Table of contents : Preface Introduction to the Handbooks of Japanese Language and Linguistics Contents Contributors Introduction Part I: Prehistory and reconstruction 1 Reconstruction of Japonic and para-Japonic based on external sources 2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan 3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan Part II: Phonology 4 Reconstruction of Old Japanese phonology 5 Old Japanese writing and phonology 6 Late Middle Japanese phonology, based on Korean sources 7 Late Middle Japanese phonology as reflected in early Japanese Christian documents 8 Sino-Japanese 9 The Ramsey hypothesis Part III: Grammar 10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax 11 The syntax and morphology of Early Middle Japanese 12 Late Middle Japanese grammar 13 The historical changes in the case marking system of Japanese 14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate 15 Loss of wh-movement 16 Development of adverbial particles 17 The history of demonstratives 18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’ Part IV: Lexicon, Materials and Kanbun 19 The history of basic vocabulary in Japanese 20 The Japanese lexicon as reflected in Christian materials 21 What mokkan (wooden documents) can tell us about ancient Japanese language 22 Early Japanese dictionaries 23 Kunten texts of Buddhist provenance (butten 仏典): Their characteristics and actuality 24 Kunten texts of secular Chinese origin (kanseki 漢籍) 25 Japanized written Chinese: Its features and contribution to the history of the Japanese language 26 Early modern kanbun and kanbun-kundoku 27 The influence of kanbun-kundoku vocabulary on the Japanese language Index
It was with great sadness that we received news of the passing of three dear friends and colleagues in the course of the last stages of the editing of this book. We dedicate this book to their memory. Alexander (Sasha) Vovin passed away far too young, at the age of 61, in April 2022. Sasha was a towering and inspiring figure within historical Japanese linguistics, as well as many other fields within East Asian linguistics. His contribution to this volume is its first chapter, in the section on Prehistory and reconstruction. Those who are familiar with Sasha and his work will know that he published important and influential papers and books of lasting value on subjects spanning all four parts of this book. Charles J. Quinn passed away in July 2023 at the age of 75. He worked widely on the semantics, syntax and morphology of Old and Early Middle Japanese and was also a pioneer in the teaching of Classical Japanese. His chapter in this volume in the section on Grammar is a characteristically insightful study on kakari-musubi which will be a lasting contribution to a fuller understanding of this phenomenon. Elisabeth M. de Boer was taken away by aggressive illness at the young age of 57 in August 2023. Since her doctoral dissertation from Leiden University in 2005, Elisabeth worked tirelessly on substantiating, promoting and particularly developing and fleshing out the alternative hypothesis of the reconstruction and evolution of “accent” in Japanese originally proposed by S. Robert Ramsey in the 1970s and 1980s, and she did far more than anyone to move this hypothesis into the mainstream. Her contribution to this volume in the section on Phonology achieves this and it should from now on rightly be referred to as the “Ramsey/de Boer hypothesis”.
In memoriam Alexander Vovin Charles J. Quinn Elisabeth M. de Boer
Preface The project of compiling a series of comprehensive handbooks covering major fields of Japanese linguistics started in 2011, when Masayoshi Shibatani received a commission to edit such volumes as series editor from De Gruyter Mouton. As the planning progressed, with the volume titles selected and the volume editors assigned, the enormity of the task demanded the addition of a series co-editor. Taro Kageyama, Director-General of the National Institute for Japanese Language and Linguistics, was invited to join the project as a series co-editor. His participation in the project opened the way to make it a joint venture between NINJAL and De Gruyter Mouton. We are pleased to present the Handbooks of Japanese Language and Linguistics (HJLL) as the first materialization of the agreement of academic cooperation concluded between NINJAL and De Gruyter Mouton. The HJLL Series is composed of twelve volumes, primarily focusing on Japanese but including volumes on the Ryukyuan and Ainu languages, which are also spoken in Japan, as well as some chapters on Japanese Sign Language in the applied linguistics volume. – Handbook of Historical Japanese Linguistics – Handbook of Japanese Phonetics and Phonology – Handbook of Japanese Lexicon and Word Formation – Handbook of Japanese Syntax – Handbook of Japanese Semantics and Pragmatics – Handbook of Japanese Contrastive Linguistics – Handbook of Japanese Dialects – Handbook of Japanese Sociolinguistics – Handbook of Japanese Psycholinguistics – Handbook of Japanese Applied Linguistics – Handbook of the Ryukyuan Languages – Handbook of the Ainu Language Surpassing all currently available reference works on Japanese in both scope and depth, the HJLL series provides a comprehensive survey of nearly the entire field of Japanese linguistics. Each volume includes a balanced selection of articles contributed by established linguists from Japan as well as from outside Japan and is critically edited by volume editors who are leading researchers in their individual fields. Each article reviews milestone achievements in the field, provides an overview of the state of the art, and points to future directions of research. The twelve titles are thus expected individually and collectively to contribute not only to the enhancement of studies on Japanese on the global level but also to the opening up of new perspectives for general linguistic research from both empirical and theoretical standpoints. The HJLL project has been made possible by the active and substantial participation of numerous people including the volume editors and authors of individual chaphttps://doi.org/10.1515/9781614512851-202
X
Preface
ters. We would like to acknowledge with gratitude the generous support, both financial and logistic, given to this project by NINJAL. We are also grateful to John Haig (retired professor of Japanese linguistics, the University of Hawai‘i at Mānoa), serving as copy-editor for the series. In the future, more publications are expected to ensue from the NINJAL-Mouton academic cooperation. Masayoshi Shibatani, Deedee McMurtry Professor of Humanities and Professor of Linguistics, Rice University/Professor Emeritus, Kobe University Taro Kageyama, Director-General, National Institute for Japanese Language and Linguistics (NINJAL)
Masayoshi Shibatani and Taro Kageyama
Introduction to the Handbooks of Japanese Language and Linguistics Comprising twelve substantial volumes, the Handbooks of Japanese Language and Linguistics (HJLL) series provides a comprehensive survey of practically all the major research areas of Japanese linguistics on an unprecedented scale, together with surveys of the endangered languages spoken in Japan, Ryukyuan and Ainu. What follows are introductions to the individual handbooks, to the general conventions adopted in this series, and an overview of the minimum essentials of contemporary Standard Japanese. Fuller descriptions of the languages of Japan, Japanese grammar, and the history of the Japanese language are available in such general references as Martin (1975), Shibatani (1990), and Frellesvig (2010).
1 Geography, population, and languages of Japan Japan is situated in the most populous region of the world – Asia, where roughly one half of the world population of seven billion speak a variety of languages, many of which rank in the top tier among languages of the world in terms of number of native speakers. Japanese is spoken by more than 128 million people (as of 2013), who live mostly in Japan but also in Japanese emigrant communities around the world, most notably Hawaii, Brazil, and Peru. In terms of the number of native speakers, Japanese ranks ninth among the world’s languages. Due partly to its rich and long literary history, Japanese is one of the most intensely studied languages in the world and has received scrutiny both within the domestic grammatical tradition and in traditions outside Japan such as the Chinese philological tradition, European structural linguistics, and the tradition of generative grammar originating in America. The Handbooks of Japanese Language and Linguistics intend to capture the achievements garnered over the years through analyses of a wide variety of phenomena in a variety of theoretical frameworks. As seen in Map 1, where Japan is shown graphically superimposed on Continental Europe, the Japanese archipelago has a vast latitudinal extension of approximately 3,000 kilometers ranging from the northernmost island, roughly corresponding in latitude to Stockholm, Sweden, to the southernmost island, roughly corresponding in latitude to Sevilla, Spain.
https://doi.org/10.1515/9781614512851-203
XII
Masayoshi Shibatani and Taro Kageyama
Map 1: Japan as overlaid on Europe. Source: Shinji Sanada. 2007. Hōgen wa kimochi o tsutaeru [Dialects convey your heart]. Tokyo: Iwanami. p. 68
Contrary to popular assumption, Japanese is not the only language native to Japan. The northernmost and southernmost areas of the Japanese archipelago are inhabited by people whose native languages are arguably distinct from Japanese. The southernmost sea area of Okinawa Prefecture is dotted with numerous small islands where Ryukyuan languages are spoken. Until recent years, Japanese scholars tended to treat Ryukyuan language groups as dialects of Japanese based on fairly transparent correspondences in sounds and grammatical categories between those language groups and mainland Japanese, although the two are mutually unintelligible. Another reason that Ryukyuan languages have been treated as Japanese dialects is that the Ryukyuan islands and Japan form a single nation. In terms of nationhood, however, Ryukyu was an independent kingdom until the beginning of the seventeenth century, when it was forcibly annexed to the feudal domain of Satsuma in southern Kyushu. A more recent trend is to treat Ryukyuan as forming a branch of its own with the status of a sister language to Japanese, following earlier proposals by Chamberlain (1895) and Miller (1971). Many scholars specializing in Ryukyuan today even confer language status to different language groups within Ryukyuan, such as the Amami language, Okinawan language, Miyako language, etc., which are grammatically distinct to the extent of making them mutually unintelligible. The prevailing view now has Japanese and Ryukyuan forming the Japonic family as daughter languages of Proto-Japonic. HJLL
Introduction to the Handbooks of Japanese Language and Linguistics
XIII
follows this recent trend of recognizing Ryukyuan as a sister language to Japanese and devotes one full volume to it. The Handbook of the Ryukyuan Languages provides the most up-to-date information pertaining to Ryukyuan language structures and use, and the ways in which these languages relate to Ryukyuan society and history. Like all the other handbooks in the series, each chapter delineates the boundaries and research history of the field it addresses, presents the most important and representative information on the state of research in that field, and spells out future research desiderata. This volume also includes a comprehensive bibliography of Ryukyuan linguistics. The situation with Ainu, another language indigenous to Japan, is much less clear in terms of its genealogical relationship to Japanese. Various suggestions have been made relating Ainu to Paleo-Asiatic, Ural-Altaic, and Malayo-Polynesian or to such individual languages as Gilyak and Eskimo, besides the obvious candidate of Japanese as a sister language. The general consensus, however, points to the view that Ainu is related to Japanese only indirectly, if at all, via the Altaic family with its Japanese-Korean sub-branch (see Miller 1971; Shibatani 1990: 5–7 for an overview). Because Ainu has had northern Japan as its homeland and because HJLL is also concerned with various aspects of Japanese linguistics scholarship in general, we have decided to include a volume devoted to Ainu in this series. The Handbook of the Ainu Language outlines the history and current state of the Ainu language, offers a comprehensive survey of Ainu linguistics, describes major Ainu dialects in Hokkaido and Sakhalin, and devotes a full section to studies dealing with typological characteristics of the Ainu language such as polysynthesis and incorporation, person marking, plural verb forms, and aspect and evidentials.
2 History Japan’s rich and long literary history dates back to the early seventh century, when the Japanese learned to use Chinese characters in writing Japanese. Because of the availability of abundant philological materials, the history of the Japanese language has been one of the most intensely pursued fields in Japanese linguistics. While several different divisions of Japanese language history have been proposed, Frellesvig (2010) proposes the following four linguistic periods, each embracing the main political epochs in Japanese history. 1. 2. 3.
Old Japanese Early Middle Japanese Late Middle Japanese
This division reflects a major boundary between Pre-modern and Modern Japanese brought about by some radical changes in linguistic structure during the Late Middle Japanese period. Modern Japanese is often further subdivided into Early Modern (Edo, 1603–1868), Modern (Meiji, 1868–1912; Taishô, 1912–1926), and Present-day Japanese (Shôwa, 1926–1989; Heisei, 1989–2019; Reiwa 2019–). The Handbook of Historical Japanese Linguistics will present the latest research on better studied topics, such as segmental phonology, accent, morphology, and certain salient syntactic phenomena such as focus constructions. It will also introduce areas of study that have traditionally been underrepresented, ranging from syntax and Sino-Japanese (kanbun) materials to historical pragmatics, and demonstrate how these contribute to a fuller understanding of the overall history of Japanese, as well as outlining larger-scale tendencies and directions of change that have taken place within the language over its attested history. Major issues in the reconstruction of prehistoric Japanese and in the individual historical periods from Old Japanese to Modern Japanese are discussed, including writing and the materials available for historical study, influences of Sino-Japanese on Japanese, the histories of different vocabulary strata, the history of honorifics and polite language, generative diachronic syntax, and the development of case marking.
3 Geographic and social variations Because of the wide geographical spread of the Japanese archipelago from north to south, characterized by high mountain ranges, deep valleys, and wide rivers as well as numerous islands, Japanese has developed a multitude of dialects, many of which differ from each other in a way more or less like current descendants of the Romance language family. Like historical studies, the research tradition of dialect studies has a unique place in Japanese linguistics and has attracted a large number of students and amateur collectors of dialect forms as well as professional linguists. The Handbook of Japanese Dialects surveys the historical backdrop to theoretical frameworks of contemporary studies in Japanese geolinguistics and includes analyses of prominent research topics in cross-dialectal perspective, such as accentual systems, honorifics, verbs of giving, and nominalizations. The volume also devotes major attention to sketching the grammars of dialects from the northern island of Hokkaido to the southern island of Kyushu, allowing a panoramic view of differences and similarities among representative dialects throughout Japan. Besides having a physical setting that has fostered geographic variation, the society of Japan has exhibited differing types of social structure over the years, starting from the time of the nobility and court life of the Old and Early Middle Japanese periods, through the caste structure of the feudalistic Late Middle and Early Modern Japanese periods, to the modern democratic society of the Modern and Present-day Japanese periods. These
Introduction to the Handbooks of Japanese Language and Linguistics
XV
different social structures have spawned a variety of social dialects, including powerand gender-based varieties of Japanese. The Handbook of Japanese Sociolinguistics examines a wide array of sociolinguistic topics ranging from the history of Japanese sociolinguistics, including foreign influences and internal innovations, to the central topics of variation due to social stratification, gender differences, and discourse genre. Specific topics include honorifics and women’s speech, critical discourse analysis, the pragmatics of political discourse, contact-induced change, emerging new dialects, Japanese language varieties outside Japan, and language policy.
4 Lexicon and phonology The literary history of Japan began with early contacts with China. Chinese apparently began to enrich the Japanese lexicon even in pre-historic periods, when such deeply assimilated words as uma ‘horse’ and ume ‘plum’ are believed to have entered the language. Starting in the middle of the sixth century, when Buddhism reached Japan, Chinese, at different periods and from different dialect regions, has continuously contributed to Japanese in an immeasurable way affecting all aspects of grammar, but most notably the lexicon and the phonological structure, which have sustained further and continuous influences from European languages from the late Edo period on. Through these foreign contacts, Japanese has developed a complex vocabulary system that is composed of four lexical strata, each with unique lexical, phonological, and grammatical properties: native Japanese, mimetic, Sino-Japanese, and foreign (especially English). The Handbook of Japanese Lexicon and Word Formation presents a comprehensive survey of the Japanese lexicon, word formation processes, and other lexical characteristics seen in the four lexical strata of contemporary Japanese. The agglutinative character of the language, coupled with its intricate system of vocabulary strata, makes it possible for compounding, derivation, conversion, and inflection to be closely intertwined with syntactic structure, giving rise to theoretically intriguing interactions between word formation processes and syntax that are not easily found in inflectional, isolate, or polysynthetic types of languages. Theoretically oriented studies associated with these topics are complemented by ones oriented toward lexical semantics, which also bring to light theoretically challenging issues involving the morphology-syntax interface. The four lexical strata characterizing the Japanese lexicon are also relevant to Japanese phonology, as each stratum has some characteristic sounds and sound combinations not seen in the other strata. The Handbook of Japanese Phonetics and Phonology describes and analyzes the basic phonetic and phonological structures of modern Japanese with a main focus on standard Tokyo Japanese, relegating the topics of dialect phonetics and phonology to the Handbook of Japanese Dialects. It includes
XVI
Masayoshi Shibatani and Taro Kageyama
several chapters dealing with phonological processes unique to the Sino-Japanese and foreign strata as well as to the mimetic stratum. Other topics include word tone/accent, mora-timing, sequential voicing (rendaku), consonant geminates, vowel devoicing and diphthongs, and the appearance of new consonant phonemes. Also discussed are phonetic and phonological processes within and beyond the word such as rhythm, intonation, and the syntax-phonology interface, as well as issues bearing on other subfields of linguistics such as historical and corpus linguistics and research on the L2 acquisition of Japanese phonology.
5 Syntax and semantics Chinese loans have also affected Japanese syntax, though it is unclear to what extent they have affected Japanese semantics beyond the level of lexical semantics. In particular, Chinese loans form two distinct lexical categories in Japanese – verbal nouns, forming a subcategory of the noun class, and adjectival nouns (keiyō dōshi), which are recognized by some as forming major independent lexical categories along with noun, verb, and adjective classes. The former denote verbal actions and, unlike regular nouns denoting objects and thing-like entities, can function as verbs by combining with the light verb suru, which is obviously related to the verb suru ‘do’. The nominal-verbal Janus character of verbal nouns results in two widely observed syntactic patterns that are virtually synonymous in meaning; e. g., benkyoo-suru (studying-DO) ‘to study’ and benkyoo o suru (studying ACC do) ‘do studying’. As described in the Handbook of Japanese Lexicon and Word Formation, the lexical category of adjectival noun has been a perennial problem in the analysis of Japanese parts of speech. Property-concept words that fall into this class, such as kirei ‘pretty’ and kenkoo ‘health/healthy’, do not inflect by themselves, unlike native Japanese adjectives, and, like nouns, require the inflecting copula da to perform the predication function, hence the label of adjectival noun for this class. However, many of these cannot head noun phrases – the hallmark of the nominal class – and some even yield nouns via -sa nominalization, which is not possible with regular nouns. The Handbook of Japanese Lexicon and Word Formation and the Handbook of Japanese Syntax make up twin volumes because many chapters in the former deal with syntactic phenomena, as the brief discussion above on the two Sino-Japanese lexical categories clearly indicates. The syntax handbook covers a vast landscape of Japanese syntax from three theoretical perspectives: (1) traditional Japanese grammar, known as kokugogaku (lit. national-language study), (2) the functional approach, and (3) the generative grammar framework. Broad issues analyzed include sentence types and their interactions with grammatical verbal categories, grammatical relations (topic, subject, etc.), transitivity, nominalizations, grammaticalization, voice (passives and causatives), word order (subject, scrambling, numeral quantifiers, configurationality),
Introduction to the Handbooks of Japanese Language and Linguistics
XVII
case marking (ga/no conversion, morphology and syntax), modification (adjectives, relative clause), and structure and interpretation (modality, negation, prosody, ellipsis). These topics have been pursued vigorously over many years under different theoretical persuasions and have played important roles in the development of general linguistic theory. For example, the long and sustained study of the grammatical relations of subject and topic in Japanese has had a significant impact on the study of grammatical relations in European as well as Austronesian languages. In the study of word order, the analysis of Japanese numeral quantifiers has been used as one of the leading pieces of evidence for the existence of a movement rule in human language. With regard to case marking, the way subjects are case marked in Japanese has played a central role in the study of case marking in the Altaic language family. Recent studies of nominalizations have been central to the analysis of their modification and referential functions in a wide variety of languages from around the globe, with far-reaching implications for past studies of such phenomena as parts of speech, (numeral) classifiers, and relative clauses. And the study of how Japanese prosody plays a crucial role in interpretation has become the basis for some important recent developments in the study of wh-questions. The Handbook of Japanese Semantics and Pragmatics presents a collection of studies on linguistic meaning in Japanese, either as conventionally encoded in linguistic form (the field of semantics) or as generated by the interaction of form with context (the field of pragmatics). The studies are organized around a model that has long currency in traditional Japanese grammar, whereby the linguistic clause consists of a multiply nested structure centered in a propositional core of objective meaning around which forms are deployed that express progressively more subjective meaning as one moves away from the core toward the periphery of the clause. Following this model, the topics treated in this volume range from aspects of meaning associated with the propositional core, including elements of meaning structured in lexical units (lexical semantics), all the way to aspects of meaning that are highly subjective, being most grounded in the context of the speaker. In between these two poles of the semantics-pragmatics continuum are elements of meaning that are defined at the level of propositions as a whole or between different propositions (propositional logic) and forms that situate propositions in time as events and those situating events in various modes of reality including non-actual worlds, e. g., those hoped for (desiderative meaning), denied (negation), hypothesized (conditional meaning), or viewed as ethically or epistemologically possible or necessary (epistemic and deontic modality). Located yet closer to the periphery of the Japanese clause are a rich array of devices for marking propositions according to the degree to which the speaker is committed to their veracity and for marking differing perceptual and cognitive modalities as well as for distinguishing information that is presupposed versus affirmed. These studies in Japanese syntax and semantics are augmented by cross-linguistic studies that examine various topics in these fields from the perspectives of language universals and the comparative study of Japanese and other languages. The Handbook
XVIII
Masayoshi Shibatani and Taro Kageyama
of Japanese Contrastive Linguistics sets as its primary goal uncovering principled similarities and differences between Japanese and other languages around the globe and thereby shedding new light on the universal and language-particular properties of Japanese. Topics ranging from inalienable possession to numeral classifiers, from spatial deixis to motion typology, from nominalization to subordination, and other topics closely related to these are taken up within the framework of typological universals. Additionally, various aspects of Japanese such as resultative-progressive polysemy, entailment of event realization, internal-state predicates, topic constructions, and interrogative pronouns, are compared and contrasted with other specific languages, including Ainu, Koryak, Chinese, Korean, Newar, Thai, Burmese, Tagalog, Kapampangan, Lamaholot, Romanian, French, Spanish, German, English, Swahili, Sidaama, and Mayan languages.
6 Psycholinguistics and applied linguistics HJLL includes two volumes containing topics related to a wider application of Japanese linguistics and to those endeavors seeking grammar-external evidence for the psycho-neurological reality of the structure and organization of grammar. Incorporating recent research on the study of the cognitive processes and brain mechanisms underlying language use, language acquisition, and language disorders, the Handbook of Japanese Psycholinguistics presents the current state of scholarly understanding of the mechanisms of language acquisition and language processing. In particular, the volume seeks answers to the question of how Japanese is learned/acquired as a first or second language, and pursues the question of how Japanese sentences are comprehended and produced. The chapters in the acquisition section allow readers to acquaint themselves with issues pertaining to the question of how grammatical features (including pragmatic and discourse features) are acquired and how the language domain of the brain develops, with respect to both language particular and universal features. Specific topics dealt with include Japanese children’s perceptual development, the conceptual and grammatical development of nouns, Japanese Specific Language Impairment, narrative development in the L1 cognitive system, and L2 Japanese acquisition and its relation to L1 acquisition. The language processing section focuses on both L1 and L2 Japanese processing, covering topics such as the role of prosodic information in production/comprehension, the processing of complex grammatical structures such as relative clauses, processing issues related to variable word order, and lexical and sentence processing in L2 by speakers of different native languages. The Handbook of Japanese Applied Linguistics complements the Psycholinguistics volume by examining language acquisition from broader sociocultural perspectives, including language as a means of communication and as a social behavioral system, emphasizing pragmatic development as central to both L1 and L2 acquisition
Introduction to the Handbooks of Japanese Language and Linguistics
XIX
and to overall human development. Topics approached from these perspectives include the role of caregiver speech in early language development, literacy acquisition, and the acquisition of writing skills. Closely related to L1 and L2 acquisition and development are studies of bilingualism/multilingualism and the teaching and learning of foreign languages, including Japanese as a second language, where topics are discussed such as cross-lingual transfer from L1 to L2, learning errors, and proficiency assessment of second language acquisition. Chapters dealing with topics more squarely falling in the domain of applied linguistics cover issues in corpus/computational linguistics (including discussions of CHILDES for Japanese and the YK corpus, both widely used in research on Japanese as a second language), clinical linguistics (including discussions of language development in children with hearing impairment and other language disorders, Down syndrome, and autism), and translation and interpretation. Technically speaking, Japanese Sign Language is not a variety of Japanese, but in view of the importance of this language in Japanese society and because of the rapid progress in sign language research in Japan and abroad and for what it has to offer to the general theory of language, chapters dealing with Japanese Sign Language are also included in this volume.
7 Grammatical sketch of standard japanese The following pages offer a brief overview of Japanese grammar as an aid to a quick grasp of the structure of Japanese that may prove useful in studying individual, thematically organized handbooks in this series. One of the difficult problems in describing non-European languages using familiar technical terms derived from the European grammatical tradition concerns mismatches between what the glosses may imply and what grammatical categories they are used to denote in the description. We will try to illustrate this problem below by way of a warning not to take all glosses at their face value. But first some remarks are in order about the conventions of transcription of Japanese, glossing of examples, and their translations used in this series.
7.1 Writing, alphabetic transcription, and pronunciation Customarily, Japanese is written by using a mixture of Chinese characters (for content words), hiragana (for function words such as particles, suffixes, and inflectional endings), katakana (for foreign loans and mimetics), and sometimes the Roman alphabet. Because Japanese had no indigenous writing system, it developed two phonogram systems for representing the phonological unit of “mora,” namely hiragana and katakana, by simplifying or abbreviating (parts of) Chinese characters. Hiragana and katakana syllabaries are shown in Table 1, together with the alphabetic transcriptions adopted in the HJLL series.
XX
Masayoshi Shibatani and Taro Kageyama
Table 1: Alphabetic transcriptions adopted in HJLL. transcription hiragana katakana
a あ ア
ka か カ
sa さ サ
ta た タ
na な ナ
ha は ハ
ma ま マ
ya や ヤ
ra ら ラ
wa わ ワ
transcription hiragana katakana
i い イ
ki き キ
si し シ
ti ち チ
ni に ニ
hi ひ ヒ
mi み ミ
– – –
ri り リ
– – –
transcription hiragana katakana
u う ウ
ku く ク
su す ス
tu つ ツ
nu ぬ ヌ
hu ふ フ
mu む ム
yu ゆ ユ
ru る ル
– – –
transcription hiragana katakana
e え エ
ke け ケ
se せ セ
te て テ
ne ね ネ
he へ ヘ
me め メ
– – –
re れ レ
– – –
transcription hiragana katakana
o お オ
ko こ コ
so そ ソ
to と ト
no の ノ
ho ほ ホ
mo も モ
yo よ ヨ
ro ろ ロ
o を ヲ
n ん ン
Because of phonological change, the columns indicated by strikethroughs have no letters in contemporary Japanese, although they were filled in with special letters in classical Japanese. If all the strikethroughs were filled, the chart would contain 50 letters for each hiragana and katakana, so the syllabary chart is traditionally called Gojū-on zu (chart of 50 sounds). To these should be added the letter ん or ン representing a moraic nasal [N], on the rightmost column. The “50-sound chart,” however, does not exhaust the hiragana and katakana letters actually employed in Japanese, because the basic consonant sounds (k, s, t, h) have variants. The sound represented by the letter h is historically related to the sound represented by p, and these voiceless obstruents (k, s, t, and p) have their respective voiced counterparts (g, z, d, and b). Table 2 shows letters for these consonants followed by five vowels. Table 2: Letters for voiced obstruents and bilabial [p]. ga が
za ざ
da だ
ba ば
pa ぱ
katakana
ガ
ザ
ダ
バ
パ
transcription hiragana katakana
gi ぎ ギ
zi じ ジ
zi ぢ ヂ
bi び ビ
pi ぴ ピ
transcription hiragana katakana
gu ぐ グ
zu ず ズ
du づ ヅ
bu ぶ ブ
pu ぷ プ
transcription hiragana
Introduction to the Handbooks of Japanese Language and Linguistics
It is important to note that Tables 1 and 2 show the conventional letters and alphabetical transcription adopted in the text of the HJLL series; they are not intended to represent the actual pronunciations of Japanese vowels and consonants. For example, among the vowels, the sound represented as “u” is pronounced as [ɯ] with unrounded lips. Consonants may change articulation according to the vowels that follow. The following will require particular attention. There are two Romanization systems widely used in Japan. One, known as the Hepburn system, is more widely used in public places throughout Japan such as train stations, street signs, as well as in some textbooks for learners of Japanese. This system is ostensibly easier for foreigners familiar with the English spelling system. Another, the Kunreishiki (the cabinet ordinance system), is phonemic in nature and is used by many professional linguists. The essential differences between the two Romanization systems center on palatalized and affricate consonants, as shown in Table 3 below with some representative syllables for which the two Romanization renditions differ: Table 3: Two systems of Romanization. Hiragana
IPA
Hepburn
Kunreishiki
し しゃ しゅ しょ じ and ぢ じゃ じゅ じょ ち ちゃ ちゅ ちょ つ づ and ず ふ
shi sha shu sho ji ja ju jo chi cha chu cho tsu zu fu
si sya syu syo zi zya zyu zyo ti tya tyu tyo tu zu hu
XXII
Masayoshi Shibatani and Taro Kageyama
Except for the volumes on Ryukyuan, Ainu, and Japanese dialects, whose phonetics differ from Standard Japanese, HJLL adopts the Kunreishiki system for rendering cited Japanese words and sentences but uses the Hepburn system for rendering conventional forms such as proper nouns and technical linguistic terms in the text and in the translations of examples. Japanese sentences cited in HJLL look as below, where the first line transliterates a Japanese sentence in Kunreishiki Romanization, the second line contains interlinear glosses largely following the Leipzig abbreviation convention, and the third line is a free translation of the example sentence. (1) Taroo wa Ziroo to Tookyoo e it-te Taro TOP Jiro COM Tokyo ALL go-GER ‘Taro went to Tokyo with Jiro and bought socks.’
kutusita o kat-ta. sock ACC buy-PST
The orthographic convention for rendering Japanese is to represent a sentence with an uninterrupted sequence of Sino-Japanese characters and katakana or hiragana syllabaries without a space for word segmentation, as in 太郎は次郎と東京へ行って 靴下を買った for (1). In line with the general rules of Romanization adopted in books and articles dealing with Japanese, however, HJLL transliterates example sentences by separating word units by spaces. The example in (1) thus has 10 words. Moreover, as in it-te (go-GERUND) and kat-ta (buy-PAST) in (1), word-internal morphemes are separated by a hyphen whenever necessary, although this practice is not adopted consistently in all of the HJLL volumes. Special attention should be paid to particles like wa (topic), to ‘with’ and e ‘to, toward’, which, in the HJLL representation, are separated from the preceding noun or noun phrase by a space (see 7.3). Remember that case and other kinds of particles, though spaced, form phrasal units with their preceding nouns.
7.2 Word order As seen in (1), Japanese is a verb-final, dependent-marking agglutinative language. It is basically an SOV language which marks nominal dependent arguments by particles (wa, to, e, and o above) and whose predicative component consists of a verbal stem with a variety of suffixes, auxiliary verbs, and semi-independent predicate extenders pertaining to the speech act of predication (see section 7.6). While a verb is rigidly fixed in sentence final position, the order of subject and object arguments may vary depending on pragmatic factors such as emphasis, background information, and cohesion. Thus, sentence (2a) with the unmarked order below, in principle may vary in multiple ways as shown by some possibilities in (2b)-(2d).
Introduction to the Handbooks of Japanese Language and Linguistics
XXIII
(2) a. Taroo ga Hanako ni Ziroo o syookai-si-ta. Taro NOM Hanako DAT Jiro ACC introducing-do-PST ‘Taro introduced Jiro to Hanako.’ b. Taroo ga Ziroo o Hanako ni syookai-si-ta. c. Hanako ni Taroo ga Ziroo o syookai-si-ta. d. Ziroo o Taroo ga Hanako ni syookai-si-ta. Adverbs, likewise, can be rather freely placed, though each type of adverbs has its own basic position. (3) a. Saiwainimo Hanako ga gohan o tai-te kure-te luckily Hanako NOM rice ACC cook-GER GIVE-GER ‘Luckily Hanako had done the favor of cooking the rice (for us).’ b. Hanako ga saiwainimo gohan o tai-te kure-te i-ta. c. Hanako ga gohan o saiwainimo tai-te kure-te i-ta.
i-ta. BE-PST
Notice that while the verbal complex in the sentence above is not as tightly organized as a complex involving suffixes, a sentence adverb cannot be placed within the verbal complex, showing that the sequence of tai-te kure-te i-ta forms a tighter constituent which, however, permits insertion of the topic particle wa after each of the gerund-forms. (See section 7.4 below on the nature of gerund-forms in Japanese.) As the normal position of sentence adverbs is sentence initial, manner and resultative adverbs have an iconically-motivated position, namely before and after the object noun phrase, respectively, as below, though again these adverbs may move around with varying degrees of naturalness: (4)
Hanako ga isoide gohan o tai-te kure-ta. Hanako NOM hurriedly rice ACC cook-GER GIVE-PST ‘Hanako hurried did the favor of cooking the rice (for us).’
(5) Hanako ga gohan o yawarakaku tai-te Hanako NOM rice ACC softly cook-GER ‘Hanako did the favor of cooking the rice soft (for us).’
kure-ta. GIVE-PST
The fact that an object noun phrase can be easily separated from the verb, as in (2b.d), and that adverbs can freely intervene between an object and a verb, as in (5), has raised the question whether Japanese has a verb phrase consisting of a verb and an object noun phrase as a tightly integrated constituent parallel to the VP in English (cf. ✶cook hurriedly the rice – the asterisk marks ungrammatical forms).
XXIV
Masayoshi Shibatani and Taro Kageyama
7.3 NP structure Noun phrases, when they occur as arguments or adjuncts, are marked by case particles or postpositions that are placed after their host nouns. Because case markers can be set off by a pause, a filler, or even longer parenthetic material, it is clear that they are unlike declensional affixes in inflectional languages like German or Russian. Their exact status, however, is controversial; some researchers regard them as clitics and others as (non-independent) words. Elaboration of Japanese noun phrases is done by prenominal modifiers such as demonstratives, genitive noun phrases, or adjectives, as below, indicating that Japanese is a consistent head-final language at both nominal and clausal levels. (6)
a. kono Taroo no kaban this Taro GEN bag lit. ‘this Taro’s bag’ b. Taroo no kono kaban Taro GEN this bag lit. ‘Taro’s this bag’
Japanese lacks determiners of the English type that “close off” NP expansion. The literal translations of the Japanese forms above are ungrammatical indicating that English determiners like demonstratives and genitive noun phrases do not allow further expansion of an NP structure. Also seen above is the possibility that prenominal modifiers can be reordered just like dependents at the sentence level. The order of prenominal modifiers, however, is regulated by the iconic principle of placing closer to the head noun those modifiers that have a greater contribution in specifying the nature and type of the referent. Thus, descriptive adjectives tend to be placed closer to a head noun than demonstratives and genitive modifiers of non-descriptive types. Interesting is the pattern of genitive modifiers, some of which are more descriptive and are placed closer to the head noun than others. Genitives of the same semantic type, on the other hand, can be freely reordered. Compare: (7)
a.
b.
(8)
a.
Yamada-sensei no kuroi kaban Yamada-professor GEN black bag ‘Professor Yamada’s black bag’ ✶ kuroi Yamada-sensei no kaban (O.K. with the reading of ‘a bag of Professor Yamada who is black’) Yamada-sensei no gengogaku no koogi Yamada-professor GEN linguistics GEN lecture ‘Professor Yamada’s linguistics lecture’
Introduction to the Handbooks of Japanese Language and Linguistics
(9)
b.
gengogaku no Yamada-sensei no koogi (O.K. with the reading of ‘a lecture by Professor Yamada of linguistics’)
a.
Yamada-sensei no kinoo no koogi Yamada-professor GEN yesterday GEN lecture lit. ‘Professor Yamada’s yesterday’s lecture’ ‘Yesterday’s lecture by Professor Yamada’ Kinoo no Yamada-sensei no koogi
b. (10) a.
b. (11) a.
XXV
b.
✶
oomori no sio-azi no raamen big.serving GEN salt-tasting GEN ramen lit. ‘big-serving salt-tasting ramen noodles’ sio-azi no oomori no raamen atui sio-azi no raamen hot salt-tasting GEN ramen ‘hot salt-tasting ramen noodles’ sio-azi no atui ramen
Numeral classifiers (CLFs) pattern together with descriptive modifiers so that they tend to occur closer to a head noun than a possessive genitive phrase. (12) a.
b.
Taroo no
san-bon
no
enpitu
Taro GEN three-CLF GEN pencil ‘Taro’s three pencils’ ✶ san-bon no Taroo no enpitu
Numeral classifiers also head an NP, where they play a referential function and where they can be modified by a genitive phrase or an appositive modifier, as in (13a, b). They may also “float” away from the head noun and become adverbial, as in (13c). (13) a.
b.
c.
Taroo wa gakusei no san-nin o mikake-ta. Taro TOP student GEN three-CLF ACC see.by.chance-PST ‘Taro saw three of the students by chance.’ Taroo wa gakusei san-nin o mikake-ta. Taro TOP student three-CLF ACC see.by.chance-PST lit. ‘Taro saw student-threes by chance.’ Taroo wa gakusei o san-nin mikake-ta. Taro TOP student ACC three-CLF see.by.chance-PST ‘Taro saw students, three (of them), by chance.’
XXVI
Masayoshi Shibatani and Taro Kageyama
As in many other SOV languages, so-called relative clauses are also prenominal and are directly placed before their head nouns without the mediation of “relative pronouns” like English which or who or “complementizers” like that. Predicates in relative clauses are finite, taking a variety of tense and aspect. The subject may be replaced by a genitive modifier. Observe (14a). (14) a.
b.
Boku mo [Taroo ga/no kat-ta] hon o kat-ta. I ADVPART Taro NOM/GEN buy-PST book ACC buy-PST ‘I also bought the book which Taro bought.’ Boku mo [Taroo ga/no kat-ta] no o kat-ta. I ADVPART Taro NOM/GEN buy-PST NM ACC buy-PST ‘I also bought the one which Taro bought.’
The structure used as a modifier in the relative clause construction can also head a noun phrase, where it has a referential function denoting an entity concept evoked by the structure. In Standard Japanese such a structure is marked by the nominalization particle no, as in (14b).
7.4 Subject and topic Some of the sentences above have noun phrases marked by the nominative case particle ga and some by the topic marker wa for what appear to correspond to subject noun phrases in the English translations. This possibility of ga- and wa-marking is seen below. (15) a. Yuki ga siro-i. snow NOM white-PRS ‘The snow is white.’ b. Yuki wa siro-i. snow TOP white-PRS ‘Snow is white.’ As the difference in the English translations indicates, these two sentences are different in meaning. Describing the differences between topic and non-topic sentences has been a major challenge for Japanese grammarians and teachers of Japanese alike. The difference in the English translations above, however, is indicative of how these two sentences might differ in meaning. Sentence (15a) describes a state of affairs involving specific snow just witnessed, whereas (15b) is a generic statement about a property of snow unbounded by time. Thus, while (15a) would be uttered only when the witnessed snow is indeed white, (15b) would be construed true even though we know that there are snow piles that are quite dirty.
Introduction to the Handbooks of Japanese Language and Linguistics
XXVII
A similar difference is seen in verbal sentences as well. (16) a.
b.
Tori ga tob-u. bird NOM fly-NONPST ‘A bird is flying/is about to fly.’ Tori wa tob-u. bird TOP fly-NONPST ‘Birds fly.’
Non-topic sentences like (15a) and (16a) are often uttered with an exclamation accompanying a sudden discovery of a state of affairs unfolding right in front of one’s eyes. The nonpast tense forms (-i for adjectives and -(r)u for verbs) here anchor the time of this discovery to the speech time. The nonpast tense forms in (15a) and (16b), on the other hand, mark a generic tense associated with a universal statement. These explanations can perhaps be extended to time-bound topic sentences seen in (17b) below. (17) a.
b.
Taroo ga hasit-ta. Taro NOM run-PST ‘Taro NOM ran.’ Taroo wa hasit-ta. Taro TOP run-PST ‘Taro ran.’
That is, while (17a) describes an occurrence of a particular event at a time prior to the speech time, (17b) describes the nature of the topic referent – that Taro was engaged in the running activity – as a universal truth of the referent, but universal only with respect to a specifically bound time marked by the past tense suffix. Topics need not be subjects, and indeed any major sentence constituent, including adverbs, may be marked as topic in Japanese, as shown below. (18) a.
b.
c.
Sono hon wa Taroo ga yon-de i-ru. that book TOP Taro NOM read-GER be-NONPST ‘As for that book, Taro is reading (it).’ Kyoo wa tenki ga yo-i. today TOP weather NOM be. good-NONPST ‘As for today, the weather is good.’ Sonnani wa hayaku wa hasir-e na-i. that.way TOP quickly TOP run-POTEN NEG-NONPST ‘That quickly, (I) cannot run.’
XXVIII
Masayoshi Shibatani and Taro Kageyama
7.5 Complex sentences Like other Altaic languages, compound sentences in Japanese do not involve a coordinate conjunction like English and. Instead, clauses are connected by the use of inflected verb forms, as in (19a) below, where the -i ending is glossed in the HJLL series as either INF (infinitive) or ADVL (adverbal) following the Japanese term ren’yō-kei for the form. While the -i ending in the formation of compound sentences is still used today, especially in writing, the more commonly used contemporary form involves a conjunctive particle -te following the -i infinitive form, as in (19b) below. In HJLL, this combination is glossed as GER (gerund), though the relevant Japanese forms do not have the major nominal use of English gerund-forms. (19) a.
b.
Hana wa sak-i, tori flower TOP bloom-INF bird ‘Flowers bloom and birds sing.’ Hana wa sa-i.te, tori flower TOP bloom-GER bird ‘Flowers bloom and birds sing.’
wa uta-u. TOP sing-NONPST wa uta-u. TOP sing-NONPST
Both the -i and -ite forms play important roles in Japanese grammar. They are also used in clause-chaining constructions for serial events (20a), and in complex sentences (20b)(20d), as well as in numerous compound verbs (and also in many compound nouns) such as sak-i hokoru (bloom-INF boast) ‘be in full bloom’, sak-i tuzukeru (bloom-INF continue) ‘continue blooming’, sa.i-te iru (bloom-GER be) ‘is blooming’, and sa.i-te kureru (bloom-GER GIVE) ‘do the favor of blooming (for me/us)’. (20)
a. Taroo wa [ok-i/ok.i-te], [kao o ara-i/arat-te], Taro TOP rise-INF/rise-GER face ACC wash-INF/wash-GER [gohan o tabe-ta]. meal ACC eat.PST ‘Taro got up, washed his face, and ate a meal.’ b. Taroo wa [sakana o tur-i] ni it-ta. Taro TOP fish ACC catch-INF DAT go-PST ‘Taro went to catch fish.’ c. Taroo wa [aruk-i nagara] hon o yon-da. Taro TOP walk-INF SIMUL book ACC read-PST ‘Taro read a book while walking.’ d. Taroo wa [Hanako ga ki-ta no] ni awa-na-katta. Taro TOP Hanako NOM come-PST NM DAT see-NEG-PST. ‘Taro did not see (her), even though Hanako came.’
Introduction to the Handbooks of Japanese Language and Linguistics
XXIX
(20d) has the nominalized clause marked by the particle no followed by the dative ni, also seen in (20b) marking the purposive form. In modern Japanese the no-ni sequence has been reanalyzed as a concessive conjunction.
7.6 Context dependency The context dependency of sentence structure in Japanese is much more clearly pronounced than in languages like English. Indeed, it is rare that Japanese sentences express all the arguments of a verb such as a subject (or topic) and an object noun phrase included in the sentences used above for illustrative purposes. A typical dialog would take the following form, where what is inferable from the speech context is not expressed. (21) a.
Speaker A:
Speaker B:
b.
Tokorode, Murakami Haruki no saisin-saku by.the.way Murakami Haruki GEN newest-work yon-da ka. read-PST Q ‘By the way, have (you) read Haruki Murakami’s latest work?’ Un, moo yon-da. uh-hu already read-PST ‘Uh-hu, (I) have already read (it)’.
In (21a) A’s utterance is missing a subject noun phrase referring to the addressee, and B’s response in (21b) is missing both subject and object noun phrases. In some frameworks, sentences like these are analyzed as containing zero pronouns or as involving a process of “pro drop,” which deletes assumed underlying pronouns. This kind of analysis, however, ignores the role of speech context completely and incorporates information contextually available into sentence structure. In an analysis that takes seriously the dialogic relationship between speech context and sentence structure, the expressions in (21) would be considered full sentences as they are.
7.7 Predicative verbal complexes and extenders Coding or repeating contextually determinable verb phrases, as in (21b), is less offensive than expressing contextually inferable noun phrases, presumably because verb phrases have the predication function of assertion, and because they also code a wide range of other types of speech acts and of contextual information pertaining to the predication act. Declarative sentences with plain verbal endings like the one in (21b) are usable as “neutral” expressions in newspaper articles and literary works, where
XXX
Masayoshi Shibatani and Taro Kageyama
no specific reader is intended. In daily discourse, the plain verbal forms “explicitly” code the speaker’s attitude toward the hearer; namely, that the speaker is treating the hearer as his equal or inferior in social standing, determined primarily by age, power, and familiarity. If the addressee were socially superior or if the occasion demanded formality, a polite, addressee honorific form with the suffix -masu would be used, as below. (22) Hai, moo yom-i-masi-ta. yes already read-INF-POL-PST ‘Yes, (I have) already read (it).’ Referent honorific forms are used when the speaker wishes to show deference toward the referent of arguments – subject honorific and object honorific (or humbling) forms, depending on the type of argument targeted. If (21b) were to be uttered in reference to a social superior, the following would be more appropriate: (23) Un, (Yamada-sensei wa) moo yom-are-ta. uh-hu (Yamada-professor TOP) already read-SUB.HON-PST ‘Uh-hu, (Professor Yamada has) already read (it).’ This can be combined with the polite ending -masu, as below, where the speaker’s deference is shown to both the referent of the subject noun phrase and the addressee: (24) Hai, (Yamada-sensei wa) moo yom-are-masi-ta. Yes (Yamada-professor TOP) already read-HON-POL-PST ‘Yes, (Professor Yamada has) already read (it).’ As these examples show, Japanese typically employs agglutinative suffixes in the elaboration of verbal meanings associated with a predication act. The equivalents of English auxiliary verbs are either suffixes or formatives connected to verb stems and suffixed forms in varying degrees of tightness. These are hierarchically structured in a manner that expresses progressively more subjective and interpersonal meaning as one moves away from the verb-stem core toward the periphery. For example, in the following sentence a hyphen marks suffixal elements tightly bonded to the preceding form, an equal sign marks a more loosely connected formative, which permits insertion of certain elements such as the topic particle wa, and a space sets off those elements that are independent words following a finite predicate form, which may terminate the utterance.
Introduction to the Handbooks of Japanese Language and Linguistics
XXXI
(25) (Taroo wa) ik-ase-rare-taku=na-katta rasi-i mitai (Taro TOP) go-CAUS-PASS-DESI=NEG-PST CONJEC-NONPST UNCERT des-u wa. COP.POL-NONPST SFP ‘(Taro) appears to seem to not want to have been forced to go, I tell you.’ The final particle wa above encodes the information that the speaker is female. A male speaker would use yo or da yo, the latter a combination of the plain copula and yo, instead of desu wa above, or combinations such as da ze and da zo in rough speech. Non-declarative Japanese sentences, on the other hand, frequently suppress auxiliary verbs, the copula, and the question particle, especially in casual speech, where intonation and tone of voice provide clues in guessing the intended speech act. Casual interrogatives take the form of (26a) with a nominalization marker bearing a rising intonation, marked by the question mark in the transcription, whereas fuller versions have the interrogative particle ka or a combination of the polite copula and ka, as in (26b). (26) a.
b.
Moo kaer-u already return-NONPST ‘Going home already?’ Moo kaer-u already return-NONPST ‘Going home already?’
no? NM no (des.u) NM (COP.POL-NONPST)
ka. Q
Requests are made with the aid of an auxiliary-like “supporting” verb kureru ‘GIVE (ME THE FAVOR OF …)’, its polite form kudasai, or its intimate version tyoodai, as seen in (27a). Again, these forms are often suppressed in a highly intimate conversation and may result in a form like (27b). (27) a.
b.
Hayaku kaet-te kure/kudasai/tyoodai. soon return-GER GIVE.IMP/GIVE.POL-IMP/GIVE.INTI ‘(Please) come home soon (for me/us).’ Hayaku kaet-te ne. soon return-GER SFP ‘(Please) come home soon, won’t you?’
XXXII
Masayoshi Shibatani and Taro Kageyama
The use of dependent forms (e. g., the gerund -te form above) as independent sentences is similar to that of subjunctive forms in European languages as independent sentences, as illustrated by the English sentence below. (28) If you would give me five thirty-cent stamps. Conditionals are used as independent suggestion sentences in Japanese as well. For example, (29a) has a fuller version like (29b) with the copula as a main-clause verb, which can also be suppressed, giving rise to the truncated form (29c). (29) a.
b.
c.
Hayaku kaet-tara? quickly return-COND lit. ‘If return quickly.’ ‘Why don’t you go home quickly?’ Hayaku kaet-tara ikaga des-u ka. quickly return-COND how COP.POL-NONPST Q lit. ‘How would it be it if (you) went home quickly?’ Hayaku kaet-tara ikaga? quickly return-COND how ‘Why don’t (you) go home quickly?’
Understanding Japanese utterances requires full recourse to the elements of speech context, such as the nature of the speaker and the hearer and the social relationship between them, the information “in the air” that is readily accessible to the interlocutors, and the formality of the occasion. Indeed, the difficult part of the art of speaking Japanese is knowing how much to leave out from the utterance and how to infer what is left unsaid.
8 Conclusion Many of the interesting topics in Japanese grammar introduced above are discussed in great detail in the Lexicon-Word Formation volume, the Syntax volume, and the present Semantics and Pragmatics volume of the HJLL series.. The Historical Linguistics volume also traces developments of some of the forms and constructions introduced above. The Sociolinguistics volume gives fuller accounts of sentence variations motivated by context and discourse genre.
Introduction to the Handbooks of Japanese Language and Linguistics
Appendix: List of abbreviations for HJLL 1 2 3 A ABL ACC ACOP ADJ AND ADV ADVL ADVPART AGR AGT ALL AN ANTIP AP APPL ART ASP ATTR AUX AUXV C CAUS CLF COHORT COM COMP COMPL CONC CONCL COND CONJEC CONJCT CONT COP CVB DAT D DECL DEF DEM DET DESI DIST
first person second person third person agent-like argument of canonical transitive verb ablative accusative adjectival copula adjective adnominal adverb(ial(izer)) adverbial adverbial particle agreement agent allative adjectival noun antipassive adverbial particle, adjective phrase applicative article aspect attributive auxiliary auxiliary verb consonant causative classifier cohortative comitative complementizer completive concessive conclusive conditional conjectural conjunctive continuative copula converb dative demonstrative declarative definite demonstrative determiner desiderative distal
XXXIII
XXXIV
Masayoshi Shibatani and Taro Kageyama
DISTR DO DU DUR EMPH ERG ETOP EVID EXCL EXPL FOC FUT GEN GER H HON HUM IMP INCL IND INDEF INF INS INT INTERJEC INTI INTR IO IRR ITERA k-irr L LB LM LOC MPST MVR N n-irr NCONJ NEC NEG NM NMLZ NMNL NOM NONPST NP OBJ
Introduction to the Handbooks of Japanese Language and Linguistics
OBL OPT P PART PASS PST PCONJ PERF PL POL POLCOP POSS POTEN PP PRED PRF PRS PRES PROG PROH PROV PROX PST PSTCONJ PTCP PURP Q QD QUOT r‐irr REAL RECP REFL RES RESP S SBJ SBJV SFP SG SIMUL s‐irr SPON SPST STAT TOP TR UB UNCERT
oblique optative patient-like argument of canonical transitive verb, preposition, postposition particle passive past present conjectural perfective plural polite polite copula possessive potential prepositional/postpositional phrase predicative perfect present presumptive progressive prohibitive provisional proximal/proximate past past conjectural participle purposive question/question particle/question marker quadrigrade (yodan) quotative r‐irregular (ra‐hen) realis reciprocal reflexive resultative respect single argument of canonical intransitive verb, sentence subject subjunctive sentence final particle singular simultaneous s-irregular (sa-hen) spontaneous simple past stative topic transitive upper bigrade (kami-nidan uncertain
Languages ConJ EMC EMJ EOJ J-Ch LMC LMJ JPN MC MJ MK ModJ OC OJ pJ pK SJ Skt
contemporary Japanese Early Middle Chinese Early Middle Japanese Eastern Old Japanese Japano-Chinese Late Middle Chinese Late Middle Japanese Japanese Middle Chinese Middle Japanese Middle Korean Modern Japanese Old Chinese Old Japanese proto-Japanese proto-Korean Sino-Japanese Sanskrit
References Chamberlain, Basil H. 1895. Essay in aid of a grammar and dictionary of the Luchuan language. Transactions of the Asiatic Society of Japanese, vol. 23 supplement. Frellesvig, Bjarke.2010. A history of the Japanese language. Cambridge: Cambridge University Press. Martin, Samuel E. 1975. A reference grammar of Japanese. New Haven: Yale University Press. Miller, Roy A. 1971. Japanese and the other Altaic languages. Chicago: University of Chicago Press. Shibatani, Masayoshi. 1990. The languages of Japan. Cambridge: Cambridge University Press.
Contents Preface
IX
Masayoshi Shibatani and Taro Kageyama Introduction to the Handbooks of Japanese Language and Linguistics Contributors
XI
XLI
Bjarke Frellesvig and Satoshi Kinsui Introduction 1
Part I: Prehistory and reconstruction Alexander Vovin 1 Reconstruction of Japonic and para-Japonic based on external sources Thomas Pellard 2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan Akiko Matsumori 3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan
39
69
Part II: Phonology Teruhiro Hayata 4 Reconstruction of Old Japanese phonology J. Marshall Unger 5 Old Japanese writing and phonology
101
133
Sven Osterkamp 6 Late Middle Japanese phonology, based on Korean sources
145
Masayuki Toyoshima 7 Late Middle Japanese phonology as reflected in early Japanese Christian documents 169 Marc Hideo Miyake 8 Sino-Japanese
179
11
XXXVIII
Contents
Elisabeth M. de Boer 9 The Ramsey hypothesis
195
Part III: Grammar Yuko Yanagida 10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax 223 Yoshiyuki Takayama 11 The syntax and morphology of Early Middle Japanese Hirofumi Aoki 12 Late Middle Japanese grammar
253
269
Takashi Nomura 13 The historical changes in the case marking system of Japanese
301
Charles J. Quinn 14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate 313 Akira Watanabe 15 Loss of wh-movement
347
Tomohide Kinuhata 16 Development of adverbial particles Tomoko Okazaki 17 The history of demonstratives
371
393
Yoshihiko Ikegami 18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’ 405
Part IV: Lexicon, Materials and Kanbun John R. Bentley 19 The history of basic vocabulary in Japanese
437
Contents
Toru Maruyama 20 The Japanese lexicon as reflected in Christian materials
453
Takashi Inukai 21 What mokkan (wooden documents) can tell us about ancient Japanese language 463 Shōju Ikeda 22 Early Japanese dictionaries
473
Masayuki Tsukimoto 23 Kunten texts of Buddhist provenance (butten 仏典): Their characteristics and actuality 493 Teiji Kosukegawa 24 Kunten texts of secular Chinese origin (kanseki 漢籍)
503
Shingo Yamamoto 25 Japanized written Chinese: Its features and contribution to the history of the Japanese language 509 Fumitoshi Saito 26 Early modern kanbun and kanbun-kundoku
523
Valerio Luigi Alberizzi 27 The influence of kanbun-kundoku vocabulary on the Japanese language 539 Index
565
XXXIX
Contributors Valerio Luigi Alberizzi holds a Ph.D. in Japanese Language and Linguistics from the University of Ca’ Foscari, Venice (Italy). He is currently Assistant Professor at the Training Center for Foreign Languages and Diction, Tokyo University of the Arts, Tokyo. He has taught courses on modern and classical Japanese language in the Department of East Asian Studies, University of Ca’ Foscari, Venice, and in the Department of Modern Foreign Languages and Literatures at the Alma Mater Studiorum University of Bologna. In the last ten years he has also been in charge of the courses in Contrastive Linguistics and Second-language Acquisition at Shirayuri College, Tokyo; from 2012 to 2017 he was the Italian language program coordinator at the Global Education Center, Waseda University, Tokyo. His interests range from historical linguistics to foreign-language teaching using new technologies. Hirofumi Aoki is Professor at the Kyushu University Faculty of Humanities. He completed his Ph.D. at Kyushu University in 1999. His research interest is in the history of Japanese grammar. He is the author of Go-keisei kara mita nihongo-bunpō-si [A historical study of Japanese grammar from the perspective of word formation] (Hituzi Syobo, 2010), Nihongo rekisi tōgo-ron zyosetu [An introduction to Japanese historical syntax] (Hituzi Syobo, 2016). He is co-author of Bunpō-ka, goi-ka, kōbun-ka [Grammaticalization, Lexicalization, and Constructionalization] (Kaitakusha, 2020). John R. Bentley received his Ph.D. in Japanese Language from the University of Hawai‘i at Manoa in 1999. He currently is Professor of Japanese at Northern Illinois University where he teaches all levels of Japanese, and serves as the Department Chair. He is the author of books and articles on Japanese language history, Ryukyuan linguistics, Japanese early historiography, and Kokugaku. He is the author of ABC Dictionary of Ancient Japanese Phonograms (University of Hawaii Press, 2016) and An Anthology of Kokugaku Scholars: 1690–1868 (Cornell, 2017). His research interests center on the history of Japanese language and its influence on culture. † Elisabeth M. de Boer (1966–2023) was senior lecturer in Japanese linguistics at the Ruhr Universität Bochum. Her research concentrates on historical changes in the Japanese tone systems, on which she published various articles as well as the monograph The historical development of Japanese tone (2010, Harrasowitz). She headed a research project funded by the European Research Council on links between dialect diversification in Japanese (concentrating on the Japanese tone systems) and migration routes spreading the Japanese language through Japan. Her book on this topic What tonal data can tell us about the history of the Japanese language: Concentrating on evidence from the Tokyo type tone systems will be published by John Benjamins Publishing Company. Teruhiro Hayata is Professor emeritus, Faculty of Letters, Kyushu University. His research interest encompasses morphology and phonology of Manchu. He is also interested in the phonology and morphology of ancient Japanese. He is the author of “Accusativus subjecti in Manchu” in Papers in Linguistics, Vol. 32 (Kyushu University, 2011), “A historical overview of the alternation of n~0 in Manchu” in ALTAI HAKPO / Journal of the Altaic Society of Korea (Altaic Society of Korea, 2012), and Jōdai Nihongo no on’in [The phonology of Old Japanese] (Iwanami Shoten, 2017). Shōju Ikeda is Professor emeritus at Hokkaido University. He is a President of the Society for Research in Kunten language. He has been conducting research on linguistic studies of early dictionaries in Japan. He has published articles in anthologies and journals including Kokugogaku and Kuntengo to Kunten-shiryo.
https://doi.org/10.1515/9781614512851-205
XLII
Contributors
Yoshihiko Ikegami is Professor emeritus at the Department of Language and Information Sciences, Graduate School of Arts and Sciences, University of Tokyo. Currently, he is also honorary president of the Japanese Cognitive Linguistics Association and honorary member of the Japanese Association for Semiotics. He received his B.A. and M.A. in English at Tokyo University and M.Phil. and Ph.D. in linguistics at Yale University. He was guest professor at Indiana University (semiotics), Universität München (poetics) and Freie Universität Berlin (Japanese linguistics) and lectured at a number of universities mainly in Europe and in China. Among his numerous publications are Suru to naru no gengogaku: Gengo to bunka no taiporojii e no shiron (Taishukan, 1981) [Sprachwissenschaft des Tuns und Werdens: Typologie der Japanischen Sprache und Kultur, a German translation (2007) available from LIT Verlag, Berlin] and The Empire of Signs: Semiotic Essays on Japanese Culture (ed.) (John Benjamins, Amsterdam, 1991). Takashi Inukai received his Ph.D. in Linguistics from the University of Tsukuba in 1993. His research interests include grammatology and Old Japanese. He is also a phonetician. He published many books including Jōdai moji-gengo no kenkyū [Studies in the written language of Old Japanese] (Kasama-Shoin 1992 revised 2005), Mokkan ni yoru Nihongo shokishi [A historical study on the Japanese writing system based on mokkan] (Kasama-Shoin 2005 revised 2008) Gishiki de utau Yamato-uta [Ancient Japanese ballads sung in ceremonies] (Hanawa-Shobō 2017). Tomohide Kinuhata is a Professor in the Faculty of Humanities at Fukuoka University. His research interests encompass the history of Japanese grammar, Ryukyuan linguistics, and formal semantics/ pragmatics. His publications include “Historical development from subjective to objective meaning: evidence from the Japanese question particle ka” (Journal of Pragmatics 44(6–7), 2012), “Presupposition and assertion” (in W. Jacobsen and Y. Takubo (eds.), Handbook of Japanese Semantics and Pragmatics, Mouton, 2020), “Syntax/Semantics discrepancy in the grammaticalization of resultatives: Evidence from KarimataMiyako Ryukyuan” (Glossa 7(1), 2022), “Scope ambiguity and the loss of NPI feature: Evidence from the history of Japanese scalar particle dani” (H. Kishimoto, O. Sawada, and I, Imani (eds.), Polarity-sensitive expressions: Comparison between Japanese and other languages, Mouton, to appear). Teiji Kosukegawa received his M.A. from Hokkaido University in 1986, worked as an assistant professor there from 1987 to 1993, then taught Japanese linguistics at the University of Toyama from 1993 to 2022, and is currently a professor emeritus at the University of Toyama. His fields of research are mainly Japanese linguistics, with special attention to vernacular reading of Classical Chinese text in the Sinosphere (China, Japan, Korea, and Vietnam). He is the co-author of the bibliography for volumes 1, 5, 6, 7, 9, 11, and 12 of the Tōyō Bunko Zenpon Sōsho [Reproduction series of good books held by the Tōyō Bunko] (Bensei Shuppan, 2014–2015). Toru Maruyama is Professor emeritus of Nanzan University, Nagoya, Japan. His academic interest in general is the study of Jesuit linguistic works published in the sixteenth and seventeenth centuries in Africa, Brazil, India and Japan. How Father João Rodrigues described the Japanese language has been his major concern for a long time. In recent years he has also been involved in the analysis of Konkani Christian Doctrine published in India in 1622 and Konkani-Portuguese dictionary manuscripts of that time. Akiko Matsumori is a professor of linguistics at the Department of English at Japan Women’s University. Her research interests focus on synchronic and diachronic phonology (especially accent) of Japanese dialects, and historical tonology of Ryukyuan languages. She is an author and editor of Nihongo akusento nyūmon [Introduction to Japanese Accent] (Sanseido, 2012). Her other publications include “On the Reconstruction of the Proto-accentual System of Japanese” Frellesvig, B. and J.Whitman (eds.), ProtoJapanese: Issues and Prospects (John Benjamins, 2008) and “A Prosodic Unit, Recursive Structure and Nature of Accent of Miyako Ryukyuan” (The Linguistic Review. Vol. 29, No.1, De Gruyter Mouton, 2019).
Contributors
XLIII
Marc Hideo Miyake completed his Ph.D. in linguistics at the University of Hawai‘i at Mānoa in 1999. He is the author of Old Japanese: A Phonetic Reconstruction (Routledge, 2003) and its sequel, “Philological evidence for ✶e and ✶o in Pre-Old Japanese” (Diachronica 20.1 (2003), 83–137). He taught Chinese and Japanese language history and East Asian areal linguistics at the University of Oregon and then taught linguistics and the Japanese language at the University of Hawai‘i at Hilo. His current research interests focus on the decipherment of extinct Asian languages (Tangut, Jurchen, Khitan, and Pyu). His most recent publications include a decipherment of a Pyu-Sanskrit bilingual inscription (with Arlo Griffiths, Bob Hudson, and Julian K. Wheatley; Bulletin de l’École française d’Extrême-Orient, 2017) the first reconstructions of Pyu phonology (Bulletin of Chinese Linguistics, 2018; Language and Linguistics, 2021) and grammar (Linguistics of the TibetoBurman Area, 2019). Takashi Nomura is Professor emeritus at Tokyo University. His research interest is in historical change of Japanese, especially grammar. He is the author of Nihongo hyōjunkei no rekishi: Hanashikotoba, kakikotoba, hyōki [The history of Japanese standard form: Spoken, written, inscribed] (Kodansha, 2019) and others. Tomoko Okazaki (Ph.D., Osaka University, 2004) is Professor at the College of Letters, Ritsumeikan University. Her research is concerned with cohesion, nouns and conjunction, especially the historical changes of Japanese demonstrative pronouns and adverbs. She is the author of Nihongo shijishi no rekishitekikenkyū [A historical study of Japanese demonstratives] (Hituzi Syobo, 2010, Kindaichi Award) and Workbook nihongo no rekishi [Workbook, A history of the Japanese language] (co-authored with Yūta Mori, Kurosio Publishers, 2016). Sven Osterkamp is Professor of Japanese Language and Literature at Ruhr University Bochum. His research interests include Japonic historical linguistics, the early history of studies pertaining to East-Asian languages and writing systems in Europe, and the typology of writing systems. His recent publications include “The Last Days of Old Japanese: Early Heian gloss texts and the periodization of Japanese language history” in J. Kupchik et al. (eds.), Studies in Asian Historical Linguistics, Philology and Beyond (Brill, 2021) and “
e Ubiity of Polygray and its Significan for
e Typology of iti Systems” (with Gordian Schreiber) in Written Language and Literacy 24.2 (John Benjamins, 2021). Thomas Pellard is a researcher at the Centre national de la recherche scientifique (CNRS) in Paris and teaches linguistics at the Institut national des langues et civilisations orientales (INALCO). His research interests focus on the synchronic diversity and the diachronic diversification of the languages of Japan. He has carried out linguistic fieldwork on all Japonic languages and has published widely on the description of the endangered Ryukyuan languages, the historical-comparative reconstruction of Japanese and Ryukyuan, the linguistic archaeology on the Ryukyu Islands, and the phylogeny of the Japonic language family. † Charles J. Quinn (1948–2023) (Ph.D. University of Michigan, 1987) was Associate Professor in East Asian Languages & Literatures at The Ohio State University, where he taught Japanese language, pedagogy, and linguistics. He was visiting researcher at Keio University, Tohoku University, and the National Institute of Multimedia Education, and with Jane M. Bachnik co-authored and edited Situated Meaning: Inside and Outside in Japanese Self, Society and Language (Princeton, 1994). His articles on deriving inflected adjectives, and on particles zo, ka and adnominal-concordant kakari-musubi have appeared in Japanese/Korean Linguistics. “Mo than expected” (in Satoko Suzuki, ed. Emotive Communication in Japanese, Benjamins 2006) argues that Old Japanese final/mirative mo and inclusive mo were cognate. He was lead developer of Classical Japanese Portal, a web-based multimedia learning tool.
XLIV
Contributors
Fumitoshi Saito is Professor at the Graduate School of Humanities, Nagoya University. He completed his master’s degree at the University of Tokyo in 1988, and received his Ph.D, in Japanese Linguistics from Nagoya University in 2011. His research interest is in Kanbunkundoku (Japanese readings of Chinese texts) in the Edo and Meiji Period. He is the author of Kanbun-kundoku to kindai nihongo no keisei [Kanbun-kundoku and the formation of the modern Japanese language] (Bensei Shuppan, 2011). Yoshiyuki Takayama is Professor at the Faculty of Education, University of Fukui. He completed his Ph.D. at Osaka University in 1999. His research interest is historical development of Japanese syntactic structure and functions of auxiliary verbs. He is the author of Nihongo Modality no Shi-teki Kenkyu [A Historical Study of Japanese Modality] (Hituzi Syobo, 2002), Nihongo Bunposhi no Shikai [The Perspective of Historical Research on Japanese Grammar] (Hituzi Syobo, 2021). Masayuki Toyoshima is Professor emeritus at the department of Japanese Literature and Linguistics of the Graduate school for Humanities at Sophia University, Tokyo, Japan. His research interest is in the bibliographical and linguistic studies of the Early Christian documents in Japan, especially of the materials produced by the Jesuits in Japan in its early evangelization period (1549 to 1612). Among his publications are (Toyoshima ed.) Kirishitan to shuppan [Christian documents and printing] (Yagi shoten, 2012), which was awarded the 35th Japan Society of Publishing Studies Award (2013), Tōyō Bunko no kirishitan-ban [Printed Christian documents in the Tōyō Bunko] (Tōyō Bunko, 2021), Tōyō Bunko no meihin [Treasures of the Tōyō Bunko], (Yamakawa shuppansha). He has also contributed as an editor in establishing several Japan Industrial Standards (JIS) for the Japanese language, such as JIS X0208:1997 “7-bit and 8-bit double byte coded KANJI sets for information interchange”, JIS X4061:1996 “Collation of Japanese characters”. Masayuki Tsukimoto is Professor emeritus at the Graduate School of Humanities and Sociology, the University of Tokyo. His field of expertise is the study of Japanese language history based on Kunten materials. He is the author of the following publications: Kotengo kenkyū no shōten [The focus of studies of Classical Japanese] (Musashino Shoin, 2010); “Kokunten no kaihen ni tsuite: Fujiwara Yorinaga katen ‘Inmyō Ronsho’ o megutte [On the reorganization of old Kunten: The “Inmyō Ronsho” annotated by Fujiwara Yorinaga]” Kokugo to Kokubungaku 85-8:1–13, 2008. J. Marshall Unger chaired departments from 1988 to 2004 at the University of Hawai‘i, University of Maryland, and The Ohio State University, where he is emeritus professor. He has been a visiting professor at Kōbe University, Tsukuba University, University of Tōkyō, Minpaku, and NINJAL. He is the recipient of a Guggenheim, a Ford, and two Japan Foundation fellowships. Among his many publications are five books on Japanese linguistics (two translated into Japanese), and two on premodern Japanese mathematics (wasan). † Alexander Vovin (1961–2022) obtained his Ph.D. from the Institute of Oriental Studies, Leningrad, in 1987. He held positions in the US, including a long stint at the University of Hawaiʻi, and was from 2014 until his death in April 2022 professor at L’École des hautes études en sciences sociales (EHESS). He published widely on Asian linguistics, making important contributions to the synchronic, diachronic and comparative study of many of the languages in the region. His work on Japanese includes important research articles and monographs and reference works such as the two-volume A descriptive and comparative grammar of Western Old Japanese (GlobalOriental 2005, 2009, 2nd edition Brill 2020), as well as the majestic, but sadly incomplete translation of and commentary on the Man’yōshū. Among the recognitions he received during his lifetime is the NIHU Prize in Japanese Studies which he was awarded in 2015.
Contributors
XLV
Akira Watanabe received his Ph.D. from MIT in 1993. He is now Professor of Linguistics in the department of English at the University of Tokyo. His major research interests lie in the theory of universal grammar and parametric syntax. He has published articles on wh-in-situ, Case, switch reference, negative concord, models of parameter setting, the structure of DP, phi-feature agreement, syntax of adjectives and adpositions, and diachronic changes in the history of Japanese. He is also the author of Case Absorption and Wh-Agreement (Kluwer, 1996). He is currently involved in the project on degree modification. Shingo Yamamoto is a Professor at the School of Arts and Sciences at Tokyo Woman’s Christian University. He completed his Ph.D. at Hiroshima University in 2006. His research interest is the historical study of the style of sentences written by native speakers of Japanese using only kanji. Recently, he has also been interested in the topological study of words used by Japanese speakers to read Chinese classics. He is the author of Heian-Kamakura-jidai ni okeru hyōhaku, ganmon no buntai no kenkyū [Research on the style of hyōhaku and ganmon texts from the Heian and Kamakura periods] (Kyūko Shoin, 2006). Yuko Yanagida is Professor of Linguistics at the University of Tsukuba. She completed her Ph.D. at Cornell University in 1995. Her research focuses on the historical syntax of Japanese. She applies the insights of linguistic typology and generative syntactic theory to the empirical investigation of changes in the case and pronominal systems, changes in word order and related topics in earlier Japanese. Her recent publications include “The origin of dative subjects and psych predicate constructions in Japanese” (Journal of Historical Linguistics 12, 2022), “Two types of Alignment change in nominalizations: Austronesian and Japanese” with Edith Aldridge (Diachronica 38, 2021).
Editors Bjarke Frellesvig is Professor of Japanese Linguistics at the University of Oxford and Fellow of Hertford College, Oxford. He was elected a Fellow of the Royal Danish Academy of Sciences and Letters in 2019. He obtained his Ph.D. from the University of Copenhagen in 1993 and has, after a spell at the University of Oslo, been at Oxford since 1999. The main focus of his research has been the attested history and pre-history of the Japanese language, including phonology, morphology and syntax. He is the author of amongst others A history of the Japanese language (Cambridge University Press, 2010) and is the director and editor-in-chief of the Oxford-NINJAL Corpus of Old Japanese (https://oncoj.ninjal.ac.jp/). Satoshi Kinsui is the director of Osaka Study Center, The Open University of Japan. He is also a member of the Japan Academy. His research interest is in Japanese historical grammar and a study of role languages and character languages. He is the author of Nihongo sonzai hyōgen no rekishi [History of existence expressions in Japanese] (Hituzi Syobo, 2006), Bācharu nihongo: Yakuwarigo no nazo [Virtual Japanse: Enigmas of role languages] (Iwanami Shoten, 2003), and Kore mo Nihongo aru ka: Ijin no kotoba ga umareru toki [Is this also Japanese?: When an alien’s language is born] (Iwanami Shoten, 2014).
Bjarke Frellesvig and Satoshi Kinsui
Introduction 1 Background
The Japanese language has a fairly continuous written history from the beginning of the 8th century AD until the present day. This is conventionally divided into four periods, as also explained in the general introduction by the series editors; the English terms are those generally used in historical linguistics and for Japanese finally established in Martin 1987:1 Old Japanese (OJ) (Japanese: jōdaigo) 700–800 Early Middle Japanese (EMJ) (chūkogo) 800–1200 Late Middle Japanese (LMJ) (chūseigo) 1200–1600 Modern Japanese (ModJ) (kindaigo, gendaigo) 1600– While certainly reflecting or corresponding to significant periods within the language itself, the beginning and endpoints of these periods have conspicuously been fixed in close correlation with major political periods within Japanese history. Other, finer or cross-cutting, periodizations are of course possible, but this one is used in the chapters in this volume, except where authors explicitly use different periodizations that are more relevant or useful to their subject matter. The scholarly study of earlier stages of the Japanese language goes back at least to the philological work of the kokugaku scholars of the Edo period whose work still informs much traditional scholarship in Japan. With the introduction and spread of modern linguistic theories and methodologies from the late 19th century onwards, Japanese scholars also started applying these methods to investigate earlier forms of Japanese in addition to the philological tradition. From the beginning of the 20th century pioneers such as YAMADA Yoshio (1873–1958), HASHIMOTO Shinkichi (1882–1945), HATTORI Shirō (1908–1995), ARISAKA Hideyo (1908–1952), KINDAICHI Haruhiko (1913– 2004), and OHNO Susumu (1919–2008) to name but a few, made great strides in charting and describing the attested history of the Japanese language as well as its reconstructed prehistory. Outside of Japan, early and still influential work was done particularly from the early post-war period by for example Günther Wenck (1916–1992), Roy Andrew Miller (1924–2014), and Samuel Martin (1924–2009).
1 Until Martin (1987), some scholars writing in English used “Late Old Japanese” for EMJ, following the terminology of Miller (1967). The 1993 German translation of Miller (1967) uses the German equivalents of the now current English terminology which is used in this volume. https://doi.org/10.1515/9781614512851-001
2
Bjarke Frellesvig and Satoshi Kinsui
2 Structure of the volume Following the work of these early scholars, large amounts of scholarship have been conducted on the history and prehistory of Japanese, both inside and outside of Japan. In this book we focus on the attested history of Japanese, with only a few chapters addressing reconstruction and pre-history. Today, most basic facts about earlier attested Japanese are fairly well established, including its script, phonology, morphology, and basic syntax, set out and easily accessible in English in Frellesvig’s A history of the Japanese language (2010). In this volume we have solicited chapters from experts covering the main areas which either are treated in less detail, or not at all, in Frellesvig (2010), or where little or no consensus exists. The book consists of 27 chapters and is divided into four overall parts as follows: Prehistory and reconstruction: While this volume mainly is concerned with the attested history of Japanese itself, this short section sets the stage by addressing important aspects of the pre-history, particularly the language family which Japanese belongs to and which also includes the Ryukyuan languages, often referred to as “Japonic” or “Japanese-Ryukyuan”. Chapter 1 “Reconstruction of Japonic and para-Japonic based on external sources” by Alexander Vovin addresses contributions to reconstructing Japonic from sources external to this family. Chapters 2 “Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan” by Thomas Pellard and 3 “Towards the prosodic reconstruction of proto-Japanese-Ryukyuan” by Akiko Matsumori both consider the specific role of Ryukyuan material in the reconstruction and the results they yield. Phonology: The segmental phonology of Japanese and changes over time are in general quite well understood (see section 2 below on transcription). This part of the book addresses a variety of specific issues within the historical phonology of Japanese. In Chapter 4 “Reconstruction of Old Japanese phonology”, Teruhiro Hayata gives a detailed reconstruction and interpretation of the phonological system of OJ. Chapter 5 “Old Japanese writing and phonology” by J. Marshall Unger pursues the use of OJ writing, including kungana, for a further understanding of the phonology of the time. Chapters 6 “Late Middle Japanese phonology, based on Korean sources” by Sven Osterkamp and 7 “Late Middle Japanese phonology as reflected in early Japanese Christian documents” by Masayuki Toyoshima discuss LMJ phonology on the basis of detailed use of Korean and Christian materials, respectively. Chapter 8 “Sino-Japanese” by Marc Hideo Miyake gives an in-depth description of the phonology of what is loosely termed ‘Sino-Japanese. Finally, Chapter 9 “The Ramsey hypothesis” by Elisabeth M. de Boer presents and elaborates on an interpretation of the late EMJ prosodic system which was originally developed by S. Robert Ramsey as an alternative to the traditional interpretation. Grammar: In studies of pre-modern Japanese grammar, case marking, focus constructions (kakari-musubi) and some changes in morphological categories are core areas of recent research and attention. Here, Chapter 10 “Differential argument marking in Old
Introduction
3
Japanese: Morphology, semantics, and syntax” by Yuko Yanagida discusses core case marking in OJ. Chapter 11 “The syntax and morphology of Early Middle Japanese” by Yoshiyuki Takayama addresses predicate structure, modality and copula. Chapter 12 “Late Middle Japanese grammar” by Hirofumi Aoki discusses a number of grammatical changes that took place in the period leading up to LMJ. Chapter 13 “The historical changes in the case marking system of Japanese” by Takashi Nomura considers changing in subject marking over time. Chapters 14 “Voicings of kakari-musubi: Shifting from cleft construction to referential predicate” by Charles Quinn and 15 “Loss of Wh movement” by Akira Watanabe address the focus construction known as kakari-musubi and its changes over time from different perspectives. Chapter 16 “Development of adverbial particles” by Tomohide Kinuhata describes the emergence and development of the so-called adverbial particles from OJ through to ModJ. Chapter 17 “The history of demonstratives” by Tomoko Okazaki explains the historical changes in the use of demonstrative pronouns and adverbs over time. Finally in this section, Chapter 18 “Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’” by Yoshihiko Ikegami analyses subjective construal in expressions involving verbs of visual perception in OJ from a semiotic perspective. Lexicon, materials and kanbun: The chapters in this section deal with lexicon and various important materials, including annotated kanbun (texts written in classical Chinese), which have traditionally been studied and used less than the literary texts in the historical study of Japanese,. In Chapter 19 “The history of basic vocabulary in Japanese”, John R. Bentley provides a systematic description of the basic vocabulary of Japanese, particularly OJ, to a large extent based on the earliest available lexicographical materials. Chapter 20 “The Japanese lexicon as reflected in Christian materials” by Toru Maruyama sets out characteristic features of the vocabulary contained in and described in the Christian materials from the end of LMJ. Chapter 21 “What mokkan (wooden documents) can tell us about ancient Japanese language” by Takashi Inukai describes the nature and significance of mokkan dating from the late 7th to early 9th century which are being excavated in large numbers and which constitute a large and as yet largely untapped wealth of primary sources for OJ language and in particular writing practices. In Chapter 22 “Early Japanese dictionaries”, Shoju Ikeda gives a detailed overview of the history of dictionaries of Japanese in Japan. The final five chapters in the book provide a comprehensive view of the importance for the study of the history of Japanese language and writing of interaction with and annotation of kanbun, texts written in Classical Chinese. Chapters 23 “Kunten texts of Buddhist provenance (butten 仏典): Their characteristics and actuality” by Masayuki Tsukimoto and 24 “Kunten texts of secular Chinese origin (kanseki 漢籍)” by Teiji Kosukegawa address the two main types of materials with kunten, annotations used as aids to kanbun-kundoku (translation, or rendition, of Chinese text into Japanese). In Chapter 25, “Japanized written Chinese: Its features and contribution to the history of the Japanese language”, Shingo Yamamoto explores what may be learnt about Japanese from
4
Bjarke Frellesvig and Satoshi Kinsui
Chinese written by native Japanese speakers and exhibiting some influence from Japanese. Finally, Chapters 26 “Early modern kanbun and kanbun-kundoku” by Fumitoshi Saito and 27 “The influence of kanbun-kundoku vocabulary on the Japanese language” by Valerio Luigi Alberizzi detail the lasting influence on the Japanese language of the practice of kanbun-kundoku.
3 Transcription The basic principle used in transcribing older Japanese words and forms used in this volume is to write them in a time-appropriate phonemic transcription. This may sound obvious, but it is in fact a practice not universally adopted within Japanese historical linguistics. Instead, much scholarship uses a transliterated form of the normative historical kana spelling (rekishiteki kanazukai) which may have the advantage of making forms more readily recognizable to readers of Japanese without knowledge of the historical phonology, but which also, first, is unrealistic in that it posits forms that have never been used in Japanese speech, and second, denies Japanese a phonological diachrony and its texts a realistic vocalization. In this volume, words will be transcribed differently depending on the age of the text, or the period, they are cited from. For example, ModJ mae ‘front’ is in the normative historical kana spelling written まへ and therefore often for pre-modern texts transliterated mahe (or sometimes mafe), neither of which, however, have ever been phonemic forms of that word; here it will be transcribed phonemically as mapye (pre-800), mape (800 – 950), mawe (950 – 1100) or mae (post-1100), depending on the approximate date of the citation. We follow the phonemic interpretation of the Old Japanese sound system reflected in what is often called the Frellesvig-Whitman system of transcription. Thus, the earliest OJ had 88 distinct syllables, transcribed as shown in (1). Other transcription systems for OJ have also been used; see for reference (2) which contrasts some of these. Note, however, that both Hayata in Chapter 5 and Unger in Chapter 6 arrive at different phonological analyses and therefore use different transcription conventions for OJ, and further that Vovin in Chapter 1 uses a somewhat different system of transcribing OJ, including the so-called modified Mathias-Miller notation of the kô-otsu syllable distinctions, see (2), and explicitation of the feature of pre-nasalisation for the lax stops which Vovin writes as mb, nd, ŋg, nz, as opposed to the rest of the volume where they are written b, d, g, z and the pre-nasalisation interpreted as being the result of phonetic realization rules (i.e., /b, d, g, z/ => [mb, nd, ŋg, nz]).
5
Introduction
(1) Distinct syllables in the earliest OJ .a ka ga sa za ta da .i ki gi si zi ti di kwi gwi .u ku gu su zu tu du .e ke ge se ze te de kye gye .o ko go so zo to do kwo gwo swo zwo two dwo
na ni nu ne no nwo
pa pi pwi pu pe pye po
ba bi bwi bu be bye bo
ma mi mwi mu me mye mo mwo
ya
ra ri
wa wi
yu ye
ru re
we
yo ro wo ywo rwo
(2) Examples of different transcription systems for OJ for the kô-otsu syllable distinctions Gloss
FrellesvigWhitman ‘sun’ pi ‘fire’ pwi ‘blood’ ti ‘woman’ mye ‘eye’ me ‘hand’ te ‘child’ kwo ‘this’ ko ‘ear (of rice)’ po
Neutral index notation pi1 pi2 ti me1 me2 te ko1 ko2 po
Yale Modified Mathias-Miller pyi pî piy pï ti ti mye mê mey më te te kwo kô ko kö po po
Ohno pi pï ti me më te ko kö po
The texts from the OJ period are all written exclusively in Chinese characters, used in a mixture of logographic and phonographic writing, the proportion of which differed widely between different texts. In the citation of examples from OJ, the important distinction between these two modes of writing is shown by use of different types: phonographically written text portions are set in italics, whereas logographically written text is set in normal (or, roman) type. For example, in (3), the first three lines from poem 3582 in book 15 of the Man’yōshū (MYS), the words opobune and kimi are written logographically by the characters 大船 and 君, respectively, but all other words and particles are written phonographically, such that the characters 乎安流美爾伊太之伊 麻須 all are used phonographically (i.e., as man’yōgana). (3)
大船乎 安流美爾 伊太之 伊麻須 君 opo-bune wo ar-umi ni idasi imasu kimi big-boat acc rough-sea dat put.out go.resp my.lord ‘you, my lord, who will take your big boat out on the rough sea’ (MYS 15.3582)
6
Bjarke Frellesvig and Satoshi Kinsui
In this example here, the original script is included for illustration, but in general original script is only included where it is directly relevant to subject matter or argument or is itself under discussion. The main regular phonemic sound changes that took place since OJ, that is, those which affected the syllables in (1), are few and easy to summarize. See further Frellesvig (2010: 414–415 for a summary and p. 176 for an easy guide to translating the historical kana spelling into a time-appropriate phonemic transcription). (4)
Summary of main regular phonemic sound changes since OJ Loss of /y/ before /e/ Cye > Ce (pre-800) .ye > .e (pre-950) Loss of /w/ before /i, e, o/ and lenition of /p/ Cwi > Ci, Cwe > Ce (pre-800) Cwo > Co (pre-950) -pu- > -u- (pre-1000) -p- > -w- elsewhere (-pi > -wi, -pe > -we, -pa > -wa, -po > -wo; pre-1000) -.wo > -.o (around 1000) -.wi > -.i, -.we > .-e (around 1100 ) (word non-initial -.wi, -.we) .wi > .i, .we > .e (around 1300) (word-initial .wi, .we) p- > f- (?around 1300) > h- (?around 1700)
Of the vowel sequences which arose from some of the changes in (3), /Vi, Vu/ resulted in long syllables with diphthongs, of which /Vu/ were monophthongized perhaps as early as the late 14th century to give long vowels (e.g. ‘today’ OJ /kyepu/ > EMJ /kepu/ > /keu/ > LMJ /kyoo/), whereas /Ve, Vo/ remained dissyllabic /V.e, V.o/, manifested with an automatic onset glide at the syllable boundary [Vje], [Vwo] (e.g. ‘front’ OJ /mapye/ > EMJ /mape/ > /mawe/ > /ma.e/ [maje]; this non-phonemic, automatic onset glide was only lost quite recently (19th, 20th century). The dating of the phonemic sound changes is largely uncontroversial as they mainly led to mergers which were reflected in changes in writing practice (so that for example ‘millet’, OJ apa, and ‘foam’, OJ awa, after the emergence of the kana letters, would be written differently in kana as あは and あわ until after the change of /-p-/ to /-w-/ when both became awa and could be written the same), except for the lenition of initial /p/ > /f/ > /h/ which found no expression in writing but only resulted in novel association of sound values with the kana letters はひふへほ in initial position and where dating is therefore more tentative, as for example ‘boat’ could be written ふね at any time regardless of whether the phonemic shape at that time was pune (?pre-1300), fune (?pre-1700) or hune (?post-1700).
Introduction
7
Acknowledgements: We owe a large debt of gratitude to several colleagues who have provided valuable input and help with this book. John Whitman was involved in an early stage of the book, both in planning its design and scope and also in taking part in early editorial work; thank you, John. Emi Sakamoto provided invaluable help with content, format and citation of references and bibliographies in all the chapters. George Wollaston did a thorough, detailed final check through all chapters, catching a large number of inconsistencies and mistakes which had been overlooked by everybody else. John Haig translated several chapters originally written in Japanese, as well as early copy-editing and proof-reading. Stephen Horn stepped in at a crucial time to help edit Chapter 12 by Aoki and John A. Bundschuh did the same for Chapter 26 by Saito. Marc Miyake proofread Chapter 1 by Vovin with the dedication, care and philological and linguistic sensitivity of a true former student of Vovin’s. Marc also contributed immensely and essentially to checking and correcting the setting and fonts used for East Asian script in the volume. Finally, the series editors, Matt Shibatani and Taro Kageyama, guided and helped us through the process and gave much valuable input. We are very grateful for all the help we have received along the way; needless to say, all remaining editorial mistakes or infelicities are our responsibility.
References Frellesvig, Bjarke. 2010. A history of the Japanese language. Cambridge: Cambridge University Press. Martin, Samuel E. 1987. The Japanese language through time. New Haven: Yale University Press. Miller, Roy Andrew. 1967. The Japanese language. Chicago: University of Chicago Press. Miller, Roy Andrew. 1993. Die japanische Sprache: Geschichte und Struktur. Translated from Miller 1967 by Jürgen Stalph et al. München: Iudicium.
Part I: Prehistory and reconstruction
Alexander Vovin
1 Reconstruction of Japonic and para-Japonic based on external sources 1 Prehistorical ethnolinguistic setting and sources The Japonic language family can be roughly divided into two groups: peninsular Japonic and insular Japonic. Since all peninsular Japonic languages (spoken in antiquity in the south and in the center of the Korean peninsula) are attested only very fragmentarily, the establishment of a taxonomy based on linguistic criteria does not seem possible; therefore, the division into peninsular and insular branches is rather geographic and intuitive (insular languages must have come to Japan from or via the Korean peninsula). There are the following peninsular Japonic languages, which are probably better called para-Japonic: 1) the pseudo-Koguryŏ language, called ‘pseudo’ in order to differentiate it from the real Koguryŏ language related to Korean, known almost exclusively from place names found in the Hangang river basin. 2) the Paekche Japonic language, poorly known, as only a few Japonic glosses are extant. The evidence is mostly based on ZS, LS, and Japanese NSK. 3) the Silla Japonic language, known mostly from place names in SGSG. 4) the Karak (Mimana) Japonic language, poorly known only on the basis of glosses in SGSG. The insular Japonic languages are much better known because all of them have texts. They comprise: 1) Japanese, consisting of the following extant varieties: a. Central Japanese, including Western Old Japanese and closely related but not completely identical Middle Japanese, from which all later historical and modern varieties of Japanese are derived (except Hachijō). b. Eastern Japanese, represented by the Eastern Old Japanese dialect continuum with the Hachijō language (possibly an offshoot of the Suruga dialect of Eastern Old Japanese) as the only modern survivor. c. Kyushu Japanese, represented just by one short poem in Kyushu Old Japanese; no modern survivors. 2) Ryukyuan consisting of two subbranches: a. Northern Ryukyuan comprising two languages: i. Amami Ryukyuan. ii. Okinawan. The Old Ryukyuan language (attested from the second half of the fifteenth century) appears to be an ancestor of modern South-Central Okinawan dialects). https://doi.org/10.1515/9781614512851-002
12
Alexander Vovin
b.
Southern Ryukyuan comprising three languages: i. Miyako Ryukyuan. ii. Yaeyama Ryukyuan (there is a possibility that there is more than one language here). iii. Yonaguni Ryukyuan. Some scholars believe that it should be included in Yaeyama Ryukyuan, but a number of reflexes are strikingly different, and no mutual comprehension is possible.
The position of Tammura (Chejudo) Japonic remains unclear because so far we have only two reliable glosses, although it seems more likely that it belongs to the insular branch. It is possible that the word kam- ‘deity’ speaks in favor of an insular affiliation, since it is not attested in peninsular languages, but given the paucity of the materials on the latter, this might be a case of an absence of evidence rather than a proof of absence. External sources for the reconstruction of different varieties of Japonic and paraJaponic can be divided into three groups: (1) Japonic loans in neighboring languages (2) Foreign transcriptions of different varieties of Japonic (3) Materials on para-Japonic languages in Chinese, Japanese, and Korean sources While (1) and (2) are applicable to the reconstruction of Japonic proper, (3) can be used only for the reconstruction of para-Japonic languages related to Japonic. It cannot be used for the reconstruction of the Japonic proper because a reconstruction ‘from above’ would be a methodological violation. There are two serious limitations to any reconstruction based on external sources. First and foremost, it is bound to remain fragmentary, because both loanwords and foreign sources can present only partial evidence that is relevant to the reconstruction of selected features but can never be used for the reconstruction of the system as a whole. Second, we should keep in mind that the phonological features presented both in loans and foreign transcriptions are filtered through foreign systems. Thus, by definition, any analysis of loans and foreign transcriptions can play only an auxiliary role in the reconstruction.
2 Japonic loans in neighboring languages Being located on the eastern fringe of the Eurasian continent, Japonic from the time when it was already in situ on the Japanese islands starting from the Yayoi period (3rd or maybe even the 7th century BC) did not have too many neighboring languages. As direct contact with Chinese occurred no earlier than the 3rd century AD, and did not become more or less permanent until the 7th century AD, these languages were limited to Korean
1 Reconstruction of Japonic and para-Japonic based on external sources
13
on the Asian mainland and various pre-Japonic local languages in the Japanese islands.1 Given the fact that the Jōmon culture lasted in the Japanese islands for many millennia, it seems natural to suppose that there were many languages or even language families unrelated to Japonic that existed in the islands (Whitman, p.c.). However, the actual surviving evidence is limited to only two languages: the Ainu language that is historically attested from Tōhoku to Hizen province (肥前國) in north-western Kyushu (Vovin 2009), and not just confined to northern Tōhoku as the traditional point of view holds, and the Hayato language attested in the Ōsumi province (大隅國) in south-western Kyushu. It is also believed that there was another non-Japonic language, Kumaso in Southern Kyushu,2 but the question is moot because not a single word of the Kumaso language is attested. Other non-Japonic languages disappeared completely or almost completely, probably surviving in place names that cannot be identified either as Japonic or Ainu. Although this situation is strange, I trust that the wide spread of the Ainu language in the Japanese islands may suggest that once upon a time Ainu obliterated other indigenous languages in exactly the same way as Japanese later obliterated Ainu.
2.1 A Japanese loan in the Hayato language Only two words survive from the Hayato language, and both are recorded in the fragments of Ōsumi Fudoki (大隅國風土記逸文) (Akimoto 1958: 526). They are given in Table 1 below with possible identifications. Table 1: Hayato words.3 gloss
word
spelling
WOJ/EOJ
Ryukyuan
comb shoal
kusera pisi
久西良 必志
WOJ kusi EOJ pînzi3
✶
kusi pisi
✶
Ainu – pís ‘shore’
The first of these words probably means ‘comb’: although the meaning ‘head hair’ was suggested as well, the gloss 髪梳 ‘head.hair comb’ in Ōsumi Fudoki strongly suggests the first meaning. The word probably was not borrowed from WOJ of the Kinki region, but from local Kyushu Old Japanese, as the diminutive suffix -ra attached to WOJ kusi ‘comb’ indicates: the Kyushu Old Japanese word appears in MYS 3.278, composed by Isikapa-no asomi Kimiko, but in reference to Sika bay on Sika island, which was located
1 Once Ryukyuan moved into the Ryukyuan Islands around the 9th century AD, and reached the Sakishima Islands in the southern part of the Ryukyuan archipelago, the Austronesian languages of Taiwan also became ‘neighbors’, at least technically. 2 There is a strong possibility that Kumaso and Hayato represent the same ethnic group (Hudson 1999: 194). 3 Attested in MYS 14.3448.
14
Alexander Vovin
in present-day Fukuoka bay. Unfortunately it is spelled logographically as 髪梳, but the meter of the poem indicates that this word must have been trysillabic. Japanese scholars normally read 久西良 as kusira (Akimoto 1958: 526), (Omodaka et al. 1967: 258), (Murayama 1975: 252), but the problem is that the phonogram 西 is never used for the syllable /si/ in man’yōgana, only for the syllable /se/. Therefore, I read 久西良 as kusera. This can potentially indicate that WOJ kusi ‘comb’ underwent vowel raising ✶e > i, and therefore pJR ✶kuse should be tentatively reconstructed. In spite of superficial reliability, no counterevidence for ✶kusi rather than ✶kuse is presented by Ryukyuan cognates: Iejima kuśi:, Hirara fusï, Ikema futsï, Ishigaki Fusï, Hateruma Futsï (Hirayama 1966: 334), with Sakishima data with the vowel -ï indicating ✶i and not ✶e. However, the reconstruction of the pJR ✶e : ✶i distinction in proto-Ryukyuan after coronals is not viable (see also below on ‘island, territory’ in the section of Japonic loanwords in Korean), so the Hayato evidence may be our only clue for reconstructing ✶kuse rather than ✶kusi. The second word, pisi ‘shoal’ is probably a native Hayato word, although its relationship with similar forms in Ryukyuan4 and Ainu opens a host of interesting problems, unless it is accidental. However, these problems fall outside the scope of the present chapter.
2.2 Japanese loans in the Ainu language Ainu has multiple loans from Japanese, but it seems that there is no detailed study of them except a concise article of mine published in Russian more than a quarter of a century ago (Vovin 1990). There are different chronological layers of these loanwords, ranging from very recent ones from the twentieth century to very old ones going back to Old Japanese and pre-Old Japanese. I am not aware of any Japanese loans in Ainu that have features of Old Japanese not motivated by internal reconstruction within Old Japanese. Since all Ainu language groups that have survived to the point that they were at least somehow documented are found to the north of the Yamato region, it probably would not be an exaggeration to say that the major donors on the Japanese side after the pre-Old Japanese stage and preceding massive Japanese emigration to Hokkaidō starting from the Meiji period can be identified as Eastern Old Japanese (EOJ), as well as Tōhoku dialects in a later period. For the identification of different strata of Japanese loanwords in Ainu, it is useful to keep in mind not only specific phonological features, which, if they are available, can certainly pin down the relative depth of a loan, but also geographical distribution. Thus, for example, Japanese loans found exclusively in the Ainu dialects along the southern shore of Hokkaidō are more than likely recent loans from the Tōhoku dialects, and more
4 Thorpe (1983: 327) suggested that pR ✶pisi ‘shoal’ consists of ✶se ‘shoal’ and ✶pi- ‘to dry’. If so, pisi might be a Ryukyuan loan in Hayato. This might be tempting, especially from the point of view that it would indicate the presence of Ryukyuans in southern Kyushu before their southward migration, but cf. OJ pï- ‘to dry’, which makes Thorpe’s etymology unlikely. In any case, no hasty conclusions should be based on just one word.
1 Reconstruction of Japonic and para-Japonic based on external sources
15
specifically the Aomori dialect. Ditto for loanwords found in just one modern dialect of Hokkaidō, e.g. Obihiro huton ‘futon’ < ModJ huton ‘id.’ On the other hand, the loans that have found their way to more remote areas of north-eastern Hokkaidō and to even more remote Sakhalin and the Kuriles,5 are much more likely to be older loanwords, unless they exhibit some peculiar phonological features indicating a later provenance. In this chapter I survey only the oldest loans and in addition only those that have any bearing on the reconstruction of Japonic. I will start in Table 2 with loanwords that are very suggestive for the reconstruction of pre-Old Japanese or Eastern Old Japanese consonantism. The following system is adopted for rendering pitch accent in Hokkaidō dialects: the diacritic ʹ (accent aigu) over a vowel is used to indicate HIGH pitch. The LOW pitch is not marked, because in modern dialects it is pretty much automatic before or after HIGH pitch, unless it appears on the possessive form of a noun, in which case I mark it with the diacritic ˋ (accent grave). I do not mark pitch accent on monosyllabic nouns if there is no information on the inherent pitch accent, which is revealed only in combination with the following possessive suffix. Among Hokkaidō dialects, Bihoro has lost its pitch accent system, but I do not cite Bihoro forms separately below. Possessive forms, if any, are cited as -CX, with C indicating the last consonant appearing in a possessive form and X indicating the rest of the possessive form. Table 2: Reflexes of the [pre-]OJ bilabial voiceless stop /p/ in loanwords in Ainu.678 gloss
Hokkaidō Ainu
Kuril Ainu
Sakhalin Ainu
EOJ6
WOJ
bone skin, fir, bark custom, manner bag bowl lid chopsticks hoe salt barnyard millet cave
pone kapa pur-i pukurô pati puta pasi kupa sipo pîye pora
5 The Japanese contact with Sakhalin and Kuril Ainu before the end of the Russo-Japanese War (1904– 1905 AD) was limited to occasional travel and trade. One should also keep in mind that materials on Kuril Ainu are very fragmentary: they are limited to several Russian and Polish sources from the eighteenth to nineteenth centuries and one Japanese source from the early twentieth century. For details on these sources, see Vovin (1993: 3–7). 6 The lack of attestations in EOJ is due to the very limited size of the EOJ corpus. 7 Sakhalin Ainu -h in Auslaut and in a preconsonantal position corresponds to Hokkaidō Ainu -p, -t, and -k, and in the Auslaut position after /i/ this /h/ can materialize as [s], e.g.: Hokkaidō Ainu sik ‘eye’ vs. Sakhalin Ainu sis ‘id.’ 8 pukurú is Yakumo accentuation; pukúru is found in the rest of Hokkaidō.
16
Alexander Vovin
On the basis of the data presented in the table above, I trust it is quite safe to say that pre-OJ or EOJ loans in Ainu present undeniable evidence for the initial ✶p- either in EOJ or pre-OJ. The situation with the Inlaut of the same phoneme is much trickier. The prevalent point of view today is that OJ intervocalic -p- was phonetically [-b-] or [-β-] (Frellesvig 2011: 35), going back to Wenck (1959). I personally find this point of view untenable for a variety of reasons (Vovin 2005a: 37–38), but I will limit myself here only to one reason directly connected to the present chapter. Namely, if OJ -p- was really [-b-] or [-β-], why would it be borrowed by Ainu as -pp- instead of -p- [-b-/-β-]?9 Unfortunately, there is only one example above that demonstrates this point uncontroversially: OJ sipo ‘salt’ would be borrowed as Ainu ✶sipo and not sippo, if [pre-]OJ intervocalic -p- was indeed phonetically [-b-] or [-β-]. Although there are no other uncontroversial examples with Japonic intervocalic labial voiceless consonants reflected as double consonants in Ainu, there is supporting evidence from dentals. Table 3: Reflexes of the [pre-]OJ bilabial voiceless stop /t/ in loanwords in Ainu.10 gloss
As we can see from Table 3 above, in most Japanese loanwords attested in both Hokkaidō and Sakhalin branches of the Ainu language family, Japonic intervocalic -t- is reflected as pA ✶-tt-. Although the case of puta ‘lid’ above does not follow this pattern, it is reliably demonstrated in three other cases. Even though there is a variation between ✶ -tt- and -✶t- in tútci ~ túci ‘hammer’ and mútci ~ múci ‘whip’, the accent type HL in the cases of both túci ‘hammer’ and múci ‘whip’ demonstrates that they are secondary 9 An anonymous reviewer pointed out that “Ainu itself, at least Hokkaidō Ainu, has intervocalic stop voicing (K. Kindaichi and Chiri 1936: 8, Tamura 1998: 40), so a plain /p/ /t/ /k/ is the expected Ainu reflex for an earlier Japanese medial consonant even if the latter did undergo intersonorant voicing. The sippo outcome for ‘salt’ in Ainu may just show that the J variety that provided the source for the Ainu form blocked intersonorant voicing because of devoicing of the high vowel in the first syllable. Shōgakkan (1974 9: 365) gives such a form, シポ [sipo] for Yamagata”. However, there are several problems with the anonymous reviewer’s statement. First, Yamagata sipo is an irregular form that violates regular correspondences. A form like Iwate sɯo would be a regular Tōhoku reflex of pJR ✶sipo, therefore Yamagata sipo is in all likelihood a back loan from Ainu. Second, devoicing of high vowels in Tōhoku must be a very recent phenomenon, because the Russian materials on eighteenth century Tōhoku do not reflect it. Third, there is no high vowel devoicing in Ainu. 10 The palatalization ✶ti > ci in Ainu is not suggestive of any late Japanese influence of the Muromachi period, since it can be traced all the way back to proto-Ainu, with the lack of distinction between /ci/ and /ti/ recovered only via internal reconstruction (Vovin 1993: 12–16).
1 Reconstruction of Japonic and para-Japonic based on external sources
17
forms that developed from pA ✶tútci ‘hammer’ and ✶mútci ‘whip’ with original closed syllables. Similar to the case of [pre-]OJ sipo ‘salt’ > Ainu sippo ‘id.’, it is incomprehensible why [pre-]OJ -t- if it was indeed voiced [-d-] or lax [-ð-] would be borrowed as pA ✶ -tt- rather than Ainu lax -t-, phonetically either voiced [-d-] or lax [-ð-]. However, the same tendency of borrowing [pre-]OJ intervocalic voiceless as Ainu double consonants cannot be successfully established with velars, as presented in Table 4 below: Table 4: Reflexes of the [pre-]OJ bilabial voiceless stop /k/ in loanwords in Ainu. gloss
However, example #5 ‘cow’ was certainly borrowed from Tōhoku be[:]ko [be[:]go] ‘cow’, and probably more specifically from Aomori, because in Aomori Japanese intervocalic voiceless obstruents became phonetically voiced, and the same picture could have been observed for Ainu until very recent times, when phonetically intervocalically voiced stops started to be pronounced voiceless under the influence of Japanese. Another important evidence that Ainu provides for the reconstruction of OJ is its corroboration of the prenasalized nature of OJ voiced intervocalic stops that go back to N[+nasal] + C[+obstruent] voiceless consonants (Table 5): Table 5: Reflexes of [pre-]OJ prenasalized voiced obstruents in Ainu. gloss
Ainu kánpi ~ kampe ~ kanpi ‘paper’ was apparently borrowed from some unattested OJ dialect that preserved the original pJR form ✶kambî, which shifted to kamî in Central Japanese with a -mb- > -m- shift. The archaic nature of the ✶kambî form is further supported by pR ✶kambî.11 11 Thorpe’s (1983) reconstruction for ✶mb is ✶b, but see below on the reconstruction of prenasalized obstruents in the section on reconstruction of Old Okinawan on the basis of external sources.
18
Alexander Vovin
Regarding the reconstruction of pre-OJ vocalism, in three cases in Table 6, Japanese loanwords in Ainu present important evidence for the reconstruction of OJ kō-otsu distinctions and their sources, as well as for vowel raising. Table 6: Ainu evidence for OJ kō-otsu distinctions and their sources. gloss
Hokkaidō Ainu
Kuril Ainu
Sakhalin Ainu
EOJ
WOJ
deity winnow chopsticks
kamúy múy pasúy
kamuy – pašuy
kamuy – –
kamï ~ kamî –
kamï ~ kamumï pasi
✶
In the first example, Ainu confirms that the pJR form was ✶kamuy.12 It is not absolutely clear what the directionality of the loan is, but since the semantics of the Ainu word is broader than in Japanese (‘bear’ in addition to ‘deity’), it might be an Ainu loan in Japanese. On the other hand, the presence of ✶kam- ‘deity’ in the Tammura (Chejudo) place name ✶kam-SAN (神山) ‘deity mountain’ (Vovin 2010: 24), may be indicative of its wider distribution in para-Japonic, and this makes a Japonic origin more likely. Although the word for ‘winnow’ is not attested phonographically in WOJ, we know that it must have been ✶mï with otsu-rui /ï/, because the kun-yomi (訓読み) of the character 箕 ‘winnow’ is used as a phonogram for the syllable /mï/ with otsu-rui /ï/ (Omodaka et al. 1967: 901). However, Ainu is the only source that disambiguates between three possible sources of WOJ mï ‘winnow’: pJR ✶muy, ✶moy, and ✶məy, clearly demonstrating that it was ✶muy. Since WOJ does not have a contrast between the kō and otsu vowels î : ï and ê : ë after coronals, the Ainu evidence for pJR ✶pasuy ‘chopsticks’ is of great importance. Not only does it demonstrate that the [pre-]OJ form was ✶pasï with otsu-rui /ï/, and not ✶✶ pasî with kō-rui /î/, it again disambiguates the pJR source of this otsu-rui /ï/, clearly demonstrating that it was ✶uy, and not ✶oy or ✶əy. The Japanese loans in Ainu in Table 7 also offer some interesting evidence for vowel raising or the lack of it. Table 7: Mid vowel raising or the lack of it in Japanese loans in Ainu. gloss
Hokkaidō Ainu
Kuril Ainu
Sakhalin Ainu
EOJ
WOJ
bag gold
pukurú ~ pukúru kónkani ~ kónkane
pukuru kongane ~ kongkan
pukuru konkaani
– –
pukurô kuŋgane MJ koŋgane
12 I follow Whitman’s 1985 proposal that the root of the Japonic word is ✶kamuy, with final ✶-y being lost in the compounding form kamu-, not the more widespread theory that the word is bimorphemic ✶ kamu-i, because the function of this ✶-i is completely unclear, and therefore it is nothing but an unaccounted segment.
1 Reconstruction of Japonic and para-Japonic based on external sources
19
Table 7 (continued) gloss
Hokkaidō Ainu
Kuril Ainu
Sakhalin Ainu
EOJ
WOJ
bone sign, mark barrel sweet potato
poné sirósi ~ sirúsi óntaro ~ hontaro imó ~ emó
poni – – –
poni sirosi ontoro imo
– – – –
pone sirusi MJ taru MJ imo
The first three examples demonstrate the typical EOJ raising of final mid vowels ✶o > u and ✶e > i (although EOJ examples of these words are not attested), but the second example also demonstrates the retention of the original pJR mid vowel ✶o (pJR ✶koy ‘yellow’), found in MJ but not in WOJ. The last three examples demonstrate the lack of raising of the pJR mid vowels ✶o and ✶e in Japanese loans in Ainu, suggesting the pJR reconstructions of ✶sirosi ‘sign’, ✶ taro ‘barrel’, and ✶emo ‘sweet potato’. Finally, the Japanese loanwords in Ainu in Table 8 offer overwhelming support for the traditional reconstruction of pJ pitch accent (H. Kindaichi 1971) that takes Kyoto-type accent to be essentially identical with pJ, and not for the innovative proposal of Tokugawa (1972) and Ramsey (1979, 1980) that insists that Tokyo-type accent is more archaic and is essentially identical with pJ. Table 8: Pitch accent in Japanese loanwords in Ainu.1314151617 gloss
proto-Japanese
Tokyo
Kyoto
proto-Ainu
metal paper cup bone skin board hammer saw bag ladle medicine
13 Adapted with some revisions from Vovin (1997: 118–119). 14 One must keep in mind that any initial mora in Tokyo that is not accented acquires an automatic low pitch. 15 In both Tokyo and Kyoto accent classes 2.2 / 2.3 and 3.3 / 3.4 have merged as 2.2 and 3.3, so the resemblance of pA accent to Tokyo rather than to Kyoto in 2.3 and 3.4 is superfluous and secondary. 16 Irregular accent in Kyoto; ✶HL should be expected. 17 Irregular accent in Kyoto; ✶LHH should be expected.
20
Alexander Vovin
2.3 Japonic loans in the Korean language Unfortunately, there are only a few Japonic loans in Korean, compared with the volume of Korean loanwords in Japonic, and even fewer of those tell us anything at all about the reconstruction of Japonic: as a matter of fact, only two are of relevance here (Table 9). Both seem to be relevant for the reconstruction of pre-para-Japonic, but not of insular Japonic. Table 9: Para-Japonic loanwords in Korean. gloss
MK
Old Korean
EOJ
WOJ
pR
island, territory bear
:syǝm :kom
PC sema PC kuma, komu
sima –
sima kuma
✶ ✶
sima kuma
pJR ✶ ✶
sima kuma
para-Japonic sema koma
✶ ✶
These two words have been traditionally claimed as Koreo-Japonic cognates (Martin 1966, #10, #117), (Whitman 1985: 138, 234); and I have previously claimed that they represent loans from Korean into Japonic (Vovin 2005b: 127–128, 130–131), (Vovin 2010: 143, 183–184). I would like now to revise my opinion. There is no evidence for reconstructing with any certainty ✶sema or ✶sima and ✶koma or ✶kuma for insular Japonic. On the basis of insular Japonic data, one may be tempted to reconstruct ✶sima, as both Amami dialects and Sakishima dialects of Ryukyuan indicate ✶sima: Yuwan śimaa, Koniya śima, Higasi Nakasone, Yonaha, Tonogusuku sïma, Hateruma sï̥ m̥a (Uchima and Arakaki 2000: 403) according to the correspondence of Amami /i/ to Sakishima /ï/, but the problem is that there is simply no difference between the reflexes of ✶si and ✶se in Ryukyuan. The meaning of pJR ✶sima is broader than of PK ✶syǝma, as it can mean not only ‘island’, but also ‘territory, land’, see, e.g. Yamatö sima (MYS 15.3648), which can only be understood as ‘land of Yamato’, not the ‘island of Yamatö’, because the Yamatö province is inland and is not an island. Consequently, it is more likely that PK ✶syema is a loanword from peninsular Japonic. Although there are different reflexes for pJR ✶ko and ✶ku in Ryukyuan, the crucial evidence from Sakishima dialects is lacking; therefore again it is not possible to establish on a purely internal basis whether insular Japonic had ✶koma or ✶kuma. Thus, the evidence from MK and OK forms is the only basis for the conclusion that in para-Japonic the word was ✶koma. There is no compelling evidence for the directionality of borrowing, but given the fact that it exhibits a system of correspondences quite parallel with that of para-Japonic ✶sema and PK ✶syema, it seems safe to assume that both words reflect the same strata, and therefore that PK ✶koma was borrowed from peninsular Japonic.
1 Reconstruction of Japonic and para-Japonic based on external sources
21
3 Foreign transcriptions of different varieties of Japonic There are very few foreign materials for Old or Middle Japanese prior to the fifteenth century. The oldest and the most important of them is the third century AD ‘Account of Wa people in the Chronicle of Wei’ (魏志倭人傳, Chin. Wei zhi Wo ren zhuan, Jpn. Gishi Wajin den).
3.1 Chinese transcriptions of Wei zhi Wo ren zhuan Voluminous research has been done on this source, both from the historical and the linguistic points of view. It is impossible to provide a complete bibliography of these studies here, so I list only the most important ones: Osada (1979, 2010), Ishihara (1985), Mizuno (1989), Mabuchi (1999), Miyake (2003), Bentley (2008). There are two kinds of problems that greatly diminish the value of Wei zhi Wo ren zhuan as an external source for the reconstruction of proto-Japanese and proto-Japonic. First is the nature of the material itself: all that is transcribed in this text are titles, personal names, and place names, which constitute the poorest possible evidence unless we know exactly what they mean. In the absence of exact semantic glosses in the Wei zhi Wo ren zhuan, all identifications with words that constitute these proper nouns must be treated as tentative. The second problem is that, with the exception of Miyake (2003) and Bentley (2008), who used Late Han Chinese readings, all other researchers rely on anachronistic readings of these characters. The final problem is that even when character readings from the appropriate period (Late Han) are used, the idiosyncrasies of Chinese transcriptions caused by the existing gaps in Late Han Chinese syllabic structure are frequently not addressed. Below in Table 10, I provide a few words recorded in the Wei zhi Wo ren zhuan that can be more or less positively identified with the following commentary and that can shed light on the problem of raising of pJR ✶e and ✶o to OJ i and u. Table 10: Possible Japonic words containing ✶e and ✶o in Wei zhi Wo ren zhuan. possible meaning
If these identifications are correct, we can find in the table above some very common basic words. However, in most cases their transcription does present a number of problems. The first of these words is ‘the sun’, found in 卑彌呼 ✶pie-mie-hɔ and 卑狗 ✶piekoʔ, which probably mean ‘sun child’ and ‘sun priestess’ respectively.18 Both Miyake (2003: 115) and Bentley (2008: 17) correctly point out that internal Japonic evidence based on the pR ✶pi ‘sun’ indicates that this word should be reconstructed with high front vowel ✶i rather than with mid front vowel ✶e. I am inclined to agree with Miyake (2003: 115) that the possible reasons for the WZ transcription of ✶i as ✶ie might have to do something either with mishearing, or with a deliberate choice of a character with a pejorative meaning, rather than with Bentley’s (2008: 18) proposal that ✶pe is a loan from Paekche ✶pe ‘west’. There is no historical evidence for contact between Wa and Paekche in the early third century AD, because most likely Paekche as a state did not yet exist at this time. There is, however, yet another possibility: as far as I can tell, in LHC there was no syllable ✶pi, but only ✶piʔ. The absence of /pi/ constitutes a gap in the syllabic system, and it is quite possible that Chinese transcribers sacrificed accuracy in vocalism due to this gap. We face quite a similar problem in the case of the suggested pJR ✶meme for ‘ear’ attested in 彌彌 ✶mie-mie and 彌彌那利 ✶mie-mie-nɑ-lis. There seems to be another gap here, as according to the Baxter-Sagart reconstruction the only words in mjij, which would reflect an earlier ✶mij (or ✶mi) are 寐 MC mjijH < OC ✶mi[t]-s ‘sleep’, which probably still had an ✶-s at least in LHC (the ✶[t] in this case means it could be ✶t, ✶p, ✶j, or maybe even ✶r), and 媢 which would have the same reconstruction except that it looks like an anomaly; it also has the EMC reading mawH < OC ✶mˤuk-s, which is more 18 Pace Bentley (2008: 19), who interprets ✶pie-mie-hɔ as OJ pîmê ‘princess, daughter of the sun’ + unidentified -hɔ.
1 Reconstruction of Japonic and para-Japonic based on external sources
23
consistent with its phonetic; the mjijH reading looks like some kind of mistake.19 Therefore, we again have here an example of sacrificing accuracy in vocalism for accuracy in consonantism. Incidentally, this also confirms the reconstruction of 卑彌呼 as ✶pi-miko ‘sun priestess’, and not as ✶pime-ko ‘princess-?’. Miyake (2003: 114) suggests that the Ryukyuan evidence indicates pR ✶memi and Bentley (2008: 19) somewhat echoes Miyake in reconstructing ✶memV rather than ✶mimi. Unfortunately, Ryukyuan evidence is inconsistent. Sakishima dialects point to ✶memi, but the evidence from Amami dialects is controversial: while Yuwan and Tokunoshima mïŋ indicate pR ✶memi, Koniya mimi (Uchima and Arakaki 2000: 453), Nase miŋ, Kamezu mimi, and Shitooke mimi (Hirayama 1966: 365) all point to pR ✶mimi. In addition, the Iejima dialect located off-shore of the Motobu peninsula in Okinawa, which has the consistent reflexes of mi < pJR ✶me and ni < pJR ✶mi (cf. Iejima mizi ‘water’ < pJR ✶mentu and Iejima niitsi ‘three’ < pJR ✶mi-tu) exhibits ninii ‘ear’ (Oshio 1999: 278). This combination of the flawed Chinese evidence with the flawed Ryukyuan evidence strongly invites the more conservative approach of reconstructing just pJR ✶mimi. The word ✶ko ‘child’ variously transcribed as ✶hɔ (呼), ✶koʔ (狗), and ✶kuɔ (觚) basically just confirms that WOJ kô ‘id’ preserved the primary pJR ✶o, but does not contribute otherwise to the reconstruction of pJR. The word ✶sema ‘territory’20 presents conflicting evidence, with the divergent transcriptions of ✶sie-maʔ (斯馬),✶ziʔ-maʔ (兕馬), and ✶siat-mɑ (泄謨). It seems that the preponderance of evidence points to ✶sema rather than ✶sima (unless ✶siat-mɑ (泄謨) was intended for ✶yias-mɑ), and in conjunction with the other external evidence for this word that has already been discussed above, it seems comparatively safe to assume pJR ✶ sema here. The word for ‘countryside, village’ transcribed as ✶pie-nɑ (卑奴) could indicate either pJR ✶pina or ✶pena. Unlike the case with pJR ✶pi ‘the sun’ above, it is difficult to pass judgment because this word in not attested in Ryukyuan. Finally, the word for ‘guard’ transcribed by Miyake as ✶mo-liai (母離) and by Bentley as ✶mə-liai again represents a problem. I agree with both Miyake’s reconstruction of the vowel in the first syllable and with his observation that this word is not attested in any WOJ text that differentiates /mô/ and /mə/ (Miyake 2003: 117).21 However, there is oblique evidence indicating that it indeed was WOJ or pre-OJ ✶môri < pJR ✶mori. EOJ -muri ‘guard’ in sakî-muri (佐伎牟理) ‘border-guard’ attested in MYS 20.4364 can only go back to pJR ✶mori, and not ✶məri, because pJR ✶ə does not raise to EOJ u. Furthermore, the character 母 ‘mother’, although it had a central vowel by the Later Han period, could also rhyme with syllables containing ✶-ɔ or ✶-o vowels (Laurent Sagart, 19 Many thanks to William Baxter and Laurent Sagart for their detailed consultations on this matter. 20 We should assume this meaning of the word ✶sema here, and not ‘island’ as Bentley believes (2008: 14), due to the contexts in WZ where it occurs, neither of which suggests ‘island’. 21 Martin (1987: 727) suspected that the WOJ form was possibly ✶môri but did not provide any supportive evidence for his point of view.
24
Alexander Vovin
p.c.). Also, the reading of the character 離 by late Han times was likely ✶rje (Laurent Sagart, p.c.). The mid vowel ✶e here cannot be attributed to any gaps in Chinese syllabic structure because the syllable ✶ri was readily available, e.g. it could be represented by 梨 (Laurent Sagart, p.c.). Before we get excited and decide to claim that WOJ nominalizer -i went back to ✶-e as sometimes claimed, we should not forget that this is the only example of the nominalizer ✶-e found in Wei zhi wo ren zhuan. In conclusion, I must say that the above Japonic words in the Wei zhi wo ren zhuan offer potential, but not conclusive evidence for the reconstruction of primary ✶e and ✶o in pJR.
3.2 Chinese transcriptions of Sui shu Wo guo zhuan Those Chinese chronicles from the Six Dynasties period (六朝), such as Liang shu (梁 書), Nan shi (南史), and Bei shi (北史) that include any transcriptions of Japonic words contain abbreviated versions of the account found in Wei zhi Wo ren zhuan, and therefore they are of little, if any value for our purposes. The first original account after Wei zhi Wo ren zhuan is that found in the ‘Account of the Wa country in the chronicle of Sui’ (Sui shu Wo guo zhuan, 隋書倭國傳). However, very little valuable linguistic material is found in the Sui shu Wo guo zhuan, because by that time the Yamatö state was slowly becoming literate, and so many place names found in this source are just renderings of the script used by the Japanese themselves, e.g. the name of Suruŋga province is written as 駿河. Among the existing transcriptions that are more or less reliable and are not obviously corrupted, only two words are of any interest. These are words corresponding to WOJ opö ‘big’ and kîmî ‘lord’ (Table 11). Table 11: Japonic words in Sui shu wo guo zhuan. gloss
SS transcription
EMC
big lord
阿輩 雞彌
✶
ʔâ ✶pəjh kie ✶mjie
✶
Miyake pre-OJ (2003) ✶
əpə kimi
✶
pre-OJ ca. 600 AD (Vovin) əpə keme
✶ ✶
I believe that the first syllable in pre-OJ ✶əpə ‘big’ is transcribed as ✶ʔâ because, while ✶ ʔâ and ✶ʔô were available in EMC, to the best of my knowledge ✶ʔə was not, so the lack of accuracy in this case can be explained by that gap in the EMC syllabic structure. It is also quite interesting that the SS transcription of the second syllable in pre-OJ ✶əpə ‘big’ confirms the existence of the syllable ✶pə in WOJ, which is otherwise confirmed by the spelling 富 pö = [pə] that is preserved only in the Kojiki. Furthermore, since EMC has a phonemic contrast between ✶p and ✶b, pre-OJ ✶əpə indicates that intervocalic ✶-p- was phonetically [-p-] and not [-b-]. The accuracy of the transcription of the second word, pre-OJ ✶keme ‘lord’, corresponding to WOJ kîmî ‘id.’ may be more arguable. Citing a personal communication
1 Reconstruction of Japonic and para-Japonic based on external sources
25
from Leon Serafim, Miyake (2003: 118) argues that there is no Ryukyuan evidence for two mid vowels ✶e in this word. But to the best of my knowledge, there is no Ryukyuan evidence for this word except OOk, MOK kimi, and Modern Shuri cimi ~ ciN. In addition, the Okinawan word throughout its history seems to denote primarily high priestesses in service of the royal court (Okinawa kogo daijiten henshū iinkai 1995: 229), (Kokuritsu Kokugo Kenkyūjo 1999: 153). Meanwhile, in both WOJ and EOJ the word kîmî ‘lord’ refers almost exclusively to males. In Japanese, the word started to be used in reference to females only starting from the Heian period, e.g. MJ pime-gimi ‘lady princess’. Thus, given its limited distribution in Ryukyuan, and the near-identical semantics with MJ, but not OJ, it seems that the Okinawan word represents a loan into Okinawan from MJ. Thus, it cannot possibly be taken as evidence for the lack of raising ✶e > i in Ryukyuan. Miyake further argues in favor of pre-OJ ✶kimi on the basis of the fact that this word is spelled as 吉寐 ✶kit ✶mjih in the Han yuan (翰苑). He also thinks that 雞 ‘rooster’ could be used as a deliberately pejorative sign (Miyake 2003: 118). I cannot see why 雞 ‘rooster’ would be pejorative. In addition, the Han yuan is a Tang period text, without exact dating (although it is sometimes believed to be written in early Tang), with only one book out of original thirty being extant, which survived only in Japan. Thus, not only can its transcription represent the WOJ phonetic shape of the word, there is also a distinct possibility that the auspicious character 吉 ‘bliss’ could have been introduced by a Japanese copyist, and both characters 吉 and 寐 could have been used under the influence of the identical man’yōgana phonograms 吉 kî and 寐 mî. Much more significant is the fact that Suiko period transcriptions ✶keme (✶kie ✶mjie, 支彌) and ✶kemi/e (✶kie ✶mï, 支美), also dealt with by Miyake (2003: 124–125), neatly fit with the transcription of this word in the Sui shu Wo guo zhuan. Thus, I reconstruct pre-OJ ✶keme ‘lord’.
3.3 Chinese and Korean transcriptions of Old Okinawan With the advent of the Ming dynasty in China (1368–1644 AD), there is a sharp increase in materials dedicated to foreign languages, due to the activities of the Bureau of Translators and Bureau of Interpreters. Although there are several vocabularies of mainland Japanese compiled starting from the fifteenth century, they are valuable materials only for a history of the Japanese language, but not for reconstruction. This should be of no surprise, since there are other earlier, mostly internal sources available for reconstruction purposes. It is quite a different matter with sources on Ryukyuan languages, or, to be more precise, on the Old Okinawan language, because Korean and some Chinese materials either predate or are contemporaneous with the earliest extant native materials, such as Old Okinawan inscriptions and the Omoro sōshi. In addition, the very fact that Old Okinawan native materials are written in a very cumbersome and unsystematic orthography based on Japanese syllabic kana that obscures certain phonological features makes the phonetic interpretation and subsequent phonological analysis of the Old Okinawan language solely based on internal sources a very lame enterprise.
26
Alexander Vovin
In particular, Chinese and Korean transcriptions of Old Okinawan quite clearly indicate that intervocalic voiced obstruents were prenasalized in the same way as they were in OJ. This is of great importance for the reconstruction not only of Old Okinawan itself, but also of pR and pJR. Thorpe, for example, believed that the nasal sonorants in Korean transcriptions were a special device only indicating voicing (Thorpe 1983: 283, 294) and reconstructed only plain voiced ✶b, ✶d, ✶g, and ✶z in his seminal dissertation (Thorpe 1983).22 Thorpe’s point of view is based essentially on an earlier and similar formulation regarding Korean and Chinese transcriptions by Iha (1934: 47), consequently echoed by Ledyard (1966: 420, n. 66) and Hagers (1997: 36) regarding Korean transcriptions and Ding (1995: 92) regarding Chinese transcriptions. However, the nasals in Korean transcriptions were more recently correctly interpreted by Tawata (2010: 369) and Osterkamp (2012) as reflecting prenasalization in Old Okinawan. To this now can be added the diachronic analysis of Chinese transcriptions presented in Table 12:23 Table 12: Prenasalized obstruents in Old and Middle Okinawan in Korean and Chinese transcription.24 gloss
22 Serafim (2008: 84, note 3) also reconstructs prenasalization only for pJR but not for Old Okinawan. 23 Based on Vovin (2012). 24 This and some other examples demonstrate that Old Okinawan retained on some occasions both primary and secondary e, unlike Modern Okinawan, which has raised both unconditionally to i.
1 Reconstruction of Japonic and para-Japonic based on external sources
25 The actual Korean spelling is kananzu, but there is frequent confusion between /n/ and /r/ in the Haedong chegukki ŏŭm ponyŏk; see also the entry ‘get muddy/cloudy’ below. 26 The actual Korean spelling is riŋgɪna, but there is frequent confusion between /n/ and /r/ in the Haedong chegukki ŏŭm ponyŏk; see also the entry ‘head [hair]’ above. 27 Notice that even Modern Okinawan preserves prenasalization in this word.
Two observations can be made on the basis of Table 12 above. First, in spite of occasional fluctuation between prenasalized and non-prenasalized forms, such as aŋgeri ~ ageri ‘raise, give’ or kanji ~ kaji ‘wind’, the majority of the forms in OOk are prenasalized. This clearly presents, alongside with WOJ, the second independent piece of evidence for reconstructing prenasalized voiced obstruents or NC clusters for pJR. Second, if we look at MOk data, we can see that in the 1721 Zhongshan zhuan xin lu there are still several words that still have prenasalized obstruents, e.g. fɪndari ‘left’ or minzɪ ‘water’, but in the 1764 Liuqiu ruxue jianwen lu the prenasalization is completely gone. Although the process of denasalization of prenasalized voiced obstruents had started in the fifteenth century (see more on this below), this stark difference between two eighteenth century Chinese vocabularies allows us to date the final disappearance of prenasalization in MOk between 1721 and 1764 AD, with the latter being a non post quem. There are nevertheless exceptions that do not demonstrate any prenasalized obstruents in OOk, although they should be expected. I list these exceptions in Table 13 below. Table 13: Lack of prenasalization in Old Okinawan in Korean and Chinese transcription.28 gloss
pJR
oil, fat
✶
OOk
MOk
MdOk
anpura
abura 아부라 (HCK 3b.3)
awura 阿㕶打 (ZZL 466), anda28 庵答 (LDJ 339)
’aNda
helmet
✶
kanputo
kawutu 嗑塢吐 (YZH 154)
kawutu 嗑塢吐 (ZZL 254) kabutu 哈不毒 (LDJ 271)
kabutu
last year
✶
kənsə
kuzo 구조 (HCK 1b.4)
–
kuzu
28 Interestingly enough, both MOk anda and MdOk ’aNda present evidence for nasalization here.
29
1 Reconstruction of Japonic and para-Japonic based on external sources
With the exception of the last and the first three examples, the denasalized voiced obstruents in the remaining examples are invariably found after initial nasal n- in the first syllable. Therefore, it is possible to view the process of denasalization as a kind of nasal dissimilation. This is probably how the denasalization started in OOk, consequently spreading into other environments. Finally, Kikaijima dialects in the extreme north of the Ryukyuan archipelago as well as Tonogusuku and Hateruma dialects in Yaeyama islands in Sakishima preserve partial evidence for prenasalization in pR as Table 14 demonstrates: Table 14: Traces of prenasalization in Kikaijima and Sakishima.30 gloss
Kikaijima30
Tonogusuku
Hateruma
OOk
fan
–
o:ŋgi, oŋgï
–
aŋgi
raise, give
–
agiruN
aŋgiruN
aŋgeri, ageri, aŋgiri, aŋgiti, agete
hurry
–
iśuŋgi
isuguN
–
hare
’usaŋi ~ usagi
uśa:ŋgï
usagi
osaŋgi, usaŋgi
mirror
kagami ~ hagami ~ khagami
kaŋgaN
kaŋgaN
–
knife
–
śiŋgu
śigu
–
throw
–
naŋgiruN
naŋgiruN
–
29 The actual Korean spelling is rigasa, but there is frequent confusion between /n/ and /r/ in the Haedong chegukki ŏŭm ponyŏk, see also the entries ‘head [hair]’ and ‘get muddy’ in table 12 above. 30 Kikaijima data are cited according to Kibe et al. (2011). I am very grateful to Prof. Nobuko Kibe for providing me with a copy of this important publication.
30
Alexander Vovin
Table 14 (continued) gloss
Kikaijima
Tonogusuku
Hateruma
OOk
request
–
niŋgai
–
–
beard, moustache
fiŋɪ ~ çinji ~ phigi
pïni
pi̥ ni
piŋgi
right
miŋi ~ migi
–
–
miŋgiri
oil, fat
ʔaNba~ʔabba~ ʔaNda ~ ʔabura
aba
aba
abura, MOk anda
As we can see, there is no complete agreement between these dialects and OOk in all cases except ‘beard, moustache’ and ‘right’, but even partial evidence is extremely suggestive.
4 Materials on para-Japonic languages in Chinese, Japanese, and Korean sources The following represents the analysis of the Japonic data of the pseudo-Koguryŏ Paekche, and Silla Japonic languages. The materials discussed here are far from being exhaustive because only those entries that have any significance for the reconstruction of pre-para-Japonic are included. The major bulk of materials comes from the Korean chronicle ‘The historical records of the Three Kingdoms’ (Samguk sagi, 三國史記), which was compiled by Kim Busik in 1145 AD, but it certainly represents a collation of several much earlier sources that are no longer extant. Some materials are also found in Chinese dynastic chronicles such as Wei zhi (魏志, ca. 290 AD), Wei shu (魏書, 551–54 AD), Liang shu (梁書, 635 AD), and Zhou shu (周書, early 7th century AD), as well as in the Nihonshoki (日本書紀, 720 AD). Most of the Samguk sagi data are found in the geographic section chiriji (地理志, vol. 34–37) of this work. The place names are highly heterogeneous, some being apparently Korean, some no less apparently Japonic, a very few possibly (but by no means certainly) Tungusic, and others unidentified. Only Japonic data will be treated in this chapter. The following table presents the data from the pseudo-Koguryŏ language that are relevant to the reconstruction of pre-para-Japonic. Three general observations can be made on the basis of Table 15 below. First, it becomes quite apparent that pseudo-Koguryŏ transcriptions are based on Late Han readings, and not on EMC readings, as is usually thought. This leads us to an inevitable conclusion that writing and at least elementary literacy on the Korean peninsula started much earlier than is usually thought. In other words, the use of writing for transcribing at least native place names in all probability goes back to the period when Chinese
1 Reconstruction of Japonic and para-Japonic based on external sources
31
Table 15: Para-Japonic words in the pseudo-Koguryŏ language.3132 gloss
pseudo-Koguryŏ
LHC reading
water, river
買 ✶me SGSG 37.3a31
✶
mɛB
garlic
買尸 ✶mel SGSG 37.5b
✶
mɛB ✶lhi
mouth
忽次, 古次 ✶koči SGSG 37.5a
✶
hare
烏斯含 ✶osegəm SGSG 37.4a
✶
huət ✶tshiC, kɔB ✶tshiC
bear
功木 ✶koŋmo(k) SGSG 37.4a
three
密 ✶mit SGSG 37.5a
EMC reading ✶
maïB
maïB śi
pJR me ‘water’
✶
pre-paraJaponic ✶
me
✶
me/ira32 ‘leek’ kuto/uy
✶ o
osanki
✶
osanki(m)?
✶
mela
✶
xuət ✶tshiC, kuoB ✶tshiC
✶
ʔɑ ✶sie gəmB
✶
ʔuo ✶sje ✶ɣâmB
✶
✶
koŋ ✶mok
✶
kuŋ ✶muk
✶
koma
✶
koma
✶
mït
✶
mjet
✶
mi
✶
mi(t)?
✶ ✶
✶
k /utoy
commanderies existed on the peninsula, and not to the period of the Three Kingdoms. Similar additional evidence can be provided on the basis of transcriptions of Koreanic that underlie the mixed bag of pseudo-Koguryŏ place names as well as on the basis of the hyangch’al script. This, however, falls outside of the scope of the present chapter and will be published elsewhere in the near future. Second, it appears that the paraJaponic pseudo-Koguryŏ language, like pR, preserves the original mid vowels ✶e and ✶ o. The only case of a major discrepancy is found in the case of the first syllable in pseudo-Koguryŏ ✶koči ‘mouth’ and pJR ✶kuto/uy ‘id.’, because pR ✶kuti ‘id.’ (Thorpe 1983: 308) quite unambiguously indicates pJR ✶u in the first syllable.33 Third, para-Japonic pseudo-Koguryŏ might have preserved some syllable final consonants as demonstrated by ‘hare’ and ‘three’ above. In addition, there are two examples presenting potential evidence for two additional phenomena, although unfortunately this evidence cannot be confirmed by additional examples. Thus, para-Japonic pseudo-Koguryŏ ✶mel ‘garlic’ may indicate that peninsular para-Japonic had the phonemic distinction between ✶l and ✶r like OK, and unlike pJR. Also, para-Japonic pseudo-Koguryŏ ✶osegəm ‘hare’ may indicate that NC clusters preserved in pJR were already simplified to voiced consonants in peninsular para-Japonic.
31 As in this case there might be multiple attestations. I only give one for illustrative purposes. Those interested in tracing all of them should use the excellent index by Song Ki-jung (2004). 32 Insufficient data to reconstruct either ✶mera or ✶mira because the word is not attested in Ryukyuan. ModJ nira may point to ✶mira rather than ✶mera, but this is by no means certain. 33 For the reconstruction of pJR ✶o/uy in the second syllable, consider WOJ kutu-N-pamî ‘bridle’ (lit. ‘mouth-GEN-insert’) and MJ kutu-wa ‘bridle’ (lit. ‘mouth-ring’) and kutu-wa-musi ‘insect that sounds like a bridle’.
32
Alexander Vovin
The extant materials from the Paekche language or languages are much scarcer than those from pseudo-Koguryǒ, and, in addition, it is more difficult to differentiate here between the potential loanwords from para-Japonic into Korean and proper peninsular para-Japonic vocabulary. Two of these words, ‘island’ and ‘bear’ were already discussed above in section one on loanwords (see Table 9 and the following discussion). Two additional words in Table 16 that have value for the reconstruction of proto-para-Japonic can be brought into the discussion. Table 16: Possible para-Japonic words in Paekche.34 gloss
Paekche
LHC reading
king
於羅瑕 ✶oraga ZS 1.512, オリコケ orikoke NSK 19.73
✶
castle
基 ✶kï SGSG 36.4a
✶
EMC reading
ʔɑ ✶rɑ ga
✶
ʔuo lâ ✶ɣa
kïə
✶
✶
✶
kjï
WOJ
pre-paraJaponic
WOJ ira‘noble’34
✶
era-
WOJ kï
✶
kï?
Again, the LHC readings appear to be a better match for Paekche words than EMC readings. We have already seen above the correspondence of LHC ✶ʔɑ to pseudo-Koguryǒ ✶o (in the case of ‘hare’ in Table 15). WOJ kï ‘castle’ certainly was not palatalized, therefore the LHC reading ✶kïə presents a better match than the EMC reading ✶kjï. The second word is especially interesting as it could present circumstantial external evidence for the reconstruction of a seventh pJR vowel ✶ï, as proposed by Frellesvig and Whitman (2008). It has become apparent recently that the Silla territory was also originally Japonic-speaking.35 I proposed to call this pre-Korean Japonic language from the Silla territory Silla Japonic (Vovin 2007). Table 17 presents the Silla Japonic data relevant for the reconstruction of pre-para-Japonic. Table 17: Para-Japonic words in Silla Japonic. gloss
Silla Japonic
LHC reading
settlement water, river
牟羅 ✶mura LS 14.805
買 me SGSG 34.5a
✶
road, way
彌知 miti SGSG 34.3b
genitive case marker
乃 ✶nə SGSG 34.3b
✶
tiger
刀良 ✶tora SGSG 34.5b
✶
✶
EMC reading
mu ✶rɑ
✶
mie ʈie
✶
mɛ
✶
B
nəB
✶
✶
tɑu ✶riɑŋ
✶
mjəu ✶lâ maï
✶
B
mjie ʈje nâiB
✶
✶
tâu ✶ljaŋ
✶
pJR mura ‘village’
✶
me
✶
mi-ti
✶
nə
✶
tora
✶
pre-paraJaponic mura
✶
me
✶
mi-ti
✶
nə
✶
tora
✶
34 Found in WOJ ira t-u kô ‘noble man’ and ira t-u mê ‘noble woman’. 35 See Unger (2005) for the historical evidence and Vovin (2007, 2013) for the linguistic evidence.
1 Reconstruction of Japonic and para-Japonic based on external sources
33
Table 17 (continued) gloss
Silla Japonic
LHC reading
face, front private
豆良 ✶tura SGSG 34.9b
武冬 mutuŋ SGSG 34.4a
✶
adjectival attributive
西 -se SGSG 34.10a
✶
✶
✶
EMC reading
doC ✶liɑŋ
✶
sei
✶
muɑ touŋ
✶
B✶
dəuC ✶ljaŋ mju
✶
siei
tuoŋ
B✶
pJR
pre-paraJaponic
tura
tura
✶
WOJ mutu ‘intimate’ WOJ -si
36
✶
mutu
✶
-se
✶
Similar to pseudo-Koguryŏ and Paekche, the LHC readings again appear to be a better match for Silla Japonic words than EMC readings, with the possible exception of ‘private’ above, in which EMC reading ✶mjuB of the character 武 seems to be a better match than the LHC reading ✶muɑB. It also appears that Silla Japonic, like para-Japonic pseudo-Koguryŏ preserves the original mid vowels ✶e and ✶o, and we are lucky to have one word with original ✶e: ✶me ‘water’, which is also attested in pseudo-Koguryŏ para-Japonic. The reconstruction of three words in the chart above deserves detailed comments. The pJR ✶mi-ti ‘road, way’ must be a compound, consisting of ✶ti ‘road’ (cf. WOJ ti-mata ‘road fork’), and probably an honorific/beautification prefix, which most likely should be reconstructed as ✶mi-, although this is by no means certain, because it can be also reconstructed as ✶me-, because the crucial Ryukyuan evidence here is conflicting (Serafim 2004). That is, Ryukyuan evidence for either ✶miti or ✶meti is also inconclusive, cf. Yuwan miči, Koniya mit, Higashi Nakasone and Yonaha m͎tsï, Hateruma and Tonogusuku mitsï (Uchima and Arakaki 2000: 453). While the first four point to ✶miti, the last two indicate ✶meti. I have already pointed out above that the character 彌 ✶mie may be used for ✶mi, due to the gap in LHC syllabic structure. The character 知 ✶ʈie is used to transcribe Silla Japonic ✶ti for the same reason: there was no syllable ✶ti in LHC. Thus, I take a conservative approach here and reconstruct ✶mi-ti for both pJR and pre-para-Japonic. The seemingly odd LHC 刀良 ✶tɑu ✶riɑŋ for ✶tora ‘tiger’ and 豆良 ✶doC ✶riɑŋ for ✶ tura ‘face, front’ are actually quite systematic. There was no syllable ✶tu in LHC, therefore mid vowel ✶o was chosen to transcribe the high vowel, and the diphthong ✶ɑu to render the mid vowel.
36 WOJ -si can also be used as a final (Martin 1987: 806–809).
34
Alexander Vovin
5 Summary of main findings The main findings in this chapter can be summarized as follows: 1) It is preferable to use LHC rather than EMC readings of the Chinese characters when dealing with para-Japonic languages, as the former reflect more precisely their phonetics. 2) It appears that there is no basis for reconstructing the contrast ✶si : ✶se in proto-Ryukyuan as Thorpe did; at least foreign transcriptional evidence does not support this contrast. There are no similar data for an ✶i : ✶e contrast after other coronals. 3) Old Japanese loans in Ainu highly suggest that OJ intervocalic -p- was phonetically [-p-] and not [-b-] or [-β-] as frequently claimed. Ditto for other stops as well. 4) Old Japanese loans in Ainu also provide interesting insights into mid vowel raising and the history of accent in Japanese. 5) Japonic loanwords in Korean also provide evidence for mid vowels ✶o and ✶e in a couple of words that cannot be reconstructed for Japonic using just the Comparative Method. 6) Early Chinese transcriptions of Japanese words also indicate some possible primary mid vowels, but what is more important, there is clear evidence for OJ intervocalic -p- being phonetically [-p-] and not [-b-] or [-β-]. 7) Both Korean and Chinese transcriptions of Old Okinawan indicate that modern Okinawan intervocalic and postsonorant voiced stops were once prenasalized pretty much as in Western Old Japanese. They also present evidence for Old Okinawan mid vowels /e/ and /o/ (both primary and secondary) that were raised to /i/ and /u/ in Modern Okinawan. 8) The main value of para-Japonic materials is that they preserve the original mid vowels ✶e and ✶o. This becomes apparent once we use the appropriate LHC and not EMC readings of Chinese transcriptional characters.
