Slavic and Beyond; Lilia Schürcks, Anastasia Giannakidou, Urtzi Etxebarria, Peter Kosta (eds.), pp. 75–128. Bošković, Željko & Serkan Şener 2014. ‘The Turkish np’. In: Crosslinguistic Studies on Noun Phrase Structure and Reference; Patricia Cabredo Hofherr and Anne ZribiHertz; Leiden: Brill, pp. 102–140. Crisma, Paola 1999. ‘Nominals without the article in the Germanic languages’. Rivista di Grammatica Generativa 24, pp. 105–125. Gajewski, Jon 2011. ‘More on superlatives in a language without articles: The case of Turkish’. Talk given at the cuny Graduate Center. Keskin, Cem 2009. Subject Agreement Dependency of Accusative Case in Turkish, or Jump-Starting Grammatical Machinery. Ph.D. dissertation, Utrecht University. Marušič, Franc & Rok Žaucer 2010. ‘Clitic doubling in a determinerless language with second position clitics’. In: Formal Studies in Slavic Linguistics: Proceedings of fdsl (Formal Descriptions of Slavic Languages) 7.5; Gerhild Zybatow, Philip Dudchuk, Serge Minor, Ekaterina Pshehotskaya (eds.); Frankfurt am Main: Peter Lang, pp. 101– 115. Longobardi, Giuseppe 2001. ‘The structure of dps: Some principles, parameters and problems’. In: Mark Baltin & Chris Collins (eds.). The Handbook of Contemporary Syntactic Theory. Oxford: Blackwell, pp. 562–601. Öztürk, Balkız 2005. Case, Referentiality and Phrase Structure (Linguistik Aktuell/Linguistics Today 77). Amsterdam & Philadelphia: John Benjamins. Sigurðsson, Hálldor Ármann 2011. ‘Conditions on Argument Drop’. Linguistic Inquiry Vol. 42, No. 2, pp. 267–304. Živanović, Sašo 2008. ‘Varieties of most: On different readings of superlative determiners.’ Proceedings of fdsl (Formal Descriptions of Slavic Languages) 6.5, pp. 337–354.

chapter 14

The Old Japanese Accusative Revisited: Realizing All the Universal Options Shigeru Miyagawa*

1

Introduction

One thing that is certain about human language is that it continuously changes. A point that is almost too obvious even to mention is that however a language changes through time, it must remain a human language. In other words, the changes cannot be random in nature. An important idea put forth by some linguists is that the changes are predetermined in that diachronic variation within a particular language directly reflects variations found across contemporary languages, that is, diachrony is synchrony when it comes to variations that can occur. In a clear expression of this idea, Lightfoot (1979: viii) observes that the formulation of “a possible grammar will provide the upper limits to the way in which a given grammar may change historically, insofar as it cannot change into something which is not a possible grammar.” In a similar vein, Joseph (1980: 346), in addressing the loss of the infinitive form in Greek, notes that “[u]niversal constraints which hold in synchronic grammars are used to explain the direction taken by certain changes in syntax.” In Miyagawa (1989) and Miyagawa and Ekida (2003), I tested this hypothesis about language change by looking at how the accusative case marking for objects of eighth century Old Japanese (oj) changed into the familiar system we find today. A noun phrase must, as a universal requirement, have case (Chomsky 1981). Across languages, we typically find two different ways of case-marking the object: morphological case marking, as in German, modern Japanese, and Latin, and what is sometimes referred to as abstract case, or the absence of any overt case form, which characterizes English, Romance languages, and a host of others. What I argued is that in oj, we find both morphological and * I wish to take this opportunity to express my gratitude to John for his work, which has had enormous impact on the field including my work as evident from this chapter. I also want to thank him for the long years of a cherished friendship. This chapter originally appeared in Case Marking, Argument Structure, and Word Order, Routledge Leading Linguists Series (2012).

© koninklijke brill nv, leiden, 2018 | doi: 10.1163/9789004351134_015

160

miyagawa

abstract case, but language change transformed the language into a strictly morphological-case-marking language, which is what we find today. The choice between morphological case and abstract case in oj is not random, as sometimes observed, but is predicted by the inflection found on the verb, a point first noted by Matsunaga (1983). Specifically, the conclusive form of the verb, which is found typically in the main clause, but also in certain subordinate clauses, allows abstract case, so that its object occurs bare without any case marking. In contrast, morphological case marking must appear if the verb has attributive inflection, which occurs commonly in subordinate clauses but can also occur in the main clause in what is called the kakarimusubi construction, as we will see later. (1) Distribution of abstract and morphological case (Miyagawa 1989) The conclusive form assigns abstract case, while the case-assigning feature of the attributive form must be manifested overtly as -wo.1 This is exemplified below from the Man’yōshū, a poetic anthology compiled in the eighth century. (2) Ware-pa imo ___ omop-u. I-top beloved think-fin ‘I think of my wife.’ (3) [sima-wo mî-ru] tökî [island-acc look-attr time ‘when I look upon the island’ In (2), the verb ‘think’ is in the conclusive form, and the object ‘beloved’ has no overt case marking, which arguably indicates that it is licensed abstractly. In (3), the verb ‘look’ is in the attributive, and the object ‘island’ is accompanied by the oj accusative case marker -wo. We can see from Matsuo’s (1938) work that both abstract case and morphological case were common in the Man’yōshū.

1 In Miyagawa (1989), I used -o as the form of the accusative case marker, for uniformity with the modern language, but in this chapter I will use -wo, which reflects the original labial glide present in oj and earlier forms of Japanese.

161

the old japanese accusative revisited

(4) Frequency of the case marker -wo in the Man’yōshū

Object np with -wo Without -wo Man’yōshū (Book 17)

51

96

The fact that -wo does not always appear has led some to suggest that the oj -wo was unstable as a case marker (Kobayashi 1970: 226), or that it was simply optional (Wrona & Frellesvig 2009). However, close examination shows that the distribution of oj -wo is highly predictable and that it clearly functions as a case marker. It is important to point out that the difference between abstract and morphological case is not a difference between main and subordinate clauses, as might be suggested from looking at (2) and (3). Although the conclusive form, which assigns abstract case, occurs largely in the main clause, it also occurs in subordinate clauses with the complementizer to. The following example is from the Tosa diary, written in the Early Middle Japanese of the tenth century.2 (5) Kono fito [uta ___ yom-am-u to] omof-u kokoro this person [poem compose-tent-fin comp think-attr mind ar-ite nar-iker-i. exist-sub cop-retr-fin ‘It turned out that this person had the intention to compose poems.’ As shown, the object uta ‘poem’ is bare, indicating that the verb has assigned abstract case. Similarly, while the attributive form occurs typically in subordinate clauses, it can also occur in the main clause in the kakarimusubi construction. This is the construction in which a kakari focus particle occurs sentence-medially, and the verb must be in the attributive form (or, with one type of kakari particle, namely koso, in the perfect form). The example below, taken from Sansom (1928), illustrates this rule for the kakari particles zo and ya (zo is used for emphasis, something akin to ‘indeed’, while ya is commonly used for rhetorical questions).

2 The text for the Tosa diary is Ikeda 1930. This particular example is noted in Zenno 1987.

162

miyagawa

(6) a. Isi-wa kawa-ni otu. rock-top river-in fall ‘Rocks fall into the river.’ b. Isi zo kawa-ni oturu. c. Isi ya kawa-ni oturu.

Conclusive

Attributive Attributive

Among 208 examples in a sample of the Man’yōshū (Takagi et al. 1962: 55– 109), thirty-four are kakarimusubi constructions with a transitive verb and an object np. All thirty-four are found in main clauses, and, the particle -wo marks the object np without exception (Matsunaga 1983). The following is one such example. (7) Kimi-ga mi-fune-wo itu to ka mat-am-u? you-nom fine-boat-acc when qp wait-tent-attr ‘When may we await your fine boat back?’ Another feature of the analysis following Matsunaga (1983) and Miyagawa and Matsunaga (1986) is that the change from the mixed case system of oj to one that uses morphological case across the board is predictable from the way the verbal inflection changed from around the tenth century to the fifteenth century. During this period, the language lost the conclusive form, and the attributive form moved into its place. This is shown below, based on a table in Matsunaga (1983): (8) Changes in the verbal inflection

Inflection type

i

ii

iii

iv

v

Before the change Conclusive -u -i -iru -eru -u Attributive -u -u -iru -eru -uru After the change

-u -u -iru -eru -uru

Before the change, conclusive and attributive inflections had different forms in Types ii and v, and, more importantly, this difference identified the two inflections as distinct across the language. But after the change, the distinction was lost, and where there were differences earlier, the new form took on the

the old japanese accusative revisited

163

shape of the old attributive form, suggesting that it is the attributive inflection that won out. This naturally led to the spread of morphological case because the attributive form requires morphological case, and its distribution expanded due to language change, from a more restricted distribution in oj, mostly in subordinate clauses, to a much broader distribution, including the main clause, in modern Japanese. In this paper, I will take up criticisms of Miyagawa (1989) by Kinsui (1993) of the analysis of the distribution of oj abstract and morphological case. I will answer them by drawing on the work of Miyagawa and Ekida (2003) and also Yanagida (2007) and Yanagida and Whitman (2009). As we will see particularly from Yanagida 2007, but also from Yanagida and Whitman (2009), oj had a third way to license Case on the object along with abstract and morphological case. Kinsui (1993: 202) accepts the distinction I drew between abstract case and morphological case marking for oj. He concludes by saying that he “believe[s] that we can accept, as a tendency, the absence of wo on the main clause object and its presence in the subordinate clause object, as Miyagawa asserts” (1993: 209). He is, however, reluctant to accept it at face value because there are “numerous counterexamples” (1993: 208). He criticizes the theory in Miyagawa (1989) as “too rigid and unable to account for the counterexamples,” raising three main issues. (9) Kinsui’s criticisms (i) Versification Ninety percent of the poems in the Man’yōshū consist of tanka, a verse form that has five lines of five or seven moras, 5-7-5-7-7 (Nippon Gakujutsu Shinkokai 1965). In poetry it is possible that the occurrence of the case marker -wo is governed in part by the rigid versification. (ii) There are examples in which the object has the morphological -wo despite the fact that the verb is in the conclusive form. (iii) There are examples in which the object does not have morphological -wo despite the fact that the verb is in the attributive form. I will take up each of these; I will combine (i) and (ii) in the next section, and in section 3, I will take up (iii) by drawing on recent work by Yanagida (2007) and Yanagida and Whitman (2009).

164 2

miyagawa

The Conclusive Form in mj Prose: Miyagawa and Ekida 2003

Kinsui’s point that versification may sometimes dictate whether -wo appears or not appears to find confirmation in the following waka poetry example taken from the Izumi Shikibu diary, a literary work of the tenth century (see m&e). The waka versification is the 5-7-5-7-7 tanka pattern commonly found in the Man’yōshū. The translation is from Cranston (1969). (10) Ofokata ni Samidaruru to ya Omofuramu Kimi ___ koiwataru Kefu no nagame wo

Nothing remarkable— The same old rain that pelts us Every year, you think? These are my tears of love Falling in a deluge all day long!