Abbreviations LHC MdOk MOk OK OOk pA PC pJR pR WOJ
Late Han Chinese Modern Okinawan Middle Okinawan Old Korean Old Okinawan proto-Ainu Paekche Old Korean proto-Japonic proto-Ryukyuan Western Old Japanese
1 Reconstruction of Japonic and para-Japonic based on external sources
35
Primary sources HCK LGY LRJ LS MYS NSK SGSG SLL SS WS WZ YZH ZS ZZL
Haedong chegukki ŏŭm pŏnyŏk (海東諸國紀語音翻譯) 1501 AD Liuqiu guan yiyu (琉球館譯語) second half of the 15th c. Liuqiu ruxue jianwen lu (琉球入學見聞録) 1764 AD Liang shu (梁書), 635 AD Man’yōshū (萬葉集), between 759 and 785 AD Nihonshoki (日本書紀), 720 AD Samguk sagi (三國史記), 1145 AD Shi Liuqiu lu (使琉球録) 1535 AD Sui shu Wo guo zhuan (隋書倭國傳), 621 AD Wei shu (魏書), 551–54 AD Wei zhi Wo ren zhuan (魏志倭人傳), ca. 290 AD Yinyun zihai (音韻字海) second half of the 16th c. (?) Zhou shu (周書), early 7th century AD. Zhongshan zhuan xin lu (中山傳信録) 1721 AD
References Akimoto, Kichirō (ed.). 1958. Fudoki [Gazetteers]. (Nihon koten bungaku taikei [Series of Japanese classical literature], vol. 2.) Tokyo: Iwanami Shoten. Bentley, John R. 2008. The search for the language of Yamatai. Japanese Language and Literature 42. 1–43. Ding, Feng. 1995. Deciphered text of the Liuqiu guan yiyu. Ryūkyū no hōgen 20. 86–105. Frellesvig, Bjarke. 2011. A history of the Japanese language. Cambridge: Cambridge University Press. Frellesvig, Bjarke and John B. Whitman. 2008. Evidence for seven vowels in proto-Japanese. In Bjarke Frellesvig and John Whitman (eds.), Proto-Japanese: Issues and prospects, 15–41. Amsterdam & Philadelphia: John Benjamins. Hagers, Steven. 1997. The importance of a Korean source on the Ryūkyūan language. In Bjarke Frellesvig and Roy Starrs (eds.), Japan and Korea. Contemporary studies, 34–43. Aarhus: Aarhus University Press. Hirayama, Teruo. 1966. Ryūkyū hōgen no sōgōteki kenkyū [A comprehensive study of Ryukyuan dialects]. Tokyo: Meiji Shoin. Hudson, Mark J. 1999. Ruins of identity. Honolulu: University of Hawaiʻi Press. Iha, Fuyu. 1934. Haedong chegukki fusai no ko Ryūkyūgo no kenkyū (Ŏŭm pŏnyŏk shakugi) [A study of the Old Ryukyuan language in the appendix to Haedong chegukki (A commentary on Ŏŭm pŏnyŏk)]. In Fuyu Iha, Nantō hōgen sikō [Historical research on Ryukyuan dialects], 37–125. Tokyo: Rakurō Shoten. Ishihara, Michihiro. 1985. Gishi Wajin den, Kō kansho Wa den, Sōsho Wakoku den, Zuisho Wakoku den [Account of Wa people in the chronicle of Wei, account of Wa in the later Han chronicle, account of Wa country in the Song chronicle, account of Wa country in the Sui chronicle]. Tokyo: Iwanami Shoten. Kibe, Nobuko, Haruo Kubozono, Kayoko Shimoji, Wayne Lawrence, Akiko Matsumori and Kōko Takeda. 2011. Kikaijima hōgen chōsa hōkokusho [Research report on the Kikaijima dialects] (NINJAL Collaborative Research Project Reports 11–01). Tokyo: National Institute for Japanese Language and Linguistics. Kindaichi, Haruhiko. 1971. On’in henka kara akusento henka e [From phonological change to accentual change]. Kindaichi hakushi beiju kinen ronshū [A Festschrift on the occasion of the eighty-eighth birthday of Dr. Kindaichi (Kyōsuke)], 929–956. Tokyo: Sanseidō.
36
Alexander Vovin
Kindaichi, Kyōsuke and Chiri Mashiho. 1936. Ainu gohō gaisetsu [A sketch of Ainu language grammar]. Reprinted in 1974, Chiri Mashiho chosakushū [Selected writings of Chiri Mashiho] 4. 3–197. Tokyo: Heibonsha. Kokuritsu Kokugo Kenkyūjo (eds.). 1999. Okinawago jiten [A dictionary of the Okinawan language]. Tokyo: Ōkurasho. Ledyard, Gari K. 1966. The Korean language reform of 1446. Berkeley, CA: University of California at Berkeley dissertation. Mabuchi, Kazuo. 1999. Kodai Nihongo no sugata [The appearance of Old Japanese]. Tokyo: Musashino Shoin. Martin, Samuel E. 1966. Lexical evidence relating Japanese to Korean. Language 42(2). 185–221. Martin, Samuel E. 1987. The Japanese language through time. New Haven & London: Yale University Press. Miyake, Marc H. 2003. Philological evidence for ✶e and ✶o in pre-Old Japanese. Diachronica 20(1). 83–137. Mizuno, Yū. 1989. Hyōshaku gishi wajin den [An annotated account of Wa people in the chronicle of Wei]. Tokyo: Yūsankaku Shuppan. Murayama, Shichirō. 1975. Hayato no kotoba [The language of Hayato]. In Taryō Ōbayashi (ed.), Hayato. Nihon kodai bunka no tankyū [Hayato. The search for the ancient Japanese culture], 249–263. Tokyo: Shakai Shisōsha. Okinawa kogo daijiten henshū iinkai (eds.). 1995. Okinawa kogo daijiten [A dictionary of the Old Okinawan language]. Tokyo: Kadokawa Shoten. Omodaka, Hisataka et al. (eds.). 1967. Jidai betsu kokugo dai jiten: Jōdai hen [A large dictionary of the national language by periods: Old Japanese]. Tokyo: Sanseidō. Osada, Natsuki. 1979. Yamatai koku no gengo [The language of Yamatai]. Tokyo: Gakuseisha. Osada, Natsuki. 2010. Yamatai koku no gengo: Yayoigo no fukugen [The language of Yamatai: The reconstruction of the Yayoi period language]. Tokyo: Gakuseisha. Oshio, Mutsuko. 1999. Okinawa Iejima hōgen jiten [A dictionary of the Iejima dialect in Okinawa]. Iemura: Iemura Kyōiku Iiinkai. Osterkamp, Sven. 2012. Prenasalization in Japonic languages as seen in pre-modern foreign transcriptions: A look at the sources and their interpretation. Lecture presented at the East Asian Linguistics Seminar, Oriental Institute, University of Oxford, 21 February. Ramsey, S. Robert. 1979. The Old Kyōto dialect and the historical development of Japanese accent. Harvard Journal of Asiatic Studies 39(1). 157–175. Ramsey, S. Robert. 1980. Nihongo no akusento no rekishiteki henka [Historical changes of Japanese accent]. Gengo 9(2). 64–76. Serafim, Leon A. 2004. The Shuri Ryukyuan exalting prefix ✶myi- and the Japanese connection. Japanese Language and Literature 38(2). 301–322. Serafim, Leon A. 2008. The uses of Ryukyuan in understanding Japanese language history. In Bjarke Frellesvig and John Whitman (eds.), Proto-Japanese: Issues and prospects, 79–99. Amsterdam & Philadelphia: John Benjamins. Shōgakkan (eds.). 1974. Nihon kokugo daijiten [A large dictionary of the Japanese language]. 20 vols. Tokyo: Shōgakkan. Song, Ki-jung. 2004. Kodae kugŏ ŏhwi pʼyogi hancha ŭi chabyŏl yongnye yŏnʼgu [The study of the usage of individual Chinese characters in Ancient Korean]. Seoul: Sǒul Taehakkyo Ch’ulp’anbu. Tamura, Suzuko. 1998. Ainugo shiryō no seiri to bunseki kenkyū [Management and analytic research on Ainu language materials]. Tokyo: Report to the Monbukagakushō [Ministry of Education, Culture, Sports, Science and Technology]. Tawata, Shin’ichirō. 2010. Okinawago on’in no rekishiteki kenkyū [A historical study of Okinawan phonology]. Hiroshima: Keisuisha. Thorpe, Maner L. 1983. Ryukyuan language history. Los Angeles, CA: University of Southern California unpublished dissertation.
1 Reconstruction of Japonic and para-Japonic based on external sources
37
Tokugawa, Munemasa. 1972. Towards a family tree for accent in Japanese dialects. Papers in Japanese Linguistics 1(2). 301–320. Uchima, Chokujin and Kumiko Arakaki. 2000. Okinawa hokubu nanbu hōgen no kijutsuteki kenkyū [A descriptive study of Northern and Southern Ryukyuan dialects]. Tokyo: Kazama Shobō. Unger, J. Marshall. 2005. When was Korean first spoken in Southeastern Korea? Journal of Inner and East Asian Studies 2(2). 87–105. Vovin, Alexander. 1990. Японские заимствования в айнском языке [Japanese loanwords in the Ainu language]. Pis’mennye Pamiatniki i Problemy Istorii Kul’tury Narodov Vostoka [Written monuments and problems in the cultural history of Oriental peoples] (Doklady i Soobshcheniia XXIII Godichnoi Sessii LO IV AN SSSR [Presentations and reports from the 23rd yearly session of the Leningrad branch of the Soviet Union Institute of Oriental Studies of the Academy of Sciences]), 58–66. Vovin, Alexander. 1993. A reconstruction of Proto-Ainu. Leiden: E. J. Brill Publishers. Vovin, Alexander. 1997. The origin of register in Japanese and the Altaic theory. In Ho-min Sohn and John Haig (eds.), Japanese-Korean Linguistics 6. 113–133. Stanford: CSLI Publications. Vovin, Alexander. 2005a. The descriptive and comparative grammar of Western Old Japanese, part 1: Sources, script and phonology, lexicon, nominals. Folkestone: Global Oriental. Vovin, Alexander. 2005b. Koguryŏ and Paekche: Different languages or dialects of Old Korean? Journal of Inner and East Asian Studies 2(2). 107–140. Vovin, Alexander. 2007. Cin-Han and Silla words in Chinese transcription. In Sang-Oak Lee, Choong-yon Park and James H. Yoon (eds.), Promenades in language: To honor Professor Chin-Woo Kim –Essays on the occasion of his retirement, 603–628. Seoul: Hanguk Munhwasa. Vovin, Alexander. 2009. Man’yōshū to Fudoki ni mirareru fushigina kotoba to jōdai Nihon rettō ni okeru Ainugo no bunpu [Strange words in the Man’yōshū and Fudoki and the distribution of the Ainu language in the Japanese islands in prehistory]. Kyoto: Kokusai Nihon Bunka Kenkyū Sentā. Vovin, Alexander. 2010. Koreo-Japonica: A re-evaluation of a common genetic origin. Honolulu: University of Hawaiʻi Press. Vovin, Alexander. 2012. Ryūkyū sogo no gochū ni okeru yūsei shiin no saiken ni tsuite [On the reconstruction of medial voiced consonants in proto-Ryukyuan]. Lecture presented on at the National Institute for Japanese Language and Linguistics, Tachikawa, Tokyo, 7 August. Vovin, Alexander. 2013. From Koguryŏ to T’amna: Slowly riding south with speakers of Proto-Korean. Korean Linguistics 15(2). 222–240. Whitman, John B. 1985. The phonological basis for the comparison of Japanese and Korean. Cambridge, MA: Harvard University unpublished dissertation. Wenck, Günther. 1959. Japanische Phonetik [Japanese phonetics]. B. 4. Wiesbaden: Otto Harrassowitz.
Thomas Pellard
2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan 1 The Ryukyuan languages The Ryukyuan languages1 are spoken in the Ryukyu Islands, a chain of around 50 inhabited islands stretching from southeast of Kyushu to northeast of Taiwan and naturally delimited by the Kuroshio Current. Ryukyuan subdivides into at least five languages, which are not mutually intelligible: Amami (Ama.), Okinawan (Oki.), Miyako (Miy.), Yaeyama (Yae.), and Yonaguni (Yonag., a.k.a. Dunan). Amami and Okinawan belong to Northern Ryukyuan, while Miyako, Yaeyama, and Yonaguni form together the Southern Ryukyuan group (Pellard 2015). All Ryukyuan languages are highly endangered: fluent speakers are usually in their late sixties or older and are bilingual in Japanese, while younger generations are monolingual in Japanese. The Ryukyuan languages form a sister branch of Japanese within the JapaneseRyukyuan family2 (Pellard 2015). Ryukyuan and Japanese most likely split during the first half of the first millenium AD, and the ancestor of Ryukyuan was then spoken on Kyushu for several centuries before it migrated to the Ryukyus around the 10th century (Pellard 2013a, 2015, 2016a). The following will address, after a brief historiographic overview, those phonological (vowels, consonants, accent/tone), grammatical (verbs, adjectives, case, kakarimusubi), and lexical (pronouns, demonstratives) topics for which the contribution of Ryukyuan is, or has been claimed to be, important for the reconstruction of protoJapanese-Ryukyuan (pJR), the common ancestor of Japanese and Ryukyuan.
2 Ryukyuan historical linguistics Though the close similarity between Japanese and Ryukyuan (usually Okinawan) had long been repeatedly noticed and their genetic relationship recognized before (Oster-
1 See Shimoji and Pellard (2010) and Heinrich, Miyara and Shimoji (2015) for a general introduction. 2 The position of the Hachijō language remains to be ascertained (Hirako and Pellard 2013). Though most specialists would group it with Japanese, and specially Eastern Old Japanese (Alexander Vovin and John Kupchik, p.c. December 2017), the relevant evidence has yet to be published. Acknowledgments: This chapter has greatly benefited from comments by (in alphabetic order) Anton Antonov, Guillaume Jacques, John Kupchik, Wayne Lawrence, Marc Miyake, Kōhei Nakazawa, Sven Osterkamp, Leon Serafim, Yukinori Takubo, J. Marshall Unger, Alexander Vovin, and the editors of this volume. https://doi.org/10.1515/9781614512851-003
40
Thomas Pellard
kamp 2015), the first historical-comparative study of Ryukyuan is Chamberlain’s (1895) comparative grammar of (Tokyo) Japanese and (Shuri) Okinawan. Chamberlain’s pioneer work is however mainly of historiographic interest today, as it suffers from the limitations of its pre-Neogrammarian methodology and conceptions of linguistic change,3 and from the shortage of descriptive linguistic studies on both Old Japanese and Ryukyuan at that time.4 It was Shirō Hattori who first undertook a strict application of the modern comparative method to Japanese and Ryukyuan on a large scale, taking advantage of the blossoming of descriptive studies on Ryukyuan in the postwar period, in which Hattori himself played an important role. After numerous and important contributions to the reconstruction of pJR (e.g. Hattori 1932, 1958, 1976, 1977), Hattori’s comparative work culminated with the publication of his magnum opus “Nihon sogo ni tsuite” (1978–1979) as a series of 22 articles (reprinted in Hattori 2018), which remains the basis of most current research. Following Hattori’s breakthrough, Thorpe (1983) was the first to undertake a systematic reconstruction of proto-Ryukyuan (pR).5 Though Thorpe’s work remains the most extensive reconstruction of pR available, it suffers from several limitations, especially in the domain of prosody, since it did not take into account several distinctions attested in Ryukyuan but absent from Japanese (§5). Other large-scale reconstructions include Martin’s (1987) timeless classic, which consistently lists Ryukyuan cognates, though it is primarily concerned with Japanese and does not accept Hattori’s findings on the pJR vocalism. The OJ grammar by Vovin (2005, 2009) and his comparative study with Korean (Vovin 2010) also provide detailed and important comparisons with Ryukyuan. Shimabukuro’s (2007) reconstruction of accent has been strongly criticized (Lawrence 2016) and should be used with caution. Bentley’s (2008) proto-Southern Ryukyuan reconstruction is a useful work, though it suffers from important limitations (Pellard 2010). Other noteworthy reconstructions include Urakami’s (1992a, 1992b) proto-Amami and Pellard’s (2009) and Pellard and Hayashi’s (2012) proto-Miyako.
3 Ryukyuan and the proto-Japanese-Ryukyuan vowel system The reconstruction of the pJR vowel system is the field which has most benefited from a comparative perspective from Ryukyuan. The pR vowel system can be reconstructed 3 Chamberlain arrived in Japan in 1873, just before the rise of the Junggrammatiker and the methodological upturn of the comparative method. 4 See Hattori (2018) for a historiographic overview of the beginnings of Japanese-Ryukyuan comparatism. 5 Unfortunately, not only was Thorpe’s dissertation never published, but he also never carried on his work.
2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan
41
following Thorpe (1983) with five vowels: ✶i, ✶u, ✶e, ✶o, ✶a. In cases where the distinction between ✶u and ✶o is unclear, Thorpe (1983) writes ✶U, but the unambiguous notation ✶ {u,o} is adopted here instead. The comparative data from Ryukyuan leads to the addition of two mid vowels, ✶e ✶ and o, to the four-vowel system (✶i, ✶u, ✶ə, ✶a) traditionally reconstructed for pJR on the basis of OJ (e.g. Miller 1967, Matsumoto 1975, Whitman 1985, Martin 1987). Ryukyuan also preserves some distinctions which are not directly attested in OJ but can be inferred to have existed earlier on the basis of internal reconstruction, e.g. the different sources of OJ Cwi. On the other hand, reconstructing a seventh pJR vowel ✶ɨ, as proposed by Frellesvig and Whitman (2008), finds no support in Ryukyuan, and the seventh vowel ✶ ü posited by Hattori (2018 [1979]) to account for some Japanese-Ryukyuan correspondences proves to be unnecessary too (Pellard 2013b).
3.1 The mid vowels ✶e and ✶o
The most important contribution of Ryukyuan to the reconstruction of pJR is perhaps the reconstruction of two mid vowels ✶e and ✶o and a six-vowel system instead of the previous standard four-vowel system. This hypothesis was first proposed by Hattori (1976, 2018 [1979]) and subsequently further developed by Thorpe (1983), Serafim (2008), and Pellard (2008, 2013b). Ryukyuan exhibits two different correspondences for both OJ i and u (Tables 1 and 2): some of the correspondents of the OJ high vowels unconditionally merged with the correspondents of the OJ mid vowels (Cye, Ce, Cwo, Co). This split correspondence leads to the reconstruction of those as mid vowels, i.e. respectively pJR ✶e and ✶o. In Northern Ryukyuan, the original pJR high vowels ✶i and ✶u usually trigger fortition of a preceding voiceless stop, while the mid vowels ✶e and ✶o trigger aspiration or spirantization. The high vowel ✶i also often triggers palatalization, while ✶e has a central or non-palatalized reflex. On the other hand, in Southern Ryukyuan ✶i often has a centralized or fricative reflex. In both groups, the original high vowels ✶i and ✶u tend to undergo syncope or nasalization in some environments. Table 1: pJR ✶i vs. ✶e in Ryukyuan. ‘ditch’ ‘water’ ‘daytime’ ‘garlic’
pJR
OJ
pR
Ama.-Yam.
Oki.-Yon.
Miy.-Hir.
Yae.-Ish.
Yonag.
minsə mentu ✶ piru ✶ peru
mizo midu piru piru
✶
mizo medu ✶ piru ✶ peru
midzo mɨdzɨ çiɾu ɸɨɾu
dʑúː mìdʑíː pˀìɾúː pʰìɾúː
mdzu midzɿ pɿː piɿ
ńdʑú mídzɿ ̀ pɿ̂ ːɾɿ ̀ pîɴ
ndú míɴ tsˀuː çìɾú
✶ ✶
✶
42
Thomas Pellard
Table 2: pJR ✶u vs. ✶o in Ryukyuan. ‘horse’ ‘sea’ ‘mortar’ ‘medicine’
pJR
OJ
pR
Ama.-Yam.
Oki.-Shur.
Miy.-Hir.
Yae.-Ish.
Yonag.
uma omi ✶ {u,o}su ✶ kusori
uma umi usu kusuri
✶
uma omi ✶ {u,o}su ✶ kusori
mˀaː ʔumi ʔusɨ kˀusuɾi
m̀ ˀmà ʔùmì ʔùːsì kùsùì
mma im usɿ fusuɿ
ḿmá íɴ úsɿ ́ ɸùɕíɾɿ ́
mmà ùnnáɡâ ùtɕî tsˀùɾî
✶ ✶
✶
The distinction between high and mid vowels was retained in most environments in pR, though not equally in all modern varieties. Since the distinction has often been transphonologized onto the preceding onset consonant, it is not always faithfully preserved when there is no onset, i.e. in word-initial (Anlaut) position, especially when the following consonant is voiceless (e.g. pR ✶{i,e}ki ‘breath’, ✶{i,e}si ‘stone’, ✶{u,o}si ‘cow’, ✶{u,o}ta ‘song’). Similarly, the loss of the pJR onset consonant in non-initial ✶-pu, ✶ -po, and ✶-wo led to the merger of ✶u and ✶o in these environments (e.g. pR ‘fall down’ ✶ ta{u,o}re-, ‘fan’ ✶a{u,o}gi, ‘pole’ ✶sa{u,o}). Also, evidence for a distinction between ✶ju and ✶jo is scarce.
3.2 Two origins of OJ Cwi syllables: ✶ui and ✶əi
The B-type Cwi syllables are usually thought to be secondary and to result from the fusion of earlier vowel sequences (Frellesvig and Whitman 2008, Hayata this volume). From the so-called “apophonic” alternations in OJ stems, it is possible to reconstruct two different origins for OJ Cwi: ✶ui for those instances of Cwi which alternate with Cu (e.g. ‘moon’ tukwi ~ tuku-), and ✶əi for those which alternate with (B-Type) Co (e.g. ‘tree’ kwi ~ ko-). In Ryukyuan, the cognates of the Cwi-apophonic forms of OJ can be divided into two sets according to whether they have a vowel ✶i or ✶e in pR. These two sets show a regular correspondence with the two different origins ✶ui and ✶əi of OJ Cwi posited from internal reconstruction. In other words, pJR ✶ui and ✶əi merge as OJ Cwi but are kept distinct in Ryukyuan as different vowels or as a distinction on the preceding consonant (Table 3),6 and the Ryukyuan data corroborates the internal reconstruction based on OJ.
6 See also Arisaka (1934), Hattori (1932, 2018), Serafim (2008), and Pellard (2013b). Polysyllabic verb stems ending in ✶ui (or ✶oi) have been analogically reshaped in the Southern Ryukyuan languages, and the distinction between the different origins of OJ Cwi is thus faithfully preserved in nouns only. Thorpe (1983: 355) proposes that the bound stem of ‘moon’ reconstructs as pJR tuko- rather than tuku-, which would entail the reconstruction of the free form as ✶tukoi, but the basis for his reconstruction is not clear (Pellard 2013b). In any case, that reconstruction is compatible with the system proposed here, as explained in §3.3. That example is retained here for lack of another Cwi ~ Cu- vs. Cwi ~ Co- minimal pair with abundant Ryukyuan cognates.
2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan
43
Table 3: Correspondences of Old Japanese Cwi in Ryukyuan.
‘moon’ ‘tree’ ‘mouth’ ‘to fall’
OJ
pR
Ama.-Yam.
Oki.-Yon.
Miy.-Hir.
Yae.-Ish.
Yonag.
tukwi ~ tukukwi ~ kokuti ~ kutuoti ~ otos-
✶
tuki ke ✶ kuti ✶ {u,o}te
tsɨki kʰɨː kˀuti ʔutɨ-
ɕìtɕǐː kʰǐː kùtɕíː ɸùtìɾúɴ̀
tsɿkɿ kiː futsɿ utiɿ
tsɿ ̀kɿ ́ kíː ɸútsɿ ̀ útíɾúɴ
tˀìː kʰìː tˀíː ùtìɾùɴ
✶
It is noteworthy that Ryukyuan distinguishes between ✶ui and ✶əi even in those contexts where OJ has no distinction between Ci and Cwi, like ‘fall’ oti- ~ otos- in Table 3. The Ryukyuan data can thus enhance our reconstructions in the following cases: when a word is not phonographically attested in OJ, when a word lacks an apophonic alternate form, or when a word contains a syllable without distinction between Ci and Cwi. On the other hand, since pJR ✶i and ✶ui, as well as pJR ✶əi, ✶ai, and ✶e, respectively merge in Ryukyuan, only the comparison between OJ and Ryukyuan can elucidate many reconstructions. To give only one example, the pR form ✶miti ‘path’ could theoretically reconstruct in pJR as ✶miti, ✶muiti, ✶mitui, or ✶muitui. On the other hand, OJ miti ‘path’ could reconstruct as pJR ✶miti, ✶meti, ✶mete, ✶mitui, ✶mitəi, ✶metui, or ✶metəi.7 The comparison of the two data sets leads by intersection to reduce the number of possibilities to just two: pJR ✶miti or ✶mitui (to which we should further add ✶mitoi as explained in §3.3).
3.3 One more origin of Old Japanese Cwi syllables: ✶oi
Some words for which we need to reconstruct a vowel ✶o in pJR, and subsequent vowel raising in OJ, exhibit the Cwi ~ Cu alternation pattern. This entails reconstructing them not with pJR ✶ui but with ✶oi (Vovin 2011, Pellard 2013b).8 This is the case of verb stems like OJ ‘exhaust’ tukwi- ~ tukus- or sugwi- ~ ‘pass’ sugwi- which correspond to pR ✶tukosand ✶sugos-. In the case of sugus-, the reconstruction pJR ✶sugos- is also supported by the existence of the EOJ form sugwos- (MYS 14.3564) and of the EMJ form sugos-, which survives in ModJ. Another example is ‘yellow’ OJ KWI- ~ ku-,9 whose reconstruction as pJR ✶koi is here again supported by the EMJ form ko-gane ‘gold’. 7 If we ignore for the sake of the discussion Frellesvig and Whitman’s (2008) proposal that mid vowel raising was limited to non-final position. 8 This solves the oddness of the systematic lack of a diphthong ✶oi in most previous reconstructions. This idea was first suggested by Frellesvig and Whitman (2008: 39), who only gave one possible example of ✶oi, OJ isi ~ iswo ‘rock, stone’, but they did not include ✶oi into their reconstructed system (Frellesvig and Whitman 2008: 16), nor did they propose a full treatment of the issue. Vovin (2011) and Pellard (2013b) independently reached the same conclusion from not completely identical evidence around the same time (April 2010), though Vovin (2011) appeared in print first. 9 The stem kwi is not phonographically attested in OJ, but the fact that it alternates with ku- shows that it must have been a B-type kwi.
44
Thomas Pellard
Among the words exhibiting an alternation between Co and Cwi in OJ, and which are thus traditionally reconstructed with an earlier ✶əi, some do not show a distinction between A-type Co and B-type Cwo. The o in such examples could thus theoretically be interpreted as coming not from ✶ə, the origin of OJ Co, but from ✶o, one of the origins of OJ Cwo. This hypothesis is reinforced by the fact that some of such words constitute exceptions to the correspondence between OJ Cwi (~ Co-) and pR ✶e described above (Pellard 2013b). Such words should thus be reconstructed with ✶oi rather than ✶əi. This is the case of ‘fire’ OJ pwi ~ po-, whose pR cognate form reconstructs as ✶pi and not ✶pe, as expected if it came from pJR ✶pəi (Table 4). Table 4: pJR ✶oi vs. ✶əi in Ryukyuan. ‘fire’ ‘yellow’ ‘tree’
pJR
OJ
pR
Ama.-Yam.
Oki.-Yon.
Miy.-Hir.
Yae.-Ish.
poi koi ✶ kəi
pwi ~ pokwi ~ kukwi ~ ko-
✶
pi ki ✶ ke
çikˀiːkʰɨː
pˀǐ: tɕˀikʰǐː
pɿ: kɿkiː
pɿ ́ː kɿkíː
✶ ✶
✶
All such examples are found after a labial consonant (i.e. {✶p,✶b,✶m,✶w}), but this is not surprising since this is precisely the exact environment where the distinction between Co and Cwo is neutralized in OJ10 and for which we may be in doubt about whether a given example should reconstruct as ✶ə or ✶o. Thorpe (1983: 230–232) argues that in fact there is no distinction between the different pJR origins of OJ Cwi after ✶p and probably ✶ w, but there is at least one potential counterexample to that hypothesis: pR ✶{u,o}po- ‘big, many’ corresponds to OJ opo- ‘big, great, many’ and has an alternate form pR ✶{u,o}pe‘big, many’, which corresponds to the OJ verb stem opwi- ‘grow’. Thorpe’s (1983: 257, 263–264) solution is to analyze pR ✶{u,o}pe- as containing the suffix ✶-pe ‘layer’, but this is debatable. The limited number of relevant examples in OJ proper and the paucity of cognates in Ryukyuan are anyway problematic.
3.4 The vowels Cwo and Co The distinction between OJ A-type and B-type o-syllables (< pJR ✶o vs. ✶ə) has been claimed to be preserved in some Amami dialects of the Ōshima and Kakeroma Islands (Hattori 1959: 58–61,11 Sibata and Mitsuishi 1979, Urakami 1993), i.e. pJR ✶ə > OJ Co :: Ama. u vs. pJR ✶o > OJ Cwo :: Ama. o. While the Amami data might seem striking at first 10 Except mwo and mo, which are distinguished in the Kojiki only. Another context where the distinction is neutralized is the initial, onsetless, position, but there are no examples of apophonic alternations involving onsetless o since apophony is restricted to stem-final syllables. 11 Hattori later abandoned his own proposal.
2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan
45
sight, the correspondences are actually complex, and it is not possible to establish the existence of a distinction between pJR ✶ə and ✶o in Amami (M. Nakamoto 1976: 94–103, Thorpe 1983: 34–37, Matsumori 1991, Nakama 1992: 67–72). The major problem for the hypothesis that some Amami dialects preserve the distinction between ✶o and ✶ə is that there are many counterexamples, perhaps about as many as examples that follow the putative sound law. Both pJR ✶ə and ✶o can correspond to either o or u in a single dialect, but there are differences in the exact correspondences between dialects (Table 5). The correspondences involved are fairly complex and need a more systematic treatment which will have to remove loans from the data, and to concentrate on the segmental (and suprasegmental) environments. Table 5: Reflexes of pJR ✶o and ✶ə in Amami. ‘heart’ ‘nine’ ‘origin’ ‘sound’ ‘cloud’ ‘feces’ ‘partner, bridegroom’ ‘thigh’
4 Consonants The pR consonant system can be reconstructed following Thorpe (1983) with 13 consonants: ✶p, ✶t, ✶k, ✶b, ✶d, ✶g, ✶s, ✶z, ✶m, ✶n, ✶r, ✶j, ✶w. Thorpe (1983) also adds two archiphonemes: the first part of geminate obstruents ✶Q and a placeless nasal ✶N, which developed through vowel loss, in a way similar to but independent of Japanese. While the comparison with Ryukyuan is crucial for reconstructing the pJR vowel system, the reconstruction of pJR consonants is rather straightforward. Though it has been proposed that the Ryukyuan consonant systems were archaic in several aspects, the contribution of Ryukyuan to the reconstruction of the pJR consonants seems limited to corroborating the OJ evidence.
4.1 Approximants A long-standing issue is whether the approximants w and y of OJ should be reconstructed as stops in pJR, i.e. respectively ✶b and ✶d. Indeed, in Southern Ryukyuan, and in some Northern Ryukyuan varieties too, OJ initial w- corresponds to a stop b- (Table 6), and in
46
Thomas Pellard
Yonaguni OJ initial y- also corresponds to a stop d- (Table 8). On the basis that lenition is more natural and frequent, it has been proposed that b- (e.g. Whitman 1985: 15–18, Martin 1987, Nakama 1996, Vovin 2010: 36–40)12 and sometimes also d- (e.g. Martin 1987) of Southern Ryukyuan were preservations of pJR ✶b and ✶d.13 However, several arguments can be offered against reconstructing pJR ✶b and ✶d, the lenition hypothesis, instead of ✶w and ✶j, the fortition hypothesis. 4.1.1 Reconstructing initial ✶wIn Southern Ryukyuan, Japanese w- regularly corresponds to b- in initial position, while in Northern Ryukyuan, most varieties have w-, but a few dialects also exhibit a stop b- or ɡ- in some contexts (Table 6). Table 6: Correspondences of JPN initial w-.
‘I, we’ ‘sit’ ‘get drunk’ ‘husband’
EMJ
Ama.-Ad.
Ama.-YHig.
Oki.-Ben.
Oki.-Kud.
Miy.-Hir.
Yae.-Ish.
Yonag.
wa wi wep-i wopito
waɴ jijui ju:jui ɡutu
wánù bjūɴ̀ ɸùìjúɴ̀ ɸūtù
ɡwaɴ biːɴ ʔuiɾuɴ ʔutu
wanaː ɡiːɴ ɡuiːɴ ɡutu
baɴ bɿː bjuːɿ butu
bánú bɿ ́ɾúɴ̀ bíːɴ búdù
bànû biɾbìɾùɴ bùtú
Some of the Northern Ryukyuan stop reflexes are clearly secondary developments since, for instance, ‘pig’ pR ✶uwa is wˀaː or waː in most Northern Ryukyuan varieties, but gwaː in Okinawa-Benoki, which suggests that a late ✶w- > gw- fortition process has occurred after the loss of the initial vowel and glottal stop in that dialect. Similarly, ʔaɡwa ‘millet’ in Okinawa-Benoki can only be due to a late fortition of pR ✶awa since it corresponds to OJ apa. In the case of Okinawa-Benoki, the absence of a w :: b correspondence before back round vowels suggests that the earlier ✶ʔu (< ✶#o) and ✶wu (< ✶#wo) seen in other dialects merged to ʔu, and that the loss of the initial ✶w prevented it from taking part into the later glide fortition process w > b. The reverse explanation, namely that ✶b was lost in front of u first, is not well motivated. The restricted distribution of ɡ in Amami-Aden is suspect too: it seems that the change ✶w > ɡ was conditioned by the following ✶o. For Okinawa-Kudaka, this ɡ only appears before non-low vowels, a mirror image of Okinawa-Benoki. In the case
12 Vovin informs me (p.c. December 2017) that he does not strongly support that hypothesis anymore. 13 One motivation for reconstructing earlier stops was to make some comparisons of Japanese with other languages seem phonetically closer. Phonetic similarity is however irrelevant: only regularity of correspondences matter, and whether we reconstruct these segments as glides or stops has thus no bearing on the question of genetic comparisons.
2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan
47
of Amami-Yoron Higashi-ku too, positing a change ✶b > ɸ in front of ✶o has no explanation, while a change ✶w > ɸ has several cross-linguistic parallels.14 Moreover, the lenition hypothesis needs to assume several fortition processes, ✶ b > w > ɡ in Northern Ryukyuan, which of course nullifies any argument about fortition being less frequent and natural than lenition. Fortition of glides is actually not unfrequent, and specifically the change w > b in initial position is far from being rare cross-linguistically (Kümmel 2007: 159–161).15 The absence of ✶bu in the proto-system posited by the lenition hypothesis is also problematic since there is no obvious explanation for it. Typologically, bu is not a disfavored sequence, while on the other hand a sequence wu (distinct from u) is often absent in phonological systems due to the lack of perceptual distance between the onset and the vowel (Maddieson and Precoda 1992). Reconstructing earlier ✶w instead of ✶b thus provides a straightforward explanation for this systematic gap. The fact that some loanwords of Chinese origin also exhibit a w :: b correspondence adds further support to the fortition scenario (Karimata 1999, Pellard 2009: 354, Vovin 2010: 37). For example, the word ‘teacup, bowl’ (MC drae ’wanX 茶碗, JPN chawan) has a labial stop in Southern Ryukyuan (Miy.-Hir. & Yae.-Ish. tɕabaɴ, Yonag. sabaɴ). This can be easily explained if that word was borrowed after the loss of intervocalic ✶-w- and before the fortition of ✶w. The lenition hypothesis has no explanation to offer here. Moreover, some words fail to exhibit a stop when in the second position of compounds, like ‘husband’ in the compound ‘couple, husband and wife’ (Table 7, Bentley 2008: 199–201, Pellard 2009: 354). Nakama (1996) rightly points out that compounds often preserve archaic forms and that compounds in Miyako usually show a b, but all the examples he adduces are transparent ones, while ‘couple’ contains a bound form ‘female’ and often an archaic suffix, which shows it has been lexicalized early. Such data goes against Nakama’s conclusion, and we can assume that ✶w underwent fortition to b in initial position and within transparent compounds, but not in lexicalized ones, where it disappeared like other instances of medial ✶-w-. Table 7: Initial ✶w- in free and bound forms. ‘husband’ ‘couple’
JPN
pR
wopito (EMJ) me-woto (LMJ)
✶
woto me-woto(-ra)
✶
Miy.-Hir.
Yae.-Ish.
Yonag.
butu mjuːtu(ɾa)
búdù mjúːtú
bùtú mìtúdâ
14 Compare with the synchronic variation observed between wu and ɸu for pR ✶wo in Tokunoshima dialects (e.g. Amami-Okazen ‘husband’ {w,ɸ}ùtúː < pR ✶woto :: EMJ wopito, personal field notes) and with the diachronic change ✶w > (✶β >) f found in Old Irish (Hock 1991: 163). 15 No phonetic motivation seems to have been proposed for initial glide fortition, but Hock (1991: 162) proposes instead a structural explanation, i.e. a generalisation of the principle of initial strength from obstruents to sonorants.
48
Thomas Pellard
Whitman (1985: 17) and Vovin (2010: 36–40) adduce interesting data from the Miyako-Hirara dialect of Miyako, where either v(u) or u corresponds to Japanese u, and they propose to reconstruct ✶bu for the former and ✶u for the latter. However, a word of caution is needed since most other sources on the Miyako-Hirara dialect (e.g. Nevskij 1922) reveal important discrepancies in the relevant forms. The data presented seem to contain at least some transcriptions which do not faithfully reflect the synchronic phonology of the Miyako-Hirara dialect but have been etymologically re-spelled in order to make them more similar to their Japanese cognates. For instance, the very syllable vu in question here is probably a dubious transcription of a syllabic v, and the letter u is added because the Japanese cognate usually exhibits such a vowel. Among the unexpected cases of initial v adduced, most do not have a v in standard sources (‘song’, ‘move’, ‘doubt’, ‘back’). The remaining examples are marked as subject to variation between v and u (‘quail’, ‘inside’) and/or almost all involve a dental stop in the second syllable. Determining whether there is actually a special correspondence in the Miyako-Hirara dialect for Japanese ut- and ud- will require more investigation. There are two problematic examples left. First, the word ‘rabbit (zodiac sign)’ is not attested in most sources on Miyako-Hirara, but in Miyako-Irabu, Ōgami, Tarama, and other Miyako dialects it is uː, with a vowel, as expected, and not a fricative. Second, while Whitman (1985: 17) and Vovin (2010: 37–38) reconstruct Miy.-Hir. vː :: JPN ur-i ‘sell’ as ✶bur-i, vs. Miy.-Hir. uːɿ :: JPN uri ‘gourd’ as ✶uri, it is possible to reconstruct ✶ur-i vs. ✶ori instead (Pellard 2009: 352, cf. MK wǒy < ✶ò{r,n}í ‘gourd’). For the time being, it is thus safer to not consider the Miyako-Hirara data as evidence supporting the reconstruction of pJR ✶b. 4.1.2 Reconstructing initial ✶jThe Southern Ryukyuan language Yonaguni is unique in exhibiting a stop d- corresponding to initial j- in other Japonic languages (Table 8). As with the case of ✶w, evidence from both loanwords and compounds can be adduced against the reconstruction of ✶d. Table 8: Initial d- in Yonaguni.
‘mountain, forest’ ‘hot water’ ‘night’
OJ
Ama.-Yam.
Oki.-Yon.
Miy.-Hir.
Yae.-Ish.
Yonag.
yama yu yworu
jama juː juɾu
jàmǎː jǔː jùɾǔː
jama juː juːɿ
jámá júː júːɾɿ ́
dàmà dùː dùɾù
In Yonaguni, several loanwords of Chinese origin exhibit a correspondence j :: d in initial position (Table 9 Whitman 1985: 19, Pellard 2009: 354–355, Vovin 2010: 41–42), which can only be explained by a late fortition. The d- ~ -j- alternation seen in examples such as daː ‘house’ ~ baja ‘my/our house’ (< ✶ba-ja ‘we-house’), ndja ‘your house’
2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan
49
(< ✶ndi-ja ‘you-house’) also supports the fortition hypothesis (Nakama 1996: 214, Bentley 2008: 166, Pellard 2009: 354). Table 9: Loanwords of Chinese origin in Yonaguni. MC ‘vegetable’ ‘desire’ ‘carelessness’
野菜 欲 油断
yaeX tshojH yowk yuw dwanX
JPN
Yonag.
yasai yoku yudan
dasai duɡu dudaɴ
Here again, the lenition hypothesis posits a system without a syllable ✶di, since there is no yi in Japanese16 and no correspondence JPN i :: Yonag. di, but is unable to explain that gap. From the point of view of the fortition hypothesis, this systematic gap is easily accounted for by a perceptual constraint and is also supported by typological evidence. A direct change ✶j > d seems to be cross-linguistically rather rare, but ✶j > { ɟ, ʝ,(d)ʑ,(d)ʒ} is cross-linguistically common (Kümmel 2007: 159–161, 165–167), and it is important to note that Yonaguni has undergone several consonant fortitions (Table 10), including ✶ (d){z,ʑ} > d. We thus only need to postulate a simple and common change ✶j > ✶(d)ʑ in Yonaguni in order to explain the observed correspondence, since the second part of the change, namely ✶(d){z,ʑ} > d, needs to be posited anyway. This change fits neatly within the general trend of fortition seen in Yonaguni. Table 10: Fortitions in Yonaguni. fortition
Yonag.
JPN
gloss
(d){z,ʑ} > d ki (> {✶tɕi,✶tsɿ}) > ti ✶ si > tɕ⁽ˀ⁾i ✶ su > tɕ⁽ˀ⁾i
kʰàdí ìtî ɸùtɕí tɕˀìnî
kaze iki posi sune
‘wind’ ‘breath’ ‘star’ ‘shank’
✶ ✶
Moreover, there is at least one piece of philological evidence for the fortition hypothesis ✶ j > d (Bentley 2008: 167, Vovin 2010: 43–44), which also happens to confirm the fricative intermediary stage posited above. The name of Yonaguni Island, modern dunaɴtɕˀima, is recorded in the 15th c. Korean chronicle Sŏngchong sillok (10th year [1479], 6th month, 10th day) as 閏伊是麼, read zyun.i sima in MK. This testifies of the intermediary stage of the fortition process ✶j > ✶(d)ʑ > d.
16 A distinct syllable yi might be reconstructed, at least in pre-OJ, mainly from morphophonological alternations, but it was not distinguished in writing (Ramsey and Unger 1972, Vovin 2009: 420–426).
50
Thomas Pellard
4.2 Initial ✶p-
The main arguments for reconstructing Modern Japanese (ModJ) h as earlier ✶p are well known since Ueda (1898). Comparative evidence from Ryukyuan played an important role in Ueda’s demonstration, and indeed Ryukyuan fully exemplifies the different stages of the development chain posited, i.e. p > pᶲ > ɸ > h (Table 11). Though determining when the lenition of initial ✶p started in Japanese is not a simple matter, its reconstruction as a stop ✶p and the archaic character in this matter of Ryukyuan enjoys a broad consensus (e.g. Hashimoto 1928, Arisaka 1955: 569–571, Hattori 1976, M. Nakamoto 1976: 165–184, Whitman 1985: 17, Nakama 1992: 82–92). Table 11: Initial ✶p- in Ryukyuan. ‘wing’ ‘day’ ‘spatula’ ‘winter’ ‘bone’
OJ
Ama.-Yam.
Oki.-Kud.
Oki.-Ize.
Oki.-Yon.
Miy.-Hir.
Yonag.
pane pi pyera puyu pone (EMJ)
hane çiː heɾa, ɸeɾa ɸuju ɸunɨ
pᶲani pˀiː pᶲiɾa pˀuju pᶲuɴ
ɸání ɸíː ɸíɾá ɸújúː ɸúní
pàní pˀíː pʰíɾà pùjúː púnì
pani pɿː piɾa fuju puni
hàní tɕˀíː çìɾâ ɸùjú ɸùnî
Still, doubts have been expressed about the archaic character of Ryukyuan p, and it has been suggested that it is the result of the late fortition of a fricative ✶ɸ (Yanagida 1989, K. Nakamoto 2011).17 However, that rather speculative hypothesis is not backed up by concrete evidence and suffers from many problems (Karimata 2009, Pellard 2016a). For instance, while fortitions of voiced approximants like w and j are cross-linguistically well attested (§4.1), the fortition of a voiceless fricative {ɸ,f} > p in initial, prevocalic, position is not,18 and the fortition hypothesis requires that such a rare sound change happened several times in Ryukyuan. On the other hand, the sound change p > (pᶲ >) {ɸ,f} > h is amply attested cross-linguistically (Kümmel 2007: 57, 65–66, 68, 194). The hypothesized link between a fortition ɸ > p and the tensification of consonants due to vowel raising is easily falsified by looking at the Okinawa-Kudaka and Okinawa-Nakijin Yonamine data, where we not only found a tense pˀ before originally high vowels as expected by the fortition hypothesis, but also a lax pʰ before originally non-
17 One motivation was to show that Ryukyuan and Japanese separated late, after the 8th c. This of course fails to account for all the evidence other than ✶p summarised in Pellard (2015, 2016a). More importantly, even if ✶p had lenited to ✶ɸ in pR, this could be an innovation parallel to Japanese, especially given the cross-linguistic frequency of that change, and it would have no bearing on the dating of the Japanese-Ryukyuan split. 18 The rare examples of {ɸ,f} > p found in Kümmel (2007: 147–148) actually involve a merger of the fricative with a pre-existing p, which would not be the case of Ryukuan under the fortition hypothesis. The two would be structurally completely different.
2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan
51
high vowels. The fact that Yonaguni, the Ryukyuan language which has undergone the greatest number of fortitions, has a reflex h also constitutes evidence against the fortition hypothesis. The putative cases of ✶k{u,o}{r,w} > ✶kw > ✶ɸ > {p,ɸ,h} adduced by K. Nakamoto (2011) can be instead interpreted as cases of ✶k{u,o}{r,w} > ✶kw > ✶p > {p,ɸ,h}, a typologically much more plausible sound change (Kümmel 2007: 274).
4.3 Word-final nasals Southern Ryukyuan exhibits word-final nasals without any correspondent in Japanese, and these have been claimed to go back to pJR (Starostin 1975, Murayama 1981). However, though the discussion has usually centered around the Hateruma dialect of Yaeyama, unexpected word-final nasals are also found19 in Yonaguni and, even more interestingly, in the Shiraho dialect, which is a subdialect of Hateruma spoken on Ishigaki Island.20 A comparison of the three (Table 12) reveals that they all exhibit unexpected word-final nasals, but not in the same words. Crucially, though both Yaeyama-Hateruma and Yaeyama-Shiraho show a correspondence for some words, the final nasals of Yonaguni never correspond to those of either Yaeyama-Hateruma or Yaeyama-Shiraho. Table 12: Word-final nasals in Yaeyama-Hateruma, Yaeyama-Shiraho, and Yonaguni.
me (OJ) kozo (OJ) patwo (OJ) ne (OJ) pi (OJ) uma (OJ) pitapi (EMJ) naba (LMJ) keburi (OJ) kwo (OJ) tunwo (OJ) nuka (EMJ)
One could theoretically reconstruct several different word-final nasals in pJR in order to try account for this situation, but reconstructing here four different nasals seems unrealistic. Anyway, this would not resolve the parsimony problem involved here: if we 19 Such unexpected word-final nasals are found sporadically in some other dialects too (Martin 1987: 74–75, Wayne Lawrence, p.c. December 2017). 20 After its destruction by a giant tsunami in 1771 and the death of almost all its inhabitants, the Yaeyama-Shiraho village was repopulated by a group of Yaeyama-Hateruma Islanders (Karimata 2008: 2).
52
Thomas Pellard
reconstruct these word-final nasals at the pJR level, then we need to assume they have been independently lost in Japanese and all other Ryukyuan branches and subbranches. This suffices to cast a reasonable doubt on the antiquity of such nasals. Moreover, the same root in Yaeyama-Hateruma sometimes fails to exhibit a final nasal in compounds (Oyler 1997: 89–102, Bentley 2008: 118, Chart 40). Compare for instance miɴ ‘eye’ with miː-kaŋɡaɴ ‘glasses’ (‘eye-mirror’), sɿkɿɴ ‘moon, month’ with mata-sɿkɿ ‘next month’ (‘again-moon’), or mmaɴ ‘horse’ with mma-nu-kaɴ ‘mane of a horse’. Oyler (1997) proposed an alternative explanation of morphological reanalysis and analogy. He links the development of word-final nasals to the frequent attachment of the focus marker =ndu (< ✶=nu=du ‘NOM=FOC’) to nominals. Aso (2010: 217) mentions that this focus marker has an allomorph =du after nasal-ending words, which adds credence to the hypothesis of Oyler. However, such a scenario is hard to postulate for Yonaguni since its focus marker is simply =du (Yamada, Pellard and Shimoji 2015). On the other hand, Karimata (2010: 5–6) proposes a phonetic explanation for these excrescent nasals and links their emergence to the strong aspiration and devoicing (or breathiness) found in Yaeyama-Hateruma. Devoicing would trigger nasalization, which would get then reinterpreted as a full nasal consonant, e.g. paː > pḁː > pãː > paã > paɴ. However, that explanation is problematic for several reasons. First, it does not solve the problem of the irregular character of these nasals. Second, it fails to explain the existence of excrescent nasals in words where neither aspiration nor devoicing is expected, such as ✶uma > mmaɴ. Last, such an explanation cannot be applied to the case of Yonaguni since it lacks the Yaeyama-Hateruma-type devoicing/breathiness.
5 Word-prosody (accent and tone) The comparison of the Ryukyuan prosodic systems enables us to reconstruct two classes for monosyllables, three for disyllables, and three or possibly four for trisyllables in pR (Hattori 1958, Matsumori 1998, 2000a, 2000b, 2001, 2012, Uwano 2000). These classes are customarily referred to by the letters A, B, C, and D. Turning to the comparison with Japanese,21 the correspondences of monosyllables are straightforward and involve a simple merger of classes 1.1 and 1.2 in pR. For disyllables and trisyllables, the correspondences are more complex (Figure 1). Several Ryukyuan varieties present two different correspondences for the same Japanese tone class, and the distinction can be realized as a difference of tone, vowel length, or both (Table 13).22
21 See Martin (1987), de Boer (2010) and Matsumori (this volume) for a more detailed presentation of the Japanese material. Note that the original class 3.3 is nowadays rejected due to too many irregularities, but the numeration has not been changed. 22 Note that in Amami-Asama, reflexes of the C-class dissyllables have a HːL pattern in case of a glottalized onset and a LːH pattern otherwise. See Uwano (2002) on the emergence of such new tonal distinctions.
2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan
53
Japanese
Ryukyuan (A) Monosyllables
(B) Disyllables
(C) Trisyllables
Figure 1: Tone class correspondences between Japanese and Ryukyuan. Table 13: Reflexes of classes 2.4 and 2.5 in Ryukyuan.
2.4 :: B 2.5 :: B 2.4 :: C 2.5 :: C
‘corner’ ‘rain’ ‘breath’ ‘shadow’
Ama.-Asa.
Ama.-YHig. Oki.-Yon.
Oki.-Shur.
Miy.-Nis.
Yonag.
kʰàdǔː ʔàmɨ̌ː ʔíːkì kʰàːɡɨ ́
hàdû ʔàmî ʔíkì hágí
kàdù ʔàmì ʔìːtɕì kàːɡì
kádu ámì ítsɿ káɡí
kʰàdù àmì ìtî kʰàŋî
LRː LRː HːL, LːH HːL, LːH
LF LF HL HL
hàdǔː ʔàmǐː ʔítɕˀì háɡì
LRː LRː HL HL
LL LL LːH LːH
HL HL HH HH
LL LL LF LF
Until recently, the very existence of two different reflexes in Ryukyuan for the classes 2.3, 2.4, 2.5, 3.4, 3.5, 3.6, and 3.7 (Figure 1) had been ignored or even denied in Japanese dialectology, and Ryukyuan was presented as deriving from a system similar, if not identical, to that of the Ōita dialect of Kyushu (e.g. Kindaichi 1960, Uemura 1997). However, it has been clearly and repeatedly demonstrated (Hattori 1958, 2018) that there is not a single dialect anywhere in the Ryukyus exhibiting such a system, and cumulative empirical evidence has confirmed the correspondences of Figure 1 (e.g. Matsumori 1998, 2000b, 2001, Uwano 2000, Igarashi et al. 2012). The only really problematic correspondences are those symbolized by dashed lines in Figure 1. There are only few 2.3 words actually corresponding to the C-class, and the D-class contains only a handful of words and is only distinguished in Amami. For other one-to-many correspondences, since no conditioning factor has been found hitherto to explain the correspondences of Figure 1, the only methodologically valid solution is to posit more tonal distinctions in pJR. Of course, some of those extra classes might be the result of secondary developments, but there also is a possibility that it is some of the Japanese classes that are secondary, and not the Ryukyuan ones. In any case, the reality of the tonal correspondences cannot be simply ignored when reconstructing pJR.23 Several scholars have argued that the vowel length found in C-class nouns is original and goes back to pJR (Hattori 2018, Vovin 1993, Shimabukuro 2007), or at least to protoNorthern Ryukyuan (Matsumori 1998, Lawrence 2016).24 On the other hand, a strong 23 Igarashi and Hirako (2016) have recently claimed that the Japanese dialect of Kishima in Saga prefecture has two different reflexes for 2.5 words, and that this distinction corresponds to the Ryukyuan B- vs. C-class. Though a fully detailed account has yet to be published, this discovery potentially has major implications for Japanese-Ryukyuan historical linguistics. 24 Lawrence (p.c. December 2017) actually considers vowel length to be “a secondary development, occurring at the locus of pitch change”. He adds that “for two-mora nouns, A-class and C-class would
54
Thomas Pellard
argument can also be made on the basis of comparative and internal reconstruction for a secondary development of vowel length due to tonal or accentual shift (de Boer 2010: 208–246, Pellard 2016b, Matsumori 2017). The general absence of C-class verbs and adjectives would be surprising if the C-class originated from earlier vowel length, but on the other hand verbs and adjectives often display fewer tonal or accentual patterns than nouns, as is the case in many Japanese dialects, including EMJ and ModJ. The lack of long vowels in compounds and derivations from C-class roots, while etymologically long vowels are usually preserved, and in disyllables reduced to monosyllables due to the loss of an intervocalic consonant indicates that the C-class vowel length is probably a secondary development (Table 14, Pellard 2016b). Table 14: Long and short vowels in Northern Ryukyuan. pR Ama.-Asa.
6 Grammar 6.1 Verbs The major problem for the comparative grammar of the Japanese-Ryukyuan languages is that the modern Ryukyuan verb paradigms often exhibit forms that do not strictly phonologically correspond to the Japanese ones. For instance, the conclusive and adnominal forms in Table 1525 are unexpected, e.g. the Okinawa-Shuri forms should be identical with the Japanese ones, but they exhibit an unexpected palatalization of the stem, as well as endings absent from the Japanese forms.
have had the same phonetic contour (although phonologically distinct), and vowel length developed in the first place to allow a phonetic distinction between the two to be made.” 25 For the sake of clarity, the Amami-Yamatohama, Okinawa-Nakijin Yonamine, and Okinawa-Kudaka adnominal forms quoted are not the usual ones used in adnominal constructions but those used in kakari-musubi constructions, where other varieties use the adnominal form.
2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan
55
Table 15: Basic forms of ✶kak- ‘write’ in Northern Ryukyuan. Form
OJ
Ama.-Yam.
Ama.-YHig.
Oki.-Yon.
Oki.-Shur.
Oki.-Kud.
NEG INF
kakanu kaki
kʰakaɴ kʰaki
kakannu kaki
hakaɴ hatɕi-
kakaɴ katɕi
hakaɴ haki-
CONCL ADN
kaku kaku
kʰakuɴ kʰakuɾu
kakjuɴ kakjuɾu
hatɕuɴ hatɕuːɾu
katɕuɴ katɕuɾu
hakiɴ hakiɾu
PROV IMP
kakeba kakye
kʰakɨba kʰakɨ
kakiba kaki
hakiːba haki
kakiwa kaki
hakiba haki
It is now well-established that the conclusive and adnominal forms of Northern Ryukyuan derive from the univerbation of an earlier imperfective construction where a main “infinitive” verb (✶-i) was followed by the stative verb ✶wor- ‘be staying’ (Hattori 1977, Uchima 1984, Serafim 2007). This univerbation of ✶-i + wor- explains the otherwise unexpected palatal element usually seen in Northern Ryukyuan conclusive and adnominal forms. Though it is plausible that the Japanese forms might also come from the fusion of an infinitive with an auxiliary verb,26 the hypothesis that both the Japanese and Northern Ryukyuan forms are cognate (i.e., homologies, Thorpe 1983: 238–258, Vovin 2009: 609–611, 2010: 82–83) rather than parallel innovations (homoplasies), faces important problems (Serafim 2007). The crucial evidence is found in Old Okinawan (Omoro sōshi, OS), which preserves the original simplex verb forms, without palatalization or special endings, that directly correspond to the Japanese ones (Takahashi 1991, Karimata 2015), like the adnominal form of ‘bloom’ 〈saku〉 (OS 14.989) vs. Oki.-Shur. satɕuɾu. Old Okinawan also exhibits the complex extended forms, like the imperfective adnominal form of ‘blow’ 〈fukiyoru〉 (OS 11.618) that correspond to Oki.-Shur. ɸutɕuɾu (now replaced by modern ɸutɕoːɾu). This indicates that the modern Northern Ryukyuan forms are the result of a late process, which is independent from the possible ultimate origins of the Japanese verb forms. On the other hand, several Northern Ryukyuan varieties possess an adnominal form restricted to some constructions, especially before more or less grammaticalized formal nouns. These forms do not include the stative auxiliary (Nakasone 1960) and reconstruct with a final ✶-o in pR (Table 16). Such archaic forms compare well with the EOJ and Hachijō adnominal forms in -o, in contrast with the -u found in OJ and most Japanese dialects (Pellard 2008, Hirako and Pellard 2013), and suggest the existence of a pJR adnominal marker ✶-o.27
26 This is the hypothesis developed by Ōno (1953), who identifies the auxiliary with the OJ verb wi-, conclusive u, ‘stay, sit’, the dynamic counterpart of the stative wor-. 27 Frellesvig’s (2012) hypothesis of a late development of the conclusive vs. adnominal opposition out of a single “finite” marker ✶-o is not incompatible with the above proposal. See also Osterkamp (2017) for new evidence of an adnominal ✶-o ending from non-Eastern OJ.
56
Thomas Pellard
Table 16: Okinawa-Nakijin Yonamine adnominal forms. old ADN ✶-o
ADN ‘stand’ ‘beat’
tʰatɕˀuːnu kʰuɾuːɕunu
tʰatˀu (†tʰatɕˀi) kʰuɾuːɕu (†kʰuɾuːɕi)
cf. ✶u
‘summer’ nàtɕˀíː < ✶natu ‘mortar’ ʔúɕì < ✶{u,o}su
Though Karimata (2014) argues for the existence of an adnominal ending -o distinct from a conclusive -u in the OS, the evidence adduced is far from being satisfactory both qualitatively and quantitatively. For instance, the orthography of the OS shows variation between 〈u〉 and 〈o〉 except after coronal obstruents, so that if an adnominal ending -o and a conclusive ending -u could have been strictly distinguished anywhere, it would have been after such consonants. Yet, no o-adnominal form is found for verbs with a stem ending with a coronal obstruent. The origins of the Southern Ryukyuan verb forms and paradigms (Table 17),28 which can be quite complex (Pellard and Yamada 2017), are more difficult to elucidate. Some finite forms seem to be directly comparable to those of Japanese, but there is evidence that at least some of them derive from the same complex auxiliary forms as Northern Ryukyuan (Thorpe 1983: 153–173, Pellard 2009: 34–346). Table 17: Basic forms of ✶kak- ‘write’ in Southern Ryukyuan. form
OJ
Miy.-Ira.
Yae.-Ish.
Yonag.
NEG INF CONCL ADN PROV IMP
kakanu kaki kaku kaku kakeba kakye
kaʕaɴ katsɿ-, kafukafum kafu kakiba kaki
kakanu kakɿ kakuɴ kaku kakjaː kaki
kʰaɡanuɴ kʰati kʰaɡuɴ kʰaɡu kʰaɡuba kʰaɡi
6.2 Adjectives Like verbs, the Ryukyuan adjectival paradigms are not directly cognate with the Japanese ones (Table 18, Uchima 1984, Nakama 1992: 555–601). Most Ryukyuan inflected forms incorporate the auxiliary ✶ar- ‘be’ attached to either a nominalized form marked with ✶-sa or an adverbial form marked with ✶-ku, and the inflectional endings are not always cognate with Japanese.
28 For the sake of clarity, I have distinguished between the Miyako adnominal and conclusive forms, but the distinction is not actually a syntactic one but rather a modal one (M. Shimoji 2009).
2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan
57
Table 18: Basic forms of ✶taka- ‘high’ in Ryukyuan. form
OJ
Ama.-Yam.
Oki.-Shur.
Miy.-Ira.
Yae.-Ish.
Yonag.
INF CONCL ADN PROV
takaku takasi takaki takakyeba
tʰahasa tʰahasaɴ tʰahasaɾu tʰahaɾɨba
takaku takasaɴ takasaɾu takasaɾeː
taʕafu taʕaham taʕaɿ taʕahaɾiba
takasaːɾi takasaːɴ takasaːɾɿ takasaːɾjaː
tʰaɡaɡu tʰaɡaɴ tʰaɡaɾu tʰaɡaɾja
6.3 Core case markers The two OJ case markers ga and no were both used to mark genitive modifiers and subjects in dependent clauses (and more rarely in main clauses), and the distinction between the two has been debated (Vovin 2005: 118–133, Frellesvig 2010: 127–129). There is some overlap in their usage, except after pronouns referring to humans (only ga) and demonstratives (only no), and the distinction between the two seems to have been based on the properties of the host noun. Several Ryukyuan varieties preserve both markers, and the distinction between the two depends on a hierarchy of nominals based on their referentiality and identifiability. Typically, personal pronouns and address terms, i.e. kinship terms and titles that can be used not only to refer to but also to address somebody, are marked with the cognate of ga, and other nominals with the cognate of no, though in some cases nominals located highest on the hierarchy can have a zero-marked form (Table 19). Table 19: Ryukyuan nominative/genitive markers.
The OJ accusative marker wo has well attested reflexes in Ryukyuan too, though some varieties have no accusative marker, and they thus have a marked nominative system (e.g. Oki., Yonag.). The reconstruction of its proto-form is not without problems at first sight, since many varieties exhibit a reflex ju. The palatal glide is however epenthetic, as is clear from Miyako accusative forms such as panau or panoː ‘flower’, in which there would be no reason for an original palatal glide to be lost intervocalically. The development of the accusative marker is paralleled by that of the topic marker pJR ✶pa > pR ✶ wa, which also has a reflex ja, but whose original onset ✶-w- is here recoverable from the alternations of a-ending stems in Northern Ryukyuan (Table 20). This suggests that the pR form of the accusative marker should thus reconstruct as ✶(w){u,o}.
bare stem topicalized form cf. ‘millet’ cf. ‘foam’ vs. ‘design’
pana pana pa apa awa aya
pana panoː oː oː aja
6.4 Kakari-musubi The focus construction type known in Japanese grammar as kakari-musubi, where the presence of a focus-marked constituent correlates with a main predicate in the adnominal or exclamatory instead of the conclusive form, is well attested in Ryukyuan. Shinzato and Serafim (2013) undertake a detailed comparison of such kakari-musubi patterns in both Japanese and Ryukyuan, though their Ryukyuan data are limited to Old and Modern Okinawan.29 They reconstruct several kakari-musubi constructions in pR and pJR, and retrace their development from a functional point of view. They also offer etymologies for the different focus markers involved. The pJR reconstruction of OJ {s,z}o :: pR ✶do, a focus marker cooccurring with a main predicate in the adnominal form, is not straightforward due to the phonological mismatch between the Japanese and Ryukyuan forms. It is possible that in this case OJ is more innovative than Ryukyuan in view of the stop initial of the EOJ cognate to (MYS 14.3409, 14.3561, 20.4385, 20.4430).30 The discrepancy in the voicing is only a minor problem, especially if we compare with the similar case of the interrogative marker OJ ka :: pR ✶ga. The pR nominalizer ✶so also plays a role in the historical investigation of kakari-musubi since Shinzato and Serafim (2013) propose that it is cognate with the second syllable of the OJ focus marker koso, whose first syllable is said to be the proximal demonstrative ko, i.e. koso is etymologized as ‘this one’. In any case, the existence of straightforward cognates of the pR nominalizer ✶so in some Japanese dialects (Northeast of Kyushu and extreme West of Honshu) suggests it already existed in pJR (Shinzato and Serafim 2013: 160, 202–207).
29 See Karimata (2011) and Shimoji (2011) on kakari-musubi in Southern Ryukyuan. 30 Shinzato and Serafim (2013: 138–139) propose to reconstruct pJR ✶t(j)ə (t(j)ö) and subsequent affrication in order to account for OJ so. They also argue that the focus marker originates in the mesial demonstrative (OJ so) and suggest a comparison with the MK distal demonstrative tye.
2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan
59
7 Lexicon The Ryukyuan languages exhibit many lexical innovations, but they also preserve many archaic words and meanings, some of which have been lost in Japanese after the OJ period. The Ryukyuan data can thus often shed a new light on etymological problems, particularly when trying to discriminate between potential cognates and loanwords from other languages (Vovin 2010, Pellard 2017).
7.1 Personal pronouns Cognates of the two OJ first person pronouns a and wa are found in Ryukyuan (Table 21 M. Nakamoto 1983: 156–161, Uchima 1984: 75–86, 133–164), though no cognates of the OJ extended forms are and ware are found. On the other hand, direct cognates of the EOJ form wan{u,wo} (MYS 14.3476a, 14.3476b, 20.4358) are well attested. Though the distinction between the two pronouns a and wa is elusive in OJ (Vovin 2005: 217–219), the attestation of both roots in several Ryukyuan varieties constitutes evidence that these are not phonological variants and do not belong to different historical stages either. It is interesting to note that while in Northern Okinawan, ✶a is used with a collective plural meaning, in Southern Ryukyuan it is ✶wa that is used in plural forms, while in the Tarama dialect of Miyako, the difference between the two has to do with politeness. Table 21: Ryukyuan first person pronouns. pJR wa a
✶ ✶
Ama.-Yam.
Oki.-Yon.
Miy.-Ira.
Miy.-Tar.
Yae.-Ish.
Yonag.
waɴ, waː—
wànǔː, waːʔa- (PL)
pan- (PL) a-, anu
ba-, baɴ a-, aɴ
bánú —
bànû- (PL) ànû
Cognates of the OJ second person pronoun na are amply attested in Northern Ryukyuan, where they usually function as honorific second person pronouns (Table 22, M. Nakamoto 1983: 156–159, Uchima 1984: 87–93). It has been suggested that na was also used as a first person pronoun in OJ because the character 己 ‘self’ is sometimes read na, and because it is used in compounds with kin terms for which there are variants involving the first person pronoun wa (e.g. na-se ‘my beloved (man)’ MYS 14.3458, FK 8; cf. a-se ‘id.’ KK 29, wa ga se ‘id.’ MYS 1.19). Whitman (1999) suggests that OJ na has undergone an intrapersonal pronoun shift from first person to second person mediated by a reflexive stage. This hypothesis can find support in the existence in Southern Ryukyuan of reflexive or logophoric pronouns cognate with OJ na (Table 22).31 31 The form naː is not derived from naɾa since no such sound change is attested in those varieties. Thus -ɾa appears to be a suffix, though its meaning and function are unclear. It is nevertheless parallel with
60
Thomas Pellard
Table 22: Ryukyuan pronoun ✶na. 2nd person reflexive/logophoric
Ama.-Yam.
Oki.-Yon.
Miy.-Ira.
Miy.-Tar.
Yae.-Ish.
naɴ, naː—
náɴ, náː—
— naɾa
— naː, naɾa
— nâː, náɾà
7.2 Demonstratives While the OJ demonstrative system consisted into a basically binary opposition between proximal ko vs. non-proximal so, with distal ka being fully integrated only from the Early Middle Japanese (EMJ) period on (Frellesvig 2010: 139–140), the existence in Southern Ryukyuan of cognates of ka (Table 23, M. Nakamoto 1983: 168–186, Uchima 1984: 54–66, 100–105) suggests it should be reconstructed in pJR. Table 23: Ryukyuan distal demonstratives.
pronoun adnominal
Miy.-Ira.
Miy.-Tar.
Yae.-Ish.
Yonag.
kaɾi kanu
kaɭ kanu
káɾì kánù
kʰàɾí kʰànù
On the other hand, no cognate of the OJ mesial demonstrative so is found in Ryukyuan (Table 24, M. Nakamoto 1983: 168–186, Uchima 1984: 54–66, 100–105). The pR mesial root ✶{u,o} nevertheless corresponds to the Hachijō distal u- (Table 24, Hirako and Pellard 2013), which demonstrates its antiquity. Shinzato and Serafim (2013: 262–265) propose it was originally a reflexive deictic, related to the OJ reflexive pronoun ono, and that it replaced the original mesial in Ryukyuan. Table 24: Ryukyuan and Hachijō mesial demonstratives.
pronoun adnominal
Ama.-Yam.
Oki.-Yon.
Miy.-Ira.
Miy.-Tar.
Yae.-Ish.
Yonag.
Hach.
ʔuɾi ʔuɴ
ʔùɾíː ʔunu
uɾi unu
uɭ unu
úɾì únù
úː ùnù
uɾe uno
that of the pR second person pronoun ✶ura and little more mysterious than the suffix of the first person pronouns ✶wa(-no) and ✶a(-no).
2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan
61
Abbreviations Ama. Ama.-Ad. Ama.-Asa. Ama.-Shib. Ama.-Yam. Ama.-YHig. FK Hach. KK Miy. Miy.-Hir. Miy.-Ira. Miy.-Nis. Miy.-Tar. MYS Oki. Oki.-Ben. Oki.-Ize. Oki.-Kud. Oki.-Shur. Oki.-Yon. OS pJR pR Yae. Yae.-Hat. Yae.-Ish. Yae.-Shir. Yonag.
Note on transcriptions and symbols All modern linguistic forms are given in a unified broad phonetic transcription, without indication of predictable vowel devoicing. Tense (“glottalized”) consonants are uniformly marked by a following ˀ and lax stops by ʰ. The velar or uvular nasal coda is transcribed with ɴ, and the fricative vowel with the “apical vowel” symbol ɿ for lack of an official symbol, though it is not always apical. When a form lacks tone diacritics, it means either that the variety has no distinctive tone or accent, that the tone pattern of the form is unknown, or that tones are irrelevant for the issue under discussion. Japanese of all historical periods is transcribed according to the system of Frellesvig (2010), Middle Korean according to the Yale system, and Middle Chinese according to Baxter and Sagart (2014). The obelus symbol (†) marks an expected but unattested form, and
62
Thomas Pellard
chevrons (⟨. . .⟩) a purely orthographical transcription without any commitment to the actual pronunciation. Alternative possibilities are indicated with the set notation between curly brackets, correspondences between languages are noted with a double colon (::), and morphological or morphophonological alternants are separated by a tilde. Forms not attested in isolation are marked with a hyphen, and a long dash in tables indicates there is no attestation. The symbol # represents a word boundary.
References Arisaka, Hideyo. 1934. Kodai Nihongo ni okeru onsetsu ketsugō no hōsoku [Laws of syllable combination in Old Japanese]. Reprinted in Hideyo Arisaka 1944, Kokugo on’inshi no kenkyū [Studies on Japanese historical phonology], 101–114. Tokyo: Meiseidō Shoten. Arisaka, Hideyo. 1955. Jōdai on’in kō [A study of Old Japanese phonology]. Tokyo: Sanseidō. Aso, Reiko. 2010. Hateruma (Yaeyama Ryukyuan). In Michinori Shimoji and Thomas Pellard (eds.), An introduction to Ryukyuan languages, 189–227. Tokyo: Research Institute for Languages and Cultures of Asia and Africa. Baxter, William H. and Laurent Sagart. 2014. Old Chinese: A new reconstruction. Oxford: Oxford University Press. Bentley, John R. 2008. A linguistic history of the forgotten islands: A reconstruction of the proto-language of the Southern Ryukyus. Folkestone: Global Oriental. Chamberlain, Basil Hall. 1895. Essay in aid of a grammar and dictionary of the Luchuan language. (Transactions of the Asiatic Society of Japan 23, supplement). Yokohama: Kelly & Walsh. de Boer, Elisabeth. 2010. The historical development of Japanese tone. Wiesbaden: Harrassowitz. Frellesvig, Bjarke. 2010. A history of the Japanese language. Cambridge: Cambridge University Press. Frellesvig, Bjarke. 2012. Chronological layering in proto-Japanese and pre-Old Japanese verb inflection. In Bjarke Frellesvig, Jieun Kiaer and Janick Wrona (eds.), Studies in Japanese and Korean linguistics, 28–47. München: Lincom Europa. Frellesvig, Bjarke and John Whitman. 2008. Evidence for seven vowels in proto-Japanese. In Bjarke Frellesvig and John Whitman (eds.), Proto-Japanese: Issues and prospects, 15–41. Amsterdam: John Benjamins. Fukuji, Tomokuni and Shin’ichi Kajiku. 2012. Kudaka-jima hōgen kiso goi jiten [Basic vocabulary dictionary of the Kudaka Island dialect]. Tokyo: Hōsei Daigaku Okinawa Bunka Kenkyūjo. Hashimoto, Shinkichi. 1928. Hagyō shiin no hensen ni tsuite [On the changes of the h consonant]. In Sanki Ichikawa (ed.), Okakura sensei kinen ronbunshū [Festschrift for Professor Okakura], 196–207. Tokyo: Okakura Sensei Kanreki Shukuga Kai. Hattori, Shirō. 1932. “Ryūkyūgo” to “kokugo” no on’in hōsoku [The sound laws of Ryukyuan and Japanese]. Reprinted in Shirō Hattori 1959, Nihongo no keitō [The genealogy of the Japanese language], 296–361. Tokyo: Iwanami Shoten. Hattori, Shirō. 1958. Amami guntō no shohōgen ni tsuite: Okinawa, Sakishima shohōgen to no hikaku [On the dialects of the Amami Islands]. Reprinted in Shirō Hattori 1959, Nihongo no keitō [The genealogy of the Japanese language], 275–294. Tokyo: Iwanami Shoten. Hattori, Shirō. 1959. Nihongo no keitō [The genealogy of the Japanese language]. Tokyo: Iwanami Shoten. Hattori, Shirō. 1976. Ryūkyū hōgen to hondo hōgen [Ryukyuan and Mainland dialects]. In Iha Fuyū Tanjō Hyakunen Kinenkai (ed.), Okinawagaku no reimei [The dawn of Okinawan studies], 7–55. Tokyo: Okinawa Bunka Kyōkai. Hattori, Shirō. 1977. Ryūkyū hōgen dōshi “shūshikei” no tsūjiteki henka [Diachronic changes of the conclusive form of verbs in Ryukyuan]. Gengo Kenkyū 72. 19–28.
2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan
63
Hattori, Shirō. 2018. Nihon sogo no saiken [The reconstruction of proto-Japanese-Ryukyuan]. With annots. by Zendō Uwano. Tokyo: Iwanami Shoten. Hayata, Teruhiro. This volume. Phonological reconstruction. In Bjarke Frellesvig and Satoshi Kinsui (eds.), Handbook of historical Japanese linguistics. Berlin and Boston: De Gruyter Mouton. Heinrich, Patrick, Shinsho Miyara and Michinori Shimoji (eds.). 2015. Handbook of the Ryukyuan languages: History, structure, and use. Berlin: De Gruyter Mouton. Hirako, Tatsuya and Thomas Pellard. 2013. Hachijōgo no furusa to atarashisa [Old and new aspects of the Hachijō language]. In Nobuko Kibe (ed.), Shōmetsu kiki hōgen no chōsa, hozon no tame no sōgōteki kenkyū: Hachijō hōgen chōsa hōkokusho [General study for research and conservation of endangered dialects in Japan: Research report on Hachijō dialect], 47–67. Tachikawa: National Institute for Japanese Language and Linguistics. Hirayama, Teruo (ed.). 1988. Minami Ryūkyū no hōgen kiso goi [The basic vocabulary of the dialects of the Southern Ryukyus]. Tokyo: Ōfūsha. Hirayama, Teruo (ed.). 1992–1994. Gendai Nihongo hōgen daijiten [Great dictionary of the modern Japanese dialects]. 9 vols. Tokyo: Meiji Shoin. Hock, Hans. 1991. Principles of historical linguistics. Berlin: Mouton de Gruyter. Hokama, Shuzen (ed.). 2000. Omoro saushi [Collection of omoro songs]. 2 vols. Tokyo: Iwanami Shoten. Igarashi, Yōsuke and Tatsuya Hirako. 2016. “Kata, tane, ase, ame” to “iki, fune, oke, nabe” ga akusento de kubetsu sareteiru Nihongo hondo hōgen: Saga-ken Kishima hōgen to Ryūkyūgo no hikaku [A Mainland Japanese dialect where “shoulder, seed, sweat, rain” and “breath, boat, bucket, pot” are distinguished by their accent: A comparison of the Kishima dialect of Saga prefecture and Ryukyuan]. Paper presented at the Thirtieth general meeting of the Phonetic Society of Japan, 17–18 September. Igarashi, Yōsuke, Yukinori Takubo, Yuka Hayashi, Thomas Pellard and Tomoyuki Kubo. 2012. Ryūkyū Miyakogo Ikema hōgen no akusento wa sankei deatte nikei dewanai [The Ikema dialect of Miyako Ryukyuan has a three-, not two-, pattern accent system]. Onsei Kenkyū 16(1). 1–15. Izena-jima Hōgen Jiten Henshū Iinkai (ed.). 2004. Izena-jima hōgen jiten [Dictionary of the Izena Island dialect]. 2 vols. Izena: Izena-son Kyōiku Iinkai. Karimata, Shigehisa. 1999. Onsei no men kara mita Ryūkyū shohōgen [The Ryukyuan dialects seen from the point of view of phonetics]. In Gengogaku Kenkyūkai (ed.), Kotoba no kagaku [The science of language], vol. 9, 13–85. Tokyo: Mugi Shobō. Karimata, Shigehisa. 2008. Ryūkyū Yaeyama hōgen no hikaku rekishi hōgengaku ni kansuru kisoteki kenkyū [Basic research on the historical comparative dialectology of the Ryukyuan Yaeyama dialects]. Research report. Nishihara-chō: University of the Ryukyus. Karimata, Shigehisa. 2009. Ryūkyūgo no ɸ > p no kanōsei o kangaeru: Nakamoto Ken “Ryūkyū hōgen no hagyō shiin p-on” e no toi [Examining the possibility of a change ɸ > p in Ryukyuan: Questioning Ken Nakamoto’s “The consonant p in Ryukyuan dialects”]. Okinawa Bunka 43(1/105). 1(78)–19(60). Karimata, Shigehisa. 2010. Minami Ryūkyū hōgen ni okeru dōka to ika [Assimilation and dissimilation in Southern Ryukyuan dialects]. Nihon Tōyō Bunka Ronshū 16. 1–38. Karimata, Shigehisa. 2011. Ryūkyū hōgen no shōtenka joji to bun no tsūtatsuteki na taipu [Focus marking particles and modality types in the Ryukyuan language]. Nihongo no Kenkyū 7(4). 69–82. Karimata, Shigehisa. 2014. Rentaikei gobi kara mita Omoro saushi no o-dan to u-dan no kana no tsukaiwake [The distinction between o-kana and u-kana characters as viewed from the verbal adnominal endings in the Omoro sōshi]. Okinawa Bunka 48(2). 187–198. Karimata, Shigehisa. 2015. Omorogo no dōshi shūshikei: Seimitsu na yomi o mezashite [The conclusive form of verbs in the language of the Omoro sōshi: Aiming for an accurate reading]. Ryūkyū Ajia Bunka Ronshū 1. 33–103. Karimata, Shigehisa. 2008. Ryūkyū Yaeyama hōgen no hikaku rekishi hōgengaku ni kansuru kisoteki kenkyū [Basic research on the historical comparative dialectology of the Ryukyuan Yaeyama dialects]. Research report. Nishihara-chō: University of the Ryukyus.
64
Thomas Pellard
Kibe, Nobuko, Haruo Kubozono, Kayoko Shimoji, Wayne Lawrence, Akiko Matsumori and Kōko Takeda. 2011. Shōmetsu kiki hōgen no chōsa, hozon no tame no sōgōteki kenkyū: Kikai-jima hōgen chōsa hōkokusho [General study for research and conservation of endangered dialects in Japan: Research report on the Kikai Island dialects]. Tachikawa: National Institute for Japanese Language and Linguistics. Kiku, Chiyo and Toshizō Takahashi. 2005. Yoron hōgen jiten [Dictionary of the Yoron dialect]. Tokyo: Musashino Shoin. Kindaichi, Haruhiko. 1960. Akusento kara mita Ryūkyūgo shohōgen no keitō [The phylogeny of Ryukyuan dialects as seen from their accent]. Tōkyō Gaikokugo Daigaku Ronshū 7. 59–80. Kokuritsu Kokugo Kenkyūjo (ed.). 1963. Okinawago jiten [Okinawan dictionary]. Tokyo: Ōkurashō Insatsukyoku. Kümmel, Martin Joachim. 2007. Konsonantenwandel: Bausteine zu einer Typologie des Lautwandels und ihre Konsequenzen für die vergleichende Rekonstruktion [Consonant change: Building blocks for a typology of sound change and its consequences for comparative reconstruction]. Wiesbaden: Reichert. Lawrence, Wayne P. 2016. Historical reanalysis in the Nakijin dialect noun accentuation system. Cahiers de Linguistique Asie Orientale 45(1). 1–25. Maddieson, Ian and Kristin Precoda. 1992. Syllable structure and phonetic models. Phonology 9(1). 45–60. Martin, Samuel E. 1987. The Japanese language through time. New Haven: Yale University Press. Matsumori, Akiko. 1991. Amami ni okeru o-dan boin no henka ni tsuite [On the changes of the vowel o in Amami]. Bungaku Kenkyū 6. 1–19. Matsumori, Akiko. 1998. Ryūkyū akusento no rekishiteki keisei katei: Ruibetsu goi 2-hakugo no tokui na gōryū no shikata o tegakari ni [The historical process of formation of Ryukyuan word prosodies: From the clues given by the special merger pattern of 2-mora classified words]. Gengo Kenkyū 114. 85–114. Matsumori, Akiko. 2000a. Ryūkyū akusento chōsa no tame no ruibetsu goi no kaihatsu: Okinoerabu-jima no chōsa kara [Development of classified vocabulary for the investigation of Ryukyuan word prosody: From the investigation of Okinoerabu Island]. Onsei Kenkyū 4(1). 61–71. Matsumori, Akiko. 2000b. Ryūkyū no takei akusento taikei ni tsuite no ichi kōsatsu: Ryūkyū sogo ni okeru ruibetsu goi 3-paku go no gōryū no shikata [A reflexion on the multi-pattern word prosodic systems in the Ryukyus: The merger of 3-mora classified words in Proto-Ryukyuan]. Kokugogaku 51(1). 93–108. Matsumori, Akiko. 2001. Historical tonology of Japanese dialects. In Shigeki Kaji (ed.), Cross-linguistic studies of tonal phenomena: Tonogenesis, Japanese accentology, and other topics, 93–122. Tokyo: Research Institute for Languages and Cultures of Asia and Africa. Matsumori, Akiko. 2012. Ryūkyūgo chōsa-yō “keiretsu-betsu goi” no soan [A draft of categorized vocabulary for Ryukyuan field research]. Onsei Kenkyū 16(1). 30–40. Matsumori, Akiko. 2017. Kita Ryūkyū ni okeru C-keiretsu 2-onsetsu meishi no gotō onsetsu no chōonka: Sono yōin ni tsuite kangaeru [Analysis of the cause of vowel lengthening in the initial syllables of the C-class disyllabic nouns in Northern Ryukyuan]. Nihongo no Kenkyū 13(1). 1–17. Matsumori, Akiko. This volume. Towards the prosodic reconstruction of proto-Japanese-Ryukyuan. In Bjarke Frellesvig and Satoshi Kinsui (eds.), Handbook of historical Japanese linguistics. Berlin and Boston: De Gruyter Mouton. Matsumoto, Katsumi. 1975. Kodai Nihongo boin soshiki kō: Naiteki saiken no kokoromi [On the vowels system of Old Japanese: An attempt at internal reconstruction]. Kanazawa Daigaku Hōbungakubu Ronshū 22. 83–152. Miller, Roy Andrew. 1967. The Japanese language. Chicago: University of Chicago Press. Mitsuishi, Yasuko. 1987. Amami Ōshima Setouchi-chō Shiba hōgen to Jōdaigo no on’in taiō shiryō [Materials on the sound correspondences between the Shiba dialect of Amami Ōshima Setouchi-chō and Old Japanese]. Kumamoto Tandai Ronshū 37(3). 55–85. Miyagi, Shin’yū. 2003. Ishigaki hōgen jiten [Dictionary of the Ishigaki dialect]. 2 vols. Naha: Okinawa Taimusu. Murayama, Shichirō. 1981. Ryūkyūgo no himitsu [The secret of Ryukyuan]. Tokyo: Chikuma Shobō.
2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan
65
Nakama, Mitsunari. 1992. Ryūkyū hōgen no kosō [Layers of history in Ryukyuan]. Tokyo: Daiichi Shobō. Nakama, Mitsunari. 1996. Nihon sogo no wagyōon [The consonant w in proto-Japanese-Ryukyuan]. Okinawa Bunka Kenkyū 22. 183–235. Nakamoto, Ken. 2011. P-on saikō: Ryūkyū hōgen hagyō shiin p-on no sujō [Reconsidering the consonant p: The origin of the consonant p in Ryukyuan dialects]. Nihongo no Kenkyū 7(4). 1–14. Nakamoto, Masachie. 1976. Ryūkyū hōgen on’in no kenkyū [Studies on Ryukyuan phonology]. Tokyo: Hōsei Daigaku Shuppan. Nakamoto, Masachie. 1983. Ryūkyū goishi no kenkyū [Studies on lexical history in Ryukyuan]. Tokyo: San’ichi Shobō. Nakanishi, Susumu (ed.). 1978–1985. Man’yōshū: Zen’yakuchū genbun tsuki [Man’yōshū: Complete annotated translation with the original text]. 5 vols. Tokyo: Kōdansha. Nakasone, Seizen. 1960. Okinawa hōgen no dōshi no katsuyō [Verb conjugations in Okinawan dialects]. Kokugogaku 41. 51–73. Nakasone, Seizen. 1983. Okinawa Nakijin hōgen jiten [Dictionary of the Nakijin dialect of Okinawan]. Tokyo: Kadokawa Shoten. Nevskij, Nikolai A. c. 1922–1928. Materialy dlja izučenija govora ostrovov Mijako [Materials for studying the dialect of the Miyako Islands]. Edited in Jarosz, Aleksandra 2015, Nikolay Nevskiy’s Miyakoan dictionary: Reconstruction from the manuscript and its ethnolinguistic analysis. Poznań: Adam Mickiewicz University dissertation. Ōno, Susumu. 1953. Nihongo no dōshi no katsuyōkei no kigen ni tsuite [On the origin of the conjugated forms of the Japanese verbs]. Kokugo to Kokubungaku 30(6). 47–56. Osada, Suma and Nahoko Suyama. 1977–1980. Amami hōgen bunrui jiten [Classified dictionary of the Amami dialect]. 2 vols. Tokyo: Kasama Shoin. Osterkamp, Sven. 2015. A sketch history of pre-Chamberlainian Western studies of Ryukyuan. In Patrick Heinrich, Shinsho Miyara and Michinori Shimoji (eds.), Handbook of the Ryukyuan languages: History, structure, and use, 61–80. Berlin: De Gruyter Mouton. Osterkamp, Sven. 2017. A mokkan perspective on some issues in Japanese historical phonology. In William McClure and Alexander Vovin (eds.), Studies in Japanese and Korean historical and theoretical linguistics and beyond, 45–55. Leiden: Brill. Oyler, Gary G. 1997. The beginning of the /N/: In search of the origin of the noun-final mora nasal in the language of Hateruma. Honolulu: University of Hawaiʻi at Mānoa Master’s thesis. Pellard, Thomas. 2008. Proto-Japonic ✶e and ✶o in Eastern Old Japanese. Cahiers de Linguistique Asie Orientale 37(2). 133–158. Pellard, Thomas. 2009. Ōgami: Éléments de description d’un parler du Sud des Ryūkyū [Ōgami: Description of a Southern Ryukyuan lect]. Paris: École des hautes études en sciences sociales dissertation. Pellard, Thomas. 2010. Review of John R. Bentley 2008, A linguistic history of the forgotten islands: A reconstruction of the proto-language of the Southern Ryukyus, Folkestone: Global Oriental. Diachronica 27(1). 170–176. Pellard, Thomas. 2013a. Nihon rettō no gengo no tayōsei: Ryūkyū shogo o chūshin ni [The linguistic diversity of the Japanese archipelago: With a focus on the Ryukyuan languages]. In Yukinori Takubo (ed.), Ryūkyū rettō no gengo to bunka: Sono kiroku to keishō [The languages and cultures of the Ryūkyū Islands: Documentation and transmission], 81–92. Tokyo: Kurosio Publishers. Pellard, Thomas. 2013b. Ryukyuan perspectives on the Proto-Japonic vowel system. In Bjarke Frellesvig and Peter Sells (eds.), Japanese/Korean linguistics 20. 81–96. Stanford: Center for the Study of Language and Information Publications. Pellard, Thomas. 2015. The linguistic archeology of the Ryukyu Islands. In Patrick Heinrich, Shinsho Miyara and Michinori Shimoji (eds.), Handbook of the Ryukyuan languages: History, structure, and use, 13–37. Berlin: De Gruyter Mouton.
66
Thomas Pellard
Pellard, Thomas. 2016a. Nichiryū sogo no bunki nendai [The date of separation of the proto-JapaneseRyukyuan language]. In Yukinori Takubo, John Whitman and Tatsuya Hirako (eds.), Ryūkyū shogo to Kodai Nihongo: Nichiryū sogo no saiken ni mukete [Ryukyuan and premodern Japanese: Toward the reconstruction of proto-Japanese-Ryukyuan], 99–124. Tokyo: Kurosio Publishers. Pellard, Thomas. 2016b. Typological and historical-comparative perspectives on tone and vowel length in Ryukyuan. Paper presented at Japanese and Korean accent: Diachrony, reconstruction, and typology, Research Institute for Languages and Cultures of Asia and Africa, Tokyo University of Foreign Studies, 2–3 July. Pellard, Thomas. 2017. A (more) comparative approach to some Japanese etymologies. In William McClure and Alexander Vovin (eds.), Studies in Japanese and Korean historical and theoretical linguistics and beyond, 55–64. Leiden: Brill. Pellard, Thomas and Yuka Hayashi. 2012. Miyako shohōgen no on’in: Taikei to hikaku [Phonology of the Miyako Ryukyuan dialects: Systems and their comparison]. In Nobuko Kibe (ed.), Shōmetsu kiki hōgen no chōsa, hozon no tame no sōgōteki kenkyū: Minami Ryūkyū Miyako hōgen chōsa hōkokusho [General study for research and conservation of endangered dialects in Japan: Research report on Miyako Ryukyuan], 13–51. Tachikawa: National Institute for Japanese Language and Linguistics. Pellard, Thomas and Masahiro Yamada. 2017. Verb morphology and conjugation classes in Dunan (Yonaguni). In Ferenc Kiefer, James P. Blevins and Huba Bartos (eds.), Perspectives on morphological organization, 31–49. Leiden: Brill. Ramsey, S. Robert and J. Marshall Unger. 1972. Evidence of a consonant shift in seventh-century Japanese. Papers in Japanese Linguistics 1(2). 278–295. Serafim, Leon A. 2007. Progressive stative predicate extensions in Ryukyuan, and their relation to earlier Japonic. In Bjarke Frellesvig, Masayoshi Shibatani and John Charles Smith (eds.), Current issues in the history and structure of Japanese, 207–218. Tokyo: Kurosio Publishers. Serafim, Leon A. 2008. The uses of Ryukyuan in understanding Japanese language history. In Bjarke Frellesvig and John Whitman (eds.), Proto-Japanese: Issues and prospects, 79–99. Amsterdam: John Benjamins. Shimabukuro, Moriyo. 2007. A reconstruction of the accentual history of the Japanese and Ryukyuan languages. Folkstone: Global Oriental. Shimoji, Kayoko. 2017. Tsukaeru Tarama futsu jiten: Tarama hōgen kiso goi [Practical dictionary of the Tarama dialect: Basic vocabulary of the Tarama dialect]. Tarama: Tarama-son Kyōiku Iinkai. Shimoji, Michinori. 2009. Epistemic modality in Irabu Ryukyuan. Shigen 5. 25–42. Shimoji, Michinori. 2011. Quasi-Kakarimusubi in Irabu. In William McClure and Marcel den Dikken (eds.), Japanese/Korean linguistics 18, 114–125. Stanford: Center for the Study of Language and Information Publications. Shimoji, Michinori and Thomas Pellard (eds.). 2010. An introduction to Ryukyuan languages. Tokyo: Research Institute for Languages and Cultures of Asia and Africa. Shinzato, Rumiko and Leon A. Serafim. 2013. Synchrony and diachrony of Okinawan kakari musubi in comparative perspective with premodern Japanese. Leiden: Global Oriental. Sibata, Takesi and Yasuko Mitsuishi. 1979. Historical relationship between the Nara-period Old Japanese and the dialect of Shiba, Kakeroma, Amami Islands. In George Bedell, Eichi Kobayashi and Masatake Muraki (eds.), Explorations in linguistics: Papers in honor of Kazuo Inoue, 484–493. Tokyo: Kenkyusha. Starostin, Sergei A. 1975. K voprosu o rekonstrukcii prajaponskoj fonologičeskoj sistemy [On the question of the reconstruction of the proto-Japanese phonological system]. Reprinted in Sergei A. Starostin 2007, Trudy po jazykoznaniju [Works on linguistics], 8–17. Moscow: Jazyki slavjanskix kul’tur. Takahashi, Toshizō. 1991. Omoro saushi no dōshi no kenkyū [A study on verbs in Omoro saushi]. Tokyo: Musashino Shoin.
2 Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan
67
Thorpe, Maner L. 1983. Ryūkyūan language history. Los Angeles: University of Southern California dissertation. Tomihama, Sadayoshi. 2013. Miyako Irabu hōgen jiten [Dictionary of the Irabu dialect of Miyako Ryukyuan]. Naha: Okinawa Taimusu. Tsuchihashi, Yutaka and Jin’ichi Konishi (eds.). 1957. Kodai kayōshū [Collection of ancient songs] (Nihon Koten Bungaku Taikei 3). Tokyo: Iwanami Shoten. Uchima, Chokujin. 1984. Ryūkyū hōgen bunpō no kenkyū [A study on the grammar of Ryukyuan dialects]. Tokyo: Kasama Shoin. Uchima, Chokujin and Kumiko Arakaki. 2000. Okinawa hokubu, nanbu hōgen no kijutsuteki kenkyū [Descriptive study of Northern and Southern Okinawan dialects]. Tokyo: Kazama Shobō. Ueda, Kazutoshi. 1898. P-on kō [On the consonant p]. Reprinted in Kazutoshi Ueda 1903, Kokugo no tame [For the Japanese language], vol. 2, 32–39. Tokyo: Fuzanbō. Uemura, Yukio. 1997. Ryūkyū rettō no gengo: Sōsetsu [The languages of the Ryukyu Islands: General introduction]. In Takashi Kamei, Rokurō Kōno and Ei’ichi Chino (eds.), Nihon rettō no gengo [The languages of the Japanese archipelago], 311–354. Tokyo: Sanseidō. Urakami, Junnosuke. 1992a. Reconstruction of the consonants of proto-Amami. Shōtoku Gakuen Gifu Kyōiku Daigaku Kiyō 23. 41–87. Urakami, Junnosuke. 1992b. Proto-Amami: Reconstruction of its vowels, synchronic description, and the formulation of the sound changes. Shōtoku Gakuen Gifu Kyōiku Daigaku Kiyō 24. 159–198. Urakami, Junnosuke. 1993. Narachō Nihongo no o-retsu kō-rui otsu-rui onsetsu ni tsuite: Amami Ōshima hōgen sogo ni miru konseki [On the A- and B-type of o-syllables of Nara period Japanese: On the vestiges seen in the proto-language of the Amami Ōshima dialects]. Shōtoku Gakuen Gifu Kyōiku Daigaku Kiyō 26. 143–161. Uwano, Zendō. 2000. Amami hōgen akusento no shosō [Aspects of accent on Amami dialects]. Onsei Kenkyū 4(1). 42–54. Uwano, Zendō. 2002. Tonogenesis in Japanese. Tōkyō Daigaku Gengogaku Ronshū (21). 3–16. Uwano, Zendō. 2014. Tokunoshima Asama hōgen no akusento shiryō (1) [Materials on the accent of the Asama dialect of Tokunoshima (1)]. Kokuritsu Kokugo Kenkyūjo Ronshū 8. 141–175. Uwano, Zendō. 2017. Tokunoshima Asama hōgen no akusento shiryō (3) [Materials on the accent of the Asama dialect of Tokunoshima (3)]. Kokuritsu Kokugo Kenkyūjo Ronshū 12. 139–161. Vovin, Alexander. 1993. Long vowels in Proto-Japanese. Journal of East Asian Linguistics 2(2). 125–134. Vovin, Alexander. 2005. A descriptive and comparative grammar of Western Old Japanese. Part 1: Sources, script and phonology, nominals. Folkestone: Global Oriental. Vovin, Alexander. 2009. A descriptive and comparative grammar of Western Old Japanese. Part 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 Hawaiʻi Press. Vovin, Alexander. 2011. On one more source of Old Japanese i₂. Journal of East Asian Linguistics 20(3). 219–228. Whitman, John B. 1985. The phonological basis for the comparison of Japanese and Korean. Cambridge: Harvard University dissertation. Whitman, John. 1999. Personal pronoun shift in Japanese: A case study in lexical change and point of view. In Akio Kamio and Ken-ichi Takami (eds.), Function and structure: In honor of Susumu Kuno, 357–386. Amsterdam: John Benjamins. Yamada, Masahiro, Thomas Pellard and Michinori Shimoji. 2013. Dunan (Yonaguni)-go no kan’i bunpō to shizen danwa shiryō [Grammatical sketch and natural discourse material of Dunan (Yonaguni)]. In Yukinori Takubo (ed.), Ryūkyū rettō no gengo to bunka: Sono kiroku to keishō [The languages and cultures of the Ryukyu Islands: Documentation and transmission], 291–324. Tokyo: Kurosio Publishers.
68
Thomas Pellard
Yamada, Masahiro, Thomas Pellard and Michinori Shimoji. 2015. Dunan grammar (Yonaguni Ryukyuan). In Patrick Heinrich, Shinsho Miyara and Michinori Shimoji (eds.), Handbook of the Ryukyuan languages: History, structure, and use, 449–478. Berlin: De Gruyter Mouton. Yanagida, Seiji. 1989. Nihongo on’inshi kara mita Okinawa hōgen no sanboinka keikō to p-on [The change to a three-vowel system and the consonant p in Okinawan dialects seen from the phonological history of Japanese]. Ehime Daigaku Kyōiku Gakubu Kiyō: Dai 2-bu Jinbun, Shakai Kagaku 21. 41–98.
Akiko Matsumori
3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan 1 Prosody-based comparative studies of Japanese and Ryukyuan The prosodic features of a language such as stress and tone are generally considered more difficult to reconstruct than its segmental features, since most writing systems do not systematically represent stress or tone. However, in the case of Japanese, historical studies have been made using the prosodic data of the currently spoken Japanese dialects. What made them possible is the regular correspondences of groups of words belonging to prosodic categories observed in present-day dialects, clearly indicating that these dialects descended from a common ancestor, plus the fact that such regional variation of prosody is still maintained in various systems of Japanese dialects throughout Japan. The discovery of such regular cross-dialectal correspondences dates back to as early as 1929, when Hattori (1929) reported the border between the Tokyo-type prosodic systems and the Kinki (Kyoto-Osaka)-type prosodic systems between two cities Kuwana and Nagashima, in the central area of Honshu (mainland Japan), showing that regular correspondences of prosodic patterns (which was then called Kata no taiō) are seen between the two types of systems in Japan. Namely, a group of words showing one prosodic pattern in a dialect of the Kinki-type system (e.g. Kyoto) regularly corresponds with a group of words in one of the Tokyo-type systems. For example, kaze ‘wind’, kuti ‘mouth’, tori ‘bird’ and usi ‘cow’, which in Kyoto are pronounced with the HH(HHH) pattern1 are all pronounced with LH(LHH) in Tokyo, while iro ‘color’, inu ‘dog’, hasi ‘bridge’ and yama ‘mountain’, which are pronounced with HL(HLL) patterns in Kyoto all appear with LH(LHL) pattern in Tokyo; similarly, hune ‘boat’, iki ‘breath’, kasa ‘umbrella’ and umi ‘sea’, which are pronounced with LH(LLH) patterns in Kyoto all appear with HL(HLL) pattern in Tokyo. This fact suggests that the regional variations of accent and tone throughout Japanese resulted from quite regular and systematic prosodic changes. Hattori then suggested that such regular correspondences should be maintained between most of the currently-spoken dialects, which would provide strong confirmational evidence that modern dialects all descended from one ancestral system: i.e.
1 The parenthesized part indicates the prosodic pattern of the phrase in which the noun is followed by a monomoraic particle; e.g. HH(HHH) indicates that the prosodic patten of a certain word is HH when pronounced in isolation, and HHH when followed by a monomoraic clitic. https://doi.org/10.1515/9781614512851-004
70
Akiko Matsumori
Nihon-sogo (literally translated as proto-Japanese, but in this article will be described as ‘proto-Japanese-Ryukyuan (hereafter pJR)’). Suggestions were then made in his subsequent papers published in the 1930s (Hattori 1931–1933) that, based on such regular cross-dialectal correspondences of words belonging to each category of prosodic classes, it is possible to ‘recontruct’ the pJR prosodic system; furthermore, Hattori suggested that we can also examine and prove how, and in what way, Japanese dialects are related, based on the observation of the systemic correspondences between these systems.
2 Development of Ruibetsu-goi 2.1 Nouns in Ruibetsu-goi A group of words which are supposed to belong to the same category in the protoJapanese prosodic system is now commonly called a ‘rui (class)’, the basic idea of which was proposed by Haruhiko Kindaichi in the 1950s. Kindaichi proposed, based on the synchronic descriptive data of some current Japanese dialects, plus some information given by the prosodic system of Early Middle Japanese (hereafeter EMJ) Kyoto dialect, reconstructed based on the tone-dot system used in the dictionary of Ruiju myōgishō, a dictionary published in the 11th and 12th centuries, a set of words that are supposed to belong to each ‘rui’ in proto-Japanese, which are now commonly called Ruibetsu-goi (Classified Vocabulary)2 and are widely used in the diachronic studies of Japanese prosody. The Ruibetsu-goi was later further developed over the following decades and become firmly established around the 1970s; the following shows only a handful of samples of the list of Ruibetsu-goi nouns employed by Kindaichi (1974): (1) Samples of Ruibetsu-goi (based on Kindaichi 1974: 62–73) monomoraic nouns Class 1.1 ti ‘blood’, e ‘handle’, ho ‘sail’, ka ‘mosquito’, to ‘door’, etc. Class 1.2 ha ‘leaf’, hi ‘day, sunlight’, mo ‘seaweed’, na ‘name’, etc. Class 1.3 he ‘fart’, hi ‘fire’, ki ‘tree’, me ‘eye’, ne ‘root’, ta ‘rice field’, te ‘hand’, etc. bimoraic nouns Class 2.1 ebi ‘shrimp’, hana ‘nose’, hane ‘wing’, kaze ‘wind’, kuti ‘mouth’, kugi ‘nail’, sake ‘wine’, sode ‘sleave’, take ‘bamboo’, tori ‘bird’, usi ‘cow’, etc. Class 2.2 aza ‘bruise’, hasi ‘bridge’, isi ‘stone’, kami ‘paper’, kawa ‘river’, kita ‘north’, oto ‘sound’, uta ‘song’, etc. Class 2.3 ami ‘net’, hana ‘flower’, hone ‘bone’, imo ‘potato’, iro ‘color’, kame ‘pot’, kusa ‘grass’, mimi ‘ear’, nomi ‘flea’, sima ‘island’, yama ‘mountain’, etc.
2 The Ruibetsu-goi was formerly presented in Kindaichi and Wada (1955, 1980).
3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan
71
Class 2.4 hari ‘needle’, hune ‘boat’, iki ‘breath’, ito ‘thread’, kasa ‘umbrella’, miso ‘beanpaste’, mugi ‘wheat’, siru ‘soup, juice’, uri ‘gourd’, usu ‘mortar’, etc. Class 2.5 ame ‘rain’, ase ‘sweat’, kage ‘shadow’, koe ‘voice’, mado ‘window’, mae ‘front’, nabe ‘pan’, oke ‘tub’, saru ‘monkey’, yoru ‘night’, etc. trimoraic nouns Class 3.1 hanazi ‘nosebleed’, kazari ‘ornament’, kemuri ‘smoke’, odori ‘dance’, sakura ‘cherry tree’, sirusi ‘mark’, tonari ‘neighbor’, yodare ‘saliva’, etc. Class 3.2 azuki ‘red bean’, hutatu ‘two’, hutari ‘two persons’, mittu ‘three’, muttu ‘six’, musume ‘daughter’, yottu ‘four’, yattu ‘eight’, etc. Class 3.3 awabi ‘abalone’, kogane ‘gold’ (according to the Waseda-go-rui), etc. Class 3.4 atama ‘head’, hasami ‘scissors’, hikari ‘light’, hukuro ‘bag’, kagami ‘mirror’, katana ‘knife’, koyomi ‘calendar’, kotoba ‘word’, otoko ‘man’, etc. Class 3.5 abura ‘oil, fat’, asahi ‘morning sunlight’, aware ‘pity’, hasira ‘pillar’, hooki ‘broom’, inoti ‘life’, itoko ‘cousin’, itutu ‘five’, kokoro ‘heart’, makura ‘pillow’, namida ‘teardrop’, sudare ‘rattan blind’, sugata ‘figure, appearance’, etc. Class 3.6 ayame ‘iris’, kitune ‘fox’, mimizu ‘earthworm’, nezumi ‘mouse’, sirami ‘louse’, suzume ‘sparrow’, usagi ‘rabbit’, unagi ‘eel’, yomogi ‘mugwort’, etc. Class 3.7 hatake ‘field’, itigo ‘strawberry’, hitotu ‘one’, hitori ‘one person’, kaiko ‘silkworm’, kuzira ‘whale’, kusuri ‘medicine’, tamago ‘egg’, tubaki ‘camelia’, etc. As for trimoraic nouns, Kindaichi, in the 1950s, proposed seven separate classes, but later changed his idea in Kindaichi (1974), rejecting the existence of Class 3.3; however, I have listed some of the words in Class 3.3 which were ‘re-established’ by the Wasedago-rui, which provide those data mainly focusing on the data acquired by documents representing the EMJ Kyoto system.3 The list of Ruibetsu-goi, once established, came to constitute a powerful descriptive tool for subsequent field research aiming for diachronic studies of accent over the past 80 years. It has been extensively used, especially in the following two or three decades after it was first proposed in 1950s (represented by Kindaichi and Wada 1955), not only for diachronic studies aiming for the historical reconstruction of accent, but also for synchronic or descriptive studies of accent in Japan; as a result, new findings of various different types of accentual or tonal systems have been newly discovered in Japanese dialects using the list. Genealogical links were also proven later between Japanese dialects and Ryukyuan dialects,4 as the descriptive studies of the areas have proceeded since 1960s and onwards, which firmly shows that the languages of the Ryukyus and all of the mainland Japanese dialects are daughters of a common ancestor: pJR (proto-Japanese-Ryukyuan).
3 For the Waseda-gorui, see Akinaga et al. (1998). 4 The problems posed by the Ryukyuan prosody will be dealt with later in Section 5.
72
Akiko Matsumori
The list of Ruibetsu-goi proposes a hypothesis about the number of prosodic classes supposed to have existed in proto-Japanese. As we see in List (1), the Ruibetsu-goi proposes that monomoraic nouns had three different classes, bimoraic nouns five, and trimoraic nouns seven in proto-Japanese. However, we should keep in mind that the number of prosodic classes proposed in Ruibetsu-goi was developed not purely based on the comparative reconstruction of the present dialects, but was supported (and was thus heavily influenced) by the system of the EMJ Kyoto dialect documented in Ruiju myōgishō, the system of which is summarized below:5 (2) EMJ Kyoto prosodic system assumed based on Ruiju myōgishō monomoraic nouns bimoraic nouns trimoraic nouns 1.1 H(H) 2.1 HH(H) 3.1 HHH(H) 1.2 H(L) 2.2 HL(L) 3.2 HHL (L) 1.3 L(L) 2.3 LL(H) 3.3 HLL (L) 2.4 LH(H) 3.4 LLL (H) 2.5 LH(L) 3.5 LLH (H) 3.6 LHH (H) 3.7 LHL (L) Considering the fact that the number of prosodic patterns generally decrease with time as a result of mergers of these patterns, the early researchers tended to regard this system (i.e. the Ruiju myōgishō system) as the oldest, due to the fact that it retains a greater number of prosodic contrasts than any other dialectal systems in Japan ever reported; in fact, for some researchers, it is ‘the’ proto-system of all the dialects in Japan. However, as stated before, due to the fact that the list of Ruibetsu-goi is not a result obtained purely by comparative reconstruction based on the data of present-day dialects, it quite often contradicts the state of accentual or tonal groups in present-day Japanese dialects. Most researchers of the present Japanese dialects realize, when actually conducting field research using this list, that we often encounter a number of exceptions, other than simple sporadic word-individual migrations between classes, especially for certain classes assumed to be trimoraic nouns of Ruibetsu-goi. Taking just a few examples, words belonging to particular classes of trimoraic nouns in Ruibetsu-goi do not necessarily appear with the same type of accentual or tonal patterns in many dialects. For example, Class 3.7, in the present-day Tokyo dialect, appears with at most three different patterns: LHH(LHHH) (itigo, karasi, kusuri, kuzira, tarai, hatake, etc.), LHL(LHLL) (tamago, hitotu, hitori, etc.), and HLL(HLLL) (kaiko, tayori, tubaki, midori, etc.). Moreover, the words in Class 3.5 is split into at least two different patterns
5 Here, the tone of a monomoraic clitic is shown in parentheses. For example, HH(H) indicates that the prosodic pattern of a certain word is HH when pronounced in isolation, and HHH when followed by a monomoraic clitic.
3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan
73
in the present-day Tokyo dialect: HLL(HLLL) asahi, hooki, inoti, karei, makura, namida, nasake, nasubi, wasabi, etc.; LHL(LHLL) itutu, itoko, kokoro, etc. (with the minor exceptions of LHH(LHHL) hasira and LHH(LHHH) abura). Similarly, in some of the Shikoku dialects including Ibuki-jima, plus the surrounding dialects of the Sanuki and Manabe type (see Section 3), the words in Class 3.5 split into two major patterns (the asahi group vs. the abura group (see later in Section 4). Such discrepancies observed in various dialects in Japan have ultimately made us realize that the fact that trimoraic nouns appear in seven different classes in the EMJ Kyoto system does not ‘guarantee’ that the proto-system reconstructed based on the present dialectal data also had exactly the same number of prosodic classes.6 Thus, it is now beginning to be realized by a number of current Japanese dialectologists focusing on accent studies that reconstruction based purely on comparative studies of the ‘present-day’ dialects has to be pursued further, which may very likely lead to the discovery of a system in which there are more prosodic distinctions than proposed by Ruibetsu-goi.
2.2 Verbs and adjectives in Ruibetsu-goi Verbs and adjectives basically have a two-way prosodic contrast in Ruibetsu-goi regardless of the length which are called Class 1 and Class 2: (3) Samples of Ruibetsu-goi for verbs (based on Kindaichi 1974: 62–73) Class 1 ik-u ‘go’, nar-u ‘ring’, sir-u ‘know’, tob-u ‘fly’, ur-u ‘sell’, yak-u ‘roast’, yob-u ‘call’, hakob-u ‘carry’, migak-u ‘polish’, nemur-u ‘sleep’, etc. Class 2 a-u ‘meet’, kak-u ‘write, draw’, nom-u ‘drink’, sum-u ‘inhabit’, tat-u ‘stand up’, tor-u ‘take’, yom-u ‘read’, tukur-u ‘make’, hirak-u ‘open’, tanom-u ‘ask for’, etc. (4) Samples of Ruibetsu-goi for adjectives (based on Kindaichi 1974: 62–73) Class 1 atu-i ‘thick’, aka-i ‘red’, ama-i ‘sweet’, asa-i ‘shallow’, omo-i ‘heavy’, usu-i ‘thin’, kanasi-i ‘sad’, yasasi-i ‘kind’, ayasi-i ‘questionable’, etc. Class 2 atu-i ‘hot’, huka-i ‘deep’, huru-i ‘old’, kara-i ‘spicy’, kuro-i ‘black’, siro-i ‘white’, kurusi-i ‘painful’, suzusi-i ‘cool’, tanosi-i ‘pleasant’, uresi-i ‘delighted’, etc. Contrary to nouns, for which accent types of Kinki areas (Kyoto-Osaka type) are more conservative in that they retain more numbers of prosodic distinctions than those
6 In fact, it is very likely that we would have reached a different set of prosodic classes than those proposed by Kindaichi in Ruibetsu-goi if the list-making were done based purely on the data obtained by the present-day Japanese dialects.
74
Akiko Matsumori
in the peripheral areas, the opposite is true for adjectives. For example, the two-way distinction as described in (4), which was attested to in the EMJ Kyoto system, is now completely lost in the present-day Kyoto dialect, so that aka-i ‘red’ (Class 1) and siro-i ‘white’ (Class 2), kanasi-i ‘sad’ (Class 1) and uresi-i ‘delighted’ (Class 2) are pronounced with the same pattern in Kyoto. Similarly, the merger of prosodic patterns from two to one for adjectives is now in progress in many areas in mainland Japanese dialects.7 On the other hand, the binary distinction between the two patterns for verbs and adjectives seems to be firmly intact in non-central areas in the Japanese archipelago such as Kagoshima and the Ryukyuan-speaking areas. Contrary to nouns, the two-way distinction is less susceptible to merger in these areas, and thus is more clearly maintained in accent systems in Ryukyu and Kyushu than in Kyoto or Tokyo Japanese.
3 Attempts to prove the genealogical relationships of present-day dialects using bimoraic words in Ruibetsu-goi 3.1 How to tell genealogical relationships of dialects based on their accent information Hattori’s aim to prove how genealogically closely related these dialects are based on the information of accent was later pursued by a number of dialectologists, but the idea proposed by Tokugawa (1962) lays the foundation for the diachronic subgrouping of dialects based on accent merger patterns; namely, Tokugawa (1962, 1981) proposed that the genealogical links between dialects (which dialects are closer than others) may be understood based on the observation of their merger patterns of bimoraic nouns in Ruibetsu-goi. The regularity of accent shift predicts that once one particular word in a class changes into another one, all the members of the same class change in the same direction, and in many of such cases a merger of two (or more) separate patterns takes place. In this way, it is generally the case that changes involve a decrease, rather than an increase, in the number of prosodic distinctions of the proto-systems. Of particular significance here is the principle that once the two separate patterns merged, they do not reverse the change and restore the previous distinctions. 7 For example, the Tokyo dialect is now undergoing the loss of the two-way prosodic distinction in adjectives, so that the longer adjectives such as kanasi-i ‘sad’ vs. uresi’-i ‘delighted’ appear with the same prosodic pattern for younger generation speakers, while shorter ones also tend to lose their distinction, especially in isolated forms (aka’-i ‘red’ and siro’-i ‘white’), while they used to be aka-i vs. siro’-i, and past tense forms (aka’-katta, siro’-katta), while the older generation would say aka’-katta vs. si’ro-katta.
3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan
75
We have seen that trimoraic nouns have a number of exceptions; however, it is generally assumed that bimoraic nouns in the Ruibetsu-goi are a fairly reliable source in identifying genealogical relationships among dialects.8 Most of the present-day dialects (except for Ibuki-jima, with which we will deal later) do not make all the distinctions made in the Ruiju myōgishō; i.e. they have fewer accentual or tonal distinctions. For example, the present-day Kyoto dialect has three distinctions for monomoraic nouns, four for bimoraic nouns, five for trimoraic nouns, etc. That is to say, these present-day dialects all resulted due to the ‘merger’ of some of their original prosodic patterns. This fact led Tokugawa (1962) to assume that the patterns of merger is a strong index in telling the genealogical relationships between dialects.
3.2 Subgrouping of dialects using bimoraic nouns of the Ruibetsu-goi The merger patterns of the prosodic classes assumed for bimoraic nouns in the Ruibetsu-goi provide suitable criteria for subgrouping the dialects, in order to provide clues to reveal the genealogical relationships between the dialects. The subgrouping of the present-day dialects depending on merger patterns of bimoraic nouns attempted by Tokugawa (1962), was further developed later; major subgrouping of the dialects based on the nature and degree of the relatedness of the dialects so far known are as follows, the name of each type was slightly modified from what was originally proposed by Tokugawa: (5) Subgroups of Japanese and Ryukyuan dialects using bimoraic nouns of the Ruibetsu-goi System types Merger type Explanation and the major location of bimoraic nouns Primary Type a. Ibukijima-type 1/2/3/4/5 The systems in which all the five-way distinctions of bimoraic nouns in the Ruibetsu-goi were retained, found in Ibuki-jima. Secondary Type b. Chūō-type 1/23/4/5 The systems in which Class 2.2 and 2.3 merged, which is mainly found in the central part of Honshu (Kyoto, Osaka, etc.) and major areas in Shikoku (Tokushima, Kōchi, etc.)
8 Except for the Ryukyuan dialects, for which different catogories for diachronic studies have been proposed and used (see Section 5).
76
Akiko Matsumori
c. Sanuki-type
13/2/4/5
d. Manabe-type
15/2/3/4
Tertiary Type e. Tarui-type
14/23/5
f. Tokyo-type (Chūrin-type)
1/23/45
g. Gairin-type
12/3/45
Others h. Southwestern Kyushu-type
12/345
g. Ryukyu-type
12/345/345
The systems in which Class 2.1 and 2.3 merged, found in the northeastern part of Shikoku (Marugame, Kawanoe, etc.) The systems in which Class 2.1 and 2.5 merged; located in some islands in the Setonai-kai (Inland Sea) (Manabe-jima, Shishi-jima, and others). The systems in which the merger of 2.1 and 2.4, in addition to the merger of 2.2 and 2.3, is observed; located in neighboring areas to the Chūō-type, around Lake Biwa in Shiga Prefecture, the Wakasa Bay area, a part of Hyōgo Prefecture, and a part of Shikoku. The systems in which, in addition to the merger of 2.2 and 2.3, Class 2.4 and 2.5 merged; located in the vast areas in Honshu, such as Tokyo, Nagoya, Nagano, etc. as well as a large area in Chūgoku area (e.g. Hiroshima, Tottori, etc.) The systems in which the merger of Class 2.1 and 2.2 and that of Class 2.4 and 2.5 took place; A number of dialects in Tōhoku areas (Hirosaki, Tsuruoka, etc.), as well as a part of the Kyushu area (Ōita), etc. are of this type. The systems in which the merger of Class 2.1 and 2.2 and that of Class 2.3, 2.4, and 2.5 took place; Areas around south-western part of Kyushu (Kashima, Nagasaki, Shimabara, Kagoshima, and the Amakusa islands). The systems with the merger of 2.1 and 2.2, plus the totally different types of split and merger patterns for Class 2.3, 2.4, and 2.5 are found (see Section 5); located in many areas in the Ryukyu archipelago, represented by Tokunoshima, Okinoerabu-jima, the northern part of the main Okinawa island, as well as various southern Ryukyu areas such as some areas in Miyakojima, Tarama-jima, some areas in the Yaeyama archipelago (e.g. Kuroshima, Kohama-jima and Iriomote-jima), and Yonaguni-jima
3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan
77
The type of merger provides a yardstick for indicating which systems are closer together than others. For example, a strong link between the Chūō-type system and Tarui-type system, scattered around in the surrounding areas of Chūō-type, would be suggested, since we can naturally assume that the Tarui-system (14/23/5) can develop from the Chūō-type system (1/23/4/5) by further merger of Class 2.1 and 2.4.9 The yardstick based on merger patterns is also useful in prohibiting us from concluding, when assuming the diachronic relationships between dialects, based on mere synchronic resemblances in prosodic systems. For example, many of the dialects with the Tokyo-type merger system (1/23/45), which are observed in vast areas in the Kantō and Chūbu areas, and the Gairin-type system (12/3/45), many of which are observed in Tōhoku areas as well as some part in Izumo in Chūgoku area and a part of Kyushu area, are, synchronically speaking, similar in that many of them have the n+1 type system (the system in which nouns of n number of moras have n+1 distinct prosodic patterns).10 However, diachronically speaking, following the basic principle that once the two separate patterns merged they do not restore previous distinctions, it is clear that they resulted in the same n+1 system through totally separate processes; i.e. neither 1/23/45 > 12/3/45 nor 12/3/45 > 1/23/45 is a possible change.11 Thus, we cannot say that the Tokyotype (1/23/45) is a daughter of the Gairin-type system (12/3/45) or vice versa.
4 Discovery of Ibuki-jima prosodic system and its implications for reconstruction Since the early stage of accent study starting from the 1930s up until the beginning of the 1960s, the EMJ Kyoto Japanese (i.e. the system of the Ruiju myōgishō) was the only known system which retained all the number of prosodic contrasts proposed in the Ruibetsu-goi (i.e. the system in which monomoraic nouns have three patterns, bimoraic nouns five, trimoraic nouns seven). That is, no system (present or past) with so many
9 When actually assuming genealogical relationships between these dialects, however, Tokugawa (1962) insisted that geographical distribution should also be taken into consideration; e.g. Tokyo-type (1/23/45) is found in Nagoya and Tokyo, while also in Chūgoku areas such as Hiroshima, Okayama, Tottori, etc. However, it is likely that they reached similar types of systems via separate development, because they are located in geographically non-contiguous areas. 10 We may say that the n+1 system have in common a synchronic prototype of Japanese, which is a relatively stable stage that has arisen as the result of many diachronic changes. For details of the n+1 system, see Kubozono (2008, 2011); the n+1 system is illustrated by the dialects of Shizukuishi, Hirosaki, and Narada, which is shown in Uwano (2012: 1425–1428). 11 For example, the dialect of Hirosaki in Aomori Prefecture, a prototypical Gairin-type system (12/3/45), and the Tokyo dialect, with 1/23/45, have the same n+1 type system. But we can confidently say that, contrary to the synchronic similarity, the dialect of Hirosaki and that of Tokyo have reached a similar n+1 system independently.
78
Akiko Matsumori
prosodic contrasts had been reported in any dialects in Japan until the 1960s. Around the time when the Ruibetsu-goi was almost established, the system in which all the fiveway distinctions for bimoraic nouns was retained was discovered in a small island close to Shikoku, Ibuki-jima, in the Inland Sea (Setonai-kai), the prosodic system of which is illustrated below (ga: nominative clitic). (6) Present-day Ibuki-jima prosodic system monomoraic nouns bimoraic nouns (μ ga) (μμ ga) 1.1 H(HH) 2.1 HH(HHH) 1.2 H(HL) 2.2 HL(HLL) 1.3 L(LH) 2.3 HM(HHM) 2.4 LH(LLH) 2.5 LH(LHL)
While the EMJ Kyoto system, i.e. the Ruiju myōgishō system, retains all the five classes for the Ruibetsu-goi’s bimoraic nouns, Ibuki-jima is the only present-day dialect which still retains the same five-way distinction for bimoraic nouns. In addition, since the Ibuki-jima dialect also maintains the 2-way distinction of verbs and adjectives, we can fairly say that the dialect is more conservative than the present-day Kyoto dialect. Of particular interest in the dialect of Ibuki-jima is that the five-way distinction of the Ruibetsu-goi uses a Mid level tone (hereafter M-tone) for Class 2.3 and Class 3.4, as well as some words in Class 3.5; thus, for example, bimoraic classes are pronounced as follows: Class 2.1 HH(HHH), 2.2 HL(HLL), 2.3 HM(HMM), 2.4 LH(LLH) and 2.5 LH(LHL). This type of tonal register including the M-tone, provides an important clue suggesting that the system attested to in the Ruiju myōgishō was not the proto-system of all the dialects in Japan. Especially noteworthy in the Ibuki-jima dialect is that, as suggested in Matsumori (1997), the words in Class 3.5 in the Ruiju myōgishō system appear with two separate patterns; HHM(M) and HLL(L).12 The two patterns are called the ‘abura’ group, and the ‘asahi’ group, hereafter. The following words belong to the patterns shown by the two groups:
12 Matsumori (1993, 1997, 1999) suggested that these two patterns resulted from the proto-form *LLH, via insertion of phrase-intial H-tone (i.e. *LLH > *HLH). Whilst one group of words resulted in HHM by the spreading of the first H-tone and the downstepping of the second one, which then resulted in a pattern having M-tone (i.e. *HLH > HHM) , the other group simply deleted the second H-tone (i.e. *HLH > HLL).
3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan
79
(7) Two types of prosodic patterns for Class 3.5 in the Ibuki-jima dialect abura group (HHM): abura ‘oil’, hibasi ‘metal chopstick for charcoals’, hooki ‘broom’, itutu ‘five’, inoti ‘life’, kyuuri ‘cucumber’, kokoro ‘heart’, namida ‘teardrop’, nasake ‘sympathy’, makura ‘pillow’ asahi group (HLL): asahi ‘morning sunlight’, aware ‘pity’, karei ‘flounder’, nisiki ‘woven brocade, fine dress’, sudare ‘rattan blind’, wasabi ‘wasabi, Japanese horseradish’ As pointed out by Matsumori (1997), the same type of split of Class 3.5 nouns into two different patterns is also found in the Sanuki type, the type in which bimoraic nouns merged as 13/2/4/5, as well as in the Manabe type, the type in which they merged as 15/2/3/4. In the Sanuki type, words in the abura group appear with HHM(HHMM) pattern, while those in the asahi group HLL(HLLL) pattern; in the Manabe type, the abura group words appear with LHL(LHLL), while those in the asahi group with HLL(HLLL). This fact led Uwano (2006) to assume that the EMJ Class 3.5 developed from a merger of two different groups that existed in the proto-system. In fact, we should not take it for granted that our reconstructed ancestral systems to have exactly the same number of classes as the EMJ Kyoto system represented by the Ruiju myōgishō, as proposed by Kindaichi, nor should we necessarily assume that the proto-system of Japanese dialects is one which exactly corresponds with the system similar to the EMJ system. A typically illustrative example is the attempt for reconstruction of the proto-Japanese system by Uwano (2006), which proposed a proto-system in which monomoraic nouns have five classes, bimoraic nouns seven, and trimoraic nouns twelve. Although his reconstructed classes need to be subject to further examination, Uwano’s attempt to reconstruct a system quite different from the Ruiju myōgishō system clearly represents the attitudes shared by many current researchers in the same field, which support that we should no longer equate the proto-system of EMJ (the Ruiju myōgishō system) with the system of Proto-Japanese.
5 The pursuit of the reconstruction of Proto-Japanese-Ryukyuan 5.1 Discovery of discrepancies between prosodic categories of Japanese and Ryukyuan dialects The recent surge of interest in Ryukyuan since around 2000, as well as the addition of Ryukyuan data, have led to a completely different stage in the diachronic study of pJR (proto-Japanse-Ryukyuan), otherwise known as Proto-Japonic. Until around the 1990s, Ryukyuan dialects were considered to be only a sub-branch of Proto-Japanese (hereafter
80
Akiko Matsumori
pJ), which were commonly considered to be closest to, and thus a sub-branch of, the accent systems in the south-western Kyushu dialects (see Uemura 1984, Kindaichi 1960, 1984, and Hirayama 1983, 1986, 1988). Some of the proponents even considered that all the Ryukyuan dialects derived from the EMJ Kyoto-type system (Kindaichi 1984). That is, the Ryukyuan dialects had all descended from a Gairin-type 12/3/45 system, which is now observed in a part of the Kyushu area (represented by the Ōita dialect). However, this picture has drastically changed since Matsumori (1998, 2000a) pointed out the discrepancy between some classes of Ruibetsu-goi and the words in each accentual or tonal category of many Ryukyuan systems, developing the idea proposed by Hattori (1958, 1979). Matsumori then suggested that, in all probability, pJ and Proto-Ryukyuan (hereafter pR) systems underwent different types of split and merger patterns after diverging from pJR (Matsumori 1998, 2000a, 2000b, 2001, 2010, 2012). The discrepancy of Ruibetsu-goi classes and those of the Ryukyuan prosodic classes are typically illustrated in the following examples from the dialect of Kin, which is located in the central part of the main Okinawa island.13 Hereafter, “cy” will be used to note [tʃ], for the following descriptions of the Ryukyuan vocabulary. (8) Ruibetsu-goi’s bimoraic nouns categorized according to the prosodic patterns in the Kin dialect (based on Matsumori 2009) a. aamii ‘net’ group: Class 2.3 words: aamii ‘net’, haanaa ‘flower’, iiruu ‘color’, kuumii ‘rice’, kuumuu ‘cloud’, kuusaa ‘grass’, maamii ‘bean’, yaamaa ‘mountain’ Class 2.4 words: iitaa ‘board’, kaasaa ‘umbrella’, kaaduu ‘corner’, miisuu ‘bean paste’, siiruu ‘soup’, waaraa ‘straw’ Class 2.5 words: aamii ‘rain’, muumuu ‘thigh’, yuuruu ‘night’ b. kaami ‘pot’ group: Class 2.3 words: kaami ‘pot’, haama ‘beach’, huuni ‘bone’, nuumi ‘flea’ Class 2.4 words: haai ‘needle’, hiira ‘spatula, shovel’, huuni ‘boat’, iicyu ‘thread’, maacyu ‘pine tree’, uusu ‘mortar’ Class 2.5 words: kaagi ‘shadow’, naabi ‘pan’, naaka ‘inside’, uuki ‘tub’ What is especially noteworthy is the fact that, as pointed out by Hattori (1958, 1979), the aamii ‘net’ group and the kaami ‘pot’ group both consist of words in Classes 2.3, 2.4, and 2.5 of the Ruibetsu-goi, as is illustrated above. While most of the words in the Ruibetsu-goi Class 2.3, such as haanaa ‘flower’, iiruu ‘color’, kuusaa ‘grass’ belong to the aamii group (8a), only a handful of words in the same class, such as haama ‘beach’, huuni ‘bone’, nuumi ‘flea’, belong to the kaami group (8b). On the other hand, words in 13 In this dialect, monosyllabic nouns have two classes (e.g. síí ‘blood’, síí nú ‘blood=NOM’ vs. huú ‘sail’, huu nú), disyllabic nouns three (e.g. háná ‘nose’, háná nú vs. háánaá ‘flower’, háána nu vs. haamá ‘beach’, haama nú, and trisyllabic nouns three (kíbúsí ‘smoke’, kíbúsí nu vs. kágámií ‘mirror’, kágámi nu vs. kataná ‘knife’ kataná nú).
3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan
81
Class 2.4 and Class 2.5 of Ruibetsu-goi are split into two major groups, one belonging to (8a), and the other to (8b). In fact, similar types of peculiar split and merger patterns exist throughout the Ryukyuan dialects for the Ruibetsu-goi’s trimoraic nouns (Matsumori 2000a), which is also represented in the following examples from the dialect of Kin. (9) Ruibetsu-goi’s trimoraic nouns categorized according to the prosodic patterns in the Kin dialect (based on Matsumori 2009) a. kagamii ‘mirror’ group: Class 3.4 words: kagamii ‘mirror’, hasamii ‘scissors’, kuyumii ‘calendar’, takaraa ‘treasure’ Class 3.5 words: andaa ‘oil, fat’, makkwaa ‘pillow’, naadaa ‘teardrop’, nuucyii ‘life’ b. katana ‘knife’ group: Class 3.4 words: katana ‘knife’, oozi ‘fan’, kutuba ‘word’, kaara ‘roof tile’, hukkui ‘bag’ Class 3.5 words: icyiku ‘cousin’, nasaki ‘sympathy’, hoocyi ‘broom’, haaya ‘pillar’ This type of discrepancy between words in Ryukyuan prosodic categories and the Ruibetsu-goi is found throughout the Ryukyuan speaking areas, from the Amami archipelago down to the Yaeyama archipelago, which strongly indicates the close genealogical link throughout the areas. That is, such ‘shared irregularities’ in the categorization of mergers are due to inheritance from a common ancestor (i.e. pR), clearly indicating that the Ryukyuan languages belong to one large group, separate from other language systems throughout Japan. In addition, the discovery of such discrepancies has cast doubt on the formerly commonly-accepted conception of Japanese dialectologists that pR is derived directly from the same proto-system as present-day two-pattern systems14 of some of the Kyushu dialects. It has also drastically changed the previously-accepted concept that the pJR system is identical to the EMJ system, from which all present-day dialects, including Ryukyuan varieties, developed. Based on the different patterns of merger (Matsumori 2001), the position of pR and pJ are illustrated in (10), in which we see that both pR and pJ are sisters, both of which are direct descendants of pJR (see also Pellard 2014, 2016): (10)
14 Typical cases of ‘two-pattern systems’ will be illustrated by the Kagoshima and Koshikijima dialects (Kubozono 2011, 2012a, 2012b) and the Ei dialect (Matsumori and Onishi 2012).
82
Akiko Matsumori
In other words, the data as we have seen above clearly shows that the Ryukyuan languages are not a daughter (or sub-branch) of any of the mainland Japanese groups, let alone southwestern Kyushu areas such as Kashima, Nagasaki, Shimabara, or Kagoshima, in which similar two-pattern systems are widely observed.15 (In fact, it is likely that they both diachronically developed similar types of two-pattern systems independently.)
5.2 Keiretsubetsu-goi: Proposal of the vocabulary list tailored for the Ryukyuan field research As shown in Section 2, the Ruibetsu-goi has been a useful descriptive tool as well as providing some crutial hints toward prosodic reconstruction of the proto-systems of the mainland Japanese dialects. However, as was clarified in the previous section, the field research using the Ruibetsu-goi is insufficient for descriptive studies of Ryukyuan prosodic systems. The need to develop a different type of vocabulary list has been recognized by several researchers of the Ryukyuan languages, since Matsumori (2000b) first attempted to propose a vocabulary list more suitable than Ruibetsu-goi for Ryukyuan prosodic studies; in addition, Matsumori (2009, 2010, 2013) proposed that the prosodic classes assumed for pR be named the A, B, and C Keiretsu (Category A, B, and C), respectively. The list is now called the Keiretsubetsu-goi (Categorized Vocabulary), literally translated as the groups of words supposed to belong to the different prosodic categories in pR. The categories in Keiretsubetsu-goi corresponds with pJ’s Ruibetsu-goi classes as illustrated below: (11)
Prosodic categories in pR seen through Ruibetsu-goi classes in pJ
monomoraic nouns:
Class 1 & 2
/
Class 3
bimoraic nouns:
Class 1 & 2
/
Class 3, 4 & 5
trimoraic nouns:
Class 1 & 2
/
Class 4 & 5
pR Categories (Keiretsu):
Category A
Category B
/
Class 3, 4 & 5
/ Class 4, 5, 6 & 7 Category C
15 It was generally considered that Ryukyuan languages are a subclass of the southwestern-Kyushu dialects because they share many crucial prosodic features: i.e. similar two-pattern prosodic systems are found in these areas. However, mere synchronic similarity of prosodic systems is insufficient to prove a strong genealogical link between them.
3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan
83
The following are samples of candidate words for each prosodic category of Keiretsubetsu-goi, the vocabulary lists belonging to each prosodic category of proto-Ryukyuan, the data of which were also obtained from the Kin dialect: (12) Candidates for Keiretsubetsu-goi categorized according to the prosodic patterns in the Kin dialect (The numbers in the parenthesis indicate the Ruibetsu-goi class) a. Candidates for Category A (A-Keiretsu) (1.1) sii ‘blood’, huu ‘sail’; (1.2) naa ‘name’, haa ‘leaf’; (2.1) usi ‘cow’, kazi ‘wind’, kizyu ‘injury, cut’, kucyi ‘mouth’, kusi ‘hip, back’, saki ‘wine’, sudi ‘sleeve’, haku ‘box’, hana ‘nose’, hani ‘wing’, hudi ‘brush’, mizu ‘water’; (2.2) cyuu ‘person’, isi ‘stone’, uta ‘song’, utu ‘sound’, kabi ‘paper’, hasi ‘bridge’; (3.1) kibusi ‘smoke’, udui ‘dancing’, kazyai ‘decoration’, tunai ‘neighbor’, hanazi ‘noseblood’, yudai ‘saliva’, koozi ‘malt’; (3.2) taacyu ‘two’, yaacyu ‘eight’, yuucyu ‘four’, miicyu ‘three’, muucyu ‘six’ b. Candidates for Category B (B-Keiretsu) (1.3) kii ‘tree’, tii ‘hand’, nii ‘load’, mii ‘eye’, yuu ‘hot water’, hii ‘fire’; (2.3) aamii ‘net’, iiruu ‘color’, kuumuu ‘cloud’, kuumii ‘rice’, haanaa ‘flower’, yaamaa ‘mountain’, maamii ‘bean’, naamii ‘wave’; (2.4) iitaa ‘board’, uuii ‘squash’, kaasaa ‘umbrella’, kaataa ‘shoulder’, muuzii ‘wheat’; (2.5) aamii ‘rain’, muumuu ‘thigh’, yuuruu ‘night’; (3.4) kagamii ‘mirror’, kuyumii ‘calendar’, taaraa ‘straw-bag for rice’, hasamii ‘scissors’, andaa ‘oil, fat’; (3.5) icyicyuu ‘five’, nuucyii ‘life’, makkwaa ‘pillow’, naadaa ‘teardrop’ c. Candidates for Category C (C-Keiretsu) (2.3) kaami ‘pot’, haama ‘beach’, maai ‘ball’, nuumi ‘flea’; (2.4) iicyu ‘thread’, uusu ‘mortar’, hiira ‘spatula, shovel’, maacyu ‘pine tree’, haai ‘needle’; (2.5) uuki ‘tub’, kaagi ‘shadow’, naabi ‘pot’, naaka ‘inside’, taabi ‘socks’; (3.4) katana ‘knife’, oozi ‘fan’, mussu ‘straw mat’, hukkui ‘bag’; (3.5) icyiku ‘cousin’, nasaki ‘sympathy’, hoocyi ‘broom’, haaya ‘pillar’; (3.6) usagi ‘rabbit’, unazi ‘eel’, garasa ‘crow’, sirami ‘louse’, hadaka ‘naked’; (3.7) hataki ‘field’, kuzira ‘whale’, kusui ‘medicine’, taare ‘tub’, tiicyu ‘one’ In addition, Ryukyuan languages have a large number of words that have no cognates in any mainland dialects or words of uncertain origin. Attempts to improve such a list of Keiretsubetsu-goi have been pursued by some researchers including Matsumori (2000b, 2012), who insist that the list, developed specifically for the descriptions of Ryukyuan prosody, should include more indigenous Ryukyuan words or longer words collected from the present-day Ryukyuan systems, which are not contained in pJ’s Ruibetsu-goi. It is clear that the Keiretsubetsu-goi will, once fully established, be a strong descriptive tool for synchronic description of Ryukyuan languages, as well as for diachronic studies of both pR and pJR.
84
Akiko Matsumori
5.3 Did the distinction between B-Keiretsu and C-Keiretsu exist in pJR? We have seen that the the distinction between Category B (B-Keiretsu) and Category C (C-Keiretsu) clearly existed in pR. Now, the next question is whether we can ascribe the distinction to the prosodic classes existing in pJR. Alternatively, one can assume that the Ruibetsu-goi Classes 2.4 and 2.5, or the Ruibetsu-goi Classes 3.4 and 3.5 merged, and then split into two different categories in pR; however, as was stated in Matsumori (2001, 2008), the reason for such an unusual split is unclear; i.e. we have not found any segmental characteristic from which the difference between these types is predictable. Based on a rigorous application of the traditional method of comparative reconstruction, we can only assume that the parent prosodic system must have had a distinction between the two, which was lost in mainland Japanese. Then, the prosodic system of pJR must have consisted of eight classes for disyllabic nouns.16 This is illustrated in the following chart, which is based on Matsumori (2001, 2008): (13)
Proto-Japanese-Ryukyuan (pJR) system consisting of eight prosodic classes for disyllabic nouns
A
PROTO-RYUKYUAN CATEGORIES
PJR PROSODIC CLASSES?
I
/
PROTO-JAPANESE CLASSES
2.1
II
B
/
2.2
III
/
IV
2.3
C
/
V
/
2.4
VI
/
VII
/
VIII
2.5
That is to say, if we give the equal weight to pJ and pR, we should ascribe the distinction between Category B (B-Keiretsu) and Category C (C-Keiretsu) to a distinction in pJR, rather than to a later development. If they did exist in pJR, the parent prosodic system in pJR must have had a system in which disyllabic nouns ultimately had eight prosodic (tonal or/and accentual) groups.
16 As for trisyllabic nouns, pJR might have had more than eight prosodic classes, as suggested by the discrepancies described in (12) between Japanese and Ryukyuan, as well as the fact that, as discussed in Section 2.1, words belonging to some of the seven classes in Ruibetsu-goi have numbers of exceptions thoughout the Japanese dialects.
3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan
85
6 Future research using the Keiretsubetsu-goi 6.1 Finding a typologically significant type of prosodic unit in the Southern Ryukyuan subbranch The Keiretsubetsu-goi list, although still incomplete, is now beginning to be used, instead of the Ruibetsu-goi, by several researchers in descriptive studies in the Ryukyus, and, as a result, has yielded important findings both synchronically and diachronically. Descriptive studies around the Miyako region in the southern part of the Ryukyu-speaking area have recently become especially advanced using the candidate list of Keiretsubetsu-goi. These advances started when Matsumori (2001, 2010) showed that, in the light of the basic idea of the Keiretsubetsu-goi, Tarama-jima, located between the Miyako archipelago and the Yaeyama archipelago, both of which were grouped into Southern Ryukyuan, and which had previously been considered to have a two-pattern prosodic system, in fact has a three-pattern system. This is followed by findings of three-patten prosodic systems in Ikema-jima (Igarashi et al. 2012, 2018), the Yonaha dialect (Matsumori 2013, 2016), and the Uechi dialect (Matsumori 2019) in the same Miyako region, all of which have long been considered to have two-pattern systems. The significance of such findings is not simply that they contain more prosodic patterns than previously thought, but that these systems all contain a type of prosodic unit quite different from those observed in the Amami-Okinawa region, or any other part of Japan. That is, the prosodic word (hereafter PWd or ω), a prosodic unit larger than a mora, syllable, or foot, is required to adequately describe the prosodic systems in the Miyako region. A PWd consists of a root of a compound, or a clitic consisting of more than two moras, etc. typical examples of which are given in (14), which is cited from Matsumori (2019): (14) Common types of PWd in Miyako Ryukyuan a. a noun: [kuusu]ω b. a member of a compound: [kuusu]ω + [bari]ω c. a bimoraic particle: [pari]ω = [kara]ω d. a noun + a monomoraic particle: [kuusu = nu]ω e. a bimoraic particle + a monomoraic particle: [kara = du]ω f. a root + a monomoraic particle: [kuusu]ω + [ pari = nu]ω
‘ABL FOC’ ‘chili pepper field GEN’ (Examples above are of the Uechi dialect in Miyako-jima.)