In the fourth line, morphological -wo fails to occur on the object kimi ‘you’ although its verb, koiwataru ‘longing for’, is in the attributive form. The absence of -wo makes it possible to maintain the versification of five or seven moras—in this case seven. I will return to this example later. To see whether poetic versification somehow skewed the distribution of -wo in the Man’yōshū, Miyagawa and Ekida looked at several major works of literature from the Heian period (a.d. 794–1185), all written in prose with some poetry sprinkled through the text; they excluded the poetry. The two primary texts taken up are the Izumi Shikibu diary and the Murasaki Shikibu diary, both written by Heian court ladies in the tenth century. Miyagawa and Ekida also drew from the Tale of Genji by Murasaki Shikibu, one of the most important literary work of the Heian period, as well as the Sarashina diary (eleventh century) to further confirm certain points observed in the other two texts. In this chapter, I will focus on the Murasaki Shikibu diary (msd) to demonstrate Miyagawa and Ekida’s point that we find the same distribution of abstract and morphological case in prose as in poetry. The msd occupies eighty-three pages in the Iwanami Bunko series (Ikeda and Akiyama 1984), and was written by Murasaki Shikibu, the celebrated Heian writer and court lady who also authored the Tale of Genji. This diary, which does not contain very many waka poems (which were, in any case, excluded for consistency), “has to do chiefly with the birth of two sons to the empress, events of political importance, since she was the daughter of Michinaga and through his royal grandchildren Michinaga got an unshakable grip on the imperial house” (Seidensticker 1981, viii).

165

the old japanese accusative revisited

2.1 The Text of the Murasaki Shikibu diary Miyagawa and Ekida found 382 pertinent sentences with a direct object in the msd. I first give the raw data below according to verbal inflection. Miyagawa and Ekida deal with four inflective forms, the conclusive, attributive, perfect, and conjunctive. I will skip the perfective inflection but will include the conjunctive, which is found in a number of environments, including as the verbal form for conjuncts. I include it because m&e use it to make an argument regarding the conclusive form. (11) Murasaki Shikibu diary: Preliminary what?

Inflection Conclusive Attributive Conjunctive

Object np with -wo Object np without -wo 30 53 90

56 46 92

These raw data from msd offer many apparent counterexamples to the distribution statement in (1). There are thirty object nps with -wo occurring with the conclusive form. Also, there are forty-six instances of an object np without -wo occurring with the attributive form. What we see with the conjunctive form is that this form freely selects between the morphological case marker and abstract case, so the two possibilities are virtually even, a point that becomes important for dealing with some of the apparent counterexamples. As noted by m&e, once the raw data above are analyzed and certain special cases are eliminated, the results are much more in tune with what I argued in Miyagawa (1989) based on the Man’yōshū (see also Matsunaga 1983; Miyagawa and Matsunaga 1986). It is possible to eliminate twenty-five of the conclusive examples that are deviant and thirty-seven of the attributive examples that are deviant.

166

miyagawa

(12) Murasaki Shikibu diary: Final

Inflection Conclusive Attributive Conjunctive

Object np with -wo 5 53 90

(9%) (85%) (49%)

Object np without -wo 56 9 92

(91 %) (15 %) (51 %)

The conclusive form overwhelmingly selects abstract case, while there is a strong tendency for the attributive form to select the morphological case marker -wo. The conjunctive form remains split virtually evenly between the morphological case marker and abstract case, showing that it optionally assigns abstract case.3 The final data show that the observation originally made in Miyagawa (1989) and Matsunaga (1983) based on the poetry of the Man’yōshū is upheld even for prose. This answers Kinsui’s criticism that poetic versification may have skewed the distribution of abstract and morphological case independent of verbal inflection. As we will see later in the chapter, one interesting possibility does arise where the occurrence of -wo may be conditioned by versification without violating any grammatical principles. Below, I will summarize some of Miyagawa and Ekida’s discussion of how we dealt with the apparent counterexamples involving the occurrence of -wo with the conclusive form, keeping the attributive form in abeyance until section 3. First, as Miyagawa and Ekida note, there are a number of lexical idiosyncrasies that force the occurrence of -wo regardless of verbal inflection. There are, for example, a number of idioms that are frozen in form with the -wo particle, and these account for some of the thirty deviant conclusive examples. I will not deal with these lexical idiosyncrasies in this paper. What I will deal with are three types of apparent counterexamples that have a common nature, namely, the conclusive form in these examples is prevented from assigning abstract case, so -wo is inserted to meet the requirement of case for the object np. To set the stage, I will briefly review the difference between conclusive and attributive forms that give rise to different case-marking possibilities. The conclusive form is a “true verb form” used to predicate an action, property, or state (Sansom 1928: 130). As a pure predicate, we can surmise that it has all of the properties of a verb, including the capability of assigning abstract case.4 In 3 On this optionality with the conjunctive, see note 6. 4 In recent theory, accusative case assignment is done not directly by the v, but by what is called

the old japanese accusative revisited

167

contrast, the attributive paradigm has “substantive” properties, which makes it similar to a nominal. Konoshima (1966), for example, notes that the attributive inflection has a nominalizing function. The following example from Sansom (1928: 136), demonstrates three substantive qualities of the attributive form.5 (13) pîtö-nö mîtöŋ gam-uru-wo sir-anz-u people-gen blaming-attr-acc know-neg-inf ‘not knowing that others blamed them’ First, the attributive form mîtöŋ gam-uru ‘the act of blaming’ has a substantive interpretation similar to that of the English gerund. Second, the particle -wo attaches to it to make the phrase an argument of the verbal form sir-anz-u ‘without knowing’. Third, the subject of mîtöŋ gam-uru has the genitive case marker, which is a hallmark of np arguments in nominal clauses. These three properties make the attributive form appear nominal in nature. It would be incorrect, however, to identify it as a pure nominal, because it has verbal and adjectival properties as well. For example, it is able to modify a noun without the prenominal genitive particle -no. In (3) above, for example, if the attributive form miru ‘look’ were a pure nominal, we would expect the prenominal modification particle -no between it and the relative head. In Miyagawa (1989), a parallel is drawn between the substantive nature of the attributive form and the gerund in English. The latter requires insertion of of for case. (14) the teaching of calculus The idea is that, due to its substantive nature, the attributive form is also unable to assign abstract case, and -wo is inserted to meet the case requirement, just as of is inserted above. In modern theory, this nominalizing behavior of the attributive inflection would naturally be viewed as nominalizing the “small v” (Yanagida and Whitman 2009), in turn depriving the predicate of the ability to assign abstract case. However, to be consistent with Miyagawa (1989), I will continue to speak of the attributive form itself as lacking this ability. Below, we will see that some of the apparent counterexamples to the idea that the conclusive form assigns abstract case are examples in which the cona “small v” that selects vp (e.g., Chomsky 1995). I will assume this, but will continue to describe case assignment in terms of verbs. 5 This example, quoted in Sansom 128, xi, is from Shoku Nihongi, a work that contains “certain Imperial edicts in pure Japanese” and was completed in 797.

168

miyagawa

clusive form is somehow prevented from exercising its normal abstract-caseassigning function, forcing -wo to be inserted. 2.2 When -wo Occurs with the Conclusive form As shown in (11), there are thirty examples of this type that seem to contradict the prediction that objects of the conclusive form should not have -wo because the verb is able to assign abstract case. Miyagawa and Ekida account for all but five of these. I will summarize three of Miyagawa and Ekida’s accounts, which have the common property of demonstrating a way in which the conclusive form is prevented from assigning abstract case. 2.2.1 Exceptional Case Marking (ecm) A defining property of abstract case is that it must be assigned under adjacency with the verb (Stowell 1981). Thus, in English, normally nothing can intervene between the verb and its object, unless the object is made “heavy” and is able to undergo heavy-np shift. (15) a. *John read yesterday a book. b. John read yesterday a book with 23 chapters. Heavy-np shift One of the msd examples in which the object has -wo despite the fact that the verb is in the conclusive form is the following. (16) Sakizaki-no miyuki-wo nadote meiboku-arite-to past-gen visits-acc why honor-cop-comp omof-i-tamaf-i-kem-u. think-inf-hon-pastspeculative-fin ‘Why did I feel my previous visits as such an honor?’

(36.11)

We can see that this is an ecm construction because it is the subordinate subject ‘previous visits’ that is marked with -wo, and the matrix verb ‘think’ is a typical ecm verb. Similar examples occur in modern Japanese (Kuno 1976). (17) a. Taroo-ga Hanako-ga tensai-da-to omotteiru. Taro-nom Hanako-nom genius-cop-comp think ‘Taro thinks Hanako is a genius.’ b. Taroo-ga Hanako-o tensai-da-to omotteiru. Taro-nom Hanako-acc genius-cop-comp think

ecm

the old japanese accusative revisited

169

The accusative subject is licensed by the matrix verb ‘think’ in both the oj example (16) and in the modern-Japanese example (17b). The accusative np in (16), by virtue of being embedded in an ecm clause, does not occur adjacent to the matrix verb, which therefore cannot assign abstract case to it. For this reason, morphological case is thus inserted on the subordinate subject. This eliminates one of the potential counterexamples from the msd. One of Kinsui’s (1993) putative counterexamples from the Man’yōshū is also an ecm construction, as he himself notes. (18) Yö-nö naka-wo usi-tö yasasitö omop-ëndömo world-acc unpleasant-comp shameful-comp think-although tobitachi kane-t-u tori n-i shi ar-an-emba. fly away cannot-perf-fin bird cop-inf ep exist-neg-ev-con ‘Although I feel the world as being unpleasant and unbearable, I cannot fly away as I am not a bird.’ 2.2.2 Emphasis A number of the apparent counterexamples involve an object that is marked by -wo because the object is emphasized; -wo is known to have this emphatic function. (19) sore-wo ware masarite if-am-u to it-acc I more than speak-tent-fin comp ‘that I speak about it more than (others do)’ (msd 78.6)

(78.6)

This happens to be an example in which the conclusive form occurs in a subordinate clause, something made possible by the complementizer form to. Is it the case that -wo is freely inserted on accusative objects whenever emphasis is desired? Note that the object in (19) occurs at a distance from the verb ‘speak’. This suggests that emphasis may have the effect of moving the object away from its original position and hence away from the verb; this makes assignment of abstract case impossible, so -wo is inserted, just as in the ecm construction. In the following example, the object is, on the surface, adjacent to the verb, but we can surmise that it has moved away from its original complement position. (20) … mi-tyau-no uti-wo tofor-ase-tamaf-u. screen-gen inside-acc pass-caus-hon-fin ‘Let … pass inside the screen.’

(43.11)

170

miyagawa

See Yanagida (2006) for an argument that in oj, all -wo-marked phrases move to a position outside the verbal projection; this may be related to a role for -wo in marking definiteness (Motohashi 1989) or specificity (Yanagida and Whitman 2009). 2.2.3 Compounding Five of the apparent counterexamples involve compound verbs. Compounding apparently deprives the verb of its ability to assign abstract case, forcing -wo to be inserted. Following is one such example. (21) Sirokane-no su-wo fitobito tuk-i-sirof-u. silver-gen cover-acc people poke-rec-fin ‘People laugh amongst themselves at the silver cover.’

(25.1)

The compound verb tuk-i-sirof- is made up of tuk- ‘poke’ and sirof-, reciprocal auxiliary m&e note that compound verbs almost always require -wo on the object regardless of the verbal inflection, suggesting that compounding somehow deprives the verb of the ability to assign abstract case. In this example, the object with -wo has moved above the subject, which also makes it impossible to assign abstract case. There are examples where the object stays adjacent to the verb but nevertheless -wo is required even if the verb is in the conclusive form. This is true for the msd, and it is also true for the other major work that m&e analyzed, the Izumi Shikibu diary. The evidence m&e give has to do with the conjunctive inflection. Recall that the conjunctive form optionally assigns abstract case, and in the literary works analyzed, objects with and without -wo are evenly split; we can see this even split in (12) (ninety to ninety-two). m&e report a similar even distribution in the Izumi Shikibu diary. However, when we look at compounds in the conjunctive inflection, there is a clear pattern of obligatory -wo. The following is data from the Izumi Shikibu diary taken from m&e. (22) Conjunctive compounds in the Izumi Shikibu diary

Object np with -wo 10

(91%)

Object np without -wo 1

(9%)

Just as with ecm and emphasis, the verb’s apparent inability to assign abstract case when it is a compound, even when it has conclusive inflection, leads to -wo

the old japanese accusative revisited

171

being inserted. Although m&e do not give an explanation, the phenomenon may be similar to what we see with gerunds in English, in that adding certain morphological structure to a verb (-ing in English, compounding or attributive inflection in Japanese) eliminates the possibility of assigning abstract case. That is not an explanation, of course, and I will leave the issue open. 2.3 Interim Conclusion See m&e for accounts of other apparent counterexamples to the idea that conclusive form assigns abstract case. All told, m&e account for all but five of the thirty apparent counterexamples in the msd. Finally, let us return to the question of versification and the poem from the Izumi Shikibu diary given in (10) above, which is repeated below. (23) Ofokata ni Samidaruru to ya Omofuramu Kimi ___ koiwataru Kefu no nagame wo

Nothing remarkable— The same old rain that pelts us Every year, you think? These are my tears of love Falling in a deluge all day long!

The fourth line contains the object kimi ‘you’ in its bare form without -wo despite the fact that the verb koiwataru ‘longing for’ is in the attributive. The absence of -wo makes it possible to fit the line into a seven-mora metrical pattern, as required by the tanka verse form. As an alternative to the idea that -wo is omitted here for the sake of the meter—or as an explanation of what permits the omission of -wo for the sake of the meter—m&e note that the use of the verb koiwataru is largely limited to poetry, and it typically occurs with kimi ‘you’ or imo ‘wife’, so that the combination kimi koiwataru formed a “poetic expression” independent of the verbal inflection. This is one possible explanation. Below, we turn to the work of Yanagida (2007; see also Yanagida and Whitman 2009), which provides a very different analysis, in line with the idea that -wo can be excluded here to respect versification because the case requirement is met by a means other than abstract or morphological case.

3

Incorporation and the Attributive: Yanagida (2007) and Yanagida and Whitman (2009)

In this section, I turn to the attributive form, specifically addressing apparent counterexamples to the hypothesis that the attributive form, being nominal in nature, cannot assign abstract case and therefore requires the morphological

172

miyagawa

case marker -wo to accompany the object. As we saw in (11), the msd contains a number of superficially deviant cases where the object occurs bare despite the verb being in the attributive inflection. While there are fifty-three instances of the attrbutive -wo, which is what we expect, there are forty-six examples where the object of the attributive is bare. Kinsui (1993) notes similar potential counterexamples from the Man’yōshū. m&e provide various explanations for all but five of the forty-six msd putative counterexamples. For example, three of them contain an object np with the adverbial particle -nado ‘such as’ and one contains an object np with -bakari ‘only’; these adverbial particles make it unnecessary for morphological case marking to appear even in modern Japanese. 3.1 Yanagida’s Discovery Yanagida (2007) proposes a unified explanation for most of the oj examples of bare objects with the attributive, one that cuts across the case-by-case accounts in m&e, based on her study of the Man’yōshū (see also Yanagida and Whitman 2009). Yanagida found that, in the Man’yōshū, there are ninety examples of transitive clauses that contain a bare object, and of these, fifty-five have the attributive form. The following is an example of the latter from her article (quoted in Yanagida and Whitman 2009; I retain their orthographic system, while adjusting the glosses in minor ways). (24) Sayôpîmê-nö kô-ga pire pur-i-si yama. Sayohime-gen child-nom scarf wave-inf-past/attr mountain ‘the mountain where Sayohime waved her scarf’ (mys 5.868) Here the verb pur- ‘wave’ is in the past attributive form pur-i-si and yet its object, pire ‘scarf’, occurs without -wo. Here is Yanagida’s discovery. (25) Yanagida’s (2007) discovery Of the fifty-five examples from the Man’yōshū where bare objects occur with the attributive form, in fifty-four (all but one) the objects are nonbranching nouns—single words, in other words. We see an example in (24) above, in which the object pire is a single word. Why should this be the case? Yanagida (2007) argues that this overwhelming tendency for bare objects of attributive forms to be nonbranching indicates a third way in which objects can be licensed to meet the case requirement. Baker (1988) shows that in a number of languages, the case requirement on the object is met not by abstract or morphological case, but by the object

the old japanese accusative revisited

173

incorporating into the verb. This is shown in the Chukchee example below, taken from Spencer (1999) and cited in Yanagida and Whitman (2009) (I have simplified the glosses). (26) a. Muri myt-ineretyrkyn kimitʔ-e. we we-are carrying load-instr ‘We are carrying the load.’ b. Ytlygyn qaa-tym-gʔe. father deer-killed-3sg ‘The father killed a deer.’ In (26a) the object occurs with an overt case marker (instrumental), but in (26b), the single-word object qaa ‘deer’ (a nonbranching n) has incorporated into the verbal structure in order to meet the case requirement. It is an important fact that incorporation occurs only with a head, because it is a kind of morphological process occurring in syntax. If Yanagida is correct in attributing some oj bare objects to head incorporation, it means that in oj we find three ways to license case on the object. (27) Licensing case on the oj object (i) Abstract case (ii) Morphological case (iii) Head incorporation While the first two ways were noted in Miyagawa (1989), the third is a new insight that cuts across most of the example-by-example explanations offered by m&e and provides a unified and dramatic account of much of the problem data related to the attributive form. In the remainder of the chapter, I will attempt to replicate Yanagida’s discovery made on the basis of poetry in the Man’yōshū by looking again at the Heian prose studied in m&e. 3.2 Replicating Yanagida’s Discovery in Heian Literature m&e list in appendices all of the potential counterexamples from the msd and also the Izumi Shikibu diary. I examined these examples to see if it is possible to replicate in tenth-century Heian prose the generalization stated in (25) that Yanagida made on the basis of poetry in the eighth-century collection the Man’yōshū. Setting aside the examples mentioned earlier of objects with an adverbial particle such as -nado ‘such’ and -bakari ‘only’, which do not require

174

miyagawa

case marking, the following are the numbers of bare objects with attributive form divided into nonbranching (single word) and branching (phrasal) types. (28) Nonbranching and branching bare objects with attributive form

Nonbranching Branching Murasaki Shikibu diary Izumi Shikibu diary

44 13

7 6

The msd clearly evidences the pattern of Yanagida’s discovery: of the 51 bare objects with the attributive form, only seven are branching. An equally striking pattern worth mentioning is that, among the examples that m&e were unable to account for with their case-by-case explanations, of which there are nine, only one is branching. Unlike the msd, the Izumi Shikibu diary does not show a clear pattern of incorporation: of the nineteen relevant examples, almost a third are branching. However, there is something noteworthy about all six of the branching examples. All involve the formal nominalizer koto. An example is given below. (29) mutukasi-ki koto ___ ifu-wo kik-osimes-i-te disturbing-attr thing say-acc hear-hon-inf-sub ‘hearing/heard that (someone) says disturbing things, and …’

(444.9)

m&e note this fact as well, and simply stipulate that a phrase with the formal nominalizer koto does not require case. If we exclude these examples, the Izumi Shikibu diary replicates Yanagida’s discovery without exception. I also note that in the msd, of the seven branching bare objects, two are of this type in which the object is headed by koto. Of the remaining six, two involve the light verb su or its related causative form, and m&e note that these verbs tend to allow the object to occur without -wo in any inflection.6 Why should the occurrence of koto (or its voiced alternate form goto) allow an object that branches to occur without -wo even when the verb is in the 6 The three remaining exceptions from the msd are the following. (i) mono-no kazukazu ___ kak-i-tar-u fumi thing-gen many write-inf-perf-attr letter ‘document in which (one) wrote many things’

(27.3)

the old japanese accusative revisited

175

attributive form? One possibility is that the clause headed by koto is not an np, but a cp, and cps do not require case. This is simply a speculation, but the idea that koto may function as either c or n has been proposed for modern Japanese by Whitman (1992). Finally, let us return one more time to the poem from the Izumi Shikibu diary (23) in which the object, kimi ‘you’, occurs without -wo despite the fact that the verb koiwataru ‘longing for’ is in the attributive, allowing the line to adhere to the seven-mora requirement. (30) Kimi ___ koiwataru The object is a single word, which means that its case requirement can be met by incorporation. In a case like this, with an attributive verb and a nonbranching object, there is an option of incorporation or assigning -wo to the object, and this option gives the poet the license to choose between them in order to respect the versification.

4

Conclusion

I demonstrated that the work of m&e and of Yanagida (2007) and Yanagida and Whitman (2009) support a systematic distribution of abstract and morphological case. Most of the putative counterexamples to the proposed distribution, including those pointed out by Kinsui (1993), find explanation in independently motivated notions such as adjacency for abstract case assignment and head incorporation to fulfill the case requirement. One interesting result that came out of applying Yanagida’s (2007) and Yanagida and Whitman’s (2009) head-incorporation analysis was an account of the limited optionality of -wo: so long as the verb is attributive and the object is nonbranching, the poet in the oj era could select either of the applicable licensing options in order to respect the versification of poetry of the time while being fully compliant with the grammar of oj. (ii)

fito-no fazi ___ mi-faber-i-si yo person-gen disgrace see-humble-inf-past/attr night ‘the night that (someone) saw a person’s disgrace’

(29.1)

(iii)

sodeguti-no afafi ___ waro-u kasane-taru fito sleeve-gen coloration bad-inf lay(inf)-perf-attr person ‘a person who layered (her) sleeves in an unpleasant manner’

(88.1)

176

miyagawa

References Baker, Mark 1988. Incorporation: A theory of grammatical function changing. Chicago: University of Chicago Press. Chomsky, Noam 1981. Lectures on Government and Binding: The Pisa lectures. Dordrecht: Foris. Chomsky, Noam 1995. The Minimalist Program. Cambridge, Mass.: mit Press. Cranston, Edward (trans.) 1969. Izumi Shikibu Diary. Cambridge, ma: Harvard University Press. Heike Monogatari-Habiyan shoo kirishitan-ban. 1966. Tokyo: Yoshikawa Kōbunkan. Heike Monogatari 1,2. 1973. Tokyo: Shōgakkan. Hirohama, Fumio 1966. ‘On o,’ Kokugogaku to Kokubungaku. Ikeda, Kikan (ed.) 1953. Genji Monogatari taisei. 1953. Tokyo: Chuōkōronsha. Izumi Shikibu Nikki. In: Tosa Nikki, Kagerofu Nikki, Izumi Shikibu Nikki, Sarashina Nikki: Iwanami koten bungaku taikei 20. 1957. Tokyo: Iwanami Shoten. Jidai-betsu kokugo daijiten (Muromachi jidai hen) 1985. Tokyo: Sanseidō. Joseph, Brian 1978/1991. Morphology and Universals in Syntactic Change: Evidence from medieval and Modern Greek. Ph.D. dissertation, Harvard University (printed and distributed by Indiana University Linguistics Club; updated version published in Outstanding Dissertations in Linguistics Series by Garland Publishing, Inc.). Joseph, Brian. 1980. ‘Linguistic universals and syntactic change’. Language 56, pp. 345– 370. Kinsui, Satoshi 1993. ‘Kotengo-no ‘wo’ ni tsuite’. In: Yoshio Nitta (ed.), Nihongo-no Kakuo Megutte. Tokyo: Kuroshio, pp. 191–224. Kitagawa, Hiroshi and Tsuchida, Bruce T. (trans.) 1975. The Tale of the Heike. Tokyo: University of Tokyo Press. Kobayashi, Yoshiharu 1970. Nihon bunpō-shi [History of Japanese Grammar]. Tokyo: Tokyo Shoin. Kogo daijiten 1982–1987. Vol. 1–3. Tokyo: Kadokawa Shoten. Kokinshū. 1975. Tokyo: Shintensha. Konoshima, Masatoshi 1966. Kokugo Joshi no Kenkyū [Study of Particles in Japanese]. Tokyo: Ofūsha. Kuno, Susumu 1976. ‘Subject Raising.’ In Masayoshi Shibatani (ed.), Syntax and Semantics 5: Japanese Generative Grammar. New York: Academic Press, pp. 17–49. Lightfoot, David. 1979. Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Mabuchi, Kazuo 1971. Kokugo On’inron [ Japanese Phonology]. Tokyo: Kasama Shoten. Matsunaga, Setsuko 1983. Historical development of case marking in Japanese. Unpublished m.a. thesis, Ohio State University. Matsuo, Osamu 1938. ‘Heian shoki ni okeru kakujoshi wo [The case marker wo in early Heian]’. Kokugo to Kokubungaku 15, pp. 1389–1412.