That the PWd is a necessary prosodic unit for proper descriptions of the prosodic systems in the Miyako region is illustrated in the following Tarama dialect examples shown in (15), in which the accent is marked by a sharp pitch fall, which is indicated by ’. Here
86
Akiko Matsumori
we can see that the phrase beginning with the Type A noun kacyuu is unaccented, while that with the Type B noun avva has the accent on the second PWd msyu, and the one with the Type C noun wa’ a on the first PWd waa: (15) Three different types of accent in the Tarama dialect in Miyako Ryukyuan (based on Matsumori 2014) Type A: kacyuu msyu =mai . . . bonito bean paste =also ‘bonito bean paste, too’ Type B: avva m’ syu =mai . . . oil bean paste =also ‘oil bean paste, too’ Type C: wa’ a msyu =mai . . . pig bean paste =also ‘pork bean paste, too’ In this dialect, the three prosodic patterns correspond with the nouns in the Keiretsubetsu goi; thus, the words in Category A shows the Type A pattern, while those in Category B, Type B, and those in Category C, Type C. As is commonly shared with many other dialects in the Miyako-Yaeyama branch, the pattern of the whole phonological phrase is determined by the category of the word comprising the first PWd; thus, for example, the phrase kacyuu msyu=mai in (15) shows the Type A pattern since the word kacyuu ‘bonito’, which is the first PWd, belongs to Category A; similarly, the phrase avva m’syu=mai shows the Type B pattern, since the first PWd avva ‘oil, fat’ belongs to Category B; in addition, the phrase wa’a msyu=mai shows the Type C pattern, since the first PWd waa ‘pig, pork’ belongs to Category C. The phrases above all consist of three PWds; e.g. ‘kacyuu msyu =mai’ consists of [kacyuu]ω [msyu]ω [=mai]ω. As we can see in (16), the accents of Type B and C are attracted to the second and the first PWd, respectively, and realized on the penultimate mora of each PWd in the Tarama dialect: (16)
Assignment of accent in the Tarama dialect in Miyako Ryukyuan
Type B:
[avva]ω
[msyu]ω
[=mai]ω
μμμ
μμ
μμ
Accent Type C:
[waa]ω
[msyu]ω
μμ
μμ Accent
[=mai]ω μμ
3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan
87
The same type of system, which also requires PWd, is found in the dialect of Yonaha on Miyako-jima, examples of which are given in (17), in which the H-tone mora is indicated by ʹ , while the L-tone mora is not marked: (17) Three different prosodic patterns in the Yonaha dialect in Miyako Ryukyuan (based on Matsumori 2013) Type A: buugI naka n kéé dú . . . sugar cane field ALL FOC ‘to the field of sugar cane’ Type B: mami náká ń kee du . . . bean field ALL FOC ‘to the field of beans’ Type C: básóó naka n kee du . . . banana field ALL FOC ‘to the field of bananas’ Matsumori (2013, 2016, 2019) analyzed this system as a type of tone language in which Type A has an LLH tonal pattern, Type B LHL, and Type C HLL; each tone is then associated with each PWd contained in every phonological phrase, as illustrated below: (18)
Assignment of tones in the Yonaha dialect in Miyako Ryukyuan
Type A:
[buugI]ω [μμμ] L
Type B:
[kee =du]ω
[μμμ]
[μμμ]
L
H
[mami]ω
[naka =n]ω
[μ μ]
[μμμ] L
Type C:
[naka =n]ω
H
[kee =du]ω [μμμ]
L
[basoo]ω
[naka =n]ω
[kee =du]ω
[μμμ]
[μμμ]
[μμμ]
H
L
L
As shown here, a PWd consists of a root of a compound, a root plus an initial mora of the following allative clitic ‘nkee’, the remaining part of the clitic plus a focus clitic ‘du’. It is noteworthy that, even though the surface prosodic features are different, the dialect of Yonaha, illustrated in (17), and that of Tarama, shown in (15), both require the
88
Akiko Matsumori
same prosodic constituent (i.e. PWd).17 Such findings, made quite recently in limited parts of the Miyako region, may open a door to a new stage of the historical studies on Ryukyuan prosody. Based on the observations so far made in the Miyako region, Matsumori (2016) argues that the prosodic system of proto-Miyako should have PWd. In fact, it is likely that we will find more (and different types of) three-pattern prosodic systems,18 in other Southern Ryukyuan areas (i.e. in the Miyako and Yaeyama archipelago) if we use the notion of PWd.19 If we were able to find that the prosodic unit PWd is shared by these dialects, we would then say that the development of PWd took place at the stage of proto-Southern-Ryukyuan (hereafter pSR), after it branched off from the pR. Thus, the development of PWd may have taken place at Stage 2 in the following diagram: (19)
The development of PWd in the history of the Ryukyuan family
Stage 1
Proto- Northern Ryukyuan (pNR)
Proto-Ryukyuan (pJR)
Amami branch Okinawa branch
Stage 2 Proto-Soutern Ryukyuan (pSR)
Miyako branch Yaeyama branch
Furthermore, the finding of PWd is typologically significant. It was recently suggested (Igarashi et al. 2012, 2018, Matsumori 2013, 2016, 2019) that the Southern Ryukyuan prosodic systems may be strikingly different from those found in any other dialects in the Japanese archipelago in that the prosodic unit PWd is required ‘in addition to’ a mora, syllable, or foot. The character of such three-pattern prosodic systems might be completely different from those already found in any other areas in Japan or in the Ryukyuan-speaking areas, in which a prosodic unit based on which an accent (or tone) is realized is either a mora or syllable.20 Not only does this finding represent a major 17 The same prosodic unit (i.e. PWd) has been found to exist in Ikema-jima (Igarashi et al. 2018), and the Uechi dialect (Matsumori 2019) on Miyako-jima. 18 In the Southern Ryukyuan areas (i.e. the Miyako-Yaeyama archipelago), Yonaguni-jima and certain dialects located in the eastern part of Iriomote-jima, represented by the Sonai dialect, are the dialects in which the three-pattern prosodic systems have long been reported to exist. 19 The previous descriptive studies of the Southern Ryukyuan varieties, represented by Hirayama (1983, 1988) and Hirayama, Ōshima and Nakamoto (1966, 1967) have not properly found the three-pattern systems in Southern Ryukyuan varieties probably because they have not examined them based on the idea of PWd. 20 The location of accent in Tokyo Japanese depends on mora-counting, while its accent-bearing unit is the syllable (Vance 2008). On the other hand, Kagoshima is a well-known example of a dialect in which the location of accent is determined based solely on syllable-counting, and the accent-bearing unit is also the syllable (Kubozono 2012b). On the other hand, for some dialects, both moras and syllables are required to determine the location of accent, which is illustrated in Koshikijima (Kubozono 2012a) in Kagoshima Prefecture.
3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan
89
breakthrough, it will certainly contribute to synchronic, as well as diachronic, studies of Japanese and Ryukyuan prosody.
6.2 Where did lengthened syllables in Northern Ryukyuan come from? In the Northern Ryukyuan branch (the Amami and Okinawa dialects),21 however, we have not found any evidence that such a prosodic unit as PWd is required; i.e. the prosodic unit of many Northern Ryukyuan systems is basically the mora. Perhaps this is related to the frequent vowel lengthening found in various Ryukyuan dialects in the Amami-Okinawa region, of which those found in the first syllables of Category C disyllabic nouns of the Keiretsubetsu-goi are the most common. The following examples are taken from the Kumejima dialect, located close to the south part of the main Okinawa island, in which we can see that the first syllables of Category C disyllabic nouns are lengthened22 as in kaami ‘pot’ or iicyi ‘breath’: (20)
The nouns corresponding to Ruibetsu-goi’s bimoraic nouns in the dialect of Kumejima in the south part of the main Okinawa island Category A nouns: kabi ‘paper’, kazi ‘wind’, haku ‘box’, hana ‘nose’, saki ‘wine’, suri ‘sleeve’, usi ‘ox’, utu ‘sound’ Category B nouns: iru ‘color’, yama ‘mountain’, hana ‘flower’, sima ‘island’, muzi ‘wheat’, wata ‘stomach’, mimi ‘ear’, misu ‘bean paste’ Category C nouns: kaami ‘pot’, iicyi ‘breath’, uusi ‘mortar’, naaka ‘inside’, haai ‘needle’, maai ‘ball’, kaagi ‘shadow’, huuni ‘bone’, haama ‘beach’, nuumi ‘flea’, naabi ‘pan’
Based on the fact that the lengthened vowels are related to Keiretsubetsu prosodic categories, Matsumori (1996b, 2017) suggested that the vowels of the first syllables of the Category C disyllabic nouns are lengthened at a stage in the development of ‘northern’ Ryukyuan varieties, clearly indicating that these lengthened vowels did not exist in pR. Matsumori (1996b, 2017) also suggested that such lengthening of the first syllables of the Category C disyllabic nouns was motivated due to some accent-related reason. It is likely that in order to distinguish among three prosodic categories, the first syllables of Category C nouns, which originally consisted of one mora, may have lengthened to have two moras. This lengthening of the initial mora of the Category C nouns, according to Matsumori (1996b, 2017), occurred at the stage of Proto-Northern-Ryukyuan (hereafter pNR), the ances-
21 As for subgroups of the Ryukyuan varieties, see Pellard (2014). 22 The same type of vowel lengthening is shared by many other dialects in the southern to central parts of the main Okinawa island.
90
Akiko Matsumori
tor of the dialects in the Amami-Okinawa region.23 Occurring initially in the first syllables of disyllabic (originally bimoraic?) nouns in Category C, the lengthening may then have proceeded to the other Keiretsubetsu-goi categories in certain dialects in the same region. The system of lengthened vowels for groups of nouns other than Category C disyllabic nouns is typically illustrated in the data from the dialect of Kin, the dialect located in the middle part of the main Okinawa island, examples of which are shown in (21). In the following, sii means ‘blood’, hana ‘nose’, kibusi ‘smoke’, huu ‘sail’, haanaa ‘flower’, kagamii ‘mirror’, haama ‘beach’, and katana ‘knife’; nu is a nominative clitic: (21)
The accent of disyllabic nouns in the Kin dialect in the main Okinawa island Monosyllabic nouns Disyllabic nouns Trisyllabic nouns Category A síí / síí nú háná / háná nú kíbúsí / kíbúsí nu Category B huú / huu nú háánaá / háána nu kágámií / kágámii nu Category C haamá / haama nú kataná / kataná nú
Here, we can see that the first syllables of Category B and Category C disyllabic nouns both consist of lengthened vowels (see more examples in (8)); e.g. the Class B disyllabic nouns such as háánaá ‘flower’ have lengthened vowels both in their first and second syllables, while Class C disyllabic nouns such as haamá ‘beach’ have lengthened vowels in their first syllables. However, in this dialect, trisyllabic nouns or longer nouns do not show such a distinction in vowel length. Although trisyllabic nouns do not show vowel length distinction in the Kin dialect, there are some dialects in the Amami region in the Ryukyus, in which vowel lengthening takes place in longer words, which is represented by the Okazen dialect in Tokunoshima in the Amami region: (22)
The accent system of the Okazen dialect in Tokunoshima Monosyllabic nouns [A] háá (leaf) háá nú (leaf-NOM) háá kárá (leaf-ABL) [B] yuú (hot water) yuu nú (hot water-NOM) yuu káara (hot water-ABL)
23 Shimabukuro (2008) attempted to refute this idea, arguing that the vowels were already long at the stage of pJR, following the idea of Hattori (1979). However, Shimabukuro’s (2008) proposal that the lengthened vowels for each category of disyllabic nouns already existed in pJR lacks convincing supporting evidence, especially in mainland Japanese dialects. Moreover, the lengthened vowels for the first syllables of Category C nouns are observed only in a very restricted area of the Ryukyuan archipelago (the Amami-Okinawa region). Until compelling evidence is demonstrated in Japanese, and/or other Ryukyuan-speaking areas such as the Miyako and Yaeyama archipelagos, the proposal made by Shimabukuro (2008) must remain only speculative.
3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan
In this dialect, the three-way distinction of Category A, B, and C is clearly made by the place of their long vowels for trisyllabic, as well as disyllabic, nouns, in addition to by the placement of accent. The long vowels in this dialect, other than those in the first syllables of Category C disyllabic nouns, may have developed after it branched off from pNR.
7 Typology of tonal change 7.1 Evaluating reconstructed systems As stated in Section 4, we should no longer equate the proto-system of EMJ (the Ruiju myōgishō system) with the tone system of pJ; i.e. it is now commonly accepted that the prosodic system of EMJ probably also developed from a certain system, with a greater number of prosodic contrasts, from which other dialects diverged. Based on the assumption that the EMJ system itself could have been the result of mergers of numbers of prosodic patterns which the previous system may have contained, various attempts to reconstruct the earlier prosodic system, typical of which is the one proposed by Uwano (2006), are now under way. We clearly need a means of evaluating the likelihood of proposed reconstructed systems, allowing comparison and determination of which competing proto-systems are most probable. One of traditional criteria for such evaluation has been the naturalness of tonal change.
7.2 Natural tendencies of tonal change It has been traditionally accepted that we should avoid proposing proto-systems which presuppose changes which are incompatible with universal tendencies of tonal change. It was Kindaichi (1973, 1974) who first attempted to propose a tendency of tonal change, which is briefly introduced in (23) (see also Matsumori 2008):
92
Akiko Matsumori
(23) General tendency of tonal change proposed by Kindaichi (1973, 1974) a. Tendency of rightward shift (e.g. HLL > HHL > HHH, LHH > LLH > LLL) b. Dissimilation at the left edge of a tonal phrase (LLL > HLL) (HHH > LHH) c. Elimination of HLH by the deletion of the second H-tone (HLH > HLL) For example, (23a) proposes that the tonal pattern HLL tends to change to HHL and subsequently to HHH, whereas change in the opposite direction (i.e. HHH>HHL>HLL) is rare. This means that ‘assimilatory’ changes tend to take place phrase-internally. On the other hand, (23b) proposes that the insertion of a new tone usually occurs at the left edge of a phonological phrase. Thus, as illustrated in (23b), an H-tone is inserted when there is a sequence of L-tones at the left edge of the phrase as in LLL > HLL. Similarly, the sequence of HHH tends to change into LHH. Notice that in this case, the tone inserted at the left edge of a phrase is always one of opposing value to the adjacent tone, which means that ‘dissimilatory’ changes tend to take place phrase-initially. Finally, as shown in (23c), Kindaichi (1973, 1974) proposed that two H-tone peaks in a phrase are generally avoided by deleting the second one, as in HLH>HLL. Although this type of change is undoubtedly the most common type of change for the tonal sequence of HLH, it is not the only type. The following remark made by McCawley (1978: 301), who postulated that the change such as HLH>HHH may have occurred in the history of Japanese, may be very important in this respect: (24)
McCawley’s remark on the change of the HLH tonal sequence I thus conjecture that the proto-Japanese accentual system that I have reconstructed arose from a true tonal system with a distinctive high or low pitch on each syllable, by means of some change which would eliminate HLH sequences, such as an assimilation of the L up to H.
He speculated that if proto-Japanese developed from some type of tone system, in which every syllable has a contrastive high or low pitch, it would be very likely that a pitch pattern such as HLH, in which the L tone is sandwiched between two H tones, may have occurred at a certain stage in the development. Thus, the tendency of change of the HLH tonal sequence has now come to assume increasing importance, in an attempt to reconstruct the prosodic systems of pJ, pR, or pJR. There are at least three different types of change from the sequence HLH so far proposed, all of which are summarized in (25): (25)
Plausible changes from HLH sequences a. HLH > HLL (Single peaking) This type of change can be regarded as a kind of dissimilation process, by which the second H-tone is simply deleted, typically reported in major areas in Koshiki-jima, Kagoshima Prefecture
3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan
b. HLH > HHH (Plateauing)
c. HLH > HHM (Downstep)24
93
An L-tone sandwiched by two H-tones is assimilated into the adjacent H tones, probably due to the rightward shift of the first H-tone, typically reported in Kakeromajima in Amami-oshima, and certain areas in Tottori Prefecture represented by Aoya and Tomari. The second H-tone is lowered into M-tone (HLH > HHM), due, perhaps, to the rightward spreading of preceding H-tone; sometimes, the second H-tone is lowered to the level of an L-tone (HLH > HHL), which will be referred to as ‘Total downstep’. This is typically found in the Nakazato dialect in Kikai-jima, Kagoshima Prefecture.
24 Here, we see that there are several competing typologically ‘plausible’ changes to consider when evaluating the validity of any proposed reconstructed system. Of course, to obtain reliable results in comparative studies of Japanese dialects, haphazard recourse to any of the tendencies in (25) would not be appropriate. To avoid that, what we need to know is what type of change is most likely to be involved than others in a particular situation. Obviously, to propose a detailed history of the intervening period from any proto-system to the present-day systems, more elaborate typological studies on tonal changes than the one proposed by Kindaichi (1974) will be needed.
8 Conclusion So far, we have seen that the descriptive studies of Japanese accent have provided excellent data for assessing the historical positions of Japanese dialects, and for understanding how actually reported systems have developed. It is fortunate that such an excellent tool as the Ruibetsu-goi was developed at a relatively early stage of the diachoronic studies of Japanese accent. The Ruibetsu-goi has not only served as a powerful tool for descriptive studies on accent for the past 80 years, but has also made a highly convincing argument possible in providing a basis for hypotheses on the historical relationships of the Japanese dialects. However, due to the fact that there are different and competing versions concerning the naturalness of tonal change, it is fair to say that there is not yet 24 Assuming that the HHM pattern observed in some dialects in Shikoku (i.e. Sanuki-type dialects) diachronically originated from HLH, by a process called “downstep” (HLH > HHM), Matsumori (1993, 1999) proposed *HLH as a proto-form of Class 1 nouns for proto-mainland Japanese.
94
Akiko Matsumori
any ‘established’ proto-system for pJ. This chapter suggests then that more substantial research is needed on the topic of typologically natural change. As for Ryukyuan, we have not yet reached to a stage at which we can attempt a solid (or reliable) reconstruction. The mass of changes in all the daughters of the pR system is much more varied than those of pJ, while the set of data collected based on the Keiretsubetsu-goi is still very limited and sketchy. However, just as the Ruibetsu-goi has facilitated our attempts in assessment of the historical relationships of the Japanese dialects, the data collected with the help of the Keiretsubetsu-goi will certainly pave the way for further solid and more reliable diachronic analyses on the history of the Ryukyuan varieties. It is now widely assumed that pR and pJ are direct daughters of pJR, and that they separated in the pre-historic period (see Matsumori 1996a, Pellard 2014, 2016), so that the data obtained from the Ryukyuan systems are much more valuable for the reconstruction of pJR (or Proto-Japonic) than previously imagined. It is obvious that, although material currently at our disposal is very limited, advancement of the research on the Ryukyuan dialects (including their prosody) will provide excellent testing grounds for the basic assumptions of historical linguistics we have established through the studies on the Japanese dialects.
References Akinaga, Kazue, Kazuaki Ueno, Kiyoe Sakamoto, Eisaku Sato and Yutaka Suzuki. 1998. Nihongo akusento-shi sōgō shiryō: Kenkyū-hen [Studies of the history of accent in Japanese: Research]. Tokyo: Tōkyōdō Shuppan. Hattori, Shirō. 1929. Kinki akusento to tōhō akusento tono kyōkai-sen [The boundary between the Kinki and eastern dialect areas]. Onsei no Kenkyu [Journal of the Phonetic Society of Japan] 3. 131–144. Reprinted in Takeshi Sibata, Masanobu Katō and Munemasa Tokugawa (eds.) 1978, Nihon no hōgen-gaku 6: Hōgen [Japanese dialectology 6: Dialects], 120–134. Tokyo: Taishukan. Hattori, Shirō. 1931–1933. Kokugo sho-hōgen no akusento gaikan (1)~(6) [General overview of accent of the Japanese dialects]. Hōgen 1~3(6). Hattori, Shirō. 1958. Amami-guntō no sho-hōgen ni tsuite: Okinawa, Sakishima sho-hōgen tono hikaku [On the dialects of the Amami archipelago: Comparison to the Okinawa and Sakishima dialects]. Jinrui-kagaku IX. 79–99. Reprinted in Hattori, Shirō, 1959, Nihongo no keitō [Genealogical origin of Japanese], 275–294. Tokyo: Iwanami Shoten. Hattori, Shirō. 1979. Nihon sogo ni tsuite 21, 22 [On Proto-Japanese 21, 22]. Gekkan Gengo 8(11). 97–107 & 8(12). 100–114. Hirayama, Teruo (ed). 1983. Ryūkyū Miyako-shotō hōgen kiso-goi no sōgō-teki kenkyū [General studies on dialectal vocabulary in the Miyako archipelago in the Ryukyus]. Tokyo: Ōfūsha. Hirayama, Teruo (ed). 1986. Amami-hōgen kiso-goi no kenkyū [A study of basic vocabulary of the Amami dialects]. Tokyo: Kadokawa Shoten. Hirayama, Teruo (ed). 1988. Minami-Ryūkyū no hōgen kiso-goi [Basic vocabulary of the southern Ryukyuan dialects]. Tokyo: Ōfūsha. Hirayama, Teruo, Ichirō Ōshima and Masachie Nakamoto. 1966. Ryūkyū hōgen no sōgōteki kenkyū [General studies of the Ryukyuan dialects]. Tokyo: Meiji Shoin.
3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan
95
Hirayama, Teruo, Ichirō Ōshima and Masachie Nakamoto. 1967. Ryūkyū Sakishima hōgen no sōgōteki kenkyū [General studies of the Sakishima dialects in the Ryukyus]. Tokyo: Meiji Shoin. Igarashi, Yosuke, Yukinori Takubo, Yuka Hayashi, Thomas Pellard and Tomoyuki Kubo. 2012. RyukyūMiyako-go Ikema-hōgen no akusento wa san-kei de atte ni-kei dewanai [The Ikema dialects of Miyako Ryukyuan has a three-, not two-, pattern accent system]. Onsei Kenkyū [Journal of the Phonetic Society of Japan] 16(1). 134–148. Igarashi, Yosuke, Yukinori Takubo, Yuka Hayashi and Tomoyuki Kubo. 2018. Tone neutralization in the Ikema dialect of Miyako Ryukyuan. In Haruo Kubozono and Mikio Giriko (eds.), Tonal change and neutralization, 83–128. Berlin/Boston: De Gruyter Mouton. Kindaichi, Haruhiko. 1960. Akusento kara mita Ryūkyū sho-hōgen no keitō [Genealogical relationships of the Ryukyuan dialects seen from accent]. Tōkyō Gaikokugo Daigaku ronshū 7. 59–80. Reprinted in Haruhiko Kindaichi (1975) Nihon no hōgen: Akusento no hensen to sono jissō [The Japanese dialects: Historical change of their accent and their situation], 129–159. Tokyo: Kyōiku Shuppan. Kindaichi, Haruhiko. 1973. Hikaku-hōgen-gaku to hōgen-chiri-gaku [Comparative historical linguistics and dialect geography]. Kokugo to Kokubungaku 50(6). 1–19. Kindaichi, Haruhiko. 1974. Kokugo akusento no shiteki kōsatu: Genri to hōhō [Diachronic analyses of Japanese accent: Principles and methods]. Tokyo: Hanawa Shobō. Kindaichi, Haruhiko. 1984. Nihongo-sogo no akusento to Ryūkyū-hōgen [Proto-Japanese accent and the Ryukyuan dialects]. Sophia Linguistica 17. 3–25. Kindaichi, Haruhiko and Minoru Wada. 1955. Kokugo akusento ruibetsu goihyō [List of the classified vocabulary of Japanese accent]. In Kokugo-gakai Kokugo-jiten Henshu-iinkai (ed.), Kokugogaku jiten, 994–998. Tokyo: Tōkyōdō Shuppan. Kindaichi, Haruhiko and Minoru Wada. 1980. Kokugo akusento ruibetsu goihyō [List of the classified vocabulary of Japanese accent]. In Kokugo-gakkai (ed.), Kokugogaku daijiten, 7–10. Tokyo: Tōkyōdō Shuppan. Kubozono, Haruo. 2008. Japanese accent. In Shigeru Miyagawa and Mamoru Saito (eds.), The Oxford handbook of Japanese linguistics, 165–191. Oxford: Oxford University Press. Kubozono, Haruo. 2011. Japanese pitch accent. In Marc van Oostendorp, Colin Ewen, Elizabeth Hume and Keren Rice (eds.), The Blackwell companion to phonology Vol. 5. 2879–2907. Malden, MA/Oxford: Wiley-Blackwell. Kubozono, Haruo. 2012a. Word-level vs. sentence-level prosody in Koshikijima Japanese. The Linguistic Review 29. 109–130. Kubozono, Haruo. 2012b. Varieties of pitch accent systems in Japanese. Lingua 122. 1395–1414. McCawley, James. 1978. Notes on the history of accent in Japanese. In Jacek Fisiak (ed.), Recent developments in historical phonology, 287–307. The Hague/Paris/New York: Mouton Publishers. Matsumori, Akiko. 1993. Nihongo akusento no sotaikei saiken no kokoromi [Towards a reconstruction of proto-systems of Japanese accent]. Gengo Kenkyu 103. 37–91. Matsumori, Akiko. 1996a. Ryukyuan: past, present and future. Journal of Multilingual and Multicultural Development 16(1/2). 19–44. Matsumori, Akiko. 1996b. Ryūkyū ni okeru ni-onsetsugo dai 4, 5-rui no gotō-chō’onsetsu o meguru sho-mondai [Problems related to the initial long syllables of Classes 4 and 5 disyllabic words in Ryukyuan]. In Hirayama Teruo Hakase Beiju Kinen Kai (ed.), Hirayama Teruo Hakase beiju-kinen ronshū: Nihongo kenkyū shoryōiki no shiten [A collection of papers for the 88th birthday of Dr. Teruo Hirayama: Aspects of Various Fields of Japanese Studies], 1130–1147. Tokyo: Meiji Shoin. Matsumori, Akiko. 1997. Tokushima-ken Waki-machi, Mikamo-chō no akusento to hondo-sogo no akusento taikei [Accent in the Waki-machi and Mikamo dialects in Tokushima Prefecture and the protoaccentual system of the mainland Japanese dialects]. Kokugogaku 189. 15–28. Matsumori, Akiko. 1998. Ryūkyū akusento no rekishiteki keiseikatei: Ruibetsu-goi ni-haku go no tokushuna gōryū no shikata o tegakari ni [Formation process of the Ryukyuan accentual systems: Based on the distinctive merger patterns of bimoraic words in the classified vocabulary]. Gengo Kenkyu 114. 84–114.
96
Akiko Matsumori
Matsumori, Akiko. 1999. Accentual reconstruction of the proto-system of mainland Japanese dialects. In Christopher Beedham (ed.), Langue and parole in synchronic and diachronic perspective, 431–447. Oxford: Elsevier. Matsumori, Akiko. 2000a. Ryūkyū no takei akusento taikei ni tsuiteno ichi kōsatsu: Ryūkyū sogo ni okeru ruibetsu-goi san-paku go no gōryū no shikata [An examination of multi-pattern accentual systems in the Ryukyuan dialects: Focusing on the merger patterns of trimoraic words in the classified vocabulary]. Kokugogaku 51(1). 93–108. Matsumori, Akiko. 2000b. Ryūkyū akusento chōsano tameno ruibetsu-goi no kaihatsu: Okinoerabu-jima no chōsa kara [Development of word lists for Ryukyuan accent research: Based on the dialects of Okinoerabu-jima]. Onsei Kenkyū 4. 61–71. Matsumori, Akiko. 2001. Historical tonology of Japanese dialects. In Shigeki Kaji (ed.), Proceedings of the symposium: Cross-linguistic studies of tonal phenomena: tonogenesis, Japanese accentology, and other topics, 93–122. Tokyo: ILCAA, Tokyo University of Foreign Studies. Matsumori, Akiko. 2008. On the reconstruction of the proto-accentual system of Japanese. In Bjarke Frellesvig and John Whitman (eds.), Proto-Japanese: Issues and prospects, 103–124. Amsterdam/ Philadelphia: John Benjamins. Matsumori, Akiko. 2009. Okinawa-hontō Kin hōgen no taigen no akusento-gata to sono keiretsu: Ryūkyū chōsayō keiretsubetsu-goi no kaihatsu ni mukete [Accentual classification of nouns in the Kin dialect of the main Okinawa island: A step towards a development of word lists for Ryukyuan accent research]. Nihon-joshidaigaku kiyō: bungakubu 58. 97–122. Tokyo: Japan Women’s University. Matsumori, Akiko. 2010. Tarama-jima no san-kei akusento to keiretsubetsu-goi [The three-pattern accent system in Tarama-jima and the categorized vocabulary for Ryukyuan accent research]. In Zendo Uwano (ed.), Nihongo-kenkyū no jūni shō [Twelve chapters on the studies of Japanese], 490–503. Tokyo: Meiji Shoin. Matsumori, Akiko. 2012. Ryūkyū-chōsa-yō ‘Keiretsubetsu-goi’ no soan [A tentative list of the Categorized Vocabulary tailored for Ryukyuan accent research]. Onsei Kenkyū 16(1). 30–40. Matsumori, Akiko. 2013. Miyako-jima ni okeru san-kei akusento taikei no hakken: Yonaha-hōgen no baai [The discovery of a three-pattern accentual system in Miyako Ryukyuan: A case of the Yonaha dialect]. Kokuritsu Kokugo Kenkyūjo ronshū 6. 67–92. Tokyo: Kokuritsu Kokugo Kenkyūjo (NINJAL). Matsumori, Akiko. 2014. Tarama-jima no akusento o saikentō suru [Reanalysing the accent of Tarama-jima]. Nihon-joshidaigaku kiyō: bungakubu 63. 13–36. Tokyo: Japan Women’s University. Matsumori, Akiko. 2016. Seichō-gengo toshite no Miyako-sogo: Tokuni sono TBU toshite kinō suru inritsu-jō no tan’i ni tsuite [Proto-Miyako as a tone language: With special focus on the prosodic unit functioning as its TBU]. In Yukinori Takubo, John Whitman and Tatsuya Hirako (eds.), Ryūkyū shogo to kodai nihongo: nichiryū sogo no saiken ni mukete [Ryukyuan and Premodern Japanese: Toward the recontruction of Proto-Japanese-Ryukyuan], 125–165. Tokyo: Kurosio Publishers. Matsumori, Akiko. 2017. Kita Ryūkyū ni okeru C-keiretsu ni-onsetsu meishi no gotō onsetsu no chōonka: sono gen’in ni tsuite kangaeru [Considering the cause of vowel lengthening of the initial syllables in the Class-C disyllabic nouns in northern Ryukyuan dialects]. Nihongo no Kenkyū 13(1). 1–17 Matsumori, Akiko. 2019. A prosodic unit, recursive structure and nature of accent of Miyako Ryukyuan. The Linguistic Review 29(1). 51–83. Berlin/Boston: Walter de Gruyter. Matsumori, Akiko and Takuichiro Onishi. 2012. Japanese dialects: Focusing on Tsuruoka and Ei. In Nicholas Tranter (ed.), The languages of Japan and Korea, 313–348. London: Routledge. Pellard, Thomas. 2014. The linguistic archaeology of the Ryukyu islands. In Heinrich Patrick, Miyara Shinshō and Shimoji Michinori (eds.), Handbook of the Ryukyuan languages, 13–37. Berlin/New York: Mouton de Gruyter. Pellard, Thomas. 2016. Nichi-ryū sogo no bunki nendai [Period of divergence between proto-Japanese and proto-Ryukyuan]. In Yukinori Takubo, John Whitman and Tatsuya Hirako (eds.), Ryūkyū shogo
3 Towards the prosodic reconstruction of proto-Japanese-Ryukyuan
97
to kodai nihongo: Nichiryū sogo no saiken ni mukete [Ryukyuan and premodern Japanese: Toward the recontruction of Proto-Japanese-Ryukyuan], 99–124. Tokyo: Kurosio Publishers. Shimabukuro, Moriyo. 2008. A reconstruction of proto-Japanese accent for disyllabic nouns: Focusing on the problem of subclasses. In Bjarke Frellesvig and John Whitman (eds.), Proto-Japanese: Issues and Prospects, 125–139. Amsterdam: John Benjamins. Tokugawa, Munemasa. 1962. Nihongo shohōgen akusento no keifu shiron: Rui no tōgō to chiri-teki bunpu kara miru [A tentative analysis on genealogical origins of the accentual systems in Japan: Based on their merger patterns of accentual classes and geographical distributions]. Gakushūin Daigaku kokugo kokubun gakkaishi 6. 1–19. Reprinted in Sibata, Takeshi, Masanobu Katō and Munemasa Tokugawa (eds.) 1978, Nihon no hōgen-gaku 6: Hōgen [Japanese dialectology 6: Dialects], 543–570. Tokyo: Taishūkan. Also reprinted in Tokugawa, Munamasa 1993, Hōgen chirigaku no tenkai [Development of dialect geography in Japan], 483–511. Tokyo: Hitsuji Shobō. Tokugawa, Munemasa. 1981. Kotoba: Nishi to higashi [Language: West and east]. Tokyo: Chūōkōron-sha. Uemura, Yukio. 1984. Ryūkyū rettō no gengo: Sōsetsu [Languages of the Ryukyuan archipelago: General review]. Gengogaku daijiten Vol. 4. 771–814. Tokyo: Sanseidō. Uwano, Zendo. 2006. Nihongo akusento no saiken [On the reconstruction of Japanese accents]. Gengo Kenkyū 130. 1–42. Uwano, Zendo. 2012. Three types of accent kernels in Japanese. Lingua 122. 1415–1440. Vance, Timothy J. 2008. The sounds of Japanese. Cambridge: Cambridge University Press.
Part II: Phonology
Teruhiro Hayata
4 Reconstruction of Old Japanese phonology 1 Introduction This chapter will consider the reconstruction of the phonology of Old Japanese (OJ), the oldest attested stage of mainland Japanese, addressing also to some extent pre-Old Japanese (pre-OJ). Discussions of proto-Japanese (pJ) must be based on Ryukyuan as well. Since the data used in this study are mainly restricted to the written sources of the central dialect of OJ (8th century), the reconstruction of pJ is out of the scope of this study, and only occasional mention will be made of the preceding stage of Japanese, pre-OJ.
2 Synchronic phonology of OJ 2.1 Levels of representation It is important to make clear the differences between the three levels of representation. In (1) two contemporary Modern Japanese (cModJ) phrases are given in three levels of representation: (1)
a. b. c.
ʻ(one) read the newspaper’ phonological siNbuN o yom-ta representation newspaper ACC read-PST phonemic siNbuN o yoNda representation phonetic [sʸĩmbũõ yonda] representation
ʻ(one) flew to Peking’ pekiN e tob-ta Peking ALL fly-PST pekiN e toNda [pekʸĩẽ
tonda]
The phonological representation in (1a) is the “representation given by the application of all readjustment rules” to the “lexical representation” (formatives provided directly by the lexicon) (Chomsky and Halle 1968: 9–11). It should be noted that the phonological (lexical) representation is a synchronic entity, not a reconstructed form. The present author assumes that one stores in the brain a lexicon that is, as it were, “a list of basic irregularities” (Bloomfield 1933: 274) and a device to relate this level of representation with the phonetic representation, e.g. a set of rules in Chomsky and Halle’s (1968) framework. The representation in (1b) is the phonemic representation in structural linguistics. Hattori (personal communication) said “(The phoneme of structural linguistics is) a https://doi.org/10.1515/9781614512851-005
102
Teruhiro Hayata
theoretical construct. (translated by T.H.)”. Hattori (1955b: 90 (=1960: 281)) also said that “the phoneme is not a fictitious concept”, but admitted that it is “a hypothetical unit”. However, he also does not say it ‘exists’ and is not known where it is; that is to say, it is a highly abstract unit. Although in generative phonology it has been claimed that “there are strong reasons to doubt its existence [the existence of a phonemic level]” (Chomsky and Halle 1968: 11), at least in Japan almost all OJ researchers have regarded this level as the most important one. Moreover, since the OJ representation distinguished by man’yōgana phonograms seems to be similar to this representation, it is convenient to use the phonographic writing (in italics in this chapter) as one kind of the representation of this level. The representation in (1c) is the phonetic representation. In general, the IPA system will be employed here except that [y] and [ʸ] will be used instead of IPA [j] and [ʲ], respectively. Although it is important to investigate the man’yōgana phonograms to clarify many problems of OJ, it is necessary to know how the phonological representation stored in the lexicon (lexical representation) and the surface phonetic representation are related – one way this relation is expressed, among other ways, in the form of classic phonological rules. Synchronic study of a language describes the phonological (lexical) representation (1a). and the set of rules relating it to its surface manifestation. Diachronic study of a language describes the diachronic changes in the phonological (lexical) representation and/or in the rules. One of the representative introductions to structural linguistics said, “the phonology should be stated without any reference to the grammar” (Gleason 1955: 66), that is, our description should meet the condition of separation of levels (Chomsky 1964: 100). However, the present author of this chapter cannot accept this condition current in the 1940s and 1950s, and does make, in the phonology, frequent reference to the grammar. Without grammatical information, morphological structures could not be referred to, intonation could not be described, and stress assignment would not be predicted. If one knows nothing about grammatical information such as verb, verbstem, verb-stem boundary, and so on, one cannot describe even the most fundamental features of a language: the conjugation of verb forms. If phonological representations must meet the condition of separation of levels, two or more stem-forms predictable from a single basic stem form would have to be postulated even for regular verbs, as in Hattori (1955a: 332 for Ryukyu (Shuri) and Kōno (1955: 393ff for Korean). The OJ verb uke2- ‘receive’, for example, would be described as having two stem-forms: 1. the non-ablaut stem uke2- and 2. the ablaut-stem uku-. Such descriptions would lead to a loss of generalization. As a consequence, the present author of this chapter does not distinguish between morphophonology (morphophonemics), which makes use of grammatical information, and phonemics, which cannot use grammatical information.
4 Reconstruction of Old Japanese phonology
103
2.2 Phonographically distinct syllables of OJ OJ had the phonographically distinct syllables shown in Table 1. Needless to say, these distinctions can not necessarily be identified with phonological distinctions. Though writers of OJ are supposed to have tried to express the surface distinctions of OJ using Chinese characters, the phonological systems of OJ and Chinese are different after all. It is unreasonable to expect Chinese characters to reproduce all the phonological distinctions of OJ. Table 1: Man’yōgana-syllables distinguished in ‘phonographic’ writing of OJ. .a ka ga sa za ta da na pa ba ma ya ra wa
ki1 gi1
pi1 bi1 mi1
.i
si zi ti di ni
ki2 gi2
pi2 bi2 mi2
ri wi
.u ku gu su zu tu du nu pu bu mu yu ru
ke1 ge1
pe1 be1 me1
.e
se ze te de ne
ye re we
ke2 ge2
pe2 be2 me2
ko1 go1 so1 zo1 to1 do1 no1
(mo1 yo1 ro1
.o
po bo
wo
ko2 go2 so2 zo2 to2 do2 no2
mo2) yo2 ro2
mo
Phonographic writing will be italicized in this chapter. Typical phonetic sources of OJ are those written syllable by syllable in Chinese characters based on their more-or-less Japanized sound values. Various Chinese characters are used to represent one and the same phonetic syllable, but distinct syllables are generally believed to be only those shown in Table 1. The subscripts ʻ1ʼ and ʻ2ʼ denote so-called kō-rui (type A) and otsu-rui (type B) syllables respectively, which merged almost completely by the 10th century. In OJ, for example, kami1 means ‘hair, top’, but kami2 means ‘god’, while from the Middle Japanese (MJ) period onwards, apart from pitch accent, kami means ‘hair, top, god’ indiscriminately. Syllables with no distinction between kō-rui and otsu-rui are given without subscripts. ‘(mo1 mo2) mo’ in Table 1 means that only one source (Kojiki) distinguishes mo1 from mo2, but all the other sources have only mo.
2.3 Phonetics of OJ While opinions vary on whether or not Chinese characters used at that time could reproduce all the phonological distinctions of OJ, the present author agrees with Hat-
104
Teruhiro Hayata
tori’s opinion that OJ also had a phonemic distinction between /’o/ [.o1] and /’ö/ [.o2] in the first syllable of words (beginning with a vowel), although there is no such distinction in the man’yōgana orthography. Consider Hattori’s remark (1976c: 13) to the effect that, although it has conventionally been believed that in OJ orthography, phonograms of different types represent different phonemic syllables, and that those of same types represent same phonemic syllables, a fundamental revision should be made to that belief. He argues that the revision need only be made to the .o syllable (淤類). See also 2.3.3 Syllables not distinct between kō-rui and otsu-rui, and the last paragraph of 3.3.2 Vowel harmony. While there may not be a one-to-one correspondence between man’yōgana syllables in Table 1 and OJ syllables in phonological (lexical) representation, the man’yōgana phonographic syllables in Table 1 correspond to phonetic syllables in IPA ([y] and [ʸ] for IPA [j] and [ʲ] respectively) in Table 2, except for the above-mentioned .o syllable. The representation ‘o’ and ‘ə’ in Table 2 line 1, corresponding to .o in Table 1, indicates phonetic values [o] and [ə] following Hattori (1976c, etc.) as mentioned above. Table 2: OJ syllables in broad ‘phonetic’ representation. [
a ka ga sa za ta da na pa ba ma ya ra wa
kʸi gʸi
pʸi bʸi mʸi
i
sʸi zʸi tʸi dʸi nʸi
rʸi wi
kᵊi gᵊi
pᵊi bᵊi mᵊi
u ku gu su zu tu du nu pu bu mu yu ru
kʸe gʸe
pʸe bʸe mʸe
e
sʸe zʸe tʸe dʸe nʸe
ye rʸe we
kᵊe gᵊe
pᵊe bᵊe mᵊe
o ko go so zo to do no po bo (mo yo ro
ə kə gə sə zə tə də nə
mə) yə rə wə
mo
]
Kamei (1950: 162) expressed the opinion that the phonetic value of the OJ vowels, at least, of the vowels in si and ti, showed no distinction between kō-rui and otsu-rui, and, taking the present-day pronunciation of Tōhoku dialects into consideration, added, “in my opinion, [it was] rather a Russian hard i (ы) than [i] in present-day Japanese”. Kamei’s opinion is attractive, but at least in the central dialect of MJ down to pre-modern Japanese, consonants followed by i and e appear to be palatalized, that is, i and e are thought to have been front vowels. Kamei’s opinion is not reflected in Table 2.
4 Reconstruction of Old Japanese phonology
105
2.3.1 Consonants The phonetic representations in Table 2 are written with syllable medials (介音 jièyīn, 介母 jièmǔ; semivowels; sounds between onset and nucleus) and nuclei represented precisely, but with onsets rather roughly. Some more details about onsets will be mentioned. 2.3.1.1 Seion and dakuon Consonants of present-day Japanese are divided into obstruents and sonorants, and obstruents into seion and dakuon.1 Sonorant consonants are nasals, liquids, and glides. The syllable initial [ŋ] of present-day Japanese is phonetically sonorant, but is usually regarded as obstruent because it is synchronically the rendaku counterpart of the obstruent /k, g/,2 and because it is diachronically a descendant of the nasal obstruent /g/ [ᵑg] of OJ. The opposition seion vs. dakuon of OJ is considered to be the opposition oral vs. nasal on various grounds. It should be noted that presumably the seion consonants were generally pronounced unvoiced, but were voiced between sonorant sounds. Assuming that seion consonants were pronounced voiced intersonorantly, we can easily explain the onbin sound changes /ki/ > /i/, /ku/ > /u/, and the /p/ > /w/ in Middle Japanese.3 Examples: ‘white-ADN(adnominal)ʼ /siro-ki/ [sʸiroɣʸi] > [sʸiroʝi] > [sʸiroyi] /siro-i/; ‘white-INF’ /siro-ku/ [sʸiroɣʷu] > [sʸiroɰu] > [sʸirou] /siro-u/; ‘river’ /kapa/ [kaβa] > [kawa] /kawa/. Although phonograms written by Chinese characters in OJ sources mostly make a distinction between seion and dakuon, Bussokuseki-ka (753) includes not a few phonograms indiscriminately used for seion and dakuon, and the two letter-like documents preserved in Shōsōin (SSL, 762) written all in phonograms show no distinction between seion and dakuon. In OJ phonograms written on excavated wooden tablets mokkan, there seems to be no exact orthographic distinction between seion and dakuon. In the MJ period the distinction became extinct generally in phonograms, as seen in Genji monogatari, Shinsen jikyō, Wamyō ruijushō, and so on. Although graphical distinction between seion and dakuon has not been stable, the phonological distinction has been kept without doubt. 2.3.1.2 Sibilants Sibilant phonemes of OJ are small in number, only /s/ and /z/. Fricatives and affricates do not contrast in OJ. In the present-day Tokyo dialect, dakuon coronals do not usually contrast between fricatives and affricates, e.g. /z/ [dz]~[z], and dakuon non-coronals do not contrast between stops and fricatives, e.g. /b/ [b] ~ [β], /g/ [ɡ] ~ [ɣ].
1 Some regard all consonants other than dakuon as seion. 2 aka-go → [akaŋo] ‘the Akha language’, aka-ko → [akaŋo] ‘young baby’ 3 These sound changes are said to have already begun to occur in late Old Japanese.
106
Teruhiro Hayata
2.3.1.3 Liquids There was only one liquid consonant /r/ in OJ. The phonetic value of /r/ is unknown, but the onset of Chinese characters used for this consonant are ones whose present-day phonetic value is [l]. Around Japanese, both Chinese and Korean have only one true liquid phoneme. So-called Altaic languages have both /l/ and /r/, and /r/ cannot occur word-initially. The fact that the single liquid sound in OJ did not occur word-initially suggests that it was an r-like sound. In present-day Korean, [l] and [r] are combinatory variants, while in present-day Japanese, they are free variants. Several decades ago no Korean words, native or loan, had initial liquids. Original initial liquids were changed into nasals. The present-day form [radio] ‘radio’, for example, had been pronounced [naʤio]. In Early Middle Japanese (EMJ) Wamyō ruijushō (10th c.) gives the then colloquial Sino-Japanese form yuwa (present Japanese: ioo) besides the native yu-no-apa for Chinese 硫黄 (New Chinese liúhuáng) ‘sulfur’. Some say that yu of yuwa is from Chinese lu 硫 with the initial y converted from l, others say yuwa is a contracted form of the native yu-no-apa (lit. bubble of hot water) with an intervocalic p > w in MJ pronunciation. Wamyō ruijushō gives another Sino-Japanese word with an initial n for original l: no(u)seu (present Japanese: nōzen (-kazura)) for Chinese 陵苕 (NC língsháo) ‘a great trumpet flower, Campsis chinensis’. It should be noted that these are very rare instances, and that all the other l-initial Chinese words seem to have been borrowed into Japanese with one and only liquid /r/, e.g. riukou (ModJ ringo ‘apple’) (Wamyō ruijushō, Ruiju myōgishō) for Chinese 林檎. There seem to be no traces of alternants for /r/ in OJ. In the MJ period, however, when the standard language became the Kyoto dialect, which was different from the Nara dialect of OJ, the passive morpheme was /(r)are/ corresponding to the OJ form /(r)aye/ in Nara. It is generally accepted that since the sound change [y] (IPA [j]) > [l/r] is phonetically unnatural, the natural change [l/r] > [y] had occurred in Nara.4 In Kyoto the original /r/ remained intact, that is, OJ of Nara was more progressive. In view of the fact that the jingle-jangle sound of beads tama shaken together was yura in OJ (cModJ [ʤara ʤara]), it is possible to infer that OJ /y/ may have derived from a sound like [ʒ] through the intervocalic sound changes like [ʒ] > [ɹ] > [y]. Of course the Kyoto type of passive morpheme became standard afterwards. OJ had the following inventory of consonants in lexicon:
4 The same is true with OJ imperative yo2 and EOJ imperative ro2. There are many instances of /r/ > /y/ > Ø in Japanese, e.g. cModJ gozarimasu > gozaimasu ‘be (polite)’, etc.; the Okinawa dialect tori > tui ‘bird’, tomari > tumai ‘harbor’, etc.; some Chinese dialects ri > yi, etc. (see Hattori 1979b: 106ff. 115 on r > i); and French [lʲ] > [y]. Contrariwise, the change [y] > [l/r] might be rare. I have never seen such a sound change. The postvocalic ru-deletion common in Japanese dialects may be another process: de aru > ʤaru > ʤa (copula), de aru > dearu > da (copula), karasumaru > karasuma (place name in Kyoto), etc.
4 Reconstruction of Old Japanese phonology
(2) p b m w
107
t s k d z g n r y
2.3.2 Vowels 2.3.2.1 o1 (kō-rui) and o2 (otsu-rui) In the Japanese academic community, it has been generally asserted that o1 is narrower than o2. Hattori, on the contrary, claimed that o2 is narrower than o1. There has been no difference of opinion as to the roundness of o1, a round vowel. Both Arisaka (1955: 390) and Hattori held the idea that o2 [ə] which in pre-OJ had been unrounded became possibly rounded [ɵ] at least just before the merger of o1 and o2, since after the merger they became rounded [o]. Hattori stated that unrounded [ə] is better than rounded [ɵ] in consideration of phonological system. In the opinion of the present author, [o] and [ə], not necessarily [o] and [ɵ], could easily merge into phonetically unmarked [o]. In the phonological interpretation in this chapter, six simple vowels and four diphthongs are posited for OJ as in (3): (3)
/i/ /u/ /ə/ /e/ /o/ /a/
Simple vowel [i] i1 [u] u [ə˔] o2 [e] e1 [o] o1 [a] a
Diphthong /ui/ /əi/
[əi] [əi]
i2 (u~ i2) i2 (o2~ i2)
/oi/ /ai/
[əi] [əe]
i2 (o1~ i2) e2 (a~ e2)
As the underlying representation of i2, the present author posits /ui/ except for forms with o2 ~ i2 and o1 ~ i2 alternations. The condition of separation of levels and the biuniqueness condition in structural linguistics are avoided here, simply because they hinder the study of OJ phonology. In the system postulated here OJ had four diphthongs /ai/, /əi/, /oi/, and /ui/ as in (3). More details about those alternations will be given in 2.3.4 Exposed forms and covert forms. As to /ai/ [əe], just for information, present-day Tokyo Japanese has only one diphthong /ai/. It is clear from accentuation that the /ai/ of Tokyo cModJ consists of one syllable though of two moras. Accented verbs in Tokyo carry an accent on the penultimate mora (not syllable) of the stem when followed by, among other forms, an adnominal past affix (ta): thus in hasir-ta → ha.sit.ta ‘ran’ the accent falls on the stem-penultimate mora si in hasir-, giving [ha.sit.ta] where boldfaced syllable designates high-pitch, while in hair-ta → hait.ta ‘entered’ the accent falls on i in hair- first, but the whole syllable with an accent is pronounced with a high pitch,
108
Teruhiro Hayata
resulting in [hait.ta]5 (eventually only the first part of the syllable bears the high pitch). Some dialects near Tokyo have haitta with a disyllabic /a.i/ showing [ha.it.ta] rather than [hait.ta]. In OJ the noun ka.i ‘oar’ must have been a disyllabic word /kayi/,6 just like .o.i /əyi/7 ‘grow old’. 2.3.2.2 Explanation for the disappearance of the distinctions between o1 and o2, e1 and e2 Probably because sources of OJ make no distinction in writing between long and short vowels, it has been believed that the only possible syllable type of OJ was (C)V, that is, that OJ had no long vowels, diphthongs, or closed syllables. Before Hattori, no one claimed that OJ had distinctive long vowels not resulting from the deletion of intervocalic consonants. He knew that distinctive long vowels are actually found in some mainland dialects and more widely observed in Ryukyuan dialects. He maintained that there must have been phonemic long vowels in OJ as well as in pJ (Hattori 1979abcd). From the typological point of view, it would be extremely unnatural for a language not to have the opposition between a long vowel and a short vowel and between a simple vowel and a diphthong as well as not to have a closed syllable or glottal adjustment as a result of a closed syllable. In pre-OJ, o2 /ə/ was the yin (feminine) counterpart to the yang (masculine) /a/ (cf. the genitive marker no2 /nə/~na /na/), and therefore o2 must have been unrounded. Hattori (1976a: 29, 1979c: 98) proposed that pJ short o1 and short e1 in the central dialect only, not in EOJ or in the Kyushu dialects, raised in OJ to u and i1 respectively. Following Hattori’s line of reasoning, as a result of the vowel raising, the remaining /o/ (o1) of OJ, which is a reflex of former /oo/ (long o1), became inferior to /ə/ (o2) in number. Consequently the remaining /o/ (o1) could not maintain the opposition to /ə/ (o2) and eventually merged with /ə/ (o2) in the transition to MJ, giving only one unmarked /o/ . In the same way, all the pre-OJ short /e/ (e1) changed into /i/ (i1), while /ee/ (long e1) remained intact and eventually merged with e2, giving only one unmarked /e/ (see also Hayata 1998: 27–30). Some dialectal differences that result from the presumed vowel raising are shown in (4): (4) OJ after vowel raising saku kanasiki1 pasiki1 (yasi)
5 Hattori said that [hait.ta] of present-day Tokyo was formerly [he:t.ta]. 6 Although yi of kayi etc. is phonographically not distinct from i, the fact that these words do not participate in jiamari (cf. 2.4.2) is, among others, reliable evidence for yi rather than i. 7 Although in .oyi phonographic discrimination between o1 and o2 is not made, an attested derived form .oyo2si ‘old’ (MYS 5.804) offers convincing evidence to prove that this o is o2.
4 Reconstruction of Old Japanese phonology
109
Hattori’s vowel raising is a very fascinating idea. However, this hypothesis would require more extensive data from internal and comparative studies. It should be added that there have been studies arguing that long vowels in present-day Ryukyuan dialects are the result of development from prosody (Kindaichi 1975: 129–159, Pellard 2012, 2013). 2.3.2.3 On the interpretation of the OJ vowel system involving w and y Few linguists still support the traditional view that OJ had eight distinct vowel phonemes. Some hold that OJ had five, and others six. Except those who reject the phonemic distinction between o1 and o2, those who embrace the five vowel system of OJ interpret the differences between o1 and o2, i1 and i2 , and e1 and e2 in terms of the presence or absence of a glide w or y as in (5). (underlined by T.H.): (5) phonographic repr. Ca Ci1 Ci2 Cu Ce1 Ce2 Miyake (1933) Kikuzawa (1935) Ca Ci Cwi Cu Ce Cwe Frellesvig (2010) Ca Ci Cwi Cu Cye Ce
Co1 Cwo Cwo Cwo
Co2 Co?Cyo Co Co
Hattori advocated the theory that OJ had six vowel phonemes. This chapter is written in terms of OJ six vowel phonemes after Hattori’s theory, but underlying representations adopted in this chapter are more abstract than Hattori’s. It seems appropriate to examine the above-mentioned interpretation involving w and y. In this interpretation, Co2 is /Co/ and Co1 is /Cwo/ as in (6). (6) /Co/ = Co2 (therefore /o/ = o2) /Cwo/ = Co1 (therefore /wo/ = o1) However, in at least all the OJ phonographically attested words in the Man’yōshū (MYS) the syllable wo never co-occurs with Co1 within a morpheme, but does co-occur with Co2. Examples: woto2-me1 ‘young woman’ 20 MYS, woto2-ko1 ‘young man’ 4 MYS, woto2tu-pi1 ‘the day before yesterday, lit. the long distance day’ 2 MYS, wodo2r- ‘jump’ 2 MYS, woso2ro2 ‘precocity, thoughtlessness’ 1 MYS. Consider Arisaka’s Laws (Arisaka 1931: 103), given in (7). (7) First law: In a morpheme8 o1 and o2 did not co-occur at all. Second law: It was not often that u and o2 co-occurred in a morpheme. In a disyllabic morpheme with u and o, this o must not be o2. Third law: It was not often that a and o2 co-occurred in a morpheme. 8
8 Arisaka’s term “ketsugō-tan’i” is translated as ‘morpheme’ in the present author’s interpretation.
110
Teruhiro Hayata
No counterexamples have been found to the first law, but the second and third laws might be thought to be tendencies. The fact that wo and Co1 did not co-occur in a morpheme suggests that wo is wo2 in the light of Arisaka’s first law. If wo is wo2, the disyllabic morpheme uwo ‘fish’ might go against the provision of the latter half of Arisaka’s second law. However, the variant form iwo rather than uwo ‘fish’ is likely to be historically older in consideration of dialectal distribution and probable sound changes. In not many words of MYS the syllable wo co-occurs with a vowel a, which is usually regarded as yang (masculine), e.g. awo ‘blue’ appears 28 times in MYS; mawos- ‘tell’ 12; woka ‘hill’ 3; wosame2- ‘reign over’ 5; and sawo ‘pole’ 2. Even if wo is wo2, the above words are compatible with Arisaka’s third law. It seems reasonable to suppose that the syllable wo was substantially an otsu-rui syllable containing o2 /ə/, that is, wo was wo2 /wə/. Generally speaking, there are two points as to the interpretation using w, as in /Cwi, Cwo/: first, adequacy for using rounded/labial /w/ in light of the underlying segments; second, naturalness of the distribution of /w/. One of the reasons to set up /Cwo/ for Co1 is that the phonetic value of o1 is considered more rounded than o2 which is considered less rounded. As mentioned above (2.3.2.1), it does not seem appropriate to believe that o2 was rounded throughout the OJ period. In the analysis adopted in this chapter, o1 was rounded and o2 was unrounded. As to Ci2, /Cwi/ in the interpretation with /w/, it is true that some instances show alternation between i2 and rounded u as in kami2 and kamu- ‘god’ and that others show alternation between i2 and unrounded o2 /ə/ as in ki2 and ko2 /kə/ ‘tree’, that is, the underlying vowels of i2 can be either rounded vowels or unrounded vowels. Last, but by no means least, there is the problem of the naturalness of the distribution of /w/. In the OJ system of five vowel phonemes with a syllable-medial glide /w/, the glide /w/ occurs not before /a/ but only before /i/, /e/, and /o/. Universally /wa/ is most unmarked syllable containing /w/, and thus in present-day Tokyo Japanese /wa/ is the only syllable containing /w/. If a language has free syllables /a, i, e, o/ and /wi, we, wo/, this language would be expected to have /wa/. A language that has /Cwi, Cwo/ but not /Cwa/ is extremely peculiar. It does not seem to be plausible to argue that OJ had such a distribution of /w/. 2.3.3 Syllables not distinct between kō-rui and otsu-rui As is shown in Tables 1 and 2, the o1-o2 distinction was made only after non-labial consonants, with the exception of a mo1-mo2 distinction found only in Kojiki, and the i1-i2, e1-e2 distinctions were made only after non-coronal consonants /k g p b m/. The fact that the o1-o2 distinction was neutralized after labial consonants suggests that roundness is the distinctive feature between o1 and o2. Hence o1 is /o/ and o2 is /ə/. However, even not preceded by labial consonants, word initial o1 and o2 are not distinct phonographically. This is quite an unnatural fact from the point of view of the overall
4 Reconstruction of Old Japanese phonology
111
phonological system. Most scholars, however, have not wondered about this. Hattori (1976b: 7) argued to the effect that, because it is very strange that there was only one kind of .o syllable in OJ, we must consider this point carefully; Arisaka did not attach much importance to this point through the absence of a macroscopic [structural] viewpoint. Hattori is of the opinion that the Chinese language at that time had no characters appropriate for distinguishing OJ .o1 from .o2 (Hattori 1976b: 10, 1976c: 13, etc.). As a result, OJ phonographic writing made no distinction between .o1 and .o2, although they were in fact two distinct phonemes (Hattori 1982: 97, 1983d: 112–113). Hattori thought that the phonogram .o followed by a syllable written in a Chinese character acceptable for masculine (e.g. 久) must represent masculine /o/, and that the phonograph .o followed by a (C)o2 (feminine) syllable must feminine /ə/ (Hattori’s /ö/). He gives the examples in (8) (Hattori 1983a: 12, 1983b: 80):9 (8) feminine /ə/ ‘raise (INF)ʼ ‘sound’ ‘think’ masculine /o/ ‘secluded place’ ‘send’ ‘delay’
Rare as it is, in the present-day Ibaraki dialect of Japanese, Ce and Ci are two distinct syllables, while .e and .i are not distinct, merging into one phoneme with phonetic value [e ~ ɩ]. /ke/ ‘hair’ and /ki/ ‘tree’ are distinct, but nyuu.en ‘entry into a kindergarten’ and nyuu.in ‘hospitalization’ are homophonous [nʸuːeɴ]. For OJ phonemes, neither the syllable-medial w nor the syllable-medial y is necessary, though the onset w and y are autonomous phonemes. Palatalized consonants [Cʸi] and [Cʸe] are produced by the phonetic rules in (9). (9) Ci → Cʸi (Ci1) Ce → Cʸe (Ce1) Consider (10), where the final syllables of Provisional/Concessive forms show alternation between Ce and Ce2 . (10) Realis/Concessive (izenkei) tugure ← tugu-re2 ‘though, as (one) tells’ aso1be2 ← aso1b-re2 ‘though, as (one) plays’
9 See also Arisaka (1957[1934]: 115).
112
Teruhiro Hayata
From (9) it is clear that one can postulate the OJ underlying representation (here in phonographic writing) of Realis/Concessive as re2, with r deleted after a consonant. A syllable beginning with a coronal consonant like re2 is neutralized in terms of the e1-e2 distinction on the surface. The underlying form ✶re2 could be reconstructed for a pre-OJ form. In some forms, the distinction between o1 and o2 written in phonograms, which do not distinguish between the two, can be reconstructed from attested compound forms whose first vowel is deleted, as in (11).1011 (11)
VV intact first V of VV deleted situ-ori NSN-Tenmu sito1ri ‘old Japanese fabric’ NSN-Jindai 10 ✶ .o2mi1 ʻ(liege) subject’ naka-tu-omi1 nakato211mi1 ‘Nakatomi (name of a clan)ʼ MYS 17.4031 That is, “weave” (tu)ori → to1ri therefore oris ✶o1r/or-/ ✶ “liege” (tu)omi1 → to2mi1 therefore omi1- is o2mi1- /əmi-/ .o1r- ‘weave’
✶
The verb ‟weave” is not wor-, but or-. If o1 is /wo/ as predicted in (6), /wo/ is incompatible with /o/, and of course incompatible with /ə/.
2.3.4 Exposed forms (roshutsukei) and covert forms (hifukukei) Many words of OJ show vowel alternation between word-final forms and word-internal forms in nouns, and between stem-final-vowel forms and stem-internal-vowel forms in (so-called) vowel-stem verbs. After Arisaka (1931), final alternant forms are called ‘exposed forms’ (roshutsukei) and internal forms ‘covert forms’ (hifukukei). Alternations of vowel-stem verbs are more regular than those of nouns. In (12), forms on the left-hand side of the full width tilde (~) are exposed forms and ones on the right-hand side covert ones. ‘#ʼ is a word boundary put only word-finally here for simplicity, and the end of a verb stem will be noted by “ ] ”. (12)
Phonographical notation exposed covert /saka/
Phonological notation exposed covert with surface forms sake2#~saka-na# /saka-i# ~ saka-na#/ ‘saké~saké-vegetables’ [sakᵊe] [sakana]
10 Phonographic naka-tu-omi1 is not attested. 11 Although there are some doubts about the phonographic notation of this poem, in the author’s opinion the nakato2mi1 is not bad from a phonological point of view.
‘receive, tr.~intr.’ ‘tree, wood~shade of a tree’ ‘a lull, calm~calm’ ‘moon~moon night’ ‘be exhausted~exhaust’
The data in (12) will be accounted for by the following synchronic rules which convert diphthongs /ai, əi, oi, ui/ into surface forms (13): (13) Non-round-otsu-rui formation Cai → Cᵊe (Ce2) Cəi Coi → Cᵊi (Ci2) Cui In (13), the underlying forms /Cu Cə Co/ are neutralized into [Cᵊi] Ci2 when followed by a vowel /i/. This neutralization is just like that of the present-day Japanese (14) and that of English (15). (14) yom+ta → yom+da → [yonda] ‘read (PST)ʼ md tob+ta → tob+da → [tonda] ‘flew’ (15)
bd
nd
telegraph (data from Chomsky and Halle 1968: 11) telegræf → télegrf → [téləgrf] e ə / [–stress] telegræf-y → telégræf-y → [təlégrəf-y] æ
In the present-day Tokyo dialect, the causative and passive forms of the verb /kak-/ ‘write’ are as follows: (16)
The changes in (16) result from the neutralization in (17): (17)
k-s →
k C-s
k-r →
k C-r
C
That is, different sequences of phonemes represent a single phone, which does not satisfy the biuniqueness condition maintained by structural linguists –the condition rejected as already mentioned in the paragraph under (3). While the view that i1 and i2 of OJ were two distinct simple vowel phonemes is out of the question, the structuralists Hattori (1959: 62) and Matsumoto (1995) maintain that the distinction between i1 and i2 of OJ was with and without a /y/ (glide) as in (18): (18) Ci1 /Cyi/ [Cʸi] Ci2 /Ci/ [Cᶤi] 【=[Cᵊi] of the present author】 Furthermore, Frellesvig (2010: 31–33) proposes the following: (19)
[Ci̯ i] Ci1 /Ci/ Ci2 /Cwi/ [Cu̯ i, Cui̯ ]
Ci1 for /Ci/ in (19) is better than Ci1 for /Cyi/ in (18). However, as already mentioned in Section 2.3.2.3, /Cwi/ has the problem of the distribution of /w/ (cf. Hattori 1983b: 81). The phonological system (18) would have undergone changes as shown in (20). (20)
the pre-OJ period when ki2 had not developed yet ki
OJ kyi (ki1)
kui
ki (ki2)
the MJ period when the distinction between ki1 and ki2 disappeared
ki
Hattori (1983c: 123) changed his previous opinion shown in (19) and claimed that the unmarked ki1[kji] ([kʸi]), in general Ci1, should not be /kji/ (/kyi/) but should be /ki/. Hattori wrote a new phonemic representation /kºi/ for the marked ki2[kïi] ([kᵊi])in order to denote that the consonant and the vowel are concatenated in an unusual way (Hattori 1983c: 123). Hattori, however, did not tell what the phonemic status of / º / in /kºi/ is. Using Hattori’s new notation, (20) will be written as in (21).
4 Reconstruction of Old Japanese phonology
(21)
the pre-OJ period when ki2 had not developed yet ki
OJ ki (ki1)
kui
kºi (ki2)
115
the MJ period when the distinction between ki1 and ki2 disappeared
ki
2.3.4.1 Phonetics of e2 It is no wonder that Hattori rejected /əi/ for OJ e2 in favor of /e/ in a phonemic (structural linguistic) representation. In the phonological representation, which is more abstract, the present author assumes /ai/ for OJ e2. In Hattori’s diachronic theory only pJ ✶ai (not ✶ a.i) developed into e2, and only e2 is a descendant of pJ ✶ai (not ✶a.i) , but in contrast, Hattori (1979c: 98) says that both pJ ✶ee and ✶ia developed into OJ e1. However, in the following synchronic analyses by Hattori, phonetic monophthong were interpreted as phonemic diphthongs. Hattori said that since in all the three Chakhar Mongolian long vowels [æː] [œː] [øː] the tongue moves toward [i], these vowels are to be analyzed as /ai/, /oi/, /ui/ respectively (Hattori 1955b: 93). As to the Tatar language, Hattori (1975) concludes that “the Tatar high vowels i, ü, and u are to be phonologically interpreted as /ej, ew, əw/ respectively, and that the high diphthong [ɨi] is phonologically /əj/”.12 Not a few dialects of cModJ have vowel phonemes which phonetically appear in free variation [aɩ~eæ~ɛː]. It is possible to interpret all these vowels as /ai/. 2.3.4.2 What is -i in exposed forms? The phonological forms /Cu Cə Co/ of OJ are neutralized into [Cᵊi] Ci2 when followed by a vowel /i/, which is thought to be a morpheme of pre-OJ (22). (22)
Ci2 /Cui/ /Cəi/ /Coi/
[Cᵊi]
Some examples of the pairs of /CV/ and /CVi/ will be given in (23). (23) kamu-kazai kamui kə-tati kəi
→ → → →
kamu-kazᵊe kamᵊi kə-tatʸi kᵊi
‘divine wind’ ‘god(s)ʼ ‘grove’ ‘tree, wood’
12 From explanatory notes on Hattori (1991).
116
Teruhiro Hayata
nago-ya ‘peacefulness, nature, state’ nagoi-isi → nagᵊi-sʸi ‘became calm (ADN)ʼ kopo-si-ku ‘be longing for (INF)ʼ kopoi-i → kopᵊi ‘long for (INF)ʼ (Although the syllables po1 and po2 were not discriminated in man’yōgana writing, the morpheme ko1po- is thought to have been /kopo-/ (ko1po1-) in light of Arisaka’s first law (7): o1 and o2 did not co-occur in a morpheme.) If the phonological (underlying) forms /Cui/, /Cəi/, /Coi/ of the surface form Ci2 [Cᵊi] are obtained from the shapes of their alternants, the question is what the underlying form would be of the surface form Ci2 [Cᵊi] that is found in morphemes without alternation between exposed forms and covert forms. In a child’s acquisition of underlying forms of Modern Japanese verbs, it is observed that the most unmarked shape is posited as the underlying form until the correct shape is acquired. In the present author’s experience of seventy years ago as well as lately, at first an incorrect form /sim-/ ‘die’ seems to be posited as an underlying form, giving simanai [die.NEG], simu [die.NONPST], until the correct form /sin-/ is acquired. Children hear sinde [die.GER], sinda [die.PST] many times, and also hear the negative forms yomanai [read.NEG], humanai [step.NEG], kumanai [assemble.NEG], tsumanai [stack. NEG], sumanai [finish.NEG], nomanai [drink.NEG], momanai [rub.NEG], etc. In cModJ. all the nasal-stem verbs are m-stem verbs except sin-. It is only natural that the verbstem si[n~m]- is apt to be understood as sim-. In OJ period also, other things being equal, it seems that the most unmarked form should be regarded as the underlying one. Among the three types of attested covert forms /Cu/, /Cə/, /Co/, which one is the most unmarked? The present author’s provisional investigation has shown that covert forms including /Co/ are extremely few in number, those including /Cu/ or /Cə/ almost the same in the number of different morphemes, but those including /Cu/ overwhelmingly greater in the number of running morphemes. This fact suggests that the underlying forms of morphemes including Ci2 overt forms that do not show corresponding covert forms can be forms including /Cu/: e.g., ki2ri ‘fog, mist’ would be /kuiri/, mi2na ‘all’ be /muina/, sugi2 ‘cedar’ be /sugui/, and so on. Hattori (1979b: 106) wrote to the effect that the ✶/-i/ in the OJ verbs әki2 ✶/әkә-i]-/ ‘occur, get up’, ake2 ✶/aka-i]-/ ‘open tr.’, make2 ✶/maka-i]-/ ‘be defeated’, etc. must be a “nominalizer”, but predecessors of other ✶/-i/ are from various consonants and others. The present author holds the opinion that synchronically the verb-stem-final /i/ is a verb-stem forming morpheme, while the covert noun final /i/ is a word-forming suffix. A few words contain (C)i2 and (C)e2 syllables word-internally. Such words always end in /i/ and therefore do not show alternations between exposed and covert forms. Hattori (1979b: 106) presumes that in the pJ word ki2ri ‘fog’, the non-word-final ki2 in fact occurred morpheme- finally, i.e. ✶kui-r-i. The quadrigade (consonant base) verb ✶ ki2r- is assumed to have existed based on attested OJ logograms. In MJ sources, Genji monogatari etc., the verb kir- ‘fog up’ is phonographically attested.
4 Reconstruction of Old Japanese phonology
117
Any of underlying /ui, oi, əi/ could produce i2 (13). Among the three, /ui/ appears to be unmarked. In the present author’s research the following words contain i2 or e2 nonword-finally and do not participate in the exposed-covert alternation. (24)
‘fog’, a cognate verb is /kuir]-/ cf. Hattori 1979b: 105–106, 1979d: 110. ‘all’ cf. Hattori 1978: 122, 123, 1979a: 113 ‘(a place name in Nara)’ ‘probably’ ‘shave, plane’ ‘grieve’ (Etymologically the noun nage2ki1 must be nagaiki1 lit. ‘long breath’, while OJ noun nage2ki1 is synchronically a nominalized form derived from the verb nage2k-.) ‘give alms’ ‘travel around’ cf. Hattori 1979b: 115. ‘love, admire’ (The adjective me2Dasi ‘admirable’, attested in the Bussokuseki-ka, must be cognate with this verb. Phonograms in Bussokuseki-ka show no distinction between seion and dakuon, hence the notation with ‘D’).
2.3.4.3 Forms ending in -a but not alternating with -e2 Many words ending in -a show no alternations, e.g. the final vowel of tama ‘pearl, jewel’ is a both in isolation and in a compound as tama-gaki1 ‘beautiful fence’. No alternation like ˟tame213~tama-gaki1 is found. In OJ synchrony, the noun tama is simply not followed by the suffix /i/. Hattori (1979d: 114) says that this tama may be explained as being traced back to tamaai, which changed into tamaa rather than ˟tame2. In guessing Hattori’s line of thinking, the 3-mora syllable maai of tamaai is super long, and is supposed to have changed into maa. In present-day Tokyo Japanese the polite form of adjective-stem+u+gozaimasu is made in a pseudo-Kansai style (-u instead of -ku), as in (25) (where long vowels are represented as geminate): (25) aka-u → a.koo siro-u → si.roo oo-u → ooo
→
oo
akoo gozaimasu siroo gozaimasu oo gozaimasu
o.oo
o.oo gozaimasu
‘it is red’ ‘it is white’ ‘it is many, much’
13 The symbol ˟ is used for non-attested, ungrammatical in distinction from the symbol structed.
✶
for recon-
118
Teruhiro Hayata
As is seen in (25), a three-mora syllable is avoided by contraction into a two-mora syllable or by a split into two syllables. Likewise, in present-day Hakata Japanese (Hayata 1985: 25, 50, 62, 74, 94; 108), the adverbial form of an adjective is formed by suffixation of u (ku in Tokyo) after the adjective stem. The negative forms of adjectives are shown in (26), where the three-mora syllable ooo is obligatorily changed into the two-mora syllable oo. (26)
aka-u → a.koo akoo nai ‘not red’ oo-u → ooo → oo oo nai ‘not many, much’
One of the polite forms of verbs in Hakata Japanese is formed by adding the infinitive morpheme i and the polite morpheme nsyar (nasar in Tokyo) to verb stems as in (27), where PST ta is suffixed: (27)
Note that in (27) the three-mora syllable iin is contracted to the two-mora syllable in rather than the equally two-mora syllable ii. From this fact it may be generally said that in contraction, CViViC is contracted to CViC rather than to CViVi . If this approximates to the truth, Hattori’s idea that /tamaai/ ‘pearl’ could be changed into /tamaa/ (tama) should be questionable. /tamaai/ ought to have changed into /tamai/ (tame2), which is not attested. 2.3.4.4 Formation of i1, e1 and the imperative forms While i2 and e2 are made by (13), i1 and e1 are made by (28). A consonant immediately followed by a front vowel or [y] is palatalized giving a kō-rui syllable as in (28), which is a revised version of (9). (28)
a. b. c.
/Ci/ /Ce/ /Cyə/
→ Cʸi
(Ci1)
→ Cʸe (Ce1)
The rule (28c) is involved in the production of the OJ imperative suffix as in (29). (29)
phonographic writing /na-i]-yə/ → nᵊe-yə neyo2 ‘sleep! go to bed!’ MYS 5.904 /nak]-yə/ → nakʸe nake1 ‘sing! weep!’ MYS 18.4050
Some linguists seem to believe that OJ had no imperative morpheme but that forms similar to verb stems had been used for imperative:
4 Reconstruction of Old Japanese phonology
119
an infinitive form i, which is the old imperative, further followed by an interjection a (this ia is contracted to e [e1]) became the form of imperative of OJ (Ohno 1955: 319) (pre-OJ) conclusive (shūshikei) i was used both as noun as well as an imperative, and for the imperative this i was later followed by an interjection a (for C-stem verbs) ~ö [o2】(for V-stem verbs), which as a result of sound change formed the OJ imperative” (Ohno 1959: 278 – translated by the present author).
Ohno’s statement will be rewritten as in (30). (30) IMP → a / C]-i __ IMP → ö [o2] / V]-i __ After the rules in (30), the rules in (31) must be applied. (31) C-ia → Cʸe (Ce1) V-iö → Vyö (Vyo2) In the opinion of the present author of this chapter, Ohno’s assertion contains many problems; in particular, it is inexplicable that the imperative morpheme should be a for consonant-stem verbs and ö [o2] for vowel-stem verbs (30). There is no explanation for why the alternation between a and ö [o2] should be conditioned not by the yang or yin vowel of the stem but by the difference between consonant and vowel of the stem-final segment. The synchronic rule in (28c) (and the derivations in (29)) would be far more natural than Ohno’s rules in (30) followed by those in (31). Furthermore, it may be safe to say that all Japanese dialects as well as written records through all ages show the existence of morphemes used exclusively for the imperative. In present-day Japanese, in addition to morphemes exclusive for the imperative, i.e. -e for consonant-stem verbs and -ro14 for vowel-stem verbs, the form verb-stem+te is also used as a kind of ‘weak’ imperative, or rather, ‘request’, used alone or followed by kure or kudasai, e.g. okite, okite kure, or okite kudasai ‘please get up!’. In the Kansai dialects, besides the imperative morphemes, the infinitive (instead of -te in Tokyo) is used for ‘request’. The same is true with Korean, which has a proper imperative morpheme as well as abridged forms. In Korean, many words and many verb-stems end in a closed syllable, but consonant-final verb stems alone cannot be used as imperative. “Imperative forms of Tungus verbs appear with an imperative ending” (Ikegami 1999: 357). In Manchu “The [verb] stem . . . at the same time serves as the Imperative” (Möllendorff 1892: 9). It may seem so at first glance, but important verbs of Manchu (‘be (exist), become, come, eat, bring, seek, take, mention, etc.’) show imperative forms by suffixation of ‘irregular’ imperative mor-
14 This -ro must be cognate with OJ yo2. See note 4.
120
Teruhiro Hayata
phemes to verb-stems. “In earlier times Manchu seems to have possessed imperative forms ending in an imperative ending” (Ikegami 1999: 357). In pre-OJ there must have been one process that would contract /Cia/ into [Cʸe], just like (28c). See (39), (40) and (41).
2.4 Problem of consecutive vowels 2.4.1 The so-called “avoidance of consecutive vowels” in OJ It has been widely believed that OJ avoided consecutive vowels, applying monophthongizations and vowel elision as much as possible. However, one should consider the truth of this belief. It is said that in consecutive vowels arising from morphological or syntactic sources, one of three processes occurs: a) the left vowel is deleted, b) the right vowel is deleted, or c) both the left and the right vowels are changed, giving a third syllable. Studies hitherto on OJ monophthongization have been efforts devoted exclusively to finding out under what conditions the above-mentioned a), b), or c) should be selected. These attempts have been pursued by Kishida (1948), Hashimoto (1948), and Yamaguchi (1971), among others. They considered important the relative height of two vowels in a sequence. Thereafter Mōri (1979 etc.) in his studies of hypermeter (jiamari) took into account not only the height of vowels participating in a concatenation of vowels, but also the position of consecutive vowels in poems and the degree of coupling. While studies hitherto have considered ordinary language and poetry language in OJ to be in a continuum, Kwon Kyoung-ae (1999) made a distinction between the two and gave good results in the field of OJ vowel deletion. However, even with this, the problem remains unsolved. The present author considers the problem of this monophthongization in connection with post-OJ forms. In etymologically inferred vowel sequences, both first vowel deletion and second vowel deletion are attested. A sound form existing at a particular time is the result of various processes of different times. Consider the following lists of words including vowel sequences whose first vowel is elided in OJ (32). The words in (32) are divided into two groups: (i) those whose reflexes in later periods of the language are attested and (ii) those whose reflexes are not attested. (32)
(ii) Base ko2-no2-ure ‘ends of twigs (treeGEN-end)ʼ kuni-uti ‘all over the country (country-inside)ʼ paya-uma ‘post horse (fast-horse)ʼ pi1me1-no2-aso1bi1 ‘womanizing (girl-GEN-play)ʼ yo2ko2-usu ‘flat-shaped mortar (horizontal-mortar)ʼ
OJ elided ko2nure
――――
――――
kunuti
――――
――――
――――
――――
――――
――――
――――
――――
payuma pi1me1naso1bi1 yo2kusu
121
After OJ (MJ, ModJ) elided non-elided
Next, consider the words in (33) whose second vowel is elided in OJ. Reflexes in later periods of the OJ words in (33), whose second vowels of vowel sequences are elided, are not transmitted to posterity, but base forms stored in OJ speakers’ brains are transmitted to posterity. OJ forms in (33) are perhaps not genuine second-vowel elided forms, but may be phonetically reduced vowel forms. The compressed, non-elided forms in the last four of (33) and the last one in (34) must be genuine OJ forms. (33)
15 Only a small number of the first vowel elided form tip-, which sounds archaic-like, are attested in MYS. 16 Henceforth the boundary of rhythm will be noted by the vertical line ‘|’.
122
Teruhiro Hayata
In (34) the first vowel and the second vowel in each vowel sequence are the same, and therefore it is impossible to decide between first vowel elision and second vowel elision from OJ forms alone. However, the fact that the later forms in (34) are not OJ forms but base forms means that second-vowel-elided OJ forms are superficial transient forms. Many words in (34) are transmitted in non-elided forms after the OJ period, just as the forms in (33) are. (34)
Let us recapitulate the OJ vowel elision and the resultant forms in later periods as in (35), (36), and (37). (35)
First vowel elision in OJ (cf. (32)) Base OJ After OJ (MJ, ModJ) elided elided non-elided (i) CV1-V2 CV2 CV2 ―――― (ii) CV1-V2 CV2 ―――― ――――
(36)
Second vowel elision in OJ (cf. (33)) Base OJ After OJ (MJ, ModJ) elided elided non-elided CV1-V2 CV1 CV1-V2 ――――
(37)
(Second) vowel elision in OJ (V1=V2) (cf. (34)) Base OJ After OJ (MJ, ModJ) elided elided non-elided CV1-V1 CV1 CV1-V1 ―――― CV1-V1 CV1 ―――― ――――
It is clear from (35) that first-vowel-elided OJ forms are fixed already in the OJ period, i.e. are not OJ synchronic forms. While first-vowel elision destroys the syllable structure of the base form CV1-V2, changing it to CV2, second vowel elision as in (36) and (37) does not destroy the syllable structure of the first syllable. The syllable structure of the second syllable is, though phonographically deleted, perhaps only phonetically reduced. At least underlying (base) forms are intact and therefore these underlying
4 Reconstruction of Old Japanese phonology
123
forms are handed down to posterity. Second vowel elided forms in OJ are in all probability superficial transient forms forced by rhythm rules. Some forms followed by the auxiliary verb ar- show both first-vowel-elided and non-elided forms in later periods as in (38). (38) First vowel elision in OJ Base OJ elided -ku-ar-kar-ku-ar-te-ar-tar-te-arni-ar-nar(compressed) -ni-ar-zu-ar -zar-zu|-ar-
Apparent compounds may be not fixed; they may be phrases. It is important to ascertain whether they are transmitted as compounds or as phrases to later periods. The present author searches for the nature of OJ monophthongization in connection with forms appearing in later periods of the language rather than seeking after conditions for selecting left or right deletion of vowels. Since sources of phonographic writings of OJ are generally scarce and skewed in distribution, statistical study is difficult, but we are still hoping to do the best we can with the materials available.
2.4.2 Summary of OJ consecutive vowels Avoidance of vowel sequences is not, in fact, a characteristic of OJ. Vowel sequences are very common in OJ prose not governed by rhythm. Even in poetry, many vowel sequences are attested. Note that forms like -ku ar- / -kar-, -zu ar- / -zar etc. are coexisting forms, not resulting from OJ synchronic processes. Lexical forms already fixed in pre-OJ and the vowel contraction and compression of OJ poetical materials by rhythm regulation have been erroneously believed to be the result of a general avoidance of vowel sequences. First-vowel deletion is true deletion, giving a compound word. First-vowel deletion destroys completely the vowel of the first syllable, leaving no traces of the vowel, e.g. to2ko2-ipa ‘lit. eternal rock’ → to2ki1pa. If the first vowel o2 of ko2 is only reduced, the form ˟to2ki2pa rather than to2ki1pa would be produced. First vowel deletion is not an OJ synchronic rule, but a pre-OJ rule.
124
Teruhiro Hayata
Second-vowel deletion leaves the vowel of the first syllable intact, but the vowel of the second syllable, while not written phonographically, can in fact not completely disappear. Two OJ forms wa|gi1|mo|ko1|to2 (← waga-imoko1-to2) ‘with my sweetheart’ NSK 61 and wa|ga-i|mo|ko1|ga (← waga-imoko1-ga) ‘my-sweetheart (NOM)ʼ MYS 20.4405 (EOJ sakimori uta) are given just for reference. The former is with the first vowel elided and the latter with both vowels retained and compressed into one syllable. The former form wagi1moko1 must have been fixed in the central dialect of OJ, but the latter form wagaimoko1 remains non-elided only because it is recorded in Eastern Old Japanese.
3 Pre-OJ: things not explained in the synchrony of OJ 3.1 i-a vowel contraction Although not admitted as a synchronic rule of OJ, the rule in (39), which has been mentioned for a long time, must be admitted as a synchronic rule of pre-OJ. (39)
Ci-a → Ce1
OJ examples that result from applying the pre-OJ rule in (39) are shown in (40), which are slightly adapted from Frellesvig (2010: 48). (40)
For the last example ke1sa ‘this morning’ in (40), Ohno, Satake and Maeda (1990) say “probably a compound of a variant changed from ko2 ‘this’ and asa ‘morning’”. A variant from ko2, however, would be ki2 rather than ki1. ki2-asa →ke1sa ([kᵊi-asa] → [kʸesa]) is very unlikely. A consonant followed by [i] with an intervening non-front vowel normally cannot be palatalized unless the intervening non-front vowel is deleted. A similar instance ke1pu ‘today’ is phonographically attested in KK 101. As to ke1pu, Ohno, Satake and Maeda (1990) say “it might be a compound of ki1 ‘this’ and apu ‘meet (合)ʼ just like ke1sa”. A morpheme ki1 meaning ‘this’ has not been attested.
4 Reconstruction of Old Japanese phonology
125
3.2 Compound by vowel contraction It seems safe to say that vowel contraction (V1 + V2 → V3) does not occur in consecutive vowels in the course of OJ compound formation, but in the pre-OJ period, such a process may have been prevalent. The examples in (41) and (43) show not vowel deletion but vowel contraction in VV sequences. (41) sak-i ar-i → sake1ri /sakeri/ ‘is in bloom’ by the pre-OJ rule described in (39) pi1-oki1 → pe1ki1 ʻ(proper name)ʼ Kojiki. vol. 2, Ōjin original annotation The proper name pe1ki1 is attested in Kojiki phonographically but its etymology is unknown. A Korean etymology is not completely impossible; cf. 辟支山 (Nihon-Shoki vol.9). Nihon- Shoki contains 日置部 (vol.6). Wamyō ruijushō (10th century) contains six place names whose etymology is not settled, each with the logographic writing 日置 and phonographic writing pioki,. There are many proper names written logographically as 日置 in Japan even now that have a variety of pronunciations and whose etymology is not settled. For all that, by postulating a rule (42), /peki/ [pʸekʸi] pe1ki1 would be obtained from pi1oki1. This rule covers the rules applied in (41). (42)
Vowel contraction in pre-OJ i + nonhigh V → nonhigh nonlow front V ( i + e, ə, o, a→ e (e1))
The above rule seems to apply in instances in (43), but (43a) does not come from the Central dialect of OJ, and (43b and c) come from EMJ; that is, none are genuine OJ materials. (43)
a. b. c.
woti-omo → wotemo ‘the other side’ MYS 14.3361 (EOJ) puki-opi → pukepi ‘Pukepi (a personal name)’ Nihongi-shiki Kōhon (9th c.) yuki-opi → yukepi ‘quiver carrying soldier’ Wamyō ruijushō (10th c.)
Pukepi in (43b) is a famous general who flourished in the Jinshin War in 672. His name appears many times in Nihon-Shoki vol. 28 (Emperor Tenmu), but phonographic writing of his name is not found in OJ sources. Among several Nihongi-shiki ‘private annotations of Nihon-Shoki’, Kō-hon 甲本 prefaced in Kōnin 10 (819) gives the name written logographically as 吹負 with pukepi in katakana; 吹 is a verb meaning “to blow” ‘puk-i’, 負 is a verb meaning “to carry” ‘op-i’. Yukepi ‘quiver carrying soldier’ in (43c) appears in Nihon-Shoki vol. 17 (Emperor Keitai) written logographically as 靫負, but regrettably no corresponding phonographic writing is attested in OJ sources. The word is found in Wamyō ruijushō (10th c.) 20 vol. version written yukepi in phonographic writing; 靫 is a noun meaning “quiver” ‘yuki’, 負is a verb ‘op-i’.
126
Teruhiro Hayata
3.3 Pre-OJ vowels 3.3.1 pJ vowel systems proposed by internal reconstruction As mentioned above, the six vowel system, /a, i, u, e, o, ə/, is postulated in this chapter for OJ. The OJ five vowel theory with glides /y, w/ was already mentioned in 2.3.2.3. A four-vowel system has been proposed for pJ, (e.g. Matsumoto 1975: 146, 1984: 31, 1995, Ohno 1977: 204, Martin 1987: 67), comprising pJ /✶i, ✶a, ✶u, ✶ə/ corresponding to OJ i, a, u, o2. The pJ four-vowel system was based on the hypothesis that, if pJ ✶ia > OJ e1, it would follow that there was no ✶e in pJ. Such a hypothesis would necessarily lead to a proto-system simpler than a later system―the evolutionary theory. Frellesvig and Whitman (2008: 15) proposed a seven vowel system for pJ based on “primarily internal reconstruction and dialect comparison” (p.16), giving /✶i, ✶ɨ, ✶u, ✶e, ✶ə, ✶o, ✶a/. Applying the comparative method to Japanese and Ryukyuan dialects, and considering written records as well, Hattori (1976a, 1979abc) also proposed seven vowels /✶i, ✶ü, ✶u, ✶e, ✶ə, ✶o, ✶a/ for pJ. After Hattori, comparative studies of Ryukyuan dialects have led to impressive results for pJ. However, reconstruction by the comparative method is beyond the scope of this chapter. Only a bare mention would be made of data for internal reconstruction of pre-OJ. Data available for internal reconstruction are generally limited, being both scarce in quantity and poor in quality. This fact may lead to making use of inappropriate data such as texts partly written with logograms, etymologically uncertain data (e.g. whether cognate or not), data philologically under dispute (e.g. data written with phonograms having two possible sound shapes), data chronologically inappropriate, data not of the central dialect, etc. Studies by internal reconstruction for pre-OJ should be carried out based on solid data, as well as studies by comparative reconstruction for pre-OJ and pJ.
3.3.2 Vowel harmony (VH) A reliable statement obviously relating to OJ vowel harmony is Arisaka’s first law (7), which says that o1 and o2 did not co-occur within a morpheme. This statement, however, would entail that the o1-o2 (/o-ə/) pair might be a yang-yin VH pair, and the domain of VH a morpheme rather than a word. Indeed o1 /o/ was yang and o2 /ə/ was yin, but in fact o1 was not the yang counterpart to the yin o2 . o2 /ə/ was the yin (feminine) counterpart to the yang (masculine) /a/. The a-o2 (/a-ə/) pair was one of the true yang-yin pairs of pre-OJ vowel harmony. Note that Arisaka’s first law concerns OJ, not pre-OJ. In OJ, vowel harmony is hardly alive, only a relic of pre-OJ. Although the so-called genitive particle appears to show the alternation between yang (masculine) /na/ and yin (feminine)/nə/ no2, /na/ occurs only in some fixed words in OJ. The only active genitive marker is /nə/ no2. The list in (44) includes nearly all the phrases with the genitive marker na in OJ sources.
4 Reconstruction of Old Japanese phonology
(44)
a.
b. c.
d.
127
covert forms ma ‘eye’: ma-na-kapi1 ‘under one’s eyes’ ta ‘hand’: ta-na-so2ko2 ‘the palm of the hand’; ta-na-suwe ‘fingertip’ nu ‘fine jade’: nu-na-to2 ← nu-na-oto2 ‘jingling sound of jewelry, sound of jewelry’ kamu ‘god’: kamu-kaze ‘divine wind’, kamu-na-gara ‘lit. nature of god’ noun without exposed-covert alternation ta ‘rice field’: ta-na-kami1 ‘place name, lit. upper part of a rice field’ truncated nouns mi1 (mi1d?) < mi1du ‘water, sea’: mi1-na-ura ‘fortune-telling by flowing of river’; mi1-na-so2ko2 ‘bottom of sea, river’; mi1-na-ki1pa ‘waterside’; mi1-na-to1 ʻ(mouth of) harbor’; u (um?) < umi1 ‘sea’: u-na-para ‘(surface of) sea’; u-na-kami1 ‘beside the sea’; u-na-kudari ‘sailing far down in a boat’ others mo1mo1 ‘hundred’: mo1mo1-na-pi1to2 ‘hundred people, many people’
Other instances of na not attested in OJ phonographically are found only after the MJ period. Incidentally, neither kamu-na-duki ‘October, lit. month of gods’ nor mi-na-duki ‘June, lit. month of water’ are attested phonographically in the OJ sources. Followed by na, nouns with exposed-covert alternation are always covert alternants. Neither i2 nor e2 are followed by na, that is, exposed forms followed by na are not attested. Since na is the masculine counterpart to the feminine no2, nouns followed by na must be masculine. High vowels i1, i2, u are not followed by na, except mi1, nu, kamu in (44), e1 and e2 are not followed by na either. The forms nu ‘fine jade’, kamu ‘god’, mi1 ‘water, sea’, at least in pJ, thought to be masculine based on VH, may be the results of Hattori’s vowel raising (Section 2.3.2.2; Hattori 1979c) from ✶no, ✶kamo, and ✶me respectively. It follows that both ✶o and ✶e were masculine vowels. Hattori (1979a: 116)17 said in comparing Ryukyuan dialects that we cannot but reconstruct ✶medu for the pJ form of the OJ mi1du ‘water’. Whether the truncated form of OJ mi1du is ✶me or ✶ med, Hattori’s vowel raising would predict ✶medu or ✶medo for the OJ mi1du in a nonabridged form. The VH pair a (a)~o2 (ə) is confirmed by the OJ vowel-system of numerals in (45), as well as the vowel-pair of GEN na (na)~no2 (nə). (45) pi1to2 ʻ1ʼ mi1 ʻ3ʼ yo2 ʻ4ʼ puta ʻ2ʼ mu ʻ6ʼ ya ʻ8ʼ
17 Hattori (1979a: 118) said that this opinion had been published already in Hattori(1976d).
128
Teruhiro Hayata
On the assumption that pJ ✶e, ✶o are masculine vowels, the following vowel system of pre-OJ (46) would be established. (46)
feminine: masculine:
i e
ə u a o
Since the symbol /e/ is misleading as a masculine vowel, replacing the masculine counterpart /e/ by /ɪ/, and the feminine /ə/ by /e/ as a feminine counterpart of /a/, we get (47). (47)
feminine: masculine:
i e u ɪ a o
The vowel system in (47) is based only on VH in pre-OJ. We are hoping the vowel system of pJ can be more firmly established by the comparative method. Comparing the Ryukyuan dialects, Hattori (1976c: 16, 1976d, 1979c: 98, etc.) maintains the existence of /✶ü/ in addition to /u/ for pJ. From Hattori’s hypothesis it follows that pJ had seven vowels. As the genitive morpheme na is fossilized in OJ, it can be used for as a diagnostic as to whether the vowels of the preceding noun are masculine or feminine. However, already in OJ, no2 (nə) has become a morpheme irrelevant to the masculine-feminine distinction of the preceding noun. Except the petrified noun+na string, any noun (phrase) can be followed by no2 (nə). It may be, however, that strings made of only feminine syllables followed by no2 (nə) are likely to be pJ feminine: e.g., tu-no2 (tu-nə) ‘wharf-GEN’, isu-no2-kami1 (isu-nə kami) ‘lit. the upper part of Isu, (makura kotoba, place name)’ NSK 94. To the present author’s knowledge, iso-no-kami ‘the upper part of Iso’ is not attested phonographically in OJ, but in the succeeding period when the distinction between o1 and o2 had become lost, the phonographic form iso-no-kami was common. In conjugation, the OJ verb meaning ‘come’ has the mizenkei (‘irrealis’) /kə/ ko2 and the ren’yōkei (‘infinitive’) /ki/ ki1, while the verb meaning ‘come into bloom’ has the mizenkei /saka/ and the ren’yōkei /saki/ saki1. It follows that the former verb ‘come’ shows a stem /k-/ followed by a vowel /ə/ o2 for mizenkei and by a vowel /i/ for ren’yōkei. Likewise, the latter verb ‘come into bloom’ shows a stem /sak-/ followed by a vowel /a/ for mizenkei and by a vowel /i/ for ren’yōkei. It should be noted that OJ occurrences of /ə/ o2 as a mizenkei vowel are extremely limited, and the mizenkei vowel is overwhelmingly /a/. This means that the verb /k-/ ‘come’ in pre-OJ must have been a feminine word in contrast to the masculine verbs such as /sak-/ ‘come into bloom’. Still in OJ, other than /k-/ ‘come’, not a few forms are found with feminine /ə/ o2 in derivation: /kik-/ ‘hear’, / kik-əs-/ ‘hear-RESPect’, /kik-əye-/ ‘hear-PASSive’; /əməp-/ ‘think’, /əməp-əs-/ ‘think-RESP’, /əməp-əye-/ ‘think-PASS’; /təyəm-/ ‘resound’, /təyəm-əs-/ ‘resound-TRANS’, etc. (Hayata 2006: 11). The feature [±masculine] could be extended within the domain of derivative affixes, but can never be extended to inflectional affixes, e.g. /əməp-əs-azu/ but ˟/əməpəs-əzu/ in OJ. In pre-OJ, when vowel harmony was active, /əməp-əs-əzu/ may have been well formed, but there is a possibility that inflectional affixes were independent words.
4 Reconstruction of Old Japanese phonology
129
4 Conclusion The studies of OJ (and a part of pre-OJ) based only on written materials include inevitably many points incompatible with the results arrived at by the comparative method based on materials from Ryukyuan as well as those from the mainland dialects. For example, many specialists of the Japanese language believe yuk- to be older than ik- ‘to go’, although Hattori (1978: 115–116) clearly demonstrated that ik- had changed into yuk- from systematic sound correspondences between Ryukyuan and mainland dialects. Reconstruction of pJ by the comparative method is making remarkable progress now. Results of comparative studies are important for pre-OJ and OJ as well.
References Arisaka, Hideyo. 1931. Kokugo ni arawareru isshu no boin-kōtai ni tsuite [Concerning a type of vowel alternation appearing in Japanese]. Onsei no kenkyū [Journal of the Phonetic Society of Japan] 4. 89–137. Arisaka, Hideyo. 1934. Kodai-nihongo ni okeru onsetsu-ketsugō no hōsoku [Laws of syllable combinations in Old Japanese]. Kokugo to koku-bungaku 11(1). 80–92. Reprinted in Hideyo Arisaka 1957, Kokugoon’inshi no kenkyū [Studies on Japanese historical phonology], 103–116. Tokyo: Sanseidō. Arisaka, Hideyo. 1955. Jōdai-on’in-kō [On Eighth Century Japanese phonology]. Tokyo: Sanseidō. Bloomfield, Leonard. 1933. Language. New York: Holt. Chomsky, Noam. 1964. Current Issues in Linguistic Theory. The Hague: Mouton. Chomsky, Noam and Morris Halle. 1968. The Sound Pattern of English. New York: Harper & Row. Frellesvig, Bjarke. 2010. A history of the Japanese language. Cambridge: Cambridge University Press. Frellesvig, Bjarke and John Whitman. 2008. Evidence for seven vowels in proto-Japanese. In Bjarke Frellesvig and John Whitman (eds.), Proto-Japanese: Issues and prospects, 15–41. Amsterdam: John Benjamins. Gleason, Henry A. 1955. An Introduction to Descriptive Linguistics. New York: Henry Holt and Company. Hashimoto, Shinkichi. 1948. Kokugo no onsetsu kōzō no tokushitsu ni tsuite [On the characteristics of Japanese syllable structure]. Kokugogaku 1. 24–43. Reprinted in Shinkichi Hashimoto 1950, Kokugo-on’in no kenkyū [Studies in Japanese phonology] (Hashimoto Shinkichi Hakase Chosakushū vol. 4 [Collected work of Dr. Shinkichi Hashimoto vol. 4]), 229–260. Tokyo: Iwanami Shoten. Hattori, Shirô. 1955a. Ryūkyūgo Ⅴ. Bunpō [Ryukyuan language, Chapter 5. Grammar]. In Sanki Ichikawa and Shirô Hattori (eds.), Sekai gengo gaisetsu (Ge) [Outline of the world’s languages (2)], 328–353. Tokyo: Kenkyusha. Hattori, Shirô. 1955b. Phonemics (1). Kokugogaku 22. 88–104. Reprinted in Shirô Hattori 1960, 279–301. Hattori, Shirô. 1959. Nihongo no keitō [The Japanese language family]. Tokyo: Iwanami Shoten. Hattori, Shirô. 1960. Gengogaku no hōhō [Linguistic methodology]. Tokyo: Iwanami Shoten. Hattori, Shirô. 1975. Phonological interpretation of Tatar high vowels. Ural-Altaische Jahr- bücher 47. 89–94. Reprinted (with misprints corrected) in 1993, Studies in Altaic languages (Selected Papers by Shirô Hattori vol. 4), 256–264. Tokyo: Sanseidō. Hattori, Shirô. 1976a. Ryūkyū-hōgen to hondo-hōgen [Dialects of the Ryukyus and of mainland Japan]. In Iha Fuyū Seitan Hyakunen Kinenkai (ed.), Okinawa-gaku no reimei [The dawn of Okinawa studies]. 7–55. Tokyo: Okinawa-bunka-kyōkai. Hattori, Shirô. 1976b. Jōdai-nihongo no boin-taikei to boin-chōwa [The Vowel system of the Eighth Century Japanese]. Gekkan Gengo 5(6). 2–14.
130
Teruhiro Hayata
Hattori, Shirô. 1976c. Jōdai-nihongo no iwayuru “hachi-boin” ni tsuite [On the so-called ‘eight vowels’ in Eighth Century Japanese]. Nihon Gakushiin Kiyō 34(1). 1–16. Hattori, Shirô. 1976d. Nihon-sogo no boin-taikei: Reigai o tsuikyū site eta atarashii kasetsu [The vowel system of proto-Japanese: A new hypothesis gained from pursuing exceptions]. Asahi Shimbun, Evening Edition. 122 [22 June]. Hattori, Shirô. 1978. Nihon-sogo ni tsuite [On proto-Japanese] 9. Gekkan Gengo 7(11). 114–123. Hattori, Shirô. 1979a. Nihon-sogo ni tsuite [On proto-Japanese] 19. Gekkan Gengo 8(9). 108–118. Hattori, Shirô. 1979b. Nihon-sogo ni tsuite [On proto-Japanese] 20. Gekkan Gengo 8(10). 105–115. Hattori, Shirô. 1979c. Nihon-sogo ni tsuite [On proto-Japanese] 21. Gekkan Gengo 8(11). 97–107. Hattori, Shirô. 1979d. Nihon-sogo ni tsuite [On proto-Japanese] 22. Gekkan Gengo 8(12). 100–114. Hattori, Shirô. 1982. Mōri Masamori-shi no ‘Man’yōshū ya-wa-gyō no onsei―i-u-no baai’ ni tsuite [On Masamori Mori’s ‘Phonetics of the ya-wa column in the Man’yōshū – the case of i-u’]. Gekkan Gengo 11(8). 94–97. Hattori, Shirô. 1983a. Hashimoto Shinkichi sensei no gaku-on: “Genchō-hishi” on’yaku-kanji no shiyōhō ni genkyū shitsutsu [The scholarship of Shinkichi Hashimoto: With reference to the use of transliteration kanji in “Genchō-hishi”]. Kokugogaku 133. 1–14. Hattori, Shirô. 1983b. Kōen ‘Hashimoto Shinkichi sensei no gaku-on’ hosetsu (1) [Lecture ‘The scholarship of Shinkichi Hashimoto’, Excursus (1)]. Gekkan Gengo 12(3). 78–81. Hattori, Shirô. 1983c. Kōen ‘Hashimoto Shinkichi sensei no gaku-on’ hosetsu (3) [Lecture ‘The scholarship of Shinkichi Hashimoto’, Excursus (3)]. Gekkan Gengo 12(5) 120–123. Hattori, Shirô. 1983d. Kako no gengo no on’in-kyōjitai saikō no hōhō (jō): ‘Jōdai-nihongo’ o rei to shite [Methods for the reconstruction of the historical synchronic phonology of a language (part 1): The case of Eighth Century Japanese]. Gekkan Gengo 12(7). 110–113. Hattori, Shirô. 1991. Hattori Shirô chosho-ronbun mokuroku [A Bibliography of Shirô Hattori’s Writings], enlarged and revised edition. Tokyo: Kyūko-shoin. Hayata, Teruhiro. 1985. Accent and morphology of the Hakata dialect. Fukuoka: Kyūshū Daigaku Shuppankai. Hayata, Teruhiro. 1998. Nihon-go no onsetsu-kōzō to o-retsu-kō-otsu no betsu [Japanese syllable structure and the kō-otsu distinction]. Onsei Kenkyū [Journal of the Phonetic Society of Japan] 2(1). 25–33. Hayata, Teruhiro. 2006. Jōdai-nihongo-boin-chōwa oboegaki [A note on Eighth Century Japanese vowel harmony]. In Chikushi Kokugogaku Danwakai (ed.), Chikushi-gogaku-ronsō Ⅱ: Nihongo-shi to hōgen [The Chikushi linguistic debate II: Dialects and the history of Japanese], 1–16. Tokyo: Kazama Shobō. Ikegami, Jirô. 1999. Manshū-go to Tsungusu-go [Manchurian and Tungus]. Manshū-go Kenkyū [Manchurian studies], 344–358. Kamei, Takashi. 1950. Jōdai-nihongo no onsetsu shi, chi no boin [The vowels of the syllables shi and chi in Eighth Century Japanese]. Gengo Kenkyu 16. 37–47 and 161–162. Reprinted in Takashi Kamei 1984, Kamei Takashi ronbun-shū [Collected work of Takashi Kamei] vol. 2, 125–140. Tokyo: Yoshikawa Kōbunkan. Kikuzawa, Sueo. 1935. Kokugo on’inron [Japanese phonological theory]. Tokyo: Kenbunkan. Kindaichi, Haruhiko. 1975. Akusento kara mita Ryūkyū-shohōgen [Dialects of the Ryukyus from the view of accent]. In Haruhiko Kindaichi, Nihon no hōgen: Akusento no hensen to sono jissō [Japanese dialects: The dialects of Japanese: Accentual change and its present state], 129–159. Tokyo: Kyōiku Shuppan. Kishida, Takeo. 1948. Kokugo on’in-henka no kenkyū [Studies on the phonological changes of the Japanese language]. Tokyo: Musashino Shoin. Kōno, Rokurō. 1955. Chōsengo [Korean language]. In Sanki Ichikawa and Shirô Hattori (eds.), Sekai gengo gaisetsu (Ge) [Outline of the world’s languages (2)], 357–439. Tokyo: Kenkyūsha. Kwon Kyoung-ae. 1999. Vowel elision in ancient Japanese: in connection with rhythm restrictions. Kokugogaku 197. 1–14. Martin, Samuel E. 1987. The Japanese Language Through Time. New Haven, CT: Yale University Press.
4 Reconstruction of Old Japanese phonology
131
Matsumoto, Katsumi. 1975. Kodai nihongo boin soshiki kō: Naiteki saiken no kokoromi [On the vowel system of Old Japanese: An attempt at internal reconstruction]. Kanazawa Daigaku Hō-bungakubu Ronshū: Bungaku-hen 22. 83–152. Matsumoto, Katsumi. 1984. Gengoshi no saiken to gengofuhen [Language history reconstruction and linguistic universals]. Gengo Kenkyu 86. 5–32. Matsumoto, Katsumi. 1995. Kodai nihongo on’inron: Jōdai tokushu kanazukai no sai-kaishaku [Phonology of Old Japanese: A reinterpretation of Eighth Century special kana use]. Tokyo: Hituzi Syobo. Miyake, Takeo. 1933. Kanazukai no kenkyū [Studies in kana use]. Kokugo Kagaku Kōza vol. 12, 46–75. Tokyo: Meiji Shoin. Möllendorff, Paul G. von. 1892. A Manchu Grammar, with analysed texts. Shanghai: The American Presbyterian Mission Press. Mōri, Masamori. 1979. Man’yōshū ni okeru tango-renzoku to tango-ketsugō-tai [Word strings and word combinations in the Man’yōshū]. Man’yō 100. 1–47. Ohno, Susumu. 1955. Man’yōjidai no on’in [Phonology of the Man’yō period]. In Manyōshū-Taisei 6 Gengo-hen [Variorum edition of the Man’yōshū 6, Volume on language], 287–330. Tokyo: Heibonsha. Ohno, Susumu. 1959. Nihongo VII. Rekishi [Japanese, Chapter 7. History]. In Sanki Ichikawa and Shirô Hattori (eds.), Sekai-Gengo-Gaisetsu (Ge) [Outline of the world’s languages (2)], 262–286. Tokyo: Kenkyūsha. Ohno, Susumu, Akihiro Satake and Kingorō Maeda (eds.). 1990. Iwanami kogo jiten [The Iwanami dictionary of earlier Japanese], 2nd and enlarged edn. Tokyo: Iwanami Shoten. Ohno, Susumu. 1977. On’in no hensen 1 [Phonological change 1]. In Iwanami-kōza Nihongo 5: On’in [Iwanami Japanese series 5: Phonology], 147–220. Tokyo: Iwanami Shoten. Pellard, Thomas. 2012. Kita-Ryūkyū-sho-hōgen ni mirareru chōboin to onchō no taiō-kankei ni tsuite: Okinawa-nanbu-hōgen ni okeru C-kei-goi no gotō-chōboin o chūshin ni [On the mutual relation between rhythm and long vowels seen in northern Ryukyuan dialects: With a focus on word-initial long vowels in C-type words in the Southern Okinawa dialect]. Paper presented at the Phonology team of “Typological and historical/comparative research on the languages of the Japanese archipelago and their environs”, National Institute for Japanese Language and Linguistics, 7 August. Pellard, Thomas. 2013. Ryukyuan perspectives on the Proto-Japonic vowel system. In Bjarke Frellesvig and Peter Sells (eds.), Japanese/Korean Linguistics 20. 81–96. Stanford: CSLI Publications. Yamaguchi, Yoshinori. 1971. Kodai-Nihon-go ni okeru boin-datsuraku no genshō ni tsuite [On the phenomenon of vowel loss in Eighth Century Japanese]. Kokugogaku 85. 1–17.
J. Marshall Unger
5 Old Japanese writing and phonology 1 Introduction There are a number of different criteria that one might invoke to estimate the date that divides the Old from the Early Middle stage of Japanese language history, but for the purposes of the present discussion, the appearance of the graphically distinctive letters that developed into today’s hiragana and katakana is a reasonable dividing line. Paleographers have traced changes in the number and form of kana graphemes, i.e. groups of kana associated with particular syllables, from that time onward. From the linguist’s perspective, more important are the changes in the transcriptive functions of the graphemes, for they provide clues to phonological changes in the language itself. Prior to the rise of kana, specimens and varieties of Japanese writing range from inscriptions on metal and in stone to mokkan (inscribed wooden slips), and, starting in the 8th century, books of various kinds. These include ostensibly historical texts such as Kojiki and Nihon shoki, the large poetry anthology Man’yōshū, gazetteers called Fudoki, collections such as the Senmyō, and works of other kinds. These show a wide range of transcriptive and graphic techniques that have attracted scholarly attention from at least the time of the kokugaku renaissance of the 18th century. It is impossible to summarize such a complex variety of primary sources and such a vast secondary literature in the limited space of this article. I therefore concentrate here on what is undeniably the most linguistically important aspect of early Japanese writing, known as the jōdai tokushu kanazukai (S. Ōno 1953), and focus mostly on research about it in English that is not as widely known as it ought to be. When the transcriptive value of a man’yōgana was related to the character’s phonographic value in Middle Chinese, we retroactively call it an ongana. Most ongana represented a single syllable; a few represented disyllabic sequences. When the transcriptive value of a man’yōgana was based on a Japanese gloss customarily applied to the character, we now call it a kungana.1 It is sometimes said that only ongana are useful for understanding OJ phonology (e.g. Miyake 2003: 24–25), but kungana transcriptions can be just as helpful in identifying phonemic distinctions and mergers over time. Every syllable of Early Middle Japanese corresponds to a set of several man’yōgana. These syllables are shown in Table 1.
1 It is not always clear whether a particular character is a kungana or a logogram taking a kun. E.g. in miramu ‘might see’ 見良牟 (MYS 19.4202), the function of 見 is uncertain because miramu is a form of mi- ‘see’, the first syllable of which was customarily used as a gloss (kun) for 見. By contrast, in tukamikakari-te ‘pluck; clutch at’ 束見懸而 (MYS 16.3816), 見 can only be a kungana. For further details of man’yōgana use, see T. Ōno 1962, 1977; Unger 2008a, 2012. https://doi.org/10.1515/9781614512851-006
134
J. Marshall Unger
Table 1: Principal syllables of Early Middle Japanese (phonemic). wa wi we wo
ra ri ru re ro
ya yu yo
ma mi mu me mo
ba bi bu be bo
pa pi pu pe po
na ni nu ne no
da di du de do
ta ti tu te to
za zi zu ze zo
sa si su se so
ga gi gu ge go
ka ki ku ke ko
a i u e o
We can tell which man’yōgana belong to each EMJ syllable’s set by examining nearminimal pairs of transcriptions of the same word or morpheme, taking note of man’yōgana used to write the same EMJ syllable. Consider, for example, the ongana for yo in the words shown in Table 2. Table 2: Minimal evidence for 余=餘=与=用=欲 (Lange 1973 corrected). MYS poem
Spelling
欲妣(比)
3622, 3993
与比
3643
余之 余思 餘之
3946, 3962 3762 3945
与之 与思 餘志
3734 3631 3978
EMJ form
Spelling
yobi ‘call’ yosi ‘good’
Sayopime ‘Sayohime (name)’
佐用比賣
872
佐欲比賣
yoru ‘night’
用流
807
欲流
yo ‘world’
与
4220
余
MYS poem
868 3669 805
yori ‘from’
用利
871
欲利
800, 802
yopi ‘evening’
与比
892, 4399
欲比
3639
yori pa ‘than’
余理波
欲理波
3655
3737
Collectively, they show that all these ongana were used interchangeably in the Man’yōshū poems cited for the OJ syllable antecedent to EMJ yo. For certain syllables, however, one finds few or no near-minimal pairs in OJ texts before a certain date. For example, in early 8th century texts, we do not find the same phonograms for the final syllables of EMJ kami ‘upper’ and kami ‘deity’. Instead, the phonograms fall into two disjoint sets, within each of which interchangeability is shown by near-minimal pairs, but which are not linked by two or more such pairs in texts of a certain date or later. This suggests the hypothesis that, at some time prior to the EMJ period, the final syllables transcribed with the phonograms in each of these sets were phonemically distinct. Note carefully that the complementary distribution of phonograms for later /mi/ is a necessary condition for this hypothesis to be true, but not a sufficient basis for proving it. Since the amount of surviving text from a given locale and period of time is finite and was transmitted (except for durable inscriptions) by hand-copying over many centuries, other possible causes of the observed complementary distribution must be
5 Old Japanese writing and phonology
135
eliminated before we can conclude that the complementary distribution of phonograms reflected phonemically distinct syllables. By tracing the uses of man’yōgana and the kana into which they graphically evolved over time, we observe the gradual loss of once strictly maintained complementary distributions. Graphic distinctions consistently made in such texts as Kojiki of 712 or the Bussokuseki-ka of 752 do not carry forward to the Nihon kōki of 840 and Wamyōshō of 934. We interpret these losses under the natural assumption that, if writers cease making certain phonemic distinctions in their speech, they will be disinclined to indicate them in writing. They may continue to write words in the manner to which they are accustomed (orthographic conservatism), so the motivating phonemic differences may appear to persist for a while after they have been lost due to sound changes, but soon previously distinct graphemes will merge.2 Therefore, when we find once well-maintained graphemic distinctions being violated for no apparent reason, it is a strong sign that the phonemic differences that originally informed the graphemic differences were no longer accessible to writers. It is on this basis that we use the step-by-step loss of graphemic distinctions in the chronological sequence of OJ texts to deduce the history of sound changes resulting in the merger of once distinct syllables until we arrive at the syllable inventory in Table 1.
2 Hashimoto’s interpretation and subsequent improvements On the basis of his study of man’yōgana usage, Shinkichi Hashimoto, building on ideas of the early 19th-century scholar Ishizuka Tatsumaro, first hypothesized that Old Japanese had the syllable inventory arrayed in Table 3, where I have used subscript numerals 1 and 2 to indicate his A (kō 甲) and B (otsu 乙) syllable classes.3 He concluded that there were two OJ syllables for a single EMJ syllable if he found two groups of words containing the EMJ syllable written with man’yōgana belonging to two mutually exclusive sets. But, perhaps dissatisfied with the asymmetries of this array of syllables, he continued to investigate the contrasts among OJ syllables not found among EMJ syllables. In a series of papers (collected in Hashimoto 1950, 1959), he explained how he came to realize that the phonograms he had previously taken to represent syllables e1 and e2 actually represented ye and e, respectively; that, contrary to what he had first claimed, there was
2 On the other hand, some writing systems do not notate every phonemic distinction. Even distinctions once notated may not be if the writing systems undergoes a major structural change. This, in fact, happened in Japan. Dialect comparisons show that phonemic differences between dakuon and seion obstruents can be traced back to the earliest stages of the language, but, when graphically distinct kana began to replace man’yōgana, writers started to ignore them; even after the invention of the dakuten diacritic, dakuon were not marked consistently until the Meiji period (Martin 1987: 84–120). 3 Yasuda (2003) argues that Hashimoto knew of and copied from Ishizuka’s work more than he admitted.
136
J. Marshall Unger
only one kind of syllable po (and bo); and that one of his two kinds of nu was actually a second variety of no. Making the necessary adjustments in Table 3, Hashimoto could say that all the inferred “extra” OJ syllables corresponded to EMJ syllables ending in i, e, or o, as shown in Table 4. Table 3: Phonemically distinct syllables of Old Japanese (Hashimoto after Ishizuka). wa wi
ra ri
ya
ru
yu
we
re
wo
ro1 ro2
yo1 yo2
ma mi1 mi2 mu
ba bi1 bi2 bu
pa pi1 pi2 pu
me1 me2 mo
be1 be2 bo1 bo2
pe1 pe2 po1 po2
na ni
da di
ta ti
za zi
sa si
nu1 nu2 ne
du
tu
zu
su
de
te
ze
se
no
do1 do2
to1 to2
zo1 zo2
so1 so2
ga gi1 gi2 gu
ka ki1 ki2 ku
a i
ge1 ge2 go1 go2
ke1 ke2 ko1 ko2
e1 e2 o
ga gi gï gu ge gë go gö
ka ki kï ku ke kë ko kö
a i
u
Table 4: Old Japanese (Hashimoto revised; italicized syllables are type A). wa wi
ra ri
ya
we
ru re
yu ye
ro rö
yo yö
wo
ma mi mï mu me më mo
ba bi bï bu be bë bo
pa pi pï pu pe pë po
na ni
da di
ta ti
za zi
sa si
nu ne
du de
tu te
zu ze
su se
no nö
do dö
to tö
zo zö
so sö
u e o
This opened the door to hypothesizing that the distinctive ways of writing the extra syllables had to do with a difference between their vowels and those of syllables with the same EMJ reflexes. Hashimoto thus set up eight OJ vowels, though he entertained the idea that three of them might involve diphthongs or intrasyllabic glides of some kind; tentatively, he used diereses (umlauts) to mark the vowels of B-type syllables as a bookkeeping device without any particular phonetic implications. This left the problem of deciding which sets of man’yōgana ought to be associated with which putative unitary vowels. Hashimoto paid careful attention to the distributions of sets of man’yōgana in the endings of verbs and adjectives, and made them the foundation of his theory because they led to a consistent labeling of the two types of i- and e-ending syllables he had discovered.4 Indeed, had the resulting labeling turned
4 Miller (1967a) blundered when he wrote that “the traditional [i.e. Hashimoto’s] labelling is significant . . . of nothing more than the order of the original philological discovery of individual OJ lexical items.” Nothing could be farther from the truth.
5 Old Japanese writing and phonology
137
out to be inconsistent, Hashimoto would no doubt have become skeptical of a purely phonemic interpretation of the grapheme data. Since o-ending syllables seldom occur in verb stems, Hashimoto had to make reference to the Middle Chinese rhyme classes associated with ongana for those syllables to decide which to call type A and which type B. Either assignment would do because there are no minimal pairs of o-ending syllables in central OJ inflectional paradigms.5 But the use of Chinese rhymes merely served as the basis for analogical reasoning and played no decisive role in establishing the core of Hashimoto’s theory. As Hashimoto had come to see in the case of po, the mere fact that different man’yōgana were used for a single EMJ syllable did not necessarily mean that it had more than one OJ antecedent. Scribal error or the fortuitous lack of a particular word in texts of similar provenance might also make it appear as if there were two mutually exclusive sets of man’yōgana. Here the American linguist Roland Lange (1968, 1973) took an important step beyond Hashimoto. When two man’yōgana are found for the same EMJ syllable in at least two different words (one minimal pair each), or twice in the same word (two minimal pairs), the only plausible reason is that both man’yōgana stood for the same syllable.6 Hashimoto compiled lists of OJ words based on negative evidence: the phonograms used to write a particular EMJ syllable in words on one list did not appear for the same EMJ syllable in the words on the other. Lange instead compiled lists of phonograms based on affirmative evidence: he used the criteria just mentioned to decide whether or not two phonograms transcribed the same OJ (not EMJ) syllable, and therefore had to be assigned to the same list. The logic here is not hard to understand. The amount of phonographic material in OJ texts from the same time and place is finite. Therefore, if we merge lists of phonograms each of which contains one or more phonograms that satisfy the criteria, we must eventually exhaust the data. The number of sets (graphemes) that remain when we reach that point, even if greater than the number of EMJ syllables, is the minimum possible for the data, and therefore necessarily also the number of phonemic OJ syllables for the corpus of contemporaneous text we are studying. Lange’s improvement on the traditional Japanese approach to determining OJ graphemes is subtle but highly important.7 Among other things, it reveals that in the 5 Although a few forms, such as the mizenkei of ku ‘come’, which is ko2, end in o, none ends in a syllable of the form Co1. 6 If one thinks that two minimal pairs might be too likely accidental, one can stiffen the criteria by demanding, for instance, that two man’yōgana be used to write the same EMJ syllable in three or more different words. But even in a large corpus, some syllables (e.g. zo) are so infrequent that it may be hard to identify even a single man’yōgana for the corresponding EMJ syllable. Heuristically, therefore, the criteria given here, which Lange devised, are an excellent compromise. 7 To use an analogy from East Asian tradition, this is like the difference between the Japanese and Chinese methods of scoring a game of go (Chinese weiqi, Korean paduk). In most but not all cases, one gets to the same result. Japanese take the difference of empty points bounded by stones of the same color less the number of stones of that color captured during the course of the game. Chinese take the difference of
138
J. Marshall Unger
8th-century central-dialect poetry of Man’yōshū, the man’yōgana used to write to, do, ro, and yo belonged, respectively, to a single grapheme, though one or more of these syllables may have been the product of a merger of pre-OJ A- and B-type syllables. For example, the data in Table 2 show affirmatively that ongana for Hashimoto’s yo1 and yo2 were used interchangeably in the poems cited. There were not many overlapping cases, and this is probably why Hashimoto was reluctant to let go of the hypothesis that a corresponding phonemic distinction for these syllables existed in Man’yōshū. But, as Lange showed, there were enough near-minimal pairs to justify putting all ongana for yo1 and yo2 into a single grapheme. Lange’s results by no means call the whole Ishizuka-Hashimoto theory into question. They do, however, suggest that the traditional methodology fails to take chance absence of certain words and scribal errors in the texts studied fully into account, and therefore overestimates how long some of the corresponding phonemic differences among syllables persisted.
3 The case of EMJ /si/ and /zi/ In practice, almost all scholars in the field, both in Japan and elsewhere, have at least tacitly acknowledged the logic of Lange’s method for many years. We see this, for instance, in the fact that, although Mabuchi (1957) marshaled considerable evidence using traditional methods for distinct A- and B-type syllables in Kojiki corresponding to EMJ /si/ and /zi/, no one – including Lange’s critics – has accepted Mabuchi’s claims. It is not hard to understand what has motivated this skeptical reaction to Mabuchi’s proposal. The heart of the argument is that, among morphemes that appear two or more times in Kojiki, a syllable reflected as EMJ /si/ is invariably written 斯 in a set of 14 words (including 3 proper nouns) and 志 in a disjoint set of 17 others. This is the sort of evidence that, for instance, Bentley (2002) regards as decisive. Since most syllables in Kojiki are written with just a single ongana, the fact that the sets of words do not overlap appears to favor the hypothesis that 斯 and 志 stood for phonemically distinct syllables. Mabuchi goes on to note that the rentaikei of the past auxiliary is written 21 times with 斯 but just twice with 志, whereas the ren’yōkei of the honorific auxiliary is written 16 times with 志 but only 4 with 斯. So too, the ren’yōkei of quadrigrade (yodan) verbs in s are written 8 times with 斯 but only 4 with 志. Mabuchi implies we are justified in treating the less numerous spellings as mistakes (goyō) since we know that all such endings must be of one type or the other (A or B). But could not one argue that, precisely because the spellings are inconsistent with verbal morphology, the usage of 斯 and 志 in these endings militates against a phonemic difference in the corresponding syllables? Indeed, Mabuchi reports
the total size of all the white and black groups on the board, not distinguishing empty and filled points or counting captures. Even in Japan, it is recognized that the Chinese method is better at clarifying certain unusual configurations.
5 Old Japanese writing and phonology
139
that 斯 and 志 are used interchangeably in the ren’yōkei of su ‘do’, the endings of siku adjectives, the conjoining particle si, the auxiliaries -rasi and -masi, and in a dozen miscellaneous common words.8 This is precisely the sort of evidence that Lange would point to as showing that 志 and 斯 belonged to the same grapheme in Kojiki. There are 5 other ongana in Kojiki for EMJ /si/ and 2 for EMJ /zi/, but they are much less frequent than 志 or 斯, which occur 216 and 205 times in the text, respectively. There are just 24 師, all in proper nouns; 14 紫, only in the name Tukusi; 6 新, only in the name Siragi; 12 色, only in the name [Ikaga-]Siko[-wo-no-mikoto]; and one 芝 for the pronoun si. For /zi/, there are 27 士 and just 4 自, all in the two words izimu and sizimu. To decide how to group these residual ongana with 志 and 斯, and which of these two to designate as type A and which as type B, Mabuchi makes use of Middle Chinese reconstructions and concepts such as Arisaka’s (1955: 1–171) so-called laws of syllable combination (boin ketsugō no hōsoku). While this bootstrapping approach is understandable, the argumentation involved is torturous and still leaves loose ends. For example, Mabuchi is forced to declare the writing of neziro ‘den’ 根白 with B-type zi a mistake because it is presumably a rendaku compound of ne ‘root’ and siro ‘white’, which is written elsewhere with 斯 and Mabuchi previously concluded stood for A-type si. This example clearly shows the advantage of approaching the analysis of OJ graphemes as Lange did. Unfortunately, the significance of his work has largely been overlooked. One reason for this may be that a production error in Lange 1973, caught by Mabuchi (1974) and Mathias (1974), obscured his results. Part of the table in Appendix B showing some of the near-minimal pairs proving that there was only one grapheme for /yo/ in the corpus was omitted in printing. Table 2 includes the missing data, which I obtained from Lange himself (see also Unger 1977 [1993]). A second reason is that Martin (1987) did not discuss the grapheme identification problem. He simply kept track of all of the A-B distinctions discussed in such works as Miller 1967b. There was also a problem in Lange’s reconstruction, shown in Table 5, that cast a shadow over his results. Table 5: Old Japanese (Lange; italicized syllables are type A). wa wi we wo
la li lu le lo
ja je ju jo
ma mi mu me mo mji mje
ba bi bu be bo bji bje
pa pi pu pe po pji pje
na ni nu ne no
nwo
da di du de do
ta ti tu te to
za zi zu ze zo
sa si su se so
(zwo)
swo
8 The noun sisi ‘boar’ is, in fact, written 斯志 and 志斯 two times each!
ga gi gu ge go gji gje gwo
ka ki ku ke ko kji kje kwo
˙a ˙i ˙u ˙e ˙o
140
J. Marshall Unger
The problem has nothing to do with Lange’s use of Karlgren’s Middle Chinese as a starting point for his reconstruction of ongana syllabic contrasts or his use of intrasyllabic glides to phonemicize the A-B distinctions – two points about which some critics have complained. The difficulty is Lange’s conclusion that the A-B designations for e-ending syllables should be reversed for ke and ge. As pointed out independently in Mabuchi 1974 and Unger 1974, this runs afoul of Hashimoto’s principle for labeling Aand B-type syllables in the first place. As it turns out, the apparent reversal after velars was due to an accidental feature of Lange’s corpus, in which ke and ge were fortuitously infrequent.9 To set things aright, I proposed Table 6 in Unger 1977 [1993], taking the opportunity also to make a change in the phonemicization of i-ending syllables. Table 6: Old Japanese (Unger, rearranged; italicized syllables are type A). wa wi we wo
ra ri ru re ro
ya ye yu yo
ma mi mu me mo mye mwi
ba bi bu be bo bye bwi
pa pi pu pe po pye pwi
na ni nu ne no
nwo
da di du de do
ta ti tu te to
za zi zu ze zo
sa si su se so
(zwo)
swo
ga gi gu ge go gye gwi gwo
ka ki ku ke ko kye kwi kwo
a i u e o
Martin (1987) basically accepted my view of OJ vocalic nuclei, which followed Lange’s, but retained those distinctions of Hashimoto that Lange had been unable to verify for his corpus and adopted some notational conventions of his own. For Martin, the phonemic distinctions were the important thing, and he was quite comfortable with intrasyllabic glides in the reconstruction. They not only implied that there were no sudden changes in the number of unitary vowels of the kind proposed in Miller 1967b, but also made the advent of Sino-Japanese borrowings with intrasyllabic glides somewhat easier to understand. It is important to emphasize that, though Lange studied ongana in Man’yōshū with an eye to comparing ongana graphemes with Middle Chinese syllabic values, his analysis of OJ graphemes was logically prior to and independent of the particular reconstruction (Karlgren’s) of Middle Chinese syllables from which he proceeded to drew inferences about what made the syllables corresponding to each grapheme phonemically distinct. Although ongana outnumber kungana, the transcriptive values of most ongana were not based on actual received pronunciations of the corresponding Chinese syllables, as
9 Curiously, in his 1971 book, Miller accepted the velar-flipped view of e-ending syllables, apparently because it echoed some ideas of Hattori’s that he mentioned in his article Miller 1967a. Later, he gave up this idea and became Lange’s fiercest critic.
5 Old Japanese writing and phonology
141
Case (2000: 7–14) carefully explains. Although Miyake (1999, 2003) was aware of this, he dwelt on Chinese historical phonology almost to the exclusion of Japanese internal evidence, simply carrying over all the syllabic distinctions listed in Martin 1987. For his corpus, the songs of Kojiki and Nihon shoki, Miyake should instead (or additionally) have done the sort of minimal-pair analysis that Lange carried out for his sampling of Man’yōshū. Had he done so, he might not have dismissed the study of kungana as having nothing of value to teach us about OJ phonology. Because kungana are based on Japanese glosses, every instance of a kungana in effect implies a word pair (the word being written and the word upon which the kungana value was based) that may be of analytic value, as Beckman (1976), inspired by Lange’s example, observed (Unger 2008b). Significantly, Beckman showed that if one applies criteria similar to Lange’s to kungana, some A-B distinctions, such as ki ≠ kwi and swo ≠ so, which Lange accepted on the basis of ongana, are not supported for Man’yōshū.10 We still await an explicit analysis in which ongana and kungana evidence from texts of similar place and time are assessed co-equally, but there is little doubt, in light of evidence from internal reconstruction and work on the early relationship of Japanese and Korean, that distinctions such as ki ≠ kwi once existed even if they were waning earlier than has been generally believed.11
4 Limitations on inferences drawn from Old Japanese writing Lange’s approach to grapheme analysis helps clarify other aspects of OJ phonology. For example, Matsumoto Katsumi once argued that, as Arisaka had observed, there appear to be no A- and B-type o-ending syllables in polysyllabic minimal OJ word pairs, and very few monosyllabic minimal pairs exist. He argued that their customary man’yōgana spellings could be accounted for even if they did not contrast phonemically. On this view, all man’yōgana corresponding to the o-ending syllables in Table 4 were in complementary distribution; therefore, no phonemic distinction between Hashimoto’s /o/ and /ö/ had actually existed. S. Ōno was unable to refute this argument, but it is obviously false provided one compiles graphemes as Lange did. For if there were no /Cwo/ ≠ /Co/ distinction for some /C/, then there would be two or more near-minimal pairs of man’yōgana spellings requiring the associated phonograms sets to be merged into a single grapheme. The problem for S. Ōno was that he, like his teacher Hashimoto, based
10 The Akihagino mokkan discovered in mid 2008, so-called because its obverse has 阿支波支乃之多波 毛美(?智)(?), which suggests the first two lines of MYS 10.2205, OJ aki pagwi no / sitaba momitinu (Frydman 2012), seems to support Beckman inasmuch as 支 normally represents OJ ki, not kwi, gi, or gwi. 11 See, for instance, Francis-Ratte 2016.
142
J. Marshall Unger
his recognition of A-B graphemes, and hence A-B syllables, entirely on negative graphic evidence.12 A similar problem arises in the case of the phonemic oppositions between what Martin calls seion and dakuon obstruents. As previously remarked in fn. 2, Martin (1987) devoted an entire, long chapter to them. The principal difficulty is that, even after the graphic development of kana, dakuten were not consistently used. Consequently, some authors have denied the existence of seidaku contrasts in the age of man’yōgana altogether. From a linguistic perspective, this seems obviously wrong since all later stages of the language do reflect seidaku contrasts. But unless one is careful to establish ongana graphemes without appeal to comparisons with MCh reconstructions, as Lange did, one lacks a rigorous method for linking knowledge of seidaku contrasts in stages succeeding Old Japanese to man’yōgana graphemes of Old Japanese.13 A third case to consider is the claim that in parts of Nihon shoki, there were ongana graphemes used differentially to indicate OJ syllable pitches. Specifically, it has been claimed that, in these texts, ongana for which one finds MCh tone 1 signified OJ high pitch whereas ongana for which one finds MCh tones 2, 3, or 4 signified low pitch. Even proponents of this theory admit, however, that there are many exceptions to this generalization.14 The question thus becomes, how shall we distinguish between “exceptions” and evidence proving that the ongana in question actually belonged to the same grapheme? Once again, the non-coöccurrence of ongana for the same syllable in different words is negative evidence, whereas multiple coöccurrences of ongana for the same syllable in the same words in contemporaneous texts can only be plausibly be ascribed to phonemic identity. Lange’s methodology is therefore important beyond what it tells us about OJ vowels. Nevertheless, the most secure inferences about OJ consonants and suprasegmentals are grounded in other kinds of evidence, including the comparison of modern dialects, Late Middle Japanese romanized materials, EMJ and LMJ texts with diacritic markings (dakuten, shōten, etc.), the reconstruction of intermediate stages in the rise of so-called onbin forms in Early Middle Japanese, and the results of internal reconstruction.
12 One may note two additional points. First, Matsumoto’s argument depends on the same negative evidence that inspired Arisaka to postulate his “law.” The use of this law in internal reconstruction is a risky business, although Lange’s criticism of the use of internal reconstruction in casting light on OJ phonology (Lange 1973: 47–59) was arguably too severe. Second, there are some polysyllabic minimal pairs involving /i/, such as kwosi (no kuni) ‘(place name)’ vs. kosi ‘came’ that Matsumoto ignored, presumably because at least one member of pair seemed to be a analyzable. 13 The problem is that the phonetic basis for the phonemic seidaku distinctions had to do with the nasalization of preceding vowels. See Unger 2008a. 14 For a critical discussion of such claims, see Mori 2003. See also De Boer 2010 on the history of pitch accent.
5 Old Japanese writing and phonology
143
References Arisaka, Hideyo. 1955. Jōdai on’in-kō [Notes on Old Japanese phonology]. Tōkyō: Sanseidō. Beckman, Mary E. 1976. Kungana evidence for the number of phonemically distinct syllables in Old Japanese. Berkeley, CA: University of California master’s thesis. Bentley, John R. 2002. The spelling of /mo/ in Old Japanese. Journal of East Asian Linguistics 11. 349–374. Case, Theresa Leyden. 2000. Kana in the eighth-century: An ancient Japanese writing system. Ithaca, NY: Cornell University dissertation. De Boer, Elisabeth. 2010. The historical development of Japanese tone: from proto-Japanese to the modern dialects [and] the introduction and adaptation of the Middle Chinese tones in Japan. Wiesbaden: Otto Harrassowitz. Francis-Ratte, Alexander T. 2016. Proto-Korean-Japanese: A new reconstruction of the common origin of the Japanese and Korean languages. Columbus, OH: The Ohio State University dissertation. Frydman, Joshua. 2012. Poetry transmission before the Man’yōshū: Mysteries of the Akihagino mokkan. Paper presented at the Association of Japanese Literary Studies, Columbus, Ohio, 12–14 October. Hashimoto, Shinkichi. 1950. Kokugo on’in no kenkyū [Studies in Japanese phonology]. Hashimoto Shinkichi hakase chosakushū [Collected works of Dr. Shinkichi Hashimoto] 4. Tokyo: Iwanami Shoten. Hashimoto, Shinkichi. 1959. Kokubunpō taikeiron [Systematic grammar of Japanese]. Hashimoto Shinkichi hakase chosakushū [Collected Works of Dr. Shinkichi Hashimoto] 7. Tokyo: Iwanami Shoten. Lange, Roland A. 1968. The phonology of 8th-century Japanese. Ann Arbor, MI: University of Michigan dissertation. Lange, Roland A. 1973. The phonology of eighth-century Japanese: A reconstruction based upon written records. Tokyo: Sophia University Press. Mabuchi, Kazuo. 1957. Kojiki no si, o, po no kana [On the kana for si, o, po in Kojiki]. Kokugogaku 31. 61–90. Mabuchi, Kazuo. 1974. Review of Lange 1973. Kokugogaku 98. 35–39. Mathias, Gerald B. 1974. Review of Lange 1973. Journal of the Association of Teachers of Japanese 9(2–3). 62–76. Martin, Samuel E. 1987. The Japanese language through time. New Haven, CT: Yale University Press. Miller, Roy Andrew. 1967a. Old Japanese phonology and the Korean-Japanese relationship. Language 43(1). 278–302. Miller, Roy Andrew. 1967b. The Japanese language. Chicago: University of Chicago Press. Miller, Roy Andrew. 1971. Japanese and the other Altaic languages. Chicago: University of Chicago Press. Miyake, Marc H. 1999. The phonology of eighth-century Japanese revisited: Another reconstruction based upon written records. Honolulu, HI: University of Hawai‘i at Mānoa dissertation. Miyake, Marc H. 2003. Old Japanese: A phonetic reconstruction. London/New York: RoutledgeCurzon. Mori, Hiromichi. 2003. Nihon-shoki seiritsu-ron shōketsu: Awasete man’yōgana no akusento yūsen-rei o ron-zu [Summary of theories about the composition of Nihon shoki, and the thesis that man’yōgana represented pitch accents]. Kokugogaku 214. 1–15. Ōno, Susumu. 1953. Jōdai kanazukai no kenkyū: Nihon shoki no kana o chūshin to shite [Studies in the ancient kana usage focusing on the kana of Nihon shoki]. Tokyo: Iwanami Shoten. Ōno, Tōru. 1962. Man’yōgana no kenkyū [Studies in man’yōgana]. Tokyo: Meiji Shoin. Ōno, Tōru. 1977. Shintei manʼyōgana no kenkyū: Kodai Nihongo no hyōki no kenkyū [Studies in man’yōgana revised: Research on Old Japanese orthography]. Tokyo: Takayama Honten. Unger, J. Marshall. 1974. Review of Lange 1973. Canadian Journal of Linguistics 19(2). 217–224. Unger, J. Marshall. 1977 [1993]. Studies in Early Japanese morphophonemics. Bloomington: Indiana Linguistics Club. Unger, J. Marshall. 2008a. Early Japanese lexical strata and the allophones of /g/. In Bjarke Frellesvig and John Whitman (eds.), Proto-Japanese: Issues and Prospects, 43–53. Amsterdam: John Benjamins.
144
J. Marshall Unger
Unger, J. Marshall. 2008b. Dating the collapse of kō-otsu distinctions and the status of /mo/. In Bjarke Frellesvig, Masayoshi Shibatani and John Charles Smith (eds.), Current issues in the history and structure of Japanese, 339–350. Tokyo: Kurosio Publishers. Unger, J. Marshall. 2012. Describing Old Japanese kanji usage: The inadequacy of traditional terminology. Scripta 4. 93–105. Yasuda Naomichi. 2003. Ishizuka Tatsumaro to Hashimoto Shinkichi: Jōdai kanazukai no kenkyū o sai-kentō-suru [Ishizuka Tatsumaro and Hashimoto Shinkichi: Re-examining research on Old Japanese man’yōgana usage]. Kokugogaku 54(2). 1–14, 109.
Sven Osterkamp
6 Late Middle Japanese phonology, based on Korean sources 1 Introduction Korean sources form one of the chief foreign corpora that are of interest for the study of Late Middle Japanese (LMJ) and later phonology and phonetics, as they contain transcriptions of contemporary Japanese into either han’gŭl – which incidentally provides us with the earliest known alphabetical renderings of Japanese – or also into Chinese characters employed for their Sino-Korean sound values, or a mixture of both.1 Chiefly dating from the late 15th to the late 18th century, they serve as witnesses of a transitional period, in which phonological and phonetic features typical of or, at the very least, common in LMJ are gradually being lost. In this chapter, we will first give an overview of the relevant sources in Section 2 and then briefly describe what we may learn from these materials about LMJ and later phonology and phonetics. Specifically, Section 3 will address issues concerning consonants, chiefly prenasalization of /b, d, g, z/ (3.1), affricatization of /t, d/ (3.2), delabialization of /f/ (3.3), as well as /-t/ as coda (3.4). Issues concerning vowels will be discussed in Section 4, concentrating on the notion of a “strongly palatalized” /e/ (4.1), on the centralization of /u/ after sibilants (4.2), the predecessors of modern /oo/ (4.3), as well as vowel devoicing (4.4).
2 The sources If taken in the broadest sense possible, Korean sources comprise the following three groups of materials, which all to a greater or lesser extent derive from the context of diplomatic relations between Korea and Japan. 1) Textbooks and dictionaries compiled for the study of Japanese at the Chosŏn period Sayŏgwoñ 司譯院, or Bureau of Interpreters (A), as well as closely related materials (B):
1 McCune–Reischauer Romanization is used in this chapter to transcribe Korean names, titles etc., whereas linguistic examples (chiefly renderings of Japanese in Korean script) are given according to the Yale system. https://doi.org/10.1515/9781614512851-007
Irop’a 伊路波 (Iroha) Ch’ŏphae sinŏ 捷解新語 (Rapid Understanding of the Recent [Japanese] Language) [CS1]2 woodblock reprint of the preceding [CS1b] Kaesu Ch.S. 改修捷解新語 (Ch.S., Revised; 1st revision) [CS2] Ch.S. (2nd revision, not extant) [CS3] Chunggan kaesu Ch.S. 重刊改修捷解新語 (Ch.S., Revised and Reprinted; reprint of CS3) [CS3b] Waeŏ yuhae 倭語類解 (Japanese Words, Classified and Explained) [WY] Iryŏp’a 伊呂波 (Iroha) Ch.S. munsŏk 捷解新語文釋 (Ch.S., Explained in [Mixed] Script; does not contain sound glosses) [CS4] Pangŏn yusŏk 方言類釋 (Local Expressions, Classified and Explained; also: Pangŏn chipsŏk 方言集釋; ms.) Samhak yŏgŏ 三學譯語 (Translations of Words from Three Fields of Study [i.e. Mongolian, Manchu, and Japanese Studies]; ms.) Hwaŏ yuhae 和語類解 (Japanese Words, Classified and Explained; ms.)
The CS textbook series as the core of this group of materials constitutes one of the larger bilingual corpora of pre-modern Japanese. Each installment of the series is of a dual nature, containing both spoken (vols. I–IX) and formal written language (correspondence in sōrōbun; vol. X, split into Xa–Xc since CS2). With the exception of Irop’a (and also
2 In the Korean and Japanese literature, the chief extant editions of Ch’ŏphae sinŏ [J Shōkai shingo] are usually referred to as wŏn’gan-bon [genkan-bon] 原刊本 (= CS1), kaesu-bon [kaishū-bon] 改修本 or cheilch’a kaesu-bon [daiichiji kaishū-bon] 第一次改修本 (= CS 2) and chunggan (kaesu)-bon [jūkan (kaishū)-bon] 重刊(改修)本 or cheich’a kaesu-bon [dainiji kaishū-bon] 第二次改修本 (= CS 3b) respectively. Special caution is necessary with the designation kaesu-bon [kaishū-bon], however, as it was first often used (and sometimes still is) with reference to CS3(b). It was only after CS2 had become generally available to scholarship in facsimile (first in Kyōdai 1987) that it shifted to refer to CS2 instead. 3 Even in recent publications this work is still dated to “?1700” (Unger 2009: 77) – probably following Martin (1987: 78) – or the “early eighteenth century” (Lee and Ramsey 2011: 135), but more and more evidence has been adduced since the late 1950s (see e.g. Ōtomo 1959: 52; Nakamura 1961: 18–20; Yasuda 1966 [1980: 364 and 383, n. 19] etc.) suggesting that this cannot possibly be correct as far as the extant printed edition is concerned. The dating adopted here – which was already arrived at by Yasuda, who narrowed the possible date of printing down to the years 1783 to 1789 – is the consensus view among specialists nowadays. 4 The only known exemplar of Iryŏp’a as an isolated work rather than appendix to CS3b/4 is the one kept at Biblioteca Apostolica Vaticana (shelf-mark Borg.cin.400). It carries a manuscript note by Franciscan brother Romualdus, likely Romuald Kocielski (1750–1791), who sent it to Rome from Beijing where he arrived in 1782. Also cf. Osterkamp (2010: 316; 2016a: 63–67).