the old japanese accusative revisited

177

Matsuo, Osamu. 1944. ‘Kyakugo hyōji-no joshi o- ni tsuite’. In: Hashimoto Hakase and Kanreki Kinenkai (eds.), Kokugogaku Ronshu. Tokyo: Iwanami, pp. 617–644. Matsuo, Osamu 1969. ‘wo-kakujoshi (kotengo, gendaigo) [The case marker wo in Classical Japanese and modern Japanese]’. In Akira Matsumura (ed.) Joshi Jodōshi Yōsetsu. Tokyo: Gakutōsha. Miyagawa, Shigeru 1989. Structure and Case Marking in Japanese. New York: Academic Press. Miyagawa, Shigeru and Setsuko Matsunaga 1986. ‘Historical development of the accusative case in Japanese’, Journal of Asian Culture ix, pp. 87–101. Miyagawa, Shigeru and Fusae Ekida 2003. ‘Historical development of the accusative case marking in Japanese as seen in classical literary texts’. Journal of Japanese Linguistics 19, pp. 1–105. Motohashi, Tatsushi 1989. Case Theory and the History of the Japanese Language. Unpublished Ph.D. dissertation, University of Arizona, Tucson, Arizona. Murasaki Shikibu Nikki, in Iwanami Bunko, Murasaki Shikibu Nikki. 1984. Tokyo: Iwanami Shoten. Nihon Bunpō Daijiten 1971. Tokyo: Meiji Shoin. The Nippon Gakujutsu Shinkōkai. 1965. The Man’yōshū. New York: Columbia University Press. Ono, Susumu, Akihiro Satake, and Kingoro Maeda (eds.) 1974. Iwanami Kogo Jiten. Tokyo: Iwanami Shoten. Pierson, Jan L. 1933. The Man’yôśû. Vol. ii. Leyden: Brill. Poser, William 1981. ‘The double-o constraint in Japanese.’ mit ms., Cambridge, ma. Sansom, George B. 1928. A Historical Grammar of Japanese. Oxford: Clarendon Press. Sarashina Nikki Sōsakuin 1979. Tokyo: Musashino Shoin. Sato, Kiyoji 1977. Kokugogaku Kenkyū Jiten (Dictionary of Japanese Grammar). Tokyo: Meiji Shoin. Shibatani, Masayoshi 1990. The Languages of Japan. Cambridge, University Press. Shogakkan Kogo Daijiten. 1983. Tokyo: Shoogakkan. Spencer, Andrew 1999. Chukchee homepage. Online course notes, University of Essex, Colchester, England. http://privatewww.essex.ac.uk/~spena/Chukchee/ CHUKCHEE_HOMEPAGE.html. Stowell, Timothy 1981. The Origins of Phrase Structure. Upublished Ph.D. dissertation, mit. Suzuki, Tomi 1973. Heike Monogatari, Amakusa-bon Heike Monogatari taihi ni yoru kakuhyōgen hattatsu katei no ichi danmen. Kokugogaku ronsetsu shiryō 10. Ronsetsu Shiryō Hozon Kai. Tokyo. Takagi, Ichinosuke, et al. (trans. and eds.) 1957. Man’yōshū, vol. i. Nihon Koten Bungaku Taikei, vol. iv. Tokyo: Iwanami Shoten. Takagi, Ichinosuke, et al. (trans. and eds.) 1962. Man’yōshū, vol. iv. Nihon Koten Bungaku Taikei, vol. viii. Tokyo: Iwanami Shoten.

178

miyagawa

Seidensticker, Edward (trans.) 1981. The Tale of Genji. New York: Alfred A. Knopf. Tosa Nikki. 1930. Tokyo: Iwanami Shoten. Whitman, John 1992. ‘String vacuous v to Comp.’ Ms., Cornell University, Ithaca, n.y. Paper presented at the 1991 glow. Wrona, Janick and Bjarke Frellesvig 2009. ‘The old Japanese case system’. Japanese/ Korean Linguistics 17, csli, Stanford, pp. 565–579. Yamada, Masahiro 2000. ‘Shugo hyōki ga no seiryoku kōdai no yōsō’. Kokugogaku 51, pp. 1–14. Yanagida, Yūko 2007. ‘Miyagawa’s (1989) exceptions: an ergative analysis’. In: Yoichi Miyamoto and Masao Ochi (eds.), fajl 4: Proceedings of the Fourth Conference on Formal Approaches to Japanese Linguistics. Cambridge: mit Working Papers in Linguistics 55, pp. 265–276. Yanagida, Yūko and John Whitman 2009. ‘Alignment and word order in Old Japanese’. Journal of East Asian Linguistics 18, pp. 101–144. Zenno, Yasushi 1987. Object case marking in Old Japanese. ucla, ms.

chapter 15

Japanese Wh-Phrases as Unvalued Operators Mamoru Saito*

1

Introduction

The purpose of this paper is to suggest a new direction in the analysis of whexpressions like dare ‘who’ and nani ‘what’ in Japanese. In the literature on wh-questions such as Lasnik & Saito (1984) and Richards (2000), these are taken to be interrogative wh-phrases exactly as their English counterparts in wh-questions. However, this view is probably too simplistic as discussed in Nishigauchi (1990), for example. These expressions are interpreted differently in different contexts as the following examples show: (1) Taroo-wa [[Hanako-ga nani-o tabeta] ka] sitteiru Taroo-top Hanako-nom what-acc ate q know ‘Taroo knows what Hanako ate.’ (2) [[Nani-o tabeta hito] mo] manzokusita what-acc ate person also was.satisfied ‘For every x, x a thing, the person that ate x was satisfied.’ In (1), nani seems to function as an interrogative wh-phrase. But it is part of universal quantification in (2). The quantificational force originates in the particles, ka and mo, instead of nani. Kuroda (1965) made this observation and called the wh-words ‘indeterminate pronouns’. He explicitly states that they are “yet unbound variables.” Nishi-

* I am happy to be able to contribute this paper to a volume in honor of John Whitman. His work has provided me with constant inspiration for over 33 years, since we became friends as first-year graduate students in Cambridge, Mass. in 1980. The material in this paper was presented in the 2013 advanced syntax course at Nanzan University. I would like to thank Mareaki Hayashi, Yuma Iwatani, and Shizuka Kato for helpful discussion. I also benefited from discussions over the years with Tomohiro Fujii, Kensuke Takita and W.-T. Dylan Tsai. Finally, thanks are due to Bill McClure for carefully going over the manuscript and making excellent suggestions for improvement. This research was supported in part by the Nanzan University Pache Research subsidy i-a-2 (2013).

© koninklijke brill nv, leiden, 2018 | doi: 10.1163/9789004351134_016

180

saito

gauchi (1990) develops this idea and proposes an analysis in terms of Heim’s (1982) unselective binding. According to this analysis, mo provides the force of a universal quantifier and nani serves as a variable in (2). He also observes a locality restriction between a wh-expression and the associated particle and proposes that a wh-phrase covertly moves to the vicinity of the particle. On the other hand, analyses in terms of particle movement are proposed in Maki (1995) and Hagstrom (1998) for wh-questions, and Takahashi (2002) for other quantificational constructions. Takahashi argues against the unselective binding analysis and proposes that mo in (2), for example, originates with nani and moves to its surface position. In this paper, I suggest that wh-expressions are operators without specific quantificational force. Then, they must move covertly to positions that allow them to probe for particles and have their quantificational force valued. The analysis I present is similar to Nishigauchi’s, which assumes covert phrasal movement. It also relies heavily on insights from Takahashi (2002). Yet, it is different in one fundamental respect from their analyses as well as the semantic analysis proposed in Shimoyama (2001). They all follow Kuroda (1965) and assume that wh-expressions are interpreted as variables. I question this widely held assumption. In the following section, I briefly go over Nishigauchi’s (1990) analysis, illustrating the basic facts of wh-expressions in Japanese at the same time. In Section 3, I discuss Takahashi’s (2002) argument against the unselective binding analysis and motivate the idea that wh-expressions in Japanese are not construed as variables but are operators. In Section 4, I outline the analysis and discuss some consequences. Section 5 concludes the paper.

2

Nishigauchi’s (1990) Unselective Binding Analysis

Wh-expressions in Japanese are employed in various contexts as illustrated in (3) with dare ‘who’. (3) a. [[Dare-ga kuru] ka] osiete kudasai who-nom come q teach please ‘Please tell me who is coming.’ b. [dare-mo-ga yomu] hon who-also-nom read book ‘the book that everyone reads’

(interrogative wh)

(universal quantifier)

181

japanese wh-phrases as unvalued operators

c. [dare-ka-ga suteta] hon who-or-nom discarded book ‘the book that someone discarded’

(existential quantifier)

d. [dare-de-mo motteiru] hon who-Cop.-also have book ‘the book that anyone has’ e. [dare-mo yomanai] hon who-also read-not book ‘the book that no one reads’

(free choice)

(negative polarity)

It should be clear from these examples that the particles in roman play crucial roles in the interpretation. Further, as shown in (2), not only the question particle ka but other particles as well can appear separated from the whexpression. Another relevant example from Takahashi (2002) is given in (4). (4) [[Dare-ga kaita hon] mo] omosiroi who-nom wrote book also interesting ‘For every x, x a person, the book that x wrote is interesting.’ In this example, mo appears on the relative head whereas dare is contained within the relative clause. While forms like dare-mo in (3b) may be lexicalized, the relation between dare and mo in (4) is clearly mediated by syntax.1 Hence, I focus on cases where a wh-expression and the associated particle appear separately as in this example. As the relevant cases are observed most extensively with the question particle ka and the universal particle mo, I mainly discuss these in what follows. As noted above, Nishigauchi (1990) proposes an analysis in terms of unselective binding. Let us first briefly review Heim’s (1982) unselective binding before discussing his analysis. Heim first assumes that quantification is represented with a tri-part clausal structure as in (5b).

1 For universal quantification, the pattern in (4) with mo is quite productive. On the other hand, the form ‘wh-word+mo’ is idiosyncratic. For example, dare-mo ‘who+mo’ is allowed, but nanimo ‘what+mo’ is not. Further, the particle mo usually cannot be followed by a Case marker, but dare-mo in (3b) must be. These facts suggest that dare-mo is lexicalized as a noun meaning ‘everyone’.

182

saito

(5) a. Everyone is smart b. [Everyx] [x is a person] [x is smart] c. [Every x: x is a person] [x is smart] The second part ‘x is a person’ is the restriction on the quantifier, and the last part is the core sentence. (5b) is expressed as (5c) in more familiar notation. Given this, (6a), an example of donkey anaphora, would be initially assigned the structure in (6b), pending the interpretation of the indefinite a painting. (6) a. Everyone who bought a painting was satisfied with it b. [Everyx] [x is a person and x bought a painting] [x was satisfied with it] c. [Everyx, y] [x is a person, y is a painting, and x bought y] [x was satisfied with y] The final structure to be obtained is (6c). Then, Heim proposes that indefinites are interpreted as variables and are unselectively bound by quantifiers. Given this, in (6b), a painting is interpreted as a variable bound by every. With a proper representation of the restriction, we derive (6c). Although the universal quantifier that unselectively binds the indefinite is explicit in (6a), it can be implicit. Thus, (7a), another typical example of donkey anaphora, is analyzed as in (7b). (7) a. If a man sees a panda, he chases it b. [Everyx, y] [x is a man, y is a panda and x sees y] [x chases y] Conditionals, Heim points out, inherently have universal force. If in (7a), for example, can be replaced by whenever. Then, the implicit universal quantifier binds the indefinites, a man and a panda, and the corresponding pronouns, he and it, as in (7b). Nishigauchi (1990) proposes that wh-expressions in Japanese, like indefinites, are interpreted as variables and that the associated quantificational particles serve as unselective binders. Thus, the embedded question of (1) receives interpretation as in (8). (8) [qx] [x is a thing] [Hanako ate x] Nani is interpreted as a variable, with the restriction that it is a thing, and ka as an interrogative operator. Similarly, (4) receives interpretation as in (9a–b).

japanese wh-phrases as unvalued operators

183

(9) a. [Everyx] [x is a book and dare wrote x] [x is interesting] b. [Everyx, y] [x is a book, y is a person, and y wrote x] [x is interesting] Mo, like every in (6a), is assumed to be a universal quantifier on the relative head. This leads to the representation in (9a). Then, dare, like the indefinite in (6), is unselectively bound by the universal quantifier and consequently, the interpretation in (9b) obtains. Nishigauchi notes at the same time that the relation between a wh-expression and the associated particle is subject to a locality requirement. The whisland effect is observed as in (10). (10) Taroo-wa [[dare-ga kuru ka] kiite] mo ikanai daroo Taroo-top who-nom come q hear even go-not will a. Taroo won’t go even if he hears who is coming. b. * For every x, x a person, Taroo won’t go even if he hears whether x is coming. Ka is interpreted as ‘whether’ and mo as ‘also, even’ when they are not associated with wh-expressions. Hence, (10) should be ambiguous between a and b depending on whether dare is associated with ka or mo. But the interpretation in b is totally impossible. Dare in this example can only be part of a question and cannot yield universal quantification with mo. This indicates that the intervening ka blocks the association of mo with dare. The wh-island effect in (10) suggests that a wh-expression moves covertly to the vicinity of the associated particle. Nishigauchi, then, proposes that a government relation between the binder and the bindee is required for unselective binding to obtain. Nishigauchi (1990) considers important data and accommodates them in his analysis. At the same time, the appeal to both unselective binding and covert movement is a little puzzling. Unselective binding applies to two independent elements: one is a quantifier and the other is a phrase that is interpreted as a variable. It is not clear why it should require movement. It is interesting in this connection that while Nishigauchi (1990, 1991) appeals to unselective binding in his analysis, he also hints that wh-expressions in Japanese have quantificational forces of their own. He states that “the quantificational force of a wh-expression” is determined by “a quantificational particle.” This is not consistent with Kuroda’s characterization of wh-expressions as “indeterminate pronouns,” which Nishigauchi aims to develop in his analysis. But I pursue this alternative conception of wh-expressions in Section 4.

184 3

saito

Evidence against the Unselective Binding Analysis

In this section, I first go over Takahashi’s (2002) argument against the unselective binding analysis of Japanese wh-expressions. Then, building on his insights, I motivate the idea, hinted at in Nishigauchi (1990, 1991), that whexpressions are operators that receive quantificational force from quantificational particles. Recall that unselective binding takes place between a quantifier and an independently generated expression that is construed as a variable. This can be observed, for example, in (6). As indicated in (6b), ‘Everyx’ binds the variable ‘x’ in the subject position and also in the restriction. Then, it enters into an unselective binding relation with the indefinite a painting and yields (6c). Nishigauchi (1990) assumes that this happens also in the binding relation between a quantificational particle and a wh-expression, as his analysis of (4), illustrated in (9), shows. Mo is assumed to universally quantify over the relative head hon ‘book’ and then, to unselectively bind dare. (9b) indeed seems to be the correct interpretation of (4). However, Takahashi (2002) points out that mo is not interpreted as a universal quantifier independently of a wh-expression. First, (11a) and (11b) differ in meaning. (11) a. Taroo-wa dono hon-mo yonda Taroo-top which book-also read ‘Taroo read every book.’ b. Taroo-wa hon-mo yonda Taroo-top book-also read ‘Taroo read also a book.’ not ‘Taroo read every book.’ Mo in (11a) serves to express universal quantification over books, but that in (11b) simply means ‘also’. Further, Takahashi observes that the analysis in (9) faces a problem when omosiroi ‘interesting’ is changed to a stage-level predicate. His example is given in (12). (12) [[Dare-ga kaita hon] mo] tosyokan-ni aru who-nom wrote book also library-in is ‘For every x, x a person, there is (also) a book that x wrote in the library.’ (12) does not mean that the library has every book of every author. It is true if there is at least one book for each author in the library. Then, (4) is construed

japanese wh-phrases as unvalued operators

185

with universal quantification over books not because mo quantifies over books but because of an independent factor.2 Takahashi concludes then that mo is interpreted as a universal quantifier only with respect to a wh-expression. In addition, wh-expressions in Japanese, in contrast with indefinites, are not interpreted as variables whenever there are quantifiers to bind them. (13a) is an example of donkey anaphora with an indefinite gakusei ‘student’. (13) a. Gakusei-ga ku-reba, boku-wa pro au student-nom come-if I-top meet ‘For every x, x a student, if x comes, I will meet x.’ b. * Dare-ga ku-reba, boku-wa pro au who-nom come-if I-top meet Heim’s (1982) analysis correctly predicts the interpretation because indefinites are construed as variables and conditionals have universal force. But when a wh-expression dare ‘who’ is substituted for the indefinite, the sentence is totally ungrammatical as in (13b). This shows that a wh-expression in Japanese creates a quantifier-variable relation only with a quantificational particle.3 Takahashi concludes that only the combination of a wh-expression and mo yields universal quantification. He then proposes the particle movement analysis illustrated in (14) for (12).

(14) [[Dare-mo-ga

kaita

hon]

]

tosyokan-ni aru (= (12))

Although this analysis has many attractive features, a few questions can be raised. It assumes, as does the analysis of Nishigauchi (1990), that mo is interpreted as a universal quantifier and dare as a variable. Then, first, it implies that there are two mo’s; one is interpreted as ‘also’ in the absence of a wh-expression as in (11b) and the other is a universal quantifier that occurs only with a whexpression. Second, as wh-expressions are assumed to be variables, it is still unclear why they cannot be unselectively bound in examples like (13b). 2 That the main predicate in (4) is individual-level seems certainly relevant. See Diesing (1992) for much relevant discussion. 3 Tsai (1999) proposes that Chinese wh-expressions are unselectively bound and shows that the Chinese counterpart of (13b) is grammatical with the intended meaning. Takita & Yang (2010) systematically discuss the differences between Chinese and Japanese wh-questions, and provide an account based on the hypothesis that the former are interpreted by unselective binding whereas the latter involve covert phrasal wh-movement.

186

saito

It seems then worth pursuing an alternative to Kuroda’s hypothesis that wh-expressions in Japanese are interpreted as variables. And this brings us back to Nishigauchi’s (1990, 1991) statement that “the quantificational force of a wh-expression” is determined by a “quantificational particle.” The idea is schematically shown in (15). quantificational force (15) [wh-expression (Op)x [[ ..... x ..... ] particle]] movement (Op.-vbl. relation) A wh-expression is an operator without specific quantificational force. It then has to move covertly to a position where it can probe for a particle that provides the required quantificational force. When it finds mo, the wh-expression is interpreted as a universal quantifier. I outline an analysis along this line in the following section.

4

Valuation of Operator Feature with Covert Movement

I propose in this section that the mechanism of feature valuation, widely assumed for φ-features and Case features, can successfully accommodate the idea illustrated in (15). I first briefly go over the mechanism of feature valuation, as proposed in Bošković (2007), and then apply it to wh-expressions in Japanese. Chomsky (2000) proposes that φ-feature agreement takes place under probe-goal relation as in (16a). (16) a.

TP T [φ: α]

→ vP

DP [φ: α] [Case: _]

v’

b.

TP DP [φ: α] [Case: NM]

T’ T

vP

Subject-verb agreement is a phenomenon where the morphological form of t reflects the φ-feature values (for person, number, and gender) of the subject. Since t itself does not have values for these features, it searches its domain for a dp. It then enters into an agreement relation with the dp and acquires its φ-feature values. Bošković’s (2007) develops this analysis, and proposes that

187

japanese wh-phrases as unvalued operators

the Case feature is valued also in a probe-goal relation. A dp is not inherently equipped with a specific Case and must have its Case feature valued. The subject dp in (16b) moves to Spec, t and probes for a Case assigner. As it finds t in the domain, the Case is valued nominative. The system of feature valuation just illustrated allows the analysis of whexpressions in Japanese as unvalued operators that need to be specified for quantificational force. The universal quantification in (12), for example, is analyzed as in (17). (17) a.

FP NP

→ mo [conj.]

b.

FP darex [Op: conj.]

F’

….. dare ..… [Op: _]

NP

mo [conj.]

..… x ….. Dare ‘who’ moves to probe for a particle and mo values its operator feature as conjunctive. As widely assumed and often noted, it does not seem to be an accident that mo is associated with universal force. As shown in (18), this particle is employed for conjunction. (18) Hanako-mo Taroo-mo Ziroo-mo soko-ni ita Hanako-also Taroo-also Ziroo-also there was ‘Lit. Also Hanako, also Taroo, also Ziroo were there. = Hanako, Taroo and Ziroo were there.’ Universal quantification is equivalent to conjunction of all individuals in the domain of discourse. Thus, if the set of persons in the domain of discourse is {a, b, c, …}, then (19) holds. (19) Everyone was there = [a and b and c and …] were there I assumed in (17) that this conjunctive meaning of mo turns dare ‘who’ into a universal quantifier. This analysis accounts for the fact that a wh-expression requires a quantificational particle. An operator cannot be interpreted if its quantificational force is unspecified. Another advantage of the analysis is that there is no need to assume that mo, for example, is ambiguous between a universal quantifier and ‘also, even’. Thus, the same mo occurs in both (20a) and (20b).