6 Late Middle Japanese phonology, based on Korean sources
147
Iryŏp’a, if this is considered in separation from CS3b) these sources have a noteworthy advantage over the other materials referred to below in that their quantity is significant enough to provide us with a relatively gapless picture of the phonological system of Japanese and also to allow generalizations. At the same time, one may expect a certain degree of standardization especially from the materials which actually made it into print and were used for educational purposes on a larger scale (i.e. CS1 to CS4 as well as WY). A persistent problem in the study of these materials is the question as to the exact identity of the variety (or varieties) of Japanese reflected in them. The correspondence in sōrōbun (Irop’a and vol. X of CS) represents a formal written and in itself super-regional variety of Japanese. Thus, it cannot be identified with any contemporary form of spoken Japanese – even if the texts would presumably have been read out in a way colored by one’s local sound system that might then be reflected in the han’gŭl sound glosses. The situation is different for the colloquial dialogues in vols. I to IX of the CS series. Some traits of Kyushu dialectal influence such as the appearance of lexical items like idoru ‘to sit’ notwithstanding, the language of CS1 is considered to be essentially that of the Kyoto region (Morita 1957: 40–41). By and large vols. I to IX, especially in the much more polished texts of CS2 and CS3b, represent a formal spoken register as used in the context of diplomatic intercourse, so that the paucity of indicators of a local origin is well in line with Nomura’s (2013: 64–69) observation that the Kyoto–Osaka area and Edo differed only little in terms of language if we compare formal, polite rather than less careful informal speech from these places. In the same monograph treating the early history of the Kyoto-based standard language, Nomura (2013: 69–74) also explicitly refers to CS3b and the language it reflects, commenting on its value as a source on the formal language of its time as used in an official context. As we have noted elsewhere (Osterkamp 2016b: 76–77), Nomura’s observations likewise apply to CS2, the 1790 Sayŏgwŏn edition of Inŏ taebang 隣語大方 (The Expert in the Neighboring Language) and various other bilingual Japanese–Korean sources, which are not treated here owing to the absence of sound glosses in them. What is unclear, however, is to what degree the Kyoto-based, but super-regional variety of formal spoken Japanese reflected in these and other sources was subject to regional variation under dialectal influence, especially on the level of phonology and phonetics. That such influence was present on the level of the lexicon can not only be gleaned from the CS series, but even more so from the dictionaries among the above-mentioned sources such as WY (see e.g. Doi, Hamada, and Yasuda 1959: 37–38 on some items associated with Kyushu dialects in WY). 2) Travelogues written by members of Korean embassies to Japan (many of which, but by far not all, are found in the collection Haehaeng ch’ongjae 海行摠載, or General Collection of Voyages Across the Sea) as well as other accounts or notices of Japan in works of Korean provenance, chiefly comprising transcriptions of proper nouns; major sources (giving the start date only in case of embassies; “C[hinese characters]” and “H[an’gŭl]” indicate the script used for transcriptions):
148
1420 1471 1596 1655 1655 1656 1719 1763 1763
Sven Osterkamp
Nosongdang Ilbon haengnok 老松堂日本行錄 (Nosongdang’s [i.e. Song Hŭigyŏng 宋希璟] Record of a Journey to Japan) (C) Haedong cheguk-ki 海東諸國紀 (An Account of the Countries East of the Sea) (C) Ilbon wanghwan ilgi 日本往還日記 (Daily Records from a Journey to Japan and Back) (C) Mun’gyŏn pyŏllok 聞見別錄 (Supplementary Records of Observations) (C) Pusang-nok 扶桑錄 (Records of Fusang [i.e. Japan]) (H [← C]) Kanyang-nok 看羊錄 (Records of a Shepherd) (H) Pusang-nok 扶桑錄 (Records of Fusang [i.e. Japan]) (H) Haesa ilgi 海槎日記 (Daily Records from a Raft on the Sea) (C+H) Kyemi susa-rok 癸未隨槎錄 (Records from a Raft Voyage in a guiwei Year [i.e. 1763]) (H)
With few exceptions these sources mainly provide us with a limited number of place names each, usually ones that are part of the embassies’ route through Japan. Some also contain a number of personal names for instance (thus also several entries in the Chosŏn period sillok, or Veritable Records), but the exact underlying forms are hard to identify at times, thus leaving us without certain comparanda. Be that is it may, the above and further similar materials may serve as a welcome corrective to those listed under 1 above – especially since the transcriptions do not usually show any sign of an intermediary stage in Japanese writing and were carried out by authors who were not necessarily proficient in Japanese.5 Care has to be taken however to identify han’gŭl transcriptions that are in fact merely re-transcriptions of pre-existing renderings in Chinese characters as such, as the latter for instance appear somewhat off at times due to limitations in the syllable inventory of Sino-Korean.6 3) Miscellaneous other renderings of Japanese in han’gŭl preserved in Japanese sources, often in form of the iroha poem, and often stemming from encounters on occasion of an embassy; major sources: 1636 Chōsen iroha 朝鮮イロハ (Iroha in Korean [Script]; ms.)7 1719 Kyakkan saisan-shū 客館璀粲集 (A Collection of Splendor from the Lodgings of the [Korean] Visitors) 1829+ Chōsenjin kenbun-sho 朝鮮人見聞書 (Record of Observations on the Koreans; ms.) 7
5 See Chŏng (2011: 70, note 7) for the distribution of the authors of 41 travelogues among various ranks, ranging from ambassadors to military officers. 6 The han’gŭl transcriptions in Pusang-nok are such re-renderings, based on the sinographic transcriptions in Mun’gyŏn pyŏllok (both 1655), as already pointed out by Yi (1984: 110–113). 7 Preserved in a manuscript in the possession of Kyoto University Library entitled Chōsen tsūshin-kō 朝鮮通信考 (shelf-mark 5–13|チ|15). The iroha was written by Mun Hongjŏk 文弘績 on occasion of the
6 Late Middle Japanese phonology, based on Korean sources
149
In the following we will survey what these Korean sources tell us about Japanese phonology and phonetics, but also what we cannot deduce from them with certainty. Interested readers are recommended to consult the comprehensive bibliography by Endō et al. (2009), as only small portions of the rich literature available on the topic can be referred to here.
3 Issues concerning consonants 3.1 Prenasalization of /b, d, g, z/ Similar to other sources on Japanese making use of segmental scripts, the Korean sources feature transcriptions of the type – i.e. clusters of letters for nasals and oral obstruents – to render prenasalized voiced obstruents, e.g. for /b/ [mb] and so on. Generally, /b, d, g, z/ were transcribed in either of the following ways: 1. obstruent preceded by (usually homorganic) nasal a. heterosyllabic: , e.g. /d/ in (CS1 I/2v) for madu ‘first of all’ b. tautosyllabic: , e.g. /g/ in (CS1 I/2v) for gozaru ‘to be’ 2. rendered by a single obstruent a. unmistakably rendering a voiced obstruent (,
): e.g. /z/ and /b/ in (CS1 I/1v) for gozaraba ‘should . . . be’ b. ambiguous, indistinguishable from transcriptions of Japanese /k, s, t/ (, , , ): e.g. /z/ in (CS1 I/3r) for fazimete ‘for the first time’; /d/ in (CS2 I/5v) for hodo ‘extent’ or in (CS2 I/3v) for madu ‘first of all’ Korean transcriptions of types 1a and 1b thus provide us with further evidence for the existence of prenasalization accompanying these obstruents in Japanese, but in fact this interpretation of -type transcriptions is not undisputed. The previously prevalent view, popularized by Ogura (1923) and Ifa (1932) for Japanese and Ryukyuan respectively, holds that is either a mere orthographic device to render a plain voiced obstruent or alternatively the nasal is thought of as a “preparation” for the following voiced obstruent. In this scenario such transcriptions are thus no sign of prenasalization. Even though this is now largely superseded by an interpretation as reflecting prenasalization (see already Hamada 1952, Ōtomo 1956, etc.), not few scholars have adopted the older interpretation (e.g. Lange 1969: 49; Ledyard 1966: 420, note 66; Thorpe 1983: 283, 294; Hagers 1997: 36). We will therefore review the facts in the following, giving old as well as new arguments against the earlier view. 1636 embassy, apparently for Hayashi Razan. See Osterkamp (2016a: 67–74) for details on this source as well as an overview over the known han’gŭl renderings of the iroha poem.
150
Sven Osterkamp
One problem the older view faces, but which has never been explained in a satisfactory manner, is why -type transcriptions for plain voiced obstruents are limited to sources on Japonic languages. Korean sources on contemporary Chinese throughout the Chosŏn period show a tripartite distinction: [-v, -a] vs. [-v, +a] vs. [+v, -a], written as , and respectively. Not a single case of for voiced initials in Chinese is known. What is more, even sources on Japanese eventually abandon (and at times adopt dakuten instead) as a means to render /b, d, g, z/ in the 19th century. This appears to be intimately related to the loss of prenasalization in the underlying variety of Japanese, so that in the original WY corresponds to in the 1837 manuscript of Hwaŏ yuhae. Korean renderings of Japanese ever since have likewise never made use of to render plain voiced obstruents. In more basic terms one would also have to argue why the transcribers should have felt the necessity to mark plain voicedness (without prenasalization) at all. In Korean /c, k, p, t/ are generally voiced between sonorants, so that no need arises to indicate voicing here. On the contrary, it is rather voiceless obstruents in intervocalic position that should be marked instead. Also, is even inserted before , which is voiced by default, so that cannot possibly indicate voicing here. If one furthermore sees as a mere orthographic device, one would have to explain why the transcribers took the trouble to use not simply a single arbitrarily chosen letter, but in fact , , and , all in accordance with the following consonant’s place of articulation. Probably the most significant fact speaking in disfavor of the older view, and one which finally leads us back to the actual data, is the distribution of -type transcriptions. If was a mere orthographic device, why was it not applied in a mechanical fashion to all instances of /b, d, g, z/ alike? The picture we get under closer scrutiny is not mechanical at all and in fact quite unbalanced – namely in a way that resembles the distribution of prenasalization both in other contemporary sources (such as Diego Collado’s Modus confitendi et examinandi, printed in 1632) and in dialects. While modern Tōhoku dialects are reported to lose pre-nasalization by and large in the order /b/ → /z/ → /d/ → /g/ (Ōhashi 2002: Chapter 2.3.1), the order than can be gleaned from philological records (not limited to Korean sources and usually reflecting Western varieties of Japanese) is rather /z/ → /b/ → /d/ → /g/, with /b/ and /z/ inversed (cf. already Ōtomo 1956: 112; Wenck 1959: 234). Consider the choice between and simple to render Japanese /b, d, g, z/ in the CS textbooks, as shown in Figure 1.8
8 Data adapted from Chin (2003: 60, 65, 68, 72). Note that the ratio for /d/ in CS2 would be higher if cases of being misprinted as were included as well.
6 Late Middle Japanese phonology, based on Korean sources
151
g: CS1 CS2
CS3 d: CS1 CS2 CS3
b: CS1
CS2 CS3 z: CS1 CS2
CS3 0%
20%
40%
60%
80%
100%
Figure 1: Distribution of and for Japanese /b, d, g, z/ in the CS series.
The rather uneven distribution of is apparent here. While prenasalization of /g/ remains stable diachronically, it decreases significantly from the 17th to the 18th century in the case of /d/. Both /b/ and especially /z/ already start out with a relatively low percentage in CS1, which is further reduced afterwards, with prenasalization of /z/ being lost entirely in the 18th century. The travelogues tell us about the same: Neither Pusang-nok (1719) nor Kyemi susa-rok (1763) indicates prenasalization for /b, d, z/, but still commonly – and in the case of Kyemi even throughout – for /g/. On the other hand, Mun’gyŏn pyŏllok (1655) still reflects more or less consistent prenasalization of /d/ besides /g/, while it is absent for /b, z/ (note however there is but a single case for /b/ here). It is exclusively the earliest sources, such as Haedong cheguk-ki (1471), that also regularly have cases for prenasalized /z/ and /b/.
3.2 Affricatization of /t, d/ and loss of the yotsugana distinction The affricatization of /t, d/ before /i, y, u/ can be dated with relatively great certainty with the help of both Korean and Chinese transcriptions. Parallel to the earlier Helin yulu 鶴林玉露 (Jade Dew in the Crane Forest; 1252) and Shushi huiyao 書史會要 (Collected Essentials on the History of Calligraphy; 1376), Korean sources up to ca. the year 1500 – such as Haedong cheguk-ki and Irop’a – indicate a plosive realization for /t, d/ in all environments. This is in contrast to the later ones dating from the 17th to 18th centuries, which now show affricates throughout before /i, y, u/. The gap between the earlier and later group can be filled by Chinese sources from the 16th century, which suggest that while the first half of the century was a period of transition with both
152
Sven Osterkamp
plosive and affricate variants, the latter had become the norm from the second half onwards (Ōtomo 1963: 698). In general, it is the affricatization of /d/ that has received most attention. However, there is also a notable phenomenon with regard to /t/, namely the confounding of post-affricatization /tu/ and fricative-initial /su/ or vice-versa. The case of aisatu ‘greeting’ being written as 相指 in CS1 (III/27v), thereby rather suggesting aisasu, is wellknown for instance. Such examples are however also found elsewhere in LMJ, including the missionary sources and shōmono (see e.g. Morita 1976: 303–305). Now the loss of the so-called yotsugana distinction is closely related to the two aforementioned sound changes: On the one hand affricatization of /d/ and prenasalization of both /d/ and /z/ – under the influence of which /z/ would have developed affricate variants – rendered the oppositions /di, zi/ and /du, zu/ unstable. On the other hand, however, /z/ started to lose its pre-nasalization considerably earlier than /d/, so that this difference in turn helped to sustain the opposition. From such a perspective we may expect CS1 to observe the distinctions rather well, by and large, as vs. for /d/ vs. /z/. In fact, however, cases of confusion do not increase over time, so that the distinction is predominantly kept even in CS3b, both in kana and in the transcriptions. WY on the other hand predominantly uses for either (see Pak 2005b for a detailed treatment). As suggested by Pak, the latter fact may have to do with the palatalization of /c, ch, cc/ (thus merging with original /cy, chy, ccy/) in late 18th century Korean, which deprived the language of a good match for Japanese /du/.9 The situation in CS2/3b is therefore likely the result of the fact that any new edition of CS was under the influence of previous ones, even if the text was revised considerably. Additionally, teaching materials often tend to conservativeness, so that the older spellings may simply have been retained.
3.3 Delabialization of /f/ The progressing delabialization of [ɸ] to [h] (~ [ç]) can clearly be observed in Korean sources (see Table 1), however they grant us only limited insight depending on the following vowel. Thus, in case of following /o, u/ the transcriptions stay the same throughout the centuries, generally yielding . The rounded vowel in Korean may have given a labial coloring to the preceding fricative /h/, but in the end we are unable to distinguish [ɸ] from [h] here. Only Irop’a (1492) also has straightforward for [ɸ] here,10 while Kyakkan saisan-shū (1719) writes for /ho/, suggesting [h]. These two may be taken as representatives of their time: as a rule, 15th century sources indicate [ɸ], whereas 18th century ones have [h] as the default. However, [ɸ] survived considerably longer before /a/ than before the front vowels /i, e/. 9 This merger is also evident in the manuscript notes added around the 1780s to the Vatican copy of Iryŏp’a (on which see Osterkamp 2016a: 65–66). 10 ㅸ and ㆄ are transcribed as (rather than Yale ) vs. here.
6 Late Middle Japanese phonology, based on Korean sources
153
Table 1: Korean transcriptions of Japanese /f/ through time.11
-a -i -u -e -o
Irop’a
“Chōsen iroha”
CS1
Kyakkan saisan-shū
CS2
CS3
WY
, , ,
,
,
, 11
,
For /a/, [ɸa] was usually indicated by (sinographic transcriptions also have , e.g. 波, besides ), until eventually getting replaced during the course of the 18th century by for [ha]. The choice between the two in the transitory stage was largely lexically determined (Yasuda 1960: 18–19). The newer [ha] is also what underlies the sporadic transcriptions with zero onset in Haesa ilgi (1763), which otherwise has [h] throughout:12 Haidomari 南風泊 as 我伊刀馬里 and Hamamatsu 浜松 as 阿馬馬즈 (sic). The front vowels /e, i/ require our greatest attention. In the teaching materials up to CS1 and are used for [ɸ], but only in the 18th century. The hitherto unnoticed “Chōsen iroha” dating from 1636, however, already has here, despite in CS1 published several decades later. Also, as Chin (2003: 92–96) has pointed out, the place names in Kanyang-nok (1656) likewise write in case of following /i/, so that [h] must already have been prevalent in at least some variety of Japanese.13 The same kind of contradiction between CS1 and other roughly contemporary sources will be addressed further below with reference to the centralization of /u/ after sibilants.
3.4 /-t/ as coda Both the Jesuit sources and cases of renjō demonstrate that /t/ had once been allowed as a coda consonant, with no need for an immediately following /t/ in onset position. Relevant transcriptional materials of later times are rare in number, but the CS series grants us some glimpses at this phenomenon. As mentioned above, the textbooks are
11 The isolated case of in CS2 (Xb/4v) is probably an erroneous retention from CS1 (X/12r). 12 Chin (2003: 94–95) interprets the phonogram 屎 in Haesa ilgi (1763) as /si/, thereby giving several cases of confusion between /hi/ [çi] and /si/ [ɕi]. However, 屎 never writes etymological /si/ here, and contemporary dictionaries such as Samun sŏnghwi 三韻聲彙 (1745) indicate /hi/ as one of its readings. Also, the han’gŭl letter is in fact even explained by this character in Hunmong chahoe 訓蒙字會 (1527). We therefore consider it to be a phonogram for Japanese /hi/ here. 13 According to Chin (2003: 94) there is also an isolated case of for /hi/, but the – as often in this work, badly disfigured – han’gŭl spelling for Hida 飛騨 rather reads 힌다 than 희다.
154
Sven Osterkamp
of a dual nature, featuring both sections in spoken and formal written language. Now /t/ as a coda is virtually limited to a handful of Sino-Japanese words in the latter, i.e. the correspondence in sōrōbun in volume X (cf. already Morita 1957: 34). The relevant cases are given in Table 2. Table 2: Words with coda /-t/ in the CS series.
Volumes I to IX generally have /-tu/ in place of /-t/ in Sino-Japanese words, suggesting that /-t/ belonged to a highly formal register while being virtually absent from commonplace speech. What is apparently the only exception in the dialogues may be taken as a confirmation of this: kiɴzit 近日 ‘in a few days’ in CS3b (VII/13v) is found in a passage set in Edo, more precisely in a sentence transmitting the words of the shōgun to the envoys.
4 Issues concerning vowels 4.1 “Strongly palatalized” /e/ Korean sources have been adduced repeatedly as alleged evidence for a strong palatal nature of Japanese /e/. For instance, Lange (1969, 1971, 1973) provides three pieces of evidence to show that “all /e/ syllables were strongly palatalized” (Lange 1969: 47): 1) the palatal realization of earlier /se, ze, (y)e/, as indicated in missionary sources, 2) the change from /eu/ to /yoo/, and 3) Korean transcriptions in Irop’a of the type , indicating a palatal glide for all /Ce/ syllables. In his view, /e/ ceased to be strongly palatalized during the 16th century, thus explaining the divergence between Irop’a and the missionary sources of ca. 1500 and ca. 1600 respectively. In Lange (1971) CS1 and CS3b are taken into account as well, but the fact that the same kind of transcriptions as in Irop’a are still found in the 17th and 18th centuries is in the end simply explained by attributing this to Kyushu dialect influence. While the facts for 1) are straightforward, the three simply tell us nothing definite about the remainder of /Ce/-type syllables. Concerning 2) suffice it to say that it has long been pointed out that the assumption of a palatal /e/ is unnecessary to explain the
6 Late Middle Japanese phonology, based on Korean sources
155
change, so that the latter is no evidence of the former.14 In the end, Lange’s claims thus stand and fall with the Korean sources – and as a closer look at these demonstrates, they tell us even less than the missionary sources in this respect, as we cannot even distinguish [je, ɕe, (d)ʑe] from [e, se, (d)ze] here. The problems start with the fact that Lange’s argument is concerned with the central language. Lange (1973: 36, note 28) borrows the authority of Hamada, who allegedly identifies the Japanese of Irop’a as the Kyoto dialect, but no such identification can be found in his writings. Instead Hamada (1965: 21) states that the identity of both the variety of Japanese and Korean reflected in Irop’a is altogether unknown. It is, therefore, problematic to treat Irop’a as a Kyoto source, but CS1 and CS3b as Kyushu sources without proper justification. The issue is further complicated by the fact that the same kind of transcription for /Ce/ syllables is likewise found in what are clearly non-Kyushu sources (1636 “Chōsen iroha”, 1719 Kyakkan saisan-shū etc.). The choice for to render Japanese /Ce/ was therefore not motivated by the phonetics of one specific dialect area, or a specific time. Instead we first have to attain a fairly reliable idea as to what /C(y)e(y)/ actually sounded like prior to the setting-in of monophthongizations in Korean.15 Consider for instance Dutch transcriptions reflecting mid-17th century Korean as summarized in Table 3.16 Table 3: 17th century Dutch transcriptions of Korean /C(y)e(y)/.17 Korean
Dutch transcriptions
/Ce/ /Cey/ /Cye/
o, oo uy, eüye17 jo, io; ja, ia, iae
/Cyey/
je, ie jee
This picture can be corroborated by pre-modern Japanese renderings of Korean (see e.g. the statistics for one of the most substantial sources provided in Hŏ 2014: 74–101): Korean /Ce(y)/ usually yields and less often in kana transcriptions. /Cye/ is already often written but also , in part depending on the position in a word. For 14 See e.g. Mathias (1974: 74) or Takayama (1992: 32). Also see Wenck (1959: 157), who proposes the following phonetic interpretation of the process: [eṷ] → [øṷ] → [i̭ oṷ] → [i̭ oː]. 15 The exact dating for monophthongization of /Cey/ etc. is much disputed and generally difficult to establish. It appears to have occurred sporadically in the 18th century, but was probably not completed on a larger scale until well into the 19th century. 16 Based on Witsen (1705: 52–53), with the exception of “jee” which is based on Hendrik Hamel’s account of Korea (Hoetink 1920: 8). 17 Cf. [œy], a reasonable match for [əj] in Dutch, also variously written , and in 17th century Dutch orthography.
156
Sven Osterkamp
/Cyey/ we finally get as the default. These data suggest that simple /e/ and also /ey/ were realized as [ə] and [əj] in Korean and thus constituted no match for anything close to [e], contra Lange.18 While /ye/ came considerably closer, the only stable correspondence for foreign [e] was, however, Korean /yey/. Being sandwiched by palatal glides, /e/ [ə] was evidently fronted by default here, yielding something close to [jej] or [je] rather than [jəj]. Put differently: As long as Korean did not have [e] as a monophthong, this /yey/ was the best match available for anything foreign resembling [e], be it [e] as such or [je] for instance. Both palatal glides in Korean /Cyey/ were needed to make sure that /e/ was fronted – and therefore the presence of the first glide in the transcriptions tells us nothing certain about Japanese phonetics. Whether a transcription such as reflects Japanese [ke] or [kje] is impossible to decide on the basis of Korean sources alone, as either could be the case. Indisputable evidence to show that “all /e/ syllables were strongly palatalized” is something these sources simply cannot provide us with.19 Finally, there is the issue of variation between versus observed in parts of the corpus. Contrary to Unger’s (2009: 79) claim, the two are far from being used interchangeably. Rather is the default, while is mostly restricted to closed syllables and syllables with nasal initial (cf. already S. Cho 1970, II: 38; Sugito 1989, etc.). In the light of this distribution and the way the two are transcribed in foreign sources on Korean, we propose that aims at close-mid [e], whereas reflects a less close (and possibly less front) vowel such as open-mid [ɛ] as a phonetic variant in closed syllables and nasal(ized) environments.20
4.2 Centralization of /u/ after sibilants The basic correspondence for Japanese /u/ in the Korean transcriptions of the earlier sources up to CS1 is , but the 18th century sources starting with CS2 use instead if preceded by a sibilant.21 We may take this to reflect the centralization of
18 Lange (1969: 49) apparently considered Korean /e/ as a possible match for non-palatalized Japanese /e/, i.e. [e]: “By consistently using 여 ([jö]) rather than 어 ([ö]) the compilers of the book show that Japanese syllables containing /e/ were perceived as being palatalized.” Similarly, Unger (2009: 79–80) claims that “both /e/ and /ey/ were evidently suitable for transcribing the nuclear Middle Japanese vowel”, i.e. [e]. Apart from a few misprints in WY, orthographic was however never equated with the Japanese vowel in question (sporadically however with /o/, cf. e.g. Pusang-nok and Kyakkan saisan-shū, both 1719). 19 The same conclusion – however with reference to the Chinese source Ribenguan yiyu – has incidentally already been arrived at by Jiang (1997). 20 For a later parallel to what we propose for the Korean sources here see Hoffmann (1864) who follows Lepsius and his Standard Alphabet to distinguish between (“e close”) vs. (“e open”, both on p. [V]) in his transcriptions of Japanese. The majority of cases for are for /e/ in closed syllables, here restricted to the shape /Ceɴ/. 21 Sporadically is also used to transcribe the vowel in other syllables, especially /nu/ and /ru/ (cf. the iroha poem in Kyakkan saisan-shū, 1719, and in Chōsenjin kenbun-sho, 1829 or later). These sources are
6 Late Middle Japanese phonology, based on Korean sources
157
/u/ after sibilants, the result of which is still observable in modern standard Japanese and han’gŭl transcriptions thereof.22 Now Chin (2003: 48–49) and Pak (2005a: Section 4) have however already pointed out that centralization can be observed in travelogues dating from as early as the mid-17th century, long before the publication of CS2.23 This is now confirmed anew with the help of the 1636 “Chōsen iroha” (see Table 4). Table 4: Korean transcriptions of Japanese /u/ through time.
/su, zu; tu, du/ other /(C)u/
Irop’a
“Chōsen iroha”
, () , ()
CS1
Kyakkan saisan-shū
CS2
CS3b
WY
That the problem is rather on the side of CS1 than on that of the other named sources is suggested by the situation found in non-Korean sources. K. Cho (2001: 98) goes as far as considering the ability to reflect the split of /u/ into two allophones a distinguishing feature of the Korean sources, stating that no such thing can be observed in other foreign sources. Fortunately, however this is actually not true, as Chinese sources could – and did – make use of the contrasts between e.g. [su] and [sɨ] (or [sz̩]) with ease. The distribution of phonograms used to transcribe /su, tu/ in 16th century Chinese sources is indicated in Table 5.24
not necessarily reliable however when it comes to look-alike letters, as they often contain scribal errors due to an insufficient level of acquaintance with han’gŭl. 22 Most previous studies speak of “delabialization” in this context. The decisive feature distinguishing the allophones of /u/ within a given variety of Japanese is however centralness (i.e. here: central vs. back) rather than labiality, regardless of whether the default phonetic realization is rounded or not. 23 Travelogues featuring sinographic transcriptions have to be handled with care here due to a skewed distribution of /u/ in Sino-Korean. As su and cu did not occur in isolation, one either had to resort to characters such as cuk 即 for /tu/, with superfluous -k (as attested in Mun’gyŏn pyŏllok), or otherwise use swu and cwu instead – leaving us with no means to tell whether the transcriber had a centralized vowel in mind or not. A third possibility was to simply switch to han’gŭl for such problematic syllables (as attested in Haesa ilgi). 24 The statistics are taken from Ōtomo (1963: 226, 314–315, 380–381, 554–555). Note that some marginal, but rather problematic cases with unexpected correspondences that cannot safely be assigned to any of the categories in the table have been ignored here. Also, the modern Pinyin readings here and elsewhere are merely given out of convenience and are, strictly speaking, irrelevant.
158
Sven Osterkamp
Table 5: Chinese transcriptions of Japanese /su, tu/ in 16th century sources. source
/su/ [su ~ sɯ]
✶
/tu/ [sɯ̈]
✶
[tu ~ tɯ]
✶
[ʦu ~ ʦɯ]
✶
[ʦɯ̈]
✶
Ribenguo kaolüe 日本國考略 (1523)
su 宿 (4), shu 踈 (3), sun 孫 (2), sui 碎 (1), shu 數 (1), song 松 (1), su 速 (1) [zun 尊 (1), zu 卒 (1)]
si 思 (3)
duo 多 (3), du 都 (2), tu 禿 (2), duo 掇 (1), tuo 脱 (1)
zu 卒 (2), zu 祖 (1), zuo 做 (1)
zi 子 (13), zhi 止 (1)
Ribenguan yiyu 日本館譯語 (1549)
sun 孫 (3), suo 唆 (2), suo 索 (2), su 宿 (1)
si 司 (20), si 寺 (1)
du 都 (13), du 讀 (4)
zu 足 (3), zu 祖 (1)
[si 司 (1)]
Riben tuzuan 日本圖纂 (1561)
suo 索 (1)
si 思 (2), si 四 (1), si 斯 (1), si 司 (1) [zi 子 (2)]
cuo 撮 (1)
zi 子 (10), zhi 治 (1), zhi 知 (1)
Riben fengtuji 日本風土記 (1592)
su 宿 (10), sun 孫 (1), suo 索 (1)
si 四 (16), 系 [= si 絲] (11), si 私 (6), shi 使 (1), shi 示 (1)
zu 卒 (1), cun 寸 (1), chu 初 (1)
zi 子 (46), zi 紫 (6), zhi 止 (3), ci 次 (1), zhi 之 (1), zeng 増 (1), zeng 繒 (1)
The preference for phonograms suggesting centralized vowels (✶[sɯ̈] and ✶[ʦɯ̈]) is overwhelming towards the end of the 16th century. Accordingly, it is not odd at all to find reflexes of centralization in Korean sources dating from the following century. What is needed however is a re-evaluation of the sound glosses in CS1, which appears to reflect a somewhat older stage of Japanese in terms of the delabialization of /f/ and centralization of /u/ after sibilants – or possibly simply another variety of the language.25
4.3 Predecessors of modern /oo/ Modern /oo/ chiefly comes from three different LMJ sources, each with several earlier sources in turn. Since Hashimoto (1928, I/C: 46–51) the prevalent view, based on missionary sources dating from around 1600, has been that the two chief ones of these
25 Note that Kang Usŏng is generally considered to have compiled CS1 during the years 1618–1636, but it is unknown whether the sound glosses in the 1676 print are his as well or only added later on.
6 Late Middle Japanese phonology, based on Korean sources
159
were pronounced as [ɔː] (← /au/; usually referred to as kaion and spelled by the Jesuits) and [oː] (← /ou/; called gōon and spelled ) respectively in LMJ, whereas the third was still disyllabic [owo] (spelled ). While this view is still found in recent studies into the historical phonology of Japanese (see e.g. Frellesvig 2010: 320 or Irwin and Narrog 2012), it has long been challenged on good grounds. In what is likely the most important study in this respect, Toyoshima (1984) has demonstrated that Hashimoto’s argumentation is flawed in several ways and therefore not tenable. A closer reading of the relevant passages of the missionaries’ grammars as well as contemporary works on phonology and their use of the terminological pair of kai versus gō leads him to a different solution with /oo/ [oː], or possibly [ɔː], as opposed to /ou/ [ow]. Sakanashi (1987: 29–30) and Ikegami (1993, I: 287–292) add to and support this view, as well, and similar ideas had been expressed even earlier, with Kawakami (1980: 6) quoting Wada Toshimasa’s assumption that might have been [ou] for instance. Neither of these authors mention any of the Korean sources, yet it is exactly this corpus which provides support for the above views. The sources from the 15th century exhibit considerable variety and in part clearly still reflect [aw], for instance, for later , indicating that they belong to a period of transition. The CS series on the other hand consistently writes , both for and .26 Some scholars have deemed such transcriptions to be motivated by kana orthography, rather than rendering something along the lines of [ow] (cf. e.g. Morita 1957: 30). There are however various reasons to reject this possibility: First of all, this would be the only case of systematic spelling pronunciations in the entire corpus. Thus, for instance, orthographic and in kana is never glossed as and but always as and , (i.e. ) in kana is never transcribed as disyllabic , and so on. Also, an equivalent of (compressed into a single syllable block orthographically: ᄆ ᆃ) is employed even in the case of moos- ‘to say (humble)’, one of the few morphemes written logographically but never in kana in the main text. Second, we find the following interesting testimony as to the pronunciation of both and in Sino-Japanese character readings in one of the travelogues, namely Sin Yuhan’s 申維翰 Haeyu-rok 海遊錄 (Records of a Voyage Across the Sea; 1719):
26 We cannot rule out the possibility however that this orthographical merger does not reflect a phonological merger in the underlying variety of Japanese, but rather the inability of the phonological system of Korean to capture a distinction that was still active. After all, the kana orthography in CS – which is generally oriented towards pronunciation and is rarely historically or etymologically motivated – does retain a distinction, and cases like koo ← kau vs. kwoo ← kwau merging into transcriptional show that such limitations in the Korean transcriptions are not inconceivable.
160
Sven Osterkamp
日本讀字之音、如東冬陽庚靑蒸之韻、則呼以二音。東字曰都于、陽字曰要于、靑字曰世伊、江字曰 乂伊。眞文元先寒刪27等韻與我國略同。[. . .]其他蕭豪韻及入聲亦用二音、而時或與我國彷彿。 [In the pronunciations as one reads characters in Japan, rimes such as dong 東, dong 冬, yang 陽, geng 庚, qing 靑, zheng 蒸 are pronounced in two syllables. The character 東 is read as two-wu 都于, the character 陽 as ywo-wu 要于, the character 靑 as syey-i 世伊, the character 江 as yey-i 乂伊. Rimes such as zhen 眞, wen 文, yuan 元, xian 先, han 寒, shan 刪 etc. are roughly the same as in our country. [. . .] Apart from this, the rimes xiao 蕭 and hao 豪 as well as rusheng 入聲 ones also have two syllables; however, at times, they closely resemble [the pronunciation] in our country.]
While the last example given for rimes in -ng is somewhat out of place here as no Sino-Japanese reading is involved, both (陽, ← yau) and (東, ← tou) are clearly given as something equivalent to [ou] or [ow] here, consisting of two distinct segments (or even syllables, if taken literally), just as in the textbooks and other sources. Obviously, this was considered as characteristic of Sino-Japanese in contrast to Sino-Korean, which like its donor retains monosyllabic forms throughout (just as both do for rimes in -n, as pointed out in the quote). Virtually the same stance on Sino-Japanese character readings is also found in the Unhae 韻解 (Explanation of Rimes; 1750) of Sin Kyŏngjun 申景濬 (1712–1781).28 Third, the transcriptions found in the various travelogues provide us with corroborative evidence that both and were at a time realized as [ow] in at least some varieties of Japanese. Unlike the CS series none of the travelogues provides the original Japanese spelling for any item transcribed, and the higher degree of irregularity likewise suggests that we are not dealing with mechanic transliterations of kana originals here. Therefore, it seems highly unlikely that we are facing spelling pronunciations here. More importantly, however, the travelogues also allow us to observe the decline of this diphthongal [ow], giving way to long [oː] during the course of the 18th century: In Pusang-nok (1719) there is variation between older and newer , whereas only the latter is found in Kyemi susa-rok (1763).29 Tsuzuku (1968) has long pointed out, based on evidence from both Japanese and foreign sources (including, but not limited to Korean ones), that up to the early 18th century the default realization of “long o” appears to have been diphthongal [ow] (Tsuzuku’s “ou”) rather than [oː]. In his scenario and would have merged into [ow] at first, becoming [oː] only in the 18th century. How can this be reconciled with the
27 The same passage in Ilbon sahaengsi ilgi 日本使行時日記 (1719; ms. in the possession of the Academy of Korean Studies, call no. “B15ID-10”) has shan 山 instead of shan 刪, both of which make sense here. 28 Space does not permit treating Sin Kyŏngjun’s work in any detail here, which likewise applies to several roughly contemporary writings by Yi Hyŏngsang 李衡祥 (1653–1733), especially his account of kana and their pronunciation as found in his Akhak p’yŏngo 樂學便考 (1712; contained in Pyŏngwa chŏnsŏ 甁窩全書 IX) and the somewhat later Chahak 字學. Both authors again strongly suggest that Korean transcriptions of the form do not originate in spelling pronunciations. 29 This pertains to cases deriving from /au/ as well as /ou/ ← /owo/; original /ou/ is not present in the data.
6 Late Middle Japanese phonology, based on Korean sources
161
Korean data? It is important to note here that besides the above-mentioned travelogues there are also cases of for original /ou/ and /au/ in WY. See among others the entries for imooto ‘younger sister’ (I/12v) or kozoo ‘young monk’ (I/53r) for the former, and kookoo ‘filial piety’ (I/22r) or oogi ‘fan’ (II/13r) for the latter. Now the crucial difference between the CS series and WY is that the latter is in han’gŭl only, whereas the former first provides the main text in kana. The same kind of situation also obtains for Korean sources on Manchu, for instance, where the presence or absence of original script leads to two quite distinct transcriptional systems. In WY the absence of kana spellings could thus have fostered the adoption of transcriptions more oriented towards the then-current, newer pronunciation – whereas CS might easily have been conservative in that newer editions were under the influence of older ones, so that spellings such as may have been retained throughout the transitional period from [ow] to [oː].
4.4 Vowel devoicing While transcriptions of Edo period Japanese in Latin and Cyrillic script are replete with evidence of vowel devoicing, the issue is rarely addressed at all with reference to Korean sources. This is because, as we will argue, devoicing is reflected only indirectly, namely as phonetic lengthening of the plosives /k, t/ (including the latter’s affricate allophone [ʨ]) if adjacent to a devoiced vowel, which is then rendered by orthographical geminates in han’gŭl: and or .30
30 Both the distribution and the likely reasons are different for and , so they are better treated separately. is virtually confined to the particle =koso in CS1, which may reflect an emphatic variant =kosso, as proposed by Fukushima (1977: 313–314). In CS2/3v cases of =koso decrease, but numerous cases for actual geminates in polite -mass- and honorific -(s)assyar- are added. In passing we may point out that the geminates in Manoel Barreto’s 1591 manuscript (Biblioteca Apostolica Vaticana, Reg.lat.459) – which has been compared with Korean sources in this respect a number of times (see e.g. Fukushima 1975 or Morita 1976: 335–338) – show only minimal overlap with the cases of here, and their distribution among the possible consonants is entirely different. In Barreto’s case about two thirds of all cases occur with Japanese /s/, which is better explained as an interference of Portuguese orthography, which has for singleton [s]. is limited to attributive =no in CS1 and only minimally attested in CS2 (and not at all in CS3b). We follow Wenck (1959: 274; 285, note 59) here, who identified this with the same phenomenon of nasal insertion or lengthening in Japanese sources and to which we may add further parallels from other foreign sources: For instance, Riben fengtuji (1592) has [a-jin-na-she-zi] 阿金那設子 to render akiɴ=no setu ‘autumn time’ (IV/3v; versus [a-qi] 阿氣 for aki in isolation, IV/3r) or [hui-tian-na-ge-re] 回天那革熱 for woiteɴ=no kaze ‘tailwind’ (IV/2v) among others. Ávila Girón’s Relación del reino de Nippón (1615) has “christan von aratamen notaqui” (see Sakuma/Okada 1965: 346), apparently transcribing Kirisitaɴ woɴ. aratameɴ=no toki ‘the time of the investigation against the Christians’.
162
Sven Osterkamp
One undisputed function of is to transcribe long consonants in Japanese, even if they are not reflected in the kana orthography.31 The many remaining cases of have posed one of the most intriguing and long-standing problems in the interpretation of the Korean transcriptions, with about as many proposed functions as scholars working on the issue. Thus, they have been claimed to: inhibit voiced readings of intervocalic consonants (Hamada 1955; Morita 1957 etc.);32 render glottalization allegedly present in Japanese voiceless obstruents as in Korean /CC/ (Ōtomo 1957; Araki 1975 etc.); mark word accent (Morita 1957; Nakayama 2000 etc.) or sentence accent (K. Cho 2005 etc.), etc. None of these proposals can explain all the data, or at least the majority of it, and is at the same time in agreement with our knowledge of Japanese from other sources. Vowel devoicing on the other hand is well attested and its distribution in other transcriptional sources closely resembles that of the remaining . A considerable portion of these occur in word-final syllables, commonly for instance with Sino-Japanese words ending in /-ku/ or /-tu/, e.g. aisatu̥ ‘greeting’ as (CS1 I/3v) or fusoku̥ ‘insufficiency’ as (CS1 IV/16v). Similarly there are also cases with native words ending in these syllables, such as fitotu̥ ‘one’ as (CS1 I/18r), but also e.g. kataku̥ ‘firmly’ as (CS1 V/8v) or ugoku̥ ‘to move’ as (WY I/29r). This position not only coincides with Collado’s (1632: 5) examples of devoicing, it is also commonly observed in other sources spanning several centuries. To name but a few examples: “Rocuguatz” for rokugwatu̥ ‘sixth month’ in 1615 (Sakuma/Okada 1965: 94); “mitz”, “jotz” etc. for mitu̥ ‘three’ and yotu̥ ‘four’ (Meister 1692: plate to p. 310); “tazits”, “Sipokf” for tazitu̥ ‘another day’ and sippoku̥ ‘table(cloth)’ (Overmeer Fisscher 1833: 102). In fact, the Korean sources themselves also contain evident references to such realizations, at least for Sino-Japanese words. In the late 18th century Iryŏp’a (3r), かく, やく, とく are transcribed as , , respectively, and this is actually well in line with the quote from Sin’s Haeyu-rok given above, according to which rusheng rimes “closely resemble [the pronunciation] in our country” at times – i.e. sometimes they were realized or at least perceived as single, closed syllables. Likewise of interest in this context are the numerous handwritten corrections in red found in the copy of WY kept at the National Library of Korea, which among other things turn all Sino-Japanese reflexes of /k/-finals into closed syllables. Thus, for instance, 蝕: → for syoku̥ (I/1r), 北: → for hoku̥ (I/1v) etc. (Reflexes of /t/-finals are already written as throughout in the print.)
31 In CS1, for instance, this applies to ko(t)ti/so(t)ti/a(t)ti ‘here/there/over there’, asa(t)te ‘day after tomorrow’, =ba(k)kari ‘only’, etc. In kana we get the conservative form without gemination, while the sound glosses with indicate the geminated form (e.g. あさて vs. in II/14v). 32 While this is not tenable for the situation in the CS series or WY in its entirety (the only exception being voiceless /C/ after moraic nasals), the transcriptions in Kyemi susa-rok (1763) appear to employ (almost exclusively ) in exactly this function.
6 Late Middle Japanese phonology, based on Korean sources
163
Another large group of comprises word-internal cases adjacent to vowels that were likely devoiced judging from other foreign transcriptions and also modern varieties of Japanese. This includes e.g. /s/-stem verbs followed by perfective -ta, where devoicing also occurs in modern standard Japanese, but note that this is not restricted to high vowels surrounded by voiceless consonants, but also extends to other environments. Consider the following examples: -masi̥ ta ‘polite perfective’ as (CS2 I/8r) cf. “-masta”, “-maschta” (Meister 1692: 187, 189), “-masta” (Thunberg 1792: 269), etc. wataku̥si ‘I’ as (CS1 I/25v) cf. “VVataxi” (Meister 1692: 189; with for [ks]), “Wataks” (Titsingh 1781: 256), “Wataks’” (Siebold 1826: 113), “Watakfs” (Overmeer Fisscher 1833: 100), etc. katazi̥ kenoo ‘gratefully’ as (CS1 I/22v) cf. “Cataschky nucky serrimesch“ (Meister 1692: 193), “kadeski, no guserrimas” (Thunberg 1792: 271; both plus gozarimasu), etc. hadu̥kasi[i] ‘ashamed’ as (WY I/21v) cf. “Fanscashy” (Meister 1692: 196), “Fatskasfi” (Titsingh 1781: 250; for “f” read “ſ” = “s”), “fatskassiku” (Thunberg 1791: 337), etc.
There are further cases of not belonging to any of these groups (some forming new groups, such as words for the number of days in -ka or also adjectival nouns ending in -ka), not all of which can be explained along the same lines. The majority of cases, however, is covered if we assume that the Korean transcribers chose for the longer duration of Korean /CC/ as opposed to /C/, since consonants adjacent to devoiced vowels had undergone compensatory lengthening phonetically. Incidentally such a lengthening effect was not only noticed by Korean transcribers, but we also have for instance Siebold’s (1826: 92) description: Characteres in U plerumque in verborum infinitivo sonum U amittunt, consonante antecedente quasi duplicata; ex. gr. Idoru-Idorr’ (sedere) Maku-Makf’ (volvere) (1) Nosomu-Nosomm’ (sperare); Tobu-Tobb’ (volare). [Most of the time characters [i.e. kana syllabograms; S.O.] in -u lose the sound u in the infinitive of verbs, the preceding consonant being as if doubled. For instance: Idoru-Idorr’ (to sit), Maku-Makf’ (to roll), Nosomu-Nosomm’ (to hope), Tobu-Tobb’ (to fly).]
Along these lines we may explain both why are found in places were Japanese is expected to have phonologically short consonants and why the CS series as well as other sources appear to be devoid of traces of vowel devoicing at first sight. More importantly this case reminds us of the importance of always considering the Korean sources in conjunction and comparison with other sources, both domestic and further foreign ones.
164
Sven Osterkamp
References Araki, Masami. 1975. Shōkai shingo no heishohō ni tsuite [On the geminate spellings in the Shōkai shingo]. Kokugo Kenkyū 38. 59–72. Chin, Namt’aek. 2003. Chōsen shiryō ni yoru Nihongo to Kankokugo no on’inshi kenkyū [A study of the phonological history of Japanese and Korean based on Korean sources]. Tokyo: University of Tokyo dissertation. Cho, Kanghŭi. 2001. Chōsen shiryō ni yoru Nihongo onsei, on’in no kenkyū [A study of Japanese phonetics and phonology based on Korean sources]. Seoul: J & C. Cho, Kanghŭi. 2005. Chosŏn charyo-ŭi han’gŭl ŭmju-e nat’ananŭn kyŏngŭm p’yogi-wa mun aksent’ŭ-wa-ŭi kwan’gye-e taehayŏ [On the relationship between the geminate spellings as seen in the sound glosses in Korean sources and sentence accent]. Ilbon Ŏmunhak 24. 115–140. Cho, Seung-bog. 1970. A phonological study of Early Modern Japanese on the basis of the Korean sourcematerials. 2 vols. (Stockholm Oriental Studies 8–9). Stockholm: Almqvist & Wiksell. Chŏng, Yŏngsil. 2011. 18-seiki shotō no Chōsen tsūshinshi to Nihon no chishikijin [Japanese intellectuals and the Korean diplomatic missions to Japan in the early 18th century]. In Ken’ichirō Aratake and Tomoe Ikeda (eds.), Bunka kōshō ni okeru kakki to sōzō [Turning points and creation in cultural interaction], 65–91. Suita: Kansai Daigaku. Collado, Diego. 1632. Ars grammaticæ Iaponicæ linguæ [A Grammar of the Japanese language]. Rome: Sacra Congregatio de Propaganda Fide. Doi, Yōichi, Atsushi Hamada and Akira Yasuda. 1959. Wago ruikai-kō [A study of the Waeŏ yuhae]. Kokugo Kokubun 28(9). 1–48. Endō, Mitsuaki, Hideto Itō, Sŭnghye Chŏng, Takashi Takekoshi, Shin’ichi Sarashina, Chinwan Pak and Xiăoyún Qū (eds.). 2009. Yŏkhaksŏ munhŏn mongnok [A bibliography on the materials for the study of foreign languages in pre-modern Korea]. Seoul: Pangmunsa. Frellesvig, Bjarke. 2010. A History of the Japanese Language. Cambridge: Cambridge University Press. Fukushima, Kunimichi. 1975. Vachikan shahon Reg.Lat.459 no Nihongo hyōki [The spelling of Japanese in the Vatican manuscript Reg.Lat.459]. Jissen Joshi Daigaku Bungakubu Kiyō 17. 1–22. Fukushima, Kunimichi. 1977. Soto kara mita Nihongo [External perspectives on Japanese]. In Atsuyoshi Sakakura (ed.), Nihongo kōza [A course in Japanese], vol. 6: Nihongo no rekishi [The history of the Japanese language], 287–324. Tokyo: Taishūkan Shoten. Hagers, Steven. 1997. The Importance of a Korean source on the Ryukyuan language. In Bjarke Frellesvig and Roy Starrs (eds.), Japan and Korea: Contemporary studies, 34–43. Aarhus: Aarhus UP. Hamada, Atsushi. 1952. Hatsuon to dakuon to no sōkansei no mondai: Kodaigo ni okeru dakushiin no onka [On the interrelation of syllable-final nasals and voiced obstruents: The sound value of voiced obstruents in ancient Japanese]. Kokugo Kokubun 21(3). 198–212. Hamada, Atsushi. 1955. Gomatsu no sokuon [On word-final /T/]. Kokugo Kokubun 24(1). 15–28. Hamada, Atsushi. 1965. Kōji gonen Chōsen-ban Iroha onmon taion-kō: Kokugoshi no tachiba kara [A study of the han’gŭl transcriptions in the Irop’a printed in Korea in 1492: From the perspective of the history of Japanese]. In Kyōto daigaku bungakubu kokugogaku kokubungaku kenkyūshitsu (ed.), Kōji gonen Chōsen-ban Iroha [The Irop’a printed in Korea in 1492], 20–32. Kyoto: Kyōto Daigaku Kokubungakkai. [First published in 1952.] Hashimoto, Shinkichi. 1928. Bunroku gannen Amakusa-ban Kirishitan kyōgi no kenkyū [A study on the Doctrina printed in Amakusa in 1592]. 2 vols. (Tōyō Bunko Ronsō 9). Tokyo: Tōyō Bunko. Hoetink, B. 1920. Verhaal van het vergaan van het jacht De Sperwer [Narrative of the shipwreck of the yacht De Sperwer]. ’s-Gravenhage [The Hague]: Nijhoff. Hoffmann, Johann Joseph. 1864. De groote studie (Ta hio of Dai gaku) / The grand study (Ta hio or Dai gaku). Leiden: Brill.
6 Late Middle Japanese phonology, based on Korean sources
165
Hŏ, Inyŏng. 2014. Chŏnil toin-ŭi Han’gugŏ pogwŏn-gwa ŭmunnonjŏk yŏn’gu [A reconstruction of the Korean language in Zen’ichi dōjin and its phonological study]. Seoul: Korea University master’s thesis. Ifa, Fuyū. 1932. Goon hon’yaku shakugi [An interpretation of the Ŏŭm pŏn’yŏk]. In Kanazawa hakushi kanreki shukugakai (ed.), Tōyōgogaku no kenkyū [Studies in oriental languages], 295–402. Tokyo: Sanseidō. Ikegami, Mineo (tr.). 1993. Rodorigesu Nihongo shōbunten [The Arte breve of Rodriguez]. 2 vols. (Iwanami Bunko, blue series 681-1/2). Tokyo: Iwanami Shoten. Irwin, Mark and Heiko Narrog. 2012. Late Middle Japanese. In Nicolas Tranter (ed.), The Languages of Japan and Korea, 246–267. London and New York: Routledge. Jiang, Chuidong. 1997. Nihonkan yakugo no ‘e’ o megutte [Concerning the vowel e in the Ribenguan yiyu]. Tsukuba Nihongo Kenkyū 2. 133–143. Kawakami, Shin. 1980. ‘Apu’ kara ‘ō’ made [From apu to ō]. Kokugakuin Zasshi 81(7). 1–9. Kyōdai [=Kyōto daigaku bungakubu kokugogaku kokubungaku kenkyūshitsu] (ed.). 1987. Kaishū Shōkai shingo [The revised edition of the Shōkai shingo]. Kyoto: Kyōto Daigaku Kokubungakkai. Lange, Roland Albin. 1969. Documentary evidence for a palatalized /e/ series in Middle Japanese. The Journal-Newsletter of the Assocation of Teachers of Japanese 6(1). 47–50. Lange, Roland Albin. 1971. Bunken shiryō ni han’ei-shita chūsei Nihongo e-retsu onsetsu no kōgaisei [The palatal nature of /Ce/-type syllables in Middle Japanese as reflected in documentary records]. Kokugogaku 85. 36–42. Lange, Roland Albin. 1973. The phonology of eighth-century Japanese: A Reconstruction based upon written records. Tokyo: Sophia University. Ledyard, Gari K. 1966. The Korean Language Reform of 1446. Berkeley, CA: University of California dissertation. Lee, Ki-Moon and S. Robert Ramsey. 2011. A History of the Korean Language. Cambridge: Cambridge University Press. Martin, Samuel E. 1987. The Japanese Language Through Time. New Haven and London: Yale University Press. Mathias, Gerald B. 1974. Review of Lange 1973. The Journal of the Association of Teachers of Japanese 9(2/3). 62–76. Meister, Georg. 1692. Der Orientalisch-Indianische Kunst- und Lust-Gärtner [The Oriental and Indian art and pleasure gardener]. Dresden: Johann Riedel. Morita, Takeshi. 1957. Shōkai shingo kaidai [Explanatory notes on the Shōkai shingo]. In Kyōto daigaku bungakubu kokugogaku kokubungaku kenkyūshitsu (ed.), Shōkai shingo, 1–55. Kyoto: Kyōto Daigaku Kokubungakkai. Morita, Takeshi. 1976. Amakusa-ban Heike monogatari nangokukai no kenkyū [A study of the explanations of difficult words appended to the Amakusa edition of the Heike monogatari]. Osaka: Seibundō. Nakamura, Hidetaka. 1961. Shōkai shingo no seiritsu, kaishū oyobi Wago ruikai seiritsu no jiki ni tsuite [On the dates of the compilation and revision of the Shōkai shingo as well as of the compilation of Waeŏ yuhae]. Chōsen Gakuhō 19. 1–23. Nakayama, Megumi. 2000. Shōkai shingo genkan-bon no hanguru onchū: Shōgaion o arawasu heisho hyōki ni tsuite [The han’gŭl sound glosses in the first edition of Shōkai shingo: On the geminate spellings expressing obstruents]. Reitaku Daigaku Kiyō 70. 99–136. Nomura, Takashi. 2013. Nihongo sutandādo no rekishi [The history of the Japanese standard language]. Tokyo: Iwanami Shoten. Ogura, Shinpei. 1923. Kokugo oyobi Chōsengo hatsuon gaisetsu [An outline of the pronunciation of Japanese and Korean]. Keijō: Ōsakayagō Shoten. Ōhashi, Jun’ichi. 2002. Tōhoku hōgen onsei no kenkyū [A study of the phonetics of the Tōhoku dialects]. Tokyo: Ōfū. Osterkamp, Sven. 2010. Review of Endō et al. (eds.) 2009. Bochumer Jahrbuch zur Ostasienforschung 34. 311–321.
166
Sven Osterkamp
Osterkamp, Sven. 2016a. Onmon iroha zakkō [Miscellaneous thoughts on the various iroha in han’gŭl]. In Yukinori Takubo, John Whitman and Tatsuya Hirako (eds.), Ryūkyū shogo to kodai Nihongo: Nichi-Ryū sogo no saiken ni mukete / Ryukyuan and premodern Japanese: Toward the reconstruction of Proto-Japanese-Ryukyuan, 57–76. Tokyo: Kurosio Publishers. Osterkamp, Sven. 2016b. Foreign sources on the history of the Japanese language: Problems and prospects. Acta Asiatica: Bulletin of the Institute of Eastern Culture 111. 75–93. Ōtomo, Shin’ichi. 1956. Shōkai shingo ni mirareru dakuon hyōki [The spelling of voiced obstruents in the Shōkai shingo]. Gengo Kenkyū 30. 111–112. Ōtomo, Shin’ichi. 1957. Shōkai shingo ni yoru kokugoon no kenkyū [A study of the pronunciation of Japanese based on the Shōkai shingo]. Bunka 21(4). 16–34. Ōtomo, Shin’ichi. 1959. Sō-Kan hitsugo ni yoru kokugoon no kenkyū [A study of the pronunciation of Japanese based on the Sō-Kan hitsugo]. Bungei Kenkyū 33. 49–59. Ōtomo, Shin’ichi. 1963. Chūgoku shiryō ni yoru Muromachi jidai no kokugo onsei no kenkyū [A study of the phonetics of Muromachi period Japanese based on Chinese sources]. Tokyo: Shibundō. Overmeer Fisscher, J. F. van. 1833. Bijdrage tot de kennis van het Japansche Rijk [A contribution to the knowledge of the Japanese empire]. Amsterdam: J. Müller & Comp. Pak, Chinwan. 2005a. Kaikō sōsai kara mita chū/kinsei Nihongo no kenkyū [A study of Middle and Early Modern Japanese as seen from the Haehaeng ch’ongjae]. Kokugo Kokubun 74(2). 1–19. Pak, Chinwan. 2005b. Chōsen shiryō no yotsugana hyōki: Kankokugo on’inshi no kanten kara [The spelling of the yotsugana in Korean sources: From the perspective of the phonological history of Korean]. Kokugo Kokubun 74(8). 1–19. Sakanashi, Ryūzō. 1987. Edo jidai no kokugo: Kamigatago [Edo period Japanese: The language of the Kyoto–Osaka region]. Tokyo: Tōkyōdō Shuppan. Sakuma, Tadashi and Akio Okada (tr.). 1965. Nihon ōkokuki / Nichiō bunka hikaku [Relation of the kingdom of Japan / A comparison of the Japanese and European cultures] (Daikōkai Jidai Sōsho 11). Tokyo: Iwanami Shoten. Siebold, Philipp Franz von. 1826. Epitome linguae japonicae [A sketch of the Japanese language]. Verhandelingen van het Bataviaasch Genootschap van Kunsten en Wetenschappen 11. 63–136. Sugito, Seiju. 1989. Genkan-bon Shōkai shingo no e-dan onsetsu boinbu e no onchū ni tsuite [On the sound glosses of the vowel in /Ce/-type syllables in the first edition of Shōkai shingo]. In Nomura sensei jushō kinen kankōkai (ed.), Nomura Masayoshi sensei jushō kinen gengogaku ronshū [Collected papers on linguistics in celebration of the conferment of the Order of the Rising Sun on professor Nomura Masayoshi], 143–158. Kasugai: Nomura Sensei Jushō Kinen Kankōkai. Takayama, Tomoaki. 1992. Nihongo ni okeru rensetsu boin no chōboinka: Sono rekishiteki imi to hassei no onseiteki jōken [The change from sequences of vowels to long vowels in Japanese: Its historical significance and the phonetic conditions for its appearance]. Gengo Kenkyū 101. 14–34. Thorpe, Maner Lawton. 1983. Ryūkyūan language history. Los Angeles, CA: University of Southern California dissertation. Thunberg, Carl Peter. 1791. Resa uti Europa, Africa, Asia, förrättad åren 1770–1779: Tredje Delen, innehållande resan til och uti kejsaredömet Japan, åren 1775 och 1776 [Travels in Europe, Africa and Asia, made in the years 1770–1779: Part three, containing the voyage to and in the empire of Japan, in the years 1775 and 1776]. Uppsala: Joh. Edman. Thunberg, Carl Peter. 1792. Observationes in linguam Japonicam [Observations on the Japanese language]. Nova Acta Regiæ Societatis Scientiarum Vpsaliensis 5. 258–273. Titsingh, Isaac. 1781. Eenige Japansche woorden [Some Japanese words]. Verhandelingen van het Bataviaasch Genootschap der Kunsten en Weetenschappen 3. 247–270. Toyoshima, Masayuki. 1984. ‘Kaigō’ ni tsuite [On the notion of open vs. closed]. Kokugogaku 136. 152–140. Tsuzuku, Tsuneo. 1968. ‘Au, ou’ kara ‘ō’ e [From au, ou to ō]. Kindaigo Kenkyū 2. 353–371. Unger, J. Marshall. 2009. Some remarks on Hankul transcriptions of Middle Japanese. Scripta 1. 77–87.
6 Late Middle Japanese phonology, based on Korean sources
167
Wenck, Günther. 1959. Japanische Phonetik [Japanese Phonetics], vol. 4. Wiesbaden: Harrassowitz. Witsen, Nicolaas. 1705. Noord en Oost Tartarye [Northern and Eastern Tartary], 2nd edn. Amsterdam: François Halma. Yasuda, Akira. 1960. Jūkan kaishū Shōkai shingo kaidai [Explanatory notes on the revised and reprinted edition of Shōkai shingo]. In Kyōto daigaku bungakubu kokugogaku kokubungaku kenkyūshitsu (ed.), Jūkan kaishū Shōkai shingo [The revised and reprinted edition of Shōkai shingo], 1–62. Kyoto: Kyōto Daigaku Kokubungakkai. Yasuda, Akira. 1966. Naeshirogawa no Chōsengo shahon-rui ni tsuite: Chōsen shiryō to no kanren o chūshin ni [On the Korean language manuscripts from Naeshirogawa: Focusing on their relationship to the Korean sources on Japanese]. Chōsen Gakuhō 39/40. 210–237. [Reprinted as: Yasuda, Akira. 1980. Naeshirogawa no Chōsengo shahon [The Korean language manuscripts from Naeshirogawa]. In Akira Yasuda, Chōsen shiryō to chūsei kokugo [Korean sources and Middle Japanese], 358–385. Tokyo: Kasama Shoin.] Yi, Wŏnsik. 1984. Chōsen tsūshinshi ni zuikō-shita wagaku yakkan ni tsuite [On the interpreters of Japanese accompanying the Korean diplomatic missions to Japan]. Chōsen Gakuhō 111. 53–117.
Masayuki Toyoshima
7 Late Middle Japanese phonology as reflected in early Japanese Christian documents 1 Principles The Jesuit Mission to Japan (1549–1611) left more than thirty works printed in Japan, of which nearly twenty are in Japanese, eleven are in Latin and Portuguese, and four are grammars and dictionaries of Japanese written in Portuguese. These are the sole extant transcriptions of the Japanese language of that era in the Latin alphabet, except for sporadic fragments of transcriptions by merchants and the Dominican mission materials that followed more than twenty years later. Thus, the Jesuit materials have been considered to be one of the major sources for the phonology of Late Middle Japanese.
1.1 What Latin transcriptions reveal The native Japanese writing system using moraic kana characters did not have a way to distinguish a moraic tsu つ from a consonantal prolongation or doubling (sokuon): e.g. もつとも (mo-tsu-to-mo) stood for both /motsutomo/ ‘even if one has’ and /mottomo/ ‘understandably’. The current Japanese writing system has acquired the “small tsu” character for representing the prolonged consonant, but in the premodern kana syllabary, あまつさへ (a-ma-tsu-sa-he) ‘even if’, may have stood for /amatsusaje/ or /amassaje/, and やつこ (ya-tsu-ko) ‘servant’ could have meant /jatsuko/ or /jakko/, and one cannot tell the difference from the kana syllables alone. According to the Jesuit materials written in the Latin alphabet, they were /amassaje/ (spelled {amassaye}; {. . .} designates original orthography hereinafter), and /jatsuko/ {yatçuco}, respectively. There can hardly be found other materials that reveal so clearly the distinction between these syllables. The same goes for un-voiced/voiced (sei-daku) pairs such as kakayaku ‘sparkle’ (ModJ kagayaku) and f/w alternations, which are common in the development of Japanese: e.g. fawa ‘mother’ (ModJ fafa > haha). As the kana system in those days specified voicing only sporadically, nor did it have a way to reflect f/w alternations according to their phonological distinction, the Latin transcription is precious as a positive testimony of the phonological (and even phonetic) state of Japanese at that time. The current Japanese writing system is still unable to represent certain phonological distinctions with kana: e.g. こうし (ko-u-shi) cannot distinguish /koo-si/ (公私 ‘public and private’) from /ko-usi/ (仔牛 ‘calf’), and ぎょう (gi-yo-u) cannot distinguish /gjoo/ (業 ‘task’) from /gjo-u/ (御宇 ‘reign’); according to the Jesuit transcriptions, they were differentiated as https://doi.org/10.1515/9781614512851-008
170
Masayuki Toyoshima
{côxi} / {covji} (/kouʒi/, voiced), and {guiô} / {guiov}. Here we can see much information that we would otherwise not know.
1.2 How to interpret a Latin transcription Jesuit transcription of Japanese is fairly consistent. One current standard interpretation of Jesuit transcription is that proposed by Hashimoto (1928), on the basis of his study of the Japanese Jesuit “Doctrina Christãa” (Christian Doctrine in Japanese, 1592). 1. {ca},{da},{ma},{na}, etc: like present-day Portuguese (PP) and Early Modern Portuguese (EP) i.e. [ka], ]da], etc. 2. {xe}, {je}: like PP and EP i.e. [ʃe], [ʒe] 3. {f}: like PP/EP [f] (labio-dental), but can also be [φ] (bi-labial) 4. {sa}, {so}: like PP/EP at word-initial (otherwise they become [za], [zo]) 5. {cha}, {chi}, {cho}, like EP, i.e. [tʃa], [tʃi], [tʃo] (cf. in PP [ʃa], [ʃi], [ʃo]) 6. {ǒ}, like oo of PP, i.e. [ɔ], but long in duration, i.e. [ɔː] 7. {ô}, like ou of PP (Lisbon), i.e. [o], but long in duration, i.e. [oː] 8. {qi}, {qe}, these spellings are not in genuine Portuguese, but are assumed to be abbreviations of {qui}, {que}, which are interpreted just like PP/EP [ki], [ke] 9. {tçu}: not found in standard Portuguese, but interpreted as [tsu] 10. {zzu}: not found in standard Portuguese, but interpreted as [dzu] (João Rodriguez spells it {dzu}) 11. {gi}: PP does not distinguish this from ji (ʒi), but they are distinguished in these transcriptions, therefore it is interpreted as [ʤi] This interpretation combines different principles: in some cases, it follows Present-day Portuguese (PP), in others Early Modern Portuguese (cf. Bechara 1991), and in others neither. As the phonological system of Early Modern Portuguese (i.e. around 1600 AD) was probably different from that of Japanese, or even from that of the Modern Portuguese, this is inevitable. The problem is how to ensure the plausibility of the interpretation without consistency in the principle(s).
1.3 Latin transcription is not always transparent Since different languages have different phonological systems, transcriptions cannot always be interpreted in one single way: transcription is an adaptation of a rudimentary principle to each linguistic system. As a rule of thumb, a transcription of a target (transcribed) language A, by a mediate (transcribing) language B, should be read according to the system of B. However, deviations from the B system are not rare, as is the case with the introduction of distinctions between {g} and {j} in Portuguese, and {b} and {v} in Spanish, for the purpose of transcribing Japanese.
7 Late Middle Japanese phonology as reflected in early Japanese Christian documents
171
Present-day Portuguese as well as Early Modern Portuguese do not distinguish {ge} from {je}, {gi} from {ji}: e.g. {lalanjeira} ‘orange’ was also spelled {lalangeira} (Cardoso 1592). But Anchieta (1595) employs {gi} and {ji} distinctively in his (Brazilian) Tupi transcription, based on certain morpho-phonological contrasts. The Japanese transcription by the Jesuits (supra) does the same, as {gi} and {ji} are distinct typically in conjugated forms: e.g. {fa-dzu} / {fa-gi-te} ‘to be ashamed’, {san-zu} / {san-ji-te}‘to visit a honorable person’, where the conjugation paradigm {dzu}-{gi} is never confused with that of {zu}-{ji} because they were morpho-phonologically distinct in the transcribed language, Early-Modern Japanese. Neither Early-Modern nor Modern Spanish distinguishes {b} from {v}: e.g. {vocabulário} was also spelled {bocabulário} (perhaps due to attraction from {boca} ‘mouth’). Nevertheless, Gilberti (1558) distinguishes {b} and {v} for the purpose of representing (Mexican) Tarascan phonological contrasts. Mentrida (1637) did the same thing for his (Filipino) Bisaya transcription. Carochi (1645) distinguishes {z} and {ç} for (Mexican) Nahuatl, even though Spanish, his describing language, had already lost this phonological distinction. This also parallels the case of the Jesuit grammarian João Rodriguez. He describes Japanese s consonants as “not so much s, as ç”, therefore with a distinction between {s} and {ç}. However, the Portuguese {s}/{ç} distinction was already fading among his contemporaries. For example, Manoel Barreto, approximately the same age as Rodriguez, has no distinction between these two: he writes {semelhança} / {semelhansa} ‘similarity’ as free variants. The printed materials by the Jesuit Mission to Japan, such as Vocabulário da língoa de Japam, also confuse {s} and {ç}. These cases reveal that spelling pairs, even when their phonological distinctions had already been lost in the describing language, can be employed to illustrate contrasts in the described language. Therefore, the distinction being encoded cannot be decoded on the basis of the phonological system of the describing language alone. One cannot argue that {z}/{ç} contrasts in the Nahuatl transcriptions are sheer spelling variants, just because Spanish had already lost the {z}/{ç} contrast as a phonological distinction. Even when the phonological system abandons a distinction, the spellings may still serve to describe a contrast in the target systems.
1.4 Idiolect The discrepancy between the two Portuguese missionaries, Manoel Barreto (1563?– 1620) and João Rodriguez (1561?–1633) the grammarian, comes from their backgrounds. Barreto was from a suburban area of the major city of O Porto, whereas Rodriguez was from a small mountainous village of Sernancelhe. Their dialects may have been considerably different, as the dialects of these places still differ not a little even today. The Jesuit transcription of Japanese employs {ô} for Japanese long vowels. The grammarian João Rodriguez explains this as “just like ou of Portuguese”. The problem is which “Portuguese” he meant. In the dialect of today’s Lisbon, {ou} is /o/ (cf. {o} is /u/),
172
Masayuki Toyoshima
and it is believed to have been so since the end of the sixteenth century. Thus Hashimoto (1928) interpreted this {ou} as /oː/, as it stands for a “long” vowel of Japanese. However, the grammarian also says it is pronounced “as if with two vowels, o and u”, suggesting a diphthong /ow/, which is exactly the current state of the dialect of Sernancelhe, his birthplace. Sernancelhe is known to be very conservative in a linguistic sense. A linguistic study of about 130 years ago (Gonçalves Viana 1883) notes that the phonological contrast of s/ç, which had been believed to have disappeared in the sixteenth century, still survived in the northern part of Portugal, and a more extensive dialectal survey in 1943 (Boléo 1943) records scattered existence of this s/ç contrast in Sernancelhe. This may be the reason that Rodriguez could describe Japanese /s/ as “not so much s, as ç”. (Toyoshima 2009) In interpreting any phonetic description, such as those in Rodriguez’s grammar, the author’s idiolect may need to be called upon.
1.5 Extant documents In interpreting the Jesuit transcription of Japanese, we must presuppose that extant printed documents by the Jesuits are reliable, but are they really? Unfortunately, in this case there exists no manuscript parallel to the printed documents, or so-called “printer’s copy”, i.e. the very manuscript to which the printing house referred in page composition of the printed edition. However, we do have a fair copy of João Rodriguez’s “História da igreja do Japão” (A history of the Japanese church), made ready for a submission to the printer and eventually left unpublished. The copy is done by somebody other than the author, to which the author (João Rodriguez, the grammarian) adds corrections, inserting items, or crossing some articles out. These changes even include spelling corrections, such as {s}/{ç} alternations, which thus constituted an “error” to the author, even though for the (dialect of the) copier they were just normally permissible alternate spellings. If Rodriguez had not made these corrections, the “História” might have been published with {s}/{ç} alternations that the author himself would not have approved. Accordingly, we should expect that some such unauthorized changes in spelling might have made it onto the printed page in other cases. Even major sources like Rodriguez’ two grammars of the Japanese language (the “comprehensive” grammar written in Portuguese and published in 1604 in Nagasaki, Japan, and the “abridged” grammar of 1620 from Macao) may have undergone orthographical mutations in the course of page composition and printing. Current Portuguese orthography of verbs uses -{ão} for the future, and -{am} for the past, and nouns have -{ão}: all three are pronounced [ɐũ] (with a shift of the accent): e.g. {melarão} ‘they will add honey’, {meláram} ‘they added honey’, {mélam} ‘they add honey’, and {melão} ‘melon’. In Rodriguez’s time, these Portuguese spellings interchange freely as there was no prescribed orthography. Maruyama (1984) points out that, in the 1604 edition, these endings are uniformly -{ão} (irrespective of tense) before folio 94,
7 Late Middle Japanese phonology as reflected in early Japanese Christian documents
173
and uniformly -{am} (even nouns) after folio 94. This is evidence that a strong spelling unification took place during its page composition and printing, as the author himself would hardly have made this kind of unification in drafting, given what we know of his idiolect. There is no reason to believe that this inclination toward excessive uniformity affected only Portuguese spellings while leaving untouched the Japanese transcriptions in which we are so interested.
1.6 Variants in phonology and variants in writing João Rodriguez was not consistent in his spellings. Rather, his Portuguese contains many spelling variants which were common to contemporaneous Portuguese writers, such as alternation between -{ão} and -{am} (supra): he even alternated his autograph signatures between {João} and {Joam}. Accordingly, a problem arises: if such script variants are also found in Japanese transcriptions, should we interpret them as phonetic (or even phonological) variants, or just arbitrary spelling variations? The Jesuits’ Japanese-Portuguese dictionary (Vocabulário da língoa de Japam, 1603, Nagasaki, Japan, henceforth DJ) is the largest of all contemporaneous Portuguese dictionaries. DJ is notorious for its unstable spellings of Japanese long vowels, such as {meǒ} / {miǒ}, {feǒ} / {fiǒ}: DJ156D18 DJ161A17 DJ086C46
Meǒchô. Amenhaã polla manhaã. ‘Tomorrow morning’. Miǒchô. Amenhaã pola menhaã. ‘Tomorrow morning’. Feǒgiǒ. l (or), fiǒgiǒ. i, Dancǒ. Consulta. ‘Conference’.
On the one hand, these variants may have stood for phonetic variations. On the other hand, the Portuguese {manhaã} / {menhaã} are sheer script variants for the same word for ‘morning’, and {polla} / {pola} are also script variants for the same preposition (and an article). Accordingly, how do we know whether Japanese {meǒ} / {miǒ} and {feǒ} / {fiǒ} are phonetically different or merely sporadic deviations in spelling like those in the Portuguese? As a dictionary, DJ is mainly concerned with translating Japanese into Portuguese, but sometimes Japanese loanwords are found incorporated with varying levels of thoroughness to Portuguese phonology: A) Japanese words transliterated as they are: DJ265D01 Toxicoxi. O passar de hum anno a outro, ou sair do inuerno, & entrar na primavera, o que acontece na derradeira noite do Xiuasu, que he a duodecima lũa, & no cabo, ou termino do Inuerno, que às vezes se desencontra do derradeiro dia do Xiuasu: estes dous tempos chamão Rixxun que são duas maneiras de começar o Faru. ‘Passing from one year to another, or leaving winter and entering spring, which happens on the last night of Xiuasu (Shiwasu ‘the twelfth lunar month’), or at the end of winter, which sometimes does
174
Masayuki Toyoshima
not coincide with the last day of Xiuasu: these two cases are called Rixxun, the start of Faru (‘spring’)’. B) Japanese words with a (Portuguese) article: DJ046A33 Chabon. Hũa esasinha como bandeja em que poem os aparelhos do Chanoyu. ‘A small tray on which instruments for the Chanoyu (‘tea ceremony’) are placed’. C) Japanese words pluralized: DJ125A02 Gijiqij. Hum paos que atrauessão de columna a columna nos Zaxiquis ao longo dos Tatamis. ‘A beam which traverses columns of Zaxiqi (‘salon’, pl.) along Tatami (‘mat’, pl.)’. D) Japanese words thoroughly re-declined according to Portuguese phonology: DJ042B36 Cataita. Hũa taboa sobre que os tintureiros fazem as pinturas dos qimõis. ‘A table on which pictures of Kimonos (‘clothing’, pl.) are drawn’. (Qimõis is a plural form of quimão, which in itself is a Portugalized form of Japanese Kimono ‘clothing’.) This last case of thorough incorporation according to Portuguese phonology is of particular interest to us here. For Japanese beǒbu ‘folded screen’, we even have a diminutive form beǒbosinho: DJ247A04 Tçuitachi xǒji. (‘standing screen’) Hũa taboa, ou porta como beǒbosinho baixo. ‘Board or door with short screen’. This word {beǒbu} has a variant form {biǒbu}. Both {beǒbu} and {biǒbu} appear in Japanese headings, as well as in Portuguese translations as loan-words, often pluralized. 1. DJ148D36 Macurabeǒbu. (‘A bedside folding screen’) Beòbus (pl.) pequenos, & curtos. ‘Small and short folding screens’. 2. DJ283A01 Xǒji, beǒbuno foneuo voru. Fazer a armação de canas, ou paos das portas de papel, Beǒbus (pl.), &c. ‘To weave frames of a paper door or a folding screen’. 3. DJ039A18 Ano biǒbuuo caqidaxe. Tirai fora aquelle biǒbu (sg.). ‘Carry out that screen’. 4. DJ231B34 Sunagono biǒbu. Biǒbus (pl.) salpicados de ouro, ou prata. ‘Screen ornamented with golden powder’. The citations above show that the Japanese usages and their Portuguese translations agree in their selection between {beǒ} and {biǒ}: i.e. when the Japanese part employs {beǒbu}, so does the Portuguese part, and vice versa. This happens in all twenty-four examples of {beǒbu} / {biǒbu} found in DJ, except for the very first appearance on DJ010D25. In addition, examples on the same page share the same {beǒ} / {biǒ} selection, e.g. three examples on fol. 23R are all {biǒbu}, and two examples on fol. 283R are both {beǒbu}. These cross-linguistic uniformities in spelling suggest that they are only scribal, and not phonetic, as there is no reason in a dictionary to unify phonetic variants in Japanese and Portuguese loan-words just because they are printed on the same page, or to harmonize the phonetic variations in the two languages. Besides, {beǒbu} / {biǒbu} appear
7 Late Middle Japanese phonology as reflected in early Japanese Christian documents
175
even in Portuguese translations, where no Japanese counterpart {beǒbu} / {biǒbu} exists, and there still the {beǒ} / {biǒ} alternation occurs, and so it cannot be evaluated as representing Japanese phonetic variations at all. The doublet spellings may appear to represent phonetic variants, but in fact they are just scribal alternations. Nevertheless, there are other views on these pairs {beǒ} / {biǒ}, {meǒ} / {miǒ}, etc., which take the alternation to represent phonetic variants, for example Takemura (2011, 2012), q.v. It is interesting to examine the grounds for their arguments. The problem of discerning between phonetic (not phonological) variants and scribal variants is always difficult, as these variants are not semantically contrastive. If the document has some principle in transcription, such as unifying the spellings in the parallel translations as in the case of DJ, that will help us to evaluate the variations.
2 State of the phonological system With the above caveats, the Jesuit transcription of Japanese around 1600 AD records the following phonological system. It is quite close to Modern Japanese (ModJ) as concerns the allowance for hiatus and the number of vowels, but it still holds the /oo/-/ow/ distinction (kaigō), the /di/-/zi/, /du/-/zu/ distinction (yotsugana), and uses nasals before voiced consonants. These are all features which disappeared several decades later.
2.1 Syllables: Hiatus avoided no longer Looking back to around 1000 AD, /f/ in non–word-initial position changed into /w/, e.g. /mafi/ ‘dance’ into /mawi/, and subsequently /w/ was dropped before /i/ and /u/, i.e. /mafi/ > /mawi/ > /mai/; /sofu/ ‘to match’ > /sowu/ > /sou/, thus resulting in a hiatus. Similarly, /w/ before /e/ was also dropped, but in this case /j/ was substituted, i.e. /we/ > /je/, as in: /kowe/ ‘voice’ > /koje/, /nafe/ ‘seedlings’ > /nawe/ → /naje/. Because of confusion between /e/ and /je/, single /e/ syllables had already disappeared. Word-initial /o/ acquired a preglide /w/, thus destroying the phonological contrast between word-initial /o/ and /wo/: /ori/ ‘weaving’ and /wori/ ‘folding’ both became /wori/. By the time of the Jesuit mission (late 15th centuries), word-initial /e/ and /o/ had disappeared and were replaced with /je/ and /wo/ respectively. By contrast, distinctions between /oo/ and /owo/ were retained from 1000 to 1600: e.g. /koori/ ‘luggage’ and /kowori/ ‘county’ or ‘ice’. Though the latter /owo/ sometimes drops the trailing /o/, thus /kowo/ becomes /kow/ (=/kou/); e.g. /kowori/-/kouri/ (in Jesuit spellings {couori} / {côri}) are doublet forms of ‘county’ or ‘ice’. Closed syllables are peculiar to this period. Syllable final /-N/ (hatsu-on) (e.g., {cunqiǒ} /kunkjoo/ or /kuNkjoo/ ‘scent’) and /-Q/ (soku-on) (e.g., {cucqiǒ} /kukkjoo/ or / kuQkjoo/ ‘ultimate’) are still maintained in ModJ. Final /-N/ comprises conditional var-
176
Masayuki Toyoshima
iants [m] (as in mb, mm, mp) and [n] (as in nd, nt, nn) just as in ModJ. In addition to these two, this period has syllable final /-t/ (t-nisshō-on) (e.g., {cutqiǒ} /kutkjoo/ ‘ultimate’: a doublet form of {cucqiǒ} /-Q/ /kuQkjoo/). Final /-t/ appears after a vowel and before consonants such as -tb, tk, td, tg, tm, tn, tr, ts, as well as in word final position -t#. Final /-Q/ and /-t/ are neutralized before /t/ and /d/: e.g. {fiddai} ‘pen case’ and {fitdai} (id.). Some suggest {tk}, {ts} also may have been /-Q/ at this time (i.e. {tk} was [kk], and {ts} was [ss]), but João Rodriguez (1604, 177R) explicitly distinguishes {tecqua} from {tetqua} ‘burning iron’ and {beccacu} from {betcacu} ‘difference’, as well as {faccot} and {facucot} ‘white (i.e., bleached) bone’. Accordingly, /-tk/ must have been phonologically distinct from /-Qk/, and so forms such as /teQka/ {tecqua} and /tetka/ {tetqua}, /kuQkjoo/ {cucqiǒ}, and /kutkjoo/ {cutqiǒ} should be considered phonologically distinct pairs. Although there is no similarly explicit evidence for {t s} being distinct from {ss}, there is no reason to treat them differently from {tk}. Doublets such as {sat satto} / {sassato} ‘sound of a wind blowing in the trees’ or {mat sugu} / {massugu} ‘straight’ are phonologically contrastive. {t s} appears most frequently in Sino-Japanese morphemes used with the verb -su ‘do’, e.g. {met su} ‘disappear’, {cat su} ‘have a thirst’, where {messu}, {cassu} spellings are also found. Moreover, [ŋ] is not explicitly attested in the Jesuit descriptions, but it might have occurred as a free variant before velar consonants (ŋg, ŋk), as well as nasal vowels. No evidence is found for the existence of [ŋ] as a variant of /g/, such as [ŋu] in [toŋu] /togu/ ‘sharpen’, which is found in some dialects of ModJ. Meanwhile, /-Q/ comprises geminates such as kk, ss, pp, or tt, but no voiced geminates gg, zz, bb, dd are attested, except for neutralizations like /fiddai/ (supra). To be sure, {zz} appears in the Jesuit spelling, but it stands for /d/ of /du/, and João Rodriguez the grammarian writes it as {dz}.
2.2 Vowels In the sixteenth century, there were five vowels just like in ModJ, and the “long” vowels are diphthongs (aa, ii, uu, ee, oo, ou). There was a distinction in long “o”, called kaigō ‘open-closed’, between /oo/ (kaion ‘open’) and /ou/ (gōon ‘closed’): e.g. /soo/ {sǒ}‘news’ (左右) vs /sou/ {sô} ‘monk’ (僧), /touzi no too/ {tôji no tǒ}‘the pagoda of Tōji temple’ (東寺の塔), as well as above-mentioned /koori/ ‘luggage’ and /kouri/ ‘county’. Medials are /j/ and /w/. The distinction between /oo/ and /ou/, or kaigō, is effective also after /j/ medials, e.g.: /kjoojou/ {qiǒyô} ‘hospitality’ (饗応) vs /kjoujoo/ {qiôyǒ} ‘care for parents’ (孝養). The distribution of /w/ medial is quite limited. It appears only after /k/ or /g/, and exclusively in these syllables: /kwa/, /gwa/, /kwoo/. e.g.: /kwanzi/ {quanji} ‘to meditate’ (観じ) vs /kanzi/ {canji} ‘to feel’ (感じ), /kwoogen/ {quǒguen} ‘wild land’ (荒原) vs
7 Late Middle Japanese phonology as reflected in early Japanese Christian documents
177
/koogen/ {cǒgen} ‘field’ (郊原), /gwan/ {guan} ‘prayer’ (願) vs /gan/ {gan} ‘wild goose’ (雁). Other combinations such as /kwou/, /gwoo/, /gwou/ are hardly attested. Meanwhile, /gwoo/ has only one example of an onomatopoeia /gwoo-to/, which is a description of a bell tolling, but this word is not recorded in Jesuit documents.
2.3 Consonants The most problematic consonant is /f/. João Rodriguez, the Portuguese Jesuit grammarian, says nothing on the Japanese /f/, as though simply writing it {f} were self-explanatory. As discussed in Section 1.2, Hashimoto interprets this literally as [f], but he also suggests interpreting it as [φ] (the bilabial), without giving any concrete grounds. Meanwhile, Diego Collado, the Spanish Dominican grammarian, dwells upon the pronunciation of the Japanese /f/, alluding to dialectal differences and describing it as “medium inter, f, et, h, os et labia plicando et claudendo, sed non integrum” (‘between f and h, with teeth and lips folded and closed, but not fully’). Since Collado’s description does not mention protruding lips, it suggests [ʍ] (non-round labial approximant unvoiced), instead of [φ] (round labial approximant unvoiced). This contrast between the Jesuit and Dominican grammars may arise from their mother tongues: Portuguese (Rodriguez) and Spanish (Collado). Spanish had just undergone the [f] to [h] shift, such as fijo /fixo/ to hijo /hixo/ ‘son’, foja /foxa/ to hoja /hoxa/ ‘leaf’, with the /h/ still being pronounced, and was sensitive to the distinction between [f] and [h], whereas the Portuguese language saw no changes at all in /f/, even today. One more problem is /s/. As mentioned in 1.4, João Rodriguez, the grammarian, had a firm distinction between {s} and {ç} in his Portuguese idiolect, and he says that the pronunciation of the Japanese /s/ is not entirely {s}, but somewhat inclined to {ç}, so it may have been different from a simple [s], although these are only phonetic distinctions, and there was no phonological distinction between /s/ and /ç/ in Japanese. The two grammars by Rodriguez say that /b/, /d/, /g/, and sometimes /z/, were accompanied by a nasal onset, except in word-initial position. This may have been realized by nasalization of the preceding vowel, e.g. /sadame/ [sãdame] ‘fate’. Rodriguez warns not to confuse this nasalization with a moraic /-n/. Possible confusions may be for example /kago/ [kãgo] ‘basket’ vs /kango/ [kango] ‘Chinese word’, /zizi/ [ʒĩʒi] ‘incessant’ vs /zinzi/ [ʒinʒi] ‘matters of a shrine’, /madara/ [mãdara] ‘spots’ vs /mandara/ [mandara] ‘mandala’. In spite of Rodriguez’ warning, confusions are found, e.g. /sangitʃoo/ vs /sagitʃoo/, both, ‘firework’ (左義長), and /indi uti/ vs /idi uti/, both ‘stoning’ (印地討ち). Distinctions between /di/-/zi/ and /du/-/zu/ were phonetically [ʤi]-[ʒi] and [ʣu]-[zu] respectively, and were already decaying, although Jesuit transcription tries to keep them separate by transcribing {gi}-{ji} and {zzu}-{zu}. As mentioned above in 1.3, in contemporaneous Portuguese (ca. 1600), there were no phonetic differences between {gi} and {ji} spellings, and orthography did not exist to distinguish them either. These spellings were deliberately employed only to maintain these phonological contrasts in the Japanese language.
178
Masayuki Toyoshima
References Anchieta, José de. 1595. Arte de gramática da língua mais usada na costa do Brasil [Grammar of the most used language on the Brazilian coast]. Coimbra: Antonio de Mariz. Bechara, Evanildo. 1991. As fases da língua portuguesa escrita [Phases of written Portuguese]. In Dieter Kremer (ed.), Dialectologie et géographie linguistique, Actes du XVIIIe congrès international de linguistique et philologie romanes [Dialectology and linguistic geography, Proceedings of the XVIIIth International Congress of Romance Linguistics and Philology], vol. 3. 68–76. Tübingen: Max Niemeyer. Boléo, Manuel de Paiva. 1943. O interesse científico da linguagem popular [Scientific interest in the popular language] First appeared in Brasileirismos, later included in Boléo, Manuel de Paiva, 1975, Estudos de linguística portuguesa e românica [Studies in Portuguese and Romance linguistics], 2 vols. Coimbra: Acta universitatis conimbrigensis, Imprensa de Coimbra. Cardoso, Jerónimo. 1592. Dictionarium latino lusitanicum et vice versa lusitanico latinum [Dictionary of Latin-Portuguese and the opposite Portuguese-Latin]. Lisboa: Alexander de Syqueira. Carochi, Horacio. 1645. Arte de la lengua mexicana [Grammar of the Mexican language]. México:Juan Ruyz. Gilberti, Maturino. 1558. Arte de la lengua de Michuacan [Grammar of the Michuacan language]. México: Juan Pablo. Gonçalves Viana, A. R. 1883. Essai de phonétique et de phonologie de la langue portugaise, d’après le dialecte actuel de Lisbonne [Essay on phonetics and phonology of the Portuguese language based on the current dialect of Lisbon]. First appeared in Romania XII, Paris, later included in Gonçalves Viana, A. R., 1973, Estudos de fonética portuguesa [Studies in Portuguese phonetics], 83–152. Lisboa: Imprensa nacional – Casa da moeda. Hashimoto, Shinkichi. 1928. Bunroku gannen amakusa-ban Kirishitan kyōgi no kenkyū [A study on the Doctrina Christãa published in 1592]. Tokyo: Tōyōbunko. Maruyama, Tōru. 1984. Rodorigesu nihon bunten ni okeru porutogarugo seishohō: /ãw/ no hyōki ni tsuite [Portuguese orthography in the Rodriguez’s Japanese grammars: On the representation of /ãw/]. Nanzan Kokubun Ronshū 8. Nagoya: Nanzan University. Reprinted in Tôru Maruyama, 2020, Kirishitan seiki no gengogaku [Missionary linguistics in Japan]. 117–125. Tokyo: Yagi Shoten. Mentrida, Alonso de. 1637. Arte de la lengua bisaya [Grammar of the Bisayan language] Valladolid: Nicolas de la Cruz. Rodriguez, João. 1604. Arte da lingoa de Japam. [Grammar of the language of Japan]. Nagasaki: Jesuit mission. Rodriguez, João. 1620. Arte breve da lingoa japoa. [A short grammar of the Japanese language]. Macao: Jesuit mission. Takemura, Asuka. 2011. Rōmaji-bon kirishitan shiryō no o-dan gō-yō-chōon hyōki [eô/iô forms in the early Christian documents]. Gobun 96 (Society for Japanese Language and Literature). 56–69. Osaka: Osaka University. Takemura, Asuka. 2012. Nippo jisho no kai-yō-chōon [eǒ/iǒ forms in Vocabulario da lingo de Japam]. Kokugo Kokubun 81(3). 1–26. Kyoto: Chuo Tosho Shuppan. Toyoshima, Masayuki. 2009. Tsūji rodorigesu no kokyō serunanserye o tazunete [Visiting Sernancelhe, the birthplace of João Rodriguez, the interpreter]. In Nihon Kindaigo Kenkyūkai (ed.), Nihon Kindaigo Kenkyū 5. 277–292. Tokyo: Hituzi Syobo.
Marc Hideo Miyake
8 Sino-Japanese 1 Introduction The term “Sino-Japanese” (SJ) can refer to anything involving both China and Japan (e.g. the Sino-Japanese Wars). However, in a linguistic context, “Sino-Japanese” refers to Chinese borrowings in Japanese. Sino-Japanese has many strata which I will group into five categories for simplicity: 1. Naturalized Sino-Japanese (early centuries CE; §3) 2. Early Sino-Japanese (ESJ; a.k.a. Go-on 呉音; c. 5th–6th centuries CE; §4) 3. Middle Sino-Japanese (MSJ; a.k.a. Kan-on 漢音; c. 7th century CE; §5) 4. Late Sino-Japanese (LSJ; mostly a.k.a. Tō-on 唐音; 2nd millennium CE; §6) 5. Neo-Sino-Japanese (Neo-SJ; modern times; §7) “Sino-Japanese” is most often equated with the on ‘sound’ readings of kanji (Chinese characters used in Japanese). Even though both on and kun ‘gloss’ readings are Japanese pronunciations of kanji, on are often called “Chinese readings” as opposed to kun readings which are “Japanese readings”. “Chinese readings” is shorthand for “Japanese readings of Chinese origin”, whereas “Japanese readings” is shorthand for “Japanese readings of native origin”. However, on are not always borrowings from Chinese. Conversely, a few kun are not of native origin (Table 1). Table 1: Combinations of etymological sources and kanji reading types. Etymology \ Reading
On
Kun
Chinese origin Japanese origin
Sino-Japanese (Go-on, Kan-on, Tō-on) Pseudo-Sino-Japanese for kokuji
Naturalized Sino-Japanese Nearly all kun
Two types of readings complicate the Chinese on / native kun dichotomy: First, there are a small number of on readings for kokuji 国字 (made-in-Japan characters) which are modelled after on readings of genuine Chinese characters: e.g. doo for the kokuji 働 ‘work’ by analogy with Early Sino-Japanese doo (< Early Middle Chinese1 ✶doŋˀ) for 動 ‘move’. These pseudo-Sino-Japanese readings are inspired by Sino-Japanese but are not themselves Sino-Japanese because they are not from China.
1 Chinese reconstructions are in my modifications of the following systems: Pulleyblank (1991) for Early and Late Middle Chinese, Schuessler (2009) for Late Old Chinese, and Baxter and Sagart (2014) for Early Old Chinese. I use ✶y to represent a palatal glide (IPA [j]) for compatibility with Japonic forms. https://doi.org/10.1515/9781614512851-009
180
Marc Hideo Miyake
Second, there are a small number of kun readings that are early Chinese loans whose foreign origin is only recognized by specialists. I regard them as naturalized Sino-Japanese (§3) rather than Sino-Japanese proper. I use the term “Sino-Japanese” here in a broad sense to refer to any Chinese loans in Japanese, regardless of whether they are classified as on or kun readings. The most recent Sino-Japanese kanji readings are neither on nor kun (§7).
2 The phonetic characteristics of modern Sino-Japanese Modern Sino-Japanese on readings fit a strict template (1). Very early and very recent Sino-Japanese borrowings may deviate considerably from this template (§3, §7). (1) First mora Second mora (C)(y)V + -Ø -i -u (after -u) -o (after -o) -ki -ku -ti, -tu -n
The reflexes of ✶-Vpu sequences are indistinguishable from those of original ✶-Vu (< Chinese ✶-Vw) sequences in post-1945 kana orthography for SJ.
182
Marc Hideo Miyake
-k was generally later Japanized as -ki and -ku which survive today. A few SJ readings whose Chinese sources ended in ✶-p and ✶-k now have -tu in modern SJ: e.g. ritu 立 < ✶-p ‘stand’ and satu 冊 < ✶-k ‘volume’. Perhaps their codas were irregularly reduced to [ʔ], which may have been an allophone of final /t/. Then that final [ʔ] was reinterpreted as a /t/ that was later expanded to -tu. -n ([ɴ] with homorganic allophones before consonants) is the only permissible final nasal in SJ (as in modern Japanese as whole). It mostly corresponds to Chinese ✶-m and ✶ -n, though it also corresponds to Chinese ✶-ŋ in later strata (§6, §7). The Heian period materials examined by Numoto (1995) imply a fading distinction between ✶-m and ✶-n. Although both ✶-m and ✶-n could be written as or , there was a tendency to write ✶-m as and ✶-n as . The onset of the second syllable of sanmi 三位 ‘third rank’ from san < ✶sam(u) 三 plus 位 i < ✶wi reflects an earlier final ✶-m(u) in the first syllable. -i, the length of some back vowels (< ✶-u), and occasionally zero may correspond to Chinese ✶-ŋ. In Miyake (2003: 193), I proposed that “OJ [Old Japanese] had a short-lived phoneme ✶ŋ found only in SJ loans which was later lost”: e.g., koo 香 (pre-1945 and a man’yōgana for OJ kaga [kaŋga], kagu [kaŋgu], and kago [kaŋgo]) was ✶kaŋu in the 8th century.2 Unger (2008: 51) cites (7), a 7th century mokkan phonetic equation3 from Tranter (2001) as evidence for reconstructing an early SJ ✶-ŋ-: ✶
(7) 熊 ✶uŋ(u) or ✶woŋ(u) = 汙 ✶u or ✶wo + 吾 ✶ŋu Early SJ ✶-ŋ- nasalized following high vowels before disappearing as in (8) and (9). -ŋi > ✶ŋĩ > ✶-ĩ > ✶-i
(8)
✶
(9)
✶
-ŋu > ✶ŋũ > ✶-ũ > ✶-u
woŋ(u) for 熊 is close to オウレ , the Go-on (§4) reading of its homophone 雄, in the Kanchi-in manuscript of Ruiju myōgishō which indicates nasalization with a kaeriten-shaped diacritic レ that I transliterate as . Another diacritic indicating nasalization in the Middle Japanese pronunciation of SJ was the anusvāra dot of the Siddham script which could be added to the kana for as well as (Frellesvig 2010: 164): e.g., イ˙ < ✶-iŋ. Nasalized vowels sometimes voiced initial obstruents in following syllables: e.g., ✶
(10) seibai 成敗 < ✶seĩ + ✶pai (cf. Late Middle Chinese ✶ɕʱieŋ pæ̀ j) ‘punishment’
2 Old Japanese forms are in Miyake’s (2003) reconstruction. 3 The phonetic interpretation of the formula is mine.
8 Sino-Japanese
183
(11) syoogun 将軍 < ✶syaũ + ✶kun4 (cf. Early Middle Chinese ✶tsɨaŋ kun) ‘general’ (12) syuzyoo 衆生 < ✶syũ + syaũ (cf. Early Middle Chinese ✶tɕuŋʰ ʂɨæŋ) ‘all living creatures’ syu 衆 < ✶syũ is one of a small number of readings whose only vowel was nasalized. Restoring all of the archaic features listed above enables me to posit a simpler, Chinese-like formula for the earliest reconstructible SJ on readings: (13) ✶C1GV1(V2/C2) G represents either ✶-w- or ✶-y-. I list possible first vowels in Table 7 in §4. The optional second vowel could be either -i or -u. The optional second consonant could be a nasal (✶-m, ✶-n, ✶-ŋ) or a stop (✶-p, ✶-t, ✶-k). Some of the earliest SJ borrowings (§3) may not conform to (13). ✶
3 Naturalized Sino-Japanese Some early Chinese loans are so integrated into the basic vocabulary of Japanese that they are considered to be kun readings of the kanji associated with them even though they are not native words (Table 1). In Miyake (1997) I called these loans “pre-Sino-Japanese”, but this may be a misnomer because some of these loans might postdate the earliest SJ borrowings: e.g. zeni ‘money’ may have been borrowed after the raising of pre-Old Japanese ✶e to Old Japanese ✶i (§4) if it is based on Early Middle Chinese ✶dzien 銭. (However, its -e- could also be from a pre-Old Japanese ✶ia if it is based on Late Old Chinese ✶dzian.) I now prefer the term “naturalized Sino-Japanese”. In any case, these loans do not conform to the phonological formulae for SJ proper (1 and Table 2). Their second syllables are not restricted to the limited set of SJ: e.g. the -ni of zeni ‘money’ is not permitted in SJ. Some have Old Chinese characteristics absent from SJ proper which is based on post-Old Chinese varieties of Chinese: e.g. tono < Old Japanese ✶tənə ‘pavilion’ may ultimately be from a Han Dynasty Chinese ✶dənh 殿 whose schwa later fronted in Late Old Chinese ✶denh. Two frequently cited examples of naturalized Sino-Japanese have first syllables that do not obviously correspond to anything in Chinese (Table 3).
4 軍 is read as gun even in other contexts when it was not preceded by an earlier nasalized vowel. The irregular g- of gun 軍 corresponding to Early Middle Chinese ✶k- may be due to an early reanalysis of the high-frequency loan 将軍 as ✶/syaũ/ + /Nkun/ instead of ✶/syaũ/ + /kun/.
184
Marc Hideo Miyake
Table 3: ‘Horse’ and ‘plum’.5 Sinograph
Gloss
Modern Japanese
Proto-Japonic5
馬 梅
horse plum
uma ume
✶
muma mumai or ✶muməi
✶
Early Middle Chinese mæˀ məy
✶ ✶
Early Old Chinese mrˁaʔ C.mˤə
✶ ✶
Vovin (2005: 59) rejected a Chinese etymology for those words and regarded their origins as “obscure”. Old Chinese ✶mrˁ- in ‘horse’ should correspond to a Proto-Japonic ✶ mur-, not ✶mum-. The ✶C. of Old Chinese ✶C.mˤə ‘plum’ was probably a voiceless consonant, not ✶m-.6 There is no agreement on which Japanese words are naturalized Sino-Japanese. Miyake (1997) attempts to evaluate previous proposals for naturalized Sino-Japanese and Sino-Korean words.
4 Early Sino-Japanese (Go-on 呉音) The earliest strata of large-scale borrowings from Chinese are usually collectively referred to as Go-on 呉音 ‘Wu sounds’. Despite their name, they are most likely borrowed via the extinct Paekche language of the Korean peninsula rather than directly from the Chinese region of Wu that Pulleyblank (1991: 3) equated with “the old southern [Chinese] capital at Jiankang (Nanking) and its surrounding territory, which had been the most prestigious cultural centre at the time of division and where the old EMC [Early Middle Chinese] standard was presumably most persistent.” Paekche speakers introduced literacy in Chinese to Japan, so the Japanese may have initially learned to pronounce Sino-Paekche: the Paekche style of pronouncing Chinese characters. Since Sino-Paekche has been lost, it is not clear what traits of Go-on are Sino-Paekche innovations. However, the following seven characteristics of Go-on point to southern Early Middle Chinese (EMC) as its ultimate source: 1. Partial preservation of the fading EMC distinction between the ✶-ie 支, ✶-i 脂, ✶-ɨ 之, and ✶-ɨy 微 rhymes that was maintained in southern EMC-based Early Sino-Vietnamese (ESV) but lost in the LMC (Late Middle Chinese) source of Kan-on (Table 4). 5 The first two Proto-Japonic reconstructions are by Vovin (2005: 59). The third reconstruction is possible within Frellesvig and Whitman’s (2008) seven-vowel system for Proto-Japonic. I do not reconstruct a morpheme boundary before ✶-i since there is no evidence for an allomorph ✶muma- or ✶mumə- without a suffix. 6 Old Chinese ✶C.mˤə is the likely source of Vietnamese mơ ‘apricot’. The tone of mơ indicates an earlier voiceless ✶m̥-. This ✶m̥- in turn may be from an even earlier ✶C̥.m- with a voiceless ✶C̥- that conditioned the devoicing of ✶m-.
2. Absence of palatal segments in reflexes of EMC ✶ɨV-rhymes (Table 5). Kan-on -yV, -i, and -e- reflect the fronting of EMC ✶ɨ to LMC ✶i. Table 5: Sonorant-final ✶-ɨV-type rhymes. -ɨə 魚
-ɨn 欣
-ɨən 元
-ɨəŋ 陽
-ɨəŋ 蒸
-ɨm 侵
-ɨəm 嚴
EMC rhyme
✶
✶
✶
✶
✶
✶
Go-on
-o
-on < ✶-ən
-on < ✶-ən
-ō < pre-1945 < ✶-aũ
-ō < pre-1945 < ✶-əũ
-on < ✶-əm -on < ✶-əm
ESV
-ưa [ɨə]
LMC
✶
Kan-on
-ân [ən]
-iə
-in
-yo < -yə ✶
✶
-in
-ươn [ɨən]
-ương [ɨəŋ]
-ien
?
-iaŋ
✶
-iəŋ
✶
-en
-yoo < pre-1945 < ✶-yaũ
✶
-yoo < pre-1945 < ✶-yəũ
✶
-im (!)
-ươm [ɨəm]
-im
✶
-iem
✶
-in < -im
-en < ✶-em
✶
3. Go-on -u from EMC ✶-u in the 尤 rhyme class. Cf. ESV -âu [əw] and Colloquial Taiwanese -u. Contrast with Kan-on -yuu < ✶-iw from LMC ✶-iw. 4. Partial preservation of the EMC distinction between ✶ə and ✶a before labial nonglides (Table 6). Table 6: ✶ə and ✶a before labials. EMC rhyme Go-on LMC Kan-on
-əm 覃
✶
-am 談
-on < -əm, -an < -am ✶
-am
✶
✶
✶
-an < -am ✶
-əp 合
✶
-oo < pre-1945 < -əp -oo < pre-1945 < ✶-ap ✶
-ap
-an < -am ✶
✶
✶
-ap 盍
-oo < pre-1945 < ✶-ap
-oo < pre-1945 < ✶-ap
5. Go-on e from EMC ✶æ and ✶ɛ. Cf. ESV e [ɛ] and Colloquial Taiwanese e. Contrast with Kan-on a from LMC ✶æ.
186
Marc Hideo Miyake
6. Southern lowering: Go-on -ai from southern EMC ✶-ay < ✶-ey. Cf. Cantonese, Meixian Hakka, and Colloquial Taiwanese -ai, ESV -ây [əy], ✶-ay in Chinese loans in Proto-Tai (in Pittayaporn’s 2009 reconstruction), and Saṃghabhara’s use of characters of this rhyme class to transcribe Sanskrit -ai in the 6th century (Pulleyblank 1984: 199). Contrast with Kan-on -ei. 7. Southern breaking: -yoo < pre-1945 < ✶-yaũ and -yaku in the EMC ✶-eŋ 青 and ✶-ek 錫 rhyme classes. Cf. southern Late Old Chinese ✶-iaŋ (Schuessler 2001: 74), Norman’s Proto-Min ✶-iaŋ and ✶-iak (Schuessler 2001: 74, 84), and Proto-Hakka ✶-iak (Schuessler 2001: 84). Contrast with Kan-on -ei and -eki. The first five characteristics are retentions indicating the age of Go-on. The last two characteristics are old southern Chinese innovations that rule out a northern Chinese origin despite the fact that Paekche was closer to north China. I have excluded what may be the most striking characteristic of Go-on: its distinct reflexes for EMC voiceless and voiced obstruent initials. This trait is not necessarily archaic or southern. I will compare Go-on and Kan-on initials in §5. Go-on often has multiple vowels for any given Middle Chinese rhyme category. Some of this variation reflects the presence or absence of Chinese sound changes in different strata of borrowing: e.g. the Go-on o-reflexes of the 覃 and 合 rhymes in Table 5 were borrowed before EMC ✶ə and ✶a merged, whereas the Go-on a-reflexes were borrowed after the merger. However, the remaining variation reflects the presence or absence of Japanese vowel shifts. Early Japanese (EJ) had six or seven vowels when the first stratum of Go-on was borrowed (Table 7). Table 7: Early Japanese vowels. i e
✶ ✶
(✶ɨ) ✶ ə ✶ a
u o
✶ ✶
The ✶ɨ reconstructed by Frellesvig and Whitman (2008) is in parentheses because it is not universally accepted. EJ ✶e and ✶o respectively raised to Old Japanese (OJ) ✶i and ✶u whereas the more controversial ✶ɨ lowered to OJ ✶ə. (OJ later developed a new ✶ɨ from EJ ✶ui, ✶oi, and ✶ɨi or ✶əi depending on whether one reconstructs ✶ɨ in EJ or not.) These changes could explain why – Go-on sometimes has u (< EJ Go-on ✶o?) for the EMC ✶-o 模 rhyme class: e.g. ku for EMC ✶kʰoˀ 苦 ‘bitter’ – Go-on sometimes has o (< EJ Go-on ✶ɨ?) for the EMC ✶-ɨ 之 rhyme class: e.g. ko for EMC ✶kɨˀ 己 ‘self’
8 Sino-Japanese
187
Readings with those correspondences could belong to an initial preraising stratum, whereas readings with more straightforward correspondences could belong to a later postraising stratum (Table 8). Table 8: ✶-o and final ✶-ɨ in two Go-on strata. -o 模 -u < OJ ✶-u < EJ ✶-o -o < OJ ✶-o
If one does not reconstruct EJ ✶ɨ, EJ ✶ə would be the sole source of Go-on o. EJ ✶-ə may reflect an ESV -ơ [əː]-like pronunciation of the 之 rhyme class in the source dialect of EMC. In any case, Go-on almost never has i for EMC ✶e.7 It is unlikely that EMC ✶e-readings were completely excluded from the first stratum of Go-on. Was the raising of EJ ✶ e completed before any Go-on were borrowed? If so, then all Go-on postdate the naturalized SJ loan kinu ‘silk’ from EJ ✶kenu which in turn was from EMC ✶kwienʰ 絹. EMC ✶e could have been approximated with a new OJ ✶e whose native EJ sources were diphthongs: ✶ia, ✶iə, and perhaps ✶iɨ (Frellesvig and Whitman 2008: 39). Go-on with medial -ya- (e.g. 経 kyoo < pre-1945 ‘sutra’ and 客 kyaku ‘guest’) must have been borrowed after the monophthongization of EJ ✶ia. Otherwise I would expect Go-on -e- instead of -ya-. Chinese ✶-Vy rhymes have two kinds of reflexes in Go-on: one borrowed prior to the monophthongization of falling diphthongs and one borrowed afterward (Table 9). Table 9: Chinese ✶-Vy rhymes in two Go-on strata. EMC
-ɨy 微
✶
-əy 咍 (later became ✶-ay)
✶
Early Go-on
-e < OJ -əy -e < OJ -əy < EJ -əi < EJ ✶-əi
Late Go-on
-i < OJ ✶-ɨ
✶
✶
✶
-ai < OJ ✶-ai (< late EMC ✶-ay)
Southern ✶-ay < ✶-ey 齊
✶
-ai < OJ ✶-ai
-ai < OJ ✶-ai
-e < OJ -e < EJ -ai (but only after pre-1945 !) ✶
✶
-ayʰ 泰
-e < OJ ✶-e < EJ ✶-ai (but only after pre1945 and !)
No stratification is possible for Go-on reflexes of EMC ✶-uy 微: -i < OJ ✶ɨ < EJ ✶ui after labials and -i < pre-1945 after velars. There are no readings like ki < pre-1945 NWLMC ✶miejŋ (not ✶mbiejŋ) > Kan-on mei (15) EMC ✶mɨen 勉 > NWLMC ✶mbien (not ✶mien) > Kan-on ben However, the initial 日 was always denasalized in Kan-on, implying there was no nasal ɲ- in NWLMC. Kan-on voiced obstruents always correspond to NWLMC prenasalized obstruents, whereas Go-on voiced obstruents always correspond to EMC voiced obstruents (Table 11).
✶
Table 11: A sample of Middle Chinese initials and their SJ reflexes. EMC Go-on NWLMC Kan-on
tt✶ tt✶
tʰt✶ tʰt✶
dd- < ✶nd✶ d- or ✶tʱ-? t✶
nn✶ nn✶
ndd- < ✶nd-
✶
The pronunciation of voiced obstruents in NWLMC is uncertain. Pulleyblank (1984: 68) reconstructed “partly devoiced” consonants such as ✶tʱ- in NWLMC corresponding to Kan-on voiceless obstruents. If he is correct, OJ speakers disregarded the voiced aspiration of NWLMC ✶tʱ- just as they disregarded the voiceless aspiration of NWLMC ✶tʰ- and borrowed both as Kan-on t-. On the other hand, Tibetan transcriptions of NWLMC do not have aspirate letters (e.g.
ཐ or དྷ) that would confirm Pulleyblank’s partly devoiced aspirates. If NWLMC still had fully voiced obstruents, was there a conscious decision to change borrowing styles and stop rendering Chinese voiced obstruents as OJ prenasalized obstruents since the latter were optimal matches for NWLMC prenasalized obstruents?9 Kan-on rhymes reflect a NWLMC system with fewer rhymes than the southern EMC dialect underlying Go-on or the Middle Chinese phonological tradition exemplified by the Guangyun 廣韻 dictionary and the Yunjing 韻鏡 rhyme tables. Some of the rhyme mergers are in Tables 4 and 6 in §4. Kan-on -i < ✶-ĩ and -ki after front vowels are from new NWLMC codas which may have been velars preceded by ✶-y- (✶-yŋ, ✶-yk) or palatalized velars (✶-ŋʲ, ✶-kʲ). These codas were an innovation found in other Tang dialects such as the sources of Sino-
9 Ramsey and Unger (1972) proposed a somewhat different theory: Chinese voiced obstruents were borrowed as voiced obstruents without prenasalization in Go-on but as voiceless obstruents in Kan-on after voiced obstruents lenited in Japanese words other than Go-on. This lenition theory lacks philological evidence (Miyake 2003: 72) and requires Go-on to be ‘immune’ to lenition.
190
Marc Hideo Miyake
Korean (which has and ) and Sino-Vietnamese (which has palatals -nh [ɲ] and -ch [c]). Kan-on tones in Numoto’s (1995) index to the Chōjō 3 (1134 CE) manuscript of the Mōgyū 蒙求 correspond in a regular manner to Middle Chinese tone categories, but the phonetic values of those categories as pronounced in Middle Japanese are controversial.
6 Late Sino-Japanese (Tō-on 唐音) Although the vast majority of SJ borrowings are from the first millennium CE, Chinese words and readings continued to flow into Japan for another millennium. Premodern second millennium CE kanji readings are called Tō-on 唐音 ‘Tang sounds’ even though they all postdate the Tang dynasty. Tō refers to China and not to the Tang dynasty, just as the Kan ‘Han’ of Kan-on refers to China and not to the Han dynasty. Another name for Tō-on is Tō-Sō-on 唐宋音 ‘Tang-Song sounds’. Sō ‘Song dynasty’ is appropriate since a few Song dynasty borrowings are still in common use: e.g. isu 椅子 ‘chair’ and huton 布団 ‘Japanese-style bedding’. Late SJ is distinguished from earlier layers by phonetic characteristics rooted in post-Tang innovations in northern Chinese varieties and in Japanese itself: e.g., – the backing of ✶-i after sibilants in Chinese: late SJ su < Song Chinese ✶tsɨ for the noun suffix 子 (cf. Kan-on < LMC ✶tsí) – the complete devoicing of obstruents and the monophthongization of ✶wa before ✶-n in Chinese: late SJ ton < Song Chinese ✶tʰɔn for 団 ‘group’ (cf. Go-on dan < EMC ✶dwan) – the backing of ✶ɣ to ✶ɦ (in a Chinese dialect without devoicing; borrowed as late SJ zero) – the borrowing of Chinese ✶-ŋ as the Middle Japanese mora nasal -n, a sound that did not exist in OJ – the borrowing of Chinese ✶-ʔ from earlier final stops as Japanese zero The last three innovations are exemplified in (16): (16)
Late SJ angya < Hangzhou Song Chinese ✶ɦaŋ kyaʔ for 行脚 ‘pilgrimage’ (cf. Go-on gyoo < OJ ✶ŋgyaũ < EMC ✶ɣæŋ 行 ‘go’ and Kan-on kyaku < NWLMC ✶ kiak 脚 ‘leg’).
The Japanized Mandarin lyrics of Shingaku 清楽 ‘Qing music’ exemplify the most recent layer of Late Sino-Japanese. Late SJ Sin 清 ‘Qing’ has a final nasal unlike earlier readings: Go-on syoo < pre-1945 and Kan-on sei.
8 Sino-Japanese
191
7 Neo-Sino-Japanese Chinese words and names are sporadically Japanized in modern times: (17) Mandarin jiaozi 餃子 ‘pot sticker’ > pre-WWII neo-SJ tyaotu, postwar neo-SJ gyooza10 (18) Mao Zedong 毛沢東 > neo-SJ Mao Tsooton Unlike previous layers of SJ, Neo-SJ has affricates corresponding to Chinese affricates. In (17), Mandarin j [tɕ] was borrowed as neo-SJ ty [tɕ]. In (18), Mandarin z [ts] was borrowed as neo-SJ t(s) [ts]. These modern borrowings do not conform to the general phonological formulae for Sino-Japanese (1 and Table 2): e.g. Mao has o after a vowel other than o. Such deviations from the Sino-Japanese norm may explain why Neo-Sino-Japanese loans are not considered on even though they are no less Chinese than Go-on, Kan-on, or Tō-on.
Appendix 1: Artificial Sino-Japanese Kan-Wa jiten ‘Sino-Japanese dictionaries’ contain Go-on (§4) and Kan-on (§5) readings for every kanji. But many of those readings are not found in actual use outside dictionaries. They are artificial readings created by lexicographers on the basis of fanqie 反切 spellings from the Chinese phonological tradition. Some of the more obscure characters may never have been pronounced in Japanese. Even common characters have invented readings that Miller (1967: 106) called “linguistic ghosts”. In the introduction to his Sino-Japanese dictionary, Tōdō (1978: 1) admitted that he supplied readings “in cases whenever readings are not found in Japanese linguistic data” on the basis of known readings of kanji that are homophones according to the Middle Chinese rhyme dictionary Guangyun and the Middle Chinese rhyme tables of Yunjing. Unfortunately, these ghost readings are not distinguished in any way from attested readings in the body of any dictionary that I have seen. Moreover, neither Guangyun nor Yunjing reflect the phonological systems of the Chinese dialects underlying Sino-Japanese (Numoto 1993). Some true Go-on and Kan-on readings that did not match lexicographers’ expectations have been mislabeled as kan’yō-on 慣用音 ‘customary readings’: e.g. “current dictionaries recognize fictitious kan’on readings such as 母 (ボウ) and 牡 (ボウ), and as a result the readings 母 (ボ) and 牡 (ボ) found in early texts have been designated ‘customary pronunciation’ or ‘popular pronunciation’ and have been excluded from the kan’on system” (Numoto 1993: 83). 10 Gyooza resembles Sino-Korean kyoja 餃子 and may not be a direct Chinese loan.
192
Marc Hideo Miyake
To avoid citing ghost readings from Sino-Japanese dictionaries, I cite (1) readings in current use in italics and (2) transliterated readings in use before 1945 in angled brackets (< >). Pre-1945 spellings generally reflect earlier pronunciation (but see Numoto 1993 for caveats). I define a reading “in use” as a reading representing a morpheme in a text as opposed to an isolated reading in a dictionary entry.
Appendix 2: Sino-Ryukyuan Like Japanese to the north, the Ryukyuan languages also have Chinese loanwords which can be called Sino-Ryukyuan (SR) by analogy with Sino-Japanese. SR vocabulary first appears in han’gŭl transcription in the Haedong chegukki 海 東諸國紀 (1471): e.g. for ✶syau ŋgwatsɨ < EMC ✶tɕiaŋ ŋuat 正月 ‘first month’, for ✶ten < EMC ✶tʰen or NWLMC ✶tʰien 天 ‘sky’,11 and for ✶dou
> > > >
LH-H LH-L LH-L HL-L HL-L
/ØØ-Ø/ /ØH-Ø/ /ØH-Ø/ /HØ-Ø/ /HØ-Ø/
3.1 katati wa 3.2 azuki wa 3.4 otoko wa 3.5 kokoro wa 3.6 karasu wa 3.7 kabuto wa
‘shape’ ‘red bean’ ‘man’ ‘heart’ ‘crow’ ‘helmet’
HHH-H LHL-L HHL-L HLL-L LHH-H LHL-L
/ØØØ-Ø/ /LHØ-Ø/ /ØHØ-Ø/ /HØØ-Ø/ /LØØ-Ø/ /LHØ-Ø/
> > > > > >
LHH-H LHH-L LHH-L LHL-L HLL-L HLL-L
/ØØØ-Ø/ /ØØH-Ø/ /ØØH-Ø/ /ØHØ-Ø/ /HØØ-Ø/ /HØØ-Ø/
The issues mentioned above make the developments that follow from the standard theory extremely unlikely. What is most troubling though, is that it is necessary to actively ignore clear evidence that the development was in fact in the exact opposite direction: that the Kyoto type tone systems developed from the Tokyo type, and not the other way around. Or, put differently, that the Kyoto type represents an innovation from the proto-system, which is more directly, or conservatively, reflected in the Tokyo type.
5 Kindaichi (1954, 1955) reconstructed detailed intermediate stages in the development of L register in Kyoto to word-initial H tone in Tokyo. See for example his reconstruction of the developments in classes 2.4 and 2.5: Kyoto type 2.4 LH-H > 2.5 LH-L
Tokyo type LL-H > LL-L > HL-L HL-L LH-L > LL-H > HL-H > HL-L
To prevent the numerous tone classes from merging, these shifts must have progressed in great synchrony with neighboring dialects in all the regions with Tokyo type tone.
199
9 The Ramsey hypothesis
2 Ramsey’s approach Forty years ago, S. Robert Ramsey published three articles in which he argued for a reversed reconstruction of the tone value of the tone dots used in MJ (Ramsey 1979, 1980, 1982). When such a reversed reconstruction is adopted, the tone system of MJ fits in very well with the modern data. It is no longer necessary to posit a process whereby H tones developed in identical locations, as parallel independent developments, in the same nouns and inflected forms throughout Japan. In Ramsey’s reversed reconstruction, these H tones are already present in MJ, in the exact same location as where they are found today in the vast majority of the dialects (namely, in the Tokyo type location in the word). In his reconstruction, the tone system of MJ functions perfectly as a model for the tone system of pJ. Table 3: The tone marks, their interpretations, and the modern Tokyo and Kyoto pitches. Tone class
Gloss
Tone mark
MJ (Ramsey)
MJ (standard)
Tokyo type
Kyoto type
2.1 tori wa 2.2 mura wa 2.3 ike wa 2.4 umi wa 2.5 saru wa
‘bird’ ‘village’ ‘pond’’ ‘sea’ ‘monkey’
上上-上 上平-上 平平-上 平上-上 平上-平
LL-L LH-L HH-L HL-L HL-H
HH-H HL-H LL-H LH-H LH-L
LH-H LH-L LH-L HL-L HL-L
HH-H HL-L HL-L LH-H LH-L
3.1 katati wa 3.2 azuki wa 3.4 otoko wa 3.5 kokoro wa 3.6 karasu wa 3.7 kabuto wa
‘shape’ ‘read bean’ ‘man’ ‘heart’ ‘crow’ ‘helmet’
上上上-上 上上平-上 平平平-上 平平上-上 平上上-上 平上平-上
LLL-L LLH-L HHH-L HHL-L HLL-L HLH-L
HHH-H HHL-L LLL-H LLH-H LHH-H LHL-H
LHH-H LHH-L LHH-L HHL-L HLL-L HLL-L
HHH-H LHL-L HHL-L HLL-L LHH-H LHL-L
(The Tokyo type tone systems in Table 3 are represented by the standard dialect, the Kyoto type tone systems are represented by the dialect of Kōchi. For more details on the marking of tone class 2.5, see footnote 13.)
According to the so-called “dialect circle theory” introduced by Yanagita Kunio (1927), innovations in the Japanese language historically tended to spread out from the cultural center of Kyoto, displacing older forms which were often preserved in the periphery. The geographical distribution of the tone systems in the modern dialects too, suggests strongly that the Kyoto type system, which is surrounded by Tokyo type systems on all sides, is an innovation. Ramsey argued that this innovation consisted of a leftward tone shift that occurred in central Japan after the MJ period. In addition to the dialect geographical argument, Ramsey pointed out certain phenomena in the modern dialects that indicate Kyoto, and not Tokyo, shifted the tones. In the following pages, I will give an overview of the issues involved, starting with a sketch of the type of tone systems in Japan and their geographical spread. I will then
200
Elisabeth M. de Boer
introduce Ramsey’s original ideas, as presented in his articles of 1979 (published in English) and 1980 (published in Japanese, in cooperation with Munemasa Tokugawa and Akiyoshi Kida).6 After this I will introduce the ideas that my own research has added, which include some philological issues.
3 The pitch accent or restricted tone systems of ModJ The ModJ tone systems largely fall into four types. In the most widely distributed type, that of the Tokyo type tone system, which is often taken to be representative of ModJ as a whole, there is a tonal opposition between H tone and Ø tone. This opposition can also be analyzed in terms of pitch accent (where syllables with Ø tone are regarded as unaccented, and syllables with H tone are regarded as accented), and when dealing with the modern stages of (especially) the Tokyo type tone systems, it does not matter very much which terminology is applied. When comparing historical stages however, I prefer an analysis of H vs. Ø tone as used for many Bantu languages, as such an analysis in terms of ‘restricted tone’ makes the comparison with MJ far more straightforward. In my presentation of Ramsey’s original ideas, I will use the term “pitch accent”, as he did in his work.7
3.1 Tokyo type All Tokyo type tone systems belong to the type that has an opposition between H tone and Ø tone. With the exception of some dialects on Noto Island (see Section 5.3), the modern Tokyo type tone systems allow no more than one phonological H tone per word or minor tonal phrase. In the standard language, Ø tones that precede the phonological H tone or accent, have high pitch, except phrase initially, where they are automatically low. Ø tones that follow the phonological H tone have low pitch, so the H tone can be recognized by a drop in pitch. There are also Tokyo type tone systems where there is no automatic rise after the phrase-initial low, and where only the H tone itself is high. What all Tokyo type tone systems have in common though, is that the H tone is followed by a fall in pitch. 6 The last article Ramsey published on the historical reconstruction of Japanese accent (Ramsey 1982) concentrated more on the sociological and historical background that prevents the standard approach from being questioned. 7 Throughout this article the terms “restricted tone” and “pitch accent”, or “the (phonological) H tone” and “the accent” will be used as equivalents, and I will occasionally refer to words that contain a phonologically distinct H tone as “accented” and words that do not contain H tone as “unaccented”.
9 The Ramsey hypothesis
201
There are always exceptions and irregularities, but according to the regular correspondences established by Kindaichi, the Tokyo type tone systems generally agree on the location in the word of this single H tone. The Tokyo type tone system is the most widely distributed tone system in Japan. It is found on all of the four main islands of Honshu, Shikoku, Kyushu and Hokkaido. Although the location of the H tone in the word (if there is one) generally agrees among the Tokyo type tone systems, three sub-types (Nairin, Chūrin, and Gairin) are distinguished, based on the fact that they have a different merger pattern of the tone classes. (The Nairin and Chūrin types are very close and will not be discussed in detail here.8) This different merger pattern is caused by the fact that, compared to the Nairin and Chūrin types, word-final H tone in certain tone classes is missing in the Gairin subtype. The two-syllable tone class 2.2, and the three-syllable tone class 3.2 for instance, which have word-final H tone in the Nairin and Chūrin tone systems, lack H tone in the Gairin tone systems, causing a merger with classes 2.1 and 3.1, which are unaccented in all dialects. As the names of the subtypes indicate, the Nairin (inner circle) type is located most centrally, surrounded by the Chūrin (middle circle) type. The Gairin type is distributed in four widely separated blocks in the periphery, northeast Kyushu, Izumo, part of Shizuoka and the Tōhoku region.
3.2 Word-melody type In the second tonal type, the location of the high tone in the word has become irrelevant. There is no longer a link between a certain tone and a particular location in the word. The distinction is now between two or three different word-melodies that are spread over the phonological word or tonal phrase as a whole. Historically, these systems developed from the Tokyo type, and can be seen as a simplification of it. They can be found in the Ryukyus and in southern Kyushu (Kagoshima) where they developed from the Gairin Tokyo type. (So there, classes 2.1 and 2.2, and classes 3.1 and 3.2 have the same word-melody).9 Oki Island, on the other hand, has a word-melody system that developed from the Chūrin Tokyo type. (So there, these classes have a different word-melody).
8 For the origin of the difference between the Chūrin and the Nairin tone systems (which is limited to tone class 1.2) see De Boer (2010: 82–85). 9 I will limit myself here to the dialects of mainland Japan, excluding the Ryukyus. A discussion of the developments in Ryukyuan can be found in De Boer (2010: 206–246) and De Boer (2017a).
202
Elisabeth M. de Boer
3.3 Kyoto type The third tonal type, the Kyoto type, has a phonological L tone, in addition to the H tone and Ø tone that are also present in the Tokyo type tone systems. Unlike the two other tones, which can occur anywhere in the word, this L tone is restricted to word-initial position, and is often referred to as “low register”. As in Tokyo, the H tone in Kyoto is followed by a fall in pitch, but the location of the H tone is one syllable earlier in the word than in the Tokyo type tone systems. When a word has initial H tone in Tokyo, such a correspondence is impossible. In that case, the correspondence is between initial H tone in Tokyo, and initial L tone in Kyoto. The Kyoto type tone systems are found in central Honshu, the largest part of Shikoku, and in a simplified form on Sado Island, where the dialect was strongly influenced by Kyoto.
3.4 Dialects without lexical tone There are also dialects where tone is not lexically distinctive, such as in the southern Tōhoku region, in Shikoku between the area with Tokyo type tone (west Shikoku) and Kyoto type tone (east Shikoku), and in central Kyushu between the area with wordmelodies (southwest) and the area with Tokyo type tone (northeast). As such dialects are mostly found in the border areas between two different tone systems, confusion most likely played a role in the loss of tonal distinctions.
4 The development of Ramsey’s ideas In his dissertation, and his later book (Ramsey 1978), Ramsey had reconstructed the history of pitch accent in Korean, based on a comparison of the accent systems of South Hamkyeng and Kyengsang with the tone dot markings in Middle Korean (MK). One of his conclusions was, that the opposition between the accent systems of South Hamkyeng and Kyengsang had developed because a leftward accent shift had taken place in Kyengsang. As in MK, the pitches of MJ (referred to as the “Old Kyoto dialect” in Ramsey’s articles) had been marked by means of tone dots (albeit with a very different system). When he became interested in the historical development of pitch accent in Japanese, the central distribution of the Kyoto type, and the fact that it had the accent one syllable earlier than in the surrounding Tokyo type tone systems, suggested to him that a leftward shift similar to the one in Kyengsang had taken place in Kyoto. This made the Kyoto type system an innovation, which was in direct opposition to the idea that was common in Japan then, and still now, that the Kyoto system is conservative, while the Tokyo system is the result of numerous innovations.
9 The Ramsey hypothesis
203
In his first article on the subject (Ramsey 1979), he compared the tones of MJ, which he transcribed as 上 = R (rising) and 平 = E (even), with the accent systems of the modern dialects, and noted that there was only one regular correspondence between MJ and any of the modern dialects: A transition from E to R in MJ, corresponded to a drop in pitch in the Tokyo type tone systems. Because a transition from H to L is distinctive, not only in all Tokyo type tone systems, but also in the Kyoto type tone systems, Ramsey concluded that HL transition must certainly have been distinctive in the ancestral system common to both. Combined with the fact that the location of an ER sequence in MJ regularly corresponded to a HL sequence in Tokyo, he concluded that E (“even”) in MJ must have had the value H, and R (“rising”) must have had the value L. So, even though the tone dot material stemmed for a large part from the old capital of Kyoto, the tone system expressed by the dots had been Tokyo-like as to the location in the word of a distinctive fall in pitch. The present-day location of the pitch fall in the Kyoto type tone systems, which is earlier in the word, had to be an innovation dating from a later period, just as the dialect geography suggested. Not only is the Kyoto type accent system surrounded by the Tokyo type, even within the Kyoto type area, there are a number of isolated rural villages in the center of the Kii peninsula which have preserved the Tokyo type location of the fall (Ramsey 1979: 172). The comparison in Table 4, between the MJ tones in the standard reconstruction and the tones of modern Kōchi, shows that (be it in nouns or verbs), the resemblances between the standard reconstruction and the pitches of the modern Kyoto type are shallow.10 Table 4: Comparison of the standard reconstruction of MJ with Kōchi. Gloss
The correspondence between Ramsey’s reconstruction and Tokyo, in Table 5 on the other hand, is completely regular. Please notice how the problem of the lacking H tones in the standard reconstruction of the MJ tone system is not there in Ramsey’s version. The H tone in hikaru moreover (which has the same tone as nouns of class 3.5) is in the
10 The representation in terms of restricted tone (H vs. Ø) in Tables 4 and 5 has been added by me.
204
Elisabeth M. de Boer
expected location in the word. (As we saw in Table 1, the correspondences of class 3.5 are problematic in the standard theory.) Table 5: Comparison of Ramsey’s reconstruction of MJ with Tokyo. Gloss
4.1 Ramsey’s arguments for a leftward shift in Kyoto, based on the tone of the particle no In MJ, most monosyllabic particles attached to nouns with “rising” tone (L in Ramsey’s reconstruction), but the particle no was different; in phrases it continued the pitch of the preceding syllable. ‘Man’ for instance, followed by wa would be otoko wa EEE-R (HHH-L in Ramsey’s reconstruction), but followed by the particle no, it would be otoko no EEE-E (HHH-H in Ramsey’s reconstruction). Likewise, ‘village’ would be mura wa RE-R (LH-L) but mura no RE-E (LH-H) and ‘pond’ would be ike wa EE-R (HH-L) but ike no EE-E (HH-H). In short, in all tone classes that ended in H tone, there was no drop to low pitch after the noun if no was attached. This is still the case in Tokyo (Ramsey 1979: 170), as can be seen in Table 6: Table 6: Effect of the particle no in Tokyo. Gloss
noun + wa
‘pond’ ‘village’ ‘man’
iké wa murá wa otokó wa
noun + no LH-L LH-L LHH-L
ike no mura no otoko no
LH-H LH-H LHH-H
A comparison with the Kyoto type dialects (see Table 7) shows that in these dialects the same tone classes cancel the accent, but here the accent does not occur word-finally, but on the penult (Ramsey 1979: 170).11 11 Wakayama data from Ramsey’s fieldwork represent the Kyoto type tone systems, as all example words in Table 6 have initial accent in Kyoto because of the secondary leftward shift.
9 The Ramsey hypothesis
205
Table 7: Effect of the particle no in Wakayama. Gloss
noun + wa
‘pond’ ‘village’ ‘man’
íke wa múra wa otóko wa
noun + no HL-L HL-L HHL-L
ike no mura no otoko no
HH-H HH-H HHH-H
If there had been no MJ data, the conclusion would have probably been that – for some unknown reason – no simply deletes the accent on a preceding noun (no matter where this accent is located). But there is MJ data available, which makes a far more detailed historical reconstruction possible. Thanks to this data, the accent loss in Tokyo can be connected to the unique effect that no had in MJ, in that it continued the pitch of the final syllable of a preceding noun. The accent loss in Kyoto on the other hand, can only be explained and connected to what we know were the qualities of no in MJ, if one assumes that the accent in Kyoto was on the final syllable too, in earlier times, and only later shifted to the penult.
4.2 Ramsey’s arguments for a leftward shift in Kyoto, based on compound nouns In both Tokyo and Kyoto, when a noun is used as the last member of a compound, it can have a pitch shape different from the one it has in isolation. As such morphological variations typically develop following phonological changes, this is an indication that some historical change took place in the tones. In compounds, the pitches in both dialects often agree with each other, even though the pitches of nouns used in isolation are completely different in the two dialects. This phenomenon (see Table 8) was first pointed out by Wada (1943).12 Table 8: Wada’s comparison of compound nouns in Kyoto and Tokyo. Tone class of 2nd element
Gloss
Compound
Kyoto dialect
Tokyo dialect
2.1 2.2 2.3 2.4 2.5
Kobe cow granite Japanese dog conical hat Shikoku monkey
12 When the first element starts with L register in Kyoto, the compound stars with L register too, but the location of the accent in the compound will be the same as in the examples in the table.
206
Elisabeth M. de Boer
Both dialects have largely preserved the distinction between classes 2.2 and 2.3 in compounds, even though these two classes have merged in both dialects in isolation. (See Table 2.) Because archaic distinctions have been preserved, and because of the agreement in the pitches in both dialects, Ramsey argued that the rules governing the tone of compound nouns must be archaic, and must have predated the split between the Kyoto type and the Tokyo type accent systems (Ramsey 1980: 71). What is remarkable is that in both dialects the pitches of the second element overall agree with Ramsey’s reconstruction of MJ. (The fact that class 2.2 is level low, and not distinguished from class 2.1 in compounds, is most likely due to the modern rule that after the drop to low that follows the accent, the pitch will not rise again within the same tonal phrase.) Most telling is the fact that in Tokyo the location of the accent remains the same, whether the words occur in compounds or in isolation, while in Kyoto the two differ. We know that one of the two dialects went through a shift, and this is clear evidence that the shift took place in Kyoto and not in Tokyo. The shift in Kyoto was accompanied by the phonetic change of unaccented words to uniformly high-pitched shapes except in compounds, where they remained low (see class 2.1 in Table 9). Table 9: Second elements in compounds preserved the pitches of MJ. MJ (Ramsey)
Although some members of tone class 2.5 had been famously marked with the “light even” tone dot on the second syllable in a small number of MJ tone dot materials, Ramsey did not include the distinction between classes 2.4 and 2.5 in his theory.13 It was thought at the time that this class was only distinguished in the Kyoto type tone systems, and Ramsey’s friend Tokugawa (1962) had argued that it was most likely an innovation that did not go back to pJ. Since then however, it has been recognized that the class left some traces in the Tokyo type dialects of the Izumo region (Okumura 1981), far removed from the Kyoto area, which indicates that the distinction does go back to pJ.
13 Use of the “light even” tone is rare. It fell out of use in the 12th century, and most materials have “rising” tone dot markings in those locations where the “light even” tone has been attested. In later materials the presence of the contour tone can be detected through the influence it had on the tone of attached particles. (See class 2.5 in Table 1.) In Ramsey’s reconstruction class 2.5 would have been HR-L or HL-H, the opposite of the tones in Table 1.
9 The Ramsey hypothesis
207
4.3 Differences in the tones of 15th century dialects, as recorded in the musical notation system of Nō At the end of his Japanese article (Ramsey 1980: 75), Ramsey quoted a description of the pitches of the word ‘dog’ inu (tone class 2.3) in several Japanese dialects in a work on Nō (Mōtan shichin-shō, 1455) by Zenpō Konparu. The pitches are given by means of the so-called goma-ten “sesame-dot” notation system. In miyako-goe “the language of the capital”, inu was (HL), which agrees with modern Kyoto. In Tsukushi (northeast Kyushu) and Bandō (the provinces east of the Osaka barrier in Ōmi), which are both areas with a Tokyo type tone system today, they were (LH), which agrees with modern Tokyo. In Shikoku however, they were (HH), which agrees with Ramsey’s reconstruction of this tone class as HH in MJ. It is clear from these markings that the difference between the Tokyo type and the Kyoto type tone systems had already developed by 1455, and Ramsey argued that the preservation of HH tone in Shikoku meant that at that time, the leftward tone shift had not yet reached Shikoku.14 Ramsey regarded the leftward shift in Kyoto as the cause of the merger between classes 2.2 (LH-L in MJ) and 2.3 (HH-L in MJ): a leftward shift would automatically have resulted in merger between these two classes as HL-L in Kyoto.15 I think, however, that in most dialects, this merger had a different cause. It is, after all, not limited to the Kyoto type tone systems, but also occurred in the Tokyo type Nairin and Chūrin tone systems. I see the merger as related to a sound change that affected all dialects in Japan. This change, which I call “the H tone reduction”, explains how the rich tone system of MJ was transformed into the modern Tokyo type tone systems, which are so restricted that they can just as well be analyzed as pitch accent systems.
5 Further application of Ramsey’s theory The following sections will introduce my ideas on the developments from the MJ tone system in Ramsey’s reconstruction to the modern dialects. I will first discuss the three Tokyo type tone systems, and then the Kyoto type tone systems. The ideas in the following sections, although based on Ramsey’s theory, are my own, and go beyond Ramsey’s original ideas as outlined in the previous sections.
14 Most of Shikoku has a Kyoto type tone system nowadays, but as there is an area where classes 2.1 and 2.3 have merged, there is the possibility that the level H pitch reported by Zenpō referred to his dialect type. 15 There is in fact one dialect (Ibukijima) that went through the leftward shift and still kept these two classes apart. See footnote 16.
208
Elisabeth M. de Boer
5.1 How did the tone system of MJ develop into the modern Tokyo type pitch accent systems? When Ramsey’s reconstruction of the MJ tones is adopted, the correspondence between MJ and the modern Tokyo type tone systems is such that, whenever there was a H tone followed by a L tone in MJ, the modern dialects have the single phonological H tone in the word in that location. This is the core of Ramsey’s theory, and it explains the predominance and wide distribution of the Tokyo type tone systems in Japan. There is, nevertheless, an enormous difference between the tone system of MJ and that of the modern Tokyo type dialects. The modern dialects are much simpler. In MJ, each syllable was marked for either H or L tone, and so the location of a rise from L to H was as distinctive as the location of a fall from H to L. In the modern dialects, only a fall from H to L is distinctive, and there is at most one H tone per word, all other tones are Ø. How and why did the system change in such a way that only a single phonological H tone per word remained, and why did only a transition from H to L remain distinctive? It is clear that the two changes must be related. An answer to these questions can be given by taking into account tonal universals. One well-established tonal universal is that a high-low interval is subject to polarization, or exaggeration. A high tone will often be significantly greater in height when followed by low tone. (A low-high interval, on the other hand, is subject to suppression, and has the opposite tendency to level out.) High tones followed by low tone, can even be raised to a higher phonological level so that a three-level system distinguishing H, M, and L develops (Hyman 1993: 75–108, Hyman 2007: 3).16 Eventually, in Japan, only the single H tone remained as a phonological tone in the modern dialects, in opposition to Ø tones, whose pitches are determined by automatic rules. The well-attested tendency for polarization of a high-low interval offers an explanation for the tonal change that must have occurred most widely in Japan namely the change from MJ to the restricted tone systems of the Tokyo type dialects, the most widely distributed tonal type in Japan. It also explains why it was the drop in pitch that became the single distinctive feature in the modern dialects, and not the rise, even though both had been distinctive in MJ. In Table 10, the gradual process of H tone loss is shown for tone classes 2.3 (ike ‘pond’), 3.4 (otoko ‘man’), and 3.5 (kokoro ‘heart’) which all contained sequences of H tone in MJ.
16 The distinctions in the dialect of Ibukijima, where tone classes 2.3 and 3.4 contain mid (M) tones, can be explained if the leftward tone shift reached this dialect when it was still at this stage (stage 2 in Table 10). Class 2.2, which was LH, shifted to HL, but class 2.3, which was MH, shifted to HM.
9 The Ramsey hypothesis
209
Table 10: Process whereby only H followed by L in MJ was preserved in the modern dialects. Tone class
Form
MJ (Ramsey)
2.3 3.4 3.5
ike wa otoko wa kokoro wa
HH-L HHH-L HHL-L
Stage 2 > > >
MH-L MMH-L MHL-L
Stage 3 > > >
LH-L LLH-L LHL-L
Modern Tokyo > > >
LH-L LHH-L LHL-L
/ØH-Ø/ /ØØH-Ø/ /ØHØ-Ø/
As result of the H tone reduction, certain tone classes merged in Tokyo. In MJ classes 2.2 LH-L and 2.3 HH-L had been distinct, but the H tone reduction caused a merger as ØH-Ø. Likewise, classes 3.2 LLH-L and 3.4 HHH-L had been distinct in MJ, and these classes merged as ØØH-Ø in Tokyo. In the Gairin tone systems though, the merger patterns are different.
5.2 Origin of the variety of tone class mergers in the Tokyo type tone systems A large part of the tone dot material for MJ originated in Kyoto, the capital of Japan at the time, and Ramsey therefore spoke of the “Old Kyoto dialect”. There is good reason to assume though, that not all tone dot material reflects the tone system of the capital. In certain manuscripts, the tone of the particles wa, ga, o, ni (typically low in most materials) is high when following words that end in a low-high sequence, such as tone classes 2.2 LH and 3.2 LLH. (This is different from the behavior of the particle no, which was high after all words ending in H tone, so also after classes 2.3 HH and 3.4 HHH.) The assimilation is seen sometimes in the Date-ke-bon of Kokin waka-shū (1226) and is standard in Kokin kunten-shō (1305). In other materials from the same period or later though, such as the Kamakura-bon of Nihon shoki (1303) and the Maeda-bon of Shūi waka-shū (1333), the assimilation does not occur, and the tone of these particles stays low. The difference, therefore, appears to be dialectal. The manuscripts where the particles attached with L tone, being most numerous, most likely represent standard Kyoto speech, while the manuscripts that show assimilation most likely represent some dialectal variant. The particle no attached to nouns with low or high tone, depending on the tone of the last syllable of the word in MJ. After words of class 2.2 LH, 2.3 HH, 3.2 LLH, and 3.4 HHH it attached high, and we have seen that in the modern dialects, this resulted in the loss of the word-final H tone before the particle no. The assimilation in variants of MJ resulted in high pitch on the particles wa, ga, o, ni, after tone classes 2.2 and 3.2. This, too, resulted in accent loss in the modern dialects. As in these dialects all particles attached with high tone after classes 2.2 and 3.2 (not just the particle no), the accent loss was complete, and tone classes 2.2 and 3.2 merged with the
210
Elisabeth M. de Boer
unaccented classes 2.1 and 3.1 at the time of the H tone reduction. This merger pattern is typical of the Gairin tone systems, which I see as the descendants of varieties of MJ that must have had tone assimilations like those attested in Kokin kunten-shō and similar materials. The Gairin type tone system closest to the centers of power in central Japan is in Shizuoka, the old province of Mikawa. It may be the tone system of this area that is reflected in some of these materials. It could also be some variety from a different location that was later supplanted. Table 11: Differences in assimilation in MJ result in different mergers in the modern dialects. Tone class
Gloss
MJ (Ramsey)
Tokyo
Oita
2.1 tori wa 2.2 mura wa 2.3 ike wa 2.4 umi wa
‘bird’ ‘village’ ‘pond’ ‘sea’
LL-L LH-L or LH-H HH-L HL-L
LH-H LH-L LH-L HL-L
ØØ-Ø ØH-Ø ØH-Ø HØ-Ø
LH-H LH-H LH-L HL-L
ØØ-Ø ØØ-Ø ØH-Ø HØ-Ø
2.5 saru wa
‘monkey’
HR-L or HL-H
HL-L
HØ-Ø
HL-L
HØ-Ø
3.1 katati wa 3.2 azuki wa 3.4 otoko wa 3.5 kokoro wa 3.6 karasu wa 3.7 kabuto wa
‘shape’ ‘bean’ ‘man’ ‘heart’ ‘crow’ ‘helmet’
LLL-L LLH-L or LLH-H HHH-L HHL-L HLL-L HLH-L or HLH-H
LHH-H LHH-L LHH-L LHL-L HLL-L HLL-L
ØØØ-Ø ØØH-Ø ØØH-Ø ØHØ-Ø HØØ-Ø HØØ-Ø
LHH-H LHH-H LHH-L LHH-L HLL-L HLL-L
ØØØ-Ø ØØØ-Ø ØØH-Ø ØHØ-Ø HØØ-Ø HØØ-Ø
The HL-H and HLH contours of classes 2.5 and 3.7 MJ were simplified to HL and HLL in (almost) all Tokyo type tone systems, where pitch – once it has fallen – will not rise again within the same word or minor tonal phrase. In these tone classes therefore, differences in the assimilation of particle tone did not result in different merger patterns, and classes 2.5 and 3.7 usually merged with classes 2.4 and 3.6.
5.3 The Tokyo type tone systems that stayed closest to MJ There is an area in Japan where the tone system stayed closer to MJ than in most other dialects. In a number of dialects on Noto Island, such as in the dialect of Nozaki, tone classes 2.5 and 3.7 have preserved the HL-H and HLH contour of these classes in Ramsey’s MJ (see Table 12). Adjectives of class B had HLH tone in the infinitive in MJ, and in Nozaki this is also still the case: The form takaku ‘high’ for instance is HLH (Kindaichi 1954).17 17 The dialect also has some innovative features. Like many dialects in Ishikawa prefecture, syllables that contain a close vowel and a voiced consonant avoid H tone, shifting the H tone to the left. According to Uwano and Nitta (1985) this innovation in Ishikawa prefecture is of relatively recent date.
9 The Ramsey hypothesis
211
Table 12: The Tokyo type tone systems that stayed closest to MJ. Tone class
Gloss
MJ (Ramsey)
Nozaki (Kindaichi 1954)
2.4 kasa wa 2.4 kasa mo 2.5 mado wa 3.6 usagi 3.7 kabuto B takaku
‘umbrella’ id. + mo ‘window’ ‘rabbit’ ‘helmet’ ‘high’
HL-L HL-H HL-H HLL HLH HLH
HL-L HL-H HL-H HLL HLH HLH
/HØ-Ø/ /HØ-H/ /HR-Ø/ /HØØ/ /HØH/ /HØH/
Although most monosyllabic particles had low tone in MJ, no was not the only particle with a more complicated tone. The particle mo ‘too’ for instance, was L (or in some materials R) following H tone, but H following L tone.18 Class 2.4 + mo for instance, had HL-H tone in MJ, similar to the tone of a noun of class 2.5 + wa, ga, o, ni etc. In case of mo, the H tone was inherent in the particle. In case of class 2.5 + wa, ga, o, ni, etc, the particle was high pitched because of the presence of a phonological rising tone on the final syllable of class 2.5. Because HL-H is still allowed in Nozaki, this difference in tone between mo and the other monosyllabic particles was preserved. Class 2.4 + mo such as kasa mo is HL-H in Nozaki, just as mado wa (class 2.5 + ga), but kasa wa is HL-L. There is, in short, a detailed agreement with the tones of MJ in Ramsey’s reconstruction, which in the standard theory must be put down to coincidence.
5.4 Ramsey’s MJ and the modern Kyoto type tone systems In Section 1, I have shown that there are serious problems in deriving the modern Kyoto type tone systems from MJ in the standard reconstruction. H tones that are present in the modern Kyoto type tone systems are not there in MJ, or in the wrong location. A plausible cause for their development is lacking.19 For unknown reasons too, tone class 3.2 developed L register in the Kyoto type tone systems, causing a merger with class 3.7.
18 This is an indication that the tone of the particle mo was originally R (as is confirmed by some rare attestations with the “light even” tone). The assimilation and loss of contour tones is determined for a large part by the tones by which they are preceded or followed (Hyman 2007). LR will readily develop into LH due to regressive assimilation with the preceding like tone. HR on the other hand, will preserve the R contour for a longer time, because the unlike tone preceding it blocks regressive assimilation. 19 Kawakami (1965) suggested that L tone before H tone could have become progressively lower, creating H tones just before the L tone in the process. For this explanation to be realistic, such a type of development must have been exceedingly natural, as H tones are present in these words all over Japan without skipping a single dialect. Kawakami’s idea is, however, in direct contradiction to the tonal universal that LH intervals tend to level out, not increase. See De Boer (2017b: 222–223).
212
Elisabeth M. de Boer
This makes it impossible to derive the Tokyo type tone systems, where such a merger did not occur, from the Kyoto type.20 If instead, Ramsey’s MJ is taken as starting point of the developments, it can be seen that, just as in the Tokyo type tone systems, all H tones that were followed by L tone in MJ were preserved but shifted one syllable to the left (see Table 13). A comparison of class 3.2 with 3.4 now reveals why class 3.2 has L register in the modern Kyoto type tone systems: In both tone classes the pre-L H tone shifted to the left, but in class 3.4 it was preceded by H tone in MJ, whereas in class 3.2 it was preceded by L tone. Table 13: H followed by L in MJ is shifted to the left in Kōchi. Tone class
Gloss
MJ
2.2 mura wa 2.3 ike wa 3.2 azuki wa 3.4 otoko wa 3.5 kokoro wa
‘village’ ‘pond’ ‘red bean’ ‘man’ ‘heart’
LH-L HH-L LLH-L HHH-L HHL-L
Kōchi > > > > >
HL-L HL-L LHL-L HHL-L HLL-L
/HØ-Ø/ /HØ-Ø/ /LHØ-Ø/ /ØHØ-Ø/ /HØØ-Ø/
There are numerous examples of leftward shift of the H tone in Bantu languages, but these languages are rich in prefixes, so that word-initial H tone, when shifted, will be realized on the prefix. In Japanese, prefixes are rare, and the leftward tone shift meant that word-initial H tone was shifted off the word. The following L tone would land on the initial syllable, resulting in L register in the modern Kyoto type tone systems. The modern dialects regularly have L register whenever MJ had L tone on the second syllable (see Table 14). In class 3.7, the word-initial H tone shifted off the word, and the word-final H tone shifted to the penultimate syllable, resulting in a merger with class 3.2. Table 14: Word-initial H followed by L in MJ is shifted to the left in Kōchi, resulting in L register. Tone class
Gloss
MJ
2.4 umi wa 2.5 saru wa 3.6 karasu wa 3.7 kabuto wa
‘sea’ ‘monkey’ ‘crow’ ‘helmet’
HL-L HL-H HLL-L HLH-L
Kōchi > > > >
LH-H LH-L LHH-H LHL-L
/LØ-Ø/ /LH-Ø/ /LØØ-Ø/ /LHØ-Ø/
20 If the idea is that the low register in class 3.2 developed independently in all the Kyoto type tone systems, after the development of the Tokyo type tone systems, it is unclear why a similar low register failed to develop in other tone classes that started with a string of H tones, such as class 3.4 or class 2.3.
9 The Ramsey hypothesis
213
The leftward shift resulted in a reversal of the tone of the initial syllable in words used in isolation in many tone classes, but (as we have seen) in Kyoto compounds, classes 2.1 and 2.2 still attach with low pitch, while classes 2.3, 2.4, and 2.5 still attach with high pitch; a remnant from the MJ stage. Finally, in MJ and in the Tokyo type tone systems, H tone can fall on any syllable in the word, but in Kyoto word-final H tone is rare in words of one and two syllables, and completely absent in longer words.21 Ramsey’s proposed leftward shift explains the origin of this gap in the system.
6 Philological issues Ramsey’s ideas have not been properly addressed or taken seriously in Japan. This rejection is based on the idea that the old tone descriptions, such as those studied by Kindaichi, are unequivocal in indicating H as the value for the “rising” tone, and L as the value of the “even” tone. This is definitely the way in which these tones were viewed in Shingon Buddhist circles from the 17th century on, as is clear from tone descriptions from that period, but this view is a late development. The tone descriptions contemporary with the production of the tone dot material are very different, and what they describe are not the tones of MJ, but the tones of LMC as they were regarded in Japan at that time.
6.1 Buddhist tone theories The standard reconstruction of the value of the tone dots agrees in every detail with Shingon Buddhist tone theories from the 17th century, where the “even” tone is described as hikusi ‘low’, for instance, and the “rising” tone as takasi ‘high’. The interpretation of the musical notation system in use in that period is clear too. The Buddhist chanting traditions however, all went through a period of collapse in the 15th century, only to be revived in the 17th century.22 As the 17th century theories and practices do not go back in an uninterrupted line to the period contemporary with the tone dot annotations, they cannot be used as evidence for the value of the tones in that period.23 21 In shorter words it developed from the word-final R contour tone in classes 1.2 and 2.5 in MJ. 22 There are tone descriptions from the 15th century in which the tones are not even described in terms of tone height, just in terms of differences in length (De Boer 2010: 500–501). 23 The similarity between the modern standard view and the ideas of the Edo period Buddhist scholars stems from the fact that both schools of thought had the same starting assumption, namely that the tone system reflected in the old tone dot materials must have phonetically resembled that of post-shift Kyoto. This assumption inevitably leads to the values given to the dots in the standard theory.
214
Elisabeth M. de Boer
Buddhist scholars from the esoteric Shingon and Tendai schools had a strong interest in the Chinese tones, as they aimed at a correct pronunciation of the Sanskrit mantra and dhāraṇī, which had been transcribed by means Chinese characters. They inferred the correct value of the Chinese tones from a prestigious 9th century description by the monk Annen, but Annen’s text left a lot of room for interpretation. (See De Boer 2010: 371–391). Annen had described the “even” tone with the Chinese character 低. Nowadays, this character is read as hikui ‘low’, but at the time contemporary with the production of the tone dot material, it was read as taru ‘to droop’. This is for instance the reading of this character in the Ruiju myōgishō dictionary. A reading as hikisi or hikusi ‘low’ does not occur. The “rising” tone had been described by Annen with the character 昂 which, according to Ruiju myōgishō, was read as agaru ‘to go up’. Again, a reading as takasi ‘high’ does not occur. The characters 低 and 昂 also occur together in Ruiju myōgishō as 低昂, for which the reading tari-agaru ‘to fall and rise’ is given. Okurigana added to the characters in the Buddhist tone descriptions confirm that they were read as verbs, indicating tonal movement, and not as adjectives, indicating a stative tone level (De Boer 2010: 400–440). Apart from the verb taru, the verb sagaru ‘to do down’ is also used to describe the “even” tone.24 The description of the “going” tone in Annen’s text was vague (most likely because it still involved breathy voice quality in LMC at the time), but it was mentioned that it was lengthened, a statement that is often repeated in the later Buddhist tone descriptions. In 13th century Shingon tone descriptions, such as by Shinpan and Ryōson (De Boer 2010: 427, 429), it is described with the character 偃, which has the reading fusu ‘to bow down, bend one’s head, stoop’ (as well as a number of related readings) in the Ruiju myōgishō dictionary. As the Kanchiin-bon of the Ruiju myōgishō dictionary, which is itself the richest source of MJ tone dot markings, stems from the same period and the same school, these tone descriptions have a direct bearing on how to interpret the value of the tone dots. The difference between the Chinese “even” tone and the Chinese “going” tone was apparently seen as one of length. The idea that the “going” tone was long, agrees with the fact that it consisted of a sequence of an “even” tone followed by “rising” tone on one syllable, as was pointed out by Kindaichi.25 Because Kindaichi interpreted the “even” 24 Some of the okurigana indicate that the verbs taru, sagaru, and agaru that were used to describe the “even” and the “rising” tones were read in the non-past form, and others that they were read in the imperfective/perfective form (agareru, tareru, sagareru). Different versions of the same text can vary in this respect, and both types of okurigana can also co-occur in the same text. The difference is not unimportant, as the imperfective/perfective can be read as “lowered” or “raised”, which would bring the old descriptions into agreement with the 17th century interpretation of the value of the tone dots. What is required therefore, is a thorough study and inventory of these materials, whereby different versions are compared with each other, in order to find out whether different types of annotations were added in different periods. 25 Kindaichi (1951: 646–648) showed that one kana readings of “going” tone characters in the Zushoryō-bon of Ruiju myōgishō (1081) were marked with a “going” tone dot, but that two kana readings of
9 The Ramsey hypothesis
215
tone as L and the “rising” tone as H, he translated fusu (‘bending down’) in the Shingon tone descriptions as ‘rising up’ (Kindaichi 1951: 691), and so did Mabuchi for the same reason (Mabuchi 1962: 437). If, however, the translation of a character that unambiguously describes the “going” tone as ‘bending down’ is accepted as is, it confirms Ramsey’s idea that the “even” tone was used to mark MJ H tone, and that the “rising” tone was used to mark MJ L tone. The abundance of contour tones in these tone theories, and their highly theoretical nature may be surprising, but a number of things have to be kept in mind: The texts do not describe the tone system of MJ, nor what kind of pitch the tone dots marked if they were applied to MJ. They described the tones of an idealized Chinese that was appropriate for the purpose of chanting magical Sanskrit formulae in a ritual context (De Boer 2008: 81–83). A Chinese moreover, that was not based on any knowledge of the spoken language, but on (mis)interpretations of an authoritative text from centuries earlier.26 It is possible to surmise how these Chinese contour tones would most likely have been used when employed to mark the mostly level tones of MJ. The auditory impression that a falling tone makes is high, and the auditory impression that a rising tone makes is low, as a falling contour stresses the high onset of the tone, and a rising contour stresses the low onset of the tone. Open vowels, for instance, are intrinsically lower than close vowels, but this difference is hardly ever phonologized. The lowering caused by voiced consonants on the other hand, frequently is. This is because the voiced consonant causes a rising tone contour on a following syllable, which is far more salient in indicating low pitch than a level low tone (Hombert 1977). It makes sense therefore, to assume that the “even” tone was used to mark MJ H tone, the “rising” tone was used to mark MJ L tone, and the long “going” tone was used to mark MJ F tone. (Although the “even” tone had a “light” variant which in some materials was used to mark a rising tone contour, in most materials the “rising” tone dot did double service, marking L tone, as well as the occasional R tone.) In short, the idea that descriptions of the tones contemporary with the production of the tone dot material agree with the value of the tones in the standard reconstruction is wrong: The descriptions do not agree with either of the reconstructions, as they such characters were marked with an “even” tone dot added to the first kana, and a “rising” tone dot added to the second kana. The Zusho-ryō-bon furthermore has nu 去, but also nuu 平上 for ‘marsh’, goma 去上, but also ugoma 平上上 for ‘sesame’, and hagi 去平, but also haagi 平上平 for ‘shank’. Shinsen jikyō (892) has hii 平上 for ‘shuttle’, which is also attested as hi 去 in Ruiju myōgishō. 26 Theoretically, the tones of the Buddhist scholars were even more complicated, as each tone was supposed to have “light” and a “heavy” variety, an idea that goes back to interpretations of Annen’s text. As Annen had described “light” with the same character (昂) that he had used to describe the “rising” tone, and “heavy” with the same character (低) that he had used for the “even” tone, “light” was regarded as “rising” and “heavy” was regarded as “falling” by later generations of scholars. In practice, fewer tones were used. The “rising” tone was regarded as intrinsically “light” (i. e. rising), and the “going” tone was regarded as intrinsically “heavy” (i. e. falling), so the “heavy rising” and “light going” varieties were discarded.
216
Elisabeth M. de Boer
described a special tone system that was appropriate for the chanting of magical formulae. Trying to reconstruct how these tones would have been used to mark the level tones of MJ though, it is easier to reconcile them with Ramsey’s MJ, especially if the description of the “going” tone as ‘bending down’ is taken into account.
6.2 Musical notation systems that developed from the tone dots There are two main types of Buddhist musical notation systems (called fushi-hakase or “teachers of melody”). The oldest type developed as extensions of the tone dots, while a later type, the goin-hakase system (“pentatonic teachers”), was devised as a true musical notation system, indicating absolute tone. (It is this later type that was used in the 17th century Shingon school.) In the earlier system, a horizontal line was added to the “even” tone, and a diagonal line was added to the “rising” tone, which were both on the left-hand side of the character. The “going” tone, which was on the right-hand side, was marked in different ways: a hook, a z-shape, or a diagonal stroke in the opposite direction of the “rising” tone. Later the strokes were used without the dots. This system, called goma-fu (“sesame notation” as the shape of the marks resemble sesame seeds) was in use in the Shingon school until late into the 14th century for genres where the recitation stayed close to the spoken language. The system was adopted in court music (gagaku), too, where it was also called goma-fu. In banquet songs for instance, a word like 2.3 toki ‘time’ was marked with two horizontal hakase marks , 2.2 hito ‘person’ was marked and 2.5 mayu ‘eyebrow’ was marked (Kindaichi 1974). Musical genres such as Nō and Heikyoku (the recitation of the Tale of the Heike) also use notation systems that are regarded as offshoots of this Buddhist fushi-hakase system. The musical notation of Nō music developed in the late 14th century, and the shapes and the names of the marks (goma-ten “sesame dots”), as well as the musical terminology, all have a very close resemblance to the goma-fu used in court music and Buddhists recitation. Adopted in Heikyoku, the Buddhist vocal notation system was called sumi-fu (“ink notation”).27 In Buddhist circles and court music the older notation system that had developed from the tone dots died out, so the value given to the marks according to the standard theory is a modern reconstruction. The notation systems of Nō and Heikyoku on the other hand, survived into the modern period.
27 In Nō and Heikyoku the marks are added to the right-hand side of the text, while in court music and Buddhist music the left-hand side is usual. There are, however, examples of right-hand side markings in Buddhist texts too, such as in the Daijiin-bon of Shiza Kōshiki.
9 The Ramsey hypothesis
217
in Nō music is H and As seen in Section 4.3, the value of the horizontal mark the diagonal mark is low. A less frequently used third goma-ten, a diagonal “forward slash” poised in the opposite direction of the mark expresses extra-H tone. In Heikyoku recitation (Ueno et al. 1998: 45–46), the mark again expresses L, and the marks and both express H. The word 2.4 kasa ‘umbrella’ LH in modern Kyoto for instance, occurs marked as , but also as . The word 2.2 hasi ‘bridge’, HL in modern Kyoto, occurs marked as , but also as . Sequences of H pitch are always indicated by means of the horizontal mark ( etc.). Only when the horizontal mark is combined with the “forward slash” mark, can it express L pitch, but such cases are rare. I see this as an indication that (just as is still the case in Nō) the original value of the mark was extra-H. In the standard theory, the horizontal goma-fu mark, having developed from the “even” tone, is given the value L and the diagonal goma-fu mark, having developed from the “rising” tone, is given the value H. This is the opposite of the value that these marks have in the Nō and Heikyoku notation systems, which developed from this same goma-fu. I see this as yet another piece of independent evidence that Ramsey’s reconstruction of the value H for the “even” tone and L for the “rising” tone is correct. The less frequently used “forward slash” mark for the extra-H tone must go back to the mark of the “going” tone. The falling contour of this tone (involving a high-low interval) is, after all, most likely to have developed an extra-H tone level.
7 Conclusion Ramsey’s MJ tone system must have been close to the tone system of pJ, as all mainland ModJ tone systems can be derived from it in a straightforward way. That does not mean the diversification of the Japanese tone systems only started after the MJ period. Merely, that the attested tone system of MJ happened to be conservative and had remained relatively close to the proto-system. One of the earliest changes to the proto-system that can be reconstructed was assimilation of the tone of monosyllabic particles after words ending in LH tone in precursors of the Gairin tone systems. Later, due to the polarizing effect of a HL interval, a tone reduction occurred, whereby only H tones that were followed by L tone survived in the modern dialects. This change generated the modern Tokyo type tone systems. When the tone reduction occurred, in each dialect certain tone classes merged. In dialects where the tone of the particles had assimilated, the result was a Gairin type merger pattern. In dialects without such assimilation, the merger patterns were of the Nairin or Chūrin type. As all Ryukyuan varieties share the Gairin merger pattern of class 2.2 with 2.1 and 3.2 with 3.1, on Kyushu, the tone reduction must have been completed before the 10th century, when it is thought that the Japanese language spread to the Ryukyus. In central
218
Elisabeth M. de Boer
Japan, most likely sometime around the late 14th or early 15th century, a leftward tone shift occurred, which caused chaos in the Buddhist tone theories and left tell-tale traces in the modern Kyoto type tone systems, such as in the tone of compound nouns. Taking Ramsey’s MJ as starting point for this shift, it is possible to predict the reflexes in the modern Kyoto type tone systems with great precision. A look at the tone of the second syllable in MJ for instance, suffices to accurately predict the register of a word in the modern dialects, something that the standard theory cannot do. In certain dialects, the bond between the H tone and a particular syllable was lost. It is hard to put a time to such developments, but in case of southern Kyushu, it most likely predated the 10th century, as this innovation is shared by the Ryukyus.
References De Boer, Elisabeth M. 2008. The Middle Chinese tones through Japanese eyes. In Redouane Djamouri and Rint Sybesma (eds.), Chinese linguistics in Leipzig (CLÉ 2), 71–86. Paris: École des hautes études en sciences sociales-Centre de recherches linguistiques sur l’Asie orientale. De Boer, Elisabeth M. 2010. The historical development of Japanese tone. I From proto-Japanese to the modern dialects. II The introduction and adaptation of the Middle Chinese tones in Japan (Veröffentlichungen des Ostasien-Instituts der Ruhr-Universität Bochum [Publications of the East Asia Institute of the Ruhr University Bochum] 59). Wiesbaden: Harrassowitz Verlag. De Boer, Elisabeth M. 2017a. Review of Handbook of the Ryukyuan languages: History, structure and use. Studies in Language 41(1). 781–790. De Boer, Elisabeth M. 2017b. Universals of tone rules and diachronic change in Japanese. Journal of Asian and African Studies 94. 217–242. Hattori, Shirō. 1937. Genshi Nihon-go no ni-onsetsu meishi no akusento [The accent of disyllabic nouns in proto-Japanese]. Hōgen 7(6). 44–58. Hattori, Shirō. 1951. Genshi Nihon-go no akusento [The accent of proto-Japanese]. In Terakawa et al. (eds.), Kokugo akusento ronsō [Essays on the accent of the national language], 45–65. Tokyo: Hōsei-daigaku Shuppan-kyoku. Hombert, Jean-Marie. 1977. Development of tones from vowel height? Journal of Phonetics 5. 9–16. Hyman, Larry. 1993. Register tones and tonal geometry. In Harry Van der Hulst and Kieth Snider (eds.), The phonology of tone: The representation of tonal register, 75–108. Berlin and New York: Mouton de Gruyter. Hyman, Larry. 2007. Universals of tone rules: 30 years later. In Carlos Gussenhoven and Thomas Riad (eds.), Tones and tunes: Studies in word and sentence prosody, 1–34. Berlin and New York: Mouton de Gruyter. Kawakami, Shin. 1965. Heian-akusento to Bumō-ki no akusento [The accent of the Heian period and the accent of Bumōki] Kokugo-kokubun 34(2). (Reprinted in Kawakami, Shin. 1995. Nihon-go akusento ronshū [Collection of articles on the accent of Japanese], 414–433. Tokyo: Kyūko Shoin.) Kindaichi, Haruhiko. 1951. Nihon shisei kogi [The old value of the four tones in Japan]. In Terakawa et al. (eds.), Kokugo akusento ronsō [Essays on the accent of the national language], 629–708. Tokyo: Hōsei-daigaku Shuppan-kyoku. Kindaichi, Haruhiko. 1954. Tōzai ryō-akusento no chigai ga dekiru made [Until there arose a difference between the accent of the east and the west]. Bungaku 22(8). 63–84. Kindaichi, Haruhiko. 1955. Kodai-akusento kara kindai-akusento e [From ancient accent to modern accent]. Kokugo-gaku 22. 15–29.
9 The Ramsey hypothesis
219
Kindaichi, Haruhiko. 1974. Koku-go akusento no shi-teki kenkyū: Genri to hōhō [Historical study of Japanese accent: Principles and methods]. Tokyo: Hanawa Shobō. Mabuchi, Kazuo. 1962. Nihon ingaku-shi no kenkyū [Study of the history of Japanese phonological study] 1. Tokyo: Nihon Gakujutsu Shin-kōkai. Okumura, Mitsuo. 1981. Koku-go akusento no ichi-mondai: Izumo-hōgen no akusento o chūshin ni [A problem in Japanese accent: Concentrating on the accent of the Izumo dialect]. In Fujiwara Yoichi sensei koki kinen ronshū: Hōgen-gaku ronsō 2: Hōgen-kenkyū no shatei [Collection of articles in honor of the 70th birthday of Professor Fujiwara Yoichi: Essays on dialectology 2: The range of dialect study], 165–176. Tokyo: Sanseido. Ramsey, S. Robert. 1978. Accent and morphology in Korean dialects: A descriptive and historical study. Seoul: The society of Korean linguistics. Ramsey, S. Robert. 1979. The Old Kyōto dialect and the historical development of Japanese accent. Harvard Journal of Asiatic Studies 39. 157–175. Ramsey, S. Robert. 1980. Nihongo akusento no rekishi-teki henka [Historical change in the accent of Japanese]. Gengo 9(2). 64–76. Ramsey, S. Robert. 1982. Language change in Japan and the Odyssey of a Teisetsu. Journal of Japanese Studies 8(1). 97–131. Tokugawa, Munemasa. 1962. Nihon sho-hōgen akusento no keifu shiron [Towards a family tree for accent in Japanese dialects]. Gakushūin-daigaku Kokugo-kokubun Gakkai-shi 6. 1–19. Ueno, Kazuaki et al. 1998. Akusento hyōshi kaisetsu [Explanation of the accent markings]. In Kazue Akinaga et al. (eds.), Nihon-go akusento-shi sōgō shiryō: Kenkyū-hen [Comprehensive materials on the history of Japanese accent: Research volume], 35–50. Tokyo: Tōkyōdō Shuppan. Uwano, Zendō and Tetsuso Nitta. 1985. Kanazawa akusento ni okeru sedai-betsu henka [A generational change in the accent of Kanazawa]. Kokugo Kenkyu 49. 1–38. Wada, Minoru. 1943. Fukugō-go akusento no kōbu seiso to shite mita ni-onsetsu meishi [Two syllable nouns from the viewpoint of their accent as second element in compounds]. Hōgen Kenkyū 7. 1–26. Yanagita, Kunio. 1927. Kagyū-kō [Thoughts on ‘snail’]. Jinrui zasshi 42(4). 125–135.
Part III: Grammar
Yuko Yanagida
10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax 1 Introduction Yanagida and Whitman (2009) propose that two major clause types in Old Japanese (OJ, 8th century), the one traditionally labeled as shūsi ‘conclusive’ and the other identified by Yanagida and Whitman as ‘nominalized’ clauses, display different alignment and word order. While the subjects of conclusive verbs are zero-marked, the subjects of nominalized verbs are realized with the genitive ga, no, or zero. Yanagida and Whitman (2009) propose that the genitive ga, ancestor of the Modern Japanese nominative, is the realization of an active case on the external argument (i.e. the agent) of transitive or active intransitive verbs. Kikuta (2012), however, addresses certain problems with Yanagida and Whitman’s hypothesis, suggesting that variable subject marking in OJ is conditioned, not by the θ-role (i.e. agent) assigned by the verb, but by the place of the subject on the animacy hierarchy. First/second person pronouns are invariably marked by ga, but inanimate nouns are marked by no. In Kikuta’s analysis, OJ has a nominative-accusative system with two differential subject markings, ga and no. It should be noted that while this opposition between ga and no has gained much attention in traditional grammar, no previous work, including Kikuta’s, has as of yet integrated a discussion of zero-marked counterparts; they are simply set aside as instances of stylistic case drop. In recent typological and theoretical literature, languages with variable case marking have been investigated from the perspective of a broader pattern of differential argument marking. Differential subject marking occurs primarily in ergative languages, while differential object marking is independent of alignment and widely attested to in both ergative and accusative languages. This chapter discusses the characteristic phenomenon of differential argument marking in OJ. According to this approach, the crucial contrast is not merely between ga and no, but between casemarked and zero-marked arguments. Zero-marked arguments cannot be characterized simply as case drop, because they have both syntactic and semantic significance. The chapter is organized as follows. Section 2 begins with a critical review of the analysis of alignment in OJ as proposed by Vovin (1997) and Takeuchi (2008). Section 3 provides the basic morphosyntactic characteristics of active alignment as discussed widely in the literature, which provides an empirical basis for the claim that nominalized clauses in OJ show active alignment. Section 4 describes the results of a comprehensive survey of variable subject marking in OJ utilizing the Oxford-NINJAL Corpus of Old Japanese (ONCOJ). The data revealed that, while the alternation between ga and no is determined by the semantics of NPs, as widely assumed, differential subject marking https://doi.org/10.1515/9781614512851-011
224
Yuko Yanagida
associated with ga and zero is closely linked to the θ-role assigned by the verb, implying a binary classification of predicates into active and inactive. Section 5 investigates the phenomenon of differential object marking vis a vis close inspection of two prose texts in OJ: Norito and Senmyō. This analysis revealed that wo marks specific objects and that the specific object moves out of VP.
2 Alignment The typological literature widely assumes that alignment systems are classified into three types. Following Dixon’s (1979) familiar terminology, S refers to the subject of an intransitive verb, A to the subject of a transitive verb, and O to the object of a transitive verb. (1) Three Types of Alignment Nominative-Accusative A
S
O
Ergative-Absolutive A
S
O
Active-Inactive A
SA
SO O
In a nominative-accusative pattern, often abbreviated as ‘accusative pattern’, A and S are marked with nominative case, and O is accusative. In ergative and active patterns, often abbreviated as ‘non-accusative’, S and O are marked with absolutive case, and A is marked ergative or active. An active-inactive pattern is often analyzed as a subtype of an ergative pattern with a split intransitivity; the subject of an active intransitive verb (Sa) is marked in the same way as the subject of a transitive verb (A), but differently from the subject of an inactive intransitive verb (So). Many languages classified as non-accusative, however, exhibit a split ergativity in which a nominative-accusative pattern shows up in certain grammatical contexts typically conditioned by person or tense/aspect (cf. Dixon 1979).
2.1 Vovin (1997) Vovin (1997) initially proposes that OJ has active-inactive alignment. Under Vovin’s analysis, the case marker i, which is treated as a nominative particle by traditional grammarians (cf. Yamada 1968), is, in fact, an active case marking the subjects of transitive and active intransitive verbs. His examples are cited in (2): (2) a.
papa i more-domo. . . mother agt guard-although ‘Though [my] mother guards [me]. . .’
(MYS 14.3393)
10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax
b.
unapi wotokwo i ame apugi… Unapi man agt sky look.up ‘The man from Unapi looked up at the sky and. . .’
225
(MYS 9.1809)
Vovin observes that the subjects of inactive intransitives are overwhelmingly unmarked in the same manner as objects of transitive verbs. Similarly, the morphological case marker wo, ancestor of the Modern Japanese accusative o, marks not only the objects of transitive verbs, but also the subjects of inactive intransitives. In particular, wo marks the subjects of adjectival predicates with -mi, which Vovin calls ‘quality stative verbs’. This is illustrated in (3–4). (3) [izami no yama wo taka-mi] kamo yamato no mi-ye-nu Izami gen mountain abs high-ger q Yamato gen see-pass-not ‘Is it because the Izami mountains are so high that I can’t see Yamato? (MYS 1.44) (4) [kuni Ø topo-mi] kamo province abs far-ger Q ‘Is it because I came too far from my country?’
(MYS 1.44)
Given these observations, Vovin claims that the unmarked zero form and wo are both absolutive in OJ. Vovin’s analysis of wo, however, relies heavily on the wo. . .-mi construction. Aside from this construction, the examples Vovin cites do not necessarily show that wo marks the subject of intransitives. For example, consider (5): (5) murasaki no nipop-yeru imwo wo niku-ku ara-ba violet gen beautiful-perf beloved abs unpleasant-ger be-cond pitoduma yuwe ni ware kwopwi-me ya mo other’s.wife due to I love-fut.excl Q even ‘If [my] beloved, who is beautiful like a violet, was not beautiful to me, would I love her even though she is another’s wife?’ (MYS 1.21) In (5), imwo ‘my beloved’ is the subject of the adjectival predicate nikuku ‘unpleasant’ (as Vovin observes), but, at the same time, it is the object of the matrix verb kwopu ‘love’. That is, the entire clause has the configuration [DPi wo [proi V] V], in which the object marked with wo appears in the higher clause, and the embedded clause contains the phonologically null subject (pro) coindexed with it. Takeuchi’s (2008) proposal that OJ has active alignment is heavily based on Vovin’s (1997) observations about wo. Example (6) is cited from Takeuchi (2008).
226
Yuko Yanagida
(6) miti no siri kwopoda wotomye wo kamwi no goto road gen back Kohada maiden abs god gen like kikoye-sika-domo api makura-maku (KK 45) be.heard-pst-but together sleep-aux.nmlz ‘Rumors about the Kohada maiden in her far-off land rumbled like thunder, but we lie together.’ Takeuchi (2008) claims that the wo-marked argument is the sole argument of the intransitive verb kiko-yu ‘can be heard’. It is important to note, however, that (6) has exactly the same structure as (5). In (6), the wo-marked argument that precedes the embedded domo-clauses is, in fact, the associative object of the matrix verb makura-maku ‘sleep together’.1 Aside from DP wo. . .-mi constructions, neither Vovin nor Takeuchi present convincing evidence that wo marks the subject of inactive intransitive verbs. On the contrary, there is substantial evidence that subjects of non-active intransitives are marked with the genitive no. (7) a. makwi no tatu ara yama naka ni tree gen stand rough mountain inside loc ‘in the rough mountains covered with trees’ b. u no pana no saku tukwi utsugi gen blossom gen bloom.adn month ‘the month when the utsugi blossom is in bloom.’
(MYS 3.241)
(MYS 18.4066)
If wo is an absolutive case marker, then we have no explanation for why the subject is never marked with wo in adnominal contexts (7a-b). For now, we may set aside the status of the NP wo. . .-mi pattern in (3). However, it is important to note that Tsuta (2004) convincingly argues that the diachronic source for -mi is the infinitive of the transitive verb mi- ‘see’. According to this analysis, the subject of the adjectival predicate is, in fact, the matrix object of the verb ✶mi- (at least in preOJ). Similarly, Yanagida and Whitman (2009) analyze the wo. . .-mi pattern as adjunct AspPs, analogous to Acc-ing gerunds, such as ‘travel being painful’ in English. These have the following structure: (8) [AspP tabi wo [VP kurusi ] mi ] kwopwi wore-ba travel obj painful mi long.for be-prov ‘travel being painful, since I long for my wife’
(MYS 15.3674)
In this analysis, -mi is the spellout of the head of [+transitive] AspP. The subject of the adjectival predicate is susceptible to a matrix object (or ECM) analysis of the verb ✶mi-. 1 As shown in Section 5.1, wo in OJ marks a much wider range of internal arguments than ModJ o (for detailed observations, see Yanagida 2006).
10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax
227
The hypothesis that wo marks the absolutive is based on the whole-language characterization of alignment typology, which assumes that the objects of transitive verbs are marked absolutive in ergative languages. However, the skewed ergative (active)-accusative pattern is widely attested, for example in Indic languages such as Hindi. Hindi, traditionally classified as ergative, in fact features active alignment; agent subjects of unergative verbs are marked with -ne, but the theme subjects of unaccusative verbs are morphologically zero (Mohanan 1994: 71). Hindi (Indo-Aryan) (9) a. Raam-ne nahaayaa. Ram-erg bathe-perf ‘Ram bathed.’
b. Raam Ø giraa. Ram.abs fall-perf ‘Ram fell hard.’
The animate objects of transitive verbs, however, are necessarily marked with the accusative ko (10). The inanimate objects are marked with ko when they are specific; otherwise, they are morphologically zero (12) (Mohanan 1994: 79–80). Hindi (Indo-Aryan) (10) a. Ilaa-ne ek bacce-ko uthaayaa. Ila-agt one child-acc lift/carry-perf ‘Ila lifted a child.’ b. Ilaa-ne haar Ø uthaayaa. Ila-agt necklace.abs lift-perf ‘Illa lifted a necklace (non-specific).’ As shown in Hindi, many ergative languages feature the phenomenon of differential object marking. Depending on the animacy, specificity, or definiteness of NPs, certain objects are marked with the accusative case, but others are not. Differential object marking in OJ will be discussed in Section 5.
2.2 What is ‘active’? Both the typological and theoretical literature have tended to classify ‘active’ as a subtype of ergative alignment because both ergative and active cases mark the agentive subjects (A) of transitive verbs, but not the patient arguments (S) of intransitive verbs. It is well known, however, that active-stative languages display considerable divergence in both morphology and syntax, which makes it difficult to find a coherent implementation of languages of this type. This section discusses the characteristic properties of active alignment that provide an empirical basis for the claim that OJ displays an active-inactive pattern.
228
Yuko Yanagida
2.2.1 The two classes of predicates Active languages divide intransitive verbs into two categories: active and inactive. The exact lexical division differs crosslinguistically, but the two classes of intransitive verbs are distinguished by case marking. Active intransitive subjects (SA, typically agent arguments of unergatives) have the same marking as transitive subjects, while inactive intransitive subjects (SO, typically patient arguments of unaccusatives) have the same marking as transitive objects. We see such a pattern in Hindi, as illustrated in (9). Dixon (1979: 80–83) divides active alignment into fluid-S and split-S systems. In fluid-S systems, verbs are divided according to the meaning of each particular token. The active pattern appears when the S argument has control over the activity, and the inactive pattern appears when control is lacking. We see this pattern in Batsbi, a fluid S language cited by Comrie (1978: 366). In split-S systems, on the other hand, the two classes of intransitive verbs have fixed membership: they are classified as active or inactive based on their prototypical meaning. Guaraní (Tupí-Guaraní Mithun 1991), a head-marking language, features a split-S system. In Guaraní, the unaccusative verb ‘die’, which involves no intention or control, is classified as active, while it is classified as inactive in most fluid S languages. In other words, this binary classification of active and inactive is based on some idiosyncratic meanings of a given word.
2.2.2 Nominal hierarchy It is important to note that active and inactive marking depends not only on the semantics of predicates but on the place of S in the nominal hierarchy (11): (11)
The Nominal Hierarchy (Silverstein 1976, revised by Dixon 1994: 85) first/second person > third person > proper nouns > human > animate > inanimate
Here, Dixon (1979: 86–87) interprets the nominal hierarchy (11) to “roughly indicate the overall ‘agency potential’ of any given NP,” and observes that “a number of languages have ‘split’ case marking exactly on this principle.” As Mithun (1991) points out, case systems based on agency are frequently restricted to nominals referring to human beings.2 According to Mithun, Koasati shows agentive case marking of pronominal prefixes within verbs, but accusative case marking of nouns. The active system in Batsbi (Tsova-Tush) is limited to the first and second persons. Central Pomo has an active system in nominals referring to humans only. The Georgian active system is likewise restricted to human beings, while the Yuki system is restricted to animates. From these
2 Mithun (1991) identifies the semantic basis of the active-marking of various West Hemisphere languages, both synchronically and diachronically.
10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax
229
cross-linguistic observations, it follows that active marking is used with NPs from the left-hand side to the right-hand side of the nominal hierarchy. That is, if a language has agent marking in the third person, it also has agent marking in the first and second persons, which is exactly the opposite of the ergative case used with NPs from the righthand to the left-hand side of the nominal hierarchy (Dixon 1979). In languages like Guaraní (Tupí-Guaraní), transitive verbs are marked either active or inactive, depending on which of the two arguments is located higher on the nominal hierarchy. The argument that ranks higher is cross-referenced on the verb (Velazquez-Castillo 1996: 17). When the subject outranks the object or two arguments are of the same rank, the agent is cross-referenced on the verb (active marking). When the object outranks the subject, it is the patient that is cross-referenced on the verb (inactive marking). Even though the thematic role assigned by the verb is identical, assignment of active case is strictly determined by the place of the subject on the nominal hierarchy. That is, the active-inactive division in Guaraní is a clause-level phenomenon defined as the type of the grammatical relation between subject and object NPs.
2.2.3 Active/Genitive syncretism It has been widely acknowledged that ergative/active patterns show syncretism between ergative/active marking and possessive marking, e.g. Mayan (Coon 2008, Coon and Salanova 2009), Inuktitut (Johns 1992), Austronesian (Kaufman 2009, Aldridge 2015), Cariban (Gildea 1998, 2000), East Caucasian (Authier 2013), Guarani (Velazquez-Castillo 1996), Indic Iranian (Bynon 2005), and many others. Johns (1992) argues that, in Inuktitut, the ergative case is homophonous with the genitive case (-up for ergative/genitive); thus, the possessive construction and the ergative/active construction are structurally identical. Johns (1992) develops a synchronic account of Inuktitut ergativity based on possessive constructions. Gildea (1998, 2000) and Bynon (2005) propose a similar analysis for Cariban and Indic Iranian languages, respectively. Following this approach, syncretism between agent marking and genitive marking arises as a result of reanalysis of a possessive construction with the copula ‘be’ as monoclausal structure. A possessor is reanalyzed as an external argument (i.e. agent), and the biclausal copular structure is reanalyzed as a monoclausal transitive clause.
2.2.4 Syntax Many researchers propose that the ergative (or active) case is assigned to the external argument (agent) in the specifier position of vP (cf. Woolford [1997, 2008], Legate [2008], Aldridge [2004, 2008] and Anand and Nevins [2006], and many others). The external argument (agent) is θ-marked and inherently case-assigned by v in a vP projection above VP, as represented in (12).
230
(12)
Yuko Yanagida
Differential subject marking at argument structure vP v’ external argument v VP [+Agt]
Legate (2008) points out that, while ergative is assigned to the external argument in the specifier position of [+transitive] v, active is assigned to the external argument in the specifier of [+Agent] v. The descriptive generalization, which supports the view that the ergative is an inherent case assigned by v, derives from the fact that ergative subjects, in some instances, occur in non-finite clauses, while structural nominative subjects do not (cf. Aldridge 2008). Derived subjects are never ergative; that is, no language promotes objects to ergative through operations such as raising or passive. Some recent researchers, however, have argued against the inherent case analysis of ergative, suggesting that ergative case is instead structural case. Baker (2014) argues that the ergative case in Shipibo is a structural case rather than inherent. Rezac, Albizu and Etxepare (2014) claim that Basque ergative is structural, based on the ergative-to-absolutive in so-called ‘defective’ T contexts, such as raising and ECM constructions. Yanagida (2018a) proposes that ergative/active case is structural when it responds to the subject-in-situ generalization (SSG), which Alexiadou and Anagnostopoulou (2001) claim to be a universal principle on structural case. The SSG states that when the subject remains in Spec, vP, the object must be externalized. (13)
The Subject-in-Situ Generalization (SSG) By Spell-out, vP can contain only one argument with an unchecked Case feature. (Alexiadou and Anagnostopoulou 2001)
A widely observed feature of syntactically ergative languages is that, while the external subject of a transitive verb remains inside vP, the object of a transitive verb appears outside VP (Bittner and Hale 1996, Manning 1996, among others). These languages include Dyirbal (the Pama-Nyungan language of Australia; Dixon 1994: 130), Kuikúro and Panare (Cariban languages; Franchetto 1990 and Gildea 1998, 2000), Vafsi (Northwest Iranian; Haig 2008: 188). These languages are split ergative languages. In a nominative-accusative pattern, a direct object remains inside VP. When the subject is marked with ergative case, the direct object moves outside VP. Importantly, object movement is not a property of absolutive DP. As noted by Dixon, the object moves regardless of whether it is absolutive (unmarked) or accusative (case-marked). In Section 5, we will show that OJ exhibits object movement characteristic of non-accusative languages when the subject is marked with ga.
10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax
231
3 Morphology 3.1 Active/inactive prefixes Crosslinguistically, active alignment can be manifested in the morphological case marked on nouns, but many active languages are strictly head-marking. As is wellknown, OJ possesses various verbal prefixes whose syntactic and semantic functions have been left unexplained by traditional grammarians. Yanagida and Whitman (2009) provide a comprehensive study of these prefixes and suggest that i- is attached to active verbs and sa- to inactive verbs. Examples (14) and (15) are cited in Yanagida and Whitman (2009: 117). N.B. all 75 occurrences for i- in the Man’yōshū are given in Yanagida (2007: 178–179). (14) a. Nara no miyakwo no Sapo kapa ni i-yuki itarite Nara gen capital gen Saho river loc i-go reaching ‘I reached the River Sahokawa in Nara.’ (MYS 1.79) b. Kume no wakugwo ga i-pure-kyemu iswo no kusane Kume gen youth agt i-touch-aux.adn rock gen grass root ‘the root of the grass that the youth of Kume would have touched.’ (MYS 3.435) (15) a. sa-ne-si tumaya ni . . . ide-tati sinwopi sa-sleep-pst.adn bedroom loc leave-out remember ‘remembering, leaving the bedroom where (I) slept’ (MYS 3.481) b. sa-niturapu wa ga opokimi sa-shine.adn I gen great.lord ‘my great lord who shines’ (MYS 3.420) c. [VP sugwi no nwo ni sa-wodoru] kigisi cypress gen field loc sa-dance pheasant ‘The pheasant dances in the cypress field.’ (MYS 19.4148) Predicates that appear with -i include: yoseru ‘put aside’, puru ‘wave’, yuku ‘go’, wataru ‘cross’, toru ‘take’, karu ‘mow’, kakuru ‘hide’, wogamu ‘pray’, maporu ‘go around’, poru ‘dig’, wakaru ‘part from’, kogu ‘row a boat’, mukapu ‘head out’, pirou ‘pick up’, mureru ‘gather’, etc. Predicates that appear with the prefix sa- include: neru ‘sleep’, niturapu ‘shine’, pasiru ‘(fish) run’, wodoru ‘(birds) dance’, wataru ‘(toads) cross’, nebapu ‘spread roots’, narabu ‘(birds) line up’, kumoru ‘get cloudy’, and nituku ‘get reddened’. Although these prefixes are somewhat vestigial in OJ, the distribution of i- and sa- strongly suggests that i- appears with active verbs, while sa- appears with inactive intransitive verbs. It should be noted that sa- occurs as a noun prefix, as in sa-ywo ‘night’, while idoes not. This division parallels exactly the distribution of agreement prefixes in active languages, such as Sateré-Mawé (Meira 2006), where inactive prefixes occur with both
232
Yuko Yanagida
nouns and inactive verbs while active prefixes occur with active verbs only.3 Furthermore, of both nouns and verbs as in (16), sa- triggers the process known as rendaku, in which the initial voiceless obstruents of a noun or a verb become voiced. (16)
a. b.
sa koromo > saNkoromo > sa-goromo ‘his clothes’ sa pasiri > saNpasiri > sa-basiri ‘his running’
✶ ✶
(16) shows that sa may reflect an original possessive S argument. The rendaku process involves an earlier syllable of the form nasal + vowel (NV). The literature has suggested that the NV sequence is the earlier form of the genitive particle no. Yanagida and Whitman (2009) cited one apparent counterexample to the generalization in MYS 5.804 in which ga and prefixal sa- appear to surface in the same clause: (17)
wotomye-ra ga sa-nasu itatwo wo osi-piraki maiden-pl agt sa-sleep door obj push-open ‘pushing open the door where the maidens sleep.’
(MYS 5.804)
Kojima, Kinoshita and Tōno (1995), however, interpret wotomyera ga ‘maidens GA’ as the genitive possessor of itatwo ‘(wooden) door’, a metonymic expression for ‘bedroom’. The entire NP, then, is interpreted as ‘pushing open the maiden’s (bedroom) door where they sleep’. This structure depicted in (18): (18) [NP wotomyera ga [[ pro sa-nasu ] itatwo]] wo osipiraki maidens gen sa-sleep door obj push open In this interpretation, wotomyera ga is not the clausemate subject of sa-nasu ‘sa-sleep.’
3.2 wataru ‘cross’ Yanagida and Whitman observe that a verb, wataru ‘cross’ appears with either i- or sa-. There are 4 examples of i-watar- in the Man’yōshū (MYS 9.1742, 10.2081, 18.4101, and 18.4126), and 6 examples of sa-watar- (MYS 5.800, 6.971, 10.1960, 10.1976, 11.2450, and 11.2804). The subject of i-watar- is [+human] and volitional in all four examples: kwo ‘the young woman’ (9.1742), tanabata ‘Vega, the weaver star’(10.2081), ama ‘the fisherfolk’ (18.4101). The subject of sa-watar- is [-human] in all six examples: taniguku ‘toads’ (5.800, 6.971), pototogisu ‘a cuckoo’ (10.1960, 10.1976), tukwi ‘the moon’ (11.2450), 3 Sateré-Mawé (Tupian) has an active system marked by two series of personal prefixes on the verb (cf. Mithun 1991). Meira (2006) shows that, in Mawé, nonactive verbs are strikingly similar to (possessed) nouns. The same set of personal prefixes appear on nouns and nonactive verbs, and these prefixes do not select active verbs.
10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax
233
takabye ‘a teal’ (11.2804). Typical examples of each pattern as cited by Yanagida and Whitman are given in (19–20). (19) ama no gapa sky gen river i–watara-sa-mu i-cross-hon-aux ‘though if one put i-cross over on it’
pasi watasera-ba sono pe yu mo bridge span-cond that over from too wo excl a bridge across the Milky Way, (they=Vega and Altair) would (MYS 18.4126)
(20) kumo-ma ywori sa-wataru tukwi no opoposiku api misi kwo-ra cloud-gap from sa-cross moon like faintly join saw child-dim ‘the girl I saw faintly like the moon sa-crossing between the clouds’ (MYS 11.2450) I-watar- ‘cross (over the bridge)’ is agentive, volitional, and telic – a stereotypical active verb. Sa-watar- is non-agentive and designates an incomplete action (the moon passing before the speaker’s view); it is a stereotypical inactive predicate. To summarize this section, OJ nominalized clauses show the vestiges of head-marking of an active-inactive division: active predicates by the prefix i- is in opposition to inactive predicates by the prefix sa-.
4 Differential subject marking In OJ, ‘nominalized clauses’, as identified by Yanagida and Whitman (2009), show three distinct ways of case marking. The genitive ga, the ancestor of Modern Japanese nominative case, is used for agentive case marking predominantly for the subjects of active verbs, while the theme subjects of inactive verbs are predominantly marked zero. The other genitive no is used independently of predicates; it can mark the subject of either an active or inactive verb. The coding property of the subject NP is determined by the location of NP in the nominal hierarchy, as stated in Table 1. Table 1: Three-way case marking patterns on the subject of nominalized verbs.
As discussed in 2.2.2, the nominal hierarchy is interpreted as ‘the agency potential of given NPs’ (Dixon 1979: 86–87). Nominals higher up in the hierarchy are more likely to serve as the subject of a prototypical transitive verb. Table 1 shows that the alternation between ga and no depends on the place of the subject in the nominal hierarchy. Ga occurs on the weak or clitic forms of personal pronouns (primarily monosyllabic forms such as a, wa [1P], na (2P), si [3P]) and kinship terms, such as ‘mother’ and ‘child’, while the other genitive no occurs on common NPs. The genitive ga marks the possessors of NP (21), the agent subject of the transitive verb (22), and the active intransitive verb (23) (See Appendix I for other nominal clause types). Possessor-possessed NP (21) [wa ga sekwo ga yadwo] I gen lover gen house ‘my lover’s house’
(MYS 20.4303)
Adnominal Clauses (22) [Saywopimye no kwo ga pire Ø puri-si] yama Sayohime gen child agt scarf wave-pst.adn mountain ‘the mountain where Sayohime waved her cloth’ (MYS 5.868) (23) kimi ga yuku miti you agt go road ‘the road my lord (you) travels’
(MYS 15.3724)
The patient subjects of inactive intransitive verbs (24), on the other hand, behave like the objects of transitive verbs (22) insofar as they are zero-marked. Zero-marked subjects appear predominantly with unaccusative verbs and strictly adjacent to the verb. (24) a. aki no nwo ni tuyu Ø opyeru pagwi wo ta-wora-zu-te fall gen field loc dew cover bush.clover obj hand-break-not-ger ‘without breaking off the dew-laden bush clover in the fall meadow’ (MYS 20.4318) b. Uramwi ywori kadi no oto Ø suru pa amawotomye kamo Urami from oar gen sound make top fisherfolk.maiden Q ‘Is the sound of the oar from Urami a maiden of the fisherfolk?’ (MYS 15.3641) Kinship terms are predominantly marked with ga, but there are some instances in which they are marked zero (25) when the predicate is inactive.
10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax
(25)
a.
b.
c.
235
tama sika-ba kimi Ø ki-masa-mu ka kiywoki pamapye ni pebble lay-cond you/lord come-aux-aux Q clean seashore loc ‘If I lay pebbles, will you/lord come on this clean seashore?’ (MYS 19.4271) puri siku yuki ni kimi Ø imasa-me ya mo fall spread snow loc you stay-aux foc Q ‘Will my lord stay despite of this heavy snowfall?’ (MYS 19.4233) kimi Ø mase-ba tokotu mikadwo to tono-wi suru kamo lord rest-prov eternal palace as on-duty do Q ‘Since the lord rests (there), shall I be permanently on night duty (in the court)?’ (MYS 2.174)
N.B. The genitive no is independent of alignment, unlike ga. It can mark both the subject of transitive verbs (26) and the subject of inactive intransitive verbs (27). (26)
ipyebito no idura to ware wo topa-ba ikani ipa-mu family gen where comp I obj ask-if how say-aux.adn ‘If your family should ask me where (you are now), how should I reply to them?’ (MYS 15.3689)
(27)
a.
b.
makwi no tatu ara yama naka ni great.tree gen stand rough mountain inside loc ‘in the rough mountains covered with trees’ u no pana no saku tukwi utugi gen blossom gen bloom.adn month ‘the month when the utsugi blossom is in bloom’
(MYS 3.241)
(MYS 18.4066)
Kikuta (2012) provides some counterexamples to Yanagida and Whitman’s hypothesis, suggesting that Japanese has never actually undergone an alignment change. She instead posits that adnominal clauses show a nominative-accusative pattern throughout the history of the language, but in OJ this pattern shows with two differential case markers: ga and no. Kikuta’s proposal is primarily based on the observation that ga marks, not only the subjects of active verbs, but also the subjects of inactive intransitive verbs. Possible counterexamples cited by Kikuta (2012) will be examined in detail below.
4.1 Psych predicates Kikuta (2012) points out that the OJ ga appears on the non-agentive theme subjects of experiencer verbs, such as wasur- ‘forget’, omop- ‘think’, mi ‘see’, as illustrated in (28). According to Kikuta, this raises a problem for Yanagida and Whitman’s (2009) hypothesis that ga marks the active case in OJ.
236
Yuko Yanagida
(28) a. imo ga kwopisiku wasura-ye-nu kamo my.lover agt miss forget-mid-neg Q ‘Did I miss my dear and cannot forget her?’ (MYS 20.4407) b. Yupuma yama kwoye-ni-si kimi ga omopo-yur-aku ni Yūma mountain cross-perf-pst you/lord agt think-mid-nmlz cop ‘you who had crossed over Mount Yūma came to mind!’ (MYS 12.3191) Note, however, that these specific psych-verb constructions with a ga-marked theme subject contain an unspecified first-person experiencer and a form of the auxiliary yu (stem ye-), which derives middles, passives, and potentials. -Yu is arguably related to the acquisitive light verb u (stem e-) ‘get’, which Whitman (2008) proposes as the source of the well-known transitivity alternations in -e- in OJ and later stages of the language. -E derives both transitives and intransitives, which is a property of acquisitives such as the English auxiliary get. Yanagida (2018b, 2022) argues that experiencer middles such as (28) may have an original causative source, i.e. ‘my dear got me to forget’ or ‘my lord got me to think.’ Yanagida suggests that these particular object experiencer predicates in OJ are semantically transitive and possess the following characteristics: 1) they are impersonal, i.e. a first-person experiencer is necessarily unexpressed; 2) the predicates take the vestigial causative light verb; and 3) an argument marked with ga is necessarily interpreted as a causer, but not as an experiencer. Accordingly, (28) can be analyzed as a causative construction. N.B. the agent subject is invariably an external argument, but the causer argument of a psych-verb is also an external argument. Thus, in many languages, the causer argument of an object-experiencer verb is marked with the ergative (see Woolford [2008] for Assamese [Eastern Indo-Aryan language]).
4.2 Active/inactive predicates If ga is an active case, we do not expect it to cooccur with non-agentive stative verbs. Kikuta (2012), however, provides possible counterexamples beside psych-predicates, as illustrated in (29–30): (29)
imo ga papi nite mase-ba. . . lover agt ash into be/go-prov ‘when my dear has gone into ashes. . .’
(30) tegwona ga ari-sika-ba. . . maid agt be-pst-prov ‘When there was a maid. . .’
(MYS 2.213)
(MYS 14.3385)
237
10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax
These verbs are generally treated as inactive: the subject has no control or intention over the activity denoted by the verb. Given the data taken from Koji (1988), as cited in Tables 2 and 3, Kikuta (2012) argues that differential case-marking in OJ is not conditioned by the semantics of the predicates, but by the location of the nominals on the animacy hierarchy. Table 2: Pronominal subject and pronominal possessives marked with ga (Kikuta 2012). wa 1p
a 1p
na 2p
ono 2p
ta 3p
si 3p
kore this
total
45 89
31 34
4 7
3 8
0 3
2 0
1 0
86 141
subject Possessive
Table 3: Nominal subject and nominal possessives marked with ga (Kikuta 2012).
subject possessive
kimi
imo
wag-imo
waga sekwo
wotomye
papa
kwo
others
total
90 39
49 97
37 26
28 25
16 11
9 5
6 9
23 76
258 288
The genitive ga is obligatory for first/second personal pronouns; w(a) and na, and the nominals intimate to the speaker, such as kimi ‘you/lord’, (wag-)imo, sekwo ‘lover’, wotomye ‘girl’, papa ‘mother’, and kwo ‘child’ (cf. Ohno 1977, Nomura 1993). The other genitive, no, in contrast, is used for nominals lower on the animacy hierarchy. However, no previous work – – including Kikuta’s – has discussed the zero case in OJ. The crucial contrast here is not merely between ga and no, but between ga and zero. If ga and zero are associated with the active/inactive division, as argued in Yanagida and Whitman (2009), we would expect ga to appear with active predicates whose subjects are not marked zero, but possibly with no if the subject NP is lower in the nominal hierarchy. The data represented in Tables 4 and 5 were obtained from the Oxford-NINJAL Corpus of Old Japanese (ONCOJ), a syntactically annotated corpus, and selected by means of an exhaustive search designed to select predicates whose subjects are marked with ga, no, and zero.4
4 I owe many thanks to Stephen Horn for his help in obtaining data from the ONCOJ.
238
Yuko Yanagida
Table 4: Verbs with high volitionality (non-conclusive form).5
ga no zero
Total
mat‘wait’
nak‘weep’
903 1255 2054
43 8 0
30 26 0
yuk- pur‘go’ ‘wave’ 24 12 0
uwe‘plant’
ki‘wear’
yos‘approach’
watar‘cross’
warap‘laugh’
7 1 0
5 1 0
4 1 0
3 3 0
2 1 0
11 0 0
Table 4 includes the total number of subjects marked with ga/no/zero with predicates in the non-conclusive form. This study reveals that verbs which most frequently appear with ga-marked subjects never appear with zero-marked subjects. Table 5, on the other hand, shows possible counterexamples in which ga appears with predicates with low volitionality. Table 5: The class of verbs with low volitionality (non-conclusive form).
ga no zero
ne‘sleep’
imas‘be/go’
wor‘sit’
ar‘be’
nure‘get wet’
sin‘die’
total
12 1 0
9 3 5
7 0 0
5 4 31
1 0 6
1 3 5
37 25 71
The verb 座 is ambiguously interpreted as either the existential ‘be’ (inactive) or ‘go’ (active). Kikuta cites (29) as a counterexample, but imas- in (29) could mean ‘go’. In OJ, the low volitionality verbs ne- ‘sleep’ and wor- ‘sit’ are, in fact, categorized as active since their subjects are marked with ga but never with zero. (Recall that the division of verbs into active/inactive subclasses involves some idiosyncratic properties of a given language [see 2.2.1]). There is only one problematic example in which ne- ‘sleep’ appears with a zero-marked subject. (31)
asipye ni pa kari Ø ne (宿) -taru kamo reed.clump loc top geese sleep/stay-perf Q ‘Wild geese might have stayed in a clump of reeds.’
(MYS 10.2135)
5 Subjects of predicates in the conclusive form are excluded since they are assigned nominative (i.e. zero) marking both S and A. Moreover, subjects with nak-, meaning ‘make a cry’ (nonvolitional) as opposed to ‘weep’ (volitional), are not included. They are all non-human animate (86 tokens). Inanimate bare subjects with yuk- ‘go/pass/come’, as in (i), are not included either (14 tokens). (i)
aki yuke-ba kurenawi nipopu autumn come-prov crimson smell ‘When autumn comes crimson smells’
(i) involves no volitional activity performed by the agent.
(MYS 13.3227)
10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax
239
Nakanishi (2005[1983]) points out that the Chinese character 宿 in (31) could be read as yador- ‘stay’ rather than ne- ‘sleep’. Given that this is the only exception we found in the ONCOJ, I simply assume that 宿 here stands for yador- ‘stay’ and that its subject appears unmarked. Although the data reviewed contain some counterexamples, specifically, 5 tokens of ga with the verb ar- ‘be’ (illustrated in [30]), overall, the data obtained from the ONCOJ support the hypothesis that ga and zero divide predicates into active/inactive in OJ.
5 Differential object marking 5.1 Zero-marked object Miyagawa (1989) proposes that, in OJ and Early Middle Japanese, adnominal and conclusive clauses have distinct case assigning mechanisms. The conclusive form of the verb is truly verbal and assigns abstract case, that is, morphologically zero, to the object in the underlying object position, while the adnominal form has no case assigning ability. The object is assigned overt structural case in the form of wo in order to avoid violation of the Case Filter. Miyagawa’s (1989) generalization is stated in (32). (32)
Miyagawa’s (1989) generalization (1989: 206) Accusative Case Assignment: The conclusive form assigns abstract case while the case assigning feature of the attributive (=adnominal) form must be manifested overtly as wo.
However, as pointed out by Kinsui (1993, 2011), and Yanagida (2007a, b) there are a number of examples in which an adnominal predicate takes an object lacking a morphological case. Miyagawa and Ekida (2003) attempt to account for these exceptions to Miyagawa (1989), but their study is not sufficient to cover all the exceptions. In response to Miyagawa (1989), Kinsui (2011: 104) suggests that the marking of objects with wo is purely stylistic. In some cases, whether wo occurs at all is determined by poetic versification with the basic line configuration of 5-7-5-7 syllables. (33)
titi-papa wo mire-ba taputwosi mye-kwo mire-ba megusi father-mother obj see-prov respect wife-child see-prov cute utukusi beautiful ‘When I see my father and mother, I feel respect; when I see my wife and children, they are lovable and beautiful. . .’ (MYS 5.800)
240
Yuko Yanagida
In (33), the first and second objects appear in the same syntactic contexts: inside a provisional clause headed by ba ‘when’. Nonetheless, the first occurrence of the object is marked with wo, and the second occurrence of the object is morphologically unmarked. It is important to note, however, that the second occurrence of mye-kwo ‘wife and child’ ends with the labialized mid-back vowel, which, according to one interpretation, is homophonous with the case particle wo. This analysis raises the possibility that deletion of the second occurrence of wo may simply be a case of haplology, or, more specifically, the poet taking advantage of haplology to preserve the meter.6 I hypothesize that poetic versification does not override language’s core grammar, but comes into play only when the grammar allows optionality. Yanagida (2007a, b) indicates that in the Man’yōshū, there are at least 90 occurrences of a transitive clause whose subject is marked with no or ga and whose object is morphologically unmarked. Fifty-five occur with adnominal predicates. These include examples like (34). (34)
a. Saywopimye no kwo ga pire Ø puri-si yama no na Sayohime gen child agt scarf wave-pst.adn hill gen name ‘the name of the hill where Sayohime waved a scarf’ (MYS 5.868) b. kanasiki kworo ga ninwo Ø posaru kamo beloved child agt cloth dry Q ‘Is my beloved drying woven cloth?’ (MYS 14.3351)
Examples like (34a-b) are clearly counterexamples to Miyagawa’s (1989) generalization. Yanagida (2007a, b), however, indicates that although zero-marked objects do occur with adnominal predicates, Miyagawa’s exceptions are predictable. The objects that follow the ga-marked subject are, without exception, non-branching noun heads immediately adjacent to the verb. Yanagida (2007a, b) and Yanagida and Whitman (2009) suggest that zero-marked nouns, such as pire ‘scarf’ and ninwo ‘cloth’, are syntactically incorporated into the verb.7 That is, given that an incorporated noun need not be assigned a structural case, as suggested by Baker (1988: 106), examples like (34a-b) are analyzed as derived intransitives (N.B. object incorporation is a salient feature of languages with active alignment as observed by Klimov (1977: 125–126) and Sapir (1911). In Section 2.2.4, it was shown that nominalized clauses display a non-accusative pattern when v assigns no structural case to the object. From a typological perspective, Miyagawa’s (1989) synchronic treatment of adnominal clauses in OJ displays strong evidence that adnominal clauses have a non-accusative pattern.
6 I extend my thanks to John Whitman for this observation. 7 Modern Japanese does not have noun incorporation in the strict sense. The patterns of incorporation discussed by Kageyama (1980), such as kosi o kakeru vs. kosikakeru, tema o toru vs. temadoru, are not productive. These expressions are possibly analyzable as lexical compounds.
10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax
241
5.2 Wo-marked objects Yanagida (2006), following Motohashi (1989), proposes that wo-marked objects in OJ are interpreted as definite, while zero-marked objects are indefinite. However, since there are examples in which specific (i.e. D-linked) wh-phrases are marked with wo, Yanagida and Whitman revise Yanagida’s (2006) original claim and propose that the accusative case occurs when objects are specific. (35) sipo pwi-na-ba tamamo kari tum-ye ipye no imo ga tide recede-perf-cond seaweed cut gather-imp house gen wife agt pamadutwo kopa-ba nani wo simyesa-mu? shore.gift want-cond what obj proffer-aux.adn ‘If the tide has gone out, cut and gather the precious seaweed! If my wife at home asks for gifts from the shore, which (other) shall I offer her?’ (MYS 3.360) (36) makwi no itatwo wo osi piraki siweya ide ko-ne wood gen door obj push open damn out come-opt noti pa nani se-mu after top what do-aux.adn ‘Pushing open the wooden door (I say) “Come out!” Then what will (I) do?’ (MYS 11.2519) In (35), the set of items that the speaker might offer his wife is defined as pamadutwo ‘gifts from the shore’. In this case, nani wo ‘what/which Obj’ picks out specific items from that set. In (36), by contrast, the universe of things the speaker might do is completely undefined by previous discourse. Frellesvig, Horn and Yanagida (2015) make a complete search for the two types of objects using the ONCOJ. They suggest that a contrast between wo-marked and zeromarked objects in OJ fits into typologically well-attested differential object marking. The minimal pair examples (37–38) are cited in Frellesvig, Horn and Yanagida (2015). (37)
kami tu se ni u wo ya-tu kaduke upper gen stream dat cormorant obj eight-clf make.dive simo tu se ni u wo ya-tu kaduke lower gen stream dat cormorant obj eight-clf make.dive ‘. . .making all eight of my cormorants dive in the upper reaches, making all eight of my cormorants dive in the lower reaches. . .’ (MYS.13.3330)
242
(38)
Yuko Yanagida
tosi no pa ni ayu si pasira-ba Sakita-kapa u ya-tu every year sweetfish part run-cond Sakita-river cormorant eight-clf kadukete kapase tadune-mu make.dive river.stream search-aux ‘Each year when the sweetfish run, making many cormorants dive, we shall scour rivers and streams’ (MYS.19.4158)
In OJ, the numeral quantifier ya-tu ‘eight-CLF’ is ambiguous; it can denote a precise cardinality or a non-specific sense of ‘many’. Example (37) describes the eight fishing cormorants in the upper reaches (specific) and eight cormorants in the lower reaches (specific). This interpretation is consistent with the presence of wo on the host noun u wo ‘cormorant’. In contrast, ya-tu ‘eight-CLF’ in (38) denotes a non-specific sense of ‘many’, hence the absence of wo on the host NP. Finally, in languages in which specificity plays an important role in object marking, specific objects tend to move out of VP, while non-specific objects remain in situ (cf. Diesing 1992). We find this pattern in OJ. Wo-marked objects necessarily move over the ga-marked subject, resulting in the configuration [Object=wo Subject=ga V]. This is illustrated in (39).8 (39) a. aki-yama wo ikani ka kimi ga pitori kwoyu-ramu autumn-mountain-obj how-Q you agt alone cross-aux ‘How do you cross the autumn mountain alone?’ (MYS 2.106) b. ware wo yamwi ni ya imo ga kwopwi-tutu aru ramu? I obj dark loc Q wife agt longing.for-cont be aux ‘Would my wife be longing for me in the dark?’ (MYS 15.3669)
As discussed in Section 2.2.4, this particular OSV order is characteristic of non-accusative alignment. The subject appears in the external argument position, namely, Spec vP (see [12]). Since v does not assign structural case, the object moves outside vP. In the next section, I will strengthen the differential object marking hypothesis by providing a close inspection of the two prose texts, Senmyō and Norito unplagued by metrical questions.
5.3 Senmyō and Norito The preceding analysis was primarily based on the Man’yōshū, a collection of poems with versification restrictions. By examining the two major prose texts in OJ, Senmyō
8 See Yanagida (2007a) for all the examples with OSV order in the Man’yōshū. For one possible counterexample to this word order restriction, see Appendix II.
10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax
243
(Shokunihongi Senmyō) and Norito (Engishiki Norito), this section attempts to show that the skewed distribution of wo is not due to poetic versification, as proposed by Kinsui (2011). Shokunihongi Senmyō is comprised of the sixty-two imperial edicts preserved in the Shokunihongi (Chronicles of Japan), an imperially commissioned Japanese history text completed in 797 CE. Volume VIII of the Engisiki Norito was compiled in the tenth century and contains Shinto rituals and practices in their pristine form. Obviously, the origin of these rituals dates to a much earlier period than that of the Norito’s compilation, and their composition is believed to reflect the language of the Nara period. The writing style of the Senmyō and Norito differs from the Man’yōshū insofar as it uses a set of writing conventions known as Senmyōgaki. In Senmyōgaki, grammatical particles, auxiliaries, and verb endings are, in some manuscripts, written phonographically in a smaller size. Lexical/content words, such as nouns and verbs, are written logographically in a larger size. Although the Senmyō contains a hybrid of phonogrammatic spellings and sections with a superficial Chinese-like style, it is known to reflect the Japanese language of the 8th century. Wrona and Frellesvig (2010) present an extensive study of the distribution of woand zero-marked objects in these two prose texts. Contrary to Miyagawa’s (1989) generalization, quantitative study shows that there is no significant difference in the use of wo vs. zero-marking between adnominal and conclusive clauses. Wrona and Frellesvig suggest that wo- and zero-marking have no semantic effects either and conclude that zero-marked objects are simply analyzed as stylistic case drops. However, one problem with the Senmyō that Wrona and Frellesvig fail to address has to do with the nature of the text. Rather than being a complete and explicit transcript of imperial proclamations, Senmyō texts contain characters of shorthand guides or notes for the orally pronounced proclamation, to be read out loud by specialized officials. They therefore omit some functional morphemes that are to be supplied by the reader, potentially including accusative wo. Thus, passages without wo in the written text do not necessarily correspond to zero-marked objects. The annotated versions of the Senmyō with so-called yomisoe ‘supplied readings’ are based on the original interpretation of Motoori (1803), and all subsequent annotated texts basically follow Motoori’s annotation. According to Ikeda (1996), Motoori’s (1803) text contains 83 tokens of supplied wo. Kitagawa’s (1982) version of the Senmyō text, on which Wrona and Frellesvig’s analysis is based, contains 85 tokens of supplied wo. Neither Motoori nor Kitagawa, however, provide explanations as to why wo is supplied in certain cases, but not in others. Given that the exact basis for yomisoe readings has never been made clear, it is extremely difficult to determine what counts as a zero-marked object. This problem is less significant in the Man’yōshū, because in most cases, the supplied wo occurs when the object is logographically written or in the kanbun ‘Chinese’ style. Crucially, the Man’yōshū, as a poetic text, follows the rules of Japanese versification, which generally require phrases to be arranged in fiveor seven-syllable phrases. This makes it possible to predict with some degree of accuracy whether objects without phonographic wo should, in fact, be read with wo.
244
Yuko Yanagida
When counting the number of wo-marked and phonographically “unmarked” objects in the Norito and Senmyō, I found that the ratio between wo-marked and unmarked objects was similar to Wrona and Frellesvig’s.9 (40) Wo-Marked Objects Senmyō Norito 498 261
Unmarked Objects Senmyō Norito 256 166
Quantitative data for Senmyō are from Kitagawa (1982), and those for Norito were taken from Kurano and Takeda (1958). The unmarked objects include tokens of supplied wo. In both the Senmyō (SM) and Norito (NT), wo-marked objects are consistently interpreted as specific. Some examples are given in (41)–(42) in which the supplied particles are in the parenthesis. (41)
a.
b.
tare (si) (no) yatukwo ka wa(ga) mikadwo who (foc) (gen) retainer Q 1p (gen) emperor wo somuki-te. . . sika suru obj betray-inf this do.adn ‘Whose retainer betrays my emperor. . . and acts in this way.’ (SM: Edict 16) pito-tu mo ware wo uramu-beki koto pa omopoye-zu one-clf foc 1p obj hate-aux that top think-not ‘I didn’t think anyone would hate me.’ (SM: Edict 16)
(42) a. yomo-(no)-kuni wo yasu-kuni to tapirakeku four-(gen)-country obj peaceful-country as tranquil irosimyesu ga yuwe ni rule gen because loc ‘because [I] rule the country in peace, as a tranquil nation’ b. opomitakara no tukuri tukuru mono wo. . . people gen grow.inf grow.adn thing obj nasi-tamapa-zu sokonapyeru pa. . . do-hon-not harm top ‘not allowing what [my] subjects grow to ripen, and doing harm’
(EN: 1)
(EN: 4)
The personal pronouns and the possessor + NP in (41–42) are inherently specific. In (42b), tukuru mono ‘crops’ is marked by wo because it refers to specific crops grown by the people of the nation (mentioned in the previous sentence). The Norito uses many instances of the expression [kusagusa Gen NP] ‘various/many NPs’. The NPs marked by
9 I ignored 是以, since there are many tokens which may or may not appear with wo. Ikeda’s (1996) data on Kitagawa (1982) excludes this phrase.
10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax
245
wo are unambiguously specific, while zero-marked NPs are non-specific. This is illustrated in (43) and (44). (43) sumyegamwi no mapye ni sirwoki uma sirwoki wi sirwoki sovereign.deity gen before loc white horse white boar white tori kusagusa (no) iromono wo sonape maturite cock various (gen) things obj prepare enshrine ‘Before the sovereign deities of the Grains, I will prepare for and provide various kinds of offerings such things as the white horse, the white boar and the white cock.’ (EN: 1) (44) mima ni mikura Ø sonapete kusagusa no mitegura Ø sonapete horse loc saddle provide many gen offerings prepare for ‘Providing a saddle for the horse and preparing for various kinds of offerings (for the deity).’ (EN: 4) In (43), the set of the items offered to the deity is defined by the previous context, while in (44), the NP mitegura ‘offering’ is not defined by the previous discourse, which explains the absence of wo. N.B. opo-ya-sima-no kuni ‘the islands of Japan’ is always used in the unmarked form in the preverbal position of the verb form sira- ‘govern’. But wo shows up when it is specific and moved out of VP, as shown in the contrast between (45) and (46). (45) Sikwi-sima ni opo-ya-simaguni-Ø sirasi-si sumyemima (no) mikoto Shiki-island loc large-eight-island govern-pst god (gen) Son ‘The Son of God (emperor) that governs the Shiki Island.’ (EN: 4) (46)
akitu mikamwi no opo-ya-simaguni wo. . . tapirakeku living Emperor gen large-eight-island obj. . . peacefully sirosimyesa-mu koto govern-aux that ‘That the emperor governs the islands of Japan peacefully.’
(EN: 27)
In (45), opo-ya-simaguni appears immediately adjacent to the verb sira- ‘govern’. The NV complex is interpreted as the predicate ‘govern (lands) in general’. In (46), opo-ya-simaguni refers to the specific islands of Japan that are governed by the emperor. Close examination, however, reveals that the bare objects that appear in Senmyō behave quite differently from those in the Norito. The Norito exhibits the same pattern as the Man’yōshū, insofar as preverbal bare objects in adnominal clauses invariably receive non-specific interpretations, as illustrated in (47).
246
(47)
Yuko Yanagida
a.
b.
sumye-mima-(no)-mikoto no oponipe Ø kikosimyesa-mu imperial (gen) Highness gen harvest partake-aux tame (no) yuwe ni reason (gen) for loc ‘in order that the Emperor partakes of rice harvest’ tasuki Ø kakuru tomo no wo wo. . . magapi Ø sash put.on companion gen man obj mistake nasa-sime-zusite do-aux-not.ger ‘making sure that officials who put on the sash do not go wrong’
(EN: 14)
(EN: 8)
The Senmyō, on the other hand, contains many counterexamples. That is, in (48), the bare objects of the adnominal verbs are unambiguously interpreted as specific. (48)
a.
b.
c.
opodi opo-mapye tu kimi no tonokadwo Ø grandfather great gen lord gen dignitary.gate arasikegasu koto naku defile that not ‘without defiling the gates of dignitaries and ancestors’ (SM: Edict 13) imperial gen lord protect serve-hon sumyera ga mikadwo Ø mamori tukape-maturu koto kapyerimi naki pito-domo that look.back not people-pl ‘people who selflessly serve and protect the emperor’ (SM: Edict 13) Nakamaro ga ipye no mono Ø kazwopuru ni pumi no NM gen house gen thing examine.adn loc letter gen naka ni Nakamaro to kaywopasi-kyeru pakarikoto no inside loc NM with lay-pst conspiracy gen pumi ari letter exist ‘When (the emperor) examined things in Nakamaro’s house, among the letters was a secret letter in which (he) laid a conspiracy with Nakamaro.’ (SM: Edict 30)
A question then arises as to why Senmyō does not pattern in the same way as Norito. It is important to note that Shoku Nihongi, in which Senmyō is included, is about the chronological history of the ritsuryō sei replicating China’s political system from the Tang Dynasty. Kotani (1986) argues that the reason for the different writing styles in Senmyō and Norito lies in the fact that they have different origins. According to Kotani, Senmyō was composed on the basis of Shōchoku ‘imperial rescript’ and written in Old Chinese. Thus, these texts contain many sentences in the kanbun ‘Chinese’ style. Even sentences in the Japanese style are based on kanbun, then transcribed from kanbun
10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax
247
to Japanese by changing the word order and adding particles or verb endings to their original kanbun counterparts. Norito, on the other hand, originates in the oral tradition, which takes the form of folktales, songs, or chants, and thus reflects a genuine oral language. Although we must leave open the issue concerning yomisoe ‘supplied reading’, it is plausible to assume that the objects with specific interpretations, such as (48), are the ones presumably read with the supplied wo. Norito, on the other hand, does not share the shorthand character of Senmyō texts. It therefore provides more reliable evidence as to the presence or absence of =wo. Examination of Norito texts shows that preverbal bare objects pattern like Man’yōshū in that they receive non-specific interpretations.
6 Conclusion This chapter has investigated two distinct levels of differential argument marking attested to in Old Japanese. Differential subject marking is associated with the semantic role assigned by the verb; specifically, agentive subjects are marked by ga, while non-agentive subjects are marked by zero or the other genitive, no. The use of ga vis-àvis zero/no is sensitive to the subject’s position on the nominal hierarchy. The human NPs higher up on the hierarchy are associated with prototypical agents, which express volition and control, while the non-human or inanimate NPs lower down on the hierarchy are not transitivity prototypes. The OJ data showed that transitivity is a clause-level phenomenon defined as the type of NPs that serve as grammatical subjects. Differential object marking in OJ, on the other hand, is associated with a specific/non-specific distinction of object NPs. The distinctive [O wo S ga V] pattern of transitive clauses is consistent with the view that objects marked by wo are specific and that specific objects move outside VP.
Appendix I Other ‘nominalized’ clause types, which show the same patterns, are inflected in the provisional (49), conditional (50), and nominal form in –(a)ku (51). (49) Realis (izenkei) conditional wa ga wore-ba ura sipo miti-ku I agt be-prov bay tide be.full-come ‘When I was present the tide was high in the bay.’
(MYS 15.3707)
248
(50)
Yuko Yanagida
Irrealis (mizenkei) conditional masakikute imo ga ipapa-ba safely wife agt bless-cond ‘if you bless me godspeed’
(51) V-aku Nominal form wotomye-ra ga ime ni tugur-aku maiden-pl agt dream loc recount-nmlz ‘what the maidens recounted in my dream’
(MYS 15.3583)
(MYS 17.4011)
Each of the nominalized clause types in (49–51) shares the active alignment properties of the adnominal clauses; the external arguments of transitive verbs are marked by ga, but not by zero.
Appendix II As Yanagida (2006: 61) indicates, there is one possible counterexample in which the subject marked by ga is followed by the wo-marked object. This is illustrated in (52). (52) yama no na to ipi-tugye to kamo Saywopimye ga hill gen name as say-tell that Q Sayohime agt kono yama no pe ni pire wo (遠) puri-kyemu this hill gen upon loc sash obj wave-aux ‘Was it for transmitting as the name of the mountain that Sayohime waved a sash upon this hill.’ (MYS 5.872) The man’yōgana 遠 in (52) is read wo. Yanagida (2006) suggests the possibility that this character is used to write the noun 緒 wo ‘long cloth/string/thong’, in which case, pirewo is a compound noun (‘long sash’) functioning as the object of the verb puri ‘wave’. Kuroda (2008), however, casts doubt on Yanagida’s proposal, suggesting an alternative interpretation for MYS 5.872. He notes that “the part of Man’yōshū in question is thought to originate in the collection of poems kept by Yamanoue Okura, one of the prominent poets of Man’yōshū, who, however, is believed to be a non-native speaker of Japanese. This fact may be relevant for this counterexample” (Kuroda 2008: 282). However, close examination of the Norito (Engishiki Norito) shows that there is a possibility that pirewo in Man’yōshū (872) is used as engo ‘related word’, associated with pire kakuru tomono wo (比禮懸伴緒)in the Norito (53).
10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax
249
(53) sumyemima (no) mikoto (no) asa no mike yupu no imperial (gen) Emperor (gen) morning gen meal evening gen mike (ni) tukape maturu pire (比禮) Ø kakuru tomo(no) (懸伴) wo (緒) meal (dat) serve-give.adn sash put.on fellow (gen) head tasuki Ø kakuru tomo(no) wo wo te (no) magapi sash put.on group (gen) head obj hand (gen) mistake asi (no) magapi Ø nasa-sime-zute foot (gen) mistake do-caus-not ‘As for the head of the group of the women, who serves for the emperor’s morning and evening meals, and who puts her sash on over her shoulder, not making any mistakes with her hands and her feet.’ (EN: 9)
Tsugita (2008: 262–264) points out that the word pire, which appears in the Man’yōshū, as in (52) and in the Norito in (53), both refer to the long sashes symbolically worn by noble women in the Nara period. According to Tsugita, the women who serve meals for the emperor ritually put sashes over their shoulder. The word wo 緒 in (53), which originally means long cloth/string/thong, here is in reference to the head of the group of women who serve the emperor his meals. The use of pire in a combination with the noun wo 緒 in the Norito favors Yanagida’s (2006) original interpretation in that wo in MYS 5.872 is used, not to write the case particle, but rather the noun 緒.
References Texts (Primary Sources) Kitagawa, Kazuhide. 1982. Shoku nihongi senmyō. Tokyo: Yoshikawa Kōbunkan. Koji, Kazuteru. 1988. Man’yōshū joshi no kenkyū [A study of particles in Man’yōshū]. Tokyo: Kasama Shoin. Kojima, Noriyuki, Masatake Kinoshita and Haruyuki Tōno. 1994–1996. Man’yōshū (1–4), Nihon Koten Bungaku Zenshū. Tokyo: Shogakkan. Kurano, Kenji and Yukichi Takeda. 1958. Kojiki norito. Nihon Koten Bungaku Taikei. Tokyo: Iwanami Shoten. Motoori, Norinaga. 1803. Shokki Rekichō Shōshikai [Commentaries on imperial edicts in the Shoku Nihongi]. Nakanishi, Susumu. 1978–1983. Man’yōshū. Tokyo: Kōdansha Bunko (reprinted in 1978–2005).
Digitalized Texts The Corpus of Historical Japanese (CHJ), the National Institute of Japanese Language and Linguistics, https://maro.ninjal.ac.jp/ The Oxford-NINJAL Corpus of Old Japanese (ONCOJ), https://oncoj.ninjal.ac.jp/
250
Yuko Yanagida
Secondary Sources Aldridge, Edith. 2004. Ergativity and word order in Austronesian Languages. Ithaca, NY: Cornell University dissertation. Aldridge, Edith. 2008. Generative approaches to ergativity. Language and Linguistics Compass: Syntax and Morphology 2(5). 966–995. Aldridge, Edith. 2015. A minimalist approach to the emergence of ergativity in Austronesian languages. Linguistics Vanguard 1(1). 313–326. Alexiadou, Artemis and Elena Anagnostopoulou. 2001. The subject-in-situ generalization and the role of case in driving computations. Linguistic Inquiry 32(2). 193–231. Anand, Pranav and Andrew Nevins. 2006. The locus of ergative case assignment: Evidence from scope. In Alana Johns, Diane Massam and Juvenal Ndayiragije (eds.), Ergativity: Emerging issues, 3–25. Dordrecht: Springer. Authier, Gilles. 2013. The origin of two genitive cases and inalienability split in Budugh (East Caucasian). Faits de Langues 41. 177–192. Baker, Mark C. 1988. Incorporation. Chicago: The University of Chicago Press. Baker, Mark C. 2014. On dependent ergative case (in Shipibo) and its derivation by phase. Linguistic Inquiry 45. 341–379. Bittner, Maria and Ken Hale. 1996. The structural determination of case and agreement. Linguistic Inquiry 27. 1–68. Bynon, Theodora. 2005. Evidential, raised possessor, and the historical source of the ergative construction in Indo-Iranian. Transactions of the Philological Society, Vol. 103(1). 1–72. Comrie, Bernard. 1978. Ergativity. In Lehmann Winfred (ed.), Syntactic typology, 329–394. Austin: University of Texas Press. Coon, Jessica. 2008. When ergative=genitive: Nominals and split ergativity. In Natasha Abner and Jason Bishop (eds.), Proceedings of the 27th West Coast Conference on Formal Linguistics, 99–107. Somerville, MA: Cascadilla Proceedings Project. Coon, Jessica and Andrés Pablo Salanova. 2009. Nominalization and predicate-fronting: Two sources of ergativity. University of Pennsylvania working papers in linguistics 15(1). 45–54. Diesing, Molly. 1992. Indefinites. Cambridge, MA: MIT Press. Dixon, R. M. W. 1979. Ergativity. Language 55. 59–138. Dixon, R. M. W. 1994. Ergativity. Cambridge: Cambridge University Press. Franchetto, Bruna. 1990. Ergativity and nominality in Kuikúro and other Carib languages. In Doris L. Payne (ed.), Amazonian linguistics: Studies in lowland South American languages, 407–428. Austin: University of Texas Press. Frellesvig, Bjarke, Stephen Horn and Yuko Yanagida. 2015. Differential object marking in Old Japanese: A corpus based study. In Dag Haug et al. (eds.), Historical linguistics: Current issues in linguistic theory, 195–211. Amsterdam: John Benjamins. Gildea, Spike. 1998. On reconstructing grammar. Oxford: Oxford University Press. Gildea, Spike. 2000. On the genesis of the verb phrase in Cariban languages. In Spike Gildea (ed.), Reconstructing grammar: Comparative linguistics and grammaticalization, 65–106. Amsterdam: John Benjamins. Haig, Geoffrey. 2008. Alignment change in Iranian languages: A construction grammar approach. Berlin/New York: Mouton de Gruyter. Ikeda, Yukie. 1996. Senmyō no wo kaku hyōji [The objective case wo in Senmyō]. Machikaneyama ronsō 30. 19–32. Osaka: Osaka University. Johns, Alana. 1992. Deriving ergativity. Linguistic Inquiry 23(1). 57–87. Kageyama, Taro. 1980. Goi no kōzō [The structure of lexicon]. Tokyo: Shohakusha.
10 Differential argument marking in Old Japanese: Morphology, semantics, and syntax
251
Kaufman, Daniel. 2009. Austronesian nominalism and its consequences: A Tagalog case study. Theoretical Linguistics 35(1). 1–49. Kikuta, Chiharu. 2012. Jōdai nihongo no ga kaku ni tsuite [On the case marker ga in Old Japanese]. Dōshisha daigaku jinbun gakkai (The Literary Association) 89. 89–123. Kyoto: Doshisha University. Kinsui, Satoshi. 1993. Kotengo no wo ni tsuite [On the particle wo in premodern Japanese]. In Yoshio Nitta (ed.) Nihongo no kaku o megutte [Perspectives on case in Japanese], 191–224. Tokyo: Kurosio Publishers. Kinsui, Satoshi. 2011. Tōgoron [Syntax]. In Satoshi Kinsui, Yoshiyuki Takayama, Tomohide Kinuhata and Tomoko Okazaki (eds.), Bunpōshi [The history of grammar], 77–166. Tokyo: Iwanami Shoten. Klimov, Georgij A. 1977. Tipologija jazykov aktivnogo stroja [Typology of languages of the active type]. Moscow: Nauka. Kotani, Hiroyasu. 1986. Mokkan to senmyō no kokugogakuteki kenkyū [A linguistic study of Mokkan and Senmyo]. Osaka: Izumi Shoin. Kuroda, S.-Y. 2008. On the syntax of Old Japanese. In Bjarke Frellesvig, John C. Smith and Masayoshi Shibatani (eds.), Current issues in the history and structure of Japanese, 263–318. Tokyo: Kurosio Publishers. Legate, Julie. 2008. Morphological and abstract case. Linguistic Inquiry 39(1). 55–101. Manning, Christopher D. 1996. Ergativity: Argument structure and grammatical relations. Stanford, CA: CSLI Publications. Meira, Sergio. 2006. Stative verbs vs. nouns in Sateré-Mawé and the Tupian family. In Grazyna J. Rowicka and Eithne B. Carlin (eds.), What’s in a verb: Studies in the verbal morphology of the languages of the Americas (Lot Occasional Series), 189–214. Amsterdam: Netherlands Graduate School of Linguistics. Mithun, Marianne. 1991. Active/agentive case marking and its motivations. Language 67(3). 510–546. Miyagawa, Shigeru. 1989. Structure and case marking in Japanese (Syntax and semantics 22). New York: Academic Press. 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. 1–95. Mohanan, Tara. 1994. Argument structure in Hindi. Stanford: CSLI Publications. Motohashi, Tatsushi. 1989. Case theory and the history of the Japanese language. Tucson, AZ: University of Arizona dissertation. Nomura, Takashi. 1993. Jōdaigo no no to ga ni tsuite [On the particles no and ga in Old Japanese]. Kokugo Kokubun 62. 1–17. Ohno, Susumu. 1977. Shukaku joshi ga no seiritsu [The development of the nominative case particle ga]. Bungaku 45. 102–117. Rezac, Milan, Pablo Albizu and Ricardo Etxepare. 2014. The structural ergative of Basque and the theory of Case. Natural Language & Linguistic Theory 32. 1273–1330. Sapir, Edward. 1911. The problem of noun incorporation in American languages. American Anthropologist 13. 250–282. Silverstein, Michael. 1976. Hierarchy of features and ergativity. In R. M. W. Dixon (ed.), Grammatical categories in Australian languages, 112–171. Canberra: Australian Institute of Aboriginal Studies. Takeuchi, Shiro. 2008. Kodai nihongo no kakujoshi wo no hyōji-iki to sono henka [The distribution and historical change of the case particle wo in earlier Japanese]. Kokugo to Kokubungaku 85(4). 50–63. Tsugita, Uruu. 2008. Shin-pan Norito shinkō [A new lecture on the Norito]. Tokyo: Ebisukosyo Publication. (Originally published in 1927, Tokyo.) Tsuta, Kiyoyuki. 2004. Mi no sekai [The world of mi]. Kokugo Kokubun 73(12). 10–29. Velazquez-Castillo, Maura. 1996. The grammar of possession. Amsterdam: John Benjamins. Vovin, Alexander. 1997. On the syntactic typology of Old Japanese. Journal of East Asian Linguistics 6. 273–290. Whitman, John. 2008. The source of the bigrade conjugation and stem shape in pre-Old Japanese. In Bjarke Frellesvig and John Whitman (eds.), Proto-Japanese, 159–174. Amsterdam: John Benjamins. Woolford, Ellen. 1997. Four-way case systems: Ergative, nominative, objective, and accusative. Natural Language & Linguistic Theory 15. 181–227.
252
Yuko Yanagida
Woolford, Ellen. 2008. Differential subject marking at argument structure, syntax and PF. In Helen de Hoop and Peter de Swart (eds.), Differential subject marking, 17–40. Dordrecht: Springer. Wrona, Janick and Bjarke Frellesvig. 2010. The Old Japanese case system: The function of wo. In Shoichi Iwasaki, Hajime Hoji, Patricia M. Clancy and Sung-Ock Sohn (eds.), Japanese/Korean Linguistics 17. 565–580. Stanford: CSLI Publications. Yamada, Yoshio. 1968. Narachō bunpō shi [The history of Nara grammar]. Tokyo: Hobunkan. Yanagida, Yuko. 2006. Word order and clause structure in Early Old Japanese. Journal of East Asian Linguistics 15. 37–68. Yanagida, Yuko. 2007a. Jōdaigo no nōkakusei ni tsuite [On ergativity in Old Japanese]. In Nobuko Hasegawa (ed.), Nihongo no shubun genshō [Main clause phenomena in Japanese], 147–188. Tokyo: Hituzi Shobo. Yanagida, Yuko. 2007b. Miyagawa’s (1989) exceptions: An ergative analysis. MIT Working Papers in Linguistics 55. 265–276. Yanagida, Yuko. 2018a. Differential argument marking and object movement in Old Japanese. In Kunio Nishiyama, Hideki Kishimoto and Edith Aldridge (eds.), Topics in theoretical Asian linguistics, 181–205. Amsterdam: John Benjamins. Yanagida, Yuko. 2018b. Differential subject marking and its demise in the history of Japanese. In Ilja Seržant and A. Witzlack-Makarevich (eds.), Diachrony of differential argument marking, 403–425. Berlin: Language Science Press. Yanagida, Yuko. 2022. The origin of dative subjects and psych predicate constructions in Japanese. Journal of Historical Linguistics 12(2). 282–316. Yanagida, Yuko and John Whitman. 2009. Alignment and word order in Old Japanese. Journal of East Asian Linguistics 18. 101–144.
Yoshiyuki Takayama
11 The syntax and morphology of Early Middle Japanese 1 Introduction This chapter discusses the syntax and morphology of Early Middle Japanese (hereafter EMJ). First, it explains the sentence structure of EMJ while comparing it with Modern Japanese (ModJ). Because auxiliaries play an important role in the layered structure of Japanese sentences, this chapter discusses “modality” and “copula” with a focus on auxiliaries. These two research topics have accumulated descriptive analyses among research on auxiliaries since the Edo period. Each is important in the syntax and morphology of EMJ. Based on the results of investigations and observations in wabun texts, the chapter explains the grammatical features of the modality and copula forms.
2 The layered structure of sentences The auxiliaries of EMJ are shown in Table 1. Table 1: The Auxiliaries of EMJ.1 Grammatical meaning
Auxiliary (affix)
Passive Causative Past Perfective Stative Negative Conjectural, Volition Present Conjectural Past Conjectural Subjunctive Evidential Negative Conjectural
ru, raru su, sasu, simu ki (simple past), keri (modal past) tu, nu tari, ri zu mu ramu kemu masi meri, nari (hearsay)1 mazi, zi
1 There are several types of nari. We distinguish them by the form to which they are connected. The one connected to the adnominal form (rentaikei) is called “rentai-nari”, and the one connected to a noun is called “taigen-nari”. Taigen is a traditional grammatical term designating a noun. These are copulas. The one that is connected to the conclusive form (shūshikei) is called “shūshi-nari”. This is a modality form that represents “hearsay”. https://doi.org/10.1515/9781614512851-012
254
Yoshiyuki Takayama
Table 1 (continued) Grammatical meaning
Auxiliary (affix)
Deontic Optative Respect
besi mawosi (< maposi), tasi ru, raru, su, sasu, simu
The auxiliaries in Table 1 express grammatical categories (ru, raru, su, sasu, and simu can also be used to express respect). The semantic structure of Japanese sentences has been found to be hierarchical. In ModJ, the structure is as shown in (1). (1) ModJ [[[[[[[stem]voice]aspect]polarity]tense]judgmental modality]communicative modality] (kagi ga) kake rare tei nakat ta daroo ne (key nom) lock pass perf neg pst conjec sfp ‘It wouldn’t have been locked In ModJ, grammatical categories below “voice” comprise a hierarchy. There are two types of modality: modality regarding the situation (“inter-propositional modality”) and modality regarding the listener (“inter-personal modality”). Auxiliaries handle inter-propositional modality, while sentence-final particles handle inter-personal modality. Next, the structure of EMJ is as shown in (2). (2) EMJ [[[[[stem]voice]aspect]polarity/tense/judgmental modality]communicative modality] kake-rare-tu-ramu na lock-pass-perf-conjec sfp ‘It would have been locked As in ModJ, a layered structure is found in EMJ. However, tense, negation, and inter-propositional modality are integral and undifferentiated. Among the auxiliaries in Table 1, several have undifferentiated grammatical meanings. Kemu expresses the meaning of ‘past’ and ‘conjectural’. Mazi and zi express ‘negation’ and ‘conjectural’. Moreover, keri can express ‘past’ and ‘hearsay’.2 (3) a.
Titose ya sugi-ni-kemu2 Thousand.years q pass-perf-pstconj.concl ‘Has it been a thousand years?’
(Tosa: 55)
2 The source of the example sentences in this chapter follow the Shinpen Nihon Kotenbungaku Zenshū (Shōgakukan). The numbers in parentheses indicate page numbers.
11 The syntax and morphology of Early Middle Japanese
b.
c.
255
Issyau no padi kore ni suguru pa ara-zi my.life gen shame this dat exceed.adn top exist-nconj.concl ‘There will be no more shame in my life’ (Taketori: 36) Mukasi wotoko ari-keri old man.nom exist-mpst.concl ‘I heard that there used to be a man’ (Ise: 120)
The hierarchisation of sentence structures appears thus to have advanced over time.
3 Modality In sentences, elements expressing recognition or judgement regarding a situation are called modality forms. The modality forms of EMJ are besi, mazi, meri, nari, mu, ramu, kemu, masi, and zi. Descriptions of meanings and usage for each individual form have accumulated among research on auxiliaries up through the present. Whether it be ModJ or classical language, research on the modality of Japanese has been influenced by research on modals. Representative modal auxiliary research includes Coates (1983) and Palmer (2001). English modal auxiliaries such as may, can, must, and will have both deontic and epistemic usages. This paper uses these studies and research on the modality of ModJ for reference while attempting to systematically describe the modality forms of EMJ. It clarifies the grammatical features of modality forms from the viewpoints of conjunctions of auxiliaries and relations to conditional clauses and interrogative sentences.3
3.1 Combinations of auxiliaries 3.1.1 Tense Modality expresses the speaker’s judgement at the moment of utterance. Whether or not tense postposes is an important indicator for classifying modality forms. With besi, mazi, meri, and shūshi-nari, the tense postposes, but with mu, ramu, kemu, masi, and zi, the tense form does not postpose. (4) a.
Potopoto suke mo si-tamawi-tu-bekari-ki kasi nearly become.priest foc do-resp-perf-deo-spst sfp ‘He almost became a priest’ (Genji, Tenarai: 6–358)
3 The following description is based on research in Takayama (2002).
256
b.
c.
d.
Yoshiyuki Takayama
Yo no pito no yurusi-kikoyu-mazikari-si ni yorite world gen people gen agree-hum-nconj-spst dat depend.ger ‘Because society would not approve’ (Genji, Momijinoga: 1–328) Mononozoki no kokoro mo same-nu-meri-ki peeping gen interest foc lose.interest-perf-evid-spst.concl ‘Her curiosity did not seem to be at all dampened’ (Genji, Yūgao: 1–150) Itodo ureu-nari-turu yuki kakitare imiziu very get.depressed-evid-perf.adn snow down heavy.inf puri-keri fall-mpst.concl ‘The melancholic snow was now falling faster and faster’ (Genji, Suetsumuhana: 1–291)
3.1.2 Modality In some cases of EMJ, modality forms combine. Such combinations of auxiliaries include besi+meri, besi+mu, and mazi+nari. When modality forms combine, besi and mazi always come first. (5) a.
b.
c.
Kokorobase aru wakaudo wa minisimite Knowledge exist.adn young.women top soak.into.the.body.ger omou-beka-meri think-deo-evid.concl ‘Young women who understand art seem deeply moved’ (Genji, Akashi: 2–240) Ikaga akasi-kurasu-bekara-mu how pass-live-deo-conjec.concl ‘How can we live every day’ (Genji, Usugumo: 2–429) Sarani tagau-mazika-naru mono-o absolutely wrong-nconj.adn-evid.adn conjct ‘Buddha’s guidance will never be wrong’ (Genji, Wakamurasaki: 1–215)
3.1.3 Negation In EMJ, negation is expressed by the auxiliary zu. When zu preposes in modality form, it expresses negation of the statement (proposition). Usually, zu does not postpose in the modality form and the modality itself is not negated. However, besi is an exception, as in the example shown in (6).
11 The syntax and morphology of Early Middle Japanese
257
(6) Nokori no inoti, iti-putuka o mo osima-zu wa remaining gen years, one-two.days acc foc hesitate-neg top aru-bekara-zu exist.adn-deo-neg.concl ‘It can only be a question of prolonging life for a few days at most’ (Genji, Tenarai: 6–285) The bekara-zu form is found in many kanbun materials but is rarely used in wabun. Besi appears about 3,000 times in the Genji monogatari, but bekara-zu only in the example shown in (6). Besi, similar to must in English, has both deontic ‘obligation’ and epistemic ‘presumption’ meanings. However, when a negation postposes besi, the meaning expressed is always deontic. Mazi and zi express a negative conjectural and include negation within the form’s meaning. Furthermore, zu does not form conjunctions.
3.2 Conditional clauses A typical conditional clause in EMJ is the form known as “irrealis stem + ba”. However, when conjugating adjective forms, it becomes “infinitive + wa”. The occurrence of modality forms in conditional clauses is limited to besi, mazi, and masi, while no others occur in conditional clauses. (7) a.
Nao, siwite noti no yo no on-utagawi also, by.all.means later gen year gen hon-doubt nokoru-beku wa remain-deo.inf top ‘If you feel worried about your future’ (Genji, Wakana, jō: 4–48) b. Sa dani aru-maziku wa such even exist.adn-nconj.inf top ‘If you don’t even have to worry about that’ (Genji, Yadorigi: 5–426) c. Mukasi no kokoro-nara-masikaba, utate old gen heart-cop-sbjv.prov, unpleasant.adv kokorootori-se-masi o disappointment-do-sbjv sfp ‘If I were old, I would have been very disappointed with my expectations’ (Genji, Wakana, jō: 4–74)
The occurrence of modality forms in conclusions of conditional sentences is also limited. Meri and nari ‘hearsay’ or ramu and kemu are unlikely to occur in conclusions. For more details, see Takayama (2002).
258
Yoshiyuki Takayama
3.3 Interrogative sentences There are various types of interrogative sentences. A typical interrogative sentence expresses the speaker’s doubt. Elsewhere, sentences that emphasise a declarative assertion in the form of an interrogative sentence are called irony sentences, which are rhetorical questions. Besi, mu, ramu, kemu, and masi occur in typical interrogative sentences. Besi occurs in interrogative sentences but tends to be used in irony sentences. Meri, shūshi-nari, mazi, and zi do not occur in typical interrogative sentences but are observed in irony sentences. (8) a.
b.
c.
d.
e.
Nani-waza site ka wa kurasu-beki anything do.ger q top spend.a.day.adn-deo.adn ‘What to do for a day’ (Genji, Wakana, jō: 4–136) Anadurawasiki pito nite ya maziri-tamawa-mu treat.unkindly.adn person cop q associate-resp-conjec.concl ‘She might be treated unkindly’ (Genji, Wakamurasaki: 1–241) Nado yamanowi no kake.panaru-ramu why mountain.well gen get.away-pconj.concl ‘Was it the shadows in the mountain well that told you my purpose was but jest?’ (Genji, Wakamurasaki: 1–230) Nani o site kaku owiide-kemu what acc do.ger in.this.way grow.up-pstconj.concl ‘How can they have contrived to grow up like this?’ (Genji, Hahakigi: 1–60) Kudari ya si-na-masi to kanete yori leave.capital q do-perf-sbjv quot before abl obosi-keri think-resp-mpst.concl ‘(Lady Rokujoo) has long thought of leaving the capital with her daughter’ (Genji, Aoi: 2–18)
3.4 Summary of this section Table 2 displays the results so far. Modality forms are divided into two types depending on whether the tense postposes. Mu, ramu, kemu, masi, and zi do not postpose the tense. These correspond to ‘genuine-modality’ (such as daroo) in ModJ. In contrast, besi, mazi, meri and nari ‘hearsay’ allow postposing of tense and correspond to the ‘quasi-modality’ of ModJ (such as rasii,
11 The syntax and morphology of Early Middle Japanese
259
Table 2: Features of EMJ Modality forms. Conjunctions besi mazi meri nari mu ramu kemu masi zi
Tense
Modality
Negation
+ + + + – – – – –
+ + – – – – – – –
+ – – – – – – –
Conditional Clause
Interrogative Sentence
+ + – – – – – – –
+ – – – + + + + –
yoo da, and soo da). In ModJ, the tense postposes in yoo da and rasii but not in daroo. Modality forms also postpose in besi and mazi and can be considered highly inclined to act as suffixes. Meanings of modality forms compose a continuous phase from the strongly situational meaning to the strongly judgemental. In both EMJ and ModJ, there are two types of modality forms. However, EMJ and ModJ differ significantly in the relation between subordinate clauses and modality forms. In EMJ, modality forms frequently occur in subordinate clauses, as in (7). (9) a.
b.
c.
Omowa-mu ko cherished-conjec.adn Son ‘Cherished son’ (Makura no sōshi: 32) Pito mo yuyusiku mi-omou-bekereba person foc hate.inf see-think-deo.prov ‘It will look disgusting to everyone’ (Genji, Sawarabi: 5–358) Tugitugi zun nagaru-meredo one.after.another turn go-evid.conc ‘It seems that poem is being sung one after another’ (Genji, Fujinouraba: 3–439)
Conversely, modality forms are unlikely to occur in subordinate clauses in ModJ. Diachronic changes in the relation between subordinate clauses and modality forms are directly linked to changes in the layered structure and constitute an important issue in grammatical history research.
260
Yoshiyuki Takayama
4 Copula 4.1 Rentai-nari and taigen-nari Nari is the representative copula form of ancient Japanese. The original form of nari was ni ari, found in the Nara period. (10)
Yononaka ya tune-ni life q forever-cop.inf ‘Did life last forever’
ari-kyeru exist-mpst.adn
(MYS 5.804)
The Heian period had two types of nari: one connected to a noun (substantive) (taigen-nari) and one attached to an attributive form (rentai-nari).4 Taigen-nari corresponds to -da in ModJ and rentai-nari to -no da. (11)
a.
b.
o-tubone wa Kiritubo-nari hon-room top Kiritubo-cop.concl ‘Her lodging was in the wing called Kiritsubo’ (Genji, Kiritsubo: 1–20) Onna mo site mi-mu tote suru-nari woman foc do.ger try-vol cop do.adn-cop.concl ‘As a woman, I also want to try’ (Tosa: 15)
Taigen-nari is found as early as the Nara period, while rentai-nari appears in the Heian period. Niari, the original form of nari, was also used in the Heian period. Rentai-nari is important in Heian-period grammar and deeply related to sentence structure. This section focuses on rentai-nari and examines its nature from the viewpoint of conjunctions with auxiliaries, kakari-musubi and relations with subordinate clauses.
4.2 The nature of rentai-nari 4.2.1 Conjunction relations Kitahara (1981) uses the conjunction with rentai-nari as an indicator to classify auxiliaries in EMJ. Auxiliaries overall are divided into three groups according to their conjunction status. Auxiliaries always preposed to rentai-nari are in Group A, those always
4 For more information on the difference between rentai-nari and taigen-nari, see Takayama (2002) Part 2, Chapter 3. The following description is based on its content.
11 The syntax and morphology of Early Middle Japanese
261
postposed are in Group B and those both preposed and postposed are Group C. (12) is a diagram of this relationship. (12)
[
A
] rentai-nari [ [
C
B
]
]
Group A is composed of auxiliaries such as ru, raru, su, sasu, simu, tu, nu, tari, ri, ki, and mazi. Group B includes, among others, mu, ramu, kemu, masi, zi, meri, and nari. Group C is composed of keri, zu, and besi. Auxiliaries in Group A express an ‘objective meaning’, while those in Group B express a ‘subjective meaning’. Group C has a dual nature, expressing both ‘objective meaning’ and ‘subjective meaning’. This is an excellent classification. However, to address the true nature of rentai-nari, more detailed observations and analyses are necessary. Below, we consider the postposing of tense, modality and sentence-final particles to rentai-nari. 4.2.1.1 Tense Tense in this era is expressed through ki and keri. Occasionally, some uses of tu also express the near-past. Keri is postposed to rentai-nari, but ki and tu are not. (13)
a.
b.
pakanaki sama-nari-si on-kaeri empty.adn state-cop-spst.adn hon-reply ‘short and guarded replies’ (Genji, Aoi: 2–19) pigoro sukosi okotaru sama-nari-turu kokoti recently a.little.adv recover.adn state-cop-perf.adn feeling ‘The sick woman who for a few days past had shown some improvement’ (Genji, Aoi: 2–34)
The expression no da in ModJ includes the no datta form. This -ta does not simply express the ‘past’, but also expresses ‘the recognition of the speaker’ ‘narrative past’ as inter-propositional modality. In the same way, keri can be handled as a type of modality. In that narikeri, in which keri is postposed to rentai-nari, can also express these two meanings, it has something in common with the no da of ModJ. (14)
a.
Kono kuruma mo, kono miya o sasite this cart foc, this palace all to.ger kuru-nari-keri to miyu come.adn-cop-mst.concl cop see.concl ‘It seemed that this woman’s cart was also coming here’ (Genji, Yadorigi: 5–488)
262
Yoshiyuki Takayama
b.
Mawe yori iku midu oba, patuse-gawa to Front abl flow.adn water acc.top, Hatsuse-river cop iu-nari-keri call.adn-cop-mpst.concl ‘The river flowing in front was the Hatsuse River’ (Genji, Tamakazura: 3–116)
4.2.1.2 Modality With rentai-nari, modality forms, with the exception of mazi, zi, and masi, postpose, as shown in (13), for example. (15)
a.
b.
c.
Pito matu-nan-meri person wait.adn-cop.adn-evid.concl ‘It looks like I’m waiting for someone’ (Yamato: 382) Ama ni nari-taru-naru-besi priestess dat become-perf-cop-deo.concl ‘Maybe she’s a nun’ (Yamato: 327) Omopi-amari ide-ni-si tama no feelings-overflow escape-perf-spst soul gen aru-nara-mu exist.adn-cop-conjec.concl ‘Maybe there is a soul that escaped from me because I thought of you’ (Ise: 208)
Examples of nameri appear mainly in prose and are not used in waka poems. In contrast, naruramu is used only in waka poems, not in prose. Keri usually expresses tense, but when postposed to rentai-nari, it expresses “exclamatory meaning” and is analysed as inter-propositional modality. Examples of narikeri and nameri abound. 4.2.1.3 Negation There are numerous cases in which the negative affix zu is attached to taigen-nari. However, there are minimal cases where it is attached to rentai-nari since texts generally use the bekinari form (bekinara-neba, bekinara-nedo). It is proper to see that the negative affix is followed by the compound form beki-nari. In example (16c), exceptionally zi postposes rentai-nari. (16)
a.
Nani-bakari pukaki te how-degree good.adn performance ‘It’s not a very good performance’
11 The syntax and morphology of Early Middle Japanese
b.
c.
263
Aramasiu kikoesawagu-beki podo nara-neba In.a.loud.voice blame.loudly-deo.adn degree cop-neg.prov ‘(Menoto) can’t chastise (Genji) loudly, so’ (Genji, Wakamurasaki: 1–244) Mina omou kokoro naki nara-zi all.men have.a.favor.adn heart not cop-nconj.concl ‘I don’t think all of them are downhearted’ (Genji, Tokonatsu: 3–227)
4.2.1.4 Sentence-final particles Sentence-final particles do not postpose to rentai-nari at all, calling for attention to the reason why. The relation between rentai-nari and sentence-final particles is discussed below. In contrast, sentence-final particles postpose with taigen-nari. (17)
a.
b.
Masite aru-maziki koto nari Kasi even.more.adn exist.adn-nconj.adn thing cop sfp ‘Now it is something that should never be better than before’ (Genji, Sakaki: 2–134) Usirom etaki waza nari ya stupid adn act cop sfp ‘It was too ill-assorted a match’ (Genji, Yūgao: 1–145)
Sentence-final particles postpose to the no da of ModJ (such as -no da/yo/ ne/zo). In relation to postposing sentence-final particles, rentai-nari and no da differ.
4.2.2 Kakari-musubi Kakari-musubi is an important construction that characterises ancient Japanese. Kakari-musubi in EMJ occurs through the postpositional particles zo, namu, ya, ka, and koso. Rentai-nari does not serve as the musubi of these particles, but auxiliaries are all kakari-musubi, as is taigen-nari. (18)
a.
b.
Geni zo aware-naru on-arisama-naru really foc sorry-and hon-state-cop.adn ‘It’s a really bad history of a woman’ Geni koso sadamegataki yo-nare really foc unspecified.adn world-cop.excl ‘A world that is truly unspecified’
(Genji, Suma: 2–172)
(Genji, Asagao: 2–473)
Niari, the original form of nari, also generates kakari-musubi. Rentai-nari and niari have a clear division of functions.
264
(19)
Yoshiyuki Takayama
a.
b.
aru pa namida no puru-ni zo ari-keru exist.adn top tear gen fall.adn-cop.inf foc exist-mpst.concl ‘It was my tears that made me happy’ (Ise:129) Urami o ou tumori ni ya ari-kemu jealousy acc receive.adn accumulation cop.inf q cop-pstconj.adn ‘because she was exposed to constant jealousy and ill will’ (Genji, Kiritsubo: 1–17)
4.2.3 Relations with clauses Examples in which rentai-nari occurs inside clauses are extremely rare. Indeed, rentai-nari does not occur inside adnominal clauses. Alternatively, taigen-nari occurs frequently. (20)
a.
b.
Ane-naru pito sister-cop.adn person ‘sister’ (Genji, Hahakigi: 1–96) Nakanakanite nagusame-gataki kesiki-nareba, On.the.contrary soothe-difficult.adn countenance-cop.prov kosirawe-kane-tamau appease-difficult-resp.concl ‘on the contrary, as (Genji) appears inconsolable, it will be difficult to appease him’ (Genji, Usugumo: 2–466) ‘All night long the mournful ceremonies proceeded’ (Genji, Aoi: 2–47)
Table 3 summarizes the grammatical features of rentai-nari. Table 3: Different of ‘Rentai-nari’ and ‘Taigen-nari’. Conjunctions Rentai-nari Taigen-nari
Tense
Modality
Negation
FP
– +
+ +
– +
– +
Kakari-musubi
Subordinate Clause
– +
– +
The nature of rentai-nari differs from that of auxiliaries in general. Moreover, it also differs from the copula taigen-nari, whose nature is actually very similar to that of auxiliaries in general.
11 The syntax and morphology of Early Middle Japanese
265
4.3 Rentai-nari and sentence structure 4.3.1 Similarity with sentence-final particles Here we discuss the relationship between sentence structure and rentai-nari. As shown in (21), rentai-nari receives noun clauses. (21) [ S ]wa [NP]nari. The interior of NP is as shown in (22). (22)
[NP no ― Adnominal]NP nari. [Mi no ito kokoro-uki]nari. myself gen really heart-hard cop ‘It’s really hard for me’
The no of (22) is a case-marking particle which indicates the genitive and does not occur in the main clause. If rentai-nari were included in a noun clause, it should take the form naru, but there are no examples of this at all. The existence of this construction (the no ~ nari construction) proves that rentai-nari receives noun clauses. Some sentence-final particles also receive noun clauses.5 (23) a. Kapa-kaze no suzusiku mo aru-ka river-breeze gen cool.inf foc exist.adn-sfp ‘What a cool river breeze’ (Kokinshū: 170) b. Iku tosi no osiku mo aru kana pass.adn year gen regrettable.inf foc exist.adn sfp ‘The year of drowsiness is truly regrettable’ (Kokinshū: 342) These structures are as shown in (24). (24)
a. b.
[NP no ― Adonominal] ka. [NP no ― Adonominal] kana.
Rentai-nari is treated as an auxiliary in Japanese classical grammar, but its syntactic structure resembles that of sentence-final particles. Moreover, in common with sentence-final particles, it does not occur inside clauses. A clear difference with sentence-final particles is that it allows the postposing of auxiliaries. Research shows that rent-
5 Kitahara (1966) is a foundational work on rentai-nari that introduces many important points. One of them is the relationship with final particles. Unfortunately, the study of final particles has not progressed so much, so it remains a a topic for further research.
266
Yoshiyuki Takayama
ai-nari has an extremely large number of compound usages with auxiliaries, amounting to roughly 70%.6
4.3.2 The viewpoint of clause linkage Let us look at the relationship with clause linkages. Heian-period prose often includes ‘clause linkages’, where multiple clauses are linked together. Rentai-nari is used among them, as shown in example (25). (25)
Nayonayo-to site, warenimoaranu sama-nareba, ito itaku tired-adv do.ger, insane state-cop.prov, very very.inf wakabi-taru pito nite, mono ni young-stat.adn person cop spirit dat kedora-re-nuru-nan-meri to, semukatanaki deprive-pass-perf-cop-evid quot, at.a.loss.adn kokoti-si-tamau feeling-do-resp.concl ‘Genji is at a loss as it seems that Yūgao is tired and insane, and being so very young, she appears to have been afflicted by a spirit.’ (Genji, Yūgao: 1–166)
One clause expresses one situation, so clause linkage expresses multiple situations. Usually, modality forms append a judgement targeting one situation, as shown schematically in (26a), but when the scope is widened through rentai-nari, clause linkages can fall into the modal scope, as illustrated in (26b).7 (26)
One important function of clause linkages is expressing the causal relationship of a situation. (25) is the scene in which Yūgao is robbed of her sanity by a Mononoke. -Ba expresses the basis of judgement and -nite expresses the reason. (25) shows that the situation’s causal relationship and the judgement there of take place visually in a single sentence. While depicting the situation within the scene from the character’s viewpoint (i.e. Prince Genji), it also skilfully depicts his mental state.
6 This data is based on the survey results of Takayama (2021). 7 Ramu and kemu have a wide scope and do not require the scope extension by rentai-nari. For details, see Takayama (2002).
11 The syntax and morphology of Early Middle Japanese
267
4.4 Summary Rentai-nari appears in the position of connecting noun clauses and modality forms and clearly divides the sentence into the “proposition” part and the “modality” part. Moreover, it functions to widen the scope of modality forms. This function enabled judgements to target the causal relationships of situations expressed by clause linkages. As a result, it became possible to report on the physical scene from a character’s viewpoint and simultaneously depict the character’s inner life. However, rentai-nari fell out of use in the mediaeval period. From the pre-war modern period on, no da has come to take on these functions instead.
References Coates, Jennifer. 1983. The semantics of the modal auxiliaries. London: Routledge. Kitahara, Yasuo. 1966. Shūshi-nari to rentai-nari: Sono bunpu to kōzōtekiimi [Final-nari and adnominal-nari: Their distribution and structural meanings]. Kokugo to Kokubungaku 43(9). 55–71. Tokyo: The University of Tokyo Press. Kitahara, Yasuo. 1981. Nihongo jodōshi no kenkyū [A study on Japanese auxiliaries]. Tokyo: Taishukan Publishing. Palmer, Frank R. 2001. Mood and modality. Cambridge: Cambridge University Press. Takayama, Yoshiyuki. 2002. Nihongo modariti no shiteki kenkyū [A historical study on modality of Japanese]. Tokyo: Hituzi Shobo. Takayama, Yoshiyuki. 2021. Rentai-nari no kinō o dō toraeruka [How can the functions of adnominal-nari be captured?]. In Hisashi Noda and Masaru Oda (eds.), Nihongo no rekishiteki taishōbunpō [Japanese historical contrastive grammar]. Osaka: Izumi Shoin.
Hirofumi Aoki
12 Late Middle Japanese grammar 1 Introduction Many grammar changes are observable in the Late Middle Japanese (LMJ) period, a transitional period between Old/Early Middle Japanese (from Old to Early Middle Japanese, abbreviated as OJ/EMJ) and Modern Japanese. This chapter will look first at two morphological changes in predicates: the merger of the adnominal and conclusive forms and the shift of bigrade conjugations to monograde. Following this, two structural changes – the development of the nominative particle ga and the development of the nominalization particle no – will be taken up. The points of divergence for each of these historical changes was in the LMJ period, but, rather than simply stopping with synchronic observations of a single period, this chapter will describe dynamically the diachronic historical changes, extending the scope of view to Modern Japanese. First, let me provide a brief explanation of the descriptive framework and terminology used in this chapter. The terms used to describe the conjugational forms “conclusive form” and “adnominal form” and the names of the conjugational categories “bigrade” and “monograde” derive from the traditional terminology used in kokugogaku ‘national language studies’. Table 1 shows the verbal conjugations of OJ/EMJ and is followed by a description of the morphological forms taken up in this chapter. The portion of the verb that does not change is called the “stem” and the part that changes the “ending”. In Table 1 stems and endings are separated by a hyphen. Verbs for which the stem ends in a consonant and endings all consist of a single vowel are called “Type A”. Quadrigrade and r-irregular verbs belong to this type. Verbs for which the stem ends in a vowel and the endings include the segments ru and re are termed “Type B”. Monograde verbs belong to this category. Normally the verbs belonging to Type B are further divided into “upper monograde” verbs like miru ‘see’ and “lower monograde” verbs like keru ‘kick’, but, since the difference between upper and lower monograde is a difference in the stem (whether it ends in i- or e-) and there is no difference in conjugational forms, this distinction is not made here. The other conjugational classes – n-irregular, k-irregular, s-irregular, lower bigrade, and upper bigrade – all have the Type A characteristics of a consonant-final stem and ending-initial vowel together with the Type B characteristic of endings that include ru and re. Accordingly, they are designated “Type AB”. The conclusive form is the form normally used at the end of a sentence. The adnominal form is the form used when the clause is linked to a noun (modifies a noun) or when the clause itself functions as a nominal. I will term the former the adnominal use and the latter the nominal use. Clauses formed with an adnominal form in the nominal use will be termed nominalized clauses. Items in bold font in (1) below present examples of the conclusive form (1a), the adnominal use of the adnominal form (1b), and the nominal use of the adnominal form (1c). https://doi.org/10.1515/9781614512851-013
Ka no naka no sina ni toriide-te ip-isi that.person gen middle gen class dat single.out-ger say-spst.adn ko no nami nar-am-u kasi to obosiid-u. this gen level cop-conjec-adn sfp comp recall.resp-concl ‘(Genji) recalled, “Might what those men talked about, singling it out as ‘the middle class’, not be this level?”’ (Genji, Hahakigi. SNKBZ 20: 94) Pusi wo pedate-te yo-goto ni koganear-u joint acc separate-ger section-every dat gold exist-adn take wo mituku-ru koto kasanar-in-u. bamboo acc find-adn occasion pile.up-perf-concl ‘The occasions when he found bamboo with gold in every section between the joints mounted up in number.’ (Taketori. SNKBZ 12: 18.) Aki no yo no aku-ru mo sir-azu nak-u Autumn gen night gen dawn-adn etop know-neg.inf cry-ADN musi pa wa ga goto mono ya kanasi-karu ram-u insect top I gen similarly thing q sad-adn pconj-adn ‘The insects that cry without even knowing that the autumn night is ending: Do they seem somehow sad, like me?’ (Kokinshū 197. SNKBZ 11: 98)
2 Merger of the conclusive and adnominal forms In OJ/EMJ the conclusive form was used to conclude a sentence. (2) a.
b.
Omina nurigome no uti ni Kaguyapime wo old.woman plastered.cell gen inside dat Shining.Princess acc idakape-te wor-i. hold.close-ger be.sitting-concl The old woman sat inside the plastered cell, holding the Shining Princess close.’ (Taketori. SNKBZ 12: 68) Sono sato ni ito namame-itar-u wonna-parakara that village dat very be.charming-stat-adn woman-sibling sum-iker-i. live-mpst-concl ‘In that village lived some sisters who were very charming.’ (Ise. SNKBZ 12: 113)
However, beginning around the Insei (cloistered government) period (1086–1185), the adnominal form began to be used in sentence-final position.
272
Hirofumi Aoki
(3) a.
Kasiko n-ar-u omuna no kasira ni kedamono there dat-exist-adn woman gen head dat animal no abura o nur-ite or-u. gen oil acc anoint-ger be-adn ‘A woman who is over there is having her head anointed with animal oil.’ (Sanpōe. SNKBT 31: 94) Kokoro ni zipi ar-ite mi no sai pito ni heart dat mercy exist-ger body gen genius people dat sugure-tar-iker-u. surpass-stat-mpst-adn ‘He had mercy in his heart, and his learning was superior to other people.’ (Konjaku. SNKBZ 36: 432)
b.
Example (3a) is a sentence occurring in a conversation and (3b) is a narrative sentence. From the Muromachi Period (1336–1573) on, concluding a sentence with the adnominal form in this way became quite general. The conclusive form and the adnominal form did merge, but considering the facts, it would be more appropriate to term the phenomenon the “generalization of the sentence-final use of the adnominal form”. Even in earlier Japanese when the sentence-final conclusive form was normal, there were times when the adnominal form was used to conclude a sentence. The following are such examples. (4) a.
b.
Natu-kusa no tuyu-wake-goromo tuke-naku ni summer-grass gen dew-separate-robe wear-neg.nmlz cop.inf wa ga koromo-de no pur-u toki mo na-ki. I gen robe-hand gen dry-adn time etop not.exist-adn ‘Although I am not wearing robes for parting the dew of summer grasses, there is no moment when my cuffs dry.’ (MYS 10.1994) Ika ni ar-u Puse no ura so mo Kokodaku how cop.inf exist-adn Fuse gen inlet foc etop this.much ni kimi ga mise-m-u to ware wo cop.inf my.lord gen show-conj-concl purp I acc todomu-ru. stop-adn ‘How must it be, this inlet of Fuse? That you stop me this much in order to show it.’ (MYS 18.4036)
12 Late Middle Japanese grammar
(5) a.
b.
273
Miko notamap-aku inoti wo sute-te ka no tama no prince say.resp-nmlz life acc discard-ger that gen jewel gen eda moti-te ki-tar-u tote Kaguyapime ni branch hold-ger come-stat-adn comp Kaguyahime dat mise-tatematuri-tamap-e to ip-eba okina mot-ite show-hum-resp-mp comp say-prov old.man hold-ger ir-itar-i enter-stat-concl ‘The Prince said: “It was risking my life that I have brought that jeweled branch here,” he said, “Show this to Kaguyahime for me,” and when he said this, the old man, taking it, went in.’ (Taketori. SNKBZ 12: 29) Kokoromi no pi kaku tukus-itu-reba momidi rehearsal gen day this.way exhaust-perf-prov autumn.leaves no kage ya sauzausi-ku to omow-edo gen shade q inadequate-inf compl think-conc mise-tatematur-am-u no kokoro ni-te youi show-hum-conjec-adn gen intention cop-ger preparation se-sase-tu-ru nado kikoe-tama-u do-caus-perf-adn conjct say.hum-resp-concl ‘“Although I wonder whether the spectacle of the autumn leaves will be lacking because we have done our utmost on the day of its rehearsal, I prepared it [the rehearsal] with the intention of showing it to you”, he said, among other things.’ (Genji, Momiji no ga. SNKBZ 20: 313)
Examples like those in (4) are primarily seen in waka and show deep emotion or exclamations. Examples like those in (5) are found in conversations and they are used when explaining or commenting on a situation. These can be thought of as uses of a nominalized clause to conclude a sentence. That is, they form not normal sentences ending in a predicate but sentences that conclude with a nominal form. Y. Yamada (1908) calls these kantai ‘emotive’ sentences and distinguishes them as a special kind of sentence taking a nominal core from juttai ‘statement’ sentences that are centered on conjugating forms (predicates). Two kinds of “emotives” are recognized. (6) a. “Emotional Emotives” Uruwasi-ki hana kana. beautiful-adn flower sfp ‘What a beautiful flower!’ Hito no onsin mo se-n-u people gen communication also do-neg-adn ‘Not dealing with people!’
274
Hirofumi Aoki
b. “Wishful Emotives” Oi-zu sina-zu no kusuri moga age-neg.concl die-neg.concl gen medicine opt ‘I want a medicine for neither ageing nor dying’ There are two types – one shows deep emotion and one shows desire – but both show this sort of special meaning by ending the sentence with a nominal. Sentences ending with an adnominal form are comparable to these emotive sentences (Y. Yamada calls this the gikanjutsuhō ‘pseudo-emotive predicate construction’. It is natural, therefore, to think of the generalization of the sentence-final use of the adnominal form as the loss of their specialness by these emotive sentences of earlier times. The explanation that the meaning of the use seen in (4) to show exclamation or excess emotion weakened is commonly found. However, it is hard to believe that a grammatical change began from a special use like that in waka and, in fact, the adnominal conclusive was still used for emotional exclamation in the waka of the Kamakura (1185–1333) and Muromachi Periods. Considering this, it would be better to consider the expansion of the use seen in (5) to show explanation or commentary as the source of the generalization of the sentence-final use of the adnominal form. The pattern of a nominalized clause, with the predicate in the adnominal form, with the copula nari attached, the so called “adnominal-nari” pattern, should be noted here. Examples of this pattern are given below. (7) a.
b.
Payate mo ryuu no puk-asur-u nar-i. Paya gust etop dragon gen blow-caus-adn cop-concl quickly kami ni inori-tamap-e. gods dat pray-resp-imp ‘“It is such that the dragon also makes sudden gusts blow. Quickly pray to the gods.”’ (Taketori. SNKBZ 12: 47) Kitune no tukaumatur-u nar-i. Ko no ki no fox gen do.hum-adn cop-concl this gen tree gen moto ni nan tokidoki ayasi-ki waza namu base dat foc sometimes suspicious-adn deed foc si-paber-u. do-pol-adn ‘This is something that the foxes do. Sometimes they do suspicious deeds at the base of this tree.’ (Genji, Tenarai. SNKBZ 25: 283)
This “adnominal-nari” pattern is related to the noda pattern in present day Japanese and is used when explaining some circumstance. That is to say, it expresses exactly the same meaning as the sentence-final adnominal form sentences used to give an explanation in (5).
12 Late Middle Japanese grammar
275
This sort of close semantic approximation is reflected in syntactic characteristics as well. Both constructions have in common the fact that a suppositional auxiliary verb cannot appear in the predicate part of the adnominal form predicate. (8) a. ✶ [ --- no --- mu] nari. (where the brackets enclose an adnominal form predicate and mu stands for one of the modal auxiliaries -mu,- ramu, -kemu, -masi, -rasi, -meri, or conclusive-nari) ✶ b. [ --- no --- mu]. (where the brackets enclose an adnominal form predicate and mu stands for one of the modal auxiliaries -mu, -kemu, -zi, -rasi, -meri, or conclusive-nari) This “adnominal-nari” pattern developed when the B part of the [A wa B nari (A TOP B COP)] nominal predicate sentence pattern (copular sentence pattern) was extended from an ordinary noun to allow a predicate. Since the “predicate” that appears in the B part is followed by the copula, it is a nominalized clause being used as a nominal form. Here, the predicate adnominal form with the marker nari attached came to form a pattern specialized to show comments and explanation. With the development of this pattern, it became difficult to interpret sentences that did not have the nari attached (namely, those with the sentence-final adnominal form alone) as showing this meaning of explanation or commentary. The development of the sentence-final “nominalized clause + nari” pattern can be seen to be closely related to the weakening of the specialness (markedness) of the sentence-final “nominalized phrase” pattern. The loss of markedness of sentences ending in an adnominal form can also be seen from examination of kakari-musubi sentences. Kakari-musubi sentences with adnominal forms can be thought to have their origin in the sentence-final use of nominalized clauses (emotive sentences), but EMJ kakari-musubi sentences completed the shift to juttai (statement) sentences. (9) a. Oyodure ka wa ga kik-itu-ru tapakoto ka wa ga lies q I gen hear-perf-adn lunacies q I gen kik-itu-ru hear-perf-adn ‘Is it lies that I hear? Is it lunacies that I hear?’ (MYS 3.420) b. Soko kiyoku sum-an-u midu ni yador-u bottom pure become.clear-neg-adn water dat dwell-adn tuki wa kumor-i na-ki yau no ikade moon top cloud.over-nmlz not.exist-adn appearance gen how ka ar-am-u q exist-conjec-adn ‘How could it be that a moon that abides on water that is not pure and clear to the bottom have the appearance of being without a cloud?’ (Genji. Tokonatsu, SNKBZ 22: 225)
276
Hirofumi Aoki
In OJ/EMJ kakari-musubi sentences like (9a), there is found the fixed word order [focus particle ka – subject marker no or ga– predicate adnominal form] (Sasaki 1992, Nomura 1993a). This fixed word order shows that, because the chunk formed by the sentence-final predicate adnominal form retains its status as a nominalized clause, the kakari particle cannot be inserted into the clause. This fixed word order restriction is no longer observed when reaching the EMJ period and sentences like (9b) with the order [no – ka – adnominal form] become acceptable. This means that the kakari-musubi relation, which was a concord between a sentence-final nominalized clause (emotive sentence) and a kakari particle in the sentence, was re-interpreted as a concord between a kakari particle in the sentence and a sentence-final adnominal form. This shows that it had shifted, that is, to a juttai ‘statement’ sentence. It is apparent, then, that the sentence-final use of nominalized clauses declined, unable to maintain its character as a substantive (its nouniness). Ultimately, this fate was a result of having been placed in the sentence-final position. Consider, for example, the case of the auxiliary verb yooda, which began as a nominal predicate construction [noun-modifying-clause + yau ‘appearance’] with the copula nari attached but, because it was in sentence-final position, was decategorized and reanalyzed as a predicate phrase with the auxiliary verb yau (>yoo) da. (10) a.
b.
Konata kanata no me ni pa sumomo wo putatu this.side that.side gen eye dat top plum acc two tuke-tar-u yau nar-i attach-stat-adn appearance cop-concl ‘It appeared as if he had a plum over each of eyes.’ (Taketori. SNKBZ 12: 48) Kon’ya wa daibu dote ga naga-i yoo tonight top quite causeway nom long-adn semblance da cop.nonpst ‘This evening the causeway has a rather long feel to it (meaning, “I want to go fast, but things don’t progress at all!”).’ (Yūshihōgen. SNKBZ 80: 42)
Containing highly abstract nouns, yooda ‘appear, seems’, hazuda ‘should, expect’, and wakeda ‘it is that’ formed strong lexical chunks and were grammaticalized, but this phenomenon of decategorization of nominal predicates can occur with a much broader range of nouns (Shin’ya 1989, Tsunoda 1996). (11) a.
Taroo ga benkyoo si-tei-ru moyoo Taro nom study do-stat-adn pattern ‘Taro looks to be studying.’
da. cop.nonpst
12 Late Middle Japanese grammar
b.
277
Taroo wa asu Tukuba ni ku-ru yotee Taro top tomorrow Tsukuba dat come-adn schedule da. cop.nonpst ‘Taro is scheduled to come to Tsukuba tomorrow.’
Since ‘ga/no conversion’ is blocked (✶Taroo no benkyoo siteiru moyoo da.) and there is nothing that functions as subject (✶Taroo wa yotei da.), moyoo da and yotei da in these sentences do not form a normal nominal predicate structure, but function, rather, as a kind of auxiliary verb like yooda ‘seems, appears’ or tumorida ‘intend to’. What is important here is that this decategorization phenomenon itself arose as a synchronic phenomenon. It is necessary to distinguish it from a historical process like that by which the forms yooda or tumorida became fixed (lexicalized) through repeated use. Namely the nouniness of a noun differs depending on whether it is functioning as an argument in a sentence or appears at the end of a sentence functioning as a predicate. A noun formed of a nominalized clause as well, at the point it comes to be used as a predicate at the end of a sentence, has already become predicate-like. (In this sense, Y. Yamada’s term gikan jutsuhō ‘pseudo emotional predicate use’ is right on target.) Given that a noun phrase of the form [predicate adnominal form + noun (+copula).], in which the head noun is overtly expressed, loses its nouniness, it is only to be expected that a noun phrase of the pattern [predicate adnominal form.] at the end of the sentence will lose its nominal qualities. In this way, a sentence-final adnominal form (nominalized clause) can be considered to have lost its character as a substantive and to have become an ordinary predicate-ending sentence. The adnominal form, having gained use as a conclusive form (due to the historical change described above), ended up overlapping the original conclusive form in function. As competing patterns, the conclusive form pattern was seen to be the old pattern and the newer adnominal form pattern was chosen. As shown in the examples of (12), the “old conclusive form” pattern had become fixed in the Kamakura Period to appear only in paired couplets and prefatory expressions. (12) a. Kataki wa sukuna-si, mikata wa oo-si, sei ni enemy top few-concl allies top many-concl multitude in magire-te ya ni mo atar-azu. hide.among-ger arrow by etop be.struck- neg.concl ‘There were few of the enemy and many allies. Hidden among the multitude, they were not hit by even one arrow.’ (Heike. SNKBZ 46: 212)
278
Hirofumi Aoki
b. Sipau wa mina kataki nar-i, mikata wa busei all.sides top all enemy cop-concl allies top outnumbered nar-i, ikani si-te nogaru be-si to cop-concl how do-ger escape. adn nec-concl comp oboe-nedo know-neg.conc ‘Although he was surrounded by the enemy, with his allies outnumbered, with no idea how best to escape. . .’ (Heike. SNKBZ 46: 141) Viewed in this way, it would appear that the new “adnominal form” drove out the old “conclusive form” in a remarkably short span of time. However, it is somewhat unreasonable to think that a pattern that began to be used in at the beginning of the Insei period had rapidly gained such general status by the Kamakura period. Accordingly, it would probably be more reasonable to think that there already was a perception of the conclusive form as a written (prescriptive) form in the Insei period when the conclusive use of the adnominal form began to appear in the written sources. There are several examples of “incomplete conclusive” uses of the conclusive form from Insei period sources. (13) a.
b.
c.
Yoru ni wa nar-in-itar-i koyoi wa ie e night dat foc become-perf-stat-concl tonight top home all wa kotosarani yuk-i tuk-azi top particularly go-inf arrive.nconj ‘Night has fallen, and I am not particularly troubled to get home tonight’ (Konjaku. SNKBZ 38: 304) Kore wa mina-pito no sirosimes-itar-u koto nar-eba This top all-people nom know.resp-stat-adn thing cop-prov koto mo naga-si todome-paber-ina-n story etop long-concl stop-pol-perf-conjec.concl ‘As this is a thing that everybody knows, the story is also a long one. I will stop.’ (Ōkagami. SNKBZ 34: 42) Kono usi kata-yama ni pitotu no iwa no ana This cow side-mountain dat one cop.adn stone gen hole ar-i sono ana ni ir-u exist-concl that hole dat enter-concl ‘As for this cow, there was a hole in the stone of a mountainside, and it went into that hole.’ (Konjaku. NKBT 22: 398)
Even earlier, although the sentence-final forms found in the normal literature resources were, in fact, conclusive forms, it has been reported that in conversations in stories, including kakari-musubi sentences, sentences ending in adnominal forms are actually more numerous than those ending in conclusive forms (e.g. Yoshida 2001, 2005).
12 Late Middle Japanese grammar
279
Accordingly, it is possible to think that consciousness of the adnominal form used to conclude a statement sentence as a “new pattern” was already signaled at an earlier stage. In other words, there was no clear demarcation like “EMJ: conclusive form, LMJ: adnominal form”, and perhaps it would be better to see the shift as a process in which the change from sentence-final nominalized clauses (emotives) to statement sentences had already begun in the EMJ period and acceptance of such a change came about in the LMJ period. Such a view also makes it easy to understand why the [adnominal form + nari] pattern developed late in the EMJ period.
3 Change of bigrade conjugations to monograde The core change in the “bigrade to monograde shift” apparently took place from the end of the LMJ to the Early Modern Japanese periods. Table 2, based on Table 1 but modified to show the forms after the merge of the adnominal and conclusive forms, shows the relevant changes. Table 2: Bigrade to monograde shif. Upper Bigrade okuru Stem Irrealis Adverbal Conclusive Adnominal Realis Imperative
Since the bigrade conjugation forms (that had varied between ki and ku or between ke and ku) settled on one form, ki or ke, this change has been called the “bigrade to monograde shift”. However, it is inaccurate to speak of a monograde conjugation and the unchanging part of the verb should be part of the “stem”. Therefore, in Table 2 the “monograde” stems include all of uki and uke and the morphological change in the “bigrade to monograde shift” can be seen to actually consist of an increase in the number of such stems. As mentioned before, there is no difference in conjugational forms between the upper monograde and lower monograde (they have the same conjugational endings) and the only difference between them lies in a difference in the stems (whether they end in –i or -e). In previous research on this phenomenon, the following points have been cited.
280
Hirofumi Aoki
(14) a. The change began in the Insei period b. The change proceeded more rapidly with the upper bigrade conjugation than with the lower bigrade conjugation. c. The change proceeded more quickly with words with fewer moras than with words with more moras. d. The change proceeded more rapidly with independent forms than with bound forms. e. The change proceeded more rapidly with the conclusive-adnominal form than with the conditional (realis) form. f. The change proceeded more rapidly in Eastern Japanese dialects. g. The change was complete by the middle of the Edo period. First, the observations in (14d) (“The change proceeded more rapidly with independent forms than with bound forms”) and (14e) (“The change proceeded more rapidly with the conclusive-adnominal form than with the conditional (realis) form”) reflect the relative frequencies of use of these forms. With regard to (14d), about the only bound forms showing the bigrade conjugational pattern are the voice auxiliaries (ra)ru ‘passive, potential, spontaneous’ and (sa)su ‘causative’, and it is clear that independent verbs were used much more frequently than the bound forms. With regard to (14e), in OJ/EMJ the realis form had a number of functions, such as showing a consequential conjunction with ba and a concessive conjunction with do or, in concordance with the kakari-musubi particle koso, it assumed a variety of functions, but has become fixed as the hypothetical conditional, which is an extension of the generic/habitual conditional. In contrast to this, the adnominal in OJ/EMJ, which functioned as a predicate forming a noun phrase, additionally came to have a conclusive use (see Section 2). The difference in the frequency of use of these forms is apparent at a glance. It is probably understandable that a linguistic change should start from items with a high frequency of use. Next, concerning (14c) (“The change proceeded more quickly with words with fewer moras than with words with more moras”), this fact reflects the essential character of this morphological change. As mentioned earlier, the biggest difference between the bigrade and the monograde conjugations is the length of the stems. Namely, the “monogradized” form has a longer stem and the change can be seen as one to stabilize the form of the word. When the word appears in the form uke in some places and as uku in other places depending on the context, it is more difficult to identify. This sort of instability is even more striking with words like e/u ‘get’ or ne/nu ‘sleep’ with single mora stems. The historical fact that the demand to settle on a single form, e.g. ne ‘sleep’, eliminating the nu ~ ne variation, was first realized starting from words with fewer moras can be seen as supporting the view that the change happened in order to stabilize lexical forms by expanding the stem. At this point, what comes to mind in observing state of affairs is the “upper monograde’ conjugation. The words belonging to this conjugation in OJ/EMJ were all verbs with single mora stems, like miru ‘see’, kiru ‘wear’, piru ‘dry up’, niru ‘boil’, and wiru ‘sit’.
12 Late Middle Japanese grammar
281
For some of these verbs, there are examples found in OJ materials showing that they had been upper bigrades at one point. For example, the form u, corresponding to wiru ‘sit’, is found in Nihon Shoki, Sujin 10, 9th month in the following use: 急居kore woba tuki-u to ipu [this ACC tuki-u COMP say] ‘this is read tuki-u [sit down with a thump]’. In addition, such forms as 廻る mwiru ‘circle’ and 干る piru ‘dry up’ are hypothesized to have once had upper bigrade conjugations. This being the case, it is possible that all these originally had upper bigrade conjugations. In other words, the “bigrade to monograde shift”, which aimed to stabilize the lexical forms by expanding the stems, was already in progress in OJ in the form of an “upper bigrade to upper monograde shift” in words with single mora stems. If this is the case, then (14a) (“The change began in the Insei period”) and (14g) (“The change was complete by the middle of the Early Modern period”) need to be revised. There is also the problem of the imbalance between the “upper monograde” and “lower monograde” conjugations. In contrast to the “upper monograde” conjugation, which properly includes a number of words, albeit all with single mora stems, the “lower monograde” conjugation does not exist. Accordingly, when considering the shift to monograde, rather than lumping the “upper monograde” and “lower monograde” together, they should probably be considered separately. When we distinguish the “upper bigrade” and “lower bigrade” conjugations in this way, it is clear that examples cited from the Insei-Kamakura period as early shifts to monograde are heavily biased toward upper bigrade. (15) a.
b.
媚 kob-u kobi-ru flatter-concl flatter-concl (Maeda-bon Irohajiruishō. Nakada and Minegishi 1962: 245) Turi si-te kaper-isi toki ni tyauzya no fishing do-ger go.home-spst.adn time dat rich.person gen mon o sugi-ru ni front.gate acc pass.by-adn conjct ‘When one passes by a rich person’s front gate upon going home after having been fishing. . .’ (Sangōshikichū. Tsukishima and Kobayashi 1980: 106)
Analogous to the “monograde” conjugation that already existed in OJ, the upper monograde conjugation was created by extending the pattern from single mora stems to two-mora stems. In contrast, there are almost no lower monograde examples before the Kamakura period; keru ‘kick’, which commonly appears in verb conjugation charts for OJ/EMJ, and heru ‘pass through’ are practically the only examples (Yamauchi 2003).
282
Hirofumi Aoki
(16) a.
b. c. d.