188

saito

(20) a. [[Taroo-ga kite] mo] boku-wa pro au Taroo-nom come even I-top meet ‘Even if Taroo comes, I will meet him.’ b. [[Dare-ga kite] mo] boku-wa pro au who-nom come even I-top meet ‘For any x, x a person, even if x comes, I will meet x.’ It is interpreted as ‘even’ in both (20a) and (20b) as indicated in the translations. It is just that the mo in (20b) assumes an additional function, that is, it values the operator feature of dare and turns the wh-expression into a universal quantifier. The analysis not only provides a motivation for the covert phrasal movement of wh-expressions but also makes it possible to account for the wh-island effect in a simple way. The relevant example, (10), is repeated below as (21). (21) Taroo-wa [[dare-ga kuru ka] kiite] mo ikanai daroo Taroo-top who-nom come q hear even go-not will a. Taroo won’t go even if he hears who is coming. b. * For every x, x a person, Taroo won’t go even if he hears whether x is coming. Even if dare moves to probe for mo, it must first move to the edge of the most deeply embedded question cp, as cps constitute derivational phases (Chomsky 2000). Then, ka values its operator feature at this point, and dare becomes an interrogative operator. As dare receives a proper interpretation at this position as an interrogative operator, further movement is prohibited by Rizzi’s (2010) criterial freezing, which requires an operator to stay at a position where its operator feature can be interpreted. In addition, further movement of dare makes the sentence nonsensical because an interrogative operator can only be interpreted as part of a question sentence. Before concluding this section, I would like to mention one issue that remains to be settled. It is known that a wh-expression must be in the ccommand domain of the associated quantificational particle. (22) illustrates this. (22) * Hanako-wa dare-ni hon-mo okutta Hanako-nom who-dat book-also sent ‘Intended: For every x, x a person, Hanako sent also a book to x.’

japanese wh-phrases as unvalued operators

189

The unselective binding analysis and the particle movement analysis both predict this fact. In the former, the unselective binder mo must bind the whexpression dare. (22) cannot even be derived under the particle movement analysis, given that movement is always upward. But if dare probes for a quantificational particle in its domain, as proposed here, then the configuration in (22) may allow its operator feature to be valued by mo. There are two possible ways to account for examples like (22). Descriptively, a wh-expression must move to a specifier position of the associated particle. (22) is out because this requirement cannot be satisfied. The requirement may be imposed by the semantics. That is, it may be that a wh-expression, even after its operator feature is valued, is interpreted in conjunction with the associated particle. Alternatively, the requirement may be a condition on the valuation of an operator feature. I assumed that an operator feature is valued by a particle in probe-goal relation. But a stricter Spec-head relation may be required.4 I leave these two possibilities open.

5

Conclusion

In this paper, I suggested an analysis of wh-expressions in Japanese. Building on Takahashi’s (2002) insights, I proposed, along the lines hinted at in Nishigauchi (1990, 1991), that a wh-expression in Japanese is an operator whose quantificational force is determined by the associated particle. Despite its similarities with Nishigauchi (1990) and Takahashi’s (2002) analysis, the analysis in this paper makes a novel claim on the status of wh-expressions in Japanese. It questions Kuroda’s (1965) proposal that Japanese wh-expressions are “indeterminate pronouns” and asserts that they are operators. Another consequence of the analysis concerns feature valuation. As illustrated with Bošković’s (2007) analysis, the cases of feature valuation proposed in the literature apply to uninterpretable features. For example, φ-features on Tense and Case on dps are not subject to semantic interpretation and only have effects on pf. On the other hand, the valuation of operator feature proposed in this paper is required for semantic interpretation. Then, if the analysis is on the right track, the mechanism of feature valuation is employed more extensively than generally assumed.

4 It is possible to state the condition without explicitly referring to a Spec-head relation. For example, the condition can say that an operator feature is valued by the label of its sister. But this still refers to a Spec-head relation in disguise.

190

saito

References Bošković, Željko 2007. ‘On the locality and motivation of move and agree: An even more minimalist theory’. Linguistic Inquiry 38, pp. 589–644. Chomsky, Noam 2000. ‘Minimalist inquiries: The framework’. In: Roger Martin, David Michaels, and Juan Uriagereka (eds.), Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik. Cambridge, Mass.: mit Press, pp. 89–155. Diesing, Molly 1992. Indefinites. Cambridge, Mass.: mit Press. Hagstrom, Paul 1998. Decomposing Questions. mit Ph.D. dissertation. Heim, Irene 1982. The Semantics of Definite and Indefinite Noun Phrases. University of Massachusetts Ph.D. dissertation. Kuroda, S.-Y. 1965. Generative Grammatical Studies in the Japanese Language. mit Ph.D. dissertation. Lasnik, Howard and Mamoru Saito 1984. ‘On the nature of proper government’. Linguistic Inquiry 15, pp. 235–289. Maki, Hideki 1995. The Syntax of Particles. University of Connecticut Ph.D. dissertation. Nishigauchi, Taisuke 1990. Quantification in the Theory of Grammar. Dordrecht: Kluwer Academic Publishers. Nishigauchi, Taisuke 1991. ‘Construing Wh’. In: C.-T. James Huang and Robert May (eds.), Logical Structure and Linguistic Structure, Dordrecht: Kluwer Academic Publishers, pp. 197–231. Richards, Norvin 2000. Movement in Language. New York: Oxford University Press. Rizzi, Luigi 2010. ‘On some properties of criterial freezing’. In: E. Phoevos Panagiotidis (ed.), The Complementizer Phrase, Oxford: Oxford University Press, pp. 17–32. Shimoyama, Junko 2001. Wh-constructions in Japanese. University of Massachusetts Ph.D. dissertation. Takahashi, Daiko 2002. ‘Determiner raising and scope shift’. Linguistic Inquiry 33, pp. 575–615. Takita, Kensuke and Barry C.-Y. Yang 2014. ‘On multiple wh-questions with why in Japanese and Chinese’ In: Mamoru Saito (ed.), Japanese Syntax in Comparative Perspective, New York: Oxford University Press, pp. 206–227. Tsai, W.-T. Dylan 1999. ‘On lexical courtesy’. Journal of East Asian Linguistics 8, pp. 39– 73.

Index 3.prs (third person)

157

Abney, Steven 138, 157 abstract case 159–161, 163, 165–171, 173, 175 acc (see accusative) accent 8, 66–75, 83–87, 87n10 accent locus 68–74 accent system 8, 66, 68, 72–74 accentual class 86 accentual shift 85 accomplishment (verb class) 126 accusative (acc) 87, 149, 157, 159, 160, 160n1, 166n4, 169, 177 achievement (verb class) 125, 126, 127n5 acoustic feature 28 active 131–133 ad (see Argument Drop) affricate 27–30 affricative 26–28, 33, 34, 35n13 affricativization 33, 34 Ainu 56, 58, 59, 65, 82, 88, 89 Akimoto, Kichirō 83n4, 88 Akinaga, Kazue 84n6, 88 aktionsart 110 Amami 4, 11, 65 Andersen, Henning 23, 23n2, 33n12, 35 Anderson, Stephen 32, 35 Aoki, Kazuo 81n1, 88 Aoun, Joseph 96, 109 Argument Drop (ad) 93–96, 96n2, 97–99, 107, 108 Arisaka, Hideyo 49n9, 53 aspect 110, 111, 111n1, 112–114, 116–118, 121–123, 137, 140 aspectual 110, 114, 116, 117, 125, 126, 140n1 aspirated consonants 66 aspiration 26 AspP 94, 99, 100, 102, 113–116 Asuka shiryōkan (mr) 45n1, 47, 49, 50n12, 51, 54 attr (see attributive) attributive (attr) 28, 47, 85, 87, 160–167, 171–175 Bailyn, John 94, 97, 102, 109 Baker, Mark 172, 176

Batchelor, John 82, 88 Bentley, John 18, 21 Bielorussian 19 Bošković, Željko 138, 139, 142, 143, 145, 145n4, 146, 146n5, 147–156, 158, 186, 189, 190 Brāhmī 49 Buck, Carl Darling 59, 64 Bulgarian 19, 59 Cantonese 39–41 Carlson, Gregory 127, 134n11, 136 Chen, Chung-yu 111, 114, 122 Chen, Zhong 111, 114, 122 Chikamatsu Jōruri fubon (cmjf) 84, 88 Chinese xi, xiii, xviii, xix, xx, xxiv, xxvi–xviii, xxix, 18, 37–41, 41n2, 42, 42n3, 43, 44, 45, 49–54, 57, 58, 58n6, 67, 77, 110, 116, 117, 122, 123, 154, 185n3, 190 Chomsky, Noam x, 159, 167n4, 176, 186, 188, 190 Chukchee 173, 177 Classical Japanese 51n13, 177 cmjf (see Chikamatsu Jōruri fubon) Cole, Peter 103, 109 comparative method 15, 56, 67, 72 conc (see concessive) concessive (conc) 87 conclusive 78n13, 160–163, 165–171 continuant 22, 28–32, 32n9, 33, 33n10, 34, 35, 35n13 control xxix, 130, 131, 133, 134 Cranston, Edward 164, 176 Crisma, Paola 154, 158 Czech 59 dat (see dative) dative (dat) 41n2, 157 Davies, William 131, 136 Derksen, Rick 17, 21 Diesing, Molly 130, 133, 136, 185n2, 190 dim (see diminutive) diminutive (dim) 87 diphthong 19, 24, 40, 41 distinctive feature 22, 29, 30, 35

192 Djamouri, Redouane xviii, xxv, 122 donkey anaphora 182, 185 dp (determiner phrase) 94, 107, 138, 139, 143, 147, 148–152, 154, 155n10, 156–158, 186, 187, 189 Dubinsky, Stanley 131, 136 durative 110, 111 E-iri ikoku tabi-suzuri 53n16 Early Middle Chinese (emc) 39–42, 49, 50, 50n10, 51, 51n13, 52–54 Early Middle Japanese (emj) xxii, xxv, 22n1, 23, 24n5, 25–27, 30, 31, 33, 34, 62, 76, 77n4, 76n2, 78, 161 Eastern Old Japanese (eoj) 16, 17, 47, 54, 65, 87 ecm (see Exceptional Case Marking) Edo period 22n1, 76, 83–85, 85n7, 86, 87, 88 Ekida, Fusae 159, 163–166, 168, 177 ellipsis xxi, 93–97, 99–109 emc (see Early Middle Chinese) emj (see Early Middle Japanese) endangered language 3–5, 9, 11 English x, xx, xxiv, xxviii–xxx, 37, 88, 89, 95, 97, 100, 102, 103, 117, 126, 126n5, 127n6, 132, 136–141, 141n2, 142, 142n3, 143–146, 149–151, 156, 157, 159, 167, 168, 171, 179 eoj (see Eastern Old Japanese) Estonian 19 etymology 15, 17, 56, 60, 61, 76, 77, 81, 82, 85, 86 ev (see evidential) evidential (ev) 87 Exceptional Case Marking (ecm) 168–170 excl (see exclamative) exclamative (excl) 87 exp (see experiential) experiential (exp) 111n1, 112, 124, 125, 125n1, 125n2, 126, 126n3, 127, 127n6, 128–130, 130n9, 131–136