“Mari o keru” to i-er-u ke ikan. Ke wa 蹴 nar-i. ball acc kick comp say-stat-adn ke what. Ke top cop-concl ‘What is the “ke” in “Mari o keru (Kick a ball)”? “Ke” is “蹴 (kick)”’ (Myōgoki. Kitano 1983: 91) 麗 peru pass.through (Nihonshoki, Jindai. SNKBZ 2: 64) 蹴散 kuwe pararakasu kick scatter (Nihonshoki, Jindai. SNKBZ 2: 64) 鞠打 mari kuwuru ball kick (Iwasaki-bon Nihonshoki, Kōgyoku. Tsukishima and Ishizuka 1978: 104)
As for keru, originally it was a lower bigrade with the forms kuwe, ku(w)u, ku(w)uru (kuwe-pararakasu ‘kick away’ (16c), mari kuwuru ‘kick a ball’ (16d), and it was through a process of contraction (kuwe > kwe > ke) that these forms yielded keru. The verb peru ‘pass through’ probably also monogradized in analogy to the upper monogrades in which verbs with single mora stems shifted to monogrades. However, it should be kept in mind that in the case of lower monogrades, this was an exceptional phenomenon that occurred in an extremely limited number of forms. Keru very soon changed to a godan ‘quinquegrade’ conjugation and peru very rarely appeared as peru, with the forms pe, pu, puru being much more common. In short, the “lower monograde” conjugation can be said to have been non-existent before the Kamakura period. This view of conjugational paradigms may seem strange when trying to find a neat, organized pattern of “upper bigrade/lower bigrade, upper monograde/lower monograde”. But the fact that the “upper/lower” in the names of the bigrade and monograde conjugations do not necessarily designate a “system” is apparent from examination of conjugational patterns in Japanese dialects. Kyushu is often cited as a region where dialects have retained the classical bigrade conjugations, but I would like to point out an imbalance between upper and lower bigrade conjugations here, as well. The upper and lower bigrade conjugations are both retained only in an extremely limited area in the north of the Hōnichi dialect area; most areas that retain a bigrade conjugation have only the lower bigrade. In these cases, the upper bigrade verbs have either shifted to monograde and then further shifted to godan or they have merged with the lower bigrades. This situation is shown in (17) (Sakono 1998). (17) A. Godan Conjugation (classical quadrigrade, upper monograde, n-irregular, and r-irregular) Lower Bigrade (classical lower bigrade) Upper Bigrade (classical upper bigrade
12 Late Middle Japanese grammar
283
B. (1) Godan Conjugation (classical quadrigrade, upper monograde, n-irregular, and r-irregular) Lower Bigrade (classical lower bigrade, upper bigrade) (2) Godan Conjugation (classical quadrigrade, upper monograde, upper bigrade, n-irregular, and r-irregular) Lower Bigrade (classical lower bigrade) As shown in (17), the conjugational category “lower monograde” does not exist in Kyushu dialects. The “bigrade to monograde shift” has not happened with lower bigrade verbs. On the other hand, the upper bigrade to monograde shift has taken place. However, since the “monograde” conjugation did not sit well, after “okin [get up.NEG], okuru [get up.ADN]” changed to [okin, okiru], it went on to [okiran, okiru], becoming a godan conjugation verb with the stem ending in –r. The “conjugation systems” found in Kyushu dialects show that “upper/lower” in the bigrade and monograde conjugations are not necessary features. Accordingly, it seems reasonable to recognize this sort of imbalance in the history of the central dialects as well. The words belonging to the upper bigrade and upper monograde conjugations are, from the start, semantically limited. Below is a list of representative words. (18) a. sug-u ‘surpass’, ot-u ‘fall’, kut-u ‘decay’, tod-u ‘close’, kwop-u ‘fall in love’, wab-u ‘grieve’, uram-u ‘feel bitter’, oy-u ‘age’, kuy-u ‘regret’, kor-u ‘learn a lesson’, pur-u ‘grow old’ b. mi-ru ‘see’, ki-ru ‘wear’, pi-ru ‘dry’, ni-ru ‘resemble’, ni-ru ‘boil’, i-ru ‘shoot’, i-ru ‘cast (metal)’, wi-ru ‘sit’, wi-ru ‘lead’ The verbs in (18a) are upper bigrade and those in (18b) are upper monograde and the majority of them are clearly intransitive. Even those that do take an object, as in mon wo todu [gate ACC close] ‘close a gate’ or puku wo kiru [clothes ACC wear] ‘put on clothes’, are not verbs that bring about a change in the object (verbs of change) but are limited to reflexive-like actions. In contrast, there is no such semantic tendency apparent in verbs in the lower bigrade conjugation, which are overwhelmingly more numerous than those in the upper bigrade category. Particularly noteworthy is the fact that they form transitive-intransitive pairs with quadrigrade conjugation verbs. These pairs may be like tatu ‘stand (intr)’: tateru ‘stand (tr)’, where the quadrigrade is intransitive and the lower bigrade transitive, or like yaku ‘burn (tr)’: yakeru ‘burn (intr)’, where the quadrigrade is transitive and the lower bigrade intransitive. In either pattern, there is a regular vowel alternation, and they show a close link between the quadrigrade conjugation, which is the pattern that includes the most verbs, and the lower bigrade conjugation, and this patterning is totally different from that shown by the upper bigrade. As described above, by considering upper and lower bigrade conjugations separately, we can distinguish a shift to monograde that took place by the Kamakura period
284
Hirofumi Aoki
(monogradization of the upper bigrade) and a shift to monograde that took place after the Muromachi period. Re-examining the Muromachi period data, we notice that all the examples are of lower bigrade monogradization, the reverse of the earlier change. It has been reported that there are no examples of upper bigrade to monograde shifts in Shōmono or Kyōgen materials (Yuzawa 1929, Hachiya 1968). The description in the Rodrigues’s Nihon Daibunten (1604–1608) is suggestive on this point. (19) All verbs whose stem ends in in E have an alternative formation in the present tense which is only rarely used in the spoken language. However, it is used in Kantō and by some persons of the capital. It is formed by adding Ru to the stem. For example, ku[ra]beru ‘compare’, ageru ‘raise up’, motomeru ‘seek’, faneru ‘spring, jump’, todokeru ‘deliver’, atayeru ‘give’, feru ‘pass through’, ideru ‘go out’, mazeru ‘mix’, and miseru ‘show. (adapted from Doi 1955: 29) It is clear from this that the shift was proceeding in the spoken language and that it was proceeding in eastern (Kantō) dialects more rapidly than in the central dialects (14f), but the description is limited to just the lower bigrade to monograde shift. If there were forms resulting from an upper bigrade to monograde shift, such as otiru ‘fall’ or sugiru ‘surpass’, the attentive Rodriguez would undoubtedly have noted them. Combined with the situation in the Shōmono and Kyōgen materials, it can be concluded that the “bigrade to monograde shift” that arose in the Muromachi period started first in the lower bigrade. Furthermore, looking at early examples of this shift, it is apparent that verbs from the ya column of the conjugational chart are overwhelmingly numerous. (20) a.
b.
c.
d.
Tae-ru to wa korau-ru koto nar-i. endure-concl comp top tolerate-adn action cop-concl ‘taeru (to endure) is “the act of tolerating”.’ (Rongoshō, Yōya, In Yuzawa. 1929: 63.) Tuma o mukae-ru ni wa rokutai ga ar-u wife acc welcome-adn purp top six.forms nom exist-concl zo. sfp ‘To welcome a wife, there are six forms.’ (Mōgyūshō. Okami and Ōtsuka 1971–1976, Vol. 6: 544) 舗設 kosirae-ru zo. prepare-concl sfp (Hyakujōshingishō. Ōtsuka 1980–1992, Vol. 8: 296.) Aki-kaze wa feiran ni tatoe-ru zo. autumn-wind top armed.conflict dat compare-concl sfp ‘The autumn wind is likened to an armed conflict.’ (Shigaku Taiseishō. Ōtsuka 2000, Vol. 1: 293)
12 Late Middle Japanese grammar
285
In studies of Shōmono materials (e.g. Yuzawa 1929, Yanagida 1973), the majority of examples (e.g. kaeru ‘change’, ueru ‘plant’, soeru ‘align’, osieru ‘teach’, kotaeru ‘respond’, sakaeru ‘florish’, kangaeru ‘think’, otoroeru ‘weaken, decline’, totonoeru ‘straighten’) given are from this sort of ya column verbs. These ya column verbs include ones that in classical Japanese belonged to the a, ha, and wa columns, i.e., verbs which in the historical kana spelling are written with a kana letter from the a, ha, and wa columns in the final syllable of the stem. That is, before the “monogradization” that took place in the Muromachi period, a ya column shift had taken place. This ya column shift phenomenon can be thought to have occurred because, taking kauru (かふる “kahuru”) ‘change’ as an example, since there was the possibility of the stem vowel becoming a long vowel (kauru > kooru), the verb changed to kayuru, avoiding the vowel crasis (Izumo 1979). This sound change was greatly influenced by the irrealis and adverbal forms of ha column verbs like atae ‘give’ and wa column verbs like uwe ‘plant’ becoming [ye] like ya column verbs such as mie ‘be visible’. However, rather than thinking of the process as being one of making the conclusive-adnominal form yuru by analogy, if we take into consideration the fact that the shift to ya column verbs started from verbs with fewer mora in their stems, it is probably more appropriate to view it as a change aimed at preserving the stem and achieving the stability of form described earlier. Viewed in this light, the question of why the change in the Muromachi period began with the lower bigrade verbs and not with the upper bigrades becomes more understandable. The changes occur in a chain such that, being drawn along by the ya column lower monograde verbs formed in way described above, the shift to monograde extended to other lower bigrade verbs and then to upper bigrade verbs.
4 Development of the nominative particle ga In OJ/EMJ, the subject of a main clause was not marked with a particle. However, in a normative sentence in Modern Standard Japanese, the subject noun phrase is obligatorily marked by a particle. (21) EMJ Ima pa mukasi, taketori no okina to ip-u now top long.ago bamboo.cutting gen old.man comp call-adn mono ar-iker-i. person exist-mpst-concl ‘Once upon a time there was a person called “Old Man Bamboo-cutter”.’ (Taketori. SNKBZ 12: 17)
286
Hirofumi Aoki
Modern Japanese Ima to nat-te wa mukasi no koto da ga, now comp become-ger top long.ago gen matter cop.npst but taketori no okina to i-u mono ga i-ta. bamboo.cutting gen old.man comp say-npst person nom exist-pst ‘It’s a long time ago now but there was a person called “Old Man Bamboo-cutter”.’ It is not the case that subject marking ga did not exist in OJ, but it did not mark the subject of a main clause, functioning instead to mark the subject of a subordinate (noun modifying) clause. At this stage, it was found only in the pattern [ga – predicate – noun]. (22) a.
b.
Wa ga sekwo ga uwe-si aki-pagwi I gen beloved gen plant-spst.adn autumn-bushclover pana sak-in-ikyer-i flower bloom-perf-mpst-concl ‘The the flowers of the autumn bushclovers that my beloved planted have bloomed.’ (MYS 10.2119) Imo ga mi-si aputi no pana pa beloved gen see-spst.adn chinaberry gen flower top tir-inu be-si wa ga nak-u namita imada scatter-perf nec-concl I gen cry-adn tears still pwi-naku ni dry-neg.nmlz dat ‘The chinaberry flowers that my beloved saw will have scattered, though the tears that I cry have not yet dried.’ (MYS 5.798)
However, while very few, there are examples in which ga appears to mark the subject of a main clause. (23) a.
b.
Asipikwino yama wo ko-daka-mi yupu-tukwi wo (Epithet) mountain acc tree-tall-inf evening-moon acc itu ka to kimi wo mat-u ga kurusi-sa when q comp lord acc wait-adn ga painful-nmlz ‘How excruciating is waiting for my lord, like waiting for the evening moon with the mountain being overgrown with trees, thinking “When will it be?”’ (MYS 12.3008) Ama no gapa kog-u puna-bito wo heaven gen river row-adn boat-person acc mi-ru ga tomosi-sa see-adn ga enviable-nmlz ‘How envy-provoking is seeing the boatmen who row the river of heaven!’ (MYS 15.3658)
12 Late Middle Japanese grammar
287
The examples in (23), however, have the structure [ga [predicate (adjective stem) + sa]] and the ga is still attached to a clause forming a noun phrase. It is the so-called gikan jutsuhō ‘pseudo-emotional predicate use’ (see Section 2). Considering the fact that the other places where ga could appear were, for example, in a nominalized clause, in a ku-nominalized clause, or in a kakari-musubi clause ending in an adnominal form, we can conclude that the nominative particle ga was constrained to appear in nominalized clauses in OJ. Entering the Heian period, however, the particle began to often be used in predicate phrases in the pattern [ga + predicate]. (24) a.
b.
[Podonaku makar-in-u be-ki na mer-i in.a.short.while go.back-perf-concl nec-adn cop.adn evid-con to omop-u] ga kanasi-ku paber-u nar-i. comp think-adn ga sad-inf exist-adn cop-concl ‘It is sad to think that it looks like I will have to go back in a short while.’ (Taketori. SNKBZ 12: 70) [Wonna no mada yo pe-zu woman gen yet worldly.relations experience-neg.concl to oboe-tar-u ] ga pito no on-moto ni comp consider-stat-adn ga person gen hon-place dat sinob-ite steal-ger ‘A woman who was considered to have not yet experienced worldly relations having stole into the abode of a person of stature. . .’ (Ise. SNKBZ 12: 213)
Example (24a) is an example of ga used in the structure [thing-nominalized clause + ga + predicate (adjective)] and (24b) the structure [person-nominalized clause + ga + predicate (verb)]. However, neither of the examples in (24) is of ga used directly linked to the predicate of a main clause in the conclusive form. Such examples of ga used to mark the subject of a main clause first began to appear later on in the Insei-Kamakura Period. Both of the examples in (25) have the structure [person-nominalized clause + ga + predicate] sentence-finally and the nominative particle ga is thought to have developed in such structures (Ishigaki 1955). (25) a.
[Tosi goziu bakari nar-u wotoko no osorosi-ge-nar-u ] age fifty about cop-adn man gen frightening-nmlz-cop-adn ga suikan-syauzoku si-te utiide no tati ga all.weather-garb do-ger hammer.forged cop.adn sword ob-itar-i carry-stat-concl ‘A man about fifty years of age, with a frightening countenance was wearing all-weather garb and carrying at his side a hammer-forged sword.’ (Konjaku. SNKBZ 37: 559)
288
Hirofumi Aoki
b.
[Ito mazusikar-iker-u] ga Kurama ni nanoka very destitute-mpst-adn ga Kurama dat seven.days mair-iker-i. come.hum-mpst-concl ‘A very destitute person came to Kurama Temple for seven days.’ (Uji shūi monogatari. SNKBZ 50: 211)
The development of the nominative particle described above is illustrated below. (26) EMJ [Mata tatau-gami no tenarawi nado si-tar-u] φ again folded-paper gen writing.practice conjct do-stat-adn φ mi-kityau no moto ni oti-tar-iker-i. hon-bed.curtain gen base dat fall-stat-Mpst-concl ‘Also some folded paper on which some writing practice or the like had been done had fallen to the base of the bed-curtains.’ (Genji, Sakaki. SNKBZ 21: 145) LMJ [Mi mo sir-an-u pana no iro-imizi-ki] see.inf etop know-neg-adn flower gen color-vivid-adn ga sak-i midare-tar-i ga bloom-inf be.profuse-stat-concl ‘Vividly beautiful flowers which he had never seen before were blooming profusely.’ (Ujishūi monogatari. SNKBZ 50: 421) Namely, where there had been no particle through the OJ/EMJ period, ga came to be inserted. Regarding the motive for this development, Nomura (1993b) draws attention to the two points that, first, it developed in an environment where nominalized clauses are found, and, second, it was the period when the sentence-final adnominal form had come into general use, and makes the very interesting proposal that, in order to avoid a sequence of two unmarked predicates in the adnominal form, ga was inserted to mark the nominalized clause. Namely, multiple uses of the adnominal form, as in the following example, could have caused difficulty in processing. (27) Sore no iro no koromo ki-tar-u waka-ku that gen color gen robe wear-stat-adn young-inf oi-tar-u putari no sou to gwenkau age-stat-adn two.people cop.adn monk com return si-tu-ru? do-pefv-adn ‘Have you returned from worship together with two monks, a young one and an old one, wearing robes of that color?’ (Konjaku. SNKBZ 35: 408)
12 Late Middle Japanese grammar
289
The nominative ga that developed in the environment [adnominal form of predicate (nominalized clause) + ga + predicate.] in sentence-final position eventually spread to be used with independent nouns as well, developing the pattern [nominal + ga + predicate.] in sentence-final position. (28) a.
b.
c.
Wara-sudi pito-sudi ga kauzi mitu ni straw-length one-length ga orange three cop.inf nar-itar-it-u. become-stat-perf-concl ‘A single length of straw has turned into three oranges.’ (Kohon Setsuwa-shū. SNKBT 42: 480) Foodyoo kore wo mi-te “sisai ga ar-u. Hōjō that acc see-ger reason ga exist-npst Sibasi” to yuu-te mat-are-ta. momentarily comp say-ger wait-hon-pst ‘Hōjō, having seen this, said, “There is a reason for this. Just a moment,” and waited.’ (Amakusaban Heike. Fukushima 1987: 391) Toi no Zirɔɔ ga issen-yo-ki de Toi gen Jirō nom one.thousand-more.than-warriors instr sasae-ta. support-pst ‘Toi no Jirō was in support with more than one thousand warriors.’ (Amakusaban Heike. Fukushima 1987: 245)
Finally, the pattern spread to nominal predicate (copula) sentences, giving the structure [nominal + ga + nominal (da).]. (29) a.
b.
Onore-ra ga oni ni-te ar-u zo to self-pl ga demon cop-ger exist-adn sfp comp siri-tamap-er-u nar-i. know-resp-stat-adn cop-concl ‘It is such that you have found out that we are demons.’ (Konjaku. NKBT 22: 340) Soregasi ga suefirogari-ya no teesyu de oriyar-u yo. I ga folding.fan-shop gen owner cop exist-npst sfp ‘It just so happens I am the proprietor of a folding fan shop.’ (Toraakira-bon Kyogen-shū, Suehirogari. Ikeda and Kitahara 1972: 70)
Accordingly, considering the time period, the ga in (29a) should be interpreted as the genitive case (M. Yamada 2010). It is thought that the particle did not develop the same kind of use it has in Modern Japanese via the extension to nominal predicate sentences until around the beginning of the Early Modern period.
290
Hirofumi Aoki
The nominative particle ga is a form that thus developed by shedding syntactic constraints, moving from the pattern [nominalized clause + ga] to the pattern [nominal + ga] and from subordinate clauses to main clauses.
5 Development of the nominalizing particle no Semantically and structurally, nominalized clauses can be divided into the following two types. (30) a.
b.
Tukaumatur-u pito no naka ni serve.hum-adn person gen middle dat kokoro-tasika nar-u wo erab-ite spirit-reliable cop-adn acc choose-ger ‘Choosing a stout-hearted one from among the people that served him, . . .’ (Taketori. SNKBZ 12: 37) Imizi-ki urewe ni sidum-u o extraordinary-adn wretchedness dat sink-adn acc mi-ru ni tawe-gata-kute see-adn conjct endure-difficult-ger ‘It being hard to bear when I saw you sinking into such extraordinary wretchedness. . .’ (Genji, Akashi. SNKBZ 21: 229)
The portion of (30a) in bold font expresses the meaning “a (person) with a reliable spirit” and the relation constituted between the “person” and the predicate kokoro-tasika-naru ‘be reliable in spirit’ is a case relation of [(pito) ga kokoro-tasika-naru [(person) gen be reliable in spirit]]. I will call this type of construction designating a concrete “person” or “thing” that has an “inner relation” (Teramura 1992) with the predicate the “relative clause” type. Example (30b), on the other hand, expresses the state of being sunk in grief “imiziki urewe ni sidumu [extreme grief into sink]” and the clause itself expresses the state, circumstances, or content that it refers to, schematically, [imiziki urewe ni sidumu = (state)]. I will call this type in which the predicate and the state or abstract item it refers to stand in an “outer relation”, the “noun clause” type. (Ishigaki (1955) refers to the former as keijōsei meishiku ‘stative nominalizations’ and the latter as sayōsei meishiku ‘active nominalizations’.) Use of the nominalized clauses formed from the adnominal form in nominal use declined from Muromachi period into the Edo period and the nominalization particle no came to be affixed to the end of the clause. No first starts to appear in the literature in works from the end of the Muromachi period and the beginning of Edo.
12 Late Middle Japanese grammar
(31) a.
b.
291
Sendo soti e watai-ta no wa nanto si-ta zo just.now you all pass-pst nm top how do-pst sfp ‘What I passed to you just now: What did you do with it?’ (Toraakira-bon Kyōgen-shū, Gan nusubito. Ikeda and Kitahara 1972: 168) Fime ga fadae ni titi ga simoto o noble.daughter gen flesh dat father nom cane acc ate-te sagas-u no koso kanasi-kere apply-ger search-adn nm foc sad-excl ‘A father searching by pressing the cane to his noble daughter’s flesh is indeed sad.’ (Kibune no Honji. Hagino 1901: 15)
This sort of no used as a nominalization particle is thought to have developed from the use of no as a pronoun, as shown in the following example from the late EMJ period. (32) Pito-duma to wa ga no to putatu omo-u person-wife com I gen nm com both think-adn ni wa nare-ni-si sode zo conjct top be.familar-perf-spst.adn sleeve foc aware nar-iker-u be.moving cop-mpst-and ‘When I think of both another man’s wife and my own, it’s the familiar sleeve that is moving for me.’ (Yoshitada-shū, Poem 457. NKBT 80: 110)
The no used as a pronoun referring to “the one (=wife) that is mine” came to be used as in (31a) in a pattern modified by a predicate “that (=coins) which I handed you earlier”. The no that fulfilled the role of a demonstrative pronoun referring to something concrete (thing, person) came to be used clause-finally with a nominalized clause as well. After this, no further came to be perceived as being used in the head noun position of a noun phrase consisting of an adnominal form in nominal use and came to be used in an abstract meaning as in (31b). The relative order of appearance for these two types of use is not evidenced in the written sources, but it is probably better to think of the process as having the spread of no as starting from the relative clause type showing a concrete person or thing and later extending to the noun clause type referring to abstract states or facts. This is because in the modern language (including dialects) the noun clause type can still be seen in the form of nominalized clauses, as in . . .suru ga masi [do NOM better] ‘it would be better to . . . .’ and . . .suru ni kagiru [do DAT limit] ‘no choice but to . . .’.
292
Hirofumi Aoki
The nominalization particle no became established in the late Muromachi – early Edo periods, as shown in (31), but the use of no as a marker was not obligatory even into the late Early Modern period. (33a) is an example with a noun phrase using no and (33b) is an example of a noun phrase consisting of a nominalized clause; both come from the same work. (33) a.
b.
Kure ni ka-u no o wasure-ta kara last.night dat buy-npst nm acc forget-pst because kyoo wa abura o ka-i ni yar-oo ya today top oil acc buy-inf purp send-vol sfp ‘Last night my husband forgot to buy oil, so today I’ll send him to buy some.’ (Ukiyoburo. NKBT 63: 219) Sotti no korob-u wa sosoo de mo you gen stumble-npst top carelessness cop etop sum-oo ga ore ni mizu o kake-te put.down-vol conjct I dat water acc splash-ger sosoo de sum-u ka carelessness cop put.down-npst q ‘Your stumbling might be put down as carelessness, but will your splashing water on me be put down as carelessness?’ (Ukiyoburo. NKBT 63: 101)
The use of the nominalization particle started with noun phrases in major argument positions, ga (nominative) and o (accusative); expressions in which the nominalized clause construction has become fixed are commonly found with, for example, the dative (ni) case . . .suru ni oyobanai [do DAT not.extend.to] ‘need not . . .’ or . . .suru ni turete [do DAT accompany] ‘(do, happen) in proportion to . . .’ and the comparative (yori) case . . .suru yori . . . suru hoo ga ii [do than do alternative NOM good] ‘it is better to . . . than to . . .’. Considering these facts, when contemplating the motivation for the development of the nominalization particle no, rather than thinking of it as embodying the functional change of the the decline of the nominalized clause formed with the adnominal form, it would be better to see it as being on roughly the level of speakers coming to see it as simply being better for a noun phrase to have a head noun. Supporting this view is the fact that it was after the beginning of the Meiji period that the nominalized clause stopped being used and marking with no became obligatory. The nominalized clauses taken up above are ones used as argument noun phrases in a sentence. Nominalized clauses were also, as seen in Section 2, used sentence-finally. This was the [nominalized clause + copula (nari)] (adnominal-nari) pattern used to provide explanation or commentary. Just as nominalized clauses used within a sentence came to have the nominalization particle no affixed, in the case of sentence-final use as well the pattern [noun phrase (. . .no) + copula (da)] with no affixed came to be used.
12 Late Middle Japanese grammar
(34) a.
b.
c.
293
Kitune no tukaumatur-u nar-i fox gen do.hum-adn cop-concl ‘This is something that the foxes do.’ (Genji, Tenarai. SNKBZ 25: 283) (repeated from part of (7b)) Zyukusui nar-an-eba funmyoo ni wa deep.sleep cop-neg-prov clear cop.inf top oboe-n-u nar-i remember-neg-adn cop-concl ‘As one is not in a deep sleep, one doesn’t remember clearly.’ (Chūka Jakubokushi-shō. SNKBT 53: 11) Edokko no kane o omae-gata ga musir-i Edo-dweller gen money acc you-pl nom pluck-inf tot-te ik-u no da. take-ger go-npst nm cop.npst ‘It is such that you go on expropriating the money of the people of Edo.’ (Ukiyodoko. SNKBZ 80: 283)
In (34a), as is clear from the appearance of the no used to mark the subject of a subordinate clause, the structure is one of a nominalized clause kitune no tukaumaturu with the copula nari attached. This construction was preserved into the LMJ and Early Modern Japanese periods and the no that had come to be used in a case-marked position in a sentence where “nouniness” was called for also came to be used in the sentence-final position (34c). However, as shown by the fact that the pattern [nominalized clause + copula] such as suru da [do COP] continued to be used in the late Edo period, like the sentence-internal (case-marked) case, it was not until entering the Meiji period that the no became obligatory. Forms like suru da [do cop] or samui da [cold cop] are frequently found as dialectal forms even today. It is important to note here that structure of the pattern [adnominal form + no + da] is not “noun phrase + da”. As with the forms yooda or moyooda in Section 2, ga-no conversion, which should be possible in an adnominal clause, is not permitted and it is clear that noda has a function corresponding to an auxiliary. (35) Gakusee ga Issyookenmee benkyoo si-teir-u. student nom all.one’s.might study do-stat-concl Siken ga/✶no aru noda. test nom/✶gen exist.npst noda ‘The student is studying as hard as he can. He has a test.’ It was pointed out in Section 2 that noun phrases appearing in sentence-final position lose their “nouniness” and noda became lexicalized together with the development of
294
Hirofumi Aoki
the nominalization particle no. We can say that this meaning of commentary or explanation was impressed on the form noda. It is apparent, then, that nominalized clause thus came to be used both sentence-internally as argument noun phrases and sentence-finally with a copula in the meaning of giving an explanation, with both uses coming to have the nominalization particle no affixed. In addition to these uses, the nominalization particle no also came to be used in conjunctive clauses, as shown in the following examples. (36) a.
b.
c.
Ora ga uti zyaa ore ga karada ga kik-anee I gen home loc.top I nom body nom work.well-neg.npst kara mori ga hitotu deki-nee noni because care nom one be.able-neg.npst although tosigo daa consecutive.year.child cop.npst ‘At my house, although I can’t do a bit of child-care because my body is infirm, there are children for consecutive years.’ (Ukiyoburo. NKBT 63: 124) Anmari warat-ta node syooben ga si-ta-ku nat-te excessive laugh-pst since pee nom do-des-inf become-ger ki-ta. come-pst ‘Since I laughed so much, I started to have to pee.’ (Shichihenjin. Kokumin Tosho Kabushikigaisha 1928: 254) Koo kane ga nee nonara dootomo this.way money nom not.exist.npst if whatever.way si-yoo ze do-vol sfp ‘If they have this little money, they might do anything!’ (Otoshibanashi Jukushigaki. Mutō and Oka 1975–1979, Vol. 15: 88)
It seems a bit strange that the nominalization particle should be used in a conjunctive clause but, noni ‘although’ and node ‘since’ can be thought of as having brought about a structural change. The conjunctions tokoroga ‘however’ and tokorode ‘by the way’ were also originally of the structure [noun phrase + case particle]. Noni and node can also be considered to have undergone a reanalysis as a result of this restructuring. (37)
→
[NP [NP predicate in adnominal form + no] ni ] [predicate] [ConjP [PredP predicate in adnominal form] no ni] [predicate]
That is, noni and node can be taken as having undergone the same restructuring as conjunctions that noda underwent as a predicate.
12 Late Middle Japanese grammar
(38)
→
295
[PredP [NP predicate in adnominal form + no ] da] [PredP [PredP predicate in adnominal form] no da]
That is, the predicate in the adnominal form that had formed a noun phrase changes to become a predicate phrase. Viewing these forms in this way also allows explanation of the semantic and syntactic commonalities. Since noda has as its core meaning “sequentiality” and “givenness”, it developed into a clause with the meaning of “background circumstances” and “actual circumstances” (Tanomura 1991). Noni expresses a concessive conjunctional meaning based on this “givenness”, that is, a “mismatch” with the speaker’s expectation based on his pre-existing knowledge. The meanings of “discomfort” or “dissatisfaction” can be said to stem from this kind of “mismatch”. Node is said to be better suited than kara ‘since’ to “objective description”. This can be said to be a resultative conjunction meaning based on “sequentiality”. The expansion from a modal meaning of “surprise” or “discovery” to “explanation of circumstances” and that from an emotional meaning of “dissatisfaction” as a sentence-internal concessive conjunction to an “objective” resultative conjunction meaning can be taken as being parallel. In addition, since the noda pattern has this “sequentiality” and “givenness”, it has the syntactic characteristic that it cannot appear in a pattern like daroo noda following a conjectural form. This is also true of node and noni, with forms like daroo noni or daroo node being unacceptable. The meaning of nonara ‘if that is the case’ can be explained in the same way. It is used as a hypothetical conditional clause when referring to an actual situation and “givenness” is clearly part of the core meaning. Accordingly, we can probably say that nonara is the conditional form of noda. In describing aspects of the historical expansion of the nominalization particle no above, we have concentrated on the syntactic environment – sentence-internal or sentence final – and, if sentence-internal, case-marked element or conjunctional element, but it is possible to grasp aspects of its development in the same way. Although the nominalization particle itself was established at the beginning of the Early Modern period, it did not become completely entrenched all the way through the Early Modern period and it took a very long time, until the beginning of the Meiji period, for it to finally become thoroughly entrenched. Accordingly, the situation in which the nominalizing particle no could either be present or not continued in the Early Modern period for at least 200 years. Examples with argument noun phrases were given in (33), but below are examples of sentence-final [adnominal form + 0 + da] and [adnominal form + 0 + ni].
296
Hirofumi Aoki
(39) a.
b.
Uba komori nado no tagui ga dehoodai no wet.nurse nanny conjct gen sort nom nonsense gen monku o tukur-u niyotte ano yoo ni iyasi-ku verse acc make-adn by that semblance cop.inf vulgar-inf nar-u zya te become-npst cop comp ‘It’s by wet nurses and nannies and that sort making nonsense verses that it [the song] becomes so vulgar.’ (Ukiyoburo. NKBT 63: 242) Zentee kokora made issyoni ko-yoo initially hereabout all together come-vol to it-ta ni saki e ki-te simat-ta. comp say-pst although prior all come-ger close-pst ‘Although I initially said, “Let’s go there together,” I have come ahead.’ (Hasshōjin. Kokumin Tosho Kabushikigaisha 1928: 74)
During this period, it can be surmised that people, while recognizing the fact it was fine for no to either be there or not, gradually came to wonder what its presence meant. That is, it can be thought that a consciousness of the meanings of auxiliary verb noda and the conjunctional particle noni gradually developed. Grammatical change, then, needs to be broken down into and perceived as a number of stages: origin, development, entrenchment. The form originated in the Muromachi period but, since the motivation for using it was at the level of it simply being better for there to be a head noun in a noun phrase, its use did not spread rapidly. As the forms noda and noni exerted their predicativeness in main clauses and conjunctional clauses, together with recognition as a form used in a sentence to express “circumstances”, a consciousness developed of using no in all the environments where nominalized clauses “adnominal form + 0” had been used. This sort of change can be seen taking place over a span of three or four centuries.
6 Conclusion This chapter has described four historical changes: “the generalization of the adnominal form”, “the bigrade to monograde shift”, “the development of the nominative particle ga”, and “the development of the nominalization particle no”. Each of these changes reached a transition point in the LMJ period, but the changes actually extended into both the preceding and following periods. An important aspect of the descriptions in this chapter has been giving explanations of high generality regarding grammatical phenomena viewed from a dynamic perspective. Take the sentence-final predicate form for example. The usual explanation is that the conclusive form was used through the OJ/EMJ period and was replaced
12 Late Middle Japanese grammar
297
by the adnominal form from LMJ onwards. It is, however, difficult to conceive of this change taking place over a span of a mere 100 or so years. Decategorization of nominalized clauses (subordinate clauses becoming main clauses) had already arisen in the EMJ period and we should probably consider that this gradually came to be accepted by people and it was between the Kamakura and Muromachi periods that it became established. I think it is essential to take up and consider the language behind the phenomena appearing in the written materials. Regarding the bigrade to monograde shift, when we include present day Japanese dialects in our purview, it is clear that the shift proceeded very differently for upper monograde verbs with stems ending in -i and lower monograde verbs with stems ending in –e and, accordingly, it was stated that the shift should be divided into pre- and post-Muromachi monogradization. There is a theory that links this shift to the generalization of the sentence-final adnominal form, but this chapter argues against that view. Concerning the development of the nominative particle ga, the description given here owes much to previous work by Ishigaki (1955) and Nomura (1993b), but the fact that the shift toward freedom from syntactic constraints progressed gradually, starting in the EMJ period and finally being completed in the Muromachi period, is highly intriguing. With regard to the development of the nominalization particle no, the theory that it developed in response to the decline of nominalized clauses has commonly been cited in the past. Furthermore, the reason for the decline of nominalized clauses has often been sought in the expansion in function of the adnominal form (generalization of the sentence-final adnominal form). However, although it may have been in decline, considering the fact that, historically, nominalized clauses continued to be used until very late, the fact that there was a large variation depending on the syntactic environment, and the fact that there are many places where the construction is still used in dialect forms, it is difficult to perceive a decline in nominalized clauses. Rather, based on the perceived equivalence in meaning and use between the [predicate adnominal form + 0] and [predicate adnominal form + no] patterns, it would be better to hypothesize a process in which awareness of the meaning when no was present became imprinted on the form itself. This is the awareness found in argument noun phrases like no ga [no nom]. no o [no acc], and no wa [no top], in conjunctional particles like noni ‘in spite of the fact that’, node ‘since’, and nonara ‘if that is the case’, and in auxiliary verbs like noda ‘it’s a matter of’ and nodaroo ‘it’s probably a matter of’. Another grammatical phenomenon that changed with its turning point located in the Late Middle Japanese period was that of kakari-musubi. This phenomenon has been mentioned several times in this chapter in relation to generalization of the sentence-final adnominal form and the development of the nominative particle. Its more complete description is a topic for future research.
298
Hirofumi Aoki
References Doi, Tadao (trans.). 1955. Nihon dai-bunten [Great grammar of Japanese]. Tokyo: Sanseidō. (A Japanese translation of Rodrigues 1604–1608).) Fukushima, Kunimichi [Exegesis]. 1987. Amakusaban Heike monogatari [Amakusa edition of the Heike monogatari], Final Volume. Tokyo: Benseisha. Hachiya, Kiyoto. 1968. Kyōgen daihon ni mirareru ichidan katsuyōka no genshō [The phenomenon of monogradization seen in kyōgen scripts]. Kokugogaku 74. 52–62. Hagino, Yoshiyuki. 1901. Shinpen otogi sōshi [Otogi zōshi, new edition]. Tokyo: Seishidō Shoten. Ikeda, Hiroshi and Yasuo Kitahara. 1972. Ōkura Toraakira-bon Kyōgen Shū no kenkyū: Honbun hen [Research on Ōkura Toraakira’s Collection of Kyōgen: Text]. Tokyo: Hyōgensha. Ishigaki, Kenji. 1955. Joshi no rekishiteki kenkyū [A historical study of particles]. Tokyo: Iwanami Shoten. Izumo, Asako. 1979. Chūsei ni okeru bungo a/ha/wa gyō shimonidan katsuyō ni zokusuru dōshi no ya-gyō shimonidanka genshō ni tsuite [On the phenomenon of the monogradization in Late Middle Japanese as ya-column conjugation of verbs which in classical Japanese belong to bigrade verbs with a/ha/wa-column conjugation]. In Nakada Norio hakase kōseki kinen kokugogaku ronshū kankōkai (ed.), Nakada Norio hakase kōseki kinen kokugogaku ronshū [Collection of papers on Japanese linguistics in commemoration of the achievements of Dr Norio Nakada], 15–45. Tokyo: Benseisha. Kokumin Tosho Kabushikigaisha (ed.). 1928. Kindai Nihon bungaku taikei [Early modern Japanese literature series], Vol. 22. Tokyo: Kokumin Tosho Kabushikigaisha. Kitano, Tayuru (ed.). 1983. Myōgoki [Myōgoki]. Tokyo: Benseisha. Masamune, Atsuo (ed.). 1954. Ruiju myōgishō [Ruiju myōgishō]. Tokyo: Kazama Shobō. Mutō, Sadao and Masahiko Oka (eds.). 1975–1979. Hanashibon taikei [Series of collections of short stories (from the Edo period)]. Tokyo: Tokyodō Shuppan. Nakada, Norio and Akira Minegishi (eds.). 1962. Iroha jiruishō: Kenkyū narabi ni sakuin [Iroha jiruishō: Research and index]. Tokyo: Kazama Shobō. NKBT = Takagi, Ichinosuke et al. (eds.). 1957–1969. Nihon koten bungaku taikei [Japanese classical literature series]. Tokyo: Iwanami Shoten. Nomura, Takashi. 1993a. Jōdaigo no ‘no’ to ‘ga’ ni tsuite (jō) (ge) [On no and ga in Old Japanese part I, II]. Kokugo Kokubun 62(2). 1–17. / Kokugo Kokubun 62(3). 30–49. Nomura, Takashi. 1993b. Jōdai kara chūsei no ‘no’ to ‘ga’ [No and ga from Old Japanese to Late Middle Japanese]. Nihongogaku 65(5). 23–33. Okami, Masao and Mitsunobu Ōtsuka (eds.). 1971–1976. Shōmono shiryō shūsei [Collection of shōmono materials]. Osaka: Seibundō Shuppan. Ōtsuka, Mitsunobu (ed.). 1980–1992. Zoku shōmono shiryō shūsei [Continued collection of shōmono materials]. Osaka: Seibundō Shuppan. Ōtsuka, Mitsunobu (ed.). 2000. Shin shōmono shiryō shūsei [New collection of shōmono materials]. Osaka: Seibundō Shuppan. Rodrigues, Ioão. 1604–1608. Arte da lingoa de Japam [A grammar of the Japanese language]. Jesuit Society: Nagasaki. Sakono, Fuminori. 1998. Kyūshū hōgen no dōshi no katsuyō [The conjugation of verbs in Kyushu dialects]. Gobun Kenkyū 85. 1–11. Sasaki, Takashi. 1992. Jōdaigo ni okeru ‘-ka -ha’ no kōbun [The syntax of -ka -ha in Old Japanese]. Kokugo Kokubun 61(5). 17–33. Shin’ya, Teruko. 1989. ‘Bunmatsu meishi’ ni tsuite [On ‘sentence final nouns’]. Kokugogaku 159. 75–88. SNKBT = Satake, Akihiro et al. (eds.). 1989–2005. Shin Nihon koten bungaku taikei [New Japanese classical literature s eries]. Tokyo: Iwanami Shoten. SNKBZ = 1994–2002. Shinpen Nihon koten bungaku zenshū [New Japanese classical literature, complete collection]. Tokyo: Shōgakkan.
12 Late Middle Japanese grammar
299
Tanomura, Tadaharu. 1991. Gendai Nihongo no bunpō I: ‘Noda’ no imi to yōhō [Grammar of modern Japanese, I: Meaning and usage of noda]. Osaka: Izumi Shoin. Teramura, Hideo. 1992. Teramura Hideo ronbunshū I [Collected works of Hideo Teramura I]. Tokyo: Kurosio Shuppan. Tsukishima, Hiroshi and Harumichi Ishizuka. 1978. Tōyōbunko-zō Iwasaki-bon Nihonshoki [Iwasaki Manuscript of the Nihonshoki in the Oriental Library Collection]. Tokyo: Kichōbon Kankōkai. Tsukishima, Hiroshi and Yoshinori Kobayashi (eds.). 1980. Nakayama Hokekyō-ji-zōhon, Sangō Shikichū: Sōsakuin oyobi kenkyū [The Sangō Shikichū manuscript in the Nakayama Hokekyō-ji: Index and research]. Tokyo: Musashino Shoin. Tsunoda, Tasaku. 1996. Taigenjimebun [Mermaid constructions]. Nihongo bunpō no shomondai [Various issues in Japanese grammar]. 139–161. Tokyo: Hituzi Syobo. Yamada, Masahiro. 2010. Kakujoshi ‘ga’ no tsūjiteki kenkyū [A diachronic study of the case particle ga]. Tokyo: Hituzi Syobo. Yamada, Yoshio. 1908. Nihon bunpōron [A treatise on Japanese grammar]. Tokyo: Hōbunkan. Yamauchi, Yōichirō. 2003. Katsuyō to katsuyōkei no tsūjiteki kenkyū [A diachronic study of conjugation and conjugational forms]. Osaka: Seibundō. Yanagida, Seiji. 1973. Katsuyō kara mita shōmono no goi [Vocabulary of shōmono as viewed from conjugations]. Ehime Daigaku Kyōiku Gakubu Kiyō Dai II Bu Jinbun – Shakaikagaku 5(1). 1–21. Yoshida, Shigeaki. 2001. Bunmatsu yōgen no katsuyōkei ni tsuite [On the conjugational forms of sentence-final predicates]. Yamanobe no Michi 45. 1–16. Yoshida, Shigeaki. 2005. ‘Musubi’ no katsuyōkei ni tsuite [On conjugational forms in musubi]. Kokugo to Kokubungaku 82(11). 47–57. Yuzawa, Kōkichirō. 1929. Muromachi Jidai no gengo kenkyū [Studies on the language of the Muromachi period]. Tokyo: Ōokayama Shoten.
Takashi Nomura
13 The historical changes in the case marking system of Japanese 1 Case particles no and ga In this chapter I will describe changes in the case marking system of Japanese from the ancient times to the modern, focusing on subject marking. The history of the Japanese language may be divided into Old Japanese (the 7th and 8th century), Early Middle Japanese (9th to 12th century), Late Middle Japanese (13th to 15th century), Early Modern Japanese (from the 16th to the middle of the 19th century), and Late Modern Japanese (from the 19th century), but as there is little change in Japanese grammar between the 16th century and the Japanese of today, I will not address Modern Japanese (16th century onwards). This chapter draws on Nomura 1993a, 1993b, 1993c and 1996; see also Miyagawa 1989. From Old to Modern Japanese there has been the following three means of expressing a subject. a) Use of a kakari (focus) particle or an adverbial particle. b) Use of a case particle. c) No morphological marking (no particle). In (a), there is, as in (c), no use of a case particle, but issues of word order arise, and I will not address (a) in this chapter but will focus on (b). Morphological case marking of a subject was done by the particles ga and no in Old Japanese. Although ga and no have both the functions of marking a subject and of noun modification (genitive), differences in their functions may be observed. Ga marks the subject in a subordinate clause and unites a noun phrase with a following noun (possessive). (1) a.
b.
Subject Wa-ga yuku miti. 1-GA go path ‘the path along which I go’ Possessive wa-ga yadwo 1-GA house ‘my house’
Wa-ga in (1a) is the subject of the predicate yuku in a subordinate clause (here a noun-modifier (relative) clause). Wa-ga in (1b) is related to yadwo as possessive case. https://doi.org/10.1515/9781614512851-014
302
Takashi Nomura
The functions of no are more varied than those of ga and include the following. (2) a.
b.
c.
d.
e.
Subject Tomo-no yuku miti. friend-NO go path ‘the path along which my friend goes’ Possessive Tomo-no yadwo friend-NO house ‘the house of my friend’ Noun modifier Sugwi-no ita Japan.cedar-NO board ‘the board of Japan cedar’ Copula Tanabiku kumo-no awo.kumo hanging cloud-NO blue.clouds ‘the hanging and blue clouds’ Metaphorical expression (‘like’) Suga-no ne-no omopi midarete sedge-gen root-NO thinking being.distracted ‘being distracted like root of a suga’
(2a-c) exemplify functions observed also in present-day Japanese. (2c) expresses various types of noun-modifiers. (2d) has the function of connecting identical objects, like a copula. (2e) is an expression peculiar to Old Japanese, meaning ‘be like, as’, e.g., ‘(as complicated) as the root of sedge. Although ga has only two functions (1a) and (1b), no has all of the functions in (2a-e). That is, no can express all the functions that a noun can have in a clause. This is related to the difference in the kind of noun to which ga and no attach. The nouns to which ga attaches are primarily first-, second-, and third-person pronouns and proper nouns (wa-ga ‘I’, na-ga ‘you’, kimi-ga ‘my lord, you’, imo-ga ‘my sweet, you’, si-ga ‘it, he’, Tekwona-ga: ‘proper name’ etc.). No attaches to other nouns.
2 Semantics and syntax of ga and no Let us consider nouns in terms of the logical notation “f(x)”. For example, “∃x(f(x))” expresses that ‘there is something which is f’. So, the expression “∃x(dog(x))” means ‘there is a dog’. As it is not necessary to think about quantification now, I will consider only the expression “f(x).”
13 The historical changes in the case marking system of Japanese
303
The “f ” in “f(x)”, i.e., the portion which corresponds to “dog” in the previous example “dog(x)”, expresses the “attribute” of the individual or the set to which this expression refers. On the other hand, “x” is a variable. As Quine (1952, 1953) stated, it is pronouns or proper nouns which can provide the value of this variable (eventually Quine also reduced proper nouns to the notation “f(x)”). It seems that the properties of OJ ga can be well accounted for within Quine’s approach: In short, in the expression “∃xdog(x)∧walking(x)” (“the dog is walking”), “that which is walking” is an individual. The attribute “dog” is not “walking.” Also, the “I” who is the proprietor of “my house” is an individual. Since pronouns and proper nouns directly refer to individuals, they can only occur with ga. And when pronouns and proper nouns appear with ga, they can only perform the functions of subject (1a, 2a) and possessive (1b, 2b), but not the functions illustrated in (2c-e). This is because (2c-e) are attributive functions. On the other hand, no can be concerned with both the x (individual) of f (x) and the f (attribute) of f (x). Although x as a friend is “going” on in tomo no yuku miti (2a), in sugwi no ita (2c), sugwi is f (attribute) of f(x), that is, the material of a board. Though the case of copula, tanabiku kumo no awokumo in (2d), expresses x=y in f(x)∧g(y), this use of no has mostly fallen out of use since pre-modern times as x=y is redundant. Based on this we may understand the properties of ga and no: Ga relates a noun to other words as an individual, whereas no relates a noun to other words either as an individual or as an attribute. This holds for both subject marking, (1a, 2a), and noun modification (1b, 2b-e). Now, there is an argument which attempts to explain the fact that ga is mainly attached to first and second person pronouns based on the hypothesis that ga marks “ergative” or “active” subjects (e.g., Yanagida 2007; see also Yanagida this volume). However, that hypothesis forces us to explain independently the fact that the words to which ga or no attach in noun-modifier usage are the same as when they mark subjects. On the other hand, saying that the words to which ga attaches are pronouns and proper nouns is a sufficient explanation. Ga and no have the same syntactic restrictions when used as subject markers: Ga and no appear only as subject markers in clauses with the verb in the adnominal form (“adnominal clause” in the following), as exemplified in (3a-d) which show the different types of clauses which can have the predicate in the adnominal form in OJ, or in the provisional form (3e) or the conditional form (3f) (both ending in -ba). (3) a.
b.
Noun modifying clause (relative clause) wa-ga sekwo-ga pusa tawori-kyeru wominapyesi-kamo 1-poss husband-sbj bunch snap-pst.adn patrinia-sfp ‘It is a very ominaeshi, which my husband has snapped’ (MYS 17.3943) Nominalized clause moropito-no aswobu-wo mire-ba people-sbj play.adn-acc see-prov ‘when I see the people playing’ (MYS 5.843)
304
c.
d.
e.
f.
Takashi Nomura
Adnominal conclusive (rentaikei shūshi) wa-ga koromode-no puru toki-mo naki 1-poss sleeve-sbj dry.up time-part not.exist.adn ‘My sleeves have no chance to dry up.’ (MYS 10.1994) Kakari-musubi with predicate in the adnominal form Tabi-ni-ya kimi-ga yadori seru-ramu journey-loc-Q you-sbj stay do-conjec.adn ‘I suppose that you are staying at a lodge on a journey.’ (MYS 10.1918) Conditional clause Nagwo-no umi-ni sipo-no paya pwi-ba Nago-gen sea-loc tide-sbj early dry.up-cond asari si-ni ide-mu fishing do-to go.out-int ‘If the tide ebbs away earlier at Nago beach, we will go fishing at once.’ (MYS 18.4034) Provisional clause opokimi-no make-no manimani simamori-ni emperor-poss order-gen in.accordance.with island.defender-dat wa-ga tatikure-ba 1-sbj come-prov ‘When I came to this isle in accordance with the order of the emperor.’ (MYS 20.4408)
Adnominal, provisional, and conditional clauses are tightly unified, closed units, like “forts”, into which constituents which break up the internal structure cannot intrude, much like syntactic “islands” are closed units out of which constituents cannot be moved. For example, the topic marker pa establishes a clear break after it and so it cannot be used inside adnominal, provisional, or conditional clauses. On the other hand, ga and no are so strongly connected with the following word (as in wa-ga kwo or sugwi-no ita) that it is difficult to use ga or no as subject markers except in tightly unified clauses. These were syntactic properties of ga and no. In OJ and EMJ, ga is frequently found with the first-person pronoun wa. It seems that wa-ga forms a close-knit unit with the following word. As for second person, kimi ‘you, my Lord’, imo ‘you, my sweet’ etc., which were not originally pronouns, are usually used with ga, whereas na-ga (na second person pronoun) was rare. For third-person reference, expressions based on common nouns like so-no yama ‘that mountain’ were commonly used. Although ga could be attached to proper nouns, it was taboo to point out a person by using proper nouns in ancient days, and therefore use of proper nouns in fact tended to be avoided. Thus, the form wa-ga becomes to be conspicuous.
13 The historical changes in the case marking system of Japanese
305
3 Dependent form and independent form There was also another form of the first-person pronoun: ware. Wa is a dependent (bound, or compound) form used with another morpheme, as in wa-ga. On the other hand, ware is an independent form, or a free word. In addition to wa ~ ware, there were also pairs like na ~ nare (second person), ta ~ tare ‘who’, ko ~ kore ‘this’, so ~ sore ‘that’. Wa, na, ta, ko, and so were not used independently, but in complex forms such as wa-ga kwo ‘my child’, na-ga kwo ‘your child’, ta-ga kwo ‘whose child’, ko-no kwo ‘this child’ (which cannot mean ‘this one’s child’), so-no kwo ‘that child’. Accordingly, wa, na, ta, ko, and so could not be used as bare subjects (or bare objects). On the other hand, ware, nare, tare, kore, and sore could be used as bare subjects (and objects). Thus, we can say for example (4). (4) Ware ko-no miti-wo ika-mu. I this-gen road-acc go-int ‘I will go along on this road.’ But (5) is thought to be ungrammatical. (5)
Wa ko-no this-gen
✶
I
miti-wo road-acc
ika-mu. go-int
The dependent forms wa, na, etc. must be accompanied by ga (or another particle) or must be form a compound with another word, e.g., wa-gari ‘my place’. Of course, ✶waregari is thought to be ungrammatical. Note that the topic marker pa is found with both, e.g., ware-pa and wa-pa. Such pairs of dependent and independent forms, like wa ~ ware, were found widely in OJ and EMJ. For example, ma- is the dependent counterpart of the independent form me ‘eye’. It forms compound such as mabuta ‘eyelid’ (puta ‘lid’) and manazasi ‘gaze’ (na old particle corresponding to no, sasi ‘pointing’). Similar to ma ~ me, te (hand) is used independently, while ta is used in compounds, e.g., taduna ‘bridle’ (tuna ‘rope’); another example is independent pune ‘ship’ and dependent puna (e.g., in punabito (ship-man). In these cases, the vowels alternate. Another type are pairs like independent ami ‘net’ and dependent a, as in ago ‘fisherman’, analyzed as a + kwo ‘net + boy’, or independent asi ‘foot’ and dependent a, as in the verb agaku ‘struggle, paw the ground’ (kaku ‘scratch’). Kusuri (medicine) is an independent form, whereas kusu forms compounds, e.g., kusu-dama ‘medicine ball’ (tama ‘ball’). In these cases, the independent forms may be analyzed as a + mi, a + si, kusu + ri, that is, “dependent form + suffix”. Likewise, we can consider ware, nare, tare etc. to be of the type “dependent form + suffix”, whereas wa, na, and ta are the dependent form itself. Thus, wa-ga or na-ga were easy to unite with a following predicate when they were used as the subject in a clause, because they consisted of “dependent form + par-
306
Takashi Nomura
ticle”. It may be thought that this also explains why they can appear as a subject only in clauses with a strong internal unity. Wa-ga can be the subject in a clause with a question particle ka, e.g., (6). There are 20 such examples in the Man’yōshū. (6) kimi-ga yuki mosi pisa-ni ara-ba ume yanagwi tare-to My.lord-gen going if long-dat be-cond plums willows who-with tomoni ka wa-ga kaduraka-mu along q I-nom decorate-will ‘If you are gone, with whom may I decorate my hair with branches of plums and willows?’ (MYS 19.4238) In this example, wa-ga and the verb are adjacent and only wa-ga appears between the focused ka marked constituent and the verb. We find 10 instances of this kind out of the 20 examples of wa-ga as subject in a clause with ka. However, there are nine instances which have the word order [ka modifier wa-ga verb], i.e., with an adverbial modifier between ka and wa-ga, as in example (7). (7) Nani-ka kokodaku wa-ga kwopwiwataru what-q very.long I-nom being.in.love ‘Why am I in love a long time.’
(MYS 4.658)
Especially in these instances, it is conspicuous that wa-ga has a strong affinity for adjacency to the verb (see further Yanagida 2007). However, there is also one example in which an adverbial modifier intervenes between wa-ga and the verb, (8), so the adjacency requirement of wa-ga and verb is not absolute. (8) nanisutoka imo-ni apa-zu-te wa-ga pitori ne-mu why wife(=you)-dat meet-neg-ger 1-nom alone sleep-int ‘Why must I sleep alone without sharing the bed with you.’ (MYS 4.733) And in fact, there are other examples of wa-ga as a subject with a constituent between wa-ga and the verb. In particular, (9a) has a bare object (sime) between wa-ga and the verb. (9) a.
b.
wa-ga sime yupi-si yeda nara-me-yamo 1-nom mark.of.rope tie-pst branch be-conjec-q ‘Isn’t it the branch which I marked with a rope?’ wa-ga putari mi-si . . . tukwinokwi 1-nom two.persons see-pst . . . zelkova ‘the zelkova tree which I saw with you’
(MYS 3.400)
(MYS 2.210)
13 The historical changes in the case marking system of Japanese
307
Considering subject wa-ga, in a similar manner to wa-ga kwo ‘my child’, we see that wa-ga forms a tightly bound clause both in a noun-modifying clause wa-ga iku miti ‘the path I walk along’ and in a conditional or provisional clause wa-ga ike-ba ‘when I go’. In addition, wa is a dependent form that unites strongly with a following word or morpheme, in contrast with ware. Summing up the above two observations, it is apparent that wa-ga had the characteristic that it forms a tightly bound unit with the word following it. Therefore, it should be unacceptable for another word to intervene between wa-ga and the verb. However, given the existence of examples like ikani-ka kimi-ga pitori kwoyu-ramu ‘How do you go across alone’ (MYS 2.106), with pitori between ga-marked subject and verb, and also considering that the relation of wa-ga and mi-si in (9b) is no more than the relation between subject and predicate, it may be thought that it would have been fully possible for putari to occur between wa-ga and mi-si at the time of 7th or 8th century.
4 Change of the function of ga Now, there is a special usage of ga in the Man’yōshū, as shown in (10). (10) kusamakura tabi-ni-mo tuma-to aru-ga tomosisa (Epithet) journey-loc-even wife-com be.adn-gen envy ‘I envy the man with his wife even on a journey.’ (MYS 4.634) Tomosi is an emotive adjective of a meaning like ‘being enviable’ and tomosi-sa is nominalized by sa. Although tabi-ni-mo tuma-to aru-ga tomosisa is a noun phrase formally, the whole meaning becomes like ‘those who can be together with a wife on a trip are enviable’. That is, aru-ga seems to be a subject and tomosisa seems to be a predicate meaning ‘be enviable’. The important point in this usage is that attaching to a verb in the adnominal form, ga seems to mark the subject in this sentence. This pattern developed further and ga came to be able to be related as a subject not only to nouns derived by sa, but to the adjective itself (even without sa), as in (11). (11) Podo.naku makari-nu-beki-nameri-to omopu-ga soon go.out-perfect-ought-conjec-quot think.adn-nom kanasiku paberu-nari sad being-be ‘It is sad that you must be gone soon.’
(Taketori)
Also in this case, ga is used to mark the adnominal clause . . . to omopu. The condition of ga being used inside a tightly bound clause remains. Of course, kanasi is an emotive
308
Takashi Nomura
adjective, but in (12) the predicate to which ga relates is tokimeki tamapu, which is not an emotive adjective predicate. (12) Ito yangotonaki kipa ni-wa ara-nu-ga, sugurete very noble class be-part be-neg.adn-nom especially tokimeki-tamapu ari-keri prosperous-hon.adn be-pst.concl ‘Although it was not so noble, there was a very prosperous woman.’ (Genji, Kiritsubo) Since tokimeki-tamapu is the predicate of an adnominal clause (Ito . . . tokimeki tamapu), the restriction that the appearance of ga must be an inside of an adnominal clause (a clause with the predicate in the adnominal form) is maintained. Note further that the adnominal clause Ito . . . tokimeki tamapu is nominalized and is the subject of arikeri ‘was’, which is in the conclusive form. It can form the subject of a predicate in the conclusive form because there is no ga attached to the predicate tokimeki-tamapu. Furthermore, as time passes, we find examples like the following. (13) Kakuen-to ipu pito utayomi naru-ga ki-tari (personal name)-quot say person song.maker being-nom com-perfect ‘The man named Kakuyen who is a waka composer has come.’ (Ujishūi monogatari) In this case, since Kakuen to ipu pito is a bare subject, it is allowed to be the subject of a predicate in the conclusive form, ki-tari. But comparison with (12) seems to show that utayomi naru-ga ki-tari is parallel to sugurete tokimeki tamapu ari-keri, suggesting that a subject marked by ga is allowed to be related to a predicate in the conclusive form. It seems gradually to have become accepted for ga to mark the subject of a predicate in the conclusive form. Furthermore, as in (14), not only [adnominal form predicate + ga], but also [noun + ga] is found used as the subject of a predicate in the conclusive form. This is an extension of the use of ga. (14) Wara pito-sudi-ga kauzi mitu-ni nari-nu. straw one-CLF-nom mandarin.orange three-be become-perfect ‘One piece of straw changes to the 3 mandarin oranges.’ (Ujishūi monogatari) Moreover, examples of no marking the subject of a predicate in the conclusive form begin to appear.
13 The historical changes in the case marking system of Japanese
309
(15) Toshigoro ari-keru samurai-no tuma-ni gusi-te around.a.year be-pst warrior-nom wife-com follow-ger inaka-e ini-keri. country-direction return-pst ‘The warrior being around a year has gone with his wife to the countryside.’ (Ujishūi monogatari) It seems that the pattern [adnominal form predicate + ga] created an opportunity for ga to extend to be able to mark the subject of predicate in the conclusive form in a main clause, and that this also somehow brought no along to be able mark the subject of a verb in the conclusive form. This seems to have happened around the very early part of the LMJ period. Above, I have described the restrictions on the occurrence of subjects marked by ga or no. The rise of [ga/no . . . conclusive form predicate] was one cause for relaxation of the restrictions on the occurrence of ga/no marked subjects. Other than that, there were other factors during the LMJ period which led to a further relaxation of the restrictions, one of them being the ongoing change of the form of a main clause predicate from the conclusive to the adnominal form. As is well known, the predicate appears in the end of a sentence in Japanese. By default, the predicate at the end of a main clause will be in the conclusive form, but it can be in the adnominal form in some circumstances, for example with kakari-musubi, exemplified in (3d) above. There are also cases of ‘adnominal conclusive’ (rentaikei shūshi), as exemplified in (3c); these are considered to have some exclamatory force. In such sentences, the subject can be marked with ga/no. There are about 20 adnominal conclusive sentences found in Man’yōshū, whereas there are about 550 kakari-musubi sentences with the particle ka, which is the most frequent kakari particle in Man’yōshū, and so adnominal conclusive sentences are proportionally very few in comparison. However, in Genji monogatari there are about 400 examples of adnominal conclusive sentences, whereas there are around 1,700 examples of kakari-musubi with the particle namu (which is the most numerous kakari particle in Genji monogatari), showing that the relative frequency of adnominal conclusive sentences had increased significantly compared with the time of Man’yōshū. Accordingly, subject-marker ga/no also increased along with the increase in adnominal conclusive sentences. During the last years of Heian period Japanese (EMJ), written Japanese had gradually fossilized to a fixed classical style. Therefore, in the written materials from the 13–14th century (LMJ), that classical written style was adhered to and the number of examples of adnominal conclusive are not so frequent. However, in the spoken language, predicates of main clauses had by the 16th century completely changed to the adnominal form, and so it is clear that the relative frequency of adnominal conclusive increased through Man’yōshū (OJ) > Genji (EMJ) > 13/14th century (early LMJ) > 16th century (end of LMJ). As the frequency increased, the exclamatory nuance of adnominal conclusive was lost.
310
Takashi Nomura
On the other hand, in EMJ the kakari-particle in kakari-musubi constructions lost the important function of focus and came to express no more than emphasis. Thus, the kakari particles came to have more or less the same function as the present-day particles ne, na and sa (as in Taroo-ga-ne, Taroo-ga-na or Taroo-ga-sa), that is to say, with their presence or absence contributing little of significance to a sentence. For example, there are two versions of a passage from Genji monogatari, one with and one without the kakari-particle namu, with little or no change in meaning. (16) a. b.
“Raugawasiki koto-wa pabera-zi”-to-namu tanomi paberu. discourteous matter-top be-neg.fut-quot-NAMU hope polite.adn “Raugawasiki koto-wa paberazi”-to-Ø tanomi paberu. ‘I hope that discourteous matters will not take place.’ (Genji, Tamakazura)
Sentences such as (16b) with a dropped kakari particle have the same form as other sentences with adnominal conclusive and thus dropping the kakari-particles which had little functional contribution may also be thought to have contributed to the increase in sentences with adnominal conclusive. When adnominal conclusive increases, ga/no as subject markers should occur more often in main clauses. This would be expected to apply equally for ga and no, but in fact, subject marking by ga in main clauses increased greatly in the 16th century while the frequency of no-marking decreased. There must be a reason that ga appears easily in main clauses. We have already looked at a case like (13), in Ujishūi monogatari from the 13th century, in which [adnominal form predicate + ga] is used as the subject of a verb in the conclusive form in a main clause. In Genji monogatari, [adnominal form predicate + ga] was not yet used in that way and when a verb in the adnominal form was a subject of a verb in the conclusive form, it would usually appear without a particle, as in (17) where the clause with sitaru is the subject of otitarikeri. (17) Tataugami-no tenarawi-nado si-taru mikityau-no folding.papers-gen exercises-and.so.on do-perfect screen-gen moto-ni oti-tari-keri. foot-loc drop-perfect-pst ‘A folding paper on which someone wrote letters for practice is lying at the foot of a curtained screen.’ (Genji, Sakaki) However, since the 13th century, the number of examples of [adnominal form predicate + ga] subjects, such as (13), overwhelmingly came to exceed the number of [adnominal form predicate] subjects with no particle, such as (17). For example, while the number of the [adnominal form predicate + ga] subjects of Heike monogatari (kakuichi-bon, 14th century) is about 300, it only has three examples of [adnominal form predicate] subjects with no particle. As the result of the increase in [adnominal form predicate + ga] subjects,
13 The historical changes in the case marking system of Japanese
311
the kind of noun to which ga attached came to increase: Since clauses with an adnominal predicate are like relative clauses, they can refer to various objects. So, when clauses with the verb in the adnominal form are subjects, they may be thought to be semantically equivalent to various substantial nouns. For instance, although (17) has the formal subject si-taru without a particle, the substantial subject noun is a tataugami. In other words, in instances with [adnominal form predicate + ga] as the subject, although it is formally [adnominal form predicate + ga] which is the subject, it could substantially correspond to any noun, and thus ga lost the restrictions described in Sections 1 and 2 above, to occur only with certain nouns, but came to occur freely with all kinds of noun. Consider further examples like (18). (18) Podomonaku mata tatisowi-nu-beki-ga kutiosiku-mo soon again chase-perfect-ought-nom disappointed-part aru-beki-kana be-ought-excl ‘I am disappointed to be chasing after her soon.’ (Genji, Yūgao) The substantial content of the ga-marked subject here is the fact expressed by podomonaku mata tatisowi-nu-beki ‘I will soon be chasing after her again’. Consequently, subjects of the form koto-ga ‘thing, matter, affair’ actually started appearing in the 13th century, as in (19) which is written by Nichiren (the leader of one of the Buddhist sects) in the 13th century. (19) Zentisiki-ni au koto-ga daiiti-no kataki koto nari. virtue-dat meet matter-nom number.one-gen difficult matter be ‘It is most difficult to meet a virtue.’ (Nichiren) Koto is a kind of abstract noun. As a Buddhist theoretical leader, Nichiren makes frequent use of abstract nouns and in order to make an abstract noun a subject, ga is often used. (20) gokai-ga ari-kere-ba-koso [. . .] five.commandments-nom be-pst-cond-part ‘Only when there were five precepts [. . .]’ (Nichiren, Shōwateihon Nichiren-shōnin ibun) (21) pito-no aisuru-niyorite ibau-ga aru-naru-besi people-nom love-according influence.and.confidence-nom be-be-OUGHT ‘According to people’s love you certainly get influence and confidence.’ (Nichiren)
312
Takashi Nomura
Originally no attached to many more nouns than ga and there also were more examples of no than of ga, but as ga came to be used with verbs in the adnominal form, it therefore eventually came to be used with any noun. That is, although both ga and no both could mark subjects of main clauses freely around the 13th century, the restrictions on the nouns to which they could attach were lost only for ga when used to mark subject. Thus, ga came to be used to mark subjects of main clauses with no constraints and lost its function as a genitive and became a nominative case marker.
5 Conclusion To summarize, the establishment of ga as a nominative case particle may be considered a major historical change within Japanese syntax, together with the generalization of adnominal conclusive and the disappearance of kakari-musubi.
References Miyagawa, Shigeru. 1989. Structure and case marking in Japanese (Syntax and Semantics 22). New York: Academic Press. Nomura, Takashi. 1993a. Jōdai no no to ga ni tsuite, jō [On no and ga of Old Japanese, 1]. Kokugo Kokubun 62(2). 1–17. Nomura, Takashi. 1993b. Jōdai no no to ga ni tsuite, ge [On no and ga of Old Japanese, 2]. Kokugo Kokubun 62(3). 30–49. Nomura, Takashi. 1993c. Kodai kara chūsei no no to ga [No and ga from Old to Late Middle Japanese]. Nihongogaku 12(11). 23–33. Nomura, Takashi. 1996. Ga - shūshikei e [Ga with the conclusive form of verbs]. Kokugo Kokubun 65(11). 524–540. Yanagida, Yuko. 2007. Jōdai nihongo no nōkakusei ni tsuite [On ergativity in Old Japanese]. In Nobuko Hasegawa (ed.), Nihongo no shubun genshō [Main clause phenomena in Japanese], 147–188. Tokyo: Hituzi Syobo. Yanagida, Yuko. This volume. Differential argument marking in Old Japanese: Morphology, semantics and syntax. In Bjarke Frellesvig and Satoshi Kinsui (eds.), Handbook of historical Japanese linguistics. Berlin and Boston: De Gruyter Mouton. Quine, Willard van Orman. 1952. Methods of logic. London: Routledge and Kegan Paul. Quine, Willard van Orman. 1953. From a logical point of view: Nine logico-philosophical essays. Cambridge, MA: Harvard University Press.
Charles J. Quinn
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate 1 Introduction This study surveys basic features of the adnominally concluded kakari-musubi (“km”) construction in Old and Early Middle Japanese (“oj”, “emj”), and examines its functional variability in discourse context. Part 1 characterizes km as a family of related constructions in which a focus-presupposition, cleft-like variant dominated, while nevertheless producing many non-cleft offspring. A cleft-like km was a device for treating certain information as news, in relation to a given1 situation in which it has a role. Part 1 introduces basic terms and concepts as it analyzes representative km tokens synchronically, noting how the functions of km’s components relate systemically to their uses in other constructions. It also establishes how those components relate to each other, within an individual km token and in relation to that token’s discourse context. Part 1 concludes with a comparison, in the same discourse context, of an actual km and a grammatical but non-km paraphrase, to reveal how the km meets the context’s communicative needs, while the non-km does not. Part 2 begins by confirming why a cleft-type km, with its adnominal musubi, would have been basic to km. While basic, however, it was open to the rhetorical innovation of a musubi that was informative, i.e. one whose contents were not given: still presupposed by the speaker, but news to the audience. This distinction is inspired by Prince’s (1978) construct of informative presupposition in English it-clefts. Part 2 restricts itself to declarative zo km with a rentai-inflected predicator (“pRT”) musubi. It first confirms, with reference to these tokens’ oj and emj discourse contexts, the existence of zo-focused km with both given and informative musubi. In exploring the pragmatic and rhetorical fit of these informative musubi to their contexts, two effects become clear: (1) an authoritative emphasis of the factuality of the entire km’s contents, and (2) a consequent functional shift from a cleft-like meaning to that of another oj/emj construction, the rentaikei shūshi ‘adnominal conclusive’, which we characterize as a referential predicate (“rp”). A rp was a designative predication (using zero copula, n(i)ari, zo, or ka) of an adnominally inflected nominal clause (“pRT”), for the purpose of characterizing a presupposed situation at issue (“sai”) – typically Acknowledgements: I thank the volume editors for their support, in particular Bjarke Frellesvig. For years of feedback, insights and encouragement, I am grateful to Shelley Fenno Quinn, Shoichi Iwasaki, Leon Serafim, Rumiko Shinzato, Kaoru Horie, Jim Unger and John Whitman. 1 Information known to both speaker and audience in a discourse, after Prince (1978). https://doi.org/10.1515/9781614512851-015
314
Charles J. Quinn
the one at hand in the discourse. A rp gave its situation at issue a proper identification. km tokens that shifted to this meaning are dubbed “km-as-rp”. The phenomenon itself has long been acknowledged in Japanese philology and in theories of km that claim it originated as a rp variant. km-as-rp appears to have been general for any declarative km with an informative musubi. We attribute this to the interaction of a generalized conversational implicature (implicit in declarative km) with an informative musubi. By oj and emj, this functional shift was already making km less cleft-like, less of a “term-focus” construction in the sense of Heine and Reh (1984). It is also proposed that the rp, as a situation-presupposing and identifying predication, was itself a focus construction, of Heine and Reh’s “predicate focus” type, but with the same two-part focus-and-presupposition structure of classic km. These features are a constant in all of the rp’s rhetorical uses (explanatory, exclamation, etc.). The perspective from Japanese suggests that a tacitly presupposed sai (situation at issue) may have grounded the predicate-focus constructions in Heine and Reh’s corpus too. In conclusion we briefly consider pRT-concordant km and the rp construction in cross-linguistic context. As of emj, km’s koso variant alone was semantically opaque, or “strongly grammaticalized” in Heine and Reh’s phrase. Adnominal-concordant declarative members of the km family remained semantically transparent (“weakly grammaticalized”), even as the penchant for informative musubi, already strong in oj, continued to deliver many km-as-rp tokens. This rhetorical device would continue to bleach kps of their designative meaning and invite reinterpretation of the musubi as the sentence’s matrix predicator, as km increasingly became a monoclausal focus/rp construction. An informative musubi was a major factor in initiating and driving these changes.
1.1 km as a figure/ground cleft construction: parts, functions, sources Kakari-musubi is a construction well attested in the texts of oj and emj. A word or phrase in a sentence is marked with an emphatic kakari particle (“kp”) – zo, ka, ya, namu, or koso, and the sentence then concludes not with the unmarked conclusive form, but with one of two nominal forms. The emphasized element is called the construction’s kakari (“k”, an ‘attaching’ or ‘onset’, which anticipates a corresponding closure). The k’s associated predicator phrase is the musubi ‘tying-up’ or ‘closure’ (“m”). The focal kakari and presupposed musubi together comprise the km construction, which is illustrated in the example below with identifying (“id”) kp so (which in oj alternates with zo). The poem is one among several composed by travelers while kept in port by rough seas.
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
(1)
315
Kaze pukeba oki tu siranami kasikwomito Noko no wind blow.prov offing gen white.waves being.fearful Noko gen tomari ni [[amata ywo]2 so]K [nuru]M. harbor loc many nights id sleep.adn ‘Since the winds blow (on), in fear of the waves at sea, at the harbor at Noko [it’s [many nights]]K [that we bed down]M.’2 (MYS 15.3673)
kp so identifies amata ywo ‘many nights’, and amata ywo so ‘it’s many nights’ is the kakari (“k”), the focused part of this message. In making this identification (as in making any identification), the speaker presupposes a referent or frame-of-reference, which is referred to with the musubi (“m”), the nominal nuru ‘that (we) sleep’. For ending a sentence, the conclusive form (nu ‘lies down, sleeps’) is the unmarked option, but this km sentence ends with with the verb’s adnominally inflected form, nuru. The bare, adnominally inflected clause nuru was an ad-hoc, deictic3 nominal (or “pronominal”), which might refer to ‘(our) sleeping’ or, in other contexts, ‘those who sleep’, ‘the time when we sleep’, etc., much like its ModJ analogue, neru no. Table 1 presents the km family. Table 1: km types with kakari ‘focus’ particles interpreted predicatively. kakari [focused x]K
musubi [presupposed p(redicate)]M
[x zo]K [x ka]K [x namu]K [x ya]K
pRT. pRT. pRT. pRT.
x koso
pIZ.
‘[It’s x] [(the one) that/who/etc. p]M.’ ‘[Is it x] [(the one) that/who/etc. p]M?’ ‘[It would be x] [(the one) that/wh- p]M.’ ‘Is it [(the case) that [x, p]]NP?’ (few ‘[Is it x] ]K [(the one) that/wh- p]M?’) ‘[x, it’s this one]K [given/provided that p]M.’
Example (1)’s kp so/zo4 was one of four kps in Table 1 whose corresponding musubi was a predicator5 inflected in its rentaikei ‘adnominal form’ (“adn”). The kp koso took
2 The kakari (bracketed and tagged with “K”) in the transcriptions includes the (originally predicative) KP. The kakari’s inner brackets enclose the complement of the KP, which it designated or focalized. The musubi is bracketed and tagged with an “M”. 3 Deictic in that the form’s use, as with /finite p + no/ ‘the one(s) th-/wh- p’ nominals today, assumes a referent known, or grounded, in a discourse the speaker is party to. pRT nominals (like ModJ /finite p + no/ nominals) do not refer to a category in the manner of an unspecified common noun. On the deixis of “knownness” (epistemic deixis), see Akatsuka-McCawley (1978), Akatsuka (1985), Givón (1982), Frawley (1992, especially 412 ff), Hanks (1992) or Ochs (1996). Like deixis in other domains, it is most typically grounded in the speaker. 4 We will use zo to refer to this particle as such (vis-à-vis the other kps), but transcribe it in examples as the so or zo it appears as. 5 An inflecting word (yōgen), i.e. verb, inflecting adjective, or copula. Abbreviated “p”, to refer to a p-headed predicate phrase of any length, usually a simplex clause.
316
Charles J. Quinn
a musubi inflected in its izenkei ‘realized form’ (“iz”), a presupposing infinitive6 (“pi”). The translations attached to each km type in Table 1 are representative – a pared-down sense derived from just the designative uses of kps zo, ka, etc. on the left and the deictic nominals on the right. The idea is that from these characterizations, km tokens that call for a different construal can be consistently accounted for. Each of the five kakari types above created its own kind of focus, but by the hypothesis espoused here, they all started out as a designative, copular sort of predication (‘It is ~’, ‘It would be ~’, ‘Is it ~?, etc.). Declarative km sentences are served by kps that make some kind of identification: zo (simple identification), koso (unique or epitomizing), and namu (tentative confirmation). Interrogative km sentences employ confirmation-soliciting ya (yes/no), doubted-identity ka ([wh-] is it that ~?’; ‘Is it x that ~?’), and zo, in wh- questions only (cf. the use of copula da today in Dare da. ‘Who is it.’). Ya and ka also helped index exclamatory stances, typically of a mirative or incredulous sort. The pRT or pIZ form of the musubi (Table 1, right side) referred most generally to the frame of reference, or scope, of the kakari’s specification, as the translations indicate. When combined with the identifying function of the kakari, this interpretation of the musubi characterizes the km family as a set of variations on a cleft-like ‘It’s x that p.’ construction. Identifying this focus-presupposition structure as basic to km is useful not only in understanding individual km tokens in oj and emj, but also in looking back and reconstructing km’s origins, on the one hand, and in accounting for non-cleft km tokens in oj and emj. It also gives km a developmental profile congruent with that noted of focus constructions cross-linguistically (e.g. Heine and Reh (1984), Harris and Campbell (1995)). While a situational, or open-proposition,7 characterization of km’s adnominally inflected (pRT) musubi will resemble the relative clause in an English it-cleft (e.g. It’s you that I like.), the parenthesized ‘the one’ in the translations of Table 1 is a reminder that the pRT musubi likely began as a pRT nominal (‘the one that/wh- p’) that referred to a person, thing, place, act, time, manner, extent, etc. – some facet of our experience readily individuated in talk and a viable sentence topic.8 This would reconstruct, for 6 On this characterization, see Quinn 1987, Chapter 9. The form is labeled “evidential” in Vovin (2003) and “exclamatory” in Frellesvig (2010). These uses of pIZ are quite compatible with the form’s nonfinite, presupposing functionality. 7 A proposition with one of its terms open (i.e. inchoate, to-be-determined), which the kakari fills in, completing the proposition—thus Heine and Reh’s (1984: 147) characterization of such constructions as “completive focus”. 8 An entity-referring “(the one)” interpretation remained possible in many km tokens through emj, even as the km construction’s musubi came to be regarded as a more general or abstract reference to the assumed situation/proposition, ‘that p’. The simple relative-clause translation of a musubi (without the parenthetical ‘the one’) represents the more grammaticized km, for which the musubi was no longer a postposed topic. Frellesvig’s (2010: 250–251) glosses of the musubi in oj focus-presupposition examples are similarly entity-referring, deictic nps, e.g. miyeturuRT ‘(what) I saw. . .’; okinikyeru ‘(what) has fallen’.
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
317
example, the English cleft example just above as something along the lines of It’s you – the one that I like. This study hypothesizes that the more general, situation9-referring pRT musubi translated as ‘that p’ and described as an open proposition, evolved out of an afterthought, right-dislocated nominal reference (pRT) to the topic, a given (mutually known) discourse entity (‘the one that/wh-’). That is, when pRT was still a right-dislocated sentence topic (and not yet a grammaticalized musubi), it would have referred to an individuated, given discourse entity: ‘the one that p’, ‘that which p’, ‘what (they) p’, etc. – all of which were expressable as pRT in oj and emj – rather than to the event or state (open proposition) p per se. A less individuated, more general construal of the pRT musubi as a reference to a partly inchoate situation (or “open proposition”) would have developed in grammaticalization, as the musubi came to be seen less often as a postposed nominal topic, and more of a marked type of predicator phrase. The entity-referring, right-dislocated topic pRT that we propose grammaticalized into km’s musubi would have been an actual nominal manifestation of the constituent that, in a thoroughly grammaticalized km, we analyze as gapped in the open proposition represented as pRT (or as the that-clause of an English it-cleft). Of course, the many pRT musubi tokens that still make sense as references to ‘the one that/who/when/etc. p’, also support a relative-clause ‘that/wh- p’ construal. At the bottom of Table 1, focus with kp koso is paired with the izenkei for its musubi. This paper will not take up koso-focused km, with its exceptional musubi, but as the translation given in Table 1 indicates, we hypothesize that the focus created with koso in this kakari too began with a designative variation, an /x, ko so./ ‘x – it’s this.’ predication. There are reasons to think the pIZ musubi associated with koso began as a postposed provisional, as translated.10 With the variants listed in Table 1, the km family offered related but individually distinct ways of structuring a sentence’s contents as a message, or staging of differentiated information. Up front came the message’s main point, the kakari, concerned with identification, while the musubi brought up the rear, referring to the frame-of-reference (originally an individuated referent) assumed for identification with the kakari. This fronting of the message’s main focus, complemented by the topic-like scope coming last, effectively reversed a sentence’s unmarked message structure of “old information, then new”, where the final predicator phrase was focal. This fact inspired Ohno’s tōchi-hō ‘inversion construction’ account of km (1993, inter alia). While the kakari’s focus seems to have originally been the product of one of these five variations on a specificational ‘it is x’/ ‘is it x?’ identification (affirmed or sought) of x (the kakari’s complement), there are many km tokens in oj and emj in which the kakari’s focus does not quite work as a copula-like, designative ‘it is/is it x’. An example is the
9 “Situation” refers to states and events alike; as a predicate it is often represented with the abbreviation “p” for ‘predicate/predicator’ (verb, adjective, copula). 10 Proposed by Quinn (2015). See also Quinn (2024).
318
Charles J. Quinn
first translation paired with interrogative kp ya in Table 1, /x ya pRT./ ‘Is it [(the case) that [x, p]]np?’ In this apparently non-cleft reading (long widely accepted; see JDB 755), interrogative ya is said to apply to the ‘full sentence’. Further, as Table 1’s first translation for ya km suggests, most /x ya pRT./ km questions in oj and emj do not seem to have been simple finite clauses, but rather ‘Is it [(the case) that [full proposition x, p]]?’ questions, which ask if a presupposed situation is properly characterized with [x, p] together. Thus, this bracketed “[(the case) that [x, p]]NP” was the complement in an interrogative ‘Is it [(the case) that [x, p]]NP?’ designative predicate. To my knowledge, this “the-case-that” interpretation is not part of the long-standing ‘full-sentence’ construal of ya’s scope, but it fits well with evidence presented for zo in Part 2 below (and with namu and koso km tokens too, although these are not taken up). This is a significant point, because if an entire interrogative ya km referred to a presupposed situation, that makes it functionally similar to questions posed with a situation-referring pRT, namely /pRT ka./ ‘Is it (= the situation) that p?’11 – the ancestor of today’s extended predicate /finite p + no (cop)?/ ‘It’s that p?’ questions. Part 2 of this paper will analyze declarative km tokens with zo to show that in such non-cleft interpretations, an entire nominal pRT ‘[(the case) that [x, p]]’ was the complement of an implicated designative predication. These non-cleft, declarative km interpretations were prompted under identifiable conditions, assisted by conversational implicature, and are also attested with namu and koso. Through oj and emj, non-cleft, km tokens with a rp-like function co-existed with the cleft-like km tokens represented as basic in Table 1.
1.2 Designative kakari with zo or ka and a given-information musubi We will first flesh out the above characterizations of km with declarative examples with kp zo and an interrogative one with ka, with attention to the token’s fit in its discourse context. The first is focused with the oj variant of zo, so – a quasi-12copular, designative predicator. The male speaker is responding to a poem (the preceding MYS 2.98) from a woman, which expressed her concern for how long his professed affection would last. His reply opens with the same formulaic epithet (makura kotoba ‘pillow word’) that hers did, adusa yumi ‘the catalpa bow’, which initiates the piku ‘draw/attract’ metaphor that follows. His message is a single sentence that begins with a topic (“top”) and concludes with a km:
11 This is the rentaikei shūshi ‘adnominal-form conclusive’ or rentaidome ‘adnominally stopped’ sentence, analyzed as a “referential predicate” (“rp”) in Part 2 below. 12 The “quasi-copular” labeling of zo and ka acknowledges that while functionally designative, unlike copulas derived with ari ‘be’ (e.g. n(i)ari), zo and ka did not inflect.
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
(2)
319
Adusa-yumi [turawo toripake piku pito pa]TOP [[noti no catalpa-bow bowstring attach.inf draw.adn person rf later gen kokoro wo siru pito]13 so]K [piku]M. heart acc know.adn person id draw.adn ‘The catalpa bow – the man who strings it up and draws it, [it’s [a man who knows his heart going forward]]K [(the one) who will draw it/attract you]M.’ (MYS 2.99)
With the pa-marked topic, the speaker establishes ‘the man (pito ‘person’) who will string up and draw’ (tori.pake piku) the catalpa bow’ as given information in this discourse. With the rest of his sentence – the km construction – he comments assertively and definitively on that topic. kp so (like zo) functions as a non-inflecting, non-finite copula here, identifying the referent of the initial, pa-marked topic (and of the musubi to follow, piku) as noti no kokoro wo siru pito ‘a man who knows his heart going forward’ (i.e. whose affection will remain constant). The conceit of the bow, carried over from the woman’s poem with the pun in the verb piku ‘draw (the bow)’/‘attract (someone)’, delivers the intended message: ‘As for the man who step-by-step attracts you, it’s a man who knows his heart going forward, (the one) who attracts you.’ This is a classic cleft-like km, with given information in the musubi. The focal kakari consists of the designative predicate [noti no kokoro wo siru pito]N so ‘It’s [the man who knows his heart going forward]’; this identifies just what kind of man it is that the musubi refers to. The musubi thus serves as the km construction’s grammaticalized internal “topic”, or scope-setter, for the kakari’s identification. In making the identification that constitutes this kakari, the speaker presupposes that there is someone ‘who draws the Azuma bow/who will attract you’. An ‘It’s x.’ identification (as well as an ‘Is it x?’ or ‘Wh- is it?’ request for one) logically entails the existence of a known but in some respect unclear referent, which was expressable in oj and emj as pRT ‘that which p’, ‘the one th-/wh- p’. A km construction made explicit reference to this presupposed referent with its pRT musubi. In (2), the presupposed referent that the pikuRT musubi refers to (‘the one/the man who will draw/attract’) is identified with kp so as the type of man described with the kakari’s complement, noti no kokoro wo siru pito ‘a man who knows how he’ll feel’. In tokens where the pRT is an eventive verb phrase, the item identified in the kakari might be the agent, undergoer, or an adjunct (circumstances, purpose, reason, etc.) in the situation, i.e. a term in the open proposition referred to with the verb-headed musubi.
13 The transcriptions mark the KP’s complement (“X”) by bracketing it just under/left of the KP and within the kakari’s matrix (subscripted “k”) bracketing: [ [ X ] KP]k.
320
Charles J. Quinn
It was often the case in oj and emj that what the speaker presupposed in making an identification, i.e. the musubi’s content, was also known to his audience,14 whether through prior mention (e.g. the pa-marked topic in (2), or the shared experience that occasioned the poem of (1)) or, in still other contexts, by inference (e.g. (7) below). We refer to this combination of (a) speaker-presupposed with (b) audience-known information in a musubi as given information, following Prince (1978). A given musubi was likely a baseline of sorts for the km construction, but there were more than a few exceptions to this combination in both oj and emj. An example of emj zo will supplement the preceding oj examples with so. The bamboo cutter of the eponymous tale uses it in describing his daughter Kaguyahime to the Emperor: (3)
. . . kokoro.base mo [[yo no pito ni nizu] zo]K disposition if world gen people loc resemble.neg.inf id [paberu]M be.hum.adn ‘. . . (her) disposition too – [it’s [unlike people of (this) world]]K [(the way) that she is]M.’ (Taketori, p. 92)
Like so in examples (1) and (2), id zo ‘it is’ in (3) creates focus by identifying its complement, the adjunct [yo no pito ni nizu] ‘not resembling people in the world’, as the manner of her presupposed ‘being’ (paberuRT). How is it that so/zo provided this focus-creating designative predication to km? The quasi-copular so/zo that created focus in the km construction was also used in oj and emj as a sentence-final designative predicator, as in (4): (4)
. . . kwopu to ipu koto pa koto no nagusa so. love.concl comp say.adn act rf words cop sop id ‘. . .for (my husband) to say he loves (me), it’s just a verbal sop!’ (MYS 4.656)
This designative predicator, so, put what it identified in focus, and would also do so in the kakari of a km construction. When enough speakers developed a preference for adding a referent-clarifying pRT nominal topic onto designative predications like (4), the components would be in place for a new, two-part construction to develop, which in time became the zo-clefted km. Much the same process seems to have created km focused with kp ka, which we introduce next.15 14 The term audience applies to addressees, onlookers, or readers as needed. 15 This theory is proposed in Quinn (1997), which also argues that quasi-copular ka and so (both sentence-final and then as kps) originated in right-dislocated demonstrative pronouns ka and so, deployed in reference to the known but partly inchoate topic. This and other origin scenarios for pRT-concordant km are reviewed in Shinzato and Serafim (2013, Chapter 5).
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
321
The km construction was of course also used in posing questions. With doubted identity (“di”) kp ka, the speaker marks a phrase as being at issue identity-wise and seeks to clarify the matter. Ka was the default kp for focusing wh- questions. The Emperor uses it below, in asking his summoned ministers and courtiers ‘which mountain’ it is ‘that is closest to the heavens’: (5)
[[Idure no yama] kaKP]K [ten ni tikaki]M which.one gen mountain di heavens loc being.near.adn ‘[[Which mountain] is it]K [(the one) that is close(st) to the heavens]M?’ (Taketori, p. 107)
With di ka, he seeks the identification of idure no yama ‘which mountain’ it is that fits the reference he makes with the musubi, ten ni tikakiRT ‘(the one) that is close(st) to the heavens’. With /x ka/, x was usually a wh- phrase, with which the speaker sought an identification, as here with idure no yama ‘which mountain’. Or, when x expressed specified content, ka sought to confirm that x was the right identification for the musubi – the actual value of the open term in its proposition. This latter ‘Is it x?’ use is illustrated with the sentence-concluding tokens of ka in (6) below, in three short questions from a poem. They represent the poet’s elegantly confused attempt to identify a planted display of white chrysanthemums that mimicks the shape of a famous beach: (6)
[pana]NP ka. [aranuRT]NP ka. [nami no yosuruRT]NP ka. blossoms di be.neg.adn di waves gen press.near.adn di ‘Are those [flowers]N? Is it [(the case) that they are not]N? Are they [waves breaking shoreward]N?’ (Kokinshū 5.272)
Each of the three nominal sentences in (6) is concluded with quasi-copular di ka.16 The first consists of a single noun, pana ‘blossoms’, predicated with ka (thus bracketed as ka’s complement). di ka also predicates each of the next two nominals, aranuRT ‘(the case) that they are not’ and nami no yosuruRT ‘the surf’s pressing shoreward’. Note that each sentential question presupposes a referent, something that it entertains an identification for, just as a ka-marked kakari presupposes the referent comprising its musubi. When, as in aranuRT ka and . . . yosuruRT ka, ka’s nominal complement is pRT, ka is often referred to as a shū joshi ‘sentence-final particle’, but its identification-seeking function was basic here too.17 The three questions in (6) illustrate ka’s use as a sentence-ending designative, which asks if each of its complements (pana, aranu, . . . yosuru) is the proper identification of the presupposed referent, the “topiary” island. Most basically, di ka problematized, or
16 “skp” (sentential kp) ka in Shinzato and Serafim (2013). 17 /pRT ka./ ‘Is it (the case) that p?’ was thus an interrogative referential predicate; see §2.3 below.
322
Charles J. Quinn
put at issue, its complement’s identity, and this is just what ka continued to do when a pRT was added, in afterthought fashion, to clarify the referent of that problematizing. As with zo, from this “afterthought topic” arrangement a km construction emerged. di ka qualified and positioned itself for its role in km by the same route that identifying, quasi-copular zo did: as a quasi-copular designative predicator. A constraint on the sentence-final use of zo and ka that remained in place through emj was that they predicated nominals, i.e. a lexical noun phrase or a nominal form of a predicator, usually pRT. They did not follow the unmarked conclusive form – a fact that indicates there was still something designative, or quasi-copular, to them. And, as the preceding translations indicate, there was something like an English it-cleft in the way km tokens used a designative zo or ka to create a focal figure against the musubi’s ground of given information.
1.3 km with a given pRT musubi in discourse context Like the complement of the designative It’s ~ predicate in an English it-cleft, e.g. It’s the father (that) I don’t trust., the kakari in the km construction has been said to constitute “new information”, “message focus” and the like. And like an English cleft’s relative clause, the musubi in km has been said to represent “old information”, information that is “assumed”, “presupposed” or “given”. These characterizations in turn point us to a discourse – which, however, begs other questions, such as: assumed by whom? New or old by which criteria, and for whom? Are “new information” and “focus” different labels for the same thing? If information is presupposed in the act of making an identification, is it necessarily “old” or “given” information at that point in its discourse? Dwight Bolinger (e.g. 1972) and Ellen Prince (e.g. 1978) were pioneers in exploring how English cleft sentences fit their discourse contexts, and the analyses presented below will take some cues from their work, especially Prince (1978). For grasping a speaker’s or writer’s motivation for deploying a km construction, a sense of its discourse context is essential. We will illustrate with an emj zo km whose musubi’s contents are familiar to both speaker and audience. Genji is living seaside at Suma, in exile from the capital. Immediately preceding (7) below, the narrator has related that the Kyushu Viceroy, his term there ended, is returning with his family to the capital, and they are about to pass Suma. The Viceroy sends a letter to Genji and apologizes for not being able to pay his respects in person. The letter is quoted at length (Genji’s reading it is implied), and the Viceroy’s polite apologies and excuses, like his promise of a proper visit later, make it clear that paying Genji a visit would be the expected thing. The arrival of the letter implies that someone has delivered it, and as soon as the quoting from it ends, the narrator adds (7), informing us that it is the Viceroy’s son who has come, i.e. with (implied) the letter. Thus, as we read (7), its musubi’s content – that someone ‘has come’ to Genji – is given information – presupposed by the speaker and known to the audience.
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
(7)
323
[[Ko no Tikuzen no kami] zo]K [mawir-eru]M. child cop.adn Chikuzen gen governor id come.hum-stat.adn ‘[It was18 [(the/his) son, the Governor of Chikuzen]]K, [(the one) who had come]M.’ (Genji, Suma, 2.196)
The message of (7) is structured this way: zo identifies ko no Tikuzen no kami ‘the son, the Governor of Chikuzen’ as the person who (implied: with the letter just quoted) has called on Genji. The musubi is inflected in its adnominal form, mawireruRT, in concord with zo’s identification of the son. The message’s point, or news, then, is not that someone has come (which is already clear) but rather who has come – that it was this person, who is thus introduced to the story. His relation to Genji is then featured in the narrative for a few sentences. The referent of the pRT nominal that forms the musubi is the situation/open proposition – someone’s having come to Genji – presupposed in the process of identifying ‘the son . . .’ with zo. As with it-type cleft constructions in other languages, contrast is a common consequence of a zo-focused km. Example (7) also warrants a comment about new information. The identified complement of zo (ko no Tikuzen no kami ‘the son, the Governor of Chikuzen’) is news in more than one sense. As noted, he is newly identified as the actor in the presupposed situation referred to as mawireru ‘that (someone) has come (with the letter)’. At the same time but on another level, he also appears for the first time in this discourse. The distinction between the two is one that is easy to overlook, because “new” is often used to refer simply to the introduction of a referent to the discourse. The newness that zo (or any of the declarative kps) indicates for its complement may or may not involve the introduction of that complement to the discourse, but in declarative km tokens the kakari’s complement will always be news vis-à-vis the presupposed situation referred to with the musubi – the open proposition that it completes and closes. This local, construction-internal newness would hold even if the ko ‘the son’ in the above example had been previously introduced in the story: he would be already known to the reader as a character in the story, but until (7) would not be known as the agent of this anticipated visit. We have thus far shown how zo and ka were used in making or seeking an identification. As for the musubi in the examples surveyed, whether the focus-creating kp is id (‘identifying’) zo or di (‘doubted identity’) ka, the information that the speaker presupposes and refers to with the musubi ‘closure’ is given, i.e. not just presupposed by the speaker but also assumed by the speaker to be known to the audience. km tokens with a given musubi correspond to the type of English it-cleft that Prince (1978: 896) termed “stressed-focus it-clefts”, where “the focus represents new, often contrastive information,” and “the that-clause represents known or old information” in the
18 Zo’s predication and that of the musubi will generally be tensed in our translations according to the larger narrative frame. The KPs themselves are tenseless.
324
Charles J. Quinn
discourse (1978: 896). We will refer to km tokens of this sort as “cleft-type” km; their musubi is given. Here is one of Prince’s English examples (1978: 897, (40a)) with her analysis: (8)
40a. C.B.: So who’s Barbara? B.S.: Let me put it this way. When you last saw me with anyone, it was Barbara I was with. presupposed information: I was with someone. information that is news: Barbara was that someone.
For the same conversational context, Prince considers a “b” version of (8)’s cleft it was Barbara I was with, and comments that it is “less natural” than the original’s cleft: (9)
40b.
. . . I was with Barbara. The it-cleft in 40a extracts the new information, Barbara, from final position and places it before the old information, ø I was with. Surprisingly, the non-clefted form, 40b, which has everything in the ‘right’ order, is less natural (Prince 1978: 897).
??
The problem with (9) (her 40b), writes Prince, “is that the delineation between what is old and what is new is unclear and misleading”; the it-cleft, by contrast, “though [it] presents information (old vs. new) in an aberrant order, . . . clearly marks which is which” (Prince 1978: 897). Prince’s examples 40a and 40b have a parallel, structurally and in their respective fit and misfit to their contexts, with our earlier example (7), presented here again as (10), and a made-up, non-km example (11) predicated with the unmarked conclusive mawireri ‘had (lit. has) come.’ (10) [[Ko no Tikuzen no kami] zo]K [mawireru]M. child cop.adn Chikuzen gen governor id come.hum-stat.adn ‘[It was (the/his) son, the Governor of Chikuzen]K, [(the one) who had come]M.’ (Genji, Suma, 2.196) (11) Ko no Tikuzen no kami mawireri. child cop.adn Chikuzen gen governor come.hum-stat.concl ‘The son, the Governor of Chikuzen, had come.’ As explained earlier, in the text preceding (10) in the Genji it becomes clear that there has been a visit from someone among the entourage passing Genji’s place of exile. Just who that is is clarified with the identification of the actual visitor in (10). That is, in both (8) (Prince’s (40a)) and (10) from Genji, prior discourse has created a given situation, and a rhetorical exigence, or opening, for a clarifying identification. In (8) and (10),
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
325
respectively, the it-cleft and the zo-focused km (a) acknowledge the given situation while (b) making the relevant identification. Both the it-cleft’s relative clause (ø I was with) in (8) and the mawireruRT musubi of (10) acknowledge their content’s givenness, established in the discourse that precedes each. Their canonical counterparts (9) and (11) do the opposite – instead of referring to information that is already known as such, they present their contents as news. (9) and (11) are therefore bad fits for their respective discourse contexts, in that both (a) fail to identify the entity in need of identifying, and (b) present anew, rather than refer to, information already established. Each is thus doubly irrelevant in its discourse context. The it-cleft in (8) and the zo km of (10), by contrast, respond aptly to their respective contexts’ exigences. There are many km in oj and emj like the foregoing, where the narrator or speaker presupposes, in the act of making or seeking an identification, information that is already known to or readily inferred by the audience. Indeed, this cleft-like arrangement seems to have been a default of sorts – a kakari’s complement newly identified (or considered for it) vis-à-vis the frame of given, or mutually known, information in the musubi, presupposed by the speaker in the act of identifying the kakari. But as we confirm in Part 2, km was also deployed in contexts where what the speaker presupposed (indexed by pRT of the musubi) was, until communicated, unknown to the audience. In such instances, the entire km – the focused kakari and the presupposed musubi – would constitute news on the level of the ongoing discourse. This kind of km would have its own rhetorical effects, as well as a lasting effect on the km construction itself.
2 A musubi’s dual parameters: vis-à-vis its kakari & vis-à-vis its discourse Insofar as an act of identification entails a referent to identify, a kakari (whose business was most basically the making or seeking of identifications) entails a referent, regardless of whether or how it is expressed. In a prototypical km, a given (mutually familiar) musubi refers explicitly to this referent. This much comes with the construction itself. But km constructions and their components were not used in isolation; they were composed and deployed in an ongoing discourse. So the question arises: in communication underway, how does a musubi’s status as information presupposed by the speaker intersect with what her audience knows? In examples (1), (2), (3) and (7)/(10), the musubi, i.e. the content that the speaker presupposes in making or seeking an identification, is information given, i.e. also known to, or inferable by, the audience. This seems natural enough, but in many km tokens in oj and emj, what the speaker presupposes becomes known to her audience only as the musubi is processed. These musubi are informative. What communicative purpose(s) might a km with an informative musubi serve? How did it relate to, perhaps derive from, km in which the musubi was given? What consequences did an informative musubi have for km constructions, in how they dif-
326
Charles J. Quinn
fered from other sentence types, similar and not? These and related questions matter not only for our understanding of the km family’s history, but also for how we read and understand the texts of oj and emj. An informative musubi is most clearly at home in declarative km types, so the study below is restricted to declarative km tokens in oj and emj, and within that category, to those focused with kp zo. x zo was the most basically and persistently designative kakari, so it would be significant if its predicative status could in some contexts be cancelled or effaced. The points to be made would have been further strengthened by also showing how they also apply to km with namu and koso. But focusing on one – zo – allows room for the contextual analysis needed to reveal the differences between km tokens whose musubi is given and those whose musubi, while presupposed by the speaker, informs the audience.
2.1 km with a given (mutually known) musubi were basic Why would km with a given (mutually familiar) musubi be the basic type from which km with an informative musubi developed? First of all, a pRT nominal with a given referent was a form/meaning composite already in use when it was recruited to refer to one’s topic after a designative predication – the origin of the pRT musubi, by our theory. If the musubi began as a postposed topic, that topicality in itself suggests that it constituted given information. Furthermore, if a topic in the form of a pRT nominal was the original musubi, it would have brought its deictic referentiality to the job – a pointing to its referent as mutually familiar, to the extent that a following noun for the pRT to specify was deemed unnecessary. A bare pRT in oj was, in effect, a kind of ad-hoc pronoun, including when it was used to refer to topics. The deictic pRT nominal was used, for example, in this poem set at dawn after a layover at Kumage Bay: (12) [. . . uramwi ywori kadi no oto suru]NP pa ama bay from oar gen sound make.adn rf diver wotomye ka mo. maiden di if ‘[That which/what is making the rowing sound from the bay . . .]NP—is it diver girls?!’ (MYS 15.3641) In narration, the pRT nominal’s referent might first be introduced with a full np and then referred to with the referent-presupposing pRT nominal, as here: (13) . . . tonowibito-meku wonokoi [nama-katakunasikii]NP ideki-tari.> guard-look.adn man raw-rough.adn emerge-perf.concl ‘. . . a guard-like mani, [a somewhat rough and ready onei]NP, had emerged.’ (Genji, Hashihime, 5.129)
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
327
No linguistic form acquires deictic status with just the speaker assuming its referent’s “at-handness” or referability; it must work for audiences too, repeatedly and consistently, if the form is to become conventionalized as a deictic term. By oj, both -aku-derived (e.g. ipaku ‘what (he) says’) and pRT nominals clearly had this epistemically deictic status. By using either form, a speaker indicated her presupposition of a referent (entity, situation, etc.), and could count on each form being understood that way by her audience. It therefore seems likely that the pRT in km would have pointed to given information from the start. Its indexical value (sign and referent co-present to the communication) gave pRT the feel of discourse-anchored fact, much like no-nominals today. Their deictic referentiality would have been the reason that pRT and pRT-aku nominals (oj only) were recruited to clarify the assumed frame of reference, or topic (= musubi-to-be) after a designative utterance (= a kakari-to-be). If a musubi that referred to given information was basic to km, this “known” epistemic status would have been indexed even when a musubi’s contents came as news to the audience. Indeed, this very quality of givenness would very likely have made pRT hard to resist as a way of indicating, in a non-challengeable19 way, that the content referred to was more than just the speaker’s personal assertion. When a km with a pRT musubi was used in a discourse setting where its content was news to the audience, the deictic status of the musubi’s pRT nominal would have informed the audience of content with an already given place – its own, non-challengeable presence – in a discourse, something with the status of fact for that discourse’s participants, one of whom was the speaker. As observed above in comparing (10) with (11), being informed with a pRT musubi differed considerably from being informed with an unmarked, conclusively inflected predicate (“pSS”), which lacked a pRT-aku or pRT nominal’s indexical referentiality. With a declarative sentence inflected in its conclusive pSS form there was no indication that it referred to given information. Its asserted claim was, instead, multiply contingent: on this particular speaker introducing this particular information at this particular point in this discourse. By contrast, the deixis of the pRT-aku and pRT forms indicated that there was a referent and it was already known. This deictic referentiality would also be, we will argue, a contributing factor in a major reanalysis of km tokens with an informative pRT musubi.
2.2 km with an informative pRT musubi What would have motivated the use of km in contexts where its musubi constituted news? There is a hint to be had in English it-clefts of this kind, whose that-clause is presupposed by the speaker on the one hand, but news to the audience, on the other. These
19 See, for example, Givón (1984: 253–256) and Iwasaki (2000: 248).
328
Charles J. Quinn
were identified and labeled “informative presupposition” it-clefts by Prince (1978). She introduces this particular combination – the speaker’s knowing and presupposing, on the one hand, vs. the audience’s not knowing (thus being informed), on the other – with the following (1978: 898): 20 (14) It was just about 50 years ago that Henry Ford gave us the weekend. . . .20 That Ford’s 40-hour workweek “created” the American weekend was not and is not widely known. Since this sentence opens the article that it was culled from, in that discourse the relative clause’s underlined content has not been previously established. As Prince suggested, such sentences sound authoritative, as a direct consequence of this kind of it-cleft, which delivers its news not by introducing it in a matrix predicate (as in Henry Ford gave us the weekend just about 50 years ago.), but by referring to it as something assumed to be so, with the relative clause. That (14) is taken to refer to a fact is a consequence of the that-, who-, etc. clause’s content being non-challengeable. Referring to information in this way marks it as already known, if not to the present audience or addressee: we may distinguish a whole class of sentences . . . which I shall call informative-presupposition it-clefts. With these sentences, not only is the hearer not expected to be thinking about the information in the that-clause,21 but s/he is not expected even to know it. In fact, the whole point of these sentences is to inform the hearer of that very information. (Prince 1978: 898)
Of the informative-presupposition it-clefts in her data, Prince also observed this: Their function, or at least one of their functions, is to mark a piece of information as fact, known to some people although not yet known to the intended hearer. Thus they are frequent in historical narrative, or wherever the speaker wishes to indicate that s/he does not wish to take personal responsibility for the truth or originality of the statement being made. . . . it-clefts do [this] by strengthening the statement, by presenting it as an already known fact. (Prince 1978: 899)
The informative-presupposition variant of the English it-cleft clearly contrasts with the examples reviewed thus far of kakari-musubi, where the musubi’s contents are given, presupposed by the speaker and known to the audience. oj and emj analogues of Prince’s informative-presupposition it-clefts are readily found in km tokens in verse and in prose. We will call them “informative-musubi” kms. MYS 11.2571 appears in a large group of poems on love, and deploys an informative musubi following focus with so (oj variant of zo):
20 41a in Prince (1978: 898), quoting from the The Evening Bulletin (Philadelphia) 1/3/76, p. 31. 21 This reference to what the hearer is expected to be thinking is a point from Prince’s analysis of wh-clefts, earlier in the same article.
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
329
(15) masurawo pa tomo no sawaki ni nagusamoru men rf friends gen partying loc become.cheered.adn kokoro mo arame ware so kurusiki heart if be.conj.pi I id hurting.adn ‘While a man may have even a heart that’s cheered by a loud time with friends, it’s I (the one) who am suffering.’ (cleft-like construal) (MYS 11.2571) 22 ‘. . .it’s that I am suffering!’ (implicated situation-characterizing construal) With the km that concludes this poem (ware so kurusiki), the speaker contrasts her23 present suffering with a man’s ability to be cheered up by a loud good time with friends. The verb nagusamoru ‘become cheered/soothed’ entails a prior state of distress, thus implying that men do experience hurt. What the speaker posits here, however, is that a rousing good time with friends seems to make a man feel better. The poem presents contrast in multiple parameters: (a) men vs. (b) the speaker, (a1) easily rid of distress vs. (b1) distressed, as well as (a2) a generalization (men) vs. (b2) the speaker’s immediate and personal situation. The musubi’s reference to the speaker’s immediate experience of ‘suffering’ is informative, news proffered. Keep in mind also that ware ‘I’ has a semantic role (experiencer) in the kurusi predicate, whether separately focused with so or not. This poem lacks a headnote, and mentions no one else who is ‘hurting’, so the km [ware so]K [kurusiki]M does not work as a cleft ‘[It’s I]K [(the one) who is hurting]M.’ The contrast is not between the kakari’s ware and someone else, vis-a-vis a given condition of distress. The speaker’s point is not to identify herself as ‘the one who’s hurting’. Rather, ware so kurusiki is a way of saying that the situation the speaker finds herself in (in contrast to a man’s situation) is that I (unlike them) am hurting (while their hearts are easily soothed). What she is doing is emphatically identifying her situation as one where ‘I am hurting!’ This construal acknowledges the multiple contrasts (as the cleft interpretation does not) and makes this km more like the rentaikei shūshi bun ‘adnominal-final sentences’ (/pRT./) of oj and emj, which Sakakura (1993) and others interpret as ‘commentary that clarifies’ (kaisetsu).24 The NKBZ (3.220) and SNKBT (3.55) editions of the Man’yōshū appear to prefer such a construal: both translate this km as watasi wa kurusii no desu. ‘It’s that I am suffering!’ (my underlining) – using the rp’s descendant, an extended /p no (da)./ predicate (“ep”).
22 This arrow will indicate the pragmatically implicated interpretation. 23 The NKBZ editors indicate that the speaker is a woman. Even if it were a man, the poem would, mutatis mutandis, indicate that despite the typical man’s ability to forget hurt in revelry with friends, the speaker’s situation is different: ‘I am in pain!’ 24 Iwasaki (2000) studies these “adnominal final” sentences in emj discourse and characterizes their functionality as one of “suppressed assertion”. Wrona (2008: 231–236) reviews the same construction in oj as “attributive final” sentences. Quinn (1987: 111 ff) makes a case for interpreting it as a “referential predicate” that functioned much like today’s extended /finite p no (cop)./ predicate.
330
Charles J. Quinn
On the face of it, this reinterpretation involves a significant mismatch between this km’s lexico-syntax and the intended meaning. How do we get from ware so kurusiki, which looks quite cleft-like (‘It’s I who . . .’) – and in another context might mean just that – to watasi wa kurusii no desu ‘It’s that I am suffering!’? The situation-identifying kind of interpretation that we propose here is not limited to (15); it was regularly implicated with an informative-musubi km. Since this meaning was also what the rentaikei shūshi kaisetsu bun communicated, an introduction to that construction in its own right is in order.
2.3 Rentaikei-shūshi bun ‘adnominal-final sentence’: A situation-referring (“referential”) predicate A referential predicate (“rp”) corresponds in form to a rentaidome ‘rentai-stopped’ or rentaikei shūshi ‘adnominal-form conclusive’ sentence, a construction attested in oj and emj. The features identified in the following account of this construction are basic, and at work in any of its rhetorical applications.25 A rp was a designative predicate, declarative or interrogative, with this structure: (16) [[pRT]NP + designative predicator]RP The np complement in this designative predicate was a pRT-headed nominal, and its designative predicator was variable: declarative Ø, id zo, or cop n(i)ari, and interrogative di ka. Here are the construction’s declarative and interrogative forms, with an oj example: (17) [[pRT]NP Ø /zo /n(i)ari]RP ‘[It’s [(the case) that p]NP]RP.’ [[. . . pagwi no sureru]NP so.]RP bush.clover gen rub.stat.adn id ‘[It’s [that the bush clover rubbed it]NP]RP.’
(MYS 10.2101)
ka.]RP ‘[Is it [(the case) that p]NP?]RP’ (18) [[pRT]NP [[Sidukeku mo kwisi ni pa nami pa yose-kyeru]NP quiet.inf if shore loc rf waves rf surge.in.inf-mpst.adn ka.]RP di ‘[Is it [(the case) that the waves are surging ashore so quietly, at the beach]NP?]RP’ (MYS 7.1237)
25 Such as the oft-noted exclamation (kantai) or evocation of yojō ‘surplus of affect’.
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
331
The rp construction, like the kakari in a km, made or sought an identification, and for that reason (like a cleft km’s kakari) it presupposed a referent. For an rp, that presupposed referent was (1) a situation or matter at hand in the discourse that was (2) deemed in need of proper characterization or identification. In (17), the presupposed situation is clearly implied: a stain on the speaker’s robes; in (18) it is something the speaker hears when concentrating on what he hears outside, from his position inside a house. There is a tradition in kokugogaku of reading the zero-copula rp variant and some RT p -final km as ‘exclamatory’ (Iwasaki (2000: 259), citing Koike (1967) and Onoe (1982)). But as Iwasaki’s (and many other) examples show, neither a zero-copula /pRT./ referential predicate nor an /x+kp pRT./ km was necessarily exclamatory. Insofar as a rp constitutes a statement or question about how a given situation is properly construed, it expresses the speaker’s concern about the same, so it is not surprising that a rp might index an affective stance, such as the half-doubting surprise of recognition in (18). As an rp’s least elaborated form, the zero-copula /pRT./ was best suited for iconically indicating the surprised, reflexive response of an exclamation. In fact, the zero-copula rp was exclaimed with enough frequency that Yamada (1908) categorized it as a kantaiku ‘exclamatory nominal clause’. Note, however, that the frequency of this effect with either the zero-copula rp or other rp variants does not make any of them inherently exclamatory. The interpretation of rps proposed here26 is congruent with observations on rentaikei shūshi like Sakakura’s (1993: 227 ff) and certain accounts of the ModJ extended “no da” predicate (“ep”), /[[finite p] no] (cop)./ ‘It is [(the case) that p].’27 Sakakura (1993: 227), citing Koike’s study, refers to non-km pRT final sentences as rentaikei shūshi no kaisetu bun ‘adnominally concluded commentary sentences’ (my emphasis), and provides a number of emj examples (Sakakura 1993: 228–230) where the construction ‘explains’, i.e. clarifies, some known matter. He also remarks (Sakakura 1993: 233) that these ‘commentary sentences’ (= rps) were a close functional analogue of today’s extended predicate.28 In our terms, the similarity is that, like the rp of oj and emj, today’s ep (1) is a deictic nominal29 (2) designatively predicated in order to characterize and
26 Updating the argument in Quinn (1987: 111 ff) that, like today’s extended predicate, a rp presupposes a referent situation. 27 Particularly Hanaoka McGloin (1980), Jorden and Noda (1987), Noda (1990), and Maynard (1992). In that a rp presupposes the situation that it characterizes/identifies, it also resembles the English It’s that .. construction as analyzed in Declerck (1992) and Delahunty (1995). Shinzato and Serafim (2013: 223–225) analyze ModJ’s extended no da predicate in just this way, as a specification of a presupposed variable. Their reading of MYS 15.3753 (p. 225) indicates that their analysis would also cover a pRT rp. 28 Sakakura (1993) goes on to propose that km was born when kps were inserted into /pRT./ kaisetsu bun ‘clarifying comment sentences’, each kp making explicit a particular meaning that /pRT./ could already express alone. Shinzato and Serafim (2013: 235 ff) review his proposal and scholarly response to it. 29 Again, nominals of the form pRT and ModJ /finite p no/ are deictic in that their referents are assumed to be present to the discourse context (cf. Levinson 1983: 54). Deictic markers in Japanese predicate morphology make spatial, temporal, and social (keigo) distinctions, but also epistemic ones (see e.g. Akatsuka-
332
Charles J. Quinn
clarify the situation that its speaker presupposes. Now as then, this is done in discourse contexts where the speaker deems such clarification desirable. The line cited next immediately precedes the line quoted above in (17). The rp’s situation at issue (“sai”) is the same colored stain ((17)’s translation in parentheses): (19) Wa.ga koromo [[sureruRT]NP ni pa arazu]RP I.gen robes rub.stat.adn cop.inf rp be.neg.concl ‘[It’s not [that my robes are rub-dyed]NP]RP. ([It’s [that the bush clover rubbed them, when I was gone to the fields of Takamatsu]NP]RP.’) (MYS 10.2101) In this poem, then, the speaker deploys two rps, first asserting that it is not the case that his robes are ‘rub-dyed’, and then clarifying what is the case by using quasi-copular so to positively identify the sai (repeating (17)): [[. . . pagi no sureruRT]NP so]RP ‘[It’s [that the bush clover has rubbed them . . .]NP]RP’. As (19) shows, a rp could be predicated with the inflecting copula n(i)ari (and that copula negated), or, as (17) shows, it could simply be asserted with so (zo). A rp could also be framed as a question with ka (18), or predicated as a stand-alone nominal clause. The latter two occur together below. A red-faced young Murasaki has just come out and stands before her grandmother, obviously upset – the situation at issue. The grandmother reacts immediately with (20) Nanigoto zo ya. ‘Whatever’s the matter?’ and then refines this question with her best-guess rp, . . . paradati-tamaweruRT ka. ‘Is it that you’ve been quarreling . . .?’ Murasaki answers in (21) by identifying the situation with a zero-copula rp. (20) Nanigoto zo ya. [[Warawabe to paradati-tamaweru]NP ka]RP. what matter id cs children with quarrel.inf-hon.stat.adn di ‘Whatever is the matter? [Is it [that you’ve been quarreling with the children?]NP]RP.’ (21)
Ø ]RP. [[Suzume no ko o Inuki ga nigasi-turu]NP sparrow gen child acc Inuki gen let escape.inf-perf.adn (zero cop) ‘[(It’s) [that Inuki let the baby sparrow go]NP!]RP (Genji, Wakamurasaki, 1.280)
The grandmother seeks to know if her proposed identification of what is ‘the matter’ – the nominal warawabe to paradati-tamaweruRT ‘that you’ve been quarreling with the children’ – is the case. Murasaki answers by identifying the situation with a rp, suzume McCawley 1978, Akatsuka 1985). Epistemic deixis indexes a referent’s “knownness”, typically based on evidential distinctions in how our knowledge is grounded (e.g. direct experience vs. hearsay, etc.). Modal auxiliaries (from mu to keri) afford one such marking, but so do nominals formed with oj -aku, oj and emj pRT, and ModJ /finite p + no/. All are instances of epistemic deixis, in that they routinely refer to mutually known referents (accessible through present perception or in memory).
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
333
no ko . . . nigasi-turu.RT ‘(It’s) that Inuki let the baby sparrow get away!’ – her characterization of ‘what is the matter’. These rp examples all include (a) a given situation, the sai that they characterize as “p” and (b) refer to with a deictic pRT nominal, which is (c) identified by means of a designative predicator: copula nari, quasi-copular zo or ka, or zero copula. The rentaikei shūshi ‘adnominal conclusive’ /pRT./ was the most direct of a rp’s several predicative options, each of which had the deictic pRT np for its complement. Like its more explicit alternatives, zero-copula /pRT./ might be deployed in exclamation or not. Table 2 summarizes the rp: its topic/referent/sai; its forms; its basic function. Table 2: The Referential Predicate (rp) and its Situation-at-Issue (sai). Situation at issue (sai) / “topic”
Referential predicate [[pRT]NP + ø/nari/zo/ka]
A presupposed, often tacit matter whose nature /significance/identity warrants clarification.
[[pRT that characterizes sai]NP + designative p] ‘It is/Is it (the case) [that p]NP.’ (stated/asked)
In using a rp, a speaker (1) presupposes a situation at issue, typically the situation at hand; (2) characterizes and refers to that sai with a deictic pRT nominal; and (3) using a designative predication, identifies (or seeks to confirm) that the sai is “pRT”. This is the kind of situation-identifying meaning that, we suggest, an informative-musubi km prompted, by invited inference, or implicature. For a designatively predicated pRT nominal in oj or emj, the label referential predicate reminds us – as a structural name does not – of the sai that it refers to – a feature of its meaning that has often gone unappreciated, perhaps because it is manifested outside the rp sentence itself. A rp as just described actually resembles a cleft-type km, in two significant ways: (1) its characterization is focalized by means of designative predication, while (2) it presupposes a given referent that it identifies or seeks to identify. Thus, the focused kakari of a pRT-concordant km—/x + des kp/—is the functional analogue of a designatively focused rp, /pRT + des p/. And a km’s presupposed musubi is the functional analogue of the rp’s presupposed sai. The two constructions’ congruence is laid out in Table 3: Table 3: Kakari-musubi and Referential Predicates compared.
KM RP
Focused
Presupposed
kakari complement + des (kps zo/ka/ya/namu) pRT nominal + des (zo, ka, cop, ∅)
pRT musubi situation at issue
Thus, the rp construction can be characterized as another cleft-like focus construction, one whose focused part is concerned with (making or seeking) identification and whose presupposed part is typically given, self-evident enough to be tacitly understood.
334
Charles J. Quinn
These features make the rp a good example of what Heine and Reh (1984), in their ground-breaking study of focus constructions, called “predicate focus”, which they distinguish from (the cleft/km-like) “term focus” (148). In term focus, a proposition with a term at issue or missing is “completed” with the cleft’s identification of that term. Predicate focus, on the other hand, highlights and emphasizes the verb or predicator phrase, or might include the entire clause. Shinzato and Serafim (2013: 223–226) make a similar distinction in focus constructions, when they use Mikami’s (1953) constructs shitei ‘identification’ and kaisetsu ‘explanation/interpretation’ to distinguish two kinds of ModJ no da sentences, of “narrow” and “wide scope”, respectively. They then illustrate (226) each type with an oj example – a zo km (their (55)) and a zo-designated nominal predicate ((56), MYS 15.3753). Although the nominal predicated with so in MYS 15.3753 is not itself a pRT (. . . koromo so. ‘it’s that it’s a garment that . . .’), the long and complex pRT that specifies koromo ‘garment’ effectively explains her sending it to the addressee (the sai). At all events, the early Japanese analogue of Heine and Reh’s term focus is thus the cleft-like type of km construction, and their predicate focus has an analogue in the oj/ emj rp and today’s ep. Heine and Reh state that there is no presupposition involved in predicate focus (in Nupe, 166), but do not appear to consider a presupposed situation at hand as a possibility.
2.4 Informative-musubi in zo km tokens; the non-cleft, rp-like construal Among the short lyric poems of the oj Man’yōshū, there are many zo-focused km tokens with an informative musubi (e.g. (15) above, MYS 10.2019, 10.2093, and 11.2356, among others). These poems typically lack a contextualizing narrative, so the main way for a musubi’s referent to attain given status would be by having been introduced earlier in the same poem (as with the pa-marked topic in example (1)) or in preceding poems composed on the same occasion, but it is also clear that a zo km did not require such conditions. Man’yōshū poems with an informative-musubi km are often soliloquies or lack a specified addressee. As representatives of the oj lyric genre, these poems express affective response, typically in terms of something just discovered, recognized, or understood. Sometimes this takes the form of a zo km with an informative musubi, which invites (“ ”) an ‘It’s that p.’ reading, so that the km serves to actually identify the situation at hand for what s/he has recognized it to be. Such km tokens express what the speaker has realized is the case, “what’s going on here”. Their musubi are informative, even to their speakers. The km in MYS 11.2579 declares a change in the speaker’s own emotional state, with id so marking the adverbial ima ‘now’:
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
335
(22) Paya yukite itu si ka kimi wo api-mimu to quick go.ger when el di you acc meet-see.conjec.concl compl omopi-si kokoro ima so nagwi-nuru. think.inf-pst.adn heart now id relax.inf-perf.adn ‘Rushing to you, (with) a heart that wondered “When is it I’ll see you, and you me?”—it’s now that it has calmed.’ (cleft-like construal) (MYS 11.2579) ‘—it’s that now it has calmed!’ (implicated rp-like construal) The speaker announces that her anxious heart, so eager to meet, has just calmed, so her intensely desired meeting is imminent or underway. The epistemic side of past/perfect auxiliary ki (si is its adnominal) is an ‘established factuality’ (Jdb 236), so . . . omopi-si kokoro ‘a heart that felt/has felt . . .’ refers to that anxious eagerness as a condition that continued up to this moment’s (ima) abrupt release (nagwi ‘calmed, relaxed’ + perfective auxiliary nu). The final perfective nu (as nuruRT) indexes proximate evidence30 for the change that it expresses, in this case the speaker’s felt transition to the condition nagwi. This change was not previously alluded to, and the speaker has not felt calmed until now, so it is not given information. Structurally, her emphasis and news appear to be the kakari’s direct and blunt ima so ‘it’s now’, but the musubi that she presupposes in making that identification is also news – to the addressee, to us readers, and apparently to herself. If both its kakari and musubi are informative, this whole km sentence is informative, which invites the inference that the speaker’s point is not just when (‘now’) her anxious desire melted into relief, but that it has done so, now (that they are to meet, or have just done so). This /x so pRT/ km may look like a cleft km, but with its informative musubi makes better sense as an rp-like ‘[It’s [that now, p]].’, in which ima (the prima facie complement of kp so) and the musubi phrase together identify the presupposed situation at hand. An rp’s ‘It’s that ..’ assertion is a designative, or identifying, kind of predicate, which, as observed of the kakari in km, entails a referent, a scope, for the identification. The referent of this implicated rp (or “km-as-rp”), then, is the situation the speaker has just experienced. It is the nature of this moment, and not just the time of her calming, that she identifies with this ostensible km. How might an implicature or inference turn such a km into a kind of kp-enhanced rp? We can represent the broader, rp-like construal of an entire declarative km /[x+kp]K [pRT]M/ as ‘[It’s [that x(k), p(m)]NP]RP.’, with x and p surrendering their roles kakari (k) and musubi (m) to an implicated, monoclausal rp, in which they together characterize the situation at issue. As the complement nominal in a designative rp, [that x, p] consti30 On an evidential scale, sentence-final perfective /vRY-nu/ is associated with proximate evidence for the change expressed with the v, and in this regard contrasted markedly with perfect/past /vRY-ki/ and externally established fact /vRY-kyeri/ (labeled mpst in our glosses, after Frellesvig 2010). Quinn (2018) describes emj auxiliary options in relation to such an evidential cline.
336
Charles J. Quinn
tutes that rp’s focused information, a single characterization of the sai. In this way, an informative-musubi km implied a designative ‘[It is [(the case) that x, p]]’. When a musubi is informative, its contents are readily inferred to be part of what the speaker is declaring to be so, which also includes the kakari’s (kp’s) complement, x, as a facet of the situation referred to with pRT. The formerly open proposition p, whose missing term the kakari has filled with x, is thereby closed, and the proposition complete. Thus, what we abbreviate as “p” in the implicated rp is a situation – a proposition – that includes x, in (22)’s case, ima ‘now’. As the implicated rp ‘[It is [(the case) that x, p]].’ preserves the full proposition (x, p) and its pRT-marked deixis, it resets what is focused (from ima alone to the situation-characterizing ima nagwi-nuru) and what is referred to and presupposed (from nagwi-nuru to the situation at hand, or sai). Finally, that this km/implicated rp is delivered at this particular moment implies its relevance for the situation at hand. These factors add up to a situation-defining referential predicate, not a cleft-type km. With km, an implicated rp seems to have been an interpretation waiting to happen. Even if (22) were a cleft-type km with a given musubi, /ima so nagwi-nuruRT./ ‘[It’s [now]] RT would express a known, referable fact that is K [that it has calmed]M.’, nagwi-nuru true as of ima ‘now’. If that cleft meaning is true, it follows that a rp ima nagwi-nuru ‘[It’s [(the case) that now it’s calmed]].’ – which also treats ima nagwi-nuru (the same proposition) as fact – will be true too. That is, a generalized conversational implicature (“gci”; Grice 1991[1975]: 37–40; Levinson 2000, 2010) appears to hold between a km and its corresponding rp, even if they were usually deployed to different ends. Levinson notes that “gcis are default inferences, which can nevertheless be cancelled by context and content where they do not fit” (2010: 203). The gci that might link a km to its corresponding rp construal appears to be blocked when the musubi is given, i.e. not informative. But when its pRT musubi is informative, a km does just what its corresponding rp did: it identifies the presupposed situation at hand as pRT (including the erstwhile kakari’s complement, in the semantic role it has in p). Put the other way around, the gci is cancelled when the km’s musubi is given because in just that respect the km differs from its corresponding rp, where the pRT is informative. In sum, when a km’s musubi was informative, the gci underwrote that km’s interpretation as a situation-defining rp. To recap the argument, the corresponding rp’s situation-identifying meaning is implicit, as a gci, in the km construction, but it remains in the background, canceled, when the musubi is given and the kp (zo, koso, namu) is actually designative. But when a musubi is informative and the entire km – presupposed musubi included – is news relevant to the present discourse moment, the gci supports re-construal of the entire km as a “kp-enhanced” rp. With the informative musubi, the gci invites an ‘[It’s [that x, p]].’ interpretation of the km. An identification, rps included, entails a referent, something to be identified. Unless a referent situation is otherwise specified, the most obvious candidate for identification with a designative /pRT./ is the immanent situation at hand. As it happened, the bare pRT that concluded a km was identical with the situation-identifying rp construction’s most
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
337
basic, zero-copula ending. All in all, it is not surprising that an informative-musubi km might end up construed as a referential predicate.31 A further effect of an informative musubi was that it made the km-as-rp’s final predicator phrase its sentence’s matrix predicate, which was a factor in adnominally inflected matrix predicates losing their markedness and, in time, being used as the default conclusive. Finally, if the informative musubi in MYS 11.2579 prompts its km’s reconstrual as an rp in which ima ‘now’ is adverbial, the prima facie kakari ima so does not, in that construal, express a cleft’s designative ‘it’s now’. But so remains positioned as if designating its complement ima. What does id so contribute, if anything, when the km is reinterpreted as a rp? We might say it answers a higher (node’s) calling. Another facet of the reconstrual of an informative-musubi km as a rp seems to be that the emphatic stance indexed with so, along with its designative significance, was not simply cancelled. There are at least two factors that invite a reconstrual of kp so as the designative predicator of the implicated rp. First, whether a km is a cleft or a rp-implicating km (as in (22)), the bluntly assertive, informing effect of so is not restricted to its immediate complement. As a facet of the speaker’s interpersonal stance, that effect lingers for the full utterance – including the informative musubi and, thus, the implicated rp. What of so’s designative predicativity? In any rp (actual or implicated) there is one predication, which is designative and takes the pRT ‘(the case) that p’ nominal as its complement. In a km-as-rp, the erstwhile km’s pRT musubi, with the erstwhile kakari complement, x, together constitute the complement of a designative predication. There is thus a ready-made role for a designative so in the implicated rp: that of its identifying predicator. There is a place for both sides of zo – the interpersonal stance it indexed and its designative predication – in the rp implicated by an informative-musubi km like (22). In sum, while zo remains structurally attached to its immediate complement (ima so in (22)) in a rp-implicating km, its positioning leaves it highlighting, or emphasizing, that complement, in a non-designative, non-predicative way. The pragmatics invite a “raising” of zo to predicator of the implicated rp, and efface zo’s designative relation to its immediate structural complement. This would be a factor in zo’s desemanticization, away from predicator of what it structurally governed, to highlighter. In (22), so highlights ima, but it is (virtual) predicator of the implicated rp, ima nagwi-nuruRT ‘(It’s) that now it has eased!’ – much as if the speaker had said ima nagwi-nuruRT zo. ‘id.’32
31 As Sakakura (1993) pointed out, more than a few km tokens read like rps with a kakari particle inserted. The implicature-assisted km-as-rp reanalysis accounts for this while keeping a cleft construction in km’s origin story. 32 A perfectly grammatical rp, with id zo as its explicit, in situ predicator. As an historical development that derived a rp from a cleft-like construction (km), the implicated km-as-rp perhaps makes it easier to see how ModJ extended predicates might become de facto “in-situ constructions of it-clefts”, an interpretation of a use of today’s “noda construction” proposed in Schaffar (2002: 324). The km-as-rp, of course, kept km’s explicit marking of the kakari in place.
338
Charles J. Quinn
In sum, the informative-musubi km in MYS 11.2579 and others like it have the effect of implicating a situation-characterizing referential predicate. This implicated rp can be understood to be delivered with an emphasis from id so that does not designatively predicate ima ‘now’ alone, as a cleft km would, but expands/raises to include ima nagwi-nuruRT ‘(the case) that now it has eased’, and together they characterize and identify the situation at hand.
2.5 emj zo km tokens with informative musubi and implicated rp In emj prose fiction, particularly when relating utterances or thoughts attributed to story characters, the narration sometimes employs a zo-focused km with an informative musubi. Examples occur early in the Genji, and one of them helps define the Emperor’s circumstances after the death of his beloved (and Genji’s mother) Kiritsubo. The narrator has just informed us that those who observed the Emperor’s continuing grief found themselves in tears too, when she then refers, in the same connection, to this contrasting state-of-affairs: (23) [[Naki ato made pito no mune.aku not.exist.adn after until person gen feelings.soothe.concl mazikari-keru pito no mi.oboe ka.na. to] zo]K negconj.inf-mpst.adn person gen hon.sense di.excl comp id [Kokiden nado ni wa nao yurusi.nau notamawi-keru]M. Kokiden etc. loc rf still unforgiving.inf say.hon.inf-mpst.adn ‘It was with “Such regard for the departed – that we should be left disconsolate even after (she) is gone!” that Kokiden’s feelings, unforgiving still, were expressed.’ (cleft-like construal) (Genji, Kiritsubo, p. 102) ‘It (the situation) was such that, Kokiden’s feelings, unforgiving still, were expressed as “. . .”, .’ (implicated, rp-like construal) This is one heavily loaded km, with substantial news for the reader in both the kakari and the musubi. The quoted comment highlighted with zo is the kakari, which cites the Kokiden Lady’s reaction to the Emperor’s continuing attachment to the deceased Kiritsubo. The musubi also has communicative dynamism:33 it explains how to understand the quote highlighted with zo in the kakari, and the honorific notamawi ‘saying’ confirms (following the locative Kokiden nado ni wa ‘over at the Kokiden palace’) the source 33 Its “information quotient” vis-à-vis the discourse under way, i.e. the degree to which it adds information. On communicative dynamism, Prince cites Firbas (1964: 270), and uses the concept in characterizing informative-presupposition it-clefts.
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
339
of that quote. The musubi is thus informative; it introduces its contents to the discourse by referring to them as facts-in-place. In the NKBZ edition (1, p. 102) of the Genji, the modern translation of the musubi . . . notamawi.keru ‘that (she) said’ suggests an rp-like construal, in rendering it with an extended predicate: . . . ima nao o-yoosya suru koto naku ossyaru no de atta. ‘It (the situation) was that, still unforgiving, (she) said “. . .”.’ The musubi’s content, roughly ‘that Kokiden, for her part, said, still unforgiving’, is not surprising, given what readers know about her by this point, but even as the narrator refers to this as established fact – marked as such with both modal auxiliary keri and its adnominal inflection – the specifics of that content constitute new information for the reader. Strengthened epistemically with keri’s factuality, the predicate’s content is marked as a claim that transcends the narrator’s personal testament, which gives it an authority that a simple conclusive . . . yurusi.nau notamau ‘. . . (she) said, unforgivingly’ would lack. Prince (1978) observes something quite similar of it-clefts in English, as a rhetorical factor motivating the use (and presumably, the devising) of informative-presupposition it-clefts: The fact that it-clefts may present information as known without making any claims that the hearer is thinking about it (or, in fact, even knows it) presents the speaker with a strong rhetorical temptation: what is to prevent him/her from putting new information into the that-clause? Note that such an action would be totally in line with the general convention of putting new information last. . . . we may distinguish a whole class of sentences . . . which I shall call informative-presupposition it-clefts. With these sentences, not only is the hearer not expected to be thinking about the information in the that-clause, but s/he is not expected even to know it. In fact, the whole point of these sentences is to inform the hearer of that very information. (Prince 1978: 898)
These comments suggest what might motivate a deliberately informative musubi and turn its entire km sentence into an emphatic identification of what “the situation here” is. Something like the rhetorical attraction that Prince noted of informative-presupposition it-clefts was likely also behind informative-musubi km constructions. Their rhetorical “value added” is, on one parameter, a reducing of the speaker’s “personal responsibility” (Prince 1978: 900) – an effect they share with hedges – while on another parameter they introduce an effect we have observed in km-as-rp: Hedges do this by weakening the statement, by making it into a ‘guess’ or conjecture; it-clefts do it by strengthening the statement, by presenting it as an already known fact. (Prince 1978: 900)
Like the oj example in (22), the informative musubi in (23) is rhetorically effective in just this way. The modal past keri that caps the musubi (keruRT) provides a kind of epistemic harmony, backing the adnominal’s deixis with its modality of received fact. For this reason, keri was also frequently present in informative-musubi of km focused with kp namu, as they introduced readers to facts in place at the beginning of a story or episode. Prince also comments that an informative-presupposition that-clause tends to be longer and semantically more complex than a short one. This too seems to hold for inform-
340
Charles J. Quinn
ative pRT musubi (longer) vs. given musubi (shorter).34 One-word informative-musubi tokens do occur, but the longer a musubi is, the likelier it is to be informative. Prince suggested that this reflected a basic pragmatic principle (1978: 886–887): That is, it seems natural to use as few words as possible to refer to something known, and to present new information in greater detail; this is in line with one of Grice’s Maxims of Manner: Be brief (avoid unnecessary prolixity).
Conversely, why increase the detail of a clause’s representation of a situation if not to inform? The next informative-musubi example is from Tosa nikki ‘Tosa Diary’ (‘21st day’), and comments on a personally witnessed event. The narrator’s party is making their way back to the capital in a small flotilla when, at one put-in, a local boy offers to make himself useful, and joins them. When the boats depart he looks back and sings some lines about home and the parents he will miss, which the narrator quotes and assesses thus: (24) [. . . to utau zo]K [apare naru]M. comp chant.adn id touching cop.adn ‘[It was his singing “[. . .]”]K [that was touching]M.’ (cleft-like construal) ‘It’s [that his singing “. . .” was (so) touching]RP!’ (implicated, rp-like construal) (Tosa, p. 49) The prima facie km’s kakari highlights the song lines quoted (with complementizer to), vis-à-vis the presupposed musubi, apare naruRT ‘that (/what) was moving’. The preceding text neither mentions nor implies that anything is apare ‘moving’, so this musubi is informative, not given. Thus, the narrator’s communicative point is not to identify the boy’s singing as the one thing among others ‘that was moving’. Rather, he is exclaiming at the moving impact of the boy’s act – and thereby defining the moment as something that had aware ‘pathos’, while highlighting what made it so. The examples adduced above from oj and emj confirm that informative-musubi variants of zo-focused km were in use. They were also common – in emj, perhaps as common as informative-presupposition it-clefts are in today’s English. They appear to have been highly susceptible to a gci-enabled reconstrual as an enhanced kind of referential predicate, i.e. an emphatic characterization of a situation presently at issue. They are also a likely reason for the tradition of reading the pRT that ends a km token as ‘exclamatory’.
34 An extreme case of Prince’s point in emj is the elided musubi whose gist is so much a given that it is left unsaid—the numerous instances where the narrator or a character concludes with a truncated, kakari-only phrase, e.g. . . . to te namu. ‘It’d be “. . .” you know.’ (implicated and unsaid: ‘that I thought)’ (Genji, 2, p. 28); or . . . netau koso. ‘. . . it’s vexing indeed.’ (Murasaki shikibu nikki, NKBZ, p. 236)—a figure where the ellipsis leaves the kakari looking (once again) like a matrix predicate.
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
341
Well beyond the few tokens surveyed above, declarative km with an informative musubi offered a way to assert an rp-like characterization of the discourse’s situation at hand, with a zo (or koso) in the kakari. This kind of message was often soft-pedaled too, with the tentative identification of kp namu. In emj, unassertive, audience-assuring kms with an informative musubi are frequent when namu is the kp, even as namu continued to be used in other tokens to tentatively designate its complement vis-à-vis a given musubi. If the km-as-rp readings proposed above seem an outlandish mismatch of syntax and semantics, remember that the full-proposition scope that we allege for kp zo is an interpretation that has long been an acknowledged option for namu and ya km (on ya, see Jdb 755). What the foregoing account proposes is a way of understanding how this rp-like interpretation might relate to a more basic, cleft-like construction, where syntax and semantics are more congruent. It is worth noting that a similar, pragmatically induced expansion of the scope of particles – and its conventionalization – is at work in our default understanding of expressions like Tabako wa suwanai. ‘(He) doesn’t smoke.’ or Sake mo nomanai si . . . ‘(He) doesn’t drink, either, and . . . (has other, similar virtues).’ In the everyday, default construals of these statements, wa and mo do not apply just to their immediate complements (tabako ‘tobacco/cigarettes’ and sake ‘alcoholic beverages’ respectively), but rather to the habits that those words are part of, i.e. tabako suwanai ‘doesn’t smoke’ and sake nomanai ‘doesn’t drink’. The pragmatics behind these may differ from what underwrote the km-as-rp phenomenon, but the result is strikingly similar. Pragmatically speaking – and as Prince noted of informative-presupposition it-clefts – an informative musubi seems to have been an inevitable variant of km. So too, it seems, the reanalysis of such km tokens as situation-clarifying rps. The notion, if not the label, of km-as-rp enjoys long-standing acknowledgment in the philological literature, in the form of translation preferences such as /p no zya/ (or da)./ and /p koto (yo)./, which place the entire km in focus. Kushima (1989), Sakakura (1993) and other scholars have argued that km was a type of rp, as Shinzato and Serafim (2013: 178 ff) note. The km-as-rp hybrid differed from a “straight” rp in that its kp remained in place, highlighting if not predicating its complement, while, we have suggested, its designative force applied to the implicated rp. With the kp remaining in place, an informative-musubi km token like (22)’s ima so nagwi-nuru offered more than a straight rp (✶Ima nagwi-nuru so. ‘It’s that now it’s eased.’), by retaining the kp’s non-predicative highlighting of a particular constituent. While we have examined only declarative km-as-rp tokens with kp zo, they occur, prompted by informative musubi, with koso and namu as well. The evidence shows that km-as-rp was in place by oj, and it was presumably of long standing even then. Its functional shift would have been a factor in the kps involved being desemanticized – bleached of their designative predicativity – at first with regard to their immediate complement, but over time sentence-finally as well. This would have been a factor in each of the declarative kps’ emphasis eventually becoming non-propositional and subjective-to-intersubjective in nature.
342
Charles J. Quinn
2.6 In conclusion: km in cross-linguistic context Heine and Reh (1984) includes a comparative study of how cleft constructions appear to have evolved into highlighting constructions in a selection of African languages. They induced four stages (181–182): Stage I: A cleft structure: /{np or pp} copula – subordinate clause/ Stage II: The copula is desemanticized to a focus marker. This structure is exploited to optionally emphasize wh-words. This stage is characteristic of most weakly grammaticalized systems. Stage III: The focus construction undergoes functional shift, i.e. it is no longer possible on synchronic grounds to derive it from the cleft construction, its source. . . . wh-words are obligatorily marked for focus. Furthermore, focus marking spreads to the predicate: an entirely different type of construction is introduced to mark verb phrase focus, term and verbal focus thus entering into a relationship of complementary distribution. Stage IV: Focus marking becomes an obligatory feature of sentence structure, i.e. any declarative sentence has to be marked for either term [argument, adjunct] or predicate [including full clause] focus. . . . It is not at difficult to discern the developmental history of km in Heine and Reh’s stages. Harris and Campbell (1995: 152) report that their own cross-linguistic study (which includes km35) “finds no counterexamples” to Heine and Reh’s conclusions. Once km variants with an informative musubi were deployed – almost immediately? – the construction’s journey to Stage II would have been underway. The texts of oj and emj show a Stage I, cleft type of km still in use, in km tokens with a designatively focal and informative kakari and presupposed and given musubi, particularly with kps ka and zo, less so with namu, and least with ya (which in kp position was already in oj primarily a non-designative focus marker (Jdb 755)). At the same time, declarative km tokens, such as the zo-focused, informative-musubi tokens above, invited reinterpretation as situation-identifying referential predicates. These km-as-rp tokens turned the two-part km into a unitary and informative characterization of a presupposed situation at issue. The rp /pRT des./ and its presupposed sai referent constituted a focus construction that was congruent in its parts with the cleft-type km. The km-as-rp thus provided an additional way for “focus marking” to “spread to the predicate”, a characteristic of Heine and Reh’s Stage III. If the two-part structural and functional congruence of km focus constructions with the basic rp (and today’s extended . . . no da predicate) seems little remarked in the literature, perhaps that is because the presupposed referent of a rp or ep – their sai – is immanent, seldom deemed in need of overt, separate expression.
35 To derive the cleft km, they cite (Harris and Campbell 1995: 162) Ohno’s inversion theory (1993).
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
343
Km in oj and emj on balance appears to be, in Heine and Reh’s developmental history, a “weakly grammaticalized” family of constructions. This is primarily because the cleft pRT-concordant kms of oj and emj could still be derived synchronically, with straightforward reference to how their components were used contemporaneously outside the km construction (see Table 1). In oj and emj, the km-as-rp presumably still felt vaguely like its cleft-km siblings (its kakari still highlighted, if not designatively predicated) on the one hand, and like the rp cousin that it emulated, on the other. As a consequence of the frequency of informative-musubi tokens, which were regularly construed as emphatic rps, a kp’s designative predication of its immediate complement was frequently canceled, as the kp was taken to apply instead, virtually, to the implicated rp. As noted earlier, this would have helped push those kps toward stronger association with non-designative focus, at first vis-à-vis their immediate complement but eventually with regard to their implicated (rp) one too, so that they eventually desemanticized into a set of variously emphatic particles. Heine and Reh’s “strongly grammaticalized” label applies to the /x koso pIZ/ km, which was no longer derivable synchronically from its components, as Frellesvig (2010: 256) has noted. With the opaquely grammaticalized koso variant, the km focus family had a foot in Stage III. While most pRT-concordant members of the km family still seem weakly grammaticalized in emj, the family was well on its way from Stage II to Stage III, and the informative musubi was a significant factor in this trend.
Abbreviations cf cs des di ep gci id if iz km kp p pi rf rp rt sai ss ui
confirming focus: kakari particle namo/namu confirmation-soliciting kp ya designative = copular/specifying predicator doubted identity kp ka ModJ extended predicate construction (. . . no (da).) generalized conversational implicature identifying focus: kp so/zo inclusive focus particle mo izenkei ‘realized form’ = pi kakari-musubi kakari “focus” particle predicator (v, a, cop), predicator phrase presupposing infinitive = izenkei ‘realized form’ restricted focus particle pa referential predicate rentaikei ‘adnominal form’ situation at issue (= a discourse exigence that motivates rp use) shūshikei ‘conclusive form’ unique identification: kp koso
344
Charles J. Quinn
Symbols / . . . /
Enclosed material considered as a unit or construction. Interpretation to the right is pragmatically invited.
References Akatsuka, Noriko. 1985. Conditionals and the epistemic scale. Language 61. 625–639. Akatsuka-McCawley, Noriko. 1978. Another look at no, koto and to: Epistemology and complementizer choice in Japanese. In John Hinds and Irwin Howard (eds.), Problems in Japanese syntax and semantics, 178–212. Tokyo: Kaitakusha. Bolinger, Dwight. 1972. A look at equations and cleft sentences. In Evelyn Scherabon Firchow (ed.), Studies for Einar Haugen: Presented by friends and colleagues, 96–114. The Hague: Mouton. Declerck, Renaat. 1992. The inferential it-is-that construction and its congeners. Lingua 87. 203–230. Delahunty, Gerald P. 1995. The inferential construction. Pragmatics 5(3). 341–364. Firbas, Jan. 1964. On defining the theme in functional sentence analysis. Travaux Linguistiques de Prague 1. 267–280. Frawley, William. 1992. Linguistic semantics. Hillsdale, NJ: Lawrence Erlbaum Associates. Frellesvig, Bjarke. 2010. A History of the Japanese language. Cambridge: Cambridge University Press. Givón, Talmy. 1982. Evidentiality and epistemic space. Studies in Language 6. 23–49. Givón, Talmy. 1984. Syntax: A Functional-Typological Introduction. Vol. 1. Amsterdam: John Benjamins. Grice, Paul. 1975. Logic and Conversation. In Peter Cole and Jerry L. Morgan (eds.), Syntax and semantics, vol. 3: Speech Acts, 41–58. New York: Academic Press. Reprinted in Paul Grice 1991, Studies in the way of words, 22–40. Cambridge, MA: Harvard University Press. Hanks, William F. 1992. The indexical ground of deictic reference. In Alessandro Duranti and Charles Goodwin (eds.), Rethinking context: Language as an interactive phenomenon, 43–77. Cambridge: Cambridge University Press. Harris, Alice and Lyle Campbell. 1995. Historical syntax in cross-linguistic perspective. Cambridge: Cambridge University Press. Heine, Bernd and Mechthild Reh. 1984. Grammaticalization and Reanalysis in African Languages. Hamburg: Helmut Buske. Iwasaki, Shoichi. 2000. Suppressed assertion and the function of the final-attributive in prose and poetry of Heian Japanese. In Susan Herring, Pieter Reenen, and Lene Schøsler (eds.), Textual parameters in older languages, 237–272. Amsterdam: John Benjamins. JDB = Omodaka, Hisataka et al. (eds.). 1967. Jidai-betsu kokugo daijiten, jōdai-hen [Period-specific dictionary of the Japanese language, Old Japanese]. Tokyo: Sanseido. Jorden, Eleanor H. and Mari Noda. 1987. Japanese: The spoken language part 1. New Haven: Yale University Press. Koike, Seiji. 1967. Rentaikei shūshihō no hyōgen kōka – Konjaku monogatari, Genji monogatari o chūshin ni [The effect of adnominal conclusive expressions: With a focus on Konjaku monogatari and Genji monogatari]. Kokubungaku Gengo to Bungei 54. 12–21. Kushima, Shigeru. 1989. Rentaikei shūshihō no imi suru mono: Kakari-musubi no imi kōzō to sono hōkai [What the adnominal conclusive means: the structure of km’s meaning and its demise]. Shizudai Kokubun 34. 36–37. Levinson, Stephen C. 1983. Pragmatics. Cambridge: Cambridge University Press.
14 Voicings of kakari-musubi: Shifting from cleft construction to referential predicate
345
Levinson, Stephen C. 2000. Presumptive meanings: The theory of generalized conversational implicature. Cambridge, MA: MIT Press (a Bradford Book). Levinson, Stephen C. 2010. Generalized conversational implicature. In Louise Cummings (ed.), The pragmatics encyclopedia, 201–203. London: Routledge. Maynard, Senko K. 1992. Cognitive and pragmatic messages of a syntactic choice: The case of the Japanese commentary predicate n(o) da. Text 12(4). 563–613. McGloin, Naomi Hanaoka. 1980. Some observations concerning no desu expressions. Journal of the Association of Teachers of Japanese 15, no. 2. 117–149. Mikami, Akira. 1953. Gendai gohō josetsu: Shintakusu no kokoromi [An introduction to the grammar of modern Japanese: A syntactic approach]. Tokyo: Toko Shoin. Reprinted in 1972. Tokyo: Kurosio Publishers. Noda, Mari. 1990. The extended predicate and confrontational discourse in Japanese. Ithaca, NY: Cornell University dissertation. NKBZ = Abe, Akio et al. (eds.). 1970–1976. Nihon koten bungaku zenshū [Japanese classical literature, complete collection]. Tokyo: Shogakkan. Ochs, Elinor. 1996. Linguistic resources for socializing humanity. In John J. Gumperz and Steven C. Levinson (eds.), Rethinking linguistic relativity, 407–437. Cambridge: Cambridge University Press. Ohno, Susumu. 1993. Kakari-musubi no kenkyū [On Kakari-musubi]. Tokyo: Iwanami Shoten. Onoe, Keisuke. 1982. Bun no kihon kōsei—shiteki tenkai [Basic structure of the sentence: Historical development]. In Kenji Morioka, Yutaka Miyaji, Hideo Teramura and Yoshiaki Kawabata (eds.), Kōza nihongogaku 2: Bunpō-shi [Japanese language series: History of grammar], 1–19. Tokyo: Meiji Shoin. Prince, Ellen F. 1978. A comparison of wh- clefts and it-clefts in discourse. Language 54. 883–906. Quinn, Charles. 1987. A functional grammar of predication in Classical Japanese. Ann Arbor, MI: University of Michigan dissertation. Quinn, Charles. 1997. On the origins of Japanese sentence particles ka and zo. In John Haig and Ho-min Sohn (eds.), Japanese Korean Linguistics 6. 61–89. Stanford, CSLI. Quinn, Charles. 2015. Why izenkei in koso-focused kakari-musubi?—Some considerations. Paper presented at Kakarimusubi from a Comparative Perspective, International Workshop, National Institute for Japanese Language and Linguistics (NINJAL), 5–6 September. Quinn, Charles. 2018. The epistemic/evidential dimension in Heian Japanese. Paper presented at the 27th meeting of the Association for Japanese Literary Studies, University of California, Berkeley, 6–8 September. Quinn, Charles. 2024 (forthcoming). Izenkei in early Japanese kakari-musubi clefts: a provisional account. In M. Nakayama, Z. Xie, and M. Chan (eds.), Buckeye East Asian Linguistics 8 (BEAL 8). Knowledge Bank, The Ohio State University Libraries. Sakakura, Atsuyoshi. 1993. Nihongo hyōgen no nagare [The flow of Japanese expression through time]. Tokyo: Iwanami Shoten. Schaffar, Wolfram. 2002. Kakari musubi, noda-constructions, and how grammaticalization theory meets formal grammar. Japanese/Korean Linguistics 10. 320–333. Stanford, CSLI. SNKBT = Satake, Akihiro et al. (eds.). 1989–2005. Shin Nihon koten bungaku taikei [New Japanese classical literature series]. Tokyo: Iwanami Shoten. Shinzato, Rumiko and Leon Serafim. 2013. Synchrony and diachrony of Okinawan kakari musubi in comparative perspective with premodern Japanese (Languages of Asia, Vol. 11). Leiden: Global Oriental. Vovin, Alexander. 2003. A reference grammar of Classical Japanese prose. London: Routledge Curzon. Wrona, Janick. 2008. The Old Japanese complement system: A synchronic and diachronic study. Folkestone, Kent: Global Oriental. Yamada, Yoshio. 1908. Nihon bunpō ron [A treatise on Japanese grammar]. Tokyo: Hōbunkan Shuppan.
Akira Watanabe
15 Loss of wh-movement 1 Introduction Probably the most significant syntactic property that characterizes earlier stages of Japanese is the system of so-called kakari-musubi, where a clause-internal kakari particle determines the inflectional form of clause-final predicative elements. The traditional definition of kakari-musubi just given has a more or less morphological flavor, but it in fact belongs to matters of syntax, imposing a rigid word order restriction on the case-marked subject and displaced phrasal expressions accompanied by one of the kakari particles. This word order restriction was lost in the transition from the Nara (Old Japanese) to the Heian (Early Middle Japanese) periods, and the loss brought about several important changes in the grammatical system. This chapter presents an overall framework of analysis that helps explain why these changes took place at about the same time, as well as the cause of the loss of the word order restriction itself. The phenomena will also be situated in a cross-linguistic context. The particles to be taken up in this chapter are pa (> wa), ka, ya, so/zo, and koso. What figures prominently in the discussion of these particles are the notions of topic and focus. In the simplest case, focus can be defined as the expression in an answer to a question that corresponds to the questioned phrase. In the following exchange, for example, Mary is focused. (1) Q: Which student did you see yesterday? A: I saw Mary yesterday. The notion of focus is also extended to comparable expressions in sentences that are not necessarily answers to a question. Topic, on the other hand, is what the sentence is about, roughly speaking. In English, it can be indicated by expressions like as for and speaking of. All the historical stages of Japanese also have a dedicated particle. The topic marker status of pa in Old Japanese and Early Middle Japanese is fairly uncontroversial and will be used to investigate the nature of other kakari particles. For more detailed overviews of focus, topic, and related discourse-sensitive notions in general, see Heycock (2008), Krifka and Musan (2012), and Roberts (2011). We call expressions like which student wh-phrases for short, making use of the English spelling. Note that wh-phrases are displaced to a clause-initial position in English, as can be seen from (1). This process is called wh-movement, which is also applicable to phrases that do not necessarily contain a wh-word, under the conception of Universal Grammar widely assumed since Chomsky’s (1977) seminal work. Throughout this chapter, “wh-movement” will be used in this broader sense. https://doi.org/10.1515/9781614512851-016
348
Akira Watanabe
The following section discusses the structural analysis of relevant constructions in Old Japanese, which crucially makes use of wh-movement. Section 3 deals with the demise of this grammatical system in Early Middle Japanese, which resulted in a rather different picture of the surviving kakari particles. Section 4 concludes.
2 Wh-movement in Old Japanese In the long history of the study of kakari-musubi (see Frellesvig 2010 and Ōno 1993), it is only in recent years that the general consensus has emerged that it involves syntax in a non-trivial way. Nomura (1993) and Sasaki (1992) observed that during the seventh and eighth century period, there was a rigid word order restriction concerning the topic and the focus. How this word order restriction should be analyzed is a question for syntax. It is convenient for the sake of discussion to classify kakari-musubi clauses into interrogative and declarative and start with the former.
2.1 Interrogative sentences In interrogative sentences, the particle ka marks the wh-phrase in wh-questions and a focused phrase in yes-no questions, while the particle ya is a focus marker in yes-no questions. The focused phrase marked either by ka or by ya must appear below the pa-marked topic and above the genitive subject. Examples that illustrate the ordering of the genitive subject and the focus are given in (2). (2) a.
b.
c.
. . . iduku yu ka imo ga iri-ki-te ime ni where through KA wife gen enter- come-ger dream loc miye-turu? appear-perf ‘From where did my wife come and appear in my dream?’ (MYS 12.3117) . . . Patuse no kapa pa ura na-mi ka pune no Hatsuse gen river top shore absent-acop KA boat gen yori-ko-nu? approach-come-neg ‘Is it because Hatsuse River has no shore that no boat comes near?’ (MYS 13.3225) . . . Torikape-te kire-ba ya kimi ga omo wasure-taru? exchange-ger put.on-prov YA you gen face forget-stat ‘Is it because you change cloths often that you forget my face?’ (MYS 11.2829)
15 Loss of wh-movement
349
(2b) in addition has a topic preceding the focus. Further examples that contain a topic and a focus are provided in (3), where the subject is topicalized. (3) a.
b.
. . . Sasanami no opoyamamori pa ta ga tame ka Sasanami gen big.mountain.guard top who gen sake KA yama ni sime yupu? . . . mountain loc marker tie ‘For whom does the mountain guard at Sasanami tie the marker around the mountain?’ (MYS 2.154) Momosikwi no opomiya-pito pa itoma are ya ume wo (Epithet) gen court-person top leisure be YA plum acc kazasi-te koko ni tudwop-yeru. wear.in.hair-ger here loc gather-stat ‘Is it because courtiers have free time that they wear plum flowers in their hair and have gathered together here?’ (MYS 10.1883)
The focus nature of the ka-marked phrase can be seen in examples like (4), where its correspondence with a wh-phrase is transparent, despite the fact that it appears in a yes-no question. (4)
Awayuki ka padareni puru to miru madeni nagarape-tiru bubbly.snow KA sparsely fall comp see till glide-fall pa nani no pana so mo. top what gen flower SO MO ‘What kind of flower is it that is falling smoothly in such a way as to make it look as if bubbly snow is falling sparsely?’ (MYS 8.1420)
Notice that bubbly snow is presented as an unrealistic false answer to the matrix wh-question in (4). Nomura (1993) points out that the ka-marked phrase must come after the pa-marked topic but must precede the genitive-marked subject (with ga or no), regardless of whether it is a wh-phrase or not, on the basis of the quantitative data summarized below about the position of the subject. (Here and elsewhere, quantitative data on Old Japanese cited are based on examples where the relevant particles are phonographically written.) (5) Ordering of the subject and the ka-marked phrase in Man’yōshū (Nomura 1993) I. Genitive subject: XP-ka . . . Subj-no/ga order: about 90 examples Subj-no/ga . . . XP-ka order: 4 (or 5) examples II. Topicalized subject: XP-ka . . . Subj-pa order: 2 (or 3) examples Subj-pa . . . XP-ka order: about 50 examples III. Bare subject: XP-ka . . . Subj order: 13 examples Subj . . . XP-ka order: about 30 examples
350
Akira Watanabe
Sasaki (1992) independently reaches the conclusion that the pa-marked topic, whether functioning as a subject or not, must precede the ka-marked phrase. Sasaki further observes that the ka-marked phrase and the ya-marked phrase are subject to the same word order restriction. We can conclude from their work that Old Japanese has a clause structure that gives rise to the word order pattern summarized in (6), though some questions remain for the status of bare subjects (see Kuroda 2007 for some discussion of bare subjects, which will be taken up later on). (6) Topic – Focus – Genitive Subject This word order pattern is found in other languages as well. The best-known case is Hungarian, studied by Puskás (2000), who proposes the following structure, adapting Rizzi’s (1997) split CP system: (7) [TopP Spec Top [FocP Spec Foc [TP Subj T VP]]] Under this analysis, wh-phrases are considered to undergo movement to Spec of FocP, even though they do not act semantically as focus, strictly speaking. Watanabe (2002) suggests that essentially the same structure underlies Old Japanese, modulo its head-final nature. According to this analysis, the pa-marked topic and the ka- or ya-marked focus are placed in Spec of TopP and Spec of FocP, respectively, through wh-movement. Note that wh-movement applies in a wide variety of constructions including, but not limited to, wh-questions. The process is also called A-bar movement or operator movement. Now, kakari-musubi is traditionally conceived as a phenomenon where a clause-internal particle triggers a specific inflectional form of the predicate. It is therefore very interesting to observe that it has been known since Chung’s (1982) work on Chamorro that wh-movement induces a change in the verbal inflection. In Chamorro, wh-movement triggers “nominalization”, which means that the subject is indexed with an agreement marker that is usually reserved for possessors. In (8b), the subject is marked with -ña, which also appears in (8a) as the third-person possessor marker. (8) a.
b.
. . . Ha-fahan si Maria i sanhilo’-ña erg.3.sg-buy SI Maria the blouse-her ‘Maria bought her blouse at the store.’ Hafa f-in-ahan-ña si Maria gi what in-buy+nmlz-her SI Maria loc ‘What did Maria buy at the store?’
gi tenda. loc store tenda? store
(Chung 1982: 50)
At an appropriate abstract level, the Chamorro phenomenon is quite similar to the use of the adnominal (rentai) form for ka- or ya-marked phrases in Old Japanese, which is responsible for the fact that the subject is marked with -ga and -no, the case particles for
15 Loss of wh-movement
351
genitive. Case and agreement form a unified syntactic system under the present theoretical understanding. Thus, kakari-musubi is not so unusual as it first appears. It is a recurrent structural pattern found in languages of the world. For a range of morpho-syntactic modifications of clausal elements due to wh-movement, see Watanabe (1996).
2.2 Topic-focus articulation in declarative sentences In declarative sentences, the kakari particles so/zo and koso are employed. Since they behave rather differently, they need a separate treatment.
2.2.1 So/zo The particle so/zo displays a complex behavior. Sasaki (1992) observes that like the particles ka and ya, so/zo must appear above the genitive subject. According to his counting, there are a little more than ten unambiguous cases like (9) that verify the generalization, while there is no convincing example with the opposite order. (9) . . . Ikwoma no yama wo kwoye-te so a ga kuru. Ikoma gen mountain acc go.beyond-ger SO I gen come ‘I’ve come home from behind Mt. Ikoma.’ (MYS 15.3590) The picture looks essentially the same as in the case of ka and ya. Note also that the adnominal ending is used in (9). When we turn to the relative order of so/zo and the topic marker pa, however, the result is mixed. (11) gives Watanabe’s (2007) counting of the relevant examples in Man’yōshū. (10) Ordering of so/zo and the pa-marked topic in Man’yōshū (Watanabe 2007) i. XP-so/zo . . .Topic-pa order: 30 examples ii. Topic-pa . . .XP-so/zo order: 31 examples Most of the examples that belong to the category in (10i) are given in Sasaki (1992). Two examples of each class are presented below. (11) a.
. . . Mimiga no mine ni toki-naku so yuki pa Mimiga gen mountain loc time-without SO snow top puri-kyeru. . . . fall-mpst ‘Snow fell incessantly on Mt. Mimiga.’ (MYS 1.25)
352
Akira Watanabe
b.
c.
d.
. . . imo so topoku pa wakare-ki-ni-kyeru. wife SO far top part.with-come-perf-mpst ‘I have come very far, leaving my wife behind.’ (MYS 15.3698) Ima pa a pa wabwi so si-ni-kyeru. . . . now top I top sorrow SO do-perf-mpst ‘Now I’m in deep sorrow.’ (MYS 4.644) . . . Yosinwo no miya pa yama taka-mi kumo so Yoshino gen palace top mountain tall-acop cloud SO tanabiku. hover ‘Over the palace at Yoshino, clouds hover beautifully because the mountains are tall.’ (MYS 6.1005)
Watanabe (2007) proposes that the instances of so/zo in (10i) should be analyzed as a contrastive topic marker. Contrastive topicalization is illustrated by a Modern Japanese example in (12), where John and Mary are contrasted. (12) John wa yakyuu o si ni itta ga, Mary wa kyoositu John top baseball acc do loc went but Mary top classroom ni nokot-te benkyoo-sita. loc stay-ger study-did ‘John went to play baseball, but Mary stayed in the classroom to study.’ Old Japanese also has a construction of this sort. In (13) below, the mountain road is contrasted with the streets in Nara, where both are marked by pa. (13) Awoniyosi Nara no opoti pa yuki-yokye-do, kono (Epithet) Nara gen big.street top go-good-though this yama-miti pa yuki-asik-ari-kyeri mountain-road top go-bad-be-MPST ‘Though the big streets in Nara are comfortable to walk on, this mountain road was really bad.’ (MYS 15.3728) Note also that the first clause takes a concessive ending that consists of the izenkei form and a subordinator do. Interestingly, this construction sometimes matches up pa with so, the latter appearing in the matrix clause with the adnominal ending, as in (14). (14) . . . Nipu no kapa koto pa kaywope-do, pune so kaywopa-nu. Nyū gen river word top go.along-though boat SO go.along-NEG ‘Though rumors go along the river of Nyū, boats do not.’ (MYS 7.1173)
15 Loss of wh-movement
353
There are nine clear examples of this kind in Man’yōshū, including (14). In addition, there are cases like (15), where two independent sentences are juxtaposed to highlight contrast. (15) Wa ga pori-si Nwosima pa mise-tu. Soko pukaki Agwone I gen want-spst Noshima top show-perf bottom deep Agone no ura no tama so piripa-nu. gen bay gen stone SO pick.up-neg ‘You have shown me Nosima, which I wanted to see. I haven’t picked up precious stones at the bay of Agone, whose bottom is very deep.’ (MYS 1.12) Note again that the pa-marked topic in the first sentence is paired with a so-marked phrase in the second. In fact, (11b) is also preceded by a subordinate clause of the type exemplified by (13) and (14). Since Man’yōshū mostly consists of very short verse texts, it is not always possible to point to evidence of this kind in the 30 examples of category (10i). But if contrastive topics are placed in Spec of TopP, one can accommodate these 30 examples by means of the structure (7) with a minimal modification, which is to allow multiple specifiers for TopP. Watanabe (2007) goes on to argue that some instances of so/zo in the 31 examples of (10ii) should be analyzed as a focus marker. He points out that there are cases of (10ii) that the multiple-Spec analysis cannot deal with. The crucial observation is that the pa-marked phrase and the so/zo-marked phrase are always adjacent in (10i) with only one exception, whereas they are not adjacent in 13 out of the 31 examples of type (10ii). In (11d), for example, yama takami comes in between. If the intervening phrase cannot be treated as a topic, the only possible conclusion is that it is lower than TopP, probably adjoined to FocP through scrambling, with the so/zo-marked phrase sitting in Spec of FocP. Interestingly, Lambova (2001) observes that an adverbial cannot break up a sequence of topics but it can intervene between a topic and the following focus in Bulgarian. The Old Japanese situation looks exactly the same. An additional argument presented by Watanabe (2007) for the focus marker status of so/zo refers to the fact that so/zo can be attached to a wh-phrase. Though this is a minority pattern at that time, we find around ten examples like (16) in Man’yōshū. (16) . . . idure no imo so sokoba kwopwi-taru? which gen woman SO that.degree love-stat ‘Which woman is so much in love?’
(MYS 4.706)
In addition, the grammaticalized wh-phrase nazo ‘why’ derived from nani ‘what’ + zo appears in eight examples. Given the connection between the wh-phrase and focus, the focus use of so/zo is expected. And indeed, the focus marker analysis of so/zo is well established in the literature (Ōno 1993). We are led to conclude that the kakari particle so/zo can function either as a contrastive topic marker or as a focus marker.
354
Akira Watanabe
It may be appropriate at this point to mention the widespread view (see Ōno 1993, Okazaki 1996, Sakakura 1993, Serafim and Shinzato 2011, and the references cited there) that the original form of the particle is so, derived from the homophonous demonstrative. Relevant in this connection is the fact that German contrastive left dislocation involves the so-called d-pronoun, which belongs to the demonstrative system. Grohmann (2003) demonstrates that German contrastive left dislocation features a contrastive topic, in contrast to hanging topic left dislocation, on the basis of data like (17). (17) Hast du gestern die Anna getroffen? ‘Did you meet Anna yesterday?’ a. Nein. Den Martin, den habe ich gestern getroffen. (contrastive left dislocation) b. #Nein. Der Martin, ich habe ihn gestern getroffen. (hanging topic left dislocation) ‘No. I met Martin yesterday.’ In (17a), the d-pronoun den immediately follows the contrastive topic den Martin. A different kind of left dislocation in (17b) with ihn as the resumptive pronoun disallows the contrastive topic reading and hence is deviant in this context. Serafim and Shinzato (2011) apparently assume that the demonstrative origin of so/zo leads to the focus interpretation, but that is not obvious. Grohmann shows that contrastive left dislocation cannot support the focus interpretation, by observing that it is infelicitous as an answer to a wh-question, as in (18). (18) Q: Wen has du gestern getroffen? ‘Who did you meet yesterday?’ A: #Den Martin, den habe ich gestern getroffen. Given that a German construction with a demonstrative pronoun is compatible with the contrastive topic reading, but not with the focus interpretation, and that the Old Japanese particle so/zo allows both, the original use of so/zo may have been linked to contrastive topics, to the extent that the kakari particle so/zo is derived from the demonstrative so. If so/zo allows both the focus and contrastive topic readings, is there a way to unify the two uses? Watanabe (2007) suggests that Vallduví and Vilkuna’s (1998) notion of “kontrast” is useful in capturing the shared property. Vallduví and Vilkuna propose that if an expression a is kontrastive, a membership set M = {. . ., a, . . .} is generated and becomes available to semantic computation. Rooth’s (1992) alternative semantics makes use of a set of alternatives for the focused constituent as a crucial notion in focus semantics, and thus can be accommodated quite easily under the notion of kontrast. An answer to a wh-question chooses one candidate out of the conceivable alternative choices, namely, the members of set M. The contrastive topic reading, on the other hand, arises when a property P holds of a member a of the set M generated by kontrast and other properties P’ hold of some other members of M. See Büring (2003) and Hendriks
15 Loss of wh-movement
355
(2004) for similar proposals about contrastive topics. More recently, Wagner (2012) proposes that contrastive topicalization involves recursive application of an alternative set formation. Notice that the properties P and P’ that characterize the contrastive topic interpretation are nothing other than alternatives. Thus, the notion of kontrast is shared by focus and contrastive topic. The kakari particle so/zo, then, is a marker of kontrast.
2.2.2 Koso Let us finally turn to the analysis of the particle koso. Ishida (1939) characterized it as a contrastive topic marker. This analysis is also adopted by Ōno (1993). Watanabe (2007) argues that the syntax of koso confirms its contrastive topic marker status. Koso triggers the izenkei form. According to Ishida, the original use of this form is a subordination marker in clause chaining that can feature contrastive topicalization. Recall that the construction hosting contrastive topics, exemplified by (13) and (14), makes use of the izenkei stem in the subordinate clause. Ishida observes that koso is also matched up with the topic marker pa in the same construction, as in (19). (19) . . . pito koso sira-ne, matu pa siru-ramu. person KOSO know-neg pine.tree top know-would ‘Though people do not know, pine trees surely do.’
(MYS 2.145)
Watanabe (2007) points out that if koso is a contrastive topic marker, it should be able to appear both above and below a pa-marked topic in the multiple topic structure, giving the statistical data below. (20) Ordering of koso and the pa-marked topic in Man’yōshū (Watanabe 2007) i. XP-koso . . .Topic-pa order: 12 examples ii. Topic-pa . . .XP-koso order: 11 examples Representative examples of each order are given in (21). See also Sasaki (1992), who presents most of the type (20i) examples. (21) a.
b.
Kinopu koso tosi pa pate-sika, . . . yesterday KOSO year top end-spst ‘Though the previous year just ended yesterday, . . .’ (MYS 10.1843) . . . pune no yupu-sipo wo matu-ramu ywori pa ware boat gen evening-tide acc wait-would than top I koso masare. KOSO exceed ‘I long for you more strongly than a boat would wait for the evening tide.’ (MYS 11.2831)
356
Akira Watanabe
Furthermore, the koso-marked phrase should appear above the genitive subject. There is only one relevant example, as noted by Sasaki (1992) and Nomura (1993). The prediction is borne out by (22). (22) . . . masurawonokwo no kwopure koso wa ga yupu brave.man gen yearn KOSO I gen do.up kami no piti-te nure-kyere. hair gen be.soaked-ger loosen-mpst ‘As my brave man is yearning for me, my tied-up hair has been soaked and come loose.’ (MYS 2.118) These results follow if the koso-marked contrastive topic and the pa-marked topic are both placed in Spec of TopP. In addition, if the multiple-Spec analysis of multiple topics is on the right track for the koso-marked contrastive topic as well, multiple topics of (20) should be adjacent. Watanabe (2007) observes that with one exception, this prediction is also borne out. The examples in (21) are consistent with this word order restriction. The exception, presented below, is the very first piece in Man’yōshū. (23) . . . Yamato no kuni pa osinabe-te ware koso wore. Yamato gen country top rule-ger I KOSO stay ‘I am here to rule the country of Yamato.’ (MYS 1.1) Given that the overall number of relevant examples is smaller than in the case of so/zo, the statistical significance may be questioned. But it is encouraging to find that the data pattern is consistent with the idea that koso is a contrastive topic marker.
2.3 Issues We have seen that Old Japanese provides an articulate structure for discourse-sensitive expressions in the left periphery of the clause and that these left-peripheral positions serve as the landing sites of wh-movement. Let us take up some additional details worth mentioning before we move on to the discussion of later historical stages. In general, wh-movement can take place across a clause boundary. If Old Japanese makes use of it, therefore, we expect to find instances of long-distance wh-movement. Kuroda (2007) and Tonoike (2002) bring up the point, reporting that there is no convincing example. They do not discuss what happens in bi-clausal structure, however. The two examples mentioned by Tonoike are mono-clausal, and hence irrelevant. In order to decide whether long-distance wh-movement is possible or not, it is necessary to examine how the (underlying) structure of the form [CP . . . [CP . . . wh . . . ] . . . ] behaves when the wh-phrase in the embedded clause takes the matrix scope. The task is made
15 Loss of wh-movement
357
difficult by the fact that Man’yōshū is mostly a collection of short poems, which do not often provide conditions suitable for wh-movement from within an embedded clause. One candidate for long-distance wh-movement is the following example: (24) Ikani aru Puse how be Fuse mise-mu to show-will comp ‘How beautiful is show me?’
no ura gen bay ware wo I acc the bay of
so mo kokodakuni kimi ga SO MO this.much you gen todomuru? keep Fuse that you try so much to keep me here to (MYS 18.4036)
In this example, the wh-phrase, marked with so, precedes the matrix genitive subject and is interpreted as the direct object of the embedded verb within a purpose clause. Note also that the matrix verb is in the adnominal form. A single example may not be sufficient, but at the same time, there is no bi-clausal example that traps within an embedded clause a wh-phrase that takes the matrix scope. Given the nature of Man’yōshū, we must leave the matter here. In the discussion so far, only the possibility that wh-movement lands in the region above TP is considered. A focus position lower than TP has been suggested for other languages in Jayaseelan (2001). Aldridge (2010) claims that Old Japanese also makes use of the lower focus position. For this claim to hold water, the subject marked with ga or no should not be placed in Spec of TP but must stay within the verbal projection, since wh-phrases and focused phrases precede the subject marked with ga or no. Aldridge asserts that the incompatibility of the case particles ga and no with the conclusive form is problematic for the placement of the genitive subject in Spec of TP. The point is far from obvious, though. As we have seen above in relation to Chamorro, wh-movement can trigger various morpho-syntactic modifications of clausal properties. The use of the adnominal form with the kakari particles ka, ya, and so/zo is just one such phenomenon. It does not follow from this phenomenon that the genitive subject is not raised to Spec of TP. Even granting that the genitive subject stays within the verbal projection does not lead to the conclusion that wh-movement targets the focus position below TP. Aldridge points to the possibility of scrambled phrases in front of the raised wh-phrase as evidence that the landing site of wh-movement is lower than TP. We have seen a similar example with the particle so in (11d). One of the examples given by Aldridge is: (25) . . . sima no mi-pasi ni tare ka sumapa-mu? island gen hon-step loc who KA stay-will ‘Who stays on the steps of the island palace?’
(MYS 2.187)
The assumption adopted by Aldridge is that scrambling targets TP. If so, the wh-phrase following a scrambled phrase should be lower than TP. Once again, her assumption is far from obvious in the case of Old Japanese. If adjunction to FocP is a legitimate
358
Akira Watanabe
option in Universal Grammar, as argued independently by Lambova (2001) for Bulgarian, there is no obstacle to the placement of wh-phrases and focused expressions in Spec of FocP above TP. A typologically interesting property of Old Japanese wh-questions is that the particle ka can be attached to a clausal constituent containing a wh-phrase. Below is an example cited in Aldridge (2010). (26) . . . kari no tubasa no opopiba no iduku mori-te ka goose gen wing gen covering.wing gen where leak-ger KA simo no puri-kye-mu? frost gen fall-spst-conjec ‘What part of the great wings of the wild goose is leaking, such that the frost has fallen?’ (MYS 10.2238) Notice that ka appears after the verbal complex that heads the subordinate clause containing a wh-phrase and that the subordinate clause itself precedes the genitive subject simo. This kind of wh-question is impossible in English. For that reason, only a very awkward translation can be given for (26). Watanabe (2004) points out that this kind of large-scale pied-piping is limited to those languages that have a non-interrogative use of wh-phrases, such as Basque and Imbabura Quechua. Tlingit, studied by Cable (2010), can be added to this class of languages. And indeed, Old Japanese also fits this generalization, as can be seen from (27), where tare receives a non-wh quantificational interpretation. (27) . . . tare to ipu pito mo kimi ni pa masa-zi. who comp say person MO you dat top be.superior-neg.conject ‘Nobody could be nicer than you.’ (MYS 11.2628) See Ōshika (1991) for non-interrogative indefinite uses of wh-phrases in Old Japanese. If a constituent larger than the wh-phrase can be moved to Spec of FocP, a mismatch between the displaced phrase and the expression that semantically functions as focus is expected. As long as the displaced phrase is larger, there is nothing surprising. The opposite type of mismatch, however, is found in Old Japanese. Konoshima (1966) and Nomura (2001) observe that both ka and ya allow such a mismatch. (28) shows the point straightforwardly, because the second sentence there as a whole corresponds to the wh-phrase in the first sentence asking for a reason. (28) Nazo sika no wabwi-naki su-naru? Kedasiku mo aki-nwo no why deer gen lonesome-bell do-evid perhaps MO fall-field gen pagwi ya sigeku tiru-ramu? bush.clover YA incessantly fall-would ‘Why are deer belling lonesomely? Perhaps bush clover flowers in the autumn field are falling incessantly?’ (MYS 10.2154)
15 Loss of wh-movement
359
The propositional content of the second sentence excluding kedasiku-mo and ramu is the focus. The ya-marked phrase pagwi-ya is only a part of it. This phenomenon cannot be handled in terms of pied-piping. Something else is going on. Since short poems in Man’yōshū lack detailed contextual information, it may not be so easy to identify what expression is acting as focus in each case. But it is worth mentioning that exactly the same kind of mismatch arises in Czech and German, according to Fanselow and Lenertová (2011). This phenomenon forces us to rethink the syntax-semantics interface involving FocP. Finally, let us take up bare wh-phrases, a grammatical form that may be related to the loss of wh-movement, which will be the topic of the next section. Wh-phrases can sometimes be used without any kakari particle in Old Japanese. It has been known since Saeki (1938) that in these ka-less wh-questions, the subject is not marked with any case particle, as illustrated in (29). (29) . . . Ikani ware se-mu? how I do-will ‘What shall I do?’
(MYS 18.4046)
Furthermore, there is no example with an unambiguous adnominal ending when the wh-phrase is not accompanied by ka. According to Nomura (2011), there are slightly more than 60 ka-less wh-questions in Man’yōshū, with a couple of examples that have an unambiguous conclusive form. Wh-questions of this type form a minority pattern at the time of Old Japanese. There are more than 200 wh-questions with ka in Man’yōshū (Nomura 2001). Kuroda (2007) argues that the wh-phrase is not raised but stays in its original position in these ka-less wh-questions, on the basis of examples like (30). (30) . . . ibuseki a ga mune ta wo mi-ba yama-mu? melancholic I gen heart who acc see-if stop-will ‘Whom should I see for my melancholic heart to heal? (lit.: If I see whom will my melancholic heart heal?)’ (MYS 10.2263) The idea is that general constraints on movement prohibit the wh-phrase ta-wo from being raised out of a conditional clause. Kuroda’s conclusion is premature, however, in view of the possibility of large-scale pied-piping mentioned above. Significantly, ka can be attached to a conditional clause containing a wh-phrase, as in (31), where a genitive subject comes after the displaced conditional clause.
360
Akira Watanabe
(31) Watatumi no idure no kamwi wo inora-ba ka yuku-sa mo sea gen which gen god acc pray-if KA go-nmlz MO ku-sa mo pune no payakye-mu? come-nmlz MO ship gen fast-will ‘To which god of the sea ought I pray for my voyage back and forth to be quick? (lit.: If I pray to which god of the sea will the travel by ship back and forth be quick?)’ (MYS 9.1784) It is conceivable that the same kind of pied-piping takes place in (30). One should also note that the original position of the manner adverb ikani in (29) is likely to be below the bare subject. If so, we need to posit wh-movement for (29). Given the absence of an overt genitive subject, which should be distinguished from a bare subject, in ka-less wh-questions, simple empirical considerations will probably remain inconclusive forever. If it turns out that an argumental bare wh-phrase always precedes a bare subject, we can draw the conclusion that wh-movement takes place. But I have not found a single example of that sort. The interest of the ka-less version is that it may be a precursor to the wh-question in Modern Japanese, as Kuroda notes. Since Modern Japanese is a wh-in-situ language, when and how wh movement was lost is a very important question to ask. Let us therefore turn next to the loss of wh-movement in the history of Japanese.
3 Early Middle Japanese Nomura (1996) points out that the word order restriction found in Old Japanese was lost during the transition to the Heian period, which means that wh-movement was lost. This is a rather drastic syntactic change. This section reviews the nature of this change and the repercussions it has in other areas of grammar.
3.1 When and how wh movement was lost Though Nomura (1996) does not give statistical figures, Watanabe (2002) observes that wh-questions consistently lack an overt genitive subject in the first four volumes of The Tale of Genji. Nomura’s (2005) counting in relation to the kakari particles ya and zo also shows that examples with an overt genitive subject with ga or no are extremely rare when these particles are used. Here is the summary of Nomura’s (2005) data about the entire 54 volumes of Genji.
15 Loss of wh-movement
361
(32) Ordering of the genitive subject and the discourse-prominent phrase in Genji (Nomura 2005) I. Wh-questions (total 1354): Wh . . . Subj-no/ga order: 19 Subj-no/ga . . . WH order: 11 II. ya (total 1751): XP-ya . . . Subj-no/ga order: 4 Subj-no/ga . . .XP-ya order: 10 III. zo (total 1317): XP-zo . . . Subj-no/ga order: 3 Subj-no/ga . . . XP-zo order: 4 Though the number is small, both orders are attested, crucially. Below is one of the rare cases where it is clear that the wh-phrase remains in-situ. (33) Medurasiki podo-ni nomi aru gotaimen no ikade ka wa rare extent-in only is meeting gen how KA top oroka nara-mu? slack be-will ‘How could such an infrequent rendez-vous not be passionate?’ (Genji, Sakaki, SNKBZ vol. 2: 105) In view of these observations, Watanabe (2002) claims that wh-movement no longer takes place in Early Middle Japanese and suggests further that the scarcity of overt genitive subjects in Early Middle Japanese is due to the increase of subject topicalization. According to his counting, there are eleven instances with a topicalized subject out of 81 wh-questions in the first four volumes of Genji, while there is none with an overt genitive subject. Recall that there are more genitive subjects than topicalized subjects in Man’yōshū according to Nomura’s data in (5). The ratio is reversed. Watanabe (2002) goes on to provide the following scenario for the loss of wh-movement. Suppose that the consistent placement in front of an overt genitive subject is the clue that children use to conclude that wh-movement takes place in the grammar that they are trying to acquire. Assume further that “no movement” is the default value of the parameter concerning wh-questions. Under this set of assumptions about language acquisition, the increase of subject topicalization leads to the result that deprives children of the crucial evidence that makes them choose the wh-movement option. They thus stick to the default option, namely, wh-in-situ. And Watanabe’s claim is that this is exactly what happened in the transition to the Heian period, though lack of ninth century materials prevents us from tracing the posited historical development over an extended period of time. Increased use of subject topicalization itself may not belong to matters of grammar. It probably has to do with verbal behavior of speakers. But it can bring about a language change through the mechanism of language acquisition. See also Watanabe (2006a) for more on the mechanism of parameter setting in relation to wh-movement. Incidentally, parameter-setting considerations are also relevant for the analysis of ka-less wh-questions in Old Japanese, taken up in Section 2.3 above. If there can be
362
Akira Watanabe
only a single parameter concerning the status of wh-movement in wh-questions, the evidence available from wh-phrases accompanied by ka forces children to adopt the wh-movement option for ka-less wh-questions as well. If, on the other hand, there can be multiple parameters reserved for distinct morpho-syntactic shapes of wh-phrases, ka-less wh-questions in Old Japanese may involve wh-in-situ, given that no word order pattern involving the bare wh-phrase and an overt genitive subject exists. We need to look at a wide range of languages to resolve the issue. Currently, no consensus view (or even discussion) exists in the literature, as far as I am aware. At the same time, the overall pattern of relevant constructions is something to be taken into account in this connection. Recall that ka was not limited to wh-phrases at the time of Old Japanese. This fact may extend the coverage of the wh-movement parameter to include focus constructions and wh-questions with and without ka, leaving only a single parametric choice for all. But then, wh-movement should be imposed on ka-less wh-questions, too. The ultimate answer will be given when the theory of parameters for various wh-movement constructions is provided in such a way as to allow us to talk about how many parameters are involved in what fashion. These are the theoretical considerations that bear on the treatment of ka-less wh-questions in Old Japanese. Let us turn next to other changes that are found when Old Japanese and Early Middle Japanese are compared.
3.2 Grammatical changes related to the loss of wh-movement In the discussion above of the fact that the word order restriction found in Old Japanese no longer obtains in Early Middle Japanese, the kakari particle ka is not mentioned, unlike ya and zo. The reason is that the nature of ka itself changed drastically. Actually, ya and zo changed their nature, too. Let us start by taking them up one by one, and then move on to matters of verbal inflection. Funagi (1968) observes that wh-questions without a kakari particle increased in the Heian period. He also examines the presence or absence of a clause-final question particle. Here is the data summary that compares Man’yōshū and the first four volumes of Genji. (34) Wh-questions in Man’yōshū and Genji (Funagi 1968) Man’yōshū Genji (first four volumes) a. wh-zo . . . ø 9 1 b. wh-ka . . . ø 117 27 c. wh-ø . . . ø 22 44 d. wh-ø . . . zo –– 8 e. wh-ø . . . zo-ya –– 1 f. wh-ø . . . ya 6 –– g. wh-ø . . . ka 2 ––
15 Loss of wh-movement
363
These figures look conservative, in view of Nomura’s (2001) counting of wh-questions in Man’yōshū, noted above. The overall picture is the same, though. Nomura (2005) gives the numbers for wh-questions with and without ka in the entire 54 volumes of Genji, confirming the less productive use of ka in wh-questions. The less prominent status of ka can also be seen from Omodaka’s (1941) well-known observation that ka gets limited to wh-questions in the Heian period. This means that examples comparable to (2b) are no longer possible. The use of ka in Early Middle Japanese is not only less productive, but also quite different in nature. Isobe (1990) observes that more than half of the wh-questions with ka are rhetorical questions. (33) is an example of rhetorical wh-questions with ka. In Man’yōshū, on the other hand, ka does not produce such effects. According to Omodaka (1941), there are only 12 examples of rhetorical wh-questions with ka (though Nomura (2001) implies that there may be more). The possibility of combining ka with the topic marker as in (33) is also a new feature of Early Middle Japanese. Watanabe (2005, 2006b) proposes that the function of ka in Early Middle Japanese is fairly close to that of the anti-D(iscourse)-linking marker the hell in English, which is studied in detail by Den Dikken and Giannakidou (2002). The crucial observation is that the combination of the hell and a modal auxiliary forces the rhetorical question reading. In contrast to (35a, b), (35c) cannot be interpreted as a genuine information-seeking question. (35) a. b. c.
Who would buy that book? Who the hell bought that book? Who the hell would buy that book?
Interestingly, 70 out of the 85 rhetorical wh-questions in Genji use the modal auxiliary mu, as in (33), according to Isobe (1990). At the same time, the combination of ka and mu does not force the rhetorical question reading, as also noted by Isobe. It is possible that ka has other uses as well. The particle ya is used less productively, too. This assertion may sound surprising, given Omodaka’s (1941) observation that ya took over the function of ka in yes-no questions. Yamada (2005) points out, however, that in the conversational part of the prose writings in the Heian period, the use of ya is excluded for [+participant] subjects, which are either first or second person, and for proper name subjects, with only a couple of exceptions to the generalization. Yamada adds that no such restriction is found in the case of Man’yōshū, as shown by (36). (36) . . . ware ya tiri-na-mu noti ni miyakwo pye yuka-mu? I YA fall-perf-will after loc capital to go-will ‘Will I go to the capital after flowers have fallen to the ground?’ (MYS 20.4435) This means that ya became impossible in some well-defined grammatical environments.
364
Akira Watanabe
Yamada (2010) observes further that the same restriction applies to zo as well (excluding sentences with a predicative nominal, which retained zo even for [+participant] and proper name subjects till later periods). Of course, no such generalization holds in Man’yōshū, as shown by (37). (37) . . . are so kuyasiki. I SO repentant ‘I am repentant.’
(MYS 17.3939)
The limited distribution of the particle zo indicates that it, too, was already beginning to fall into disuse in the Heian period. Miyasaka’s earlier (1952) observation that zo is generally less frequently used than the other kakari particles in the conversational part also points in the same direction. Or maybe we should say that it was disappearing more quickly than the other particles. Future research should address the question why [+participant] and proper name subjects became incompatible with ya and zo. To sum up the discussion so far, we have seen evidence that the Heian period saw the beginning of the decline of the kakari particles ka, ya, and zo. It should also be mentioned that this conclusion has to do with their clause-internal use. As clause-final particles, they were still very active at the time. The decline of these kakari particles can be linked to the loss of wh-movement through the mechanism of wh-movement. Watanabe (2002) proposes that at the time of Old Japanese, ka was the carrier of an uninterpretable focus feature that drives agreement, which is a prerequisite for movement, and that once the movement was lost, the particle had to change its character. Watanabe (2006b), on the other hand, suggests that ka helped delimit the phrasal constituent that undergoes wh-movement. Under these hypotheses, the treatment of other particles remains to be an open question, though their different shapes may be handled just as a matter of morphological variation in combination with the difference in semantic content of the displaced phrase. Though not working on Old Japanese, Cable (2010) puts forth the proposal that when a Q(uestion) particle projects by taking a sister, QP undergoes wh-movement, on the basis of the study of Tlingit, which shows essentially the same pattern as Old Japanese, as briefly mentioned in the discussion of large-scale pied-piping above. Under this analysis, ka heads QP in Old Japanese. One of the goals of Cable’s hypothesis is to accommodate as well the fact that several wh-in-situ languages place a Q-particle at a clause-peripheral position. In these languages, he claims, the Q-particle is simply adjoined to a wh-phrase and undergoes movement without dragging along the rest. This additional connection to wh-in-situ, however, does not receive support from Early Middle Japanese, which leaves ka together with wh-in-situ, as in (33). Treating the Early Middle Japanese particle ka as a simple Q-particle that undergoes covert movement fails to account for its special semantic properties in relation to rhetorical questions. Furthermore, one of the major strategies for wh-questions at the time is to use a particle-less form, as illustrated below.
15 Loss of wh-movement
365
(38) Ikanaru mono no tudowe-ru nara-mu? how person gen gather-stat be-will ‘What kind of people have gathered together?’ (Genji, Yūgao, SNKBZ vol. 1: 135) This example belongs to type (34c) above. Thus, as far as historical stages of Japanese are concerned, we have to conclude that the particles found in the phrasal movement strategy and the particles attached to an in-situ element or found at a clause-peripheral position should not always be identified. Putting aside the connection to wh-in-situ in Cable’s proposal, each of the three ideas about the syntactic function of kakari particles may have its own share of truth. We need further cross-linguistic study to figure out what theoretical account to adopt, but the minimal conclusion is that the theory of wh-movement has to explicate the movement-related role of particles attached to displaced phrases. Once this much is presupposed, it follows that the loss of wh-movement necessarily forces the associated particles to change their nature. Note also that the exact treatment of ka-less wh-questions in Old Japanese becomes relevant again in this connection. Returning to new features of Early Middle Japanese, it is noteworthy that the non-triggered use of adnominal forms (rentai-dome) was on the rise. Yoshida (2001) observes that in the conversational part of the Heian-period prose writings, verbs that unambiguously distinguish between the conclusive and adnominal forms (excluding the copula) end the sentence with the conclusive form in 94 cases and with the adnominal form in 312 cases. Out of the 312 adnominal-form examples, 79 involve the non-triggered use, which means that the adnominal form is used despite the fact that there is no kakari particle or wh-phrase that requires the adnominal clause ending. Toki (2005) also makes a similar observation. The very famous example with an auxiliary in (39) also fits this pattern, since the non-triggered adnominal ending surpasses the conclusive form in number (54 vs. 48) in the case of the auxiliary tu, according to Toki’s counting. (39) Suzume no ko o Inuki ga nigasi-turu. sparrow gen child acc Inuki gen let.go-perf ‘Inuki has let go the baby sparrow.’ (Genji, Wakamurasaki, SNKBZ vol.1: 206) Though opinions have been divided concerning what to make of the phenomenon, it is reasonable to conclude, following Komatsu (1999), that it is a faithful reflection of colloquial speech, given the rather striking statistical data. In other words, one can say that it had become fairly common by that time to use the former adnominal form as the general sentential ending, at least for a certain class of predicates and auxiliaries (see Yoshida 2001 and Toki 2005 for differences among different predicates and auxiliaries). The former adnominal form eventually replaced the former conclusive form completely in Late Middle Japanese, resulting in the collapse of the adnominal-conclusive distinction (Frellesvig 2010, Komatsu 1999, Nomura 2011). The quantitative data indicate that the process was already under way in the Heian period.
366
Akira Watanabe
The phenomenon of rentai-dome should be tied to the incipient decline of the kakari particles ya and zo, observed by Yamada and discussed above. Nomura (2005, 2011) suggests that simple dropping of ya and zo results in the apparent non-triggered use of adnominal forms. Yoshida’s (2001) data indicate that the use of adnominal forms overwhelms that of conclusive forms for a certain class of predicates at the colloquial level, when sentences with a kakari particle are included. In view of the ultimate collapse of the adnominal-conclusive distinction, one may wonder why the former adnominal form was chosen. The decline of kakari particles due to the loss of wh-movement can be a major factor. There is a more subtle way in which the loss of wh-movement brought about rentaidome. Yamada (2010) suggests that in the case of [+participant] or proper name subjects for which zo is ruled out, the genitive case marker ga is used instead to carry phonological prominence. Though he does not give analogous comments in the case of ya, probably the same account is applicable, in view of the fact that case particles can carry stress for focusing purposes in Modern Japanese, as pointed out by Vermeulen (2012). (40) is such a Modern Japanese example, where capital letters indicate stress. (40) Sore GA matigai. that nom mistake ‘THAT is a mistake.’ We can further conjecture that the function of the remaining use of zo and ya in Early Middle Japanese is to carry stress. The same can probably be said about ka as well, in view of the stress-bearing nature of the hell. As Yamada (2005, 2010) notes, [+participant] and proper name subjects are the major associates of the genitive particle ga in Old Japanese (Nomura 1993). The replacement of zo and ya with ga in the case of [+participant] and proper name subjects, therefore, can be regarded as a plausible course of events when the kakari particles changed their nature and started to decline. In a nutshell, one can say that these changes are all keyed to the in-situ strategy for focusing and wh-questions. And at this point, the restricted replacement of ya and zo with ga can be connected to rentai-dome, because clauses that end with an adnominal form is one of the major contexts that license the genitive marker ga in Old Japanese (Nomura 1993). In order to make the replacement possible, an adnominal form must be used. Now, we should recall that so/zo can also be used as a contrastive topic marker in Old Japanese. For this use, replacement with ga is probably not an option. More generally, the word order restriction concerning topicalized expressions is not taken up in (32). The loss of wh-movement discussed up to this point is limited to raising to Spec of FocP. Though the increase of subject topicalization is mentioned as the cause of the loss of wh-movement above, the syntax of topicalization itself in Early Middle Japanese remains to be investigated in detail. I would like to just mention that the Heian data provided by Oda (1989) and Shigemi (1994) indicate that topicalized phrases are further
15 Loss of wh-movement
367
away from the predicate than phrases marked with ka, ya, zo, namu, or koso, which suggests that Spec of TopP still hosts topics in Early Middle Japanese. Relevant also in this connection is the fate of koso, which is still used in a marginal way in Modern Japanese. If it continued to be a topic marker, its survival may be approached in comparison with wa, the descendant of pa (see Frellesvig 2010 for the relevant phonological change). On the basis of verse texts in Early Middle Japanese, however, Ōno (1993) claims that koso ceased to be a topic marker. If that is true, it is not so obvious why it survived. The historical account of koso needs a separate treatment.
3.3 Social background We have seen that the increased rate of subject topicalization spawned a cascade of grammatical changes, including most importantly the loss of wh-movement. The more frequent use of subject topicalization itself is not a grammatical change. It has to do with verbal performance. One may then wonder what led to such a change in people’s behavior. I would like to end this chapter with short notes on the social background. The ninth century was not a period of social stability. Sakaue (2001) points out the discontinuity of local communities during the period, attributing the rather drastic change at least partly to climatic factors. But recent interdisciplinary work, reported in Hotate (2012), has revealed that more destructive events took place. Throughout the century, Japan was troubled with devastating earthquakes (together with tsunami) and dreadful volcanic activities, which are chronicled by Hotate. One may recall how Mt. Fuji is described in The Tale of Taketori, where it appears as an active volcano. To illustrate what can happen, let me pick just one concrete example. Sakaue describes a huge flood at the Chikuma River in 888, which did irrecoverable damage to villages along the river. Sakaue wonders why the damage was so severe at that time, because flooding was not uncommon in the region. Hotate reports that a recent study made it clear that the flood in 888 was due to the collapse of a dam created by landslides caused by a very strong earthquake in the previous year, with an epicenter off the coast of the Pacific. The direct damage of the earthquake is easy to imagine. But a single big earthquake can have manifold destructive effects in various places and at numerous social levels, as a chain of events. If there are a series of them, people panic. Of course, there is no logical necessity that social unrest of that sort should cause people to use more subject topicalization. Nevertheless, a change in verbal behavior is not surprising in such a situation.
4 Conclusion Whatever is the ultimate cause of a series of changes that took place in the transition from Old Japanese to Early Middle Japanese, loss of wh-movement is a major element
368
Akira Watanabe
in these changes, resulting in the in-situ strategy for wh-questions and focusing more generally. Concomitantly, the particles ka, ya, and zo diminished their roles in the focus system of Early Middle Japanese, which suggests that their status was closely tied to wh-movement in Old Japanese.
References Aldridge, Edith. 2010. Short wh-movement in Old Japanese. Japanese/Korean Linguistics 17. 549–563. Büring, Daniel. 2003. On D-trees, beans, and B-accents. Linguistics and Philosophy 26. 511–545. Cable, Seth. 2010. Against the existence of pied-piping: Evidence from Tlingit. Linguistic Inquiry 41. 563–594. Chomsky, Noam. 1977. On wh-movement. In Peter Culicover, Thomas Wasow and Adrian Akmajian (eds.), Formal syntax, 71–132. New York: Academic Press. Chung, Sandra. 1982. Unbounded dependencies in Chamorro grammar. Linguistic Inquiry 13. 39–77. Den Dikken, Marcel and Anastasia Giannakidou. 2002. From hell to polarity: “Aggressively non-D-linked” wh-phrases as polarity items. Linguistic Inquiry 33. 31–61. Fanselow, Gisbert and Denisa Lenertová. 2011. Left peripheral focus: Mismatches between syntax and information structure. Natural Language and Linguistic Theory 29. 169–209. Frellesvig, Bjarke. 2010. A history of the Japanese language. Cambridge: Cambridge University Press. Funagi, Shuntaro. 1968. Heian-jidai kanbun-kundoku gimonshi-gimonbun no ichi-bunkei [One form of wh-questions in the kanbun-kundoku style during the Heian period]. Kokubungaku: Gengo to Bungei 58. 17–26. Grohmann, Kleanthes K. 2003. Prolific domains. Amsterdam: John Benjamins. Hendriks, Petra. 2004. Coherence relations, ellipsis and contrastive topics. Journal of Semantics 21. 133–153. Heycock, Caroline. 2008. Japanese wa-, ga-, and information structure. In Shigeru Miyagawa and Mamoru Saito (eds.), The Oxford handbook of Japanese linguistics, 54–83. Oxford: Oxford University Press. Hotate, Michihisa. 2012. Rekishi no naka no daichi dōran [Tumultuous movements of land in the history]. Tokyo: Iwanami Shoten. Ishida, Haruaki. 1939. Kosokere keishiki no hongi [The true meaning of the kosokere form]. Kokugo to Kokubun 16(2). 66–82 & 16(3). 68–82. Isobe, Yoshihiro. 1990. Chūko wabun no yōsetsumei-gimon hyōgen [Wh-expressions in the Middle Japanese native style]. Nihon-Bungaku Kenkyū 26. 165–176. Yamaguchi: Baiko Jogakuin College. Jayaseelan, K. A. 2001. IP-internal topic and focus phrases. Studia Linguistica 55. 39–75. Komatsu, Hideo. 1999. Nihongo wa naze henka suru ka [Why does Japanese change?]. Tokyo: Kasama Shoin. Konoshima, Masatoshi. 1966. Kokugojoshi no kenkyū [A study on Japanese particles]. Tokyo: Ōfūsha. Krifka, Manfred and Renate Musan. 2012. Information structure: Overview and linguistic issues. In Manfred Krifka and Renate Musan (eds.), The expression of information structure, 1–43. Berlin: De Gruyter Mouton. Kuroda, S.-Y. 2007. On the syntax of Old Japanese. In Bjarke Frellesvig, Masayoshi Shibatani and John Charles Smith (eds.), Current issues in the history and structure of Japanese, 263–317. Tokyo: Kurosio Publishers. Lambova, Mariana. 2001. On A-bar movements in Bulgarian and their interaction. The Linguistic Review 18. 327–374. Miyasaka, Kazue. 1952. Kakari-musubi no hyōgenkachi [Expressive values of kakari-musubi]. Kokugo to Kokubungaku 29(2). 41–51. Nomura, Takashi. 1993. Jōdaigo no no to ga ni tsuite [On no and ga in Old Japanese]. Kokugo-Kokubun 62(2). 1–17 & 62(3). 30–49.
15 Loss of wh-movement
369
Nomura, Takashi. 1996. Ga–shūshikei e [Towards the co-occurrence of ga and the conclusive form]. KokugoKokubun 65(5). 524–541. Nomura, Takashi. 2001. Ya ni yoru kakari-musubi no tenkai [the development of kakari-musubi by ya]. Kokugo-Kokubun 70(1). 1–34. Nomura, Takashi. 2005. Chūko kakari-musubi no hen’yō [Changing aspects of kakari-musubi in Middle Japanese]. Kokugo to Kokubungaku 82(11). 36–46. Nomura, Takashi. 2011. Hanashikotoba no nihonshi [The history of colloquial Japanese]. Tokyo: Yoshikawakōbunkan. Oda, Masaru. 1989. Shutsugen-ichi kara mita kakarijoshi zo [The kakari particle zo from the viewpoint of where it appears]. Kokugogaku 159. 15–24. Ōno, Susumu. 1993. Kakari-musubi no kenkyū [A study on kakari-musubi]. Tokyo: Iwanami Shoten. Ōshika, Tadahisa. 1991. Man’yōshū ni okeru futeigo to futei no gimon [Indeterminate words and wh-questions in Man’yōshū]. Kokugogaku 165. 53–66. Okazaki, Masatsugu. 1996. Kokugo joshi ronkō [Studies on Japanese particles]. Tokyo: Ōfū. Omodaka, Hisataka. 1941. Man’yō no sakuhin to jidai [Works and the period of Man’yō]. Tokyo: Iwanami Shoten. Puskás, Genoveva. 2000. Word order in Hungarian. Amsterdam: John Benjamins. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane Haegeman (ed.), Elements of grammar, 281–337. Dordrecht: Kluwer. Roberts, Craige. 2011. Topics. In Klaus von Heusinger, Claudia Maienborn and Paul Portner (eds.), Semantics, vol. 2. 1908–1934. Berlin: De Gruyter Mouton. Rooth, Mats. 1992. A theory of focus interpretation. Natural Language Semantics 1. 75–116. Saeki, Umetomo. 1938. Man’yōgo kenkyū [A study on the language of Man’yō]. Tokyo: Bungakusha. Sakakura, Atsuyoshi. 1993. Nihongohyōgen no nagare [The transition of Japanese expressions]. Tokyo: Iwanami Shoten. Sakaue, Yasutoshi. 2001. Ritsuryō kokka no tenkan to Nihon [A change in course of government by Chinese-style laws and “Japan”]. Tokyo: Kōdansha. Sasaki, Takashi. 1992. Jōdaigo ni okeru –ka–ha– no kōbun [Constructions with the ka–ha order in Old Japanese]. Kokugo-Kokubun 61(5). 17–33. Serafim, Leon A. and Rumiko Shinzato. 2011. On the origins of the Old Japanese kakari particles ka, zö, and kösö, and their Okinawan counterparts. Japanese/Korean Linguistics 19. 449–464. Shigemi, Kazuyuki. 1994. Joshi no kōbunkinō kenkyū [A study on the constructional function of particles]. Osaka: Izumi Shoin. SNKBZ (= Shin Nihon Koten Bungaku Zenshū). 1994–1995. Genji monogatari [The tale of Genji], vols. 1–2. Tokyo: Shōgakkan. Toki, Rumie. 2005. Heian wabunkaiwabun ni okeru rentaikeishūshibun [Sentences with the adnominal ending in native-style conversational sentences during the Heian period]. Nihongo no Kenkyū 1(4). 16–31. Tonoike, Shigeo. 2002. Jōdainihongo ni sahō wh-idō wa atta ka? [Did Old Japanese have leftward wh-movement?] Gekkan Gengo 31(3). 86–91. Vallduví, Enric and Maria Vilkuna. 1998. On rheme and kontrast. In Peter W. Culicover and Louise McNally (eds.), The limits of syntax, 79–108. New York: Academic Press. Vermeulen, Reiko. 2012. The information structure of Japanese. In Manfred Krifka and Renate Musan (eds.), The expression of information structure, 187–216. Berlin: De Gruyter Mouton. Wagner, Michael. 2012. Contrastive topics decomposed. Semantics and Pragmatics 5(8). 1–54. Watanabe, Akira. 1996. Case absorption and wh-agreement. Dordrecht: Kluwer. Watanabe, Akira. 2002. Loss of overt wh-movement in Old Japanese. In David Lightfoot (ed.), Syntactic effects of morphological change, 179–195. Oxford: Oxford University Press.
370
Akira Watanabe
Watanabe, Akira. 2004. Parametrization of quantificational determiners and head-internal relatives. Language and Linguistics 5. 59–97. Watanabe, Akira. 2005. Minimarisuto puroguramu josetsu [Introduction to the Minimalist Program]. Tokyo: Taishukan. Watanabe, Akira. 2006a. Setting the wh-movement parameter. In Adriana Belletti, Elisa Bennati, Cristiano Chesi, Elisa Di Domenico and Ida Ferrari (eds.), Language acquisition and development: Proceedings of GALA 2005, 550–563. Cambridge: Cambridge Scholars Press. Watanabe, Akira. 2006b. The pied-piper feature. In Lisa Cheng and Norbert Corver (eds.), Wh-movement on the move, 47–70. Cambridge, MA: MIT Press. Watanabe, Akira. 2007. Topic-focus articulation in Old Japanese: So/zo and koso. Proceedings of the Workshop on Altaic Formal Linguistics II, 121–137. Cambridge, MA: MITWPL. Yamada, Masahiro. 2005. Gimonhyōgen ni okeru shukakuhyōji ga kakudai no yōsō [Aspects of the expansion of the nominative marker ga in interrogatives]. Kokugo to Kokubungaku 82(11). 57–69. Yamada, Masahiro. 2010. Kakujoshi ga no tūjiteki kenkyū [A diachronic study of the case particle ga]. Tokyo: Hituzi Syobo. Yoshida, Shigeaki. 2001. Bunmatsuyōgen no katsuyōkei ni tsuite [On inflectional forms of sentence-final predicates]. Yamanobe no Michi 45. 1–16.
Tomohide Kinuhata
16 Development of adverbial particles 1 Introduction In this chapter, we discuss the development of adverbial particles. The meanings of those particles are expressed by adverbs such as only and even as in the English translations of (1); thus, we call those particles ‘adverbial’. (1) a.
b.
Taroo wa Hanako dake o Taro top Hanako only acc ‘Taro only loves Hanako.’ Taroo wa Hanako ni sae Taro top Hanako dat even ‘Taro is kind even to Hanako.’
aisiteiru. love yasasii. kind
Particles such as dake and sae in (1) were named fuku joshi by the Japanese grammarian YAMADA Yoshio (Yamada 1908). ‘Adverbial particles’ can also be regarded as a translation of this term: Fuku means ‘adverbial’, and joshi corresponds to particles. The following sections are organized as follows. We will see the development of particles that have exclusive meanings, such as only in Section 2, and the change of scaleinducing particles, such as even in Section 3. We look at exclusive and scale-inducing particles because those meanings are universally attested across languages. Section 4 discusses the development of adverbial particles from syntactic viewpoints: The change from nouns to clitics in 4.1 and from adjoined-clausal markers to adverbial particles in 4.2. To illustrate the latter’s change, I will refer to adverbial particles other than exclusive and scale-inducing ones. By examining those changes, I hope to present general patterns of syntactic and semantic changes that contribute to understanding language change in general.
Acknowledgments: The first draft of this chapter was written in February 2013. The research made public after then is not included in the reference of this paper due to time/space constraints. In revising the draft, I benefited from the comments from the reviewers of this volume. The remaining errors are my own. https://doi.org/10.1515/9781614512851-017
372
Tomohide Kinuhata
2 Development of exclusive particles 2.1 Dake and bakari in present-day ModJ (PModJ) In present-day ModJ (PModJ), we have two types of exclusive particles, i.e. dake and bakari. At first sight, they seem to function similarly as illustrated in (2). (2) a.
The use of bakari, however, must entail multiple occurrences of an event, whereas the use of dake does not (Kikuchi 1986, Sadanobu 2001). The sentence in (2a) can be used when only one exam is under debate, but the sentence in (2b) cannot be used in that situation. For example, when the interlocutors speak about the result of one particular entrance exam of Oxford University, the speaker can use dake as in (2a) but not use the sentence in (2b). For sentence (2b) to be felicitous, the interlocutors must talk about multiple entrance exams, e.g., the entrance exams to Oxford University, Harvard University, and Yale University. The sentence (2b) means that in each entrance exam, only Taro passed. The following pairs also support the entailment of multiple events by bakari. In (3) and (4), while dake is appropriate, bakari is not. (3) a.
b.
(4) a.
b.
Taroo dake kasikoi. Taro dake smart ‘Only Taro is smart.’ #Taroo bakari kasikoi. Taro bakari smart. ‘Only Taro is smart.’ Taroo dake syussinti ga tigau. Taro dake hometown nom different ‘Only Taro comes from a different town.’ #Taroo bakari syussinti ga tigau. Taro bakari hometown nom different ‘Only Taro comes from a different town.’
Whether one is smart or not is judged by the speaker’s knowledge and is not judged multiple times. Therefore, bakari with predicates like ‘be smart’ in (3b) is inappropriate.
16 Development of adverbial particles
373
The same explanation can be adapted to the inappropriateness of bakari in (4b): Predicates such as ‘being different’ cannot occur multiple times, and thus bakari is not compatible with those predicates.1 On the other hand, dake is natural in those examples since it just separates the individual referred to from others according to the predicate used in the sentence, i.e., John is smart, but others are not in (3a).2 In this subsection, we have seen that two types of exclusive particles are used in PModJ: One is an exclusive that entails the multiplicity of events, i.e., bakari, and the other is an exclusive that does not, i.e. dake. Let us call the latter A-type and the former B-type. In the following subsections, we will see how those meanings are encoded in the process of while developing exclusive meanings.
2.2 Nomi and bakari from Old to Early Middle Japanese In Old Japanese (OJ), A-type and B-type exclusive meanings are expressed by one particle nomi.3 (5a) does not imply multiple occurrences of ‘hearing the ducks squawk’ but indicates that he hears it only today and not after tomorrow. Therefore, it is appropriate to construe the nomi in (5a) as an A-type exclusive. In (5b), on the other, the phrase miru-gotoni ‘whenever I see it’ imposes the interpretation of the event ‘I feel like crying’ occurring again and again, which enables us to consider the nomi in (5b) as a B-type exclusive. (5)
a.
b.
Ike ni naku kamo wo kyepu nomwi mite ya pond loc squawk duck acc today nomi see q kumogakuri-na-mu. die-perf-conjec ‘Do I hear the ducks squawk only today, and am I going to die?’ (MYS 3.416) Asa-kumo ni tadu pa midare . . . miru-gotoni morning-cloud loc crane top fly.around . . . see-whenever ne nomwi si nak-ayu. sound nomi emph croak-spon ‘Cranes fly around between clouds in the morning . . . every time I see it, I always feel like crying loudly.’ (MYS 3.324)
1 When the speaker judges the smartness of John several times, the sentence in (3b) can be felicitously uttered. 2 The difference between dake and bakari resembles the difference between only and always in English: Always in ‘Always, John passed the exam’ can have an exclusive interpretation only if it involves multiple exams, and ‘Always, John comes from a different town’ is deviant. 3 In the following, I refer to the particles under discussion by the kunreishiki notation of ModJ despite their different phonological appearances as given in the examples of each period.
374
Tomohide Kinuhata
Although bakari is used in OJ, it denotes the meaning of ‘degree’, mostly used with demonstratives as in (6), and does not have an exclusive meaning (Konoshima 1966, Koyanagi 2003). (6)
Kaku bakari kwopwi-mu to kanete sira-mase-ba, imo like.this bakari love-conjec quot ahead know-irr-cond beloved woba mi-zu so aru-beku ari-kyeru. acc.top see-neg foc exist-should cop-pst ‘If I knew beforehand that I would love you to this extent, I would not have seen you in the first place.’ (MYS 15.3739)
The exclusive meaning of bakari can be found in examples from texts in Early Middle Japanese (EMJ). However, the exclusive meaning bakari acquires is not B-type but A-type, expressed by dake in PModJ. As the result of bakari’s acquisition of the A-type exclusive meaning, nomi is used only as B-type exclusives in EMJ. Table 1 shows the distribution of exclusive particles and ‘degree’ particles in OJ and EMJ. Table 1: Degree particles in OJ and EMJ.
‘degree’ ‘exclusive’
A-type B-type
OJ
EMJ
nomi/bakari
bakari
nomi nomi
bakari nomi
The evidence for the difference in EMJ comes from the contexts in which those particles are used: Quite a few instances of nomi are used with phrases that show the multiplicity of events. In (7), this multiplicity is indicated by the phrase asa-yuu-no miyadukawe ‘court services in the morning and the evening’, which gives rise to the interpretation that the court attendant of Lady Kiritsubo has disturbed other mistresses, not just once. Therefore, it is natural to construe the exclusiveness conveyed by nomi in this example as B-type. (7) Asa-yuu no miyadukawe ni tukete mo pito morning-evening gen court.service dat about also people no kokoro o nomi ugokasi, gen heart acc nomi disturb ‘She always disturbs other mistresses when she serves the Emperor, whether in the morning or evening.’ (Genji (1), p. 17) On the other hand, the exclusive use of bakari rarely appears with phrases denoting multiple events but is often used in describing a person or a scene instantaneously.
16 Development of adverbial particles
375
The following example is from the scene in which Myōbu visits the place where Lady Kiritsubo is. The scene is described based on the first impression from the viewpoint of Myōbu, and thus the event does not repeatedly happen, which favors the A-type interpretation of the relevant exclusive particle. (8) Kusa mo takaku nari, . . . tukikage bakari zo, weed also high become . . . moonlight bakari foc yawemugura ni mo sawara-zu sasi’iri-taru grown.weed dat also be.blocked-neg enter-cont ‘The weeds grow high so that only the moonlight comes into the garden without being blocked by the tangled growth of wild grass.’ (Genji (1), p. 27) The distinction between nomi and bakari in EMJ has been paid much attention in historical studies of Japanese (Konoshima 1966, Koyanagi 1997), and the debate started from the following comment by FUJITANI Nariakira (1778): siroki kinu o nomi kitari ‘wearing white clothes’ means that he always wears white clothes or that many people all wear white clothes. . . siroki kinu bakari o kitari means that, among white, red, green clothes, etc., he only wears the white one without wearing other colors. (Fujitani 1778: 240, Translation by author)
This description of the difference between nomi and bakari in EMJ nicely corresponds to our characterization that nomi is used as a B-type exclusive and bakari as an A-type exclusive. Nomi, as a B-type exclusive, entails the multiple events of wearing, giving rise to interpretations such as ‘he always wears. . .’ or ‘many people all wear . . .’. As an A-type exclusive, bakari does not imply the multiplicity of events, just conveying that he does not wear red and green but white ones. In the next subsection, we will see how the distribution of nomi and bakari in EMJ is affected by the appearance of another exclusive particle dake and discuss a general trend of semantic change of exclusives.
2.3 Development of dake and the path of semantic change The particle nomi had been out of use during Late Middle Japanese (LMJ). The examples of nomi are only found in texts written in Classical Japanese (Yuzawa 1958, Konoshima 1966), a fossilized register based on the grammar of EMJ. As a result, bakari expanded its use to B-type exclusives during LMJ. The following examples are taken from Miyachi (2003). Itumo ‘always’ in (9a) and ikutabi . . . -temo ‘no matter how often’ in (9b) induce the interpretation where the following events take place many times.
376
Tomohide Kinuhata
(9) a.
Itumo konata no koto bakari moosi’idas-aruru koto degozaru. always 2sg gen thing bakari say-hon thing cop.pol ‘He always talks about you.’ (Toraakira (Jō), p. 407) Ikutabi otazune-nasare-temo tada yume zya to bakari how.often ask-hon-conc just dream cop quot bakari yuute iru. say cont ‘No matter how often the magistrate asks, he always says it is a dream.’ (Kyūōdōwa, p. 74)
b.
These examples show that LMJ and Early ModJ (EModJ) bakari were used as both A and B-type exclusives. It is not until Modern Japanese that dake takes charge of the labor of exclusives (Terada 2000, Miyachi 2003). In LMJ and EModJ, dake without rendaku voicing, i.e., /take/, had the meaning of ‘degree’ as nomi and bakari had. (10) Bumo ni waga tikara no oyobu take asa parents dat 1sg.gen power gen reach dake morning wa toku oki, top early wake.up ‘To such an extent that my father and mother appreciate my effort, I try to wake up early in the morning’ (Rongoshō: gakuji, p. 8) According to Terada (2000), dake acquired exclusive use in the late 30s of the Meiji period, i.e., around 1900. The short observation so far allows us to reformulate the distributions of the forms for exclusives, as in Table 2. Table 2: Distribution of exclusives.
‘degree’ ‘exclusive’
A-type B-type
OJ
EMJ
LMJ/EModJ
PModJ
nomi/bakari
bakari
bakari/dake
bakari/dake
nomi nomi
bakari nomi
bakari bakari
dake bakari
This table shows that the particles employed as exclusives follow the following steps. (11) ‘degree’ > A-type exclusive > B-type exclusive This sequence is not unreasonable since (a) the change can be regarded as a case of grammaticalization from nouns to clitics, as we will see in Section 4.1, and (b) B-type
16 Development of adverbial particles
377
exclusives are more complex in meaning than A-type ones: B-type involves a plurality of events in addition to the exclusive meaning in each event. (2b), for instance, is interpreted as ‘For each entrance exam, only John passed.’ Since the plurality is not deduced from the meaning of ‘degree,’ it is natural to construe it as added to the meaning of exclusives in the later stage of development. Hence, it is concluded that (11) can be a general path of semantic change of exclusives. In this section, we have seen the history of exclusive particles and pointed out that exclusives arise from nouns expressing ‘degree’, and the A-type precedes the B-type in their development.
3 Development of scale-inducing particles 3.1 Dani, sura, and sae in Old Japanese As we noted at the end of the last section, the development of exclusive particles involves grammaticalization from nouns to clitics. However, general process of language change is not observable in the development of scale-inducing particles (henceforth, ‘scale particles’ for short): In the history of Japanese, the particles expressing ‘even’ evolved from other particles. In order to see the semantic change between those particles, this subsection examines the meanings of the particles of OJ from which scale particles originated. In OJ, dani expresses the meaning of ‘at least’ with predicates with desiderative meanings (Kano 1938a). Dani is used with a volitional form of a predicate in (12a) and is embedded in the conditional antecedent in (12b). Those instances of dani are hard to be translated by ‘even’ in English. (12) a.
b.
Ime ni dani miye-mu to ware pa podoke-domo, dream loc dani see-vol quot 1sg top unlace-conc ‘Although I unlace my obi, expecting that I will see you at least in a dream,’ (MYS 4.772) Sakurabana pitome dani kimi to si cherry.blossom one.sight dani 2sg com emph mi-te-ba are kwopwi-me yamo. see-perf-cond 1sg love-conjec q ‘If I could see cherry blossoms with you at least once, I would not be madly in love with you.’ (MYS 17.3970)
When dani is translated by English ‘even’, it always accompanies negation in its predicate as in (13).
378
Tomohide Kinuhata
(13) Ime ni dani mi-zari-si monowo opoposiku miyade dream loc dani see-neg-pst spf depressingly go.to.court mo suru ka. also do q ‘I depressingly go out to the court, though I have never thought it even in a dream.’ (MYS 2.175) Even in this example, dani can mean ‘at least’ with the negation scoping over it: ‘It is not the case that I thought it at least in a dream.’ Sae (< OJ sape), a scale-inducing particle in PModJ, was not a scale particle in OJ. It encoded the meaning of ‘in addition to’ without the scale entailment, presumably originating in the infinitive form of the verb soeru (< sope-) ‘to add (tr.)’. For example, in (14), there is no scalar relation between ‘the day before yesterday’, ‘yesterday’, ‘today’, and ‘tomorrow’ with respect to the speaker’s desire to see his girlfriend: The speaker just wants to see her every day. (14) Wototupi mo kinopu mo kyepu mo the.day.before.yesterday also yesterday also today also mi-ture-domo asu sape mi-m-aku posiki kimi kamo see-perf-conc tomorrow sae see-vol-nmlz want 2sg excl ‘I have seen you the day before yesterday, yesterday, and today. But you are the person that I love to see tomorrow too.’ (MYS 6.1014) Sura is the most plausible candidate for the scale-inducing particle in OJ. In (15), sura scopes over the situation that makes the speaker madly in love, i.e. the speaker only met his girlfriend just in a dream. It is imagined from the sentence with sura what the speaker would be if he had met his girlfriend in real life. (15) Ime ni nomwi mite sura kokoda kwopuru a pa ututu dream loc only see sura excessively love 1sg top reality ni mi-te-ba masite ikani ara-mu. loc see-perf-cond much.more how cop-conjec ‘I am mad about her even when seeing her in a dream. What if I had seen her in reality?’ (MYS 11.2553) Even though it is widely acknowledged that sura is a scale particle in OJ, Konoshima (1966) advocated another interpretation for sura: exhibiting a discordance between the propositions preceding and following it. The sura in (16), according to Konoshima (1966), is not used to infer other persons’ situations but to show the inconsistency of the speaker’s ability to fly freely in the sky with the situation that he walked through the Milky Way with pain.
16 Development of adverbial particles
379
(16) Opozwora yu kaywopu ware sura na ga yuwe-ni sky abl pass 1sg sura 2sg gen reason-adv ama-no kapa-di wo nadumite zo ko-si. sky-gen river-road acc with.pains foc come-pst ‘I walked through the Milky Way with pain to meet you, even though I can freely fly in the sky.’ (MYS 10.2001) Konoshima (1966) maintains that showing the discordant situation is the original meaning of sura, and the scale interpretation was derived from it. In this subsection, we have examined the uses of dani, sae, and sura in OJ. Dani and sae were not scale particles in the same way as English even in OJ. Sura had scale meanings in it, but it might originally have had another meaning, showing a dichotomy between the relevant situations. We will see in the next subsection how dani and sae evolve into scale particles.
3.2 Development of scale-inducing meaning in dani and sae As we saw in (13), dani in OJ is interpreted to convey scale-inducing meanings like even only in negative sentences. In EMJ, however, dani is used as a scale particle not only in negative contexts but also in affirmative contexts, as shown in (17) (Kano 1938a, Kinuhata 2005). (17) Katao naru o dani menoto yau no omoubeki pito wa defective cop acc dani wetnurse like cop thoughtful person top asamasiu mao ni minasu mono’o. amazingly perfect dat consider sfp ‘A thoughtful person like a wetnurse amazedly considers the child perfect even if he is defective.’ (Genji (1), p. 138) In the Kamakura period (12th–14th century), dani was solely used as a scale particle, losing its use with desiderative predicates like (12) (Kano 1938b). This loss means that dani completes the semantic change from ‘at least’ to ‘even’. A key to explaining this change is that dani was always interpreted as ‘even’ in a sentence with negation. It should be noted here that the semantics of dani is viewed differently relative to the scope of negation. For example, dani in (13) means ‘at least’ with the scope of negation interpreted as wide, and ‘even’ with the scope of negation narrow: ‘It is not the case that I thought it at least in a dream’ and ‘Even in a dream, I didn’t see it’ almost denotes the same situation. Therefore, it is reasonable to conclude that the meaning of dani transits from ‘at least’ to ‘even’ by reanalyzing the scope of negation relative to the particle (Kinuhata 2005).
380
Tomohide Kinuhata
In the Muromachi period (14th–16th century), sae acquired the scale-inducing use, and, as a result, dani had been wiped out through the Edo period (17th–19th century), i.e., in EModJ (Yuzawa 1958, Konoshima 1966). (18) a.
b.
Madoromu koto mo nara-ne-ba yume ni sae doze nmlz also fulfill-neg-because dream loc sae mo sono omokage o mo mi-nanda. also that figure acc also see-neg.pst ‘Since he could not doze off, he did not see the (priest’s) figure even in his dream.’ (Amakusa Heike, p. 85) Fotoke ni sae yudan-sase-zu semetukau waro zya mono. Buddha dat sae relax-caus-neg exploit guy cop sfp Syuzyoo o semeru wa doori zya. human acc exploit top reasonable cop ‘Since she even exploits the Buddha without hesitation, it is natural for her to exploit human beings.’ (Satsumauta, p. 298)
The growth of the meaning ‘even’ from ‘in addition to’ is more straightforwardly accounted for than the semantic change of dani: Sae just added the scale connotation to its original meaning. This hypothesis is reinforced by the fact that also, or a Japanese counterpart mo, functions in the same way as even when the context supplies a scale implicature: Compare yume-ni-mo mi-nai ‘I don’t see it also in my dream’ with yumeni-sae mi-nai ‘I don’t see it even in my dream.’ Therefore, ‘pragmatic strengthening’ is enough to make the relevant change possible. It is not straightforward to find a general trend in the development of scale particles in Japanese. No grammaticalization process or unidirectional shift like that of A-type to B-type exclusives has been observed in the change. Nevertheless, it still deepens our understanding of how reanalysis and pragmatic strengthening change the meaning of particles.
4 Syntactic change of adverbial particles We have seen so far the historical developments of exclusive and scale particles. In those surveys, we have focused on the semantic aspect of those changes, e.g., with or without the indication of multiple events, from ‘at least’ to ‘even’, etc. We now turn our attention to the syntactic change of adverbial particles. Since syntactic positions are less diverse than meanings, it is more plausible to discover some regularities of language change.
16 Development of adverbial particles
381
4.1 From nouns to clitics In concluding Section 2, we considered the following process observed in developing exclusive particles. (19) ‘degree’ > A-type exclusive > B-type exclusive (19) not only designates the change of semantics but also presupposes the change of syntax, i.e., the change from nouns to clitics. Let us take dake as an example of this syntactic change. Dake was used as an independent noun in EMJ and LMJ, having the meaning of ‘height’ or ‘length.’ Example (20) illustrates this: In (20a), dake, without rendaku voicing, is used independently and in (20b) a noun modifies it with a genitive particle. Therefore, dake in those examples is not part of a larger noun but can function as a noun phrase (NP). (20) a.
b.
Iukei to iu akusoo ari. Take siti-syaku Yuukei quot say armed.monk exist dake seven-feet bakari ari-keru ga, about have-pst conc ‘There is an armed monk named Yuukei. His height is about 7 feet tall.’ (Kakuichi Heike (1), p. 103) Witutu ni kake-si maro ga take well’s.pipe dat measure-pst 1sg gen dake sugi-ni-kye-rasi na imo mi-zaru ma-ni pass-perf-pst-conjec sfp 2sg see-neg while-adv ‘While not seeing my sweetheart, my height had topped the length of the well’s pipe I used to measure myself.’ (Ise, p. 136)
In expressing the exclusive meanings, however, dake cannot be used as an independent noun: It cannot be modified by other nouns with genitive particles but must be a part of a complex noun with the initial syllable voiced (via rendaku) as in (21a). The example in (21b) from EMJ indicates that the compounding of dake with the preceding noun occurred before it acquired the use of exclusives: Kubi-dake means ‘(up to) the height of the neck.’ (cf. Miyachi 2010) (21) a.
Taroo wa Hanako(✶no)-dake o aisiteiru. Taroo top Hanako(gen)-dake acc love ‘Taro only loves Hanako.’
382
Tomohide Kinuhata
b.
Korya sui-me ore ga kubi-dake nazunde iru voc refined-hum 1sg gen neck-dake loving be ‘Hey you, refined one. I am deeply in love with you.’ (Meido no Hikyaku, p. 114)
While the voiced variant can be modified by a verb in the same way as unvoiced ones, as in (22), it must be bound by a preceding element, i.e., it cannot be a free morpheme. (22) Saa kore made zo. Nogaruru-dake to sigeru asima excl this until sfp escape.adn-dake quot busy between.reeds o kakiwakete, acc push ‘That’s it! We have pushed our way through the reeds as far as we can.’ (Kokuseiyakassen, p. 270) Let us call the dake exhibiting this property ‘affixal noun,’ emphasizing the structure of (21), in which dake is a subpart of a larger compound. But this is not the end of the syntactic change of dake: It evolves into a clitic. Clitics are bound forms but syntactically freer than affixal nouns that constitute a larger noun. Clitic dake can reside outside of the noun, appearing after case markers as in (23a), and can attach to the verb phrase as in (23b). On the other hand, dake in (23b) appears in between the serial verb construction, whereas an adnominal form of the verb modifies it in (22). (23) a.
b.
Kono [N seito] ni dake wa oyoba-nai. this student dat dake top reach-neg ‘I am inferior only to this student.’ Noma-zuni tabete dake i-tat-te, drink-neg eat dake cont-pst-ger ‘If you were only eating, without drinking,’
(Asunaro monogatari)
(Tarō monogatari)
The above examination leads us to conclude that the following process is involved in the historical change of dake. (24) noun > affixal noun > clitic Bakari, another exclusive particle discussed in Section 2, also seems to follow the path in (24). Bakari is assumed to originate in the nominal form of (OJ) pakaru ‘to measure’. Thus, it was an autonomous word with nominal status. When used as an exclusive particle in EMJ, it suffixes to another noun as in (25): Bakari with the initial syllable voiced attaches to the noun not intervened by particles, the whole unit of which is a compounded noun with a case marker attached.
16 Development of adverbial particles
383
(25) [N Koremitu-bakari] o uma ni nosete owasi-nu Koremitu-bakari acc horse dat ride come.hon-perf ‘He came here only taking Koremitsu with his horse.’ (Genji (1), p. 253) Thus, bakari, in this use, is not syntactically free, only attaching to nominals. Examples in which bakari follows case markers first appear in texts written around the Edo period, i.e., EModJ, which means that bakari in this period does not constitute the internal structure of a noun but is syntactically free as a clitic. (26) [N Katasumi] ni bakari i-zuto tito koko e kina. corner loc bakari sit-neg little here all come.imp ‘Come here a little, not always sitting on the corner.’ (Orandakagami, p. 89) Thus, we witness the path in (24) in the historical change of bakari. Other examples include hodo and kurai. Hodo originally meant ‘a while’ or ‘distance,’ functioning as an autonomous noun, as in podo pureba ‘the time passed’. When it denotes a more abstract meaning such as ‘degree’ or ‘about,’ it must be compounded with another noun to constitute a larger noun, as in faye-fodo ‘like flies’ or ikura-fodo ‘about how many’. While autonomous use is found in OJ, the compounded one is attested after LMJ, according to Konoshima (1966), which conforms to the change from nouns to affixal nouns in (24). In the history of kurai, the whole change in (24) can be observed. Kurai was an autonomous noun meaning ‘rank.’ The affixal use of kurai appears in EModJ, conveying the meaning ‘about’: kyuusyaku-kurai ‘about 9 feet’. It can also appear after case particles in PModJ as in [N John ]-ni kurai (at least to John’, but this use is very new according to Konoshima (1966)’. We have seen in this subsection that the process of (24) is active in forming adverbial particles.4 In the next subsection, we will go into another pattern of syntactic change, which is also a productive source of adverbial particles.
4 It is not uncontroversial whether the change from affixal nouns to clitics is unidirectional in the change of adverbial particles. Some adverbial particles such as mo (also) and koso (foc) change their position from the outside to the inside of the noun: For example, ‘[N Yamada]-o-mo.’ to ‘[N dare-mo]-o’ where o is an accusative marker. But the host noun that allows the particle to be inside the case marker is limited to particular lexical items: In the case of mo, it is almost limited to wh-words such as ‘dare (who)’. Therefore, the change from clitics to affixal nouns should be regarded as a case of lexicalization (Brinton and Traugott 2005) and is not a productive process related to adverbial particles (Kinuhata 2011).
384
Tomohide Kinuhata
4.2 From adjunct-clausal markers to adverbial particles The preceding subsection, presented a change of grammatical categories by considering the syntactic aspect of the change previously discussed, i.e., the change of exclusives. This subsection will show another syntactic development of adverbial particles, which we have yet to explore in this chapter. The change is characterized as the one from adjunct-clausal markers to adverbial particles: It consists of the change from clausal to non-clausal status and that from an adjunct to an argument position in a sentence. To illustrate, we will focus on (a) the change of naritomo from connective to an adverbial particle and (b) the emergence of indeterminate and exemplification use of yara. After looking at each pattern, we will discuss the differences and similarities of those changes and argue that they are productive processes for developing adverbial particles. Naritomo originally consisted of a copula verb nari and a concessive marker tomo. In EMJ, tomo forms a concessive clause, occasionally taking nominal predicates formed by nari. In (27), for example, onazi parakara-nari ‘be a sibling’ is the clause which tomo takes, and the concessive clause formed by tomo composes a different clause from the matrix mutubiyori-na-masi ‘I would be attached to him’. (27) Ware onna-nara-ba, onazi parakara nari-tomo, kanarazu 1sg woman-cop-cond same sibling nari-tomo surely mutubiyori-na-masi. love-perf-irr ‘If I were a woman, I would be surely attached to him even if we were siblings.’ (Genji (4), p. 28) Thus, the structure of (27) is as (28), where ‘SubC’ and ‘MainC’ represent the subordinate and the main clause, respectively. (28) [ [SubC [ onazi parakara-nari ] -tomo ] [MainC mutubiyori-na-masi ] ] In LMJ, however, there is evidence showing that the phrase, including naritomo, does not constitute an independent clause. In Amakusaban Heike monogatari, published in 1592, the number of examples increases where the noun before naritomo is interpreted as an argument of the main clause predicate, and, among those examples, the following have case markers attached to the nouns naritomo takes (Kinuhata 2007b). (29) a.
Fakanai fude no ato o naritomo tatematutte on’otodure short brush gen trace acc naritomo send news o kik-au. acc hear-vol ‘I will at least send a short letter and hear his news’ (Amakusa Heike p. 61)
16 Development of adverbial particles
b.
385
Mosi iduku no ura ni naritomo kokoroyasuu possibly somewhere gen shore loc naritomo relievedly otitui-ta naraba, get.settled-perf cond ‘If I get settled relievedly at whatever shore, . . . ’ (Amakusa Heike, p. 184)
The case markers indicate that the NPs before naritomo are the arguments of the predicates such as tatematutte ‘send’ and wotitui-ta ‘get settled’, which means that tomo does not organize a subordinate clause but marks an argument with nari in the main clause. The primary change witnessed here is the loss of clausal status of the phrase to which tomo attaches. Assuming that the phrase naritomo resides outside the verb phrase (VP) inheriting the property of tomo as a subordinator, I give the following structure, for instance, to the sentence in (29a).5 (30) [MainC [ [ fude-no ato o1] naritomo ] [VP t1 tatematutte ] ] In (30), naritomo adjoins to the VP, and the case-marked NP is assumed to move from inside the VP. Although it is important to pose the structure (30) in considering the gradual change of the particle, it is possible to interpret the structure of (29a) as the one where naritomo is inserted into the VP without the movement of case-marked NPs as illustrated in (31). (31) [MainC [VP fude-no ato-o tatematutte] naritomo]
Given the structure in (31), which allows naritomo to attach to any constituent in the VP (“attachment transformation” in the sense of Kuroda (1979)), examples are expected to appear where naritomo directly attaches to the predicate. This type of example is obtained in the transcriptions of Kyōgen plays written in 1643. (32a) is one where naritomo attaches to the verb, and, in (32b), naritomo is semantically interpreted to take scope over the VP as indicated in the translation, despite that naritomo attaches to an NP syntactically. Those examples are evidence of naritomo taking the structure as in (31). (32) a.
Nikusa mo nikusi, nabutte naritomo yar-au. hate also hateful make.fun.of naritomo do-vol ‘I hate him very much, so I’ll kind of make fun of him.’ (Toraakira (Chū), p. 67)
5 Since nari is not used as a usual copula verb in this period, it constitutes a single morpheme with tomo.
386
Tomohide Kinuhata
b.
Yai koi, midu narito nom-ase-taraba torikaese. voc come.imp water naritomo drink-caus-cond get.back.imp ‘Hey, come on, if you did some treatments like letting him drink water, get it back.’ (Toraakira (Jō), p. 204)
While the change from (28) to (30) is a change in the status of the phrase that naritomo takes, i.e., from clausal to non-clausal, the change from (30) to (31) is a change of the position that naritomo occupies, i.e., naritomo occupies a higher position in (30) but it moves to or is inserted in a lower position in (31). Thus, the change is, as a whole, from adjunct-clausal markers to adverbial particles. It is important to note here that the relevant change is not incidental but repeatedly occurring changes: Kinuhata (2007b) shows that the same path as naritomo is partially observed in their histories of demo and nara, and finds a similar change in the formation of kakari-musubi construction. These instances suggest that it is not unreasonable to posit the process from (28) through (30) to (31) as a productive one in developing adverbial particles. We now go to another example: the appearance of new uses in the particle yara. Yara is originally from ‘ya+ara+mu [kakari particle + copula verb + conjectural marker]’. Before the emergence of indeterminate use, it was used as a sentence final particle to mark interrogatives. (33) Are wa ikanaru zyaurau nite masimasu yaran. that top what noble.person cop cop.hon yara ‘Who is that noble person?’ (Kakuichi Heike (1), p. 362) As a byproduct of this construction, the interrogative sentence organized by yara could also be used to conjecture the reason for a fact the following sentence denotes, as in (34): Syntactically, yara in (34) functions to make the preceding clause adjoined to the following clause. (34) Yo ni wa ikanisite more-keru yaran, awareni yasasiki world dat top how leak-pst yara sad lovely tamesi ni zo pitobito mausi-aeri-keru. instance cop foc people talk-itera-pst ‘How did people come to know, they were talking about the story, seeing it as sad and lovely.’ (Kakuichi Heike (1), p. 53) In the 15th and 16th centuries, yara developed an indeterminate use which typically marks indeterminate pronouns in a simple sentence as in (35): A postposition attaches to yara in (35a), and the indeterminate pronoun with yara functions as the subject in (35b).
16 Development of adverbial particles
(35) a.
b.
387
Nani-yarau de kono tyuu o mi-ta zo. what-yara loc this annotation acc see-pst sfp ‘I saw this annotation in something.’ (Shikishō p. 475) Dare-yarau fito no tumi o fikiukete at-ta fodoni, who-yara someone gen crime acc take cont-pst because ‘Because someone takes on the crimes of the others, . . . ’ (Mōgyūshō, p. 262)
Thus, yara changed from clausal to non-clausal status and from adjoined to argument positions.6 In the 18th century, yara gained another use to mark arguments, i.e., exemplification. In the beginning, however, the phrase to express exemplification with yara does not mark the argument of the verb but, attaching to a verb, modifies the manner of the following predicate (Iwata and Kinuhata 2011). For example, in (36a), ketumazuku ‘stumble’ and suberu ‘slip’ exemplify the manner of crawling out. (36) a.
b.
Ketumazuku yara suberu yara hauhau haidete, stumble yara slip yara crawl go.out ‘Stumbling and falling, we crawled and went out.’ (Shinjū Yoigōshin, p. 439) Sugito ni hitai utu yara ateru yara yooyooni cedar.door dat forehead bump yara hit yara at.last osihiraki, push.open ‘He, at last. opened the door made of cedar, knocking it and bumping his forehead against it.’ (Heike Nyogonoshima, p. 513)
(36) has a syntactic pattern in which yara, constituting a clause, adjoins the following sentence. It is in the 19th century that we can find more examples of yara attaching to nouns than to verbs. Particularly, the examples in which yara is used with case markers, as in (37), are not found until the 20th century. (37) Tamoto kara kaneire yara tokei yara o muzoosani tukamidasite, sleeve abl wallet yara watch yara acc casually take.out ‘He casually took out his wallet, his watch, etc. from his kimono sleeves,’ (Seinen) At this stage, yara functions to mark an argument of the main clause, which indicates that it moves the same shift as the indeterminate use.
6 Kinuhata and Whitman (2011) consider that the indeterminate use of ka follows the same change as the indeterminate yara.
388
Tomohide Kinuhata
So far, we have pointed out the syntactic change shared by both naritomo and yara. There is one difference that should be noted here, however. Though both particles came to mark arguments as seen above, the relative order of case markers and the adverbial particles differs: While naritomo follows case markers as in (29), yara precedes it as in (37), which means that the latter does not evolve into a clitic.7 If the same structural change as naritomo had taken place in the development of yara, case markers should have appeared before yara via syntactic movement as in (30). This assumption suggests that the change that yara went through is different from that of naritomo. Kinuhata (2007a) proposes that the factor that derives the new uses of yara is not purely syntactic but rather discourse pragmatic. For example, in the development of the indeterminate pattern, the phrase with yara co-refers to the same object as the noun in the following sentence as in (38a), and then the yara-phrase replaces the noun as in (38b), reanalyzed as having a nominal status. (38) a.
b.
[Iduku no onzausi yaran]1. [Iro siroku mime kao mo where gen son.of.noble yara color white figure face also yoki syaunen]1 no uma o odorasite kuru nari. good boy nom horse acc spur come cop ‘Who is that noble boy, a boy with white skin and good looks is coming spurring a horse.’ (Chūkajabokushishō, p. 182) [Iduku no onzausi yaran]1 no uma o odorasite kuru nari.
Kinuhata (2007a) also discusses what triggers the difference under discussion: While the relation between the adjoining clause and the matrix clause had been tight in the case of naritomo since tomo originates in a connective particle, the relation had been loose in the case of yara since yara comes from a sentence final particle. We have seen in this subsection that naritomo and yara evolve from adjunct-clausal markers to markers for arguments. Although yara had not acquired the status of a clitic and thus might not be considered an adverbial particle, the change from clausal to nonclausal is productive in Japanese and can be regarded as a rich source of adverbial particles.
5 Conclusion We have seen various developments of adverbial particles, which include the semantic changes of exclusive and scale-inducing particles and the syntactic changes from nouns to clitics and from adjoined-clausal markers to adverbial particles. As for the 7 As a result, naritomo can attach to, and semantically scope over, the verb phrase, whereas yara cannot. This is because yara resides inside the noun in (37). See also Section 4.1.
16 Development of adverbial particles
389
former changes, while we found a general cline from ‘degree’ to B-type exclusives in the emergence of exclusive meanings, finding such regularity in the change of scale-inducing particles is not straightforward. As for the syntactic change, both patterns are productive and worthy of investigation in that the change from nouns to clitics is a case of grammaticalization, and the change from clausal markers to adverbial particles has not been paid much attention in the literature. In concluding this chapter, I emphasize that all these changes are not deduced from a particular theory of language change but are generalized inductively from historical data of Japanese. Therefore, this chapter also shows that exploring historical data on a particular language can reveal properties of language change that have hitherto been unnoticed in the linguistic discussion.
References Brinton, Laurel J. and Elizabeth C. Traugott. 2005. Lexicalization and language change. Cambridge: Cambridge University Press. Fujitani, Nariakira. 1778. Ayuishō [Monograph on Ayui]. (Cited from Ayuhishō shinchū [New annotation on Ayuishō]. Kazama shobō 1993.) Iwata, Miho and Tomohide Kinuhata. 2011. Yara ni okeru reiji yōhō no seiritsu [The emergence of yara expressing “exemplification”]. Journal of Japanese grammar 11(2). 60–76. Kano, Kyozaburo. 1938a. ‘Dani’ ‘sura’ no yōhōjō no sai ni tsuite [On the distinction in use between dani and sura]. Kokugo to Kokubungaku 15(6). 49–64. Kano, Kyozaburo. 1938b. Insei-Kamakuraki ni okeru dani sura sae [Dani, sura and sae in the Insei-Kamakura period]. Kokugo to Kokubungaku 15(10). 155–172. Kikuchi, Yasuto. 1986. ‘Bakari’ ‘dake’ [Bakari and dake]. In Tetsuya Kunihiro (ed.), Imibunseki [Analysis of meaning], 57–59. Tokyo: Linguistic Laboratory of Tokyo University. Kinuhata, Tomohide. 2005. Fukujoshi dani no imi to kōzō to sono henka [Focus particle dani: Its syntax, semantics and their evolution from the Nara to the Heian period]. Journal of Japanese grammar 5(1). 158–175. Kinuhata, Tomohide. 2007a. Fukasetsu kara toritate e no rekishi henka no futatsu no patān [Two patterns of the historical change from adjuncts to focus]. In Hirofumi Aoki (ed.), Nihongo no kōzō henka to bunpōka [Japanese syntactic change and grammaticalization], 65–91. Tokyo: Hituzi Syobo. Kinuhata, Tomohide. 2007b. Syntactic change from connective to focus particles in Japanese. In Naomi McGloin and Junko Mori (eds.), Japanese/Korean Linguistics 16. 395–406. Stanford: CSLI. Kinuhata, Tomohide. 2011. Kakarijoshi, fukujoshi [Focus particles and adverbial particles]. In Satoshi Kinsui, Yoshiyuki Takayama, Tomohide Kinuhata and Tomoko Okazaki (eds.), Shirīzu nihongoshi 3: Bunpōshi [Series of the history of Japanese 3: The history of grammar], 167–189. Tokyo: Iwanami Shoten. Kinuhata, Tomohide and John Whitman. 2011. Genesis of indefinite pronouns in Japanese and Korean. In Willium McClure and Marcel den Dikken (eds.), Japanese/Korean Linguistics 18. 88–100. Stanford: CSLI. Konoshima, Masatoshi. 1966. Kokugojoshi no kenkyū: Joshishi no sobyō [Studies on Japanese particles: A sketch of the history of particles]. Tokyo: Ōfūsha. Koyanagi, Tomokazu. 1997. Chūko no ‘bakari’ to ‘nomi’ [Bakari and nomi in Early Middle Japanese]. Kokugakuinzasshi 98(12). 14–28. Koyanagi, Tomokazu. 2003. Gentei no toritate no rekishitekihenka: Chūkoizen [Historical change of exclusive particles: Before Early Middle Japanese]. In Yoshiko Numata and Hisashi Noda (eds.), Nihongo no toritate: Gendaigo to rekishihenka chiriteki hen’i [Japanese adverbial particles: Contemporary Japanese, historical change and geographical variation], 159–177. Tokyo: Kurosio Publishers. Kuroda, S.-Y. 1979. Generative grammatical studies in the Japanese language. New York: Garland Publishing Company. Miyachi, Asako. 2003. Gentei no toritate no rekishitekihenka: Chūseiikō [Historical change of exclusive particles: After Middle Japanese]. In Yoshiko Numata and Hisashi Noda (eds.), Nihongo no toritate: Gendaigo to rekishihenka chiriteki hen’i [Japanese adverbial particles: Contemporary Japanese, historical change and geographical variation], 179–202. Tokyo: Kurosio Publishers. Miyachi, Asako. 2010. Dake no rekishitekihenka saikō: Meishi no keishikika, bunpōka toshite [Reexamination on the historical change of dake: As a study for formalization and grammaticalization of nouns]. In Ikudō Tajima (ed.), Nihongogaku saizensen [Frontiers in Japanese linguistics], 421–442. Osaka: Izumi Shoin.
16 Development of adverbial particles
391
Sadanobu, Toshiyuki. 2001. Tansaku to gendainihongo no ‘dake’ ‘sika’ ‘bakari’ [Japanese “dake,” “shika” and “bakari” from the viewpoint of probing]. Journal of Japanese grammar 1(1). 111–136. Terada, Hiroe. 2000. Meijiki Tōkyōgo ni okeru ‘dake’ no genteiyōhō [The exclusive use of dake in Tokyo Japanese in the Meiji period]. Kokugokenkyū 63. 65–86. Yamada, Yoshio. 1908. Nihon bunpōron [Japanese grammar theory]. Tokyo: Hōbunkan. Yuzawa, Kokichiro. 1958. Muromachijidai gengo no kenkyū [A Study on the language of Muromachi period]. Tokyo: Kazama Shobō.
Tomoko Okazaki
17 The history of demonstratives 1 Introduction This chapter explains the historical changes in the use of demonstratives, especially demonstrative pronouns and adverbs, in Japanese. First, I present a comparison between demonstratives of Modern Japanese and Early Middle Japanese, highlighting the points of historical change. In Modern Japanese, both demonstrative pronouns and adverbs can be methodically arranged according to the same system termed ko, so, a (Table 1). In addition to pronouns (1) and adverbs (2) and (3), this system also contains sets of modifiers (4). It was first codified in Kanae Sakuma (1951). Table 1: Demonstrative forms in Modern Japanese. Series
Demonstrative pronouns
Demonstrative adverbs
kosoa-
kore/koko/kono, etc. sore/soko/sono, etc. are/asoko/ano, etc.
koo/koosite/konnani, etc. soo/soosite/sonnani, etc. aa/aasite/annani, etc.
Demonstrative pronouns (1) Kore wa watasi this top my ‘This is my book.’
no gen
hon desu. book cop
Demonstrative adverbs (2) Batto wa koosite moteba umaku uteru. bat top like.this hold.prov well can.hit ‘If you hold the bat like this, you’ll hit better.’ (3) Huzisan ga konnani utukusii to Mt. Fuji nom like.this pretty comp ‘I did not think Mt. Fuji was this pretty.’ (4) Ano yama ga Huzisan That mountain nom Mt. Fuji ‘That mountain is Mt. Fuji.’ https://doi.org/10.1515/9781614512851-018
desu. cop
wa sira-nakatta. top think-neg.PST
394
Tomoko Okazaki
However, in Early Middle Japanese, the three ko-, so-, ka (a-) series of demonstrative pronouns and the two kaku-, sa- series of demonstrative adverbs appear to form entirely separate systems (Table 2). Table 2: Demonstrative forms in Early Middle Japanese (Heian period). Series
Demonstrative pronouns
Series
Demonstrative adverbs
koso ka(a-)
kore/koko/kono, etc. sore/soko/sono, etc. kare/kasiko/kano (are/asiko/ano)
kakusa-
kaku(kau)/ kayau, etc. sa(sika)/sayau, etc.
In Old and Early Middle Japanese, according to their rules of usage, demonstrative pronouns and adverbs are part of the same system. They exhibit parallel changes but maintain a profound historical relationship. I define terms that are used to describe the uses of the demonstratives given in this chapter in the following way. a. Deictic use refers to an object that can be felt or directly perceived. Example: Sore tot-te! hat take-ger ‘Take that!’ b. Non-deictic use includes the following anaphoric use and recognitional use. – Anaphoric use occurs when an antecedent is shared with an indicated object or correlating to an indicated expression in either a written or spoken linguistic context. Example: Kinoo hon o kat-ta. Yesterday book acc buy-pst Sono hon wa taka-katta. that book top expensive-pst ‘I bought a book yesterday. The book was expensive.’ – Recognitional use describes an element that directly relates to a past experience (long-term) and is part of a memory. Example: (hitorigoto) Kinoo no ano (monologue) yesterday gen that ryoori wa yo-katta food top good-pst. ‘(Monologue) That food yesterday was the best.’
17 The history of demonstratives
395
In Modern Japanese, both the ko- and so- series have deictic and anaphoric uses, and the a- series has deictic and recognitional uses. There are domains within the deictic use that are described as follows. – ko- series (proximal): something close to the speaker – so- series (mesial): something somewhat distant from the speaker or closer to the listener – a- series (distal): something far from the speaker
2 Hashimoto (1966) Hashimoto (1966) is particularly important for this chapter because this study elucidated the system of demonstratives in Old Japanese during the Nara period. Most accounts of Old Japanese demonstratives posit a three-term ko-, so-, kasystem. The facts of the use of this series in Old Japanese are as follows. 1) The ko- series is almost always used deictically, referring to what is or has been directly perceived by the speaker (5), although a few examples show anaphoric use (6). (5) pototogisu ko yu naki-wataru cuckoo here abl cry-cross.concl ‘A cuckoo is crossing here while crying.’ (6) paribukurwo kore pa tabari-nu needle bag this top get.resp-perf.concl ‘The needle bag, I got this.’
(MYS 8.1476)
(MYS 18.4133)
2) The so- series is mostly used anaphorically. There are no clear examples of a spatial deictic use of so-, except for a few that have reference to the listener (7). (7) soko koko mo onazi kokoro so you I etop same heart foc ‘You and I have the same heart.’
(MYS 19.4189)
3) The ka- series is found in only three Old Japanese poems (MYS 14.3565, MYS 18.4045, and MYS 20.4384). This rarity could lead to the conclusion that this series was undeveloped. However, due to the limited material available, we do not know how much of the ka-series was actually used. The examples of the ka- series are entirely used deictically, referring to something that is far from the speaker.
396
Tomoko Okazaki
(8) a ga mopu kimi ga mi-pune I gen think.of.adn you gen hon-boat ‘Is it the boat of my beloved you, that.’
kamo kare q that
(MYS 18.4045)
Hashimoto (1966) concluded that there was a binary contrast between the ko- series and the so- series in Old Japanese (where the ko- series indicates an object that is being directly felt by the speaker, and the so- series indicates an object that is not being directly felt). This study proposed that the demonstrative system began to shift from a two-way to a three-way opposition. Although Hashimoto’s contribution is important, he did little research on demonstratives after Early Middle Japanese, and his work did not cover demonstrative adverbs. In the next section, we will examine the historical changes in the demonstrative systems, combining demonstrative pronouns and demonstrative adverbs.
3 Old and Early Middle Japanese (Nara and Heian periods) This section discusses the demonstrative system in Old and Early Middle Japanese. Much of its content is derived from Kinsui, Okazaki and Jo (2002) and Okazaki (2010). The following two points are important features of demonstratives pronouns and adverbs in Old Japanese. (9) A. Underdevelopment of the ka- series (distal) B. Two-way opposition between the ko-, ka-, kaku- series and the so-, sa- series Table 3: Demonstrative forms in Old Japanese (Nara period). Series
Demonstrative pronouns
Series
Demonstrative adverbs
kosoka-
ko/kore/ko no/koko/koti so/sore/so no/soko kare/ka no
kakusa-
kaku sika
The underdevelopment of the ka- series (Table 3) appears in Old Japanese only (ka, ka no, kare) in the form of a low frequency of use. On the contrary, words beginning with (k) in the ko-, ka-, kaku- series indicate referents that can be directly perceived, while those beginning with (s), as in the so-, sa- series, indicate a referent that cannot be directly perceived. In other words, in Old Japanese, only demonstratives that begin with /k/ have deictic use. Demonstratives that begin with /s/ have mainly anaphoric use. Even though the ko-,
17 The history of demonstratives
397
kaku- series also has anaphoric use, it is so close to deictic use that this is referred to as text deixis/discourse deixis. Incidentally, recognitional use was the most interesting type of all at this time. What appears in Modern Japanese in the a- series appears in the so-, sa- series in Old Japanese (10). (10) so no ywo pa yutani ara-masi that gen night top slowly be-sbjv ‘I should have taken it easy that night.’
monowo excl
(MYS 12.2867)
In Modern Japanese, the speaker can point out some elements that are experienced directly in the past with the use of (a) or indirectly by the use of (s). However, in Old Japanese, there was no such division. Something directly experienced could be introduced only indirectly in discourse, indicated as an unforeseen element by the use of the so-, sa- series. After this, Early Middle Japanese did not change much with respect to (9B). However, the facts for (9A) did change; in Early Middle Japanese, with rapid increases in the ka- series, many examples of recognitional use can be seen in addition to numerous examples of deictic uses of the so-, sa- series (11) and (12). Moreover, the a- series of demonstrative pronouns can be seen in Early Middle Japanese, but they are not seen often. Deictic use: so-, sa- series (11) soko wa di ni koso that top JI cop.inf foc ‘I think that this is ‘JI’.’ (12)
ara-me exist-conjec.excl
(Genji; Utsusemi)
sa na-se-so Like.that proh-do-proh ‘Don’t do such a thing.’
(Makura 146)
How was a mesial indication established at this time? This is difficult to infer from the underdeveloped ka- series in Old Japanese. However, the following does appear to be true for Early Middle Japanese. (13)
Demonstrative pronouns: ko- (proximal), ka- (mesial and distal) Demonstrative adverb: kaku- (proximal, mesial and distal)
Although a differentiation was created in the ko- series and ka- series of demonstrative pronouns in relation to deictic use, in Early Middle Japanese, there was only the kaku- series of demonstrative adverbs. Therefore, the kaku- series indicated all objects of a domain. Deictic use: kaku-, ka- series.
398
Tomoko Okazaki
(14) ka- series (mesial): kare wa tare ga zo that top who gen foc ‘Whose is that?’
(Genji; Sakaki)
(15) kaku- series (proximal): kau sitawi arukaba Like.this cherish follow.cond ‘If I follow after you like this. . .’
(Genji; Suetsumuhana)
(16) kaku- series (mesial): Kaku na Like.this proh ‘Don’t do such a thing.’
(Genji; Aoi)
(17) kaku- series (distal): kau ayasiki kakine ni nan saki paberi-keru Like.this shabby hedge dat foc bloom be-mpst.adn ‘It’s something that will bloom in that sort of shabby hedge.’
(Genji; Yūgao)
As stated, the three ko-, so-, ka- (a-) series of demonstrative pronouns and the two kaku-, sa- series of demonstrative adverbs of Early Middle Japanese are not the same system in terms of morphemes. In addition, no two series had exactly the same system in Early Middle Japanese, even from the classification of the deictic use of the domains (proximal ko-, mesial so-, and distal a-), as in Modern Japanese. I will organise and classify the usage of demonstratives (demonstrative pronouns and adverbs) in Old and Early Middle Japanese, omitting the ka- series, as it is underdeveloped in Old Japanese. Table 4: Use of demonstratives in Old Japanese (Nara period). Series Usage Anaphoric use Deictic use Recognitional use
(k) ko- kaku(P) (A)
(s) so- sa(P) (A)
✓ ✓ ✓ ✓ ✶ ✶
✓ ✓ ✶ ✶ ✓ ✓
Demonstrative pronouns are represented with (P) and demonstrative adverbs are represented with (A). “✓” when there is a usage, “✶” when there is no usage.
17 The history of demonstratives
399
Table 5: Use of demonstratives in Early Middle Japanese. Series Usage Anaphoric use Deictic use Recognitional use
(k) ko- kaku- ka(a)(P) (A) (P) ✓ ✓ ✓ ✓
✶
✶
(s) so- sa(P) (A) ✓ ✓ ? ? ✓ ✓
✶
✓ ✓
Demonstrative pronouns are represented with P and demonstrative adverbs are represented with A. “✓” when there is a usage, “✶” when there is no usage, “?” when there is a slight usage.
As shown in Tables 4, 5, and 6, when viewed through the rules of demonstrative usage, Old and Early Middle Japanese and Modern Japanese can be defined by their demonstrative systems, which are organized similarly. As stated, the ka- series was developed during Early Middle Japanese, where we begin to see numerous transformations in demonstrative systems, such as the deictic use of demonstratives starting with (s). However, in Late Middle Japanese, we see even greater transformations. Table 6: Use of demonstratives in Modern Japanese.
Usage
Series
Anaphoric use Deictic use Recognitional use
ko(P) (A)
so(P) (A)
a(P) (A)
✓ ✓ ✓ ✓
✓ ✓ ✓ ✓
✶ ✶ ✓ ✓ ✓ ✓
✶
✶
✶
✶
Demonstrative pronouns are represented with (P) and demonstrative adverbs are represented with (A). “✓” when there is a usage, “✶” when there is no usage.
4 Late Middle (the Kamakura and Muromachi periods) and Modern Japanese (Edo period and onwards) During Late Middle Japanese, the demonstrative systems began to undergo great transformations toward the ko-, so-, a- system seen in Modern Japanese. The changes that occurred in Late Middle Japanese can clearly be seen in a comparison of Kakuichi-bon Heike monogatari, from the Kamakura period, and the Amakusaban Heike monogatari, translated into colloquial Japanese in the late Muromachi
400
Tomoko Okazaki
period. Below are anaphoric examples comparing the same scenes from both versions of the tale. (18) a.
b.
(19) a.
b.
kaku- series: kakaru on-me be.like.this.adn hon-situation ‘this situation’ ko- series: konoyauna me like.this.adn situation ‘this situation’ sa- series: sayauno asobimono like.that.adn dancing.girl ‘A dancing girl like that’ so- series: sonoyauna asobimono like.that.adn dancing.girl ‘A dancing girl like that’
(Kakuichi Heike; Tenga Noriai)
(Amakusa Heike; 1–2)
(Kakuichi Heike; Giō)
(Amakusa Heike; 2–1)
The demonstrative adverbs of the kaku-, sa- series, such as kakaru and sayau, undergo a transformation to konoyau, sonoyau, representing the ko-, so- series. The a- series of demonstrative adverbs in Late Middle Japanese also began to be appear, as shown below (20). In other words, the demonstrative adverbs of Late Middle Japanese begin to change from kaku-, sa- to ko-, so-, a-. (20) anoyauni kokoro no tuyoi onna like.that heart gen strong woman ‘A strong woman like that’
(Kyōgen; Setsubun)
Furthermore, the deictic use also underwent a transformation toward the end of Late Middle Japanese. The following examples are provided. (21) a.
a- series (mesial): are wa ikani that top what ‘What is that?’
(Kakuichi Heike; Fukuhara-inzen)
17 The history of demonstratives
b.
(22) a.
b.
so- series (mesial): sore wa nani zo that top what foc ‘What is that?’
401
(Amakusa Heike; 2–9)
kaku- series (mesial): kau na-se-so like.that proh-do-proh ‘Don’t do that.’ sa- series (mesial): sau na-si-so like.that proh-do-proh ‘Don’t do that.’
(Kakuichi Heike; Ichininokake)
(Amakusa Heike; 4–7)
In (21) and (22), within the a- series and the kaku- series, the mesial demonstratives are and kau (kaku) are replaced by sore and sau. This clearly indicates that towards the end of Late Middle Japanese, the deictic use of the so-, sa- series was established. During Late Middle Japanese, the recognitional use of the so-, sa- series was lost. Usage changed to almost that of Modern Japanese. Table 7: Use of demonstratives in Late Middle Japanese. Series
ko- so- a(P)(A) (P)(A) (P)(A)
kaku- sa(A) (A)
Anaphoric use Deictic use Recognitional use
✓ ✓ ✓ ✓ ✶ ✶ ✓ ✓ ✓ ✓ ✓ ✓ ✶ ✶ ✶ ✶ ✓ ✓
✓ ✓ ✓ ✓
Usage
✶
✶
※: Demonstrative pronouns are represented with (P), and demonstrative adverbs are represented with (A). “✓” when there is a usage, “✶” when there is no usage.
In the Edo period, the kaku kau and sa sau transformations were absorbed into the so- series, becoming kou, sou. In addition, aa (23) developed in the latter half of the Edo period by analogy with ko-, so-, a-. Moreover, the other words of the kaku-, saseries underwent changes in the written language. With this, they became the ko-, so-, a- demonstrative system of Modern Japanese. (23) aa wa ika-ne like.that top do-neg ‘They can’t do that.’
(Ukiyoburo)
402
Tomoko Okazaki
5 Typological perspective Finally, I discuss Japanese demonstratives within the typology of other demonstrative systems of the languages of the world. In this chapter, I note that the domain of the so- series in Modern Japanese is mesial. Levinson (2004) argues that Japanese so- has two functions, namely, first, simply to indicate that the referent is close to addressee and, second, (as with Turkish şu) to draw the addressee’s attention to a new referent. It is not necessary to discuss Levinson’s proposal that one function is drawing the addressee’s attention, but it is appropriate to conclude that the so- series has two functions and that it is less spatial. Among the two functions of the so- series, that is, referring to something somewhat distant from the speaker or to something closer to the listener, the former is especially complex and unstable. By contrast, the domains for the ko- and a- series are speaker-centered distance systems, that is, they describe domains of things that are close to the speaker and others that are far from the speaker, and their use is stable. What does this indicate? Linguists have argued that deixis is the source of reference (Fillmore 1982; Levinson 2004). However, as far as the historical evidence of the so- series is concerned, there is no trace of such a change. As noted above, there are no clear examples of a spatial deictic use of the so- series in Old Japanese, and the deictic use was established during Late Middle Japanese. In this regard, Hashimoto (1966) reported that there are a few examples of a spatial deictic use of a demonstrative with reference to the listener, but it largely appears anaphorically in Old Japanese, adding that its usage has evolved. His explanation of the historical changes of so- is appropriate. In addition, Kinsui (1999) stated that deixis is a prototype use for any use of the koor a- series, and it is unlikely that deixis is the prototype use for the so- series. However, as Kinsui (1999) found, after the derivation of usage from the prototypical usage, the prototypical use declines, and the derivative is mistaken for the original. This is also a possibility. In conclusion, the results of observed historical facts suggest that two systems within a deictic system produced a third that is not reducible to either.
References Fillmore, Charles J. 1982. Towards a descriptive famework for spatial deixis. In Robert J. Jarvella and Wolfgang Klein (eds.), Speech, place, and action: Studies in deixis and related topics. 31–59. Chjichester: John Wiley & Sons Ltd. Hashimoto, Shirō. 1966. Kodaigo no shiji-taikei: Jōdai o chūshin ni [The demonstrative system in older Japanese: With a focus on Old Japanese]. Kokugo Kokubun 35(6). 329–341.
17 The history of demonstratives
403
Kinsui, Satoshi. 1999. On the relation between the deictic use and the non deictic use of the Japanese demonstratives. Journal of Natural Language Processing 6(4). 67–91. Kinsui, Satoshi, Tomoko Okazaki and Migyon Jo. 2002. Shijishi no rekishiteki / taishōgengogakuteki kenkyū: Nihongo, kankokugo, torukogo [Historical and contrastive studies of demonstratives: Japanese/ Korean/Turkish]. In Naoki Ogoshi and Hideki Kimura (eds.), Comparative linguistics (The Science of Language 4), 217–247. Tokyo: University of Tokyo Press. Levinson, Steven C. 2004. Deixis. In Laurence R. Horn and Gregory L. Ward (eds.), The handbook of pragmatics (Blackwell handbooks in linguistics 16), 97–121. Oxford: Blackwell. Okazaki, Tomoko. 2010. Nihongo shijishi no rekishiteki kenkyū [Historical study of Japanese demonstratives]. Tokyo: Kurosio Publishers. Sakuma, Kanae. 1951. Gendai nihongo no hyōgen to gohō: Kaitei-ban [Grammar and expressions in Modern Japanese: Revised edition]. Tokyo: Kōseikaku Publishers.
Yoshihiko Ikegami
18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’ 1 Introduction 1.1 The theoretical background The general theoretical background (broadly in line with cognitive linguistics) for the specific topic I propose to discuss in the present study is summarized as follows: i. The speaker of language is, in an important sense, a cognizing subject who construes the situations to be encoded in ways most appropriate to the purposes of his utterances. ii. There are relativistic, as well as universal aspects of construal. Thus, being faced with one and the same situation, the speaker of one language may construe it in one way, while the speaker of another language may prefer to construe it in another way. The sum total of the construals favored and preferentially opted for by the speakers of the language will generate what Whorf (1956) called ‘fashions of speaking’ characterizing the speakers of the language in question. iii. A major contrasting pair of construals I propose to consider in particular is ‘subjective construal’ and ‘objective construal’. In subjective construal, the speaker takes the ‘subject-object merger’ stance, while in objective construal, the speaker takes the ‘subject-object opposition’ stance. iv. I assume that at the present stage of what we might call ‘evolution’ of language, subjective construal is the marked choice, while objective construal is the unmarked choice. The extent to which the speaker indulges in subjective construal, however, may differ from one language to another. The speaker of Japanese seems to take the stance of, and to indulge in, subjective construal more readily than the speakers of Western languages. The preferential choice of one type of construal stance rather than the other will result in relativistic aspects of the ‘fashions of speaking’ across individual languages. (For more detailed discussion of these points, the readers are referred to Ikegami (2005, 2008, 2015) among others.)
1.2 Aims of the present study The main part of the present study addresses the question of subjective construal insofar as it can be documented in the earliest existent records of the Japanese language. Framing the question in this way, however, entails having to work under heavy self-imposed restrictions. First of all, you can expect to find hardly any access to the https://doi.org/10.1515/9781614512851-019
406
Yoshihiko Ikegami
spoken language of the period. You have to work exclusively on written materials. Moreover, the written materials you can work on are severely limited in terms of genre, namely waka poems (Man’yōshū) or otherwise historical chronicles (Kojiki and Nihonshoki) and fudoki (descriptions of the natural and cultural features of local regions together with their histories). To have to do with poetic text exclusively, for example, is certainly a fatal blow in linguistic research. In our present case, however, there is a sense in which we can turn an apparent misfortune into a blessing. Underlying a piece of waka is a specific construal of the world by the author which may very well go beyond the customary practice by the average speakers of the author’s language. (One may recall in this connection Chomsky’s formulation in 1964, in which the average speaker’s activity is characterized as ‘rule-governed creativity’ and the poet’s activity as ‘rule-changing creativity’. According to Chomsky, the latter lies outside the scope of linguistics and linguists are advised not to bother about it.) A set of specific linguistic practices characteristic of a particular waka-poets or a particular school of waka-poets is referred to by the term ‘kafū’ (literally, ‘waka-style). Just as language changes with time, so does the ‘kafū’. And as any history of literature can tell us, changes in ‘kafū’ can take place and proceed in much shorter spans and sometimes even in drastic ways. Since ‘kafū’ is definable in terms of a set of specific construals favored and indulged in by a particular waka-poet or a particular school of waka-poets, it should be a particularly fitting place to test the validity and effectiveness of the kind of approach proposed in the present paper. It may be argued that the present paper will then turn out to be nothing more than a study of stylistic changes in the traditional lines. Rather than trying to relate stylistic changes to the specificities of the private circumstances under which the author was living or of the social circumstances under which a group of authors were living – as is usually done in stylistic studies, the present study, in line with its basically linguistic orientations, will focus on how the waka-poets (as speakers of Old Japanese) fared when they found themselves perceptually in direct contact with events, processes, and states which engrossed their whole attention – in other words, being involved in situations to which they were irresistibly tempted to apply ‘subjective construal’. Their mediating sensory channels can be various – visual, auditory, olfactory, palatory, and tactile. It can be safely assumed, however, that the most important part is played by visual perception, followed perhaps by auditory perception.1 (This is linguistically reflected in our data by the fact that we find the verbs compounded with the auxiliary of ‘spontaneity’, yu for sight (miyu ‘be visible to’) and for hearing (kikoyu ‘be audible to’), but none for the other senses.) Taking the relative primacy of vision into account, then, we can reformulate our question as ‘construing for verbalization what the authors of waka saw before
1 Readily attributing primacy to vision is justly criticized as ‘Eurocentrism’ in recent publications on linguistic typology (e.g. Aikhenwald and Storch 2013: 3, 18). For pre-modern (and also Modern) Japanese, however, the primacy of vision is well attested. See Section 2 and note 6.
18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’
407
their eyes, what was it that the authors were really seeing?’ To capture this and to trace the way their stance evolved over time – these are the main points I am going to address in the present paper. Thus, our main concern will be with those waka-pieces generally subsumed under the rubric ‘jokeika’, i.e. waka-pieces which depict the scenery, where the stance of the waka-poet is characterizable as ‘subject-object merger’, himself being thoroughly embedded in the scene, being at one with what he visually perceives.2
2 Vision and the verbs of visual perception in Man’yōshū 2.1 On human visual perception in general Humans are genetically endowed with capacities of perceiving the environments they live in. These capacities serve them as important channels through which they get to know more about their environments and to deal better with them. Aikhenvald and Storch (2013) offer a general description of the verbs of perception across languages. With a focus on vision and the verbs of visual perception, the main points they make can be summarized as follows: Every language has a way of talking about seeing, hearing, smelling, tasting and touching (Aikhenvald and Storch 2013: 1). Among them, vision is the preferred sense in many, especially Western, cultures (but not universally), followed by hearing, which is highly valued in certain other cultures.3
Visual perception can be either intentional (agent-controlled) or unintentional (experience-controlled) and some languages provide two different lexical items for the two cases. (Cf. the contrast between look at and see in English.) Aikhenvald and Storch (2013: 19–20) point out specifically that in many languages the distinction between the two cases is a corollary of the constructions in which one and the same single lexeme is used. Yet a third type, called ‘source-based copulative (state) construction’ (as in English ‘the painting looks old’) is mentioned by Viberg (1983). It is pointed out by Aikhenvald and Storch (2013: 20), however, that the copular status of the perception verb in this construction is debatable (and hence the term ‘copulative’ is inappropriate) and that
2 Notice that in clear contrast to the waka-pieces on one’s own love (called sōmonka), jokeika nearly consistently zero-encode the speaker, coupled with the zero-encoding of the verbs of perception (cf. Sasaki 2007). For the rather conspicuous use of the verb miyu, see Section 2.4. 3 Aikhenvald and Storch (2013: 26–27) cite and discuss a number of Australian and Amazonian languages in which the supremacy of vision is not verified. In some Amazonian languages, verbs of cognition are found to be recruited from verbs referring to auditory rather than visual perception.
408
Yoshihiko Ikegami
there are cases in which a third verb, in addition to the verbs with controlled and noncontrolled meanings, is used here (as in English sound side by side with listen and hear). Perception leads to cognition. It is, therefore, well expected that the expression of perception is closely associated with the expression of cognition, either of them spanning grammar as well as lexicon. Instances of semantic shifts from either visual or auditory perception to cognition appear to be commonly observed. On the cultural side, vision, with its privileged access to knowledge, is readily thought of as the source of power. Thus, a special kind of visual power can be attributed to the head of a community, the shaman, and the like. Related to this implication are a set of taboos associated with vision; this is because vision can be intrusive and therefore dangerous. Very commonly the taboo takes the form of prohibition on looking at someone’s face (especially, with serious eye-contact) or looking into a certain room, box, and the like.4 In the social domain, the verb of visual perception can semantically be transferred to mean ‘visiting’ and ‘guarding’ someone or rather contrastively, ‘looking upon someone with lust’. In certain cultures, the idea is current ‘that what is in one’s perceptual domain belongs to them’ (Aikhenvald and Storch 2013: 27; see also Morishige 1951: 2.5) What can we say about vision and the verbs of visual perception in Man’yōshū against the backdrop as sketched above?
2.2 Linguistic aspects of visual perception in Man’yōshū First, the linguistic aspects. In Man’yōshū, visual perception is encoded most commonly by the two verbs, miru and miyu, both of them presumed to be derived from the lexical base, me (‘eye’).6 4 Japan has its share of the theme of ‘forbidden chamber’ in its myths and folktales. See Kawai (1991). 5 Cf. “‘Seeing people and things’ implies ‘having them in one’s possession’” (Morishige 1951: 2; translation mine). See also note 10. 6 A close etymological tie between the word for ‘eye’ and the word for ‘seeing’ is attested across languages (Aikhenvald and Storch 2013: 29) and Japanese is no exception. Note, however, that the same does not necessarily apply to other senses. In Japanese, the organ for hearing is mimi, while the verb for auditory perception is kiku and the organ for smelling is hana, while the verb for olfactory perception is kagu. This is taken as an indication of the supremacy of vision in ancient Japanese culture (Nakanishi 1972: 24). On the other hand, the statement by Aikhenvald and Storch (2013: 26) that in numerous languages, ‘hear, listen’ also means ‘obey’ applies to Old Japanese and the use is still commonly observed in Modern Japanese as well (as when one says, Oya no iu koto o kike (‘Listen to/obey what the parents say’)). I may add also that the Chinese character for ‘listen’ (聴) borrowed into Japanese has the ‘ear’ radical on its left-hand side. Special mention should perhaps be made on the use of the verb nipopu in Old Japanese. It primarily refers to the color red, especially when it is produced as a soft reflection of light (as when cherry blossoms are glowing in the morning sun). Presumably, it construes the coloring as something emanating from an entity (later, even extended to cover a woman emanating beauty). At the same time, the verb
18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’
409
According to the recently published volume, Nihon koten bunrui goi hyō [Classified Lexicon of the Japanese Classics] (Miyajima, T. Suzuki and Abe. (eds.) 2014), miru and miyu are used in Man’yōshū 805 and 201 times respectively, the former ranking second in the frequency of occurrence in the text, next only to the highest ranking word ari (‘be’, ‘exist’). This, incidentally, suggests how ‘seeing something’ plays an important part in triggering off the imaginative activities of the poets of Man’yōshū.7 Grammatically, miru is basically a transitive verb, being used in the syntactic frame, ‘Subject (perceiver) [ga] + Object (something perceived) [wo] + miru’, while miyu is an intransitive verb, being used in the syntactic frame, ‘Subject (something perceived) [ga] [+ Experiencer (perceiver) + ni] + miyu’. (The pair of brackets indicates linguistic items that can be zero-encoded.) As a transitive verb, miru is normally agentive, referring to an intentional act ‘controlled’ by the perceiver (i.e. ‘look at’, ‘cast one’s glance at’), although it can also be non-agentive contextually (i.e. ‘happen to see’).8 Miyu is, etymologically, mi[ru] + yu, where yu is an auxiliary verb known to have been used in Old Japanese in three senses: ‘passive’, ‘spontaneity’ (‘jihatsu’, defined as ‘a natural development of an action / a process / a state without explicit human intervention’), and ‘possibility’. In combination with the lexical base mi-, yu in the form miyu is most prominently used in senses which vacillate between ‘spontaneity’ and ‘possibility’ in Man’yōshū, quite close to the ways in which the verb mieru is used in Modern Japanese (as in hosi ga mieru ‘A/The star is visible [to me]’). Besides the verb miyu, the same auxiliary yu is found in Man’yōshū with a small set of perceptual and cognitive verbs, kikoyu, which is composed of the lexical base kiko- (‘hear(ing)’) and yu and the verb omopoyu,
is also used to mean ‘smell (of), give off fragrance’, so that it covers two apparently different senses. The apparently synaesthesic use of the word nipopu here is discussed by Nakanishi (1972: 125) together with the word sayaka, which is used in the sense of ‘(either visually or auditorily) distinct, clear’, thus covering two apparently different senses. Cases like these offer the question, ‘How many meanings?’ (Aikhenvald and Storch 2013: 17). Aikhenvald and Storch, carefully avoiding the pitfall of ‘Eurocentrism’, propose to treat analogous cases in Australian languages as involving a generic meaning ‘perceive’. The same interpretation could very well apply to the cases in Man’yōshū as discussed by Nakanishi (1972), who, while admitting the plausibility of generic interpretation, concludes nevertheless by saying that the semantic ‘center’ is in vision. Whether or not this is due to the recognition of the surpassing prominence of vision among the authors in Man’yōshū is a question still to be explored. 7 In the sum-total of the seventeen works (including poetic works like Man’yōshū, Kokinshū and Shin-kokinshū and prose works like Genji monogatari, Makura no sōshi, Heike monogatari and Tsurezuregusa), miru ranks eleventh, preceded by ari ‘be’, su ‘do’, ipu ‘say’, and omopu ‘think (of)’, to mention the verbs only. Incidentally, the verb nagamu (‘look attentively at’), often homonymically associated with the noun, naga-ame (‘long, enduring rain’) is not recorded in Man’yōshū. In Kojiki, however, a few examples of the verb nagamu are found (Kitayama 1969). Kokinshū lists only four examples, (cf. 3.4.2) but Shin-kokinshū lists as many as fifty-seven. 8 As when someone says in Modern Japanese, Yoru nohara o aruite ita toki, nagarebosi o mita ‘When walking across the field at night, I saw a falling star’. The use of miru here is parallel to the one in which the speaker says, Asi o otta ‘I broke my leg’, referring to an accidental injury. The idea is that a volitive subject is held responsible to what happened to him as much as what he himelf did intentionally.
410
Yoshihiko Ikegami
which is composed of the lexical base omopo- (‘think(ing)’) and yu. All these three verbs are intransitive, their grammatical subject (encoded as a noun or as a nominal phrase or clause) referring to something (an entity, an action, a process or a state being either visually or auditorily perceived or being thought about).9 Semantically, both of the verbs, miru and miyu are generally used in Man’yōshū referring to visual perception. This applies to the verb miyu in particular, especially when the verb is used in the formulaic way (see 2.4). Miyu is also commonly applied to the experience of ‘seeing’ in the dream; yume ni miyu ‘be seen in the dream’ is almost a set phrase. Even in those cases where the dictionary definitions offer glosses like ‘come’ (as in “. . . maturamu mono wo miyenu kimi kamo” (15.3771: the author, thinking of a man in exile who is surely waiting for a pardon to be granted, laments over his not really coming back), ‘meet’ (as in “. . . ponoka ni miyete wakarenaba” (8.1526: the author laments over the briefness of his encounter), and ‘think, know’ (as in “. . .puru yuki to pito no miru made” (5.839: the author compares the scattering plum blossoms to falling snow), the implication of ‘visual perception’ is still clearly felt. In Genji monogatari, it is possible to find comparable examples in which a semantic shift away from visual perception can more clearly noticed. There is, however, an interesting set of linguistic contexts which apparently facilitated the semantic shift of the verbs, miru and miyu from ‘actually visual’ to ‘fictionally non-visual’. This is when these verbs of perception are used in conjunction with the particle to preceding them. MYS 5.839, discussed in the preceding paragraph, is an example in point involving the verb, miru. This, in fact, served as the starting point from which the device of ‘mitate’ was later to be developed. (See Section 3.5.1). For the verb miyu, there is another type of linguistic context available, namely connecting to the adnominal form of an adjective, as in the following example: Siranupi Tukusi no wata pa mwi ni tukete Shiranui Tsukushi gen cotton top body dat attach.ger imada pa ki-nedo atatakani mi-yu not.yet top wear-neg.conc warm see-pass.concl ‘The cotton of Shiranui Tsukushi – I have not yet worn it, but it looks warm’ (MYS 3.336) The author is certainly looking at the cotton, but he appears to be more interested in telling about the warmth he will be blessed with, when he wears it. What is being talked about here is the author’s personal (and hence ‘subjective’) feeling. In this sense, the verb miyu still retains some of its own features as a verb of visual perception. It is,
9 The term, ‘source’ is sometimes applied to the grammatical subject as used here (cf. Aikhenvald and Storch (eds.) 2013), the idea being that ‘something perceived’ functions as ‘stimulus source’ for the ‘perceiver’. Thus, in the Modern Japanese sentence, Yama ga mieru, yama is characterized as ‘source’.
18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’
411
however, only a step from here to those cases in which the particular feeling in question is assumed to be entertained by anybody – something surely to be shared by everybody (and hence, something ‘intersubjective’). There is a shift of meaning of the verb miyu from ‘perceptual’ to ‘cognitive’ (i.e. judgment). Genji monogatari is found to abound in examples of the latter kind (Otake 1969).
2.3 Cultural aspects of visual perception in Man’yōshū We already have a substantial amount of discussion on the cultural implications of the act of seeing at the age of Man’yōshū. According to Tsuchihashi (1965: 266), “In ancient times, the act of ‘seeing’ was not a simple act of perception; it was an act affecting people’s life / soul” (translation mine). There was, on the one hand, a popular belief, known as ‘the evil eye’, to the effect that some people had a power to harm others by simply casting a glance at them. This was the dark aspect associated with the act of ‘seeing’. In contrast, it also had a bright aspect, namely ‘seeing’ believed to be endowed with spiritual power which could strengthen the spiritual force of the object (irrespective of whether it is a person or a non-person) that was seen. Little is as yet known about the precise nature of this effect, called tama-furi (lit. ‘soul shaking’). Apparently, the communion between one who saw and something which was seen was conceived to be mutual rather than just one way. The seer, on his part, shared the spiritual power evoked in what he saw, the result being a state of ‘subject-object merger’ in which the seer was totally embedded, and merged with what he saw (cf. Uchida 1983: 4, 5, 13, Nagafuji 1973: 621). A ritualized form of such communion at the most elevated level is kunimi (lit. ‘country / land’ + ‘seeing’). (For a convenient survey of the studies on kunimi, see Kanno 1985.) The emperor climbs a hill, stands on top of it and looks over the land he rules. By visually surveying the land he rules, the emperor is not only assured that he is the one who rules the land, but also his glance is believed to be endowed with power to energize the spirits of the land and to make the land prosper. The second waka piece in the first volume of Man’yōshū (incidentally, not in the form of ‘tanka’ (‘short waka’, with 5-7-57-7 mora) but of ‘chōka’ (‘long waka’) is introduced as one composed by the emperor Jomei on the occasion of his kunimi on top of Kaguyama, a hill not far from the royal palace in Yamato (today’s city of Nara and its environs):
412
Yoshihiko Ikegami
Yamato ni pa murayama aredo toriyoropu Yamato dat top many.mountains exist.conc beautifully.dressed.adn Ameno Kaguyama, nobori-tati kunimi wo sureba kunipara celestial Kaguyama climb-stand.inf kunimi obj do.prov plainland pa keburi tatitatu, unapara pa kamome tatitatu. Umasi top smoke keep.rising wide.sea top seagulls fly.over wonderful.adn kuni so Akitusima Yamato no kuni pa. country excl Akitsushima Yamato gen country top ‘Many are the mountains in Yamato. However, the best one is the splendidly dressed Heavenly Kaguyama. Climbing up and standing on top of it, I perform kunimi. Then I see smoke rising (from the kitchens) all over the land and seagulls taking flight all over the sea. What a wonderful country this land of Yamato is, this Akitsushima (‘lit. Dragonflyland’)’ (MYS 1.2) The emperor commends the land, and his commendation is supposed to invoke the auspicious spirits of the land. Notice that some of the features verbally mentioned by the emperor could not possibly have been visually perceived from the top of Kaguyama (e.g. the sea mentioned in the present piece). The emperor thus ‘foresees’ the land as it will be seen through his act of commendation.10 (Cf. Uchida 1983: 4, also Nagafuji 1973: 22.) It is generally assumed that this ritualistic role was then taken over by ‘professional’ poets in the court. Kakinomoto Hitomaro is considered to have been one of them, who accompanied the emperor’s local tours and composed waka-poems at various places on the way admiring and praising the features of the land ruled by the emperor – a practice called kunibome ‘land/country lauding’. (Cf. Tsuchihashi 1965.) After the demise of the ritualistic practice, a lot of waka-poets (whether at court or among ordinary people) still liked to compose about natural scenes (e.g., Mt. Fuji as in 3.318) and sometimes natural scenes with people actively working in them which they found impressive. Borrowing a term from the recent discussion on evidentiality, we can characterize ‘jokeika’ as involving ‘mirativity’ in its various aspects (cf. Delancey 1997 and relevant papers in Cognitive Linguistics 16(3) (2012) in a debate on ‘mirativity’). The idea of power associated with the act of seeing is found in the culture in less ritualized forms as well. There is, in Old Japanese, a word, magupapi (etymologically presumed to be ‘ma’ (a variant 10 Notice that the function of language in kunimi is just the reverse of the one in curse, where the linguistic statement is inauspicious, being intended to bring harm to people. Whether auspicious or inauspicious, the underlying idea is the same, namely, what one utters in language will be realized in actuality – a superstitious belief in the mysterious power of language widely found across cultures. In ancient Japan, the belief was in fact quite real – one and the same lexical form, koto is known to have meant either ‘speech, language’ or ‘event’ (the idea being that what one utters in language will be realized in actuality). The remnant of the traditional belief in auspicious language is still found in norito (transcribed in Chinese characters by a combination of the character meaning ‘commend, congratulate’ plus the one meaning ‘language, word’), prepared linguistic statements addressed to gods and read by priests in Shinto shrines.
18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’
413
of ‘me’ ‘eye’) + kupu ‘eat’ + api ‘meet, match’, meaning ‘letting each other’s eyes meet, looking directly at each other’ (referring to an act by a couple of man and woman) which implies a mutual consent to marriage. Closely related to this is the belief that letting someone else know one’s name is to let oneself be placed under their control. This is in fact the theme of the very first piece of the first volume of Man’yōshū. The emperor Yūryaku, going out in the field on a spring day, meets a girl with a basket. He addresses her with the following words: “. . . kono woka ni na tumasu kwo, ipye kikana, na norasane, . . .” ‘You, a girl who gathers greens on the hill, let me know where you live, let me know your name . . . .’