index 36, 40, 41, 43, 46, 53, 55–57, 60, 64, 65, 77n6, 79, 161, 178 French ix, 60 fricative 26–30, 33, 34 Fudoki 83n4, 88 Fujii, Takashi 124, 135n13, 136 Fukui, Rei 67, 75 fut.nom (future nominalization) 157 future 112, 157 Gajewski, Jon 146, 146n5, 147, 158 gen (see genitive) Gengo kokka (gk) 84, 88 genitive (gen) 78, 148, 149, 157, 167 German xxiii, xxvi, 41, 127n6, 133, 138, 141– 143, 145–150, 150n7, 156, 157, 159 Germanic 108, 152, 154, 158 gk (see Gengo kokka) glossing xi, xxv, 45, 85n8 go’on/go-on 37, 39, 40, 51–53, 77, 77n4 Goldberg, Lotus 93, 94, 102, 109 Gong, Hwang-cherng 39, 43 Greek xi, xxii, xxvi, 58, 159, 176 Gribanova, Vera 93–96, 96n2, 97, 97n3, 98– 100, 102, 103, 105–109 -guo 110, 111, 111n1, 112–117, 122

habitual 114n3, 124, 135, 135n13, 136 Hachijō 56, 60 Hagstrom, Paul 180, 190 Halle, Morris x, 28, 36 Hamkyeng 66, 67, 70–75 hapax legomenon 85, 86 Hashimoto, Shinkichi 57, 64, 78 Hashimoto, Takako 50, 53 Hattori, Shirō 57, 64 Hayashi, Yuka 12 Hebrew 94, 102 Heian (period) 17, 49n9, 50–52, 54, 81, 84, 85, 164, 173, 176 Heike mabushi (hkmb) 84, 88 Heim, Irene 180–182, 185, 190 hf (see Hitati Fudoku) Fact.nom (factive nominalization) 157 high register 58, 84 feature 22, 28, 29, 29n8, 30, 33, 35, 103, 140n1, Hirai, Meari 4n4, 12 160, 186–189, 189n4 Hirara 57, 64 Foshuo guanding jing 51 Hirayama, Teruo 48, 54 Frellesvig, Bjarke xvi, xvii, xx, xxii, xxvi, Hirohama, Fumio 176 xxix, 15, 18, 21, 23, 23n3, 23n4, 24n5, 35, historical phonology 22, 35, 43–45, 75

193

index Hitati Fudoku (hf) 83, 88 hkmb (see Heike mabushi) hoj (= Frellesvig 2010) 23n4, 24n5, 27, 27n7, 28, 30, 34, 36 Hoji, Hajime 96, 109 Hu, Wenze 118, 122 Huang, C.-T. James 93, 109, 116n4, 119, 123, 190 Hyman, Larry 71, 75 Iejima 57, 62, 63, 65 Ikeda, Kikan 161n2, 164, 176 Ikema 3, 3n1, 4, 4n3, 5, 8, 9, 12, 63 4 ilp (see individual level predicate) Inaoka, Kōji 81n1, 88 individual level predicate (ilp) 127–130, 133–135, 185n2 Indo-European xi, 17, 64 inf (see infinitive) infinitive (inf) 87 initial consonant clusters 66 Inoue, Kazuko 130n10, 136 Intokuki 53n16 inverse scope 138, 150–152 Irish 94 Ishigaki 59, 61–64 Itō, Chiyuki 67, 75 Iwasaki, Shoichi xxiii, 3n1, 4n3, 12 Izena 57, 59, 61–64 Jacobsen, Wesley x, xvii, 124, 136 Jakobson, Roman 28, 36 Japanese ix–xi, xiii, xvi–xxx, 4, 8, 9, 12, 15–18, 20–24, 24n5, 26–28, 30, 32, 33, 35–41, 41n1, 43–49, 49n9, 51, 51n13, 52, 53, 53n16, 54–57, 57n1, 58, 58n6, 59–65, 68, 71, 75, 76, 76n1, 77, 77n4, 78, 78n9, 79, 82, 83, 85–89, 94, 109, 124, 126, 127n6, 128, 130, 131, 135n13, 136, 137, 148, 149, 154, 159, 160n1, 161, 163, 167n5, 168, 169, 171, 172, 175–180, 182–185, 185n3, 186, 187, 189, 190 Japanese reading 17, 57 Japonic xxvii, 4, 15–21, 40, 41, 44, 46, 48, 54, 56, 57, 59–61, 61n10, 62, 62n12, 65, 88, 89 Jaxontov, Sergei 112, 113, 118, 121, 123 Jeju dialect 4, 11

Jōben-bon of Shūi waka shū (siwks) Johnson, Kyle 109 Joseph, Brian 159, 176

84, 88

Kageyama, Tarō 130n10, 136 Kagoshima accent 84, 85 kakarimusubi xxiii, xxvi, xxvii, xxx, 160–162 Kallestinova, Elena 102, 109 Kamakura 49n9, 84 Kamikatetsu 57n1, 59 kan’on/kan-on 37–39, 49, 50, 52n14, 77, 77n4 Kane, Daniel 39, 41, 41n2, 44 Kanto dialects 22n1 katakana 45, 51 Kazenin, Konstantin 99, 100, 102, 103, 109 Keskin, Cem 149, 158 Kikaijima dialect 11 Kiku, Chiyo 64 Kindaichi, Haruhiko 124–126, 126n5, 136, 137 Kinsui, Satoshi 163, 164, 166, 169, 172, 175, 176 Kitagawa, Hiroshi 176 Kitahara, Yasuo 19, 21 kkwks (see Kokin waka shū) Kobayashi, Yoshiharu 161, 176 Koguryo 41, 46 Kojiki 58, 59n7, 60, 64 Kojiki kayō 48n7, 50 Kokin waka sshū (kkwks) 46, 48, 84, 88 Kōno, Rokurō 49n9, 54 Konoshima, Masatoshi 167, 176 Korean ix–xi, xiii, xvi–xxx, 4, 11, 12, 15, 16, 20–22, 36–42, 42n3, 43, 44, 49, 49n9, 53n16, 55, 58n6, 59–62, 65, 66, 66n1, 67, 69–72, 74, 75, 77, 79, 89, 94, 137, 178 Korean accent 66, 71, 75 Korean dialect 71, 75 Koreanic 15, 16, 18, 20, 41 Koreo-Japonic 15, 19, 21, 40, 44, 60–62, 65 Kratzer, Angelika 127, 128, 130, 133, 134, 136 Krifka, Manfred 135n13, 136 kungana 17, 45 Kuno, Susumu x, xvi, xx, 129, 130n10, 136, 168, 176 Kurano, Kenji 64 Kuroda, S.-Y. 117, 123, 129, 136, 179, 180, 183, 186, 189, 190 Kuroita, Katsumi 83n5, 88

194 Kusumoto, Kiyomi 124, 137 Kyengsang 66, 68–75 Kyoto dialect 84, 85 Kyūshū dialect 57n5 Lange, Roland A. 79 Lasnik, Howard 179, 190 Late Middle Chinese (lmc) xiii, 38, 41–43, 49, 49n9, 50, 50n10, 52n14, 54 Late Middle Japanese (lmj) 22, 22n1, 24n5, 24n6, 25–28, 30, 32–34, 57 Latin ix, xviii, xxv, xxvii, 53n16, 59, 60, 159 -le 110, 111, 111n1, 112–119, 122 Lee, Ki-Moon 66n1, 67, 75 Lee, Yeonju 71 Li, Audrey 96, 109 Liao Chinese 39–41 Lightfoot, David 159, 176 lmc (see Late Middle Chinese) lmj (see Late Middle Japanese) loc (see locative) locative (loc) 41, 87, 110, 117–120, 122 locative inversion 110, 117–120, 122 Longobardi, Giuseppe 154, 158 low register 84–86 Lü, Shuxiang 116, 123 Mabuchi, Kazuo 176 Machida, Ken 124, 137 Macro-Yaeyaman 4n2 Maeara, Tōru 64 majority (superlative) reading (mr) 138, 145–148, 157 Maki, Hideki 180, 190 man’yōgana 40, 83 Man’yōshū (mys) 17, 45n1, 48n7, 50, 51n13, 59, 79, 83, 85, 88, 161–163, 165, 166, 169, 172, 173 Mandarin xi, xviii, xix, xxvi, xxviii, 39– 41, 110, 113n2, 116 Martin, Samuel E. x, 66n1, 67, 75, 78, 79, 82, 83, 86 Marušič, Franc 156, 158 Matsumoto, Yo 130n10, 137 Matsunaga, Setsuko 160, 162, 165, 166, 176, 177 Matsuo, Osamu 176, 177 Matsuyama, Jirō 83n5, 88

index McClure, William 124, 127n5, 137 MdJ (see Modern Japanese) mellow (feature) 29 Middle Chinese xiii, 37–42, 42n3, 43–45, 49, 54 Middle Japanese (mj) xxii, xxv, 22, 22n1, 23, 24, 27, 57, 61, 62, 76, 77n4, 84, 85, 87, 87n10, 161, 164 Middle Korean (mk) xxvii, xxix, 16–20, 42, 55, 59, 61, 61n11, 62, 63, 66, 66n1, 67, 69–72, 74, 77n8 Milsark, Gregory 127, 137 Min 39, 40 Miyagawa, Shigeru 159, 160, 160n1, 162–168, 173, 177 Miyagi, Shin’yū 64 Miyake, Marc 46, 49n9, 51–54 Miyako 3, 4, 8, 12, 64 Miyako Ryukyuan 3, 4 Mizutani, Shinjō 50, 54 mj (see Middle Japanese) mk (see Middle Korean) Modern Japanese (MdJ) 17, 22n1, 25–27, 27n7, 28, 30, 32, 32n9, 33n10, 34, 47, 48, 81, 82, 84, 87, 137, 159, 163, 168, 169, 172, 175, 177 modern Korean 42 modern Kyoto dialect 84 modern Sino-Japanese 41n1 modern Sino-Korean 37, 38, 42, 43 mokkan 45, 45n1, 46–48n, 48n5, 49–51, 53– 55 Mori, Hiromichi 50, 50n11, 54 Moriyama, Takurō 127n5, 137 morphological case 159–166, 169, 171–173, 175 Motohashi, Tatsushi 170, 177 mr (see Asuka shiryōkan) mr (see majority reading) Mt. Fuji xiii, 80–83, 85, 85n7, 86n9 Myōgoki 76 mys (see Man’yōshū) Nakamura, Yukihiko 64 Nakanishi, Susumu 64 Nakasone, Seizen 64 Nakijin 57, 59–64 Nara period 85 Ndendeule 94