2.4 On the verb miyu in Man’yōshū Before concluding Section 2, we have to pay special attention to the verb miyu, which, as has already been noted by a number of researchers, demonstrates certain peculiar behaviors in its use in Man’yōshū. After Man’yōshū, the verb continues to be used through the eight successive imperial anthologies of waka poems (ranging from Kokinshū in 905 to Shinkokinshū in 1205) called Hachidaishū (Eight-Generation Anthologies), and the average frequency of use through Hachidaishū is nearly as much as (or even slightly higher than) the frequency of its use in Man’yōshū. According to Hirasawa (1999: 91), however, there is a drastic drop of frequency from Man’yōshū to Hachidaishū in the use of the verb in the waka-final position. In Man’yōshū, 40 out of the 203 examples of its use occur in the waka-final position, while in Hachidaishū, only 8 out of the 485 examples of its use are found in the waka-final position. (Note further that three out of the eight examples in Hachidaishū are known to rerecord the pieces in Man’yōshū with some verbal modifications.) Concluding a waka-piece with the verb, miyu is called miyu-dome (lit. ‘miyu-stoppage’), and miyu-dome is thus a distinctive stylistic characteristic of Man’yōshū. There is another interesting but somewhat puzzling peculiarity about the use of the verb miyu in Man’yōshū. When connecting to a verb, miyu is often found to connect to the conclusive form, and not, as is expected, to the adnominal form, of the preceding verb: e.g. Ama wotomye tama motomurasi okitunami kasikwoki umi ni punade seri miyu (6.1003) ‘Female divers are going to look for pearls in the wavy offing, for their boats (have put out and) are heading for the awesome sea + miyu’, where seri is se- (originally, si (infinitive of su ‘do’) + ari ‘be’, reinterpreted as se + ri (stative auxiliary, conclusive form). There has in fact been a series of lively discussions among Japanese philologists about the nature of miyu-dome – starting with Gomi (1939), followed by Inuzuka (1973 [1955]), Yoshii (1962), Satake (1980 [1964]). Kitahara (1965), Ide (1995 [1981]), Sakaguchi (1982), Uchida (1983), Takahashi (1985), Sato (1989), Hisajima (1992), Kageura (1993), Uchida (1997), Hirasawa (1999), among others. Noting, in particular, that there is something formulaic about the way the verb miyu is used – used especially at the waka-final position, itself always in the conclusive form, connecting either to a nominal phrase or
414
Yoshihiko Ikegami
to the conclusive form of the predicate verb.11 I suggest that the verb miyu in these uses, is already starting on its way to grammaticalization.12 (In this I agree with those few Japanese philologists (e.g. Ide 1995 [1981] and Hisajima 1992) who claim that miyu is, in Man’yōshū, turning into an auxiliary rather than being a full verb. Used formulaically at the waka-final position, miyu is always used in the sense of visual perception. It is used in the syntactic frame: ‘subject (referring to an activity or a state being perceived, as encoded in a clausal form with its final predicate verb in conclusive form’ + [implied, but linguistically unencoded] Indirect Object (referring to the speaker as perceiver) + miyu’. Since it refers to a ‘private’ visual experience, known only to the perceiver himself, miyu is used appropriately only by the speaker (i.e. ware ni miyu ‘be visible to me’) and since the verb is used ‘ego-centrically’ in the ‘subject-object merger’ type of stance, the speaker is always encoded as ‘zero’, the result being a ‘perceiver-less’ sentence (Uehara 1998). Cf. also Ikegami (2005: Section 2.6, 2008: Section 6, 2015: Section 5.1.2). There is another paradox involved in the use of the verb miyu. The speaker of Japanese, when talking about what he visually perceives, has actually two alternatives – either verbalizing only what he visually perceives or verbalizing his own act of visual perception in addition to what he visually perceives. Thus, if the speaker goes out of the house in the evening and sees the moon shining brightly in the sky, he may say either “I SEE THE MOON SHINING BRIGHTLY” (or alternatively, “THE MOON SHINING BRIGHTLY IS VISIBLE TO ME”) or simply “THE MOON IS SHINING BRIGHTLY”. Being definitely committed to objective (or the subject-object-contrast type of) construal, the speaker of English prefers saying the former, while the speaker of Japanese, being inclined to subjective (or the subject-object-merger type of) construal, prefers saying
11 Tsuchihashi (1965: 390–394), referring specifically to the use of the modifying adjectives followed by a nominal in makura kotoba (‘pillow words’: ornamental epithet), points out that these adjectives can occur in conclusive form as well as in adnominal form (e.g. na kupasi Samine no sima (MYS 2.220: ‘the island of Samine with its beautiful name’) and na kupasiki Inami no umi (MYS 3.303 ‘the sea of Inami with its beautiful name’) and that the former implies ‘admiration’ and ‘praise’, while the latter has an objective tone. It is also added that the parallel contrast can be observed with verbs and that the verb in infinitive form (e.g. isana twori umi (NSK 68: ‘the sea where [they] hunt whales’) implies ‘admiration and praise’. So far as I am aware, however, these possibly related cases do not seem to have been together discussed by Japanese classicists. 12 In support of those linguists (cf. Aikhenwald and Dixon (eds.) 2014), I suggest that a ‘grammatical’ (i.e. ‘fully grammaticalized’) marker of evidentiality is something obligatory in each sentence which reports the speaker’s perceptual experience. In Man’yōshū, miyu was still far from that stage – something which Aikhenwald and Dixon propose calling ‘an evidential strategy’. Cf. Ikegami (2020a). It is also known that there was an auxiliary myeri (etymologically, mi (lexical base of miru) + ari (‘be, exist’)) in Old Japanese. In Man’yōshū only one example is found, in MYS 14.3450, an azuma uta (waka composed in the eastern dialect). In the Heian period, it is favored by women writers in their diaries, its meaning being extended from visual (‘look like’) – at which level it is contrasted with nari (etymologically, na (lexical base of naku ‘cry’ + ari ‘be exist’) – to non-perceptual (‘seem’). Its use declines drastically after the Heian period. For a description of the grammaticalization of the full verb of ‘seeing’ in Thai, see Rattanaphanusorn and Thepkanjana (2007).
18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’
415
the latter. Interestingly, in Man’yōshū, we find the verb miyu used in a number of cases where it could very well have been dispensed with. Thus, we actually have two comparable waka-pieces in Man’yōshū, in one of which the verb miyu is duly encoded, while in the other miyu is not encoded: “. . .Apadi no sima ni tadu wataru miyu” (7.1160: ‘. . . cranes are flying over to Awaji Island + miyu’) and “. . . Asibye o sasite tadu naki-wataru” (6.919: ‘. . .cranes are flying crying towards the reedy shore’). What is the function of the verb miyu explicitly encoded here? By explicitly verbalizing ‘miyu’, the author makes sure that what he is talking about is something which he is visually perceiving at the moment. Thus, miyu is here functioning as a kind of evidential marker – the speaker finds it necessary to state explicitly that he is in fact an eyewitness to what he is talking about. (The poet here is comparable to the shaman of a tribe, whose utterance of a verb of visual perception is supposed to carry incomparably more weight than in the case of ordinary people.) This incidentally is what Satake (1980 [1964]: 30–31) had in mind when he wrote: “At the bottom of the author’s consciousness, there was a strong desire, when describing a scene, explicitly to say ‘miyu’ – some strong motivation which he couldn’t possibly suppress unless he encoded the progressing activity in terms of ‘miyu’. . . . It goes without saying that this is how the poet’s ancient mentality worked.” (translation mine) (cf. Ikegami 2020a.) Satake (1980 [1964]: 25–26, 30) made another important point on the nature of the verb miyu in Man’yōshū. Its primary meaning is commonly known to be ‘spontaneity’, i.e. ‘become visible (to me)’. Satake argues, however, that the meaning of the verb in its formulaic uses is ‘stative’, i.e. ‘be being seen’. His argument is based on ‘sayōsei yōgen hanpatsu no hōsoku’ (law of the incompatibility of actional yōgen’), first proposed by Ishigaki (1955 [1942]: 215–238), who claims that only ‘keijōsei yōgen’ (stative predicate) can connect to ‘sayōsei meishiku’ (actional nominal phrase). Taking note of the fact that the nominal phrase to which the verb miyu connects is actional (e.g. “ama wotomye punadeseri miyu” (‘female divers are setting sail + miyu’) 6.1003, “Kasuganwo ni keburi tatu miyu” (‘smoke is rising in Kasugano + miyu’) 10.1879), together with his own observation that the verb miyu is followed by the perfective auxiliary tu rather than nu (Satake 1980 [1964]: 26–30),13 Satake concludes that miyu constitutes ‘keijōsei 13 The distinction between tu and nu as auxiliaries of perfectivity has traditionally been made by saying either that they are used with transitive and intransitive verbs respectively or that they are used with active and non-active verbs respectively. Consequently, it was considered at one time something of a puzzle that the verb miyu is found with the auxiliary tu side by side with a few examples in which it is found with the auxiliary nu. Satake (1980 [1964]: 24–30) takes note of the fact that the auxiliary tu is also applied to a special group of stative verbs (including ari ‘be’, among others). He proposes to account for the extension of the use of tu by assuming a shift of the semantic focus of active verbs from the process of the activity to the resulting state and concludes that the verb miyu, which takes the auxiliary tu, is in fact, stative (keijōsei yōgen: mieteiru) and not active (sayōsei yōgen: mietekuru). The most persuasive study on the distinction between tu and nu is Ide (1995 [1981]). The author points out that just as dynamic verbs can be used with a focus either on the process (katei) or the result (kekka) of the dynamism they refer to, so stative verbs can also be used with a focus either on the
416
Yoshihiko Ikegami
yōgen’ (stative predicate), meaning something like ‘be being seen (by me) or mieteiru in Modern Japanese. Thus, the situation described by the verb miyu is, to all intents and purposes, one in which the perceiver is integrated with what is being perceived – one in which the speaker’s stance (as perceiver) is typically characterizable as ‘subjective (or subject-object merger type of) construal’. (Ikegami 2020a; see also Takahashi (1985: 247–268) and Hirasawa (1999: 85–97), who refer to the same point in terms of kōkan ‘mutual communion’ and ittaikan ‘sense of oneness’ respectively) rather than one in which something to be perceived, which is detached from the perceiver, is ‘coming’ to the perceiver (‘come to be perceived (by me)’), ‘become visible (to me)’ or mietekuru in Modern Japanese.)
3 Evolution of ‘seeing’: Changing construal as applied to the natural scenery 3.1 Jokeika in Man’yōshū On the basis of what we have so far reviewed about the language and culture of the period of Man’yōshū, I now propose to trace the evolutionary path of the stance of construal followed by those who aspired to be authors of jokeika (lit., ‘scenery-depicting waka’).
3.2 Encounter with ‘mirativity’ as starter We will start with those waka-pieces the main theme of which is the sense of ‘mirativity’14 which the author experiences when he ‘suddenly’ and ‘without expectation’
continuity (sonzoku) or the conclusion (kiketsu) of the state they refer to and that tu is used when their process or continuity is focused on and nu is used when their result or conclusion is focused on. This view agrees with and supports Satake’s view on the semantic status of the verb miyu both in its formulaic and non-formulaic uses. 14 ‘Mirativity’ is defined as “grammatical marking of ‘unprepared mind’, including unexpected and also surprising information” (Aikhenvald 2004: 394). The point about the definition above is that ‘mirativity’ is to be conceived of as a ‘grammatical’ category. This incidentally suggests, however, that ‘mirative meanings’ can be encoded by other linguistic forms than those of grammar, just as evidential meanings are, as is already fully appreciated, for example, encoded by lexical forms (e.g. verbs of perception serving to indicate information sources) as well as grammatical ones, so that people talk about loosely ‘evidential strategies’. What the present author is interested in in the present article is not the theoretical question of whether or not ‘mirativity’ can be attested as a grammatical category but the pragmatic question of how ‘mirative meanings’ are manipulated as a springboard for the poet’s further flight of imagination.
18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’
417
finds himself face to face with an overwhelmingly impressive natural scene of some kind. These pieces may sound, at least superficially, like jokeika of the purest and most unadulterated kind. The following well-known piece by Yamabe no Akahito serves us as an example: (1) Tagwo-no-ura yu utiidete mireba masirwo Tago-gen-cove abl coming.out.ger see.prov pure.white ni zo Puzi no takane ni yuki pa puri-kyeru cop.inf foc Fuji gen lofty.peak dat snow top fall-mpst.adn ‘Coming out from the cove of Tago, I look around and see that the snow has fallen pure white on the lofty peak of Mt. Fuji’ (MYS 3. 318) See also 3.272, 7.1133, 8.1568. The whole piece starts with the characteristic tone of kunimi-uta with the formulaic phrase, mireba ‘when looking widely around’ and proceeds to commend the noble presence of Mt. Fuji. Its main clause, however, already contains a germ of the double vision which is later to evolve into a more sophisticated one: the image of the pure white snow-covered mountain before the poet’s eye as contrasted with the image of the same mountain as it usually looks. The contrast is well implied in the use of the auxiliary kyeri which concludes the whole piece. Again, we already have quite a lot of discussion about what is the exact meaning of the auxiliary kyeri. An attractive idea has been suggested by Takeoka (1970), who proposes on the basis of the etymology of the word (ki ‘come’ + ari ‘be, exist’) that the auxiliary implies that something in the other world (anata naru sekai: ‘a world which exists far away, apart from this present world’) comes to this present world. This implies that, applied to the present author’s work, the snow-covered pure white mountain is an entity of ‘the other world’ and that the author, coming out of Tago’s cove, suddenly and unexpectedly notices this and admires it.15 What we have here is a set of features of a situation which can be characterized as ‘mirative’ (comparable to what is implied when a term eitan is used in interpreting a waka-poem) and we may even claim that part of the function attributable to the auxiliary, kyeri is ‘mirativity’, although it has in Japanese admittedly developed other related functions as well.
3.3 ‘Mirativity’ triggering off abductive reasoning In anticipation of the way the structure of jokeika is going to evolve further, there is one point in particular about a piece like (1) which is worthwhile specially noticing. As 15 In Japanese linguistics, cases like this are also commonly labelled as kizuki ‘becoming aware of, noticing’ and hakken ‘discovery’. The speaker concomitantly feels pleased (that he is clever enough to have solved the apparent puzzle), regrets or feels ashamed (that he has been so ignorant as to be puzzled), as the case may be.
418
Yoshihiko Ikegami
already mentioned, there is no doubt about (1) that its main theme is the sense of ‘mirativity’ which the author had on seeing Mt. Fuji pure white covered with snow. Actually, however, this is not the whole story. Latent through a work like (1) is also a line of logical thinking – a process of inference which is apparently triggered off by the sense of ‘mirativity’ felt by the author and which proceeds as follows: [ENCOUNTERING SOMETHING EXTRAORDINARY]: The poet finds Mt Fuji pure white. [INVOKING A PLAUSIBLE ‘RULE’]: Whenever it snows, the mountain, covered with snow, looks white. [INFERENCE]: It has snowed. This, in fact, is the kind of inference called ‘abduction’ by Peirce (1931–1958: 2.264–265).16 It is a kind of mental logic which is commonly applied by humans whenever they try to interpret something as signifying something else. Among the jokeika poets in Man’yōshū, interested in interpreting a witnessed scene as a ‘sign’ (shirushi)17 of something else, this came to establish itself as a favorite way of construing the scenery.
16 ‘Abduction’ is a kind of inference, but unlike ‘deduction’, it is fallible. It may lead to a wrong judgment, because the assumed rule may happen to be wrong. Thus, in the well-known example about the Turkish governor discussed by Peirce (1931–1958: 2.264–265), the man on horseback may or may not be the governor of the province. There is a good possibility that someone other than the governor is also honored as much in the same province. Nevertheless, ‘abduction’ is acknowledged as an extremely important means for the humans to expand the scope of knowledge on the world they live in and on how to interact with it. (Note, by contrast, that ‘deduction’, though infallible, operates only within the scope of already acquired knowledge and does not serve creative purposes.) On the prevalence of abductive reasoning underlying various aspects of Japanese culture, see Arima (1991). Cf. also note 19. 17 Toyama (1991) addresses the question of what the earliest attested native word for ‘sign’ in Japanese was. He checks the two oldest history books in Japanese, Kojiki and Nihonshoki, actually two of the earliest written documents in Japanese, compiled in 712 and 720 respectively. He finds that “shirushi” (印) is the word in question. The author notes, on the basis of the different Chinese characters used to transcribe the word, that three meanings, ‘a foreboding, omen, prefigurement’, ‘an effect (as evidencing something)’ and ‘a good omen’, can be attributed to the word shirushi as used in Kojiki and that the meanings of the word as used in Nihonshoki are more diversified – classified into as many as twenty-two senses, including ‘a foreboding, omen, prefigurement’, ‘a piece of evidence’, ‘an effect (as evidencing something), ‘a good omen’, among others. The overlapping between the two sets of meanings listed for the two books above suggests that there are in fact three basic ways in which the word, shirushi is applied by the speaker: in the first case, the speaker perceives something and associates it with something else that will take place in the future (in which case, what the speaker perceives counts as ‘foreboding, omen’), in the second case, the speaker perceives something and associates it with something else that took place in the past (in which case, what the speaker perceives counts as ‘effect’, ‘trace’ or ‘evidence’, as the case may be); and in the third case, the speaker perceives something and associates it with something else that is taking place spatially away from him (so that the speaker cannot possibly perceive it himself). As will be seen easily, these three basic ways in which the word, shirushi is applied
18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’
419
3.4 The scene before one’s eyes interpreted as a ‘sign’ of something else ‘Mirativity’ invites the poetic mind to speculate how and why it is that something can be so ‘mirative’ and this brings us to the second stage of evolution of ‘jokeika’, which, for the poets in Man’yōshū very often means that they address the question of what the mirativity signifies – what it is a sign of. Notice also that it involves two images, one serving as ‘sign (shirushi)’ of the other. A typical case in point is found in the following well-known piece by Empress Jitō: (2) paru sugwite natu ki-taru-rasi sirwotape spring pass.ger summer come-stat-pres white.mulberry no koromo posi-tari Ameno Kaguyama cop.adn cloth hang-stat celestial Kaguyama ‘Spring has passed and summer seems to have come, white cloths are hung to dry on Heavenly Kaguyama’ (MYS 1.28) See also 3.271, 7.1087, 10.1812, 1814, 1844, 2224, 15.3642, 3654, 17.4018. The author, Empress Jitō, happened to notice white cloths hung to dry on Kaguyama and this was interpreted by her as shirushi for seasonal change, namely the arrival of summer. Thus, the underlying abductive process of reasoning can be represented as follows: [ENCOUNTERING SOMETHING EXTRAORDINARY]: I see white cloths hung to dry on Kaguyama. [INVOKING A PLAUSIBLE ‘RULE’]: When summer comes, people hang white cloths to dry on Kaguyama. [INFERENCE] Summer has come. By introducing the notion of ‘abduction’ into the analysis of waka text involving two contrasted images, we will see two points showing up. The first point concerns a new perspective on the status of the commonly occurring auxiliary rasi in the waka-pieces of this type. The traditional interpretation given to this particular piece (2) is that it involves an inference on the part of the author and that this subjective stance on the part of the author is duly encoded in terms of the auxiliary rasi, an auxiliary usually understood to refer to an inference, specifically one which, unlike the auxiliary ramu, refers to an inference based on tashikana by the speaker can further be integrated with each other and generalized into a single thesis which says that when applying the word shirushi, the speaker is operating with a pair of images – an image of something real which he actually sees and an image of something absent which he sees only in his mind’s eye.
420
Yoshihiko Ikegami
shōko ‘positive evidence’. Notice, however, that there is serious ambiguity involved in characterizing the meaning of the auxiliary rasi as inference based on tashikana shōko. Insofar as the adjective tashikana is understood to mean ‘unmistakably perceived or perceivable’, there will be no problem, and as everybody agrees, something visually perceived, as in the waka piece above, serves as the most compelling kind of evidence. In fact, among those languages provided with a set of grammatical markers of evidentiality, vision normally enjoys the most privileged status (e.g. Velupillai 2013: 217). But whether or not visual evidence leads most reliably and unmistakably to the correct inference is another matter. People may indeed just misinterpret a visual piece of evidence. The auxiliary rasi will best be understood as indicating that the ‘rule’ invoked by the author is a ‘subjective’ one and can thus range from being fully infallible to being totally arbitrary. (The auxiliary ramu is also ‘subjective’; but it does not, in contrast, invoke any ‘rule’.) What this means is that to the extent that the ‘rule’ invoked by the author proves to be an infallible one, the conclusion drawn by the author and encoded in terms of rasi bears the tone of reservedness on the part of the author, who is consciously trying to avoid sounding categorically dogmatic. (This will apply, in fact, to the piece by Empress Jitō we have discussed, if we assume that hanging white cloths to dry on Kaguyama in summer was a customary practice of the day, not only known to the author, but also shared by the members of the community as well.18) The second point concerns the fact that ‘abduction’ is not simply an ‘inference’. The inference by ‘abduction’ involves positing a ‘rule’ which serves to account for something extraordinary encountered by the author and in this sense, it also involves offering an ‘explanation’ for what has been observed. Coupled with this connotation, the inference may come to bear a strong persuasive tone (something like ‘actually, this is the case’), called nintei ‘affirmation’ by some Japanese philologists (e.g. Komatsu 1983: 25).19 As discussed in the preceding section, the distinctive characteristic of the wakapieces of the second stage is that the whole text is organized along the line of abductive logic. There are two further points to be specifically discussed in relation to this type of waka-pieces.
18 In their discussion of the function of evidential markers across languages, Aikhenvald and Dixon (2014: 9) call cases like this ‘general knowledge’. 19 Notice that the whole cognitive process involved here is in fact very close to the one which the Japanese speaker undergoes when he construes the situation and decides to choose the phrase -noda for encoding. Thus, the cognitive process leading to the utterance of the sentence, Ame ga hutta noda will be describable as follows: [ENCOUNTERING SOMETHING EXTRAORDINARY]: e.g. The speaker wakes up, opens the window, and finds that the road is wet. (The speaker wonders why.) [INVOKING A PLAUSIBLE ‘RULE’ (TO ACCOUNT FOR WHAT THE SPEAKER HAS SEEN)]: If it rains, the road will get wet. [INFERENCE (OFFERED AS EXPLANATION)]: “It rained”. [Ame ga hutta noda.]
18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’
421
3.4.1 Explicit or implicit encoding of the verb miyu The first point is the explicit coding or the non-encoding of the verb miyu as predicate verb. Relevant examples in Man’yōshū can be classified into three types with subdivisions: I. (a) . . . mireba, . . . miyu. (b) . . . mireba, . . . ϕ. The first type is assumed to have two verbs of visual perception, miru (‘cast one’s glance’) in the first half-line and miyu (‘be visible’) in the second half-line. The latter verb, miyu may or may not be encoded, as in the following examples: (3)=(Ia) Nanipagata sipopwi ni tatite mi-wataseba, Naniwagata dried.beach dat stand.ger look-around.prov Apadi no sima ni tadu wataru mi-yu Awaji gen island dat crane go.over see-pass.concl ‘Standing on the dried beach of Naniwagata at low tide and looking far and wide, a flight of cranes is seen going over to the island of Awaji’ (MYS 7.1160) See also 7.1194, 1227, 17.3890. (4)=(Ib) Tagwo-no-ura yu utiidete mireba masirwo Tago-gen-cove abl coming.out. ger see.prov pure.white ni zo Puzi no takane ni yuki pa puri-kyeru cop.inf foc Fuji gen lofty.peak dat snow top fall-mpst.adn ‘Coming out from the cove of Tago, I look around and see that the snow has fallen in pure white on the lofty peak of Mt. Fuji’ (MYS 3. 318) See also 3.272, 7.1131, 8.1568. Both in (3) and in (4), the verb, miru (only implied in (4)) refers to an act of watching. The author, as perceiver, is totally embedded in, and merged with, the situation he observes. He construes the situation in the stance of ‘subject-object-merger’; hence he does not see himself and he may not be linguistically encoded, as in (4). This suggests that the explicit encoding of miyu as in (3) is a marked option. It is certainly a marked option – in the sense that it is a remnant of the ritualistic tradition deriving from kunimi, where only the ruling monarch was supposed to possess the privileged power of envigorating the land and of ‘seeing’ it prosper. After the demise of the monarchical ritual, the practice still managed to survive as a stylistic device. It was doomed eventually to die away, however.
422
Yoshihiko Ikegami
3.4.2 Explicit or implicit encoding of the sense of doubt The second type of waka-pieces are characterized in terms of the following patterns: II. (a) . . . rasi, . . . miyu. (b) . . .rasi, . . . ϕ. In contrast to the pieces of type I (in which the focus is laid predominantly on the mirativity of the scene witnessed by the author), those of type II seize the opportunity of an encounter with a mirative scene to indulge in abductive inference, speculating as to what the mirativity signifies or what it is a ‘sign’ of. (5)=(IIa)
Ayutigata sipo pwi-ni-kyerasi Titanoura Ayuchigata tide ebb-perf-mpst.pres Chitanoura ni asa kogu pune mo oki ni dat morning row.adn boat etop offing dat yoru mi-yu head.for see-pass.concl ‘At Ayuchigata the tide seems to be ebbing, for the boats rowing in Chita-noura are seen heading for the offing’ (MYS 7.1163) See also 6.1003, 1033, 7.1178, 1199, 9.1701, 10.1879, 15.3597, 17.4017, cf. 7.1227.
(6)=(IIb)
paru sugwite natu ki-taru-rasi sirwotape spring pass.ger summer come-stat-pres white.mulberry no koromo posi-tari Ameno Kaguyama cop.adn cloth hang-stat celestial Kaguyama ‘Spring has passed and summer seems to have come, white cloths are hung to dry on Heavenly Kaguyama’ (MYS 1.28) See also 3.271, 7.1087, 10.1812, 1814, 1844, 2224, cf. also 6.930, 934, 10.1938, 15.3642, 3654, 17.4018.
(6) has already been discussed in Section 3.4 as an example with an underlying structure of abductive reasoning. Essentially the same type of reasoning underlies (5): [ENCOUNTERING SOMETHING EXTRAORDINARY] I see the boats rowing in Chitano-ura heading for the offing. [INVOKING A PLAUSIBLE ‘RULE’] When the tide is ebbing in Ayuchigata, fishermen’s boats head for the offing. [INFERENCE] The tide is ebbing in Ayuchigata.
18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’
423
Notice, however, that in comparison with (6) (in which the abductive reasoning is obviously triggered by a marvelous scene witnessed by the author), the abductive reasoning practiced by the author of (5) is triggered by a scene which is certainly witnessed by the author but which is apparently much less exciting for him. It appears, in fact, that rather than choosing to admire the witnessed scene, the author is more interested in taking the witnessed scene for a sign (shirushi) and indulges himself in abductive reasoning to see what it is a sign of. He associates an image of something actually seen before his eyes with an image of something not actually seen before his eyes (which, however, he could actually see before his eyes only if he could properly displace himself). Here we have, in other words, two jikkei (i.e. scenes actually perceived or perceivable) juxtaposed in contiguity (or in ‘metonymic’ relationship). On the meaning of the auxiliary rasi, which characterizes the waka-pieces of this type, see Section 3.4, where it is discussed in relation to waka-piece (6). Here let me make some supplementary remarks. In the discussion of ‘evidentiality’ as a grammatical category, it is fairly common to talk about and distinguish between ‘inferential’ and ‘assumptive’, the former referring to inferences ‘based on visible or tangible evidence’ and the latter to inferences based on evidence other than visible results’, possibly covering such cases as “logical reasoning, assumption, or simply general knowledge” (Aikhenvald and Dixon 2014: 9). A dissenting voice, however, is also heard (e.g. Blakemore 1994: 1183), who points out that the distinction between the two posited categories is a matter of gradience rather than absolute. Positing two contrasting categories is of course well-motivated in the sense that there are in fact languages in which the two kinds of inference are encoded by two morphologically different grammatical markers of evidentiality. In the case of Man’yōshū, however, we are dealing, not with a fullfledged grammatical system of evidentiality, but with a number of ‘evidential strategies’ (i.e. “non-evidential categories frequently acquiring evidential extensions” (Aikhenvald and Dixon 2014: 19)). They involve verbs of visual perception as their integral part and rules of inference of varying degrees of validity, which are applied to the visual images encoded by the verbs of visual perception. Under the circumstances, I find it more convenient not to sharply distinguish between ‘inferential’ and ‘assumptive’ evidentiality. Underlying all cases are visually perceived images and inferences are made on their bases, irrespective of whether they are of the type of absolutely valid logical reasoning, ‘general knowledge’ or just arbitrary ad hoc ‘rules’. Interpreted in this way, rasi in Old Japanese can be conceived of as being semantically contrasted with besi in terms of either partial or maximal fulfilment of the relevant conditions entailed by the ‘rules’ on the basis of which the inference is made. Some scholars (e.g. Fujii 2012: 269–279) propose to make much of the etymologically adjectival character (‘be similar to’, ‘be like’) of the auxiliary rasi. For them, the meaning of rasi can and should be talked about in terms of the partial resemblance to the ideal image of something.
424
Yoshihiko Ikegami
3.4.3 The use or non-use of the traditional formulaic pattern III. (a) (b)
. . . ba, . . . miyu. . . .ba, . . . ϕ.
(7)=(IIIa)
Inabinwo mo yuki-sugwi-kateni omopyereba Inamino etop go-pass-unbearable seem.prov kokoro kwoposiki Kakwonosima mi-yu heart yearning.for.adn Kakonoshima see-pass.concl ‘As I find it unbearable to pass Inamino, I see Kakonoshima, of which I am fond’ (MYS 3.253) See also 3.255.
(8)=(IIIb)
Waka no ura ni sipo miti-kureba kata wo Waka gen cove dat tide rise-come.prov lagoon acc nami asibye wo sasite tadu naki-wataru being.lost reed.shore acc head.for.ger crane cry-go.across.concl ‘When the tide rises in the cove of Waka, the lagoons are lost and the cranes go crying toward the reedy shore’ (MYS 6.919) See also 3.273, 389, 6.925, 9.1161, 17.4029.
In the third type, two different lines of development converge. In type (IIIa), we find waka-pieces inheriting the features of kiryoka (i.e. waka-pieces composed on a journey), especially those supposed to have been composed by professional poets who accompanied the emperor’s journeys through the country. Originally these pieces were marked with the use of the formulaic phrase consisting of mireba at the end of the subordinate clause and miru at the end of the main clause – a correlative pattern which characterizes type (Ia). In type (IIa), however, the journeys talked about by the poets can very well be private ones, so that the poets are free to refer to their own particular occasions, often dispensing with the formulaic mireba. Thus, in the subordinate clauses, they usually refer either to the spots they or their parties pass by or come to, or to the fond thoughts of someone they care for. The predicate verbs preceding the particle ba need not be limited to miru (e.g. omopu). Type (IIIb) is superficially similar to type (Ib) (just as type (IIIa) is partially similar to type (Ia)). They commonly contain a subordinate clause marked by the particle ba, followed by the main clause which presents what is visually perceived by the author. Notice, however, that underlying type (Ia) is a formulaic correlative pattern, ‘. . . mireba, . . .miyu’ (‘if one looks . . ., . . . is visible (to one)’), while in type (IIIb), we have two sentences about the scenes observed by the poet related to each other in terms of the cause and its effect. We still can detect here a trace of inference by abduction, albeit in a somewhat transformed guise:
18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’
425
[ENCOUNTER WITH SOMETHING EXTRAORDINARY]: The poet notices that the cranes go crying toward the reedy shore. [INVOKING A PLAUSIBLE ‘RULE’]: When the tide rises (and consequently the coastal lagoons are lost), the cranes move from there toward the reedy shore. [INFERENCE]: The poet infers, on the basis of his posited rule, that it is because of the rising tide that the cranes move toward the reedy shore. We can even hear the poet soliloquizing, using a sentence in the -noda construction: Turu ga asibe o sasite tonde iku – sio ga mititekita noda ‘The cranes are flying toward the reedy shore – I see that the tide is rising.’ But notice that the wording of the waka-piece is not structured in line with the logic of abduction but in such a way as to be referring only to the posited ‘rule’, which by so being singled out, gives an impression as if it is a law of nature. Incidentally, (8) can also profitably be compared with (9) below, in which a closely similar situation is construed in terms of type (IIb): (9)=(IIb´)
Sakurada pye tadu nakiwataru Ayutigata sipo Sakurada all cranes cry.go.across Ayuchigata tide pwi-ni-kyerasi tadu naki-wataru ebb-perf-mpst.pres cranes cry-go.over.concl ‘In Ayuchigata the tide seems to be ebbing, for the cranes go crying toward Sakurada’ (MYS 3.271)
A possible internal monologue by the poet here will be: Turu ga Sakurada e tonde iku – sio ga mitite kita rasii ‘Cranes are flying toward Sakurada – it seems that the tide is ebbing in Ayuchigata’. We thus see that type (IIb) and type (IIIb) are related to each other, representing two different construals of one and the same pair of situations: underlying type (IIb) is the logic of abduction (A because B), inferring the cause (A) from what is supposed to be its effect (B) on the basis of a subjectively posited ‘rule’, while type (IIIb) assumes that the cause and effect relationship (if A, then B) between a pair of situations is fully valid and presents them as if they represent a natural course of events. Insofar as the posited rule is not necessarily infallible, poetic pieces of the former kind sound very much like human, while those of the latter kind give an impression of being instances of divine providence, over and above human involvement, reminiscent of the kind of solemnity with which the earlier poems on kunimi and jokeika in general are more or less characterized. But notice again that there is a crucial difference between the earlier jokeika and the latter set of apparent jokeika which has evolved from the tradition of waka-pieces with the underlying logic of abduction. What we find commonly in the earlier jokeika is a striking ‘mirativity’ effect of the scenery, which totally overwhelms the author, who loses himself in its presence. Waka-pieces of type (IIIb), by contrast, give us a different impression. The author is admiring the scene, it is true, but it is not the case that he is simply overwhelmed by it. He already knows
426
Yoshihiko Ikegami
some of the ways nature works (which he encodes in terms of ‘if A, then B’). But he also knows that nature works totally beyond human control and so he is not a total stranger to the feeling of awe toward nature, either. Waka-pieces of type (IIIb) seem to be characterized with such mixed feelings about nature underlying them.
3.4.4 Metonymically juxtaposing the image of the scene before one’s eyes with the image of the scene in one’s mind’s eye In the previous sections, we have reviewed a set of major patterns on which the wakapieces (especially, those involving, either explicitly or implicitly, an act of seeing) in Man’yōshū are composed. There are two major points which have emerged from the review. The first one is that seeing is not to be taken as a simple act of visual perception. It is associated with a number of heavily culture-laden ideas, making possible communions of various kinds between the subject (i.e. one who sees) and the object (i.e. someone or something seen), especially in the context of subjective (or ‘subject-object merger’ type of) construal. The second point is the frequently employed device of construing the situation in terms of double vision, juxtaposing the image of the scene before one’s eyes with the image of the scene in one’s mind’s eye. The two images juxtaposed are related to each other through metonymy, or in terms of contiguity – spatially, temporally, or causally. Isn’t it possible to regard such a device as an extension, in an artistically elevated form, of the still earlier tradition of seeing something as a ‘sign’ of something else? Notice that at the earlier stage, either of the two images associated are typically a jikkei (‘real’ scene). In the case of those waka-pieces which are classifiable as jokeika, the poet is, first of all, supposed to be (either actually or fictively) witnessing a scene (jikkei A) before his eyes and he associates it with another (supposedly) real scene (jikkei B) which is either spatially or temporally distanced from HERE / NOW and hence he cannot possibly witness before his eyes but which he could very well witness, if he were duly displaced. In the case of these waka-pieces structured on the logic of abduction, the poet starts by witnessing a (mirative) scene (jikkei A), which he construes as an effect caused by some event (jikkei B) which he could have well witnessed and described, if he had been duly displaced. In any case, the poet is supposed to be dealing with two jikkei, images of something real, whether seen or unseen, which are juxtaposed as metonymically related (spatially, temporally, or in terms of cause and effect).
3.5 Metaphorically superimposing the image of the scene in one’s mind’s eye on the image of the scene before one’s eyes The shift from the stage described in the final part of the preceding section to the next stage with the flourishing technique of ‘mitate’ is effected by introducing a patently fictive image as the second member of the pair of evoked images, specifying, at the same
18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’
427
time, that the two images refer, in fact, to one and the same situation, in other words, that the two evoked images are to be interpreted as being in metaphorical relationship to each other – as being one superimposed on the other, so to speak.
3.5.1 Superimposition in a simile-like manner But, first, there is one point to be mentioned which marks the period of transition preceding the age of ‘mitate’. This concerns the use of the phrase, ‘. . . to miru made ni’ (to such an extent that it appears as if . . .), placed between the real image and the fictive image and serving semantically to connect the former to the latter: (10) ume no pana yeda ni ka tiru to20 miru plum gen blossom branch dat q.foc scatter cop.inf see madeni kaze ni midarete yuki zo puri-kuru so.much.so wind dat scatter.ger snow foc fall-come.adn ‘Snowflakes are scattered in the wind, such that it appears as if plum blossoms have been blown off the branches’ (MYS 8.1647) See also 5.839, 844, 7.1182, 8.1420, 1598. 20 Instead of straightforwardly superimposing the fictive image on the real image, we have here the phrase, ‘. . . to miru made ni’ serving to indicate explicitly that the two images are similar to each other. The function of the phrase in question here is comparable to that of the word like in simile: ‘B to miru made ni A’ is structurally parallel to ‘A wa B no yoo da’ (A is like B). Thus, an instance like (4) can be considered a piece of ‘mitate’ at the simile level.21 Cf. also 10.1841, where the cognitive verb omopu is used in a way equivalent to the verb of visual perception miru. A waka-piece like the following, in which the author softens the positive tone of his ‘mitate’-oriented statement by the particle kamo (which implies ‘wondering if . . ., asking myself if . . .’), may also be regarded as an instance of ‘mitate’ at the ‘simile’ level:
20 Note the function of the particle to here. It is a marker of ‘subjectiveness’ of the speaker’s judgment, quite comparable to the function of the English particle as, which accompanies verbs like regard (cf. ‘regard the man as eccentric’) and construe (cf. ‘construe his silence as consent’). ‘Regard A as B’ is normally translated into Japanese as A o B to minasu. Cf. also Aoki (1996: 41, note 1), where to miru, as contrasted with wo miru is discussed. 21 It is sometimes pointed out (e.g. H. Suzuki 1968: 172–174) that the use of the device of mitate in waka was triggered by the rhetorical device of comparison in classical Chinese poetry. It is interesting, however, to note that the device in waka soon developed into a pattern of a real image and a fictive image being directly opposed to each other, while the corresponding device in Chinese poetry continued to keep a reminder (e.g. utagapuraku pa ‘although I doubt (it)’) that one of the juxtaposed images was patently fictive (cf. H. Suzuki 1968: 177–178).
428
(11)
Yoshihiko Ikegami
wa ga sono ni ume no pana tiru pisakatano I gen garden dat plum gen blossom scatter.concl eternity.gen ame ywori yuki no nagare-kuru kamo. heaven abl snow gen drift-come.adn conj ‘In my garden, plum blossoms fall scattering around. May these be snowflakes drifting down from the remote heavens?’ (MYS 5.822) Cf. 10.2052.
In 5.844 (referred to above in relation to (3)), ‘kamo’ occurs with ‘to miru made ni’ in one and the same piece. Cf. also 8.1420, where the particle of questioning, ‘somo’ cooccurs with ‘to miru made ni’. Instead of simply wondering, however, the poet may offer an account for what he thought he saw, as in the following: (12)
misonopu no momokwi no ume no tiru pana garden gen hundred.trees gen plum gen scatter.adn blossom no ame ni tobi-agari yuki to puri-kye-mu gen heaven dat jump-rise.inf snow cop.inf fall-spst-conjec.adn ‘Blossoms scattering from hundreds of plum trees in the garden – they must have flown upward to the heavens and have come down as snowflakes’ (MYS 17.3906)
When we have no mediating words relating the image before one’s eyes to the fictive image in one’s mind’s eye and the two images are simply mixed as referring to one and the same situation, we have regular mitate pieces – ‘mitate’ at the metaphor (instead of simile) level. Such are the ‘mitate’ pieces characteristic of Kokin Wakashū (edited and presented to the emperor in 905): (13)
kasumi tati ko-no-me mo paru no yuki pureba mist rise.inf tree-gen-bud etop spring gen snow fall.conj pana naki sato mo pana zo tiri-keru blossom without village etop blossom foc scatter-mpst.adn ‘In spring, when the mist lies and the buds are swelling, it snows – then the blossoms fall scattering in the village with no blossoming trees’ (Kokinshū 1.9)
The piece is headed with a foreword, “On seeing the snow fall”, which means that ‘falling of the snow’ is the image before the poet’s eye. Over this is superimposed an image of ‘scattering blossoms’, which is fictive and seen by the poet in his mind’s eye. Note, however, that although the poet could very well have started by actually watching the falling snow, the piece can hardly be classified as jokeika of the kind in Man’yōshū. Nothing like the effect of mirativity can be associated with the image of falling snow here. The poet is rather concerned here with comparing it to a highly distinct image of
18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’
429
scattering blossoms (which contrasts sharply with the former) and leading the reader into realizing something he has never thought of. What is purported here to be impressive is no longer the magnificence of the scene before one’s eyes but the wittiness with which the comparison is made. With this shift in interest on the part of the author, it is no longer crucially important that the author is fully involved in what he sees before his eyes. In (13), for example, the author may very well not have seen the falling snow before his eyes – he may very well have seen it only in imagination, in his mind’s eye. This will give the author a far freer hand than before with selecting the second image to which the first one is to be compared. He can now let his imagination work in much more unconstrained ways.
3.5.2 Toward a full-scale and elaborate superimposition – ‘mitate’ At this point, a further new device is introduced which will allow the author to behave still more freely in his imagination – the device of kakekotoba, a play of words in which the author uses words and phrases in double sense. For example, in (13) quoted above, the word paru is to be understood both in the sense of ‘swell’ (in relation to the noun, konome, which precedes it) and in the sense of ‘spring’ (in relation to the noun, yuki, which follows it). The image of the buds swelling on the trees leads naturally to the image of the blossoming trees, making it possible to superimpose the fictive image of scattering blossoms on the real image of falling snow. Thus, the device of kakekotoba on the word paru mediates between the real and the fictive image, which, by themselves, are simply incompatible to each other. ‘Mitate’ can now be practiced not necessarily on the basis of perceptual similarities but on totally fortuitous homonymies at the level of words and phrases. While this allows the authors to play with a far wider range of freer and novel associations, their works tend to be estranged from the sincere feelings they entertain and wander away toward the more elaborate linguistic manipulations and technical novelties. Let me conclude by quoting the well-known piece of this kind by Ono no Komachi in Kokinshū: (14)
pana no iro pa uturi-ni-keri na itadurani blossom gen colour top change-perf-mpst excl vainly wa ga mi yo ni puru nagame se-si mani I gen body world dat become.old gaze do-spst.adn while ‘The color of the blossoms has vainly faded away, while I grow old, vainly watching the world’ (Kokinshū 2.113)
The author laments over the irresistibly failing beauty of herself, comparing it to the inexorably fading color of blossoms in such a way that the two images are intermingled, either being superimposed on the other. They are closely tied up together through
430
Yoshihiko Ikegami
twofold kakekotoba – with the word, puru we have the meaning of ‘getting old’ and the meaning of ‘fall’ overlapping, the former being related to the word, wagami ‘myself’ preceding it and the latter to nagame (a shortened form of naga-ame ‘long rain’) following it. Secondly, the word nagame also has two meanings overlapping, one being ‘long, persistent rain’ (interpreted as shortened form of naga-ame) and the other ‘watching’ (interpreted as nominal form of the verb, nagamu) and the former connects to the intransitive predicate sesi (the verb su ‘do’ + the auxiliary ki (experienced past), while the latter relates to the subject, wagami ‘myself’. In either case we find an event in the human world contrasted with an event in the natural world and the two series of events, human and natural, are tied together closely through the double meaning on nagame, one referring to the human (i.e. ‘watching’) and the other to the natural (i.e. ‘long rain’). It is easy to see that a waka-piece so elaborately structured, will pose a real challenge to translators. Let me conclude my article by presenting several pieces of attempted translation in English: The colours of the blossoms fair Are faded now and gone for e’er, While the long rain falling on; The beauties of my looks are gone Now that I grow old, left alone. – T. Wakameda Alas! The colours of the flowers Have faded; the long continued rain My beauty aging, too, as in this world I gaze, engrossed, on things that were but vain. – Asataro Miyamori The fairest flowers most quickly fall, Beaten with rain . . . And yet more vain, My beauty fadeth once for all And will not bloom again. – C. H. Page The flowers and my love Passed away under the rain While I idly looked upon them; Where is my yester-love? – Yonejiro Noguchi As certain as color Passes from the petal Irrevocable as flesh, The gazing eyes fall through the world. – Kenneth Rexroth
18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’
431
The color of these flowers No longer has allure, and I am left To ponder unavailingly The desire that my beauty once aroused Before it fell in this long rain of time. – Robert H. Brower and Earl Miner The lustre of the flowers Has faded and passed, While on idle things I have spent my body In the world’s long rains. – Geoffrey Bownas and Anthony Thwaite Alas! The beauty of the flowers has faded and come to nothing, while I have watched the rain lost in melancholy thought. – Helen C. McCullough the colors of the blossoms have faded and passed as heedlessly I squandered my days in pensive gazing and the long rain fall – Laurel R. Rodd The flowers withered, Their color faded While meaninglessly I spent my days in the world And the long rains were falling. – Donald Keene
The device of ‘mitate’ is, in later ages, to go over the boundary of language art and spread to a number of other fields of Japanese culture as well, in such a way that can be described in terms of ‘homology’. For discussions on Japanese culture from this perspective, see Ikegami (1986, 1989, 1998, 2009, 2020b and also Ikegami (ed.) 1991).
432
Yoshihiko Ikegami
References Aikhenvald, Alexandra Y. 2004. Evidentiality. Oxford: Oxford University Press. Aikhenvald, Alexander Y., and Anne Storch. 2013. Linguistic expression of perception and cognition: A typological glimpse. In Alexander Y. Aikhenvald and Anne Storch (eds.), Perception, cognition in language and culture, 1–45. Leiden: Brill. Aikhenvald, Alexander Y., and Robert M. W. Dixon (eds.). 2014. The grammar of knowledge: A cross-linguistic typology. Oxford: Oxford University Press. Aoki, Takao. 1996. ‘Mitate’ no bigaku [Aesthetics of ‘mitate’]. Nihon no Bigaku 24. 36–62. Tokyo: Perikansha. Arima, Michiko. 1991. Creative interpretation of the text and the Japanese mentality. In Yoshihiko Ikegami (ed.), The empire of signs: Semiotic essays on Japanese culture, 33–56. Amsterdam: John Benjamins. Blakemore, Diane. 1994. Evidence and modality. In Ronald E. Asher (ed.), Encyclopedia of language and linguistics, vol. 3, 1183–1186. Chomsky, Noam. 1964. Current issues in linguistic theory. The Hague: Mouton. Delancey, Scott. 1997. Mirativity: The grammatical marking of unexpected information. Linguistic Typology 1. 33–52. Frellesvig, Bjarke. 2010. A history of the Japanese language. Cambridge: Cambridge University Press. Fujii, Sadakazu. 2012. Bunpōteki shigaku [Grammatical poetics]. Tokyo: Kasama Shoin. Gomi, Yasuyoshi. 1939. Man’yōshikukō: ‘Miyu’ no mondai [Studies in the lexicon of Man’yōshū: The problem of ‘miyu’]. Kokugo Kokubun 9(12). 39–93. Hirasawa, Ryusuke. 1999. Kokin kafū no seiritsu [The establishment of Kokin-style]. Tokyo: Kasama Shoin. Hisajima, Shigeru. 1992. Jōdai no ‘~ shūshikei + miyu’ bun no kōzō ni tsuite [On the structure of the Old Japanese ‘conclusive form + miyu’ sentence]. In Tsujimura Kyōju Koki Kinen Ronbunshū Kankōkai (ed.), Nihongoshi no shomondai [Topics in the history of Japanese], 193–210. Tokyo: Meiji Shoin. Ide, Itaru. 1995 [1981]. Jōdaigo ‘miyu’ no jodōshiteki seikaku [Auxiliary features of miyu in Old Japanese]. In Itaru Ide, Yūbunroku: Kokugoshihen [Miscellanies: History of the Japanese Language] 1, 245–253. Tokyo: Izumi Shoin. Ikegami, Yoshihiko. 1986. Semiotic mechanism of the empty center in Japanese culture and society. Asian Studies Association of Australia Review 10(1). 7–14. Ikegami, Yoshihiko. 1989. Homology of language and culture: A case study in Japanese semiotics. In Walter Koch (ed.), The nature of culture, 388–403. Bochum: Studienverlag Dr. Nobert Brockmeyer. Ikegami, Yoshihiko. 1998. Sign conceptions in Japan. In Roland Posner, Klaus Robering and Thomas A. Sebeok (eds.), Semiotics: A handbook on the sign-theoretic foundations of nature and culture, vol. 2, 1898–1910. Berlin/New York: Mouton De Gruyter. Ikegami, Yoshihiko. 2005. Indices of a subjectivity-prominent language: Between cognitive linguistics and linguistic typology. Annual Review of Cognitive Linguistics 3. 132–164. Ikegami, Yoshihiko. 2008. Subjective construal as a ‘fashion of speaking’ in Japanese. In María de los Ángeles Gómez González, J. Lachia Mackenzie and Elsa M. González Álvarez (eds.), Current trends in contrastive linguistics, 227–250. Amsterdam: John Benjamins. Ikegami, Yoshihiko. 2009. Jinbungaku kenkyū ni okeru sagyōkasetsu to shite no ‘sōdōsei’ [‘Homology’ as a working hypothesis in the humanities]. Eibungaku Kenkyū: Shibu-tōgōgō 2. 93–107. Ikegami, Yoshihiko. 2014. Superimposing the image in one’s mind’s eye on the image of the scene before one’s eyes – A favourite type of construal by Japanese speakers. A paper presented at the 14th Conference of the Association for Japanese Studies, University of Ljubliana, 27–30 August. Ikegami, Yoshihiko. 2015. ‘Subjective construal’ and ‘objective construal’: A typology of how the speaker of different languages behave differently in linguistically encoding a situation. Journal of Cognitive Linguistics 1. 1–21.
18 Vision and the verbs of visual perception in Man’yōshū: From mirativity to ‘mitate’
433
Ikegami, Yoshihiko. 2020a. hyōshiki to shiteno Man’yōshū ni okeru shūshikei-setsuzoku no miyu [miyu connecting to the conclusive form as evidential marker in Man’yōshū]. In Youko Yonekura, Osamu Yamamoto and Ryosaku Asai (eds.), Kotoba kara kokoro e – ninchi no shin’en: Yoshimura Kimihiro kyōju taishoku kinen ronbunshū [From language to mind – the depths of cognition: Festschrift for Professor Kimihiro Yoshimura on the occasion of his retirement], 2–13. Tokyo: Kaitakusha. Ikegami, Yoshihiko. 2020b. Homology, homologization and the empty center – An essay on cultural semiotics. In Hiroyuki Miyashita, Yasuhiro Fujinawa und Shin Tanaka (eds.), Form, Struktur, und Bedeutung: Festschrift für Akio Ogawa [Form, structure and meaning: festschrift for Akio Ogawa], 363–375. Tübingen: Stauffenburg Verlag. Ikegami, Yoshihiko (ed.). 1991. The empire of signs: semiotic essays on Japanese culture. Amsterdam: John Benjamins. Inuzuka, Wataru. 1973 [1955]. ‘Miyu’ no yōhō ni tsuite [On the use of miyu]. In Wataru Inuzuka, Ōchō biteki goshi no kenkyū [A study of the history of the aesthetic vocabulary of the court], 353–366. Tokyo: Kasama Shoin. Ishigaki, Kenji. 1955 [1942]. Joshi no rekishiteki kenkyū [A historical study of joshi]. Tokyo: Iwanami Shoten. Kageura, Shintaro. 1993. Miru – Man’yōshū ni okeru chikakuteki hyōgen no hen’yō [Miru: Changes in perceptual expressions in Man’yōshū]. Nihonbungaku Kenkyu 32. 112–119. Kanno, Tomiichi. 1985. Kunimiuta [The land-lauding poem]. Kokubungaku 30(11). 110–111. Kawai, Hayao. 1991. The “forbidden chamber” motif in a Japanese fairy tale. In Yoshihiko Ikegami (ed.), The empire of signs: Semiotic essays on Japanese culture, 157–180. Amsterdam: John Benjamins. Kitahara, Yasuo. 1965. ‘Nari’ to ‘miyu’ – jōdai no yōrei ni mieru iwayuru shūshikei shōsetsu no imisuru mono [‘Nari’ and ‘miyu’: What the so-called connection to the conclusive form, a usage found in Old Japanese, means]. Kokugogaku 61. 11–28. Kitayama, Masamichi. 1969. ‘Nagamu’ oboegaki [A note on ‘nagamu’]. Kokugo Kokubun 38. 86–99. Komatsu, Kozo. 1983. Jodōshi ‘rasi’ no imikinō to hyōgen [The semantic function and use of the auxiliary ‘rasi’]. Ehimedaigaku Hōbungakubu Ronsō 16. 25–52. Miyajima, Tatsuo, Tai Suzuki and Seiwa Abe (eds.). 2014. Nihon koten taishō bunrui goi hyō [A classified lexicon of Japanese classics]. Tokyo: Kasama Shoin. Morishige, Satoshi. 1951. Man’yōshū no ‘miru’: Buntaironteki kōsatsu [‘Miru’ in Man’yōshū: Stylistic considerations]. Man’yō 24. 1–31. Nagafuji, Yasushi. 1973. Kiki∙Man’yō ni okeru ‘miru’ koto ni tsuite [On ‘seeing’ in Kiki∙Man’yōshū]. Bungaku 41. 123–133. Nakanishi, Susumu. 1972. ‘Miru’ – kodaiteki shikaku [‘Miru’: Ancient visual perception]. Bungaku 43(4). 123–133. Otake, Misao. 1969. Genji monogatari ni okeru shikakuteki hyōgen no igi [The meanings of the visual expressions in Genji monogatari]. Joshidai Kokubun 52. 21–30. Peirce, Charles Sanders. 1931–1958. Collected Papers. Cambridge, MA: Harvard University Press. Rattanaphanusorn, Rungthip and Kingkarn Thepkanjana. 2007. The grammaticalization of the verb of seeing as an evidential marker in Thai. Ninchi Gengogaku Ronkō 7. 273–301. Sakaguchi, Yoritaka. 1982. Shūshikei o ukeru ‘miyu’ shiken [A private view of ‘miyu’ connecting to the conclusive form]. Beppu Daigaku Kiyō 23. 29–34. Sasaki, Nobutsuna. 2007. Man’yōshū no [‘I’ in Man’yōshū]. Tokyo: Kadokawa Gakugei Shuppan. Satake, Akihiro. 1980 [1964]. ‘Miyu’ no sekai [The world of ‘miyu’]. In Akihiro Satake, Man’yōshū Nukigaki [Miscellanies on Man’yōshū], 13–33. Tokyo: Iwanami Shoten. Sato, Kazuyoshi. 1989. ‘Yu’ [The auxiliary yu]. In Kodai Goshi Kankokai (ed.), Kodai Goshi [Lexicon of Ancient Japanese] II, 132–135. Tokyo: Ōfusha. Suzuki, Hideo. 1968. Kokinshū no mitate ni tsuite [On mitate in Kokinshū]. Bungaku 54(2). 167–179. Takahashi, Bunji. 1985. Fūkei to kyōkankaku [Scenery and synaesthesia]. Tokyo: Shunjūsha. Takeoka, Masao. 1970. ‘Keri’ to ‘ki’ to no imiyōhō [Meaning and use of ‘keri’ and ‘ki’]. Gekkan Bumpō 2(5). 46–55.
434
Yoshihiko Ikegami
Toyama, Tomonori. 1991. The notion of the sign in Japanese tradition. In Yoshihiko Ikegami (ed.), The empire of signs: Semiotic essays on Japanese culture, 25–32. Amsterdam: John Benjamins. Tsuchihashi, Hiroshi. 1965. Kodai kayō to girei no kenkyū [Studies in ancient songs and rituals]. Tokyo: Iwanami Shoten. Uchida, Kentoku. 1983. ‘Miru, miyu’ to ‘omopu, omopoyu’: Man’yōshū ni okeru sono sōkan [‘miru, miyu’ and ‘omopu, omopoyu’: Their correlations in Man’yōshū]. Man’yō 115. 1–32. Uchida, Kentoku. 1997. Shūshi-kei setsuzoku no ‘miyu’ ni tsuite [On ‘miyu’ connecting to the conclusive form]. In Yoshiaki Kawabata and Yoshio Nitta (eds.), Nihongo bunpō: Taikei to hōhō [Japanese grammar: system and methods], 115–132. Tokyo: Hituzi Syobō. Uehara, Satoshi. 1998. Pronoun drop and perspective in Japanese. Japanese/Korean Linguistics 7. 26–43. Velupillai, Viveka. 2013. An introduction to linguistic typology. Amsterdam: John Benjamins. Viberg, Åke. 1983. Verbs of perception: A typological study. Linguistics 21. 123–162. Whorf, Benjamin Lee. 1956. Language, thought, and reality. John B. Carrol (ed.). Cambridge, MA: MIT Press. Yoshii, Iwao. 1962. ‘Miru uta no hassō keishiki ni tsuite: ‘Miyu’ o chūshin ni [On the way of thinking in the poems of seeing: With special reference to ‘miyu’], Man’yō 45. 1–14.
Part IV: Lexicon, Materials and Kanbun
John R. Bentley
19 The history of basic vocabulary in Japanese 1 Introduction “Basic vocabulary” is the term used in linguistics to designate a core set of vocabulary that children first learn – body parts, kinship terms, numbers, and pronouns – which are generally considered to be highly resistant to borrowing. This set of vocabulary is important, because these terms are generally found in all cultures and languages, being grounded in the fundamental quality of human life. Most scholars in comparative linguistics have used this set of vocabulary as a starting point in their comparative linguistic work. However, as Thomason has pointed out, even basic vocabulary is prone to borrowing depending on the kind of language contact a certain language experiences. Of the four stages of contact on her borrowing scale, by the third stage of “more intense contact”, she notes that basic as well as non-basic vocabulary is prone to borrowing (Thomason 2001: 70–71). In categorizing the basic vocabulary of the dialects of the Japanese islands for study, Kuno (2005: 23) has set up three criteria: i) vocabulary found throughout the islands, ii) vocabulary found in every day speech, and iii) vocabulary also found in literary texts from earlier periods. Perhaps because the borrowing of basic vocabulary cannot be ruled out, much of the work on ‘basic vocabulary’ in Japanese has tended to focus on rigo (俚語 ‘common words’, as opposed to elegant, literary terms; cf. Kuno 2005: 15, Nakamoto 1981: 105). Thus, scholars have conducted rather extensive research on the names of insects, birds, and plants. From a pragmatic point of view, it is still important for scholars to focus on a broader, more extensive list of basic vocabulary, and in the following pages I have tried to take a balanced approach, illustrating the changes that have taken place over time, starting with the earliest period for which a large cache of written corpora is available. This period is known as Old Japanese (OJ). For many of the vocabulary dealing with botany and zoology, these words have come from two early Heian dictionaries: Shinsen jikyō (ca. 898 CE) and Wamyō ruijushō (ca. 935 CE, shortened to Wamyōshō), which chronologically belong to Early Middle Japanese (EMJ); however Shinsen jikyō still preserves the distinction between OJ kwo and ko, and many scholars generally take these data to be from the late OJ period. Wamyōshō has lost that distinction, but many of the words mirror that in Shinsen jikyō. Both dictionaries also preserve many words that disappear from our texts in the early Heian era. As such, I will label data from Shinsen jikyō as SJK, and data from Wamyōshō as WMS. Setting regular phonological changes aside, the trend in basic vocabulary over time has involved one of two changes: semantic change or substitution. In the majority of cases, substitution has involved a native OJ word being replaced by a Sino-Japanese (SJ) word. Due to spatial limitations, this chapter focuses on six lexical fields within basic vocabulary: body parts, botany, celestial bodies, kinship terms, numerals, and zoological terms. https://doi.org/10.1515/9781614512851-020
438
John R. Bentley
2 Body parts Many words for common parts of the body have remained relatively unchanged from the OJ period to ModJ, so I will only focus on words that have undergone either substitution or a semantic change (see the Basic Vocabulary list below).1 The OJ word for ‘body’ was mwi, and this has been replaced by karada or the SJ words nikutai, zindai~sintai. The OJ word is still found in some modern forms, such as miugoki ‘move, being able to move’. There is a doublet for ‘skin’, with OJ pada and OJ kapabe. The latter is now obsolete; there is now a new doublet, ModJ hada and SJ hihu. Starting from the top of the body: OJ kasira ‘head’ is replaced by atama around the late LMJ period. ModJ kasira refers to a boss or chief, ‘the head of a group’. SJK natuki refers to the skull or cranium, which is later replaced by SJ zugaikotu or koogaikotu. In OJ the four limbs were collectively called yeda ‘branches’, a form that was later replaced by te to asi ‘arms and legs’ or the SJ word sisi (四肢). In OJ there was a doublet for ‘fingers’, yubi and oyobi. Only yubi survives in the present, but in many languages of the Ryukyus the word for ‘finger’ goes back to an earlier form of oyobi. In OJ ‘the lines on the palm of the hand’ were called te no aya, but this has been replaced by a SJ hybrid, tesō. ‘Thumb’ in OJ was opo-oyobi, or literally ‘big finger’. This is later replaced by oyayubi ‘parent finger’ in the LMJ period. What we call the ‘ring finger’ was nanasi no yubi ‘the finger without a name’ in OJ, but this is later replaced by the term kusuri yubi ‘the finger [you taste, check] medicine with’. WMS konokami is ‘abdomen’, and is later replaced by LMJ sitabara. The ham, or back of the knee, is SJK yoporo, and this is later replaced by ModJ hikagami. OJ pagi ‘shin’ is replaced by sune. The OJ word still survives, however, in ModJ fukurahagi ‘calf’. Interestingly, OJ sune originally meant ‘bone marrow’. In OJ there was a doublet for ‘heel’, tubunaki or SJK tububusi, both of which have been replaced by another doublet: ModJ kakato or kurubusi. In OJ the ‘instep of the foot’ was SJK anapira, and this was replaced by ModJ asi no koo. Looking internally, the five organs are OJ kokoro ‘heart’ (modern SJ sinzoo), OJ kimwo ‘liver’ (modern SJ kanzoo), SJK yokosi ‘spleen’ (modern SJ hizoo), SJK pukupukusi ‘lungs’ (modern SJ hai), and SJK murato ‘kidneys’ (modern SJ zinzoo).
3 Botany I divide plants into two very generic groups: plants and trees, alphabetized by the English gloss, but I have kept related botanical items together. Terms with no comment after the OJ attestation have survived in the present, undergoing regular sound changes.
1 Much of the information of these six areas comes from Shinsen jikyō (1992) and Wamyōshō (1978). I have ignored the kō-otsu distinction, aside from kwo~ko, which is preserved in Shinsen jikyō. When I do make a kō-otsu distinction, it is based on attestations found in earlier texts.
19 The history of basic vocabulary in Japanese
439
3.1 Plants The generic word for grass-like plants is OJ kusa. ‘Aralia cordata’ is SJK udo or WMS tutitara, but the second term has died out. ‘Arrowhead’ is WMS kuwai, more commonly known as ModJ siroguwai. ‘Asian ginseng’ is SJK kanonikegusa. This is later replaced by ninzin ‘carrot’, so there has been semantic change and substitution. ‘Atractylodes japonica’ is SJK wokera. ‘Asparagus cochinchinensis’ is SJK sumarogusa, which is now known as kusasugi kazura. ‘Bamboo’ is OJ take. ‘Barley’ is OJ mugi, more commonly known as ModJ oomugi. ‘Bush clover’ is OJ pagi. ‘Candock’ is SJK ukigusa. ‘Cattail’ is OJ kama. ‘Chinese peony’ is WMS ebisu kusuri ‘Ebisu medicine’ or numi kusuri ‘oral medicine’ because of its medicinal properties, now known as SJ syakuyaku. ‘Chrysanthemum’ is WMS kapara yomogwi or WMS kapara-obaki. These are later replaced by a SJ form, kiku. ‘Coco grass’ is SJK kugu and is now known as ModJ siokugu. ‘Cogon grass’ is OJ ti. ‘Deutzia’ is OJ unopana, also known as ModJ utugi. ‘Dropwort’ has a doublet in SJK, pirumusiro or OJ seri, but later comes to be known as ModJ hamazeri. ‘Eleorchis japonica’ is WMS akamakusa, which has been replaced by several words: a hybrid, sawaran, or SJ takuran. ‘Eularias’ is OJ kaya. ‘Feather cockscomb’ is SJK umakusa, now known as nogeitoo. ‘Gastrodia’ has an OJ doublet: WMS wotowotosi or WMS kami no yagara. This is modern oni no yagara; the first OJ term has died out, and the second has undergone partial substitution, where kami ‘deity’ has been replaced with oni ‘ogre’. ‘Goosefoot’ is SJK akasa. ‘Houttuynia cordata’ is WMS sipuki ‘the bitter medicine’, but this has been replaced by SJ dokudami ‘stops poison’. ‘Iris’ is OJ ayamegusa, and ‘rabbit-ear iris’ is OJ kakitupata. ‘Japanese yellow rose’ is OJ yamabuki or WMS yamapubuki, but the first is likely an abbreviated form of the latter. ‘Knotweed’ is WMS inutade. ‘Miscanthus sacchariflorus’ is called OJ wogwi. ‘Lardizabalaceae’ is SJK akebi. ‘Leek’ is OJ mira or wms komira, which is now ModJ nira. ‘Lily’ is OJ yuri. ‘Madder’ is OJ akane. ‘Mugwort’ is OJ yomogwi. ‘Poisonous bracken’ is SJK oniwarabi, so called because it has matured beyond its edible state. ‘Orchid’ is OJ pudibakama, but this was later replaced by SJ ran. ‘Nettle’ is SJK ira. ‘Paris tetraphylla’ is WMS nubarikusa, now known as ModJ tukubanesoo. ‘Patrinia’ is OJ wominapyesi. ‘Polygatum odoratum’ is known by a doublet in OJ, amana and WMS wemikusa, but the name is now ModJ amadokoro. ‘Polygonatum sibiricum’ is OJ WMS opowemi or WMS yamawemi. The second name has died out, but opowemi is now known as ModJ naruko yuri. ‘Reed, rush’ is OJ asi, and ‘bulrush’ is OJ opowi ‘great rush’. This last term has been replaced by a semantic equivalent, ModJ hutoi ‘fat rush’. However, this is based on a semantic connection, and not a botanical one, because OJ WMS tukumo ‘Scirpus lacustris’ is actually ModJ hutoi. ‘Bur reed’ is SJK mukuri. ‘Candle rush’ is OJ wi. ‘Rice’ is OJ ine. ‘Rose’ is OJ ubara, later shortened to bara. A related name is OJ umara. ‘Schoenoplectus triqueter’ is SJK komokusa, later known as ModJ sankakui. ‘Sedge’ is OJ suge, and ‘mountain sedge’ is OJ yamasuge. ‘Sicklepod’ is SJK ebisukusa. ‘Small dragon root’ is WMS kakitubana, but in the present, it is known as musasi abumi. ‘Smilax’ is WMS sarutori. ‘Soy’ is WMS
440
John R. Bentley
kuromame, now more commonly known as SJ daizu. ‘Spiderwort’ is OJ tukikusa, later known as ModJ tuyukusa. ‘Taro’ is OJ umo, now commonly known as ModJ satoimo. ‘Wasabi (Japanese horseradish)’ is WMS wasabi. ‘Wheat’ is WMS komugi. ‘Wild pink’ is OJ nadesikwo. ‘Yam’ is WMS yamatuimo or WMS yamanoimo, with the latter surviving in the present. ‘Zizania latifolia’, also known as Manchurian wild rice, is OJ komo, and is ModJ makomo.
3.2 Trees The generic word for ‘tree’ is OJ kwi (> ModJ ki), and many of the names of trees include it. ‘Aphananthe aspera’ is OJ muku. ‘Ash tree’ is SJK tonerikonoki, but Wamyōshō also lists an early SJ reading, tamunoki ‘the tan tree’. ‘Black persimmon’ is WMS kurokaki. ‘Box tree’ is OJ tuge. ‘Camellia’ is OJ tubakwi. ‘Camphor’ is OJ kusu or kusunoki. ‘Catalpa’ is OJ adusa, and ‘Catalpa ovata’ is OJ pisakwi. ‘Cedar’ is OJ sugwi. ‘Cherry tree’ is OJ sakura. ‘Chinese black pine’ is OJ makwi. ‘Cleyera japonica (sakaki)’ is OJ sakakwi. ‘Cypress’ is OJ pi, and ModJ hinoki. ‘Elm’ is SJK yanire, which later is shortened to nire. ‘Eurya japonica’ is called WMS pisakaki. ‘Fir tree’ is WMS momi, and ‘temple juniper’ is OJ momu. ‘Holly tree’ is OJ pipiragwi; the genus of this tree is actually Osmanthus heterophyllus, while holly belongs to the genus Ilex. ‘Horse chestnut’ is SJK toti, now ModJ totinoki. ‘Japanese angelica tree’ is OJ tara. ‘Japanese bead tree’ is OJ aputi, but this is later replaced by SJ sendan. ‘Japanese pagoda tree’ is SJK wenisu, but this has undergone a sound change to enzu. ‘Japanese star anise’ is OJ sikimi. ‘Katsura tree’ appears in Man’yōshū as katura, but Wamyōshō preserves a doublet based on gender, wokatura ‘male katura’ and mekatura ‘female katura’. The latter form now refers to a cinnamon tree. ‘Mandarin orange’ is OJ tatibana, and ‘Trifoliate orange’ is known as OJ karatati, which is a shortened form of karatatibana ‘Chinese orange’. ‘Maple’ in OJ is kaperute, but Wamyōshō also records the LOJ form kapirutenoki. The latter name has died out; in modern Japanese, it is kaede. ‘Mulberry’ is OJ kupa, ‘mountain mulberry’ is OJ tumi, and ‘paper mulberry’ is OJ kadi, later known as kadinoki. ‘Nettle tree’ is OJ e, later known as enoki. ‘Oak’ is OJ kasi and ‘Japanese oak’ is OJ nara. ‘Oleaceae’ is WMS nezumimotinoki, but in the present, it is known simply as nezumimoti. ‘Pear’ is OJ nasi. ‘Pine’ is OJ matu. ‘Plum’ is OJ ume and WMS mume. ‘Pomegranate’ is WMS sakuro, which is ModJ zakuro. ‘Quercus dentate’ is OJ kasipa. ‘Scutellaria baicalensis Georgi’ is SJK kisa, ModJ kogane yanagi. When used for medicinal purposes, it is known as oogon. ‘Spindle tree’ is OJ mayumi. ‘Vaccinium bracteatum’, a species of the same family as the rhododendron, is OJ sasibu, now called syasyanbo. ‘Viburnum plicatum’ is WMS pemi ‘snake tree’, but it has been replaced by ModJ yabudemari. ‘Willow’ is OJ yanagwi; ‘pussy willow’ is kapa yanagwi, and ‘weeping willow’ OJ sidari yanagwi.
19 The history of basic vocabulary in Japanese
441
4 Celestial bodies The following words dealing with celestial bodies are preserved in Old Japanese texts. Only a few have been replaced by other words. ‘Firmament, heaven’ is OJ ame, now commonly called SJ ten or tengoku. ‘Sun’ is OJ pi. ‘Moon’ is OJ tukwi, while a ‘crescent moon’ is WMS yumibari ‘a bow pulled (as if to shoot an arrow)’, now ModJ mikazuki. ‘Full moon’ is OJ motidukwi, now more commonly called SJ mangetu. ‘Solar or lunar halo’ is WMS kasa. ‘Star’ is OJ posi. There are several OJ words for ‘comet’: pabakibosi ‘broom star’ and yuputudu; the first term is ModJ hookibosi, but the second term has been replaced by SJ suisei. ‘Shooting star’ is WMS yobapi posi ‘evening star’, which has been replaced by ModJ nagarebosi ‘flowing star’. ‘Sky’ is OJ swora. ‘The Milky Way’ is known as OJ amanokapa. ‘The Pleiades’ is OJ subaru. ‘Venus’ is OJ akaposi ‘bright star’, but when Venus appears in the evening sky it is known as OJ yupututu. It is now known as either SJ kinsei or SJ myoozyoo; the latter in Chinese could mean either Venus or a comet. It is interesting that Wamyōshō records that ‘comet’ is yupututu, suggesting that perhaps the original meaning of this word was yupu ‘evening’ tutu ‘cylinder’.
5 Kinship terms We are fortunate that Ryō no shūge (ca. 880), a commentary on the Taihō and Yōrō administrative codes of the Nara era, contains a list of kinship terms, found preserved in an older commentary called Koki ‘Old Record’, dated to roughly 738 CE.2 This list is contained in the protocol regarding funerals and periods of mourning, and consists of 24 kinship terms (cf. Kuroita 1955: 971–973). Many of these terms are built around basic etyma for ‘father’, ‘mother’, and ‘person’. Consider Table 1: Table 1: Kinship Terms. Paternal
Maternal
Other
ti titi ‘father’ opo-ti ‘grandfather’ opopo-ti ‘great grandfather’ wodi ‘uncle’ mama-titi ‘stepfather’
pa papa ‘mother’ opo-pa ‘grandmother’ opopo-pa ‘great grandmother’ woba ‘aunt’ mama-papa ‘stepmother’
Here opo- ‘great’, opopo- ‘great, great’, and wo ‘small’ are attached to show the direct or indirect relationship to the father and mother. Other terms are built with gender
2 For this dating see Yamanaka and Morita (1991: 130–131).
442
John R. Bentley
markers, OJ kwo ‘child, son’ and OJ imwo ‘sister’: itokwo ‘male cousin’ versus itogimwo ‘female cousin’. Also wopi ‘nephew’ and myepi ‘niece’. Here wo is a male and mye is ‘female’. Terms for brothers are missing from the list, as male relatives were subsumed under the other terms, though ‘older brother’ is given the reading kami no kwo ‘upper son’. Other texts note that ‘older brother’ is OJ ani, ‘older sister’ OJ ane, and ‘younger brother’ is OJ otopito, literally ‘younger person’. There is a doublet for ‘younger sister’, oimo and imouto. It is instructive that the doublet omo ‘mother’ is not used as the foundation of other maternal terms. Ryō no shūge mentions the following about the term ‘widow’, “When the woman [the wife] has been left by the patriarch [because he has died], she becomes a widow. In the vernacular this is known as titi ni yamarenitaru omo”.3 Considering that omo later refers to a ‘wet nurse’, it is probable that originally papa referred to one’s biological mother, while omo was a non-biological mother, or one who is adopted (married) into a family. It should be mentioned that Vovin (2010: 92–93) has argued, however, that one of the pair is a loan from a language on the peninsula (his ‘Korean’), while the other is a native Japanese word. Aside from regular sound changes, most of these terms continue in the present. Exceptions include the use of hii (< OJ pipi) for the reduplication of opo, hii-oozi ‘great grandfather’. Also, the gender distinction for ‘cousin’ was lost, and itoko now refers to all cousins. Finally, the native designation for ‘in-law’ has been replaced by SJ or hybrid forms: giri no titi or gifu ‘father-in-law’, and giri no haha or gibo ‘mother-in-law’.
6 Numerals The OJ numeral system was built on a common system of a base of ten numerals. As Frellesvig (2010: 289) has pointed out, these numerals are “partly based on vowel alternations to show doubling”. Thus, pito ‘1’ and puta ‘2’, mi ‘3’ and mu ‘6’, and yo ‘4’ and ya ‘8’. Larger numbers were created with the addition of ‘base numbers’. ‘Twenty’ was apparently a special usage, because it has its own form, pata. Related to towo ‘ten’ is the first base, swo meaning ‘tens’. Thus ‘30’ was miswo, and ‘50’ was iswo. ‘One hundred’ was mwomwo, but the base po was added to a basic number to create ‘hundreds’: yopo ‘400’ or yapo ‘800’. ‘Thousand’ was ti and functioned both as a numeral and a base number: putati ‘2000’ and yati ‘8000’. The final base was yorodu ‘10,000’. Few examples of large numbers exist in our texts, but a well-known number is yayorodu ‘80,000’. Nihon shoki preserves an example of a very large number, ‘1,792,470 years’, glossed as mwomwo yorodu tose amari nanaswo yorodu tose amari kokono yorodu tose amari putati tose amari yopo tose amari nanaswo tose amari ‘one hundred-ten thousand years seventy-thousand years nine-ten thousand years, two thousand years four hundred years and seventy years’. 3 Literally ‘the mother who had become (a victim) of the illness of the father’.
19 The history of basic vocabulary in Japanese
443
This entire cardinal system later became obsolete, replaced by the less cumbersome system of Sino-Japanese, contrasted in Table 2. Table 2: Number Systems of Japanese. #
Native
Sino-Japanese
1 2 3 4 5 6 7 8 9 10 100 1000 10,000
pito puta mi yo itu mu nana ya kokono towo mwomwo ti yorodu
iti ni san si go roku siti pati kiu zipu pyaku sen man
7 Zoological terms Because of the diversity of animal life, this subsection focuses on three groups of animals: beasts, birds, and fish. The generic OJ term for ‘beast’ is kemono ‘furthing’. Wamyōshō records, “Quadrupeds with fur are called kemono, but the six domesticated animals (chicken, dog, horse, ox, pig, and sheep) are (collectively) called kedamono”.
7.1 Beasts Male animals are OJ wokemono, and females are mekemono. These are later replaced in the late Edo period by ModJ wosu ‘male’ and mesu ‘female.’ ‘Badger’ has an OJ doublet, SJK muzina and SJK mi (< ✶mu-i), suggesting that these are formed from the same etymon, mu-. ‘Bear’ is OJ kuma, and ‘brown bear’ is OJ sikuma. ‘Boar’ is OJ wi, and ‘field boar’ is known as OJ kusawi. ‘Camel’ is WMS rakudanouma, ‘the lakda horse’, now SJ rakuda. ‘Cat’ is OJ nekwo or WMS nekoma. There are two related OJ words for ‘dog’, wenu and inu, as in mukuke inu ‘shaggy dog’. Wenu is a contraction of wo-inu ‘little dog’. ‘Donkey’ is OJ usagiuma, ‘a rabbit horse’, later replaced by SJ roba. ‘Elephant’ is OJ kisa, but this was replaced by SJ zoo. ‘Fox’ is OJ kitune. ‘Giant flying squirrel’ has an OJ doublet: musasabi and WMS momi, though the later term has died out in the modern language. ‘Gibbon’ is WMS mata, a word that has died out of the language, replaced by tenagazaru; ‘monkey’
444
John R. Bentley
is OJ saru. ‘Horse’ is OJ uma. A possible doublet is OJ kwoma, often interpreted as ‘colt’, but as this word also appears in the language of Paekche (cf. Bentley 2000: 430), it may actually be a peninsular word for ‘horse’ that was later reanalyzed in OJ as kwo ‘small’ uma ‘horse’. ‘Mole’ is WMS ukoromoti¸ and this is later replaced by ModJ moguramoti. ‘Otter’ is WMS woso, later replaced by ModJ kawauso ‘river otter’. ‘Ox’ is OJ usi. ‘Panther’ is OJ nagatukami, later replaced with SJ hyoo. ‘Rabbit’ is OJ usagi. ‘Raccoon dog’ is WMS tanuki. ‘Rat’ is OJ nezumi, while ‘mouse’ is WMS norane, but is now known as hatuka nezumi; ‘gray mouse’ is WMS turaneko, later replaced by ModJ turanezumi. ‘Sable’ is WMS temu, SJ ten. ‘Black sable’ is WMS purugi, but this has been replaced with ModJ kuroten. ‘Sheep’ is OJ pituzi. ‘Spotted deer’ is OJ ka, with ‘buck’ OJ sawosika and ‘doe’ OJ meka. Related to this is OJ opozika ‘elaphus’. Wamyōshō notes, “This animal resembles a spotted deer, but without spots on its fur”. In the modern language ‘deer’ is sika. ‘Tiger’ is OJ tora (< ✶twora). ‘Wolf’ is OJ opokami.
7.2 Birds The generic OJ word for ‘bird’ is tori; the male is known as wotori and the female metori. ‘Baikal teal’ is adi. ‘Black mallard’ is WMS kurwotori. ‘Bulbul’ is OJ piedori, ModJ hiyodori. ‘Bunting’ is OJ sitoto, but this has been replaced by ModJ hooziro or aozi. ‘Bush warbler’ is OJ ugupisu, while ‘Korean warbler’ is OJ tuki. ‘Cormorant’ is OJ u. There is a doublet for ‘crane’ in OJ: turu and tadu, and this doublet survives in the present. In OJ tadu is only found in poetry (cf. Omodaka 1967: 428). ‘Crow’ is OJ karasu. ‘Cuckoo’ is OJ pototogisu, and ‘Oriental cuckoo’ is WMS pupudori. ‘Dabchick’ is OJ nipo. ‘Eagle’ is WMS opowasi. ‘Grebe’ is WMS tuburi. There is an OJ doublet for ‘Haw finch’, pime and WMS sime. ‘Hawk’ is OJ taka, ‘hawk eagle’ SJK kumataka, ‘accipiter’ WMS pasitaka, ‘sparrow hawk’ WMS nose, and ‘small hawk’ WMS tumi. Nose dies out, replaced by haitaka. ‘Heron’ is OJ sagi, and ‘night heron’ is WMS imi ~ ibi,4 but this was replaced by ModJ gosai sagi. ‘Japanese goatsucker’ is SJK yotaka, while ‘Japanese grosbeak’ is OJ ikaruga, which has been shortened to ModJ ikaru. ‘Kingfisher’ is WMS sopi, but is now known as ModJ mebaru. ‘Kite’ is OJ tobi. ‘Magpie’ is OJ kasasagi. ‘Osprey’ is OJ misagwo. ‘Mandarin duck’ is OJ wosi, and this later becomes osidori. ‘Owl’ has an OJ doublet, SJK pukuropu and WMS sake, and the second form has died out. ‘Horned owl’ is SJK tuku. ‘Peregrine falcon’ is OJ payabusa. ‘Pheasant’ has an OJ doublet, kigisi and kizi, which are related forms, and ‘golden pheasant’ is OJ yamadori. ‘Pigeon’ is OJ yama no patwo, while a ‘domesticated pigeon’ is SJK ipebato. In ModJ these are collectively known as hato. ‘Plover’ has what
4 Wamyōshō (1978: 18, 7b) quotes Benshiki ryūjō which has 鵁伊微 where 微 can be read either mi or bi. As the intended reading is unclear, I give both.
19 The history of basic vocabulary in Japanese
445
appears to be a related doublet, SJK tadori and OJ tidori, both which persist in the present as ModJ tidori and tageri. ‘Quail’ is OJ udura, and another type of quail is called SJK kayakuki. ‘Sea gull’ is OJ kamame. ‘Skylark’ is OJ pibari. ‘Sparrow’ is OJ suzume. ‘Swallow’ is WMS tubakurame, which is often shortened to tubame. ‘Swan’ had a doublet in OJ, kukupi and WMS kopu; both have been replaced by SJ hakutyoo ‘white bird’. ‘Teal’ is OJ takape, which is later replaced by ModJ kogamo. ‘Waterrail’ is OJ kupina. ‘Woodpecker’ is SJK teratuki or WMS teratutuki, both of which have been replaced by kitutuki. ‘Wren’ is OJ sazaki, while ‘golden-crested wren’ is known as both OJ nipakunapuri and totugiwosipedori ‘the bird that teaches people (how to have) intercourse’; both have died out, replaced by SJ sekirei. ‘Wild duck’ is OJ kamwo, and ‘wild goose’ is OJ kari.
7.3 Fish There is a doublet for the generic OJ word for ‘fish’, OJ uwo ~ iwo and na, including sakana. Nihon shoki records, 魚、此云儺 “‘Fish’, this is read na”. Wamyōshō notes, “‘Fish’ is uwo; in vulgar speech (dialect) it is iwo”. A number of fish names contain one of these elements. Consider the following in Table 3: Table 3: Doublets for “fish” in OJ. iwo/uwo
‘Black porgy’ is SJK tinu, later replaced by ModJ kurodai. ‘Bonito’ is OJ katuwo. ‘Bullhead’ is WMS karakako, which has been replaced by ModJ kazika. ‘Carp’ is OJ kwopi, ‘Crucian carp’ is OJ puna, and a related type of carp is called SJK paraka. ‘Catfish’ is SJK namadu. ‘Char’ is WMS kamatuka, later replaced by ModJ iwana. ‘Crocodile’ is OJ wani. ‘Cuttlefish’ is OJ ika. ‘Dolphin’ has an OJ doublet: OJ iruka or WMS napasaba, but the latter term has died out. ‘Dusky triple-tooth goby’ is WMS titikapuri, but this is later shortened to just titibu. ‘Eel’ has an OJ doublet: WMS pasikami iwo or OJ munagi. The first term now refers to a salamander. ‘Fistularia commersonii’ is WMS apagara. ‘Greater amberjack’ is WMS parimati, which shortened over time to hamati. ‘Hemibarbus barbus’ is WMS mi, later replaced by ModJ nigoi. ‘Hemiramph’ has an OJ doublet, WMS pariwo or WMS yorodu; neither word has survived, replaced by ModJ sayori. ‘Japanese anchovy’ is WMS pisiko iwasi, and this has been replaced by ModJ katakuti iwasi. ‘Japanese trout’ is OJ ayu. ‘Konosirus punctatus’ has a doublet: OJ konosiro
446
John R. Bentley
and tunasi, but only the first survives in the present. ‘Mackerel’ is OJ saba or WMS awosaba. ‘Nibe croaker’ is OJ nibe. ‘Perch’ is OJ susuki. ‘Pufferfish’ is WMS pugu. ‘Righteye flounder’ is WMS karaebi or WMS karepi. The first name has died out. ‘Salmon’ is WMS sake. ‘Sardine’ is OJ iwasi. ‘Saurel’ is OJ adi. ‘Sciaenid’ is WMS isimoti. ‘Sea bass’ is SJK sepi, and the current term seigo is a compound formed from the OJ word. ‘Sea bass’ is OJ suzuki. ‘Sea bream’ is OJ tapi. ‘Sea cucumber’ is SJK ko, now known as ModJ namako. ‘Sea eel’ is SJK pamu, now ModJ hamo. ‘Shark’ has an OJ doublet, OJ same and WMS kotuwo, and the latter form has died out. ‘Hammerhead shark’ is SJK kasesaba, but that word has been replaced by ModJ syumokuzame. ‘Shrimp’ is SJK ebi. ‘Stingfish’ is SJK wokozi, which later changes to okoze. ‘Stingray’ has an OJ doublet, SJK tobiwo and WMS kome, but the first refers to a flying fish in the present, ModJ tobiuo. ‘Striped mullet’ is SJK tukura. ‘Three-lips fish’ is WMS paso, now ModJ hasu. ‘Trout’ is OJ masu, and ‘Biwa trout’ is WMS ame, replaced by ModJ biwamasu. ‘Tuna’ is OJ sibi. ‘Whale’ is OJ kudira. ‘Zacco platypus’ is SJK paye, later replaced by ModJ oikawa.
8 Basic vocabulary list I conclude with a Swadesh list of 100 basic vocabulary items (Swadesh 1971: 283). Due to the lack of evidence for a number of words in the Old Japanese corpus, I have modified the list by altering eight words, adding: ‘blue’, ‘child’, ‘father’, ‘fingernail’, ‘mother’, ‘sea’, ‘vomit’, and ‘year’. All the words on the list are found in records in the period of Old Japanese, and we are fortunate that most examples are preserved in man’yōgana. I have arranged the list alphabetically according to the English gloss and each OJ attestation. Where more than one OJ word is possible for the English gloss, I have provided these. In addition, when a bound and a free form exist, the bound form is listed first. All
Ash Bark Belly Big Bird Bite Black Blood
OJ mwina, ModJ mina; OJ subete, ModJ subete; OJ kotogotoku, ModJ kotogotoku. These all continue in ModJ; aside from these, there are a number of SJ forms in ModJ: zenbu, zentai, banzi. OJ papi, EMJ pawi, LMJ fai, ModJ hai OJ kapa, LMJ kawa, ModJ kawa OJ para, LMJ fara, ModJ hara OJ opo, LMJ oo, ModJ oo. This word only exists as a bound form, while the free form is an adjective meaning ‘many’. OJ tori, ModJ tori. Tori also refers to ‘chicken’ as early as the EMJ period, as in nipatori ‘a chicken’ (literally a bird in the yard), and toriniku ‘chicken meat’. OJ kam-, ModJ kamOJ kurwo-, ModJ kuro-. ‘Black like the earth’ is OJ nigurwo-. ‘Pitch black’ in OJ is kagurwo-, but this has been replaced by ModJ makkuro-. OJ ti, ModJ ti
19 The history of basic vocabulary in Japanese
Blue
Bone Breast Burn Child Cloud Cold
Come Die
Dog Dry Ear Earth Eat
Egg
Eye Fat Father
Feather
447
OJ awo-, ModJ ao-. The semantics of this word has a rather wide spectrum, from a light bluish grey to an herbal green, as in examples like OJ awogumwo ‘grey clouds’, awona ‘greens’, and awoni ‘green earth’. OJ pone, LMJ fone, ModJ hone OJ ti, ModJ ti or titi; OJ muna-, mune, ModJ mune. The bound form appears in words like munage ‘chest hair’ and OJ munasaka ‘on the breast/chest’. OJ mwoye- (intr.), ModJ moe-; OJ mwoyas- (trans.), ModJ moyas-; OJ yake(intr.), ModJ yake-; OJ yak- (trans.), ModJ yak-. OJ kwo, ModJ ko, kodomo OJ kumwo, EMJ kumo, ModJ kumo OJ samu-, ModJ samu-. The appearance of the semantic difference between a cold object and feeling cold (samu- versus tumeta-) does not happen until the EMJ period. OJ ko-, ModJ koOJ sin-, ModJ sin-. There are a number of euphemistic usages from the OJ period for dying: for the ruler and deities, OJ kamuagar- ‘divinely rise up’, LMJ kanagar-. The death of the emperor is now known as SJ hoogyo. For royalty and courtiers of the third rank and up, OJ kamusar- ‘divinely leave’, LMJ kansar-. The same group of courtiers could also OJ mius- ‘lose one’s body’, which later becomes LMJ miusinaf-, ModJ miusinaw-. For the death of courtiers of the fourth or fifth rank, mimakar- ‘expire’, which usage continued until the early 20th century. OJ inu, wenu < wo-inu ‘small dog’, ModJ inu OJ kawak- (intr.), ModJ kawak-; OJ pwi- (intr.). OJ pos- (trans.) ‘to hang something so it dries out’, LMJ fos-, ModJ hos-. OJ mimi, ModJ mimi OJ tuti, ModJ tuti OJ kup-, MJ kuw-, ModJ kuw-. OJ tab-, ModJ tabe-. This second form originally was an honorific verb ‘to give’, and as in the OJ liturgies, ‘to give food, to feed’, came to mean to eat. OJ kwo, kapigwo, LMJ tamago, ModJ tamago. This word is likely the same as kwo ‘child’, and to avoid confusion tama ‘round’ was added during the latter end of the LMJ. OJ ma-, me, ModJ ma-, me. The bound form appears in OJ manabuta ‘eyelid’ and OJ matuge ‘eyelash’. OJ abura, ModJ abura; OJ puto- ‘be fat’, ModJ futo-. OJ titi, ModJ titi. The OJ form kaso appears in Nihon shoki and Wamyōshō (1978: 2, 14a–14b) contains this explanation, “‘Father’ is kaso, while ‘mother’ is iropa. In the vernacular ‘father’ is titi, and ‘mother’ is papa”. Kaso appears as kazo in the 17th century then disappears. OJ pa, pane, LMJ fane, ModJ hane. Ōno et al. (2000: 1044) believe that OJ pa meant ‘feather’, and pane meant ‘wing.’
448
John R. Bentley
Fingernail Fire Fish
Fly Foot Fill / Full Give
Good Green Hair
Hand
Head
Hear Heart
Horn I Kill Know Leaf Lie (down) Liver
OJ tume, ModJ tume OJ po-, pwi, LMJ fo-, fi, ModJ ho-, hi. The bound form appears in words such as OJ potakupi ‘half burnt tree’ and ponaka ‘inside the flames’. OJ iwo, uwo, LMJ io, uo, ModJ uo. Rodriguez notes that both words, uo and sakana, exist for fish in the LMJ vernacular. The traditional etymology for the latter is ‘greens (with your) sake’, but recall that Nihon shoki provides evidence that na existed as a doublet for ‘fish’ (魚、 此云儺 ‘Fish’, this is read na.), which leaves open the possibility that sakana still contains an element meaning ‘fish’. OJ tob-, ModJ tobOJ asi, ModJ asi OJ mit- (trans.), ModJ mitas-; OJ miti- (intr.), ModJ mitiOJ age-, ModJ age-; OJ yar-, ModJ yar-. The second form originally meant to send someone to do something, and later included the meaning to send money, to pay. This came to mean ‘give’, and now means to give something to someone lower in status than the agent. OJ yo-, ModJ yo-. There is also a related form, OJ yoro- and yora-, but only yoro- survives in ModJ. OJ awo-, LMJ ao- , ModJ ao-. EMJ midori, ModJ midori OJ kami, ModJ kami no ke. OJ ke referred to the fur of animals, but it also appears in words for specific hair of humans, such as OJ matuge ‘eyelash’. OJ ta-, te, ModJ ta-, te. The bound form is found in OJ words like OJ taduna ‘reins, bridle (hand-rope)’ and OJ tanamata ‘the web of skin between the fingers’. OJ kasira, ModJ kasira; this word is later replaced by EMJ atama, written 顖會 and glossed atama in Wamyōshō, which originally meant the fontanel, or the soft part of an infant’s skull. OJ kik- (trans.), ModJ kik-; OJ kikoye- (intr.), ModJ kikoeOJ kokoro, ModJ kokoro; OJ kimwo originally meant organs, but also referred to emotions or maturity, as in kimwo wakasi ‘immature’ as found in the Suiko record of Nihon shoki. Kimwo later comes to refer to a specific organ, the liver, and survives as ModJ kimo ‘liver’ (cf. Omodaka 1967: 246–247). OJ tunwo, ModJ tuno OJ ware, ModJ ware, watakusi OJ koros-, ModJ korosOJ sir-, ModJ sirOJ pa, LMJ fa, ModJ ha OJ ne-, ModJ neOJ kimwo, ModJ kimo. As the name of an organ, SJ kanzoo has replaced kimo, but the native form still appears in words like ModJ kimo ga hutoi
19 The history of basic vocabulary in Japanese
Long Louse Male
Many
Meat
Moon
Mother
Mountain
Mouth
Name Neck New
Night Nose Not One Path Person
449
‘courageous’. Omodaka (1967: 246–247) states that OJ kimwo originally pointed to internal organs in general, and over time came to then just refer to the liver. OJ naga-, ModJ nagaOJ sirami, ModJ sirami OJ wo, wotokwo, ModJ otoko. The simple form wo falls out of use in the late EMJ period, though it persists as a part of personal names, such as Mitsuo (光男) and many other names. OJ opo-, LMJ oo-, ModJ oo-. As noted above, the adjective always means ‘many,’ while the bound form of opo means ‘large in size’. There is also OJ sapa ‘many’, which is no longer used in the modern language. OJ sisi, ModJ sisi. Sisi refers to ‘game’ and by extension the flesh of that game. This word later comes to refer specifically to wild boar, and is replaced by SJ niku for ‘meat’ around the 11th century. OJ tuku-, tukwi, EMJ tuku-, tuki, ModJ tuki. The bound form appears in OJ words such as tukuywo ‘moonlit night’, and tukuyomi ‘deity of the moon’. OJ papa, LMJ fafa, ModJ haha; OJ omo, but this word undergoes semantic change during the EMJ period, meaning ‘wet nurse’. There is also a dialectal version amo. Another word for mother is OJ iropa, but this does not survive in the modern language. OJ yama, ModJ yama. There is also OJ ne ‘peak’, found in a variety of compounds, such as mine ‘peak’, takane ‘high peak’, and yamane ‘mountain peak’. OJ kutu-, kuti, ModJ kutu-, kuti. The bound form appears in OJ words such as kutubami ‘a metal bridle’ and kutuuta ‘sing to oneself (< ‘mouthsong’). OJ na, ModJ na, namae OJ kubi, ModJ kubi OJ nipi-, EMJ niwi-, LMJ nii-, ModJ nii-. This form only survives in fossilized words, such as the toponym Niita ‘new rice fields’. There is also the word OJ arata-, ModJ arata-. The current adjective, ataras- appears to be a metathesized form of arata-. OJ ywo, yworu, ModJ yo, yoru. A related word is OJ ywopi, which Omodaka (1967: 801) states refers to the first half of the night. OJ pana, LMJ fana, ModJ hana OJ nasi, ModJ nai OJ pito-, LMJ fito-, ModJ hito-. This number is now primarily an ordinal number, while SJ iti is used as a cardinal. OJ miti, ModJ miti OJ pito, LMJ fito, ModJ hito
450
Rain Red Round Sand Say Sea
See Seed Sit Skin Sleep Small
Smoke
Stand Star Stone Sun Tail
John R. Bentley
OJ ama-, ame, ModJ ame. The bound form appears in OJ words like amagiri ‘rainy mist’, and amatutumi ‘stuck in the house due to rain’. OJ aka-, ModJ akaOJ maru-, ModJ maruOJ isagwo, ModJ isago. By the EMJ period the word sunako ‘sand’ begins to displace isago. In ModJ the truncated form suna is the primary word. OJ ip-, EMJ iw-, ModJ iw-; OJ mawos-, EMJ maus-, ModJ moos-. This latter form is used as an honorific. OJ umi, ModJ umi. A doublet is OJ wata, which appears to be a loan from a language on the Korean peninsula, but this word dies out during the EMJ period, only preserved in names, such as watatumi ‘deity of the ocean’. Another OJ word is una- found in compounds such as unasaka ‘shore line’ or unapara ‘the vast sea’. However, Nihon shoki preserves a story of a river in Paekche that springs from Mount Kane. In that account 源 ‘spring, source’ is glossed as unakami, making me believe the original meaning was ‘water’ and later came to be used analogous to the Middle English usage of ‘waters’, referring to a large body of water. OJ mi- (trans.), ModJ mi-; OJ miye- (intr.), ModJ mie-. OJ sane ~ tane, LMJ sane ‘pit’, tane ‘seed’. In the present tane survives in the standard language, but many dialects in Japan still use sane. OJ wir-, EMJ wir-, LMJ ir-, ModJ ir-. In the LMJ period the word suwarappears, and semantically replaces ir-, which comes to mean ‘to exist’. OJ pada, LMJ fada, ModJ hada OJ nebur-, ModJ nemur-. By the late 12th Century, nemur- replaces nebur-. OJ tipisa-, EMJ tiwisa-, LMJ tiisa-, ModJ tiisa-. OJ sukuna-, ModJ sukuna-. The first word refers to small in size, while the second refers to small in number. There is also a prefix, OJ wo, denoting a diminished characteristic of the noun: OJ woba ‘aunt’ (small mother, compared to papa ‘mother’ and opopa ‘grandmother’). Other examples are OJ woya ‘hut (small house)’ and OJ wokamo ‘small duck’. OJ keburi, LMJ kemuri, ModJ kemuri. These two words, keburi and kemuri, seem to have coexisted for centuries through LMJ, where keburi came to be a counter for chimneys or smokestacks, and kemuri continued as ‘smoke’. OJ tat- (intr.), ModJ tat-, OJ tate- (trans.), ModJ tateOJ posi, LMJ fosi, ModJ hosi OJ isi, ModJ isi; OJ ipa, ModJ iwa. In general, isi refers to a rock or stone, while ipa is a larger object, like a boulder. OJ pi, LMJ fi, ModJ hi OJ wo, ModJ o, though the word siriwo ‘rear-tail’ became general in the mid to late LMJ, resulting in ModJ sippo. This development is clear from
19 The history of basic vocabulary in Japanese
This Tongue Tooth Tree Two Walk
Warm Water
White What Who Woman Vomit
Year Yellow
dialectal data, where dialects in Iwate and Miyazaki have sirippo, and Shimane has siriho. OJ kore, kono, ModJ kore, kono OJ sita, ModJ sita OJ pa, LMJ fa, ModJ ha OJ ko-, kwi, EMJ ko-, ki, ModJ ko-, ki. The bound form appears in OJ word such as konoma ‘between trees’, and kodati ‘grove of trees’. OJ puta-, LMJ futa-, ModJ futaOJ ayum-, ModJ ayum-; OJ aruk-, ModJ aruk-. Omodaka (1967: 60) explains that anciently ayum- meant ‘step, walk’ and aruk- meant ‘move, make progress toward a place, wander’. OJ atu-, ModJ atu-, for ‘hot’, while ‘warm’ is OJ atatake-, ModJ atataka-. OJ midu, ModJ mizu. The language of OJ also preserves mina, found in compounds like minagipa ‘by the water’, minasoko ‘bottom of a body of water’, minagirapu ‘overflowing with water’, and minatwo ‘port’, literally ‘door to the water’. In the modern language, there are a number of words for water, depending on its quality or use: SJ reisui or ModJ ohiya ‘cold water’, ModJ oyu ‘hot water’, SJ yoosui ‘irrigation water’, and ModJ amamizu ‘rain water’. OJ sira-, sirwo-, EMJ sira-, siro-, ModJ sira-, siro-. The bound form appears in OJ words such as siraka ‘white hair’, and sirakumwo ‘white clouds’. OJ nani, ModJ nani OJ ta or tare, ModJ dare OJ mye, LMJ me; OJ womina, LMJ ouna, ModJ onna OJ tagur-, LMJ tagur-; OJ pak-, LMJ fak-, ModJ hak-. Tagur- changed semantically over time, and came to mean ‘to cough’ and then died out, replaced by seki ‘a cough’ in the LMJ period. OJ tosi, ModJ tosi. This word primarily refers to ‘age’ in the modern language, and the SJ word nen is used for years in a calendrical sense. OJ ku-, ki (likely kwi, but the word is unattested in phonograms), ModJ ki-. The bound form survives in OJ words like kugane ‘gold (yellow-metal)’.
Abbreviations SJ SJK WMS
451
Sino-Japanese Shinsen jikyō Wamyōshō
452
John R. Bentley
References Bentley, John R. 2000. New Look at Paekche and Korean: Data from Nihon shoki. Ehaku Yengu LanguageResearch 36(2). 417–443. Frellesvig, Bjarke. 2010. A history of the Japanese language. Cambridge: Cambridge University Press. Kuno, Mariko. 2005. Nihon hōgen kiso goi no kenkyū [Research into the basic vocabulary of Japanese dialects]. Tokyo: Ōfūsha. Kuroita, Katsumi. 1955. Ryō no shūge [A collection of commentaries on the administrative code]. Tokyo: Yoshikawa Kōbunkan. Kyotō Daigaku Bungakubu Kokugogaku Kokubungaku Kenkyū-shitsu (ed.). 1992. Shinsen jikyō: Tenji-bon [Newly compiled mirror of characters: the Tenji-bon manuscript]. Kyoto: Rinsen Shoten. Nakamoto, Masachie. 1981. Nihongo no genkei: Nihon rettō no gengogaku [The proto-form of Japanese: The linguistics of the Japanese archipelago]. Tokyo: Rikitomi Shobō. Nihon Koten Zenshū Kankōkai (ed.). 1978. Wamyō ruijushō [Japanese names of things, classified and annotated]. Nihon koten zenshū. Tokyo: Gendai Shichōsha. Omodaka, Hisataka. 1967. Jidai betsu kokugo daijiten: Jōdai-hen [Great dictionary of Japanese language divided by period: Ancient period]. Tokyo: Sanseidō. Ōno, Susumu, Satake Akihiro, Maeda Kingorō (eds.). 2000. Iwanami kogo jiten, hoteiban [Iwanami dictionary of classical Japanese]. Tokyo: Iwanami Shoten. Swadesh, Morris. 1971. The origin and diversification of language. Chicago: Aldine. Thomason, Sarah G. 2001. Language contact: An introduction. Edinburgh: Edinburgh University Press. Vovin, Alexander. 2010. Koreo-Japonica: A re-evaluation of a common genetic origin. Honolulu: University of Hawaii Press, Center for Korean Studies. Yamanaka, Hiroshi and Yasushi Morita. 1991. Ronsō: Nihon kodaishi [Controversy: Ancient Japanese history]. Tokyo: Kawade Shobō Shinsha.
Toru Maruyama
20 The Japanese lexicon as reflected in Christian materials 1 Introduction Following the route of Portuguese expansion, the Jesuit Mission Press produced linguistic works of indigenous languages of Africa, Brazil, and India. Those which we have been able to trace so far are as follows: in Africa a grammar of Kimbundu and Christian doctrines in Kimbundu and in Kongo; in Brazil grammar books of Tupi and Kiriri and Christian doctrines written in them; in India a grammar of Konkani, Christian doctrines in Tamil and in Konkani, and so forth. These are actually pioneering linguistic works, the first to ever have been created, concerning indigenous languages of the world. In Japan, Portuguese missionaries tried to propagate Christianity through printed material from the very beginning, since Francisco Xavier stated that considering the extremely high literacy of Japanese it was desirable to propagate Christianity through printed works (Letter No.90 in Schurhammer G. and J. Wicki. eds. 1944–1945). Moreover, in Japan virtually only one language was spoken throughout the entire country. In India, for example, facing so many different languages and writing systems, they might not have had so much motivation to compile printed dictionaries of any one specific language. Although there are many other documents that can be discussed concerning this theme, like Racuyoxu, a Japanese dictionary of Chinese characters published by Jesuits in 1598 (see Bailey (1962) for an excellent introduction in English), or books and dictionaries published by other religious societies, the concentration of the discussion here is placed on the two major voluminous dictionaries, namely the Japanese-Portuguese dictionary (Section 2) and the Latin-Portuguese-Japanese dictionary (Section 3) published by the Jesuits. Section 4 will briefly address the notion of “elegance” as espoused by the Jesuits, and finally Section 5 looks at the importance of printing.
2 Japanese-Portuguese dictionary The Japanese-Portuguese dictionary compiled by Jesuits in 1603–1604 contains more than 32,000 entries with explanations in Portuguese, having the following title, Vocabulario da lingoa de Iapam com a declaração em portugues, feito por alguns padres, e irmãos da Companhia de IESV. This dictionary not only contains almost twice the number of entry-words of the Jesuit Tamil-Portuguese dictionary published in 1679 with about 15,000 entries or of https://doi.org/10.1515/9781614512851-021
454
Toru Maruyama
the Konkani-Portuguese dictionary (manuscripts) written in 1626 with about 14,000 entries, but also the explanations therein are much more detailed. For example, the following is the explanation for the entry “Taixi”. For verbs three different forms, or the basic, the present, and the past forms are listed first as an entry, followed by various examples with Portuguese translations. Taixi, suru, ita. i.[i.e.] Mucǒ. Por se de fronte [To place oneself in front of]. ¶ Fitoni taixite mǒsu. Falar direito a alguem [To speak directly to someone]. ¶ Deus ni taixi tatematçurite. Por amor, or por respeito de Deos [For love or respect of God]. ¶ Iesu Christo varerani taixerarete. l, Taixi tamaite. Christo por amor de nos, etc. [Christ for love of us]. (238r)
This dictionary also has various indications for different registers like feminine words, children’s words, dialect forms, poetic words, literary words, impolite/taboo words etc. with special abbreviation codes like P for poetic words, S (= Scriptura) for literary words, B (= Baixo ‘low’) for impolite/taboo words, and X (= Ximo (Japanese) ‘low or inferior’) for dialect forms.1 For example, more than a hundred words are registered as “palaura de molheres” ‘feminine words’ as shown here. Aca [literally meaning ‘red’]. i.[i.e.] Azzuqi. Certa especie de feijões pequenos [Certain kinds of small beans]. He palaura de molheres [This is a feminine word]. (332r-supplement) Amo. Mochi [rice cake]. Palaura de molheres & mininos [literally meaning ‘words of women and children’ = Terms used by women or children]. (8v) Amo. Bollos de farinha de arroz [Cake from flour of rice]. Palaura de molheres. (332v-supplement) [Amo – shortened form of An-mochi, or rice cake with sweet bean paste] Bobo. Pudenda mulieris [female genital]. Palavras que as molheres, & meninas usão [Words which women and children use]. (23v~24r) Caca. i. Faua. Mãy [mother]. He palaura de meninos [This is a children’s word]. (29v) Vofuru [vo- prefix expressing politeness; furu ‘being old’]. Cousa velha como vestido que o senhor da a algum criado [An old thing like – for example – a dress which a master gives to a servant]. . . . Palaura de molheres. (278r) Vonaca [naka ‘middle’]. Barriga [belly]. . . . He palaura de molheres. (389r – supplement) Xixi. Ourina de mininos [Children’s urine]. Palaura de molheres. . . . (308v)
These feminine words are basically interpreted as euphemisms originally employed by court ladies, then having spread from common women to men perhaps through children or through social intercourse. We should note that some of these words are interpreted as for children’s use as well as for women’s use, underlining the probable close contact between women and children. How Portuguese missionaries succeeded in accumulating all of these words into the dictionary is unclear, but without the help of a considerable number of Japanese collaborators it would have been impossible.
1 For the whole list of these feminine words, children’s words, dialect forms, poetic words, literary words, impolite/taboo words etc., see Morita (1989).
20 The Japanese lexicon as reflected in Christian materials
455
More than 70 words registered with the code B (for Baixo, a Portuguese word meaning ‘low’) as shown below. These indicate impolite/taboo words which Portuguese missionaries should not use either in sermons or in everyday life, but which they should know as passive vocabulary, mainly for hearing and understanding confessions. Feco. Encacho. [loincloth] B. A propria palaura he Xitauobi,. . . [The more appropriate word is Xitauobi,. . .]. (85r) Miguite. Meliùs [better form], mete. Parte direita [right part or direction]. B. (159v)
Besides the words with the code B, some are indicated as those which should not be used in sermons or in everyday life with expressions like “Baixo” ‘low’, “Palaura baixa” ‘inferior words’, “falando baixamente” ‘speaking impolitely’, “gente baixa” ‘lower people’, “falando com desprezo” ‘speaking with contempt’, “abatendo” ‘lowering’, or “abaixando” ‘showing contempt’. Dialect forms are also indicated in various ways. Some are registered with the code X as follows. Voyacata. i. Ani. Irmão mais velho. [Elder brother] X. (390v.)
In this case the word following “i.” (id est ‘that is’), “Ani”, is considered to be the standard form. Others are indicated with the Latin word “alicubi” ‘in some parts’ as follows. Catacoma. Alicubi. Catacuma. Hombros, ou pescoço que serue de hum ir nelle como em caualo. . .. [Shoulders or neck which carry one like horse-riding] (42r)
Others are also identified as dialect forms registered as “Ximo” (a Japanese word meaning ‘lower’ or ‘inferior’) in contrast with the corresponding standard forms with the code “Cami” (‘upper’), central region. Tamago. Ouo [Egg]. No Cami [In the central region], Caigo. (238v) Caigo. Ouos de galinha, ou passaro [Eggs of chicken, or bird]. . . . No Ximo, bicho de seda. [In the Ximo area, (it means) silkworm.] (33r) Caico. Bicho de seda, ou sua semente no ximo dizem tambem, caigo. [Silkworm or its seed. In the Ximo area they also call it Caigo.] (32v)
The descriptions of these three items tell us why the newer form tamago became to be used as a dialect form for the legitimate form kaigo. The similarity in form and meaning between kaigo ‘egg’ and kaico ‘seed of silkworm’ must have given birth to a distinguishable temporary form tamago for ‘egg’, first in the Ximo or Kyushu area where the above three forms were competing with each other as follows: Caigo – ‘egg’ or ‘silkworm’ Caico – ‘silkworm’ or ‘seed of silkworm’ Tamago – (newly developed distinguishable form for) ‘egg’
456
Toru Maruyama
On the other hand, in the Cami or central region, the following traditional distinction was maintained: Caigo – ‘egg’ Caico – ‘silkworm’ or ‘seed of silkworm’
This distinction seems to have been maintained until the Muromachi period, when the newly created form tamago became prevalent both in rural and central areas of Japan. It goes without saying that we now only use the form tamago for ‘egg’, not kaigo, at least in standard Japanese. The people started to use the form tamago probably because the shape of an egg resembles a tama ‘ball’. The form -go is a morphological alternation of the diminutive suffix -ko. Here, however, we should also keep in mind that this description of the dictionary – the new form tamago was only prevalent in the Ximo area at that time – does not exclude the possibility of the form tamago already having existed in the central region. It goes without saying that the contents of the Vocabulario are by no means the only absolute data base for the lexical description of medieval Japanese. This Japanese-Portuguese dictionary contains about 150 words with the code “Bup.”, for Buddhism. Buddhism being one of the most prevalent religions in contemporary Japan, Portuguese missionaries had to have them in mind not only as a passive vocabulary but also as an active vocabulary. For the propagation of Christianity, they first employed various concepts from Buddhism, and they also had to be able to use them in refuting Buddhism in the debate with Buddhist priests or educated people. Cù. Bup. Hũa cousa imperfectissima, & de minimo ser como materia prima, ou vacuo [A most imperfect thing or of the minimum like materia prima, or emptiness].(62r) Zajen. Meditação [Meditation]. Bup. (402r)
There are also more than 1500 words with the code S of Scriptura, or literary words, or the equivalent expressions like “Palaura de liuros” ‘word for books’, “Palaura de cartas” ‘word for letters’ etc. More than 500 words are indicated with the code P of Poesia, for poetry, or equivalent expressions like “apud poetas” ‘by poets’, “na poesia” ‘in poems’, etc. The knowledge of various socially different styles of Japanese was indispensable for Jesuits whose mission was to convert all Japanese people, from the upper highly educated class to the lower illiterate class. For Jesuits, whose main daily obligation was to preach and to hear confessions, it was vital to learn the difference between the language style they should use and the one they should not use but should understand hearing.
20 The Japanese lexicon as reflected in Christian materials
457
2.1 Women’s language in Japanese As is explained in the previous section, various types of feminine words are registered in the Vocabulario, and they are considered as one of the roots for women’s language in modern Japanese (Kamei 1978). Basically, there are three types of lexical formation of these feminine words, i.e. replacement, prefixing and suffixing. (1) Replacement – The covered form (“Azzuqi” in the previous example) is replaced by the covering form (“Aca” in the previous example), where the relation between the two words is exclusively semantic (sharing the quality of redness in the previous example). Another example is “Tǒfu” ‘bean curd’ replaced by “Cabe” (= kabe ‘wall’), sharing the quality of whiteness. The connection between the covering form “Caramon” (= karamon ‘hot tasty thing’) and the covered form “Daicon” (= daikon ‘Japanese white radish’) must be the quality of hot taste. This “Daicon” was also replaced by the covering form “Yuki” ‘snow’, sharing the quality of whiteness. (2) Prefixing – The covering form is composed by adding the prefix o- to the covered form, such as “Vocama” (= okama) from “Camaboco” (= kamaboko ‘boiled fish paste’), or “Vonama” (= onama) from “Namasu” ‘finely chopped raw fish and vegetable soaked in vinegar’. A modern Japanese form onaka in the expression such as onaka ga suita ‘I am hungry; lit. the stomach has become empty’ has its root in this prefixing process. (3) Suffixing – The covering form is composed by dissecting the initial syllable of the covered form and adding the word “moji” ‘letter, graph’, producing the final covering form, such as “Fimoji” (= himozi ‘hunger’) from hidarui ‘hungry’. This final form himozi has survived even today, yielding the adjectival form himozii ‘to be hungry’. Another surviving example of this type is syamozi (a wooden scoop for serving rice) composed from the covered form syakusi.
2.2 Dictionaries of Japanese and Konkani with explanations in Portuguese The preface of the Japanese-Portuguese dictionary, Vocabulario da Lingoa de Iapam, says that various manuscripts of dictionaries and grammar books had been a help for those who started to study Japanese. Because of the severe persecution in place at time, paradoxically Portuguese fathers and Japanese brothers came to have more time to concentrate on compiling dictionaries instead of propagating Christianity in public. In this way they spent several years concentrating on reviewing manuscripts of dictionaries and making supplements or enlarging dictionaries. Up until now, at least four different manuscripts have been reported to have existed before the publication of Vocabulario in 1603 – Nachlass manuscripts by Duarte da Sylva who died in 1563, the manuscripts compiled in 1564 by João Fernandes at the strong urging of Luis Frois and being added to continuously for the following twenty
458
Toru Maruyama
years by Frois himself, the one compiled at the Colegio in Funai around 1581 or 1582, and lastly the manuscripts compiled at the Seminario in Arima in 1585. One of the above four, especially the last two, might have been the major basis for the Vocabulario published in Nagasaki in 1603–1604. Unfortunately none of the above are in existence today (Morita 1989). In India, on the contrary, only manuscripts of dictionaries have been confirmed to have been in existence. No publication of dictionaries by Jesuits in the sixteenth or early seventeenth century has been attested to so far. Several manuscripts of Konkani-Portuguese dictionaries have been reported to be in existence in India or in Portugal. Here we compare two of them (VLKK & VLCP), one obtained by the Portuguese Ministry of Education in 1965, the facsimile of which was published in 1973, and the other compiled by Jesuits in Salcete and supplemented by Father Diogo Ribeiro in 1626. The compilation date of the former is unknown. It is clear that the two manuscripts of Konkani-Portuguese dictionaries compiled by Jesuits about four hundred years ago have a considerable portion of their contents in common. Their relative chronology is still very debatable. However, they clearly indicate the process of how dictionaries were compiled by the Jesuits, incessantly adding and revising preceding manuscripts. For some indexed words, the addition of descriptions in one of the two is apparent. For example, for the Konkani item gau, the following explanations are seen in the two manuscripts. Gâu. â. Aldea. [village] [VLCP f56] Gâu. uâ. Aldea, tambem certa legoa que contem quatro kossos. [village, it also means certain leagues, or four kossos] Hum gâu dizem ser como duas legoas das nossas. [One gau corresponds to our two leguas.] Goem quitulem gâua âssati. Quantas legoas são daquy a Goa? [How many gaus are there from here to Goa?] [VLKK f104]
The explanation in VLCP for gau is only one word, namely, “Aldea” ‘village’, while in VLKK relatively long comments follow after that including an example sentence. It says that gau also means a unit of some distance. In any language of the world the unit of distance often derives from the designation of some block or inhabited area. This additional explanation must have been made after VLCP was written and as far as this part is concerned, VLCP precedes VLKK. Another example of addition of explanations in VLKK are as follows. Kasttu, â. Trabalho [work]. [VLCP f71v] Kasttu. â. Trabalho. kastta karitâ. Trabalhar [to work], tambem paga do trabalho [also the salary of work]. [VLKK f150]
Possible signs for these kinds of indications of indexed items are apparent in the manuscript form of VLKK as follows, although, as for the publishing year, the Vocabulario (da lingoa de Iapam) chronologically precedes VLKK.
20 The Japanese lexicon as reflected in Christian materials
459
Dalladrinni. y. Molher pobre [poor woman], palavra baixa [impolite/taboo word]. Rândda. ê. e Rânddâua. ê. Viuva [widow]. Palavra odioza [odious word]. Dari Adv. De prinçipio [from the beginning], desde [from] Kâlidari. Desde ontem [from yesterday]. Melhor se diz [It is better to say] Kâli leguina. Dâuô. ueâ. Calumnia [calumny]. Melhor he [The better is] Dâu ghâlita.
Probably before printing, through the same kinds of adding, erasing, or editing processes, the early seventeenth century Japanese-Portuguese dictionary must have been prepared with various indications for different registers like feminine words, children’s words, dialect forms, poetic words, literary words, impolite/taboo words etc. with special abbreviation codes like P for poetic words, S (= Scriptura) for literary words, B (= Baixo ‘low’) for impolite/taboo words, and X (= Ximo ‘Japanese meaning low or inferior’) for dialect forms.
3 The Latin-Portuguese-Japanese dictionary Preceding the publication of the Vocabulario, Jesuits had published the Latin-Portuguese-Japanese dictionary (henceforth LPJ dictionary) in the multilingual tradition of Calepin. As is indicated in the title, it is based on the famous Calepin Latin dictionary and it is the only edition that contains Portuguese and/or Japanese in its long tradition – a quarto edition with 908 pages. The number of entries is about 30,000 and the explanations therein are slightly simpler than the JP dictionary explained above. The following explanation, for example, is seen for the entry “Fortuna”. Fortûna. ae. Lus. Acontecimento subito, e inopinado. Iap. Vomoino focani ideqitaru cotouo yŭ. ¶ Item. (in vtroq; numero) Riquezas. Iap. Zaifô, tacara. ¶ Item, Fortunae, arum. Lus. Felicidade. Iap. Fucqi, yeiguani tomisacayuru cotouo yŭ. ¶ Fortuna aestuaria. Lus. Fortuna varia, inconstante. Iap. Xecaino giŏsŏ naqi cotouo yŭ, l, funo xôretno cotouo yŭ. (p. 296)
This marvelous LPJ dictionary, compiled by three Portuguese missionaries and two Japanese brothers in 15 months (according to Pedro Gomes’ letter of 12 October 1595), has not been studied in detail so far. Comparing various editions of Calepin’s dictionary, Hiroshi Harada (2011), one of the leading Latin philologists in Japan, so far concludes that this LPJ dictionary (1595) is probably based on the Lyon edition of Calepin dictionary published in 1580. While the Vocabulario contains various social registers, like female expressions, children’s expressions, dialectal forms, impolite or taboo words, etc., the LPJ dictionary basically only contains standard “elegant” Japanese expressions. This is possibly because the former was compiled mainly for the purpose of Catholic confession, while the latter was intended for religious sermons. (Ōtsuka 2006). How to interpret the very few examples of everyday informal Japanese expressions in the LPJ is a difficult question. For example, we can see an example of the informal Japanese sentence final expression “gia” ‘to be (copula)’ in the LPJ as follows.
460
Toru Maruyama
Ita se habet. . . .. Conobun gia. l. cacu nogotoqu nari. (LPJ p. 322)2
According to Kishimoto (2012), in contrast with more than 1000 examples of the formal sentence final expression “nari” ‘to be (copula)’, only a few examples of the vernacular informal “gia” are found in the LPJ. These might be only a residue of preceding manuscripts, which were supposed to have been erased before publication. However, at the same time, there is a possibility that the editor intentionally inserted these few examples to show the vernacular nature of the Latin expression Ita se habet. It is indeed a debatable question for the future research. Basically, however, only formal polite Japanese expressions are registered in the LPJ, which were supposed to be used in Catholic sermons. According to Harada (personal communication) this expression Ita se habet is an expression used in dialogue in Latin, which requires an informal vernacular as well as a formal written style translation, depending on context. In this sense, the explanation of the LPJ, giving both a vernacular and a written style translation, i.e. vernacular “conobun gia” and formal “cacu nogotoqu nari”, is considered to be quite appropriate. The relation between these very few oral informal Japanese expressions in the LPJ and the following statement in the preface of the LPJ is also thought-provoking – “Sometimes in Japanese explanations, there are less polite and less elegant expressions. However, these are to explicitly explain the essential nature of Latin words. (Interdum in declaratinone Iaponica aliquid minus polite, & eleganter dictum est, vt Latini vocabuli vis apertius, & enucleatius exponeretur.)” (Kishimoto 2012). The study of the relation between this Latin-Portuguese-Japanese dictionary (1595) and the Japanese-Portuguese dictionary, or the Vocabulario (1603–1604) is only in the initial stage. Little is known so far concerning this issue.
4 What is “elegant” for Jesuits? The word “elegant” is one of the key concepts of linguistic studies by Jesuits. Not only in the preface of the LPJ dictionary mentioned above, but also in the marvellous two first Japanese grammars written by João Rodrigues, this word is repeatedly used in the description of Japanese. “Elegance” for Portuguese missionaries is basically considered to be the gracefulness of style, the origin of which can be traced back to classical works. Therefore, the word “elegance” or “elegant” in their texts implies a knowledge of classical languages, namely Latin in the Western tradition, and the formal Japanese of earlier classical literature in Japan. For Rodrigues, for example, the forms traceable to the classical language are more elegant than the others.
2 Ita (= “so”) se habet (= “find oneself in this or that circumstance; be”) means “That’s it!” “It is so!” or “Exactly!”.
20 The Japanese lexicon as reflected in Christian materials
461
Actually, the expression “elegant” was used in this sense in sixteenth century Europe. In the book Dialogo em louvor da nossa lingoagem by João de Barros, the following dialogue between Father and Son is observed – (F) Quál destas á por melhór, e mais elegante ? (P) A que se mais confórma com a latina, assi em uocábulos como na orthografia. E nesta párte muita uantaiem tem a italiana e espanhól, á francesa; . . . [Filho ‘son’) Which of these (forms) is better and more elegant?] [Pai ‘father’) The one which has more conformity with Latin, not only in vocabulary but also in orthography (is more elegant). In this point Italian and Spanish have more advantage than French; . . .]
According to this tradition, Rodrigues aimed at a “proper” style of language, which followed the classical style. For him the language traceable to classics was more “proper” and more “elegant”, and so for other Portuguese missionaries as well.
5 The importance of printing The Japanese-Portuguese dictionary as well as the Latin-Portuguese-Japanese dictionary compiled by Jesuits are goldmines for the study of the history of Japanese or of the sixteenth and seventeenth century Japanese culture. Actually, all of the linguistic, literary, and religious works produced by Jesuits, especially by their printing, are indispensable for the study of medieval Japan and Japanese or of other areas with other languages of the world where Jesuits had contact in the sixteenth and seventeenth centuries. The linguistic studies by Jesuits in the sixteenth and seventeenth century are highly appreciated even from the modern linguistic point of view. But why the sixteenth and seventeenth centuries? Why Christianity? Alexander the Great, surely must have encountered a considerable number of different languages on his way to the East. Neither he nor any of his subordinate interpreters have left any linguistic work on the languages with which they were in contact. Neither the famous traveler Marco Polo nor the famous Moroccan traveler Ibn Battuta left us any linguistic work. The Indian emperor Ashoka did not produce any linguistic work, nor did any of the famous Chinese pilgrims, who translated a considerable number of religious documents into Chinese, leave us any linguistic work of indigenous languages on the way of their pilgrimage. Even in the world of Christianity, St. Paul and those who came after him, for example, propagating Christianity through history far beyond their native land, did not produce any linguistic work. Here again –Why in the sixteenth and seventeenth centuries? Why Christianity? A key to various possible interpretations is “movable type printing press”. Francisco Xavier’s following remark in his letter sent from Japan is symbolic – “Here we can propagate Christianity through printing.” In the process of printing we have to set
462
Toru Maruyama
up a linguistic norm for the language. We have to decide, for example, fixed shapes for each letter, fixed spellings for each word, fixed word order, fixed expressions, and so forth. Linguistic analyses are indispensable for setting up these linguistic norms. And the movable type printing system was utilized most effectively for proselytizing in the sixteenth and seventeenth century Christian world, not only by Protestants but also by Catholics.
References Bailey, Don C. 1962. The Rakuyōshū [Collection of fallen leaves]. Monumenta Nipponica XVI. 289–376/XVII. 214–264. Tokyo: Sophia University. Harada, Hiroshi. 2011. Kirishitanban rahojitsujisho no genten Calepinus ratengo jisho no keifu [Geneology of Calepinus’ Latin dictionary, the source of the Jesuit mission press Latin-Portuguese-Japanese dictionary]. (Private edition) Kamei, Takashi. 1978. Covering and covered forms of women’s language in Japanese. Hitotsubashi Journal of Arts & Sciences 19(1). 1–7. Tokyo: Hitotsubashi University. Kishimoto, Emi. 2012. Colloquial Japanese expressions in the Dictionarium Latino Lusitanicum, ac Iaponicum (1595). Paper presented at the seventh international conference on missionary linguistics, the University of Bremen, 28 February–2 March. Morita, Takeshi. 1989. Hōyaku nippojisho sakuin [Index for Japanese translation of Vocabulario da lingoa de Iapam]. Tokyo: Iwanami Shoten. Ōtsuka, Mitsunobu. 2006. Kirishitan no nihongo kenkyū [Study of Japanese by christians]. Kokubungaku kaishaku to kyōzai no kenkyū [Study of teaching materials of Japanese literature] 51(11). 60–69. Schurhammer, Georg and Josef Wicki (eds.). 1944–1945. XAVIER, Saint Francis. Epistolae S. Francisci Xavierii aliaque eius scripta [The letters of Saint Francis Xavier and other writings of his]. 2 vols. Roma: Historical Institute of the Society of Jesus. ✶✶✶
Vocabulario = Vocabulario da Lingoa de Iapam com a declaração em Portugues, feito por algvns Padres, e Irmaõs da Companhia de Iesv. Com Licença do Ordinario, & Superiores em Nangasaqui no Collegio de Iapãm da Companhia de Iesvs. Nagasaki. 1603–1604 LPJ = Dictionarium Latino Lusitanicum, ac Iaponicum ex Ambrosii Calepini volumine depromptum: in quo omissis nominibus proprijs tam locorum, quàm hominum, ac quibusdam alijs minùs vsitatis, omnes vocabulorũ significationes, elegantioresq; dicendi modi apponuntur: in vsum, & gratiam Iaponicae iuuentutis, quae Latino idiomati operam nauat, nec non Europeorũ, qui Iaponicũ, sermonem addiscunt. Amakusa.1595. VLCP = Vocabulario da lingoa Canarina com versam portuguesa (manuscripts) VLKK = Vocabulario da lingoa canarim, feito pellos Padres da Companhia de Iesus, que residem na Christandade de Salcete, & nouamente acressentado com uarios modos de fallar pello Padre Diogo Ribeiro da mesma Companhia.1626 (manuscripts)
Takashi Inukai
21 What mokkan (wooden documents) can tell us about ancient Japanese language 1 Origin and definition of mokkan In ancient East Asia, Chinese characters were the only medium of writing. Originally, in the 14th–11th centuries B.C., they were engraved on tortoise shells or cattle bones with a stylus. The characters were used as a fortune-telling medium concerning political and economic policies of an empire, judged through how the characters cracked when heated over a flame. Later, characters were engraved on the surface of bronzeware or stone monuments as media for praying to and glorifying the gods. Much later, Chinese characters were used as the medium for recording the ordinary Chinese language, written with a brush in Chinese ink on the surface of various materials including silk cloth, wood, and bamboo. Long sentences were written on the surface of slender wooden or bamboo boards. The size of these boards measured about 20–30 cm long, 1–2 cm wide, 3–10 mm thick, with the surfaces planed flat. Characters were arranged in a single vertical line on the surface of each board. When the messages were long, the slender boards were tied up together with a thread. This was the original form of Asian books, as we recognize in the shape of the Chinese character 冊. During the Qin dynasty in the 3rd century B.C., Chinese characters were widely used as the medium for everyday administration. Many official documents, records, accounts, and notices were written in the government offices. Bureaucratic matters were written on the surface of small wooden boards. These boards measured about 20–30 cm long, 3–4.5 cm wide, 5–15 mm thick, with the surfaces planed flat. Characters were arranged in multiple vertical lines on the board, and the string of characters frequently ran to the reverse side. Public servants stored these boards as if they were mass data in a computer in our time and exchanged them as if they were e-mails. Servants always carried a knife and, after they had finished communication, they shaved the surface of the boards, and reused them for the next business. With the invention of paper before the 2nd century A.D., a differentiation in writing habits began: long and formal writings came to be written on paper, and paper books took the place of tied bamboo board books, whereas short and daily documents were written on wooden small boards. Paper was lightweight and not bulky, but it was expensive in those days and only allowed for a maximum of two repeated uses. On the other hand, wood had recyclability, was portable and, furthermore, comparatively cheap. This was the reason why wooden documents were widely preferred for daily bureaucratic business. Among Chinese scholars, the above-mentioned boards are generically called jiǎn dú (簡牘; Jpn. kantoku). Slender boards, written in single line, are called jiǎn (簡; kan). https://doi.org/10.1515/9781614512851-022
464
Takashi Inukai
Therefore, wooden jiǎn are called mù jiǎn (木簡; mokkan), bamboo jiǎn are called zhú jiǎn (竹簡; chikukan). And small wooden boards, written in multiple lines, are called mù dú (木牘; mokutoku). In addition to them, tetrahedral sticks were used as writing materials. Occasionally hexahedrons were also used, on which somewhat long messages such as formal letters and phrases extracted from books were written among other documents, known as gū (觚; ko). Among Japanese scholars, all of them are generically termed mokkan. When Chinese characters were introduced to Japan for everyday administration in the 6th–7thth centuries, both paper and wood were available. Long and formal documents were written on paper. Chikukan were not used and ko were seldom used. Consequently, Japanese scholars use the term mokkan as almost the same definition as Chinese mokutoku, in a narrow sense. Japanese mokkan in its wide sense is a generic term, and has various uses, sizes, shapes, materials, and so on. While conducting administrative business, ancient Japanese people also wrote notes or memos, doodles, etc. and practiced calligraphy on wooden materials. They were scraps or pieces of wood such as repeatedly used mokkan, and scraps of wooden buckets for instance. These are all called mokkan in a wider sense. “Wooden documents” is the English term used among Japanese scholars. Korean scholars use the same terms as the Japanese do.
2 Outline of Japanese mokkan Over the past several decades, many wooden pieces have been unearthed from under ancient Japanese ruins. On many of them, Chinese characters written in Chinese ink can be identified. Most of these strings of characters represent daily government affairs, and at times they also provide other types of information, as mentioned above. In 1961, when dozens of them were unearthed in Nara, the old Japanese capital area, scholars named them collectively mokkan. Mokkan have since been unearthed in the older capital areas Fujiwara and Asuka and from the remains of old government offices throughout all parts of Japan including the Iba Site in Shizuoka Prefecture, the Yashiro Site in Nagano Prefecture, the Kannon-ji Site in Tokushima Prefecture, the Dazaifu Site in Fukuoka Prefecture, as well as the Taga Castle Site in Miyagi Prefecture. Currently, the total number of mokkan is estimated to amount to over three hundred thousand (as of 2019, restricted to those from ancient times). The oldest mokkan are believed to stem from the first quarter of the 7th century, and the majority of them originate from the period between the late 7th century and early 9th century. Information can be gleaned about Japanese mokkan from a variety of sources. The Research Institute for Cultural Properties, Nara [奈良文化財研究所] provides Wooden Tablet Database [木簡庫 mokkanko.nabunken.go.jp] on the Internet, which can be utilized free of charge. The journal Proceedings of the Japanese Society for the Study
21 What mokkan (wooden documents) can tell us about ancient Japanese language
465
of Wooden Documents [木簡研究] is issued annually. Many photo-books of mokkan, together with transcriptions and commentaries, have been published by each institution in charge of the site. Furthermore, bulletins and excavation reports published by local administrative organizations are readily available. See “Supplemental materials” below in the reference section for fuller details. Some interpretative books on mokkan have also been published, including for example Inukai 2005 (revised 2011), Nara Bunkazai Kenkyūjo 2020, and Yoshimura et al. 2020.
3 Significance of mokkan as data of ancient Japanese language As long as proper procedures are followed, we can regard these strings of Chinese characters written on mokkan as representing words and sentences of 7th–8th century Japanese language. Frequently, these Chinese characters are arranged in the Japanese word order, not the regular Chinese order. Also, these Chinese characters assume Japanese meanings, different from the original meanings of Old Chinese. Speakers of Old Japanese adapted Chinese characters as appropriate for Japanese language and customs. As previously mentioned, mokkan were usually reused. The surface was shaved off several times, and when they were disposed of, the mokkan were often cleaved because of confidentiality of the written matter. Therefore, most unearthed wooden pieces are scraps. This means that whether surviving characters on the wooden pieces retain their original forms or not is of great importance. If it proves that the characters are identified as a part of original strings, then it provides important evidence for the original forms in which these characters were used. Scholars have to bear these points in mind when they transcribe the forms of Chinese characters written on the wooden pieces and interpret the strings of characters. In studying mokkan, interdisciplinary approaches combined with Japanese history and archaeology are required on the part of scholars with a view to learning about the details of the remains. Take the following example. 大粮綿 十月料/□八日 国勝列一人凡至 古志列一人辛人府生 悪閇干一人正 身/三人遣 On this mokkan unearthed in the capital area Nara, the character 干 is hardly distinguishable from 于 when it is handwritten. 于 occasionally stood for the Old Japanese accusative case particle wo. However, by making use of historical knowledge we can declare this seemingly ambiguous character to be 干 because, in 7th century Japanese mokkan this character was used to represent the allocation of kitchen work in the official works imposed as national tax. The whole string of Chinese characters on this mokkan may be deciphered as follows:
466
Takashi Inukai
Cotton cloth as fee for the labor in the 10th month / ✶8th day. 国勝 [Kuni-no-suguri] team share one person,凡至 [Oposi-no-Itari]. 古志 [Kwosi] team share one person, minor official 辛人 [Karapito]. Cook 悪閇 [Asipe] share one person, himself . / Three people sent.1
This is a payment slip of the official business engaged in provision of meals. Thus, mokkan prove to be highly useful for studying the characteristics of the 7th–8th century Japanese language. There are three main reasons for this. First, mokkan are original, primary linguistic sources, in stark contrast to texts like Kojiki (Records of Ancient Matters), Nihon shoki (Chronicles of Japan) and Man’yōshū (Anthology of Ten Thousand Leaves). Texts such as those were repeatedly hand-copied, and liable to corruption over time. It is the manuscripts written in medieval times that are available today. Second, mokkan are believed to reflect the common and colloquial aspects of Old Japanese, because they were used in everyday business. In contrast, classical literature tends to reflect the formal and literary aspects, and we should fully realize that they reflect partial aspects of 8th-century Japanese language. Third, mokkan have usually been discovered together with many other historic items, which can be convincing evidence of when and why these old documents were written. Many research papers have been written on 7th–8th century Japanese language. They are, however, based upon the evidence that was found through the analyses of the classical literature. With mokkan as evidence, it is possible to uncover daily and colloquial aspects of 7th–8th century Japanese language, which can lead people to grasp the whole picture of the 7th–8th century Japanese language more correctly.
4 Case studies 4.1 A lack of letters transcribing the voiced initial-consonant syllables in the Japanese syllabic system kana In the consonant system of the Japanese language, the distinctive feature ‘voiceless’ versus ‘voiced’ is a phonological opposition. Noto ‘Noto’ place name’ vs. nodo ‘throat’ is a good modern example. The two words are distinguished just by /t/ and /d/. In this case, the feature voiceless vs. voiced shows a minimal pair distribution. In 7th–8th century Japanese, the phonological opposition was approximately the same as in present-day Japanese,2 and, in classical literature, ways of transcription were 1 国 family name, 勝 clan name or given name; 凡 family name, 至 given name; 古志 family name or given name; 辛人 given name; 悪閇 given name (means ‘so-so quality dish’). 2 Among Japanese scholars, the main remaining issue concerns the nasal glide preceding voiced consonants.
21 What mokkan (wooden documents) can tell us about ancient Japanese language
467
also the same in principle as those in present-day Japanese. When words are spelled in the Japanese syllabic kana system in the present-day Japanese writing system, voiced consonants are written by adding the mark ‶ on the upper right. For instance, と represents /to/ and ど /do/. In the 7th–8th centuries, however, they were distinguished by Chinese characters. In the Kojiki, for example, 登 was used to represent /to/, and 杼 to represent /do/. Similarly, 多 and 陀 represented /ta/ and /da/, respectively. 知 was voiceless /ti/ and 遅 was voiced /di/. 都 and 豆 represented /tu/ and /du/, respectively. /te/ and /de/ were written as 弖 and 傳, respectively. This writing system holds true in the Nihon Shoki and the Man’yōshū. However, on mokkan, the syllable /to/ is transcribed with the Chinese character 止 in almost all cases, only occasionally with 等, and never with 登, which represented /to/ in the Kojiki as described immediately above. This seems to indicate that writing with 登 was expected in formal and literary documents, and that 止 was expected in the daily and colloquial style of writing. Between classical literature and mokkan, we can notice a similar distinction in the choice of Chinese characters used for the transcription of Japanese syllables. In the Kojiki, for instance, /ki/, /no/, /ma/, and /ri/ are transcribed as 岐, 能, 麻 or 摩, and 理, respectively, and in mokkan as 支, 乃, 万 or 末, and 利. The original Chinese pronunciations of the former were suitable for the pronunciation of Japanese syllables in the 7th–8th centuries. In contrast, those of the latter were not necessarily suitable. The original Chinese pronunciation of 支 around 700 A.D. was suitable for Japanese syllable /si/, but not for /ki/. The Chinese pronunciation of the above-mentioned 止 was also suitable for /si/, not for /to/. These transcriptions of /ki/ and /to/ were based on very old Chinese pronunciations introduced via the Korean Peninsula. The original Chinese pronunciation of 万 had word-final consonant [-n], and 末 [-t]. They were not well-matched to Japanese pronunciation because Japanese syllables at the time always ended in vowels. The reason why 支, 乃, 万 or 末, and 利, and also 止, were preferred in daily style of writings can be ascribed to the fact that they were convenient when written because of fewer strokes, and that they were long rooted in Japanese writing system. Here it is noteworthy that, on mokkan, the use of those Chinese characters like 杼, which exclusively transcribe syllables in which the word-initial consonant is voiced has scarcely been found. Voiced syllables are transcribed with the same characters that are used to transcribe syllables with word-initial voiceless consonants. For example, a string of characters 加々美 was found on one of the mokkan unearthed in Hyogo Prefecture. 加 々美 is a phonographic writing of /kagami/ ‘mirror’. In this string, the first character 加 represents /ka/, and the second character 々, which is used to show the repetition of the preceding character, represents /ga/. Here the same character 加 has a twofold function: voiceless [k-] and voiced [g-]. In the Kojiki, however, 加賀美 is used for /kagami/, where 加 represents /ka/ and /ga/ is written by a different Chinese character 賀. The question of why these Chinese characters, which exclusively transcribe voiced initial-consonant syllables, were not in daily use is one of the debatable points in the
468
Takashi Inukai
history of the Japanese language. It is true that characters which express the voiced initial-consonant syllables have more strokes when written, such as the pair 賀 /ga/ and 加 /ka/, and that the minimal pairs based upon the opposition of being voiceless or voiced does not occur so often because the voiced initial-consonants did not occur in word-initial position in Old Japanese. Furthermore, originally voiceless consonants can become voiced in compound words in the Japanese phonology, such as /pa/ ‘leaf’ changing into /ba/ in a compound such as /puta-ba/ ‘cotyledon’. When /pa/ is combined with /puta/, /-p-/ becomes /-b-/ in assimilation to the surrounding vowels. People in the 7th–8th centuries may have been aware of this phenomenon. Actually, the Japanese syllabic kana system lacks letters which are exclusively used for transcribing voiced initial-consonant syllables, using a diacritic mark to indicate voicing, as described earlier. Consequently, it is evident that modern kana are the descendants of those Chinese characters that were used on mokkan for the transcription of Japanese.
4.2 Usage signifying choice from among things or people One of the mokkan which were unearthed in the remains of the capital area Nara has the string of Chinese characters 一々物. The remains were the residence of Nagaya-Ō, who was an Imperial prince and prime minister from 724 to 729 A.D. This string of characters 一々物 represents the Japanese phrase pitotu-pitotu meaning ‘choose one, single out one from among things’. The entire message conveyed on this mokkan is a request asking for sending either non-glutinous rice or millet to a noblewoman. Another mokkan unearthed from the same remains had the string of characters 一々人, representing the Japanese phrase pitori-pitori ‘single out one person from among people and dispatch him’, a phrase in a formal letter of invitation. From a grammatical point of view, we should consider that there was a sentence pattern at the time in which the choice from among things or people was expressed by repeating the same word. However, in the Man’yōshū poetry anthology compiled after the second half of the 8th century, phrases such as pitotu-pitotu or pitori-pitori are not found, despite the fact that it contains numerous sentence patterns from 7th–8th century Japanese. Similar phrases, however, occur in the literature written in and after the 10th century. In the Taketori monogatari (Tale of the Bamboo Cutter) written in the first half of the 10th century, for example, the old bamboo cutter says to the female protagonist, pitori-pitori-ni api-tatematuri-tamapine, meaning “Allow me to suggest that you choose one aristocrat (from among the five candidates) for your husband.” In the Utuho monogatari (Tale of the Hollow Tree) written in the latter half of the 10th century, the phrase pitotu-pitotu is found. This phrase occurs in a part of a Japanese poem and refers to choosing either a crane or a pine as a gift, both of which were words symbolizing and celebrating longevity in those days. All of this indicates that in 8th-century Japanese, while the above stated pattern signifies choice from among things or people expressed by repeating the same word was
21 What mokkan (wooden documents) can tell us about ancient Japanese language
469
not used in poetic speech level, it was in fact already in actual use in daily speech, and in and after the 10th century this usage spread to other linguistic speech levels including literary and poetic ones.
4.3 Usage of the Chinese character 賜 as a Japanese/Korean honorific auxiliary verb Since the Korean language and the Japanese language are syntactically similar and yet phonologically dissimilar, the parallel with unearthed Korean materials is very significant. The string of Chinese characters 勞賜時, found in the inscription on the mandorla (an almondshaped panel or decorative space) of a statue of the Buddha in the Hōryū-ji temple, is a pertinent example. It is assumed that this inscription was written towards the end of the 7th century. Originally in Chinese, the verb 賜 meant ‘(for somebody in a higher position, like an aristocrat) to give (to someone of lower status)’. In the inscription above, however, 賜 attached to the verb 勞 is used to represent the Japanese honorific auxiliary verb tamapu in Japanese word order. The verb 勞 means ‘to suffer from illness’, the subject in this case is understood to be the Emperor, and 時 is a conjunction meaning ‘when’. As a whole, the