195

index Neg/neg (negation) 111n1, 138–140, 140n1, 141, 141n2, 142–145, 157 neg-raising (nr) 138–140, 142, 142n3, 143, 145, 157 negative polarity item (npi) 138–140, 140n1, 141, 141n2, 142–145, 157 Nevskij, Nikolai A. 64 Nichūreki 49n9, 53n16 Nihonshoki (nsk) 19, 49, 50, 54, 57–59, 63, 83, 83n5, 88 Nishigauchi, Taisuke 130n10, 137, 179–185, 189, 190 nml (see nominalizer) nom (see nominative) nominalizer (nml) 87, 174 nominative (nom) 157, 187 North Korean 66 Northeastern Chinese 40, 41 Northern Mandarin 113n2 Northern Ryukyuan 57 Northwestern Chinese 49 noun phrase (np) xxiii, xxvii, 107, 117, 138, 139, 143, 146–149, 151–154, 156–158 np (see noun phrase) npi (see negative polarity item) nr (see neg-raising) nsk (see Nihonshoki) Numoto, Katsuaki 52n14, 54 obstruent 22, 24, 26–30, 32 Ōgami 57n1, 61–63 Ogihara, Toshiyuki 124, 127n5, 129, 137 Ōita 60 oj (see Old Japanese) Okami, Masao 64 Okazen 57n1, 61–63 Okinawa 4, 64, 65 Old Japanese (oj) xiii, xix–xxi, xxx, 15– 19, 20–24, 24n5, 25–28, 30–33, 33n10, 34, 40, 41, 45–47, 48n7, 48n8, 50–54, 56– 61, 61n11, 62, 63, 65, 76, 76n2, 77n8, 78, 78n10, 78n11, 79, 82–87, 87n10, 88, 89, 159, 160, 160n1, 161–163, 169, 170, 172, 173, 175, 178 Omodaka, Hisataka 17–19, 21 ongana 45 ongi mokkan 45, 45n1, 47, 51, 55 Ono, Susumu 177 Ono, Tsuyoshi 4n3, 12

onset consonants 22, 25, 26 operator 134, 153, 154, 182, 186–189, 189n4 Osada, Suma 65 Oshio, Mutsuko 65 Osterkamp, Sven 48n8, 51n13, 54 Otani, Kazuyo xxi, 94 Öztürk, Balkız 152–155, 158 pA (see proto-Ainu) Paris, Marie-Claude 114, 123 particle (pt) 19, 47, 68, 69, 73, 87, 116, 129, 161, 162, 166, 167, 172, 173, 180, 181, 183–189 passive 131–133 Paul, Waltraud 116, 123 Pellard, Thomas 4n2, 12, 16, 20, 21, 46–48, 54, 56, 57n2, 57n3, 65, 86, 89 perf (see perfective) perfective (perf) 111n1, 112, 117, 141, 165 phonemic 24, 24n5, 24n6, 25, 26, 28–33, 33n10, 33n12, 34, 35n13, 52, 70, 74, 77, 79 phonetic 25–30, 33, 34, 39, 49n9, 66, 68, 77, 87 phonogram 45, 49–51, 51n13, 52n15 phonographic 18, 58, 77, 82, 83 phonological ix, xi, xvi, xvii, 22, 24n5, 28, 30, 33n12, 35, 51, 52n15, 59, 67, 73, 74, 82, 86 Pierson, Jan L. 177 pitch 66–69, 71, 73, 87 pj (see proto-Japonic) pJ (see proto-Japonic/proto-Japanese) pJK (see proto-Japanese-Korean) pJN (see proto-Japanese) pKJ (see proto-Koreo-Japonic) pl (see plural) plural (pl) 111, 157 plurality reading (pr) 146, 147, 157 Poser, William 177 pr (see plurality reading) pr (see proto-Ryukyuan) pre-accent 69, 70, 73, 74 pre-oj (see pre-Old Japanese) pre-Old Japanese (pre-oj) 15, 19, 24, 57, 60 prenasalization 26 prenasalized 24, 52 pres (see present) pres.prog (see present progressive) 157 present (pres) 135n13, 140, 142, 146, 157 present progressive (pres.prog) 140, 157

196 pro-drop 152 progr (see progressive) progressive (progr) 110, 111, 111n1, 113, 114, 124–126, 126n5, 127, 130–133, 135n14, 140, 140n1, 157 proto-Ainu (pA) 82, 88 proto-Japanese (pJN/pJ) xvi, xvii, xx, xxviii, xxix, 15, 23, 83, 88 proto-Japanese-Korean (pJK) 19, 20 proto-Japonic (pj/pJ) 4, 15–17, 19, 23, 24, 40, 46, 48, 57–59, 61–63, 78n11, 82–84, 86, 86n9, 88 proto-Korean xx, xxi, xxvi, xxviii, 66 proto-Koreo-Japonic (pKJ) 19, 40, 77n8 proto-Ryukyuan (pr) 48, 56n1, 57–59, 62, 63 proto-Slavic 17 pt (see particle) Pulleyblank, Edwin 40, 44, 50n10, 54

index

Sato, Kiyoji 177 Satō, Eisaku 84n6, 88 Sawada 57 scrambling ix, xxx, 149, 150, 152 scs (see Shūchūshō) segmental 17, 23, 25, 26, 82, 84 Seidensticker, Edward 164, 178 senmyō-gaki 45 Serafim, Leon 40, 44 sfib (see Suruga Fudoki Ibun) sg (see singular) Shibatani, Masayoshi 130n10, 137, 177 Shimoji, Michinori 4n2, 12 Shimoyama, Junko 180, 190 Shin’yaku Kegonkyō ongi shiki 52, 53 Shinsen jikyō 57, 64 Shirafuji Noriyuki 81n1, 88 Shirai, Yasuhiro 124, 135n13, 137 Shodon 57n1, 61–63 Shoku Nihongi (sng) 81, 81n1, 83, 88, 167n5 quantity sensitivity 23 Shūchūshō (scs) 84, 88 Shuri 57, 59, 61–63 radical pro-drop 152 Siamese 155n9 raising xxi, 24, 41, 46, 46n2, 47, 48, 48n8, 57, sibilants 29, 30, 34 78n11, 93, 94, 99, 102, 105, 130, 131, 133, Sigurðsson, Hoskildur 96, 107–109, 152, 154, 138–140, 143, 145, 157 158 Ramsey, S. Robert 66, 66n1, 67, 69, 75 Silla dialect 41, 42 ranking 22, 30, 33, 33n12, 34 singular (sg) 111, 157 resultative 124, 125, 131, 133 Sino-Japanese (sj) xiii, 37–39, 41n1, 49–53, Richards, Norvin 179, 190 77, 77n4, 82, 83 rising tone 69 Sino-Korean (sk) xiii, 37–42, 42n3, 43, 49, Rizzi, Luigi 93, 109, 188, 190 49n9, 50, 51, 77 Romance 159 Sino-Vietnamese xiii, 38–40 Ruiju myōgishō 60, 64 sinograph 38, 39, 42, 43 Russian ix, x, 17, 19, 93–96, 96n2, 97, 99, 100, siwks (see Jōben-bon of Shūi waka shū) 102–104 sj (see Sino-Japanese) Ryukyuan xvi, xvii, xxviii, 3, 4, 4n2, 16–18, sk (see Sino-Korean) 20, 48, 56, 56n1, 57, 57n5, 59, 59n7, 60, Slavic 17, 19 61, 61n11, 62, 62n12, 63, 83, 86 slp (see stage level predicate) sloppy (identity reading) 99, 100, 103 Saito, Mamoru 96, 103, 109, 179, 190 Slovenian 145, 146, 156 Sakakura, Atsuyoshi 64, 81n2, 89 sng (see Shoku Nihongi) Sakamoto, Kiyoe 84n6, 88 Soga, Matsuo 124, 127, 137 Sakhalin dialect 82n3 sound change 20, 23, 24n5, 25, 32, 52, 56, sakîmôri poem 85 73 Sanskrit 40, 49 Southern Min 40 Sansom, George 161, 166, 167, 167n5, 177 Southern Ryukyuan 57, 61, 63 Saru dialect 58, 65 Soya dialect 82n3 Sasayama, Haruo 81n1, 88 Spanish 152–154

index Spencer, Andrew 173, 177 stage level predicate (slp) 130, 133, 184 Standard Japanese 4, 8, 9 stops 27–30, 32n9 Stowell, Timothy 168, 177 strict (identity reading) 99, 103, 104 strident 22, 28–33, 33n10, 34, 35n13 Sugita, Mamori 130, 137 sub (see subordinate/subordinator) subordinate/subordinator (sub) 87, 100, 111n11, 160, 161, 163, 168, 169 Sung, Li-May 103, 109 suprasegmental 68–70, 73, 74 Suruga Fudoki Ibun (sfib) 83, 88 Suyama, Nahoko 65 Suzuki, Tomi 177 Suzuki, Yutaka 84n6, 88 Svenonius, Peter 97, 109 Swahili 94 syllable post-peak phonemes 23 Taiwanese 40 Taiwanese Mandarin 113n2 Takagi, Ichinosuke 162, 177 Takahashi, Daiko 180, 181, 184, 185, 189, 190 Takahashi, Tarō 135n13, 137 Takahashi, Toshizō 64 Takeda, Yūkichi 64 Taketomi 59, 62–64 Taketori monogatari (tm) 81, 81n2, 88 Takita, Kensuke 185n3, 190 Takubo, Yukinori 3n1, 12, 128n7, 137 Tamura, Suzuko 65 Tangut 39 Tatar 19 -te iru 124, 125, 125n1, 125n2, 126, 126n5, 127– 135, 135n13, 136 Teng, Shou-hsin 120, 123 Teramura, Hideo 124, 135n13, 137 Thai xi, 155n9 Thorpe, Maner 48, 54 Tibetan xxv, 49 tm (see Taketori monogatari) Tokyo accent 83, 86 Tokyo Japanese 68 Tongkwuk cengwun 37–40, 42, 43 top (see topic) topic (top) xvii, xxiii, xxviii, 19, 87, 93, 96n2, 125, 152–p

197 Tottori 60 Tranter, Nicolas 51, 52, 54 Travis, Lisa deMena 110, 123 Tsai, W.-T. Dylan 185n3, 190 Tsukishima, Hiroshi 52, 54 Turkic 19 Turkish 16, 19, 138–141, 143–148, 148n6, 149, 151–155, 155n9, 155n10, 156, 157 Ueno, Kazuaki 84n6, 88 Uighur 49 Umeda, Hiroyuki 71, 72, 75 Unger, J. Marshall x, 51, 54, 77n8, 78n10, 79, 86 unitary vowels 24 unselective binding 180, 181, 183, 184, 185n3, 189 uta mokkan 45, 46 Uwano, Zendō 71, 72, 75 variable 103, 133, 134, 152, 153, 179, 180, 182– 186 Vendler, Zeno 125, 126, 126n5, 137 verb-stranding 93, 104, 108 vocal cord vibration 28 voicing 24, 26, 45 Vovin, Alexander 15–19, 21, 46, 49n9, 55, 56, 60, 61, 61n11, 62, 63, 65, 82, 85n8, 86, 89, 155n9 vsc (v-stranding constructions) 93, 94 vvpe (vp-ellipsis after v+v raising) 93–95, 97–99, 102, 104, 105, 108 Waji taikan shō (wts) 84, 88 Wamyō ruijū-shō 47, 48, 50, 53, 59, 64 Wamyōshō 76–78 Welsch, Bernhard 76, 79 Western Old Japanese (woj) 46, 47, 83, 85– 88 wh-phrase 179, 180 Whitman, John ix, xi, xvi, 15, 16, 18–23, 23n3, 24n5, 36, 40, 41, 43, 46, 46n2, 53, 55– 61, 61n9, 61n11, 62–65, 67, 75, 77n8, 79, 86n9, 94, 109, 163, 167, 170–173, 175, 178 woj (see Western Old Japanese) Wrona, Janick 161, 178 wts (see Waji taikan shō) Wu 39

198 Yaeyama 4, 4n2, 64 Yale system/romanization 37, 66n1 Yamada, Masahiro 178 Yamatohama 59–61, 61n11, 62, 63 Yanagida, Yūko 60, 65, 163, 167, 170–173, 175, 178 Yanbian 71, 72, 75 Yang, Barry C.-Y. 185n3, 190 Yangjuemoluo jing 52 Yonaguni 4, 4n2, 57n1, 61–63

index Yoron 57n1, 59–64 Yoshikawa, Taketori 124, 135n13, 137 Yuwan 57n1, 62, 63 zài

110, 111, 111n1, 113, 114, 114n3, 116, 117, 120, 122 Žaucer, Rok 156, 158 -zhe 110–113, 113n2, 114, 114n3, 115–122 Zhu, Dexi 116, 123 Živanović, Sašo 145n4, 147, 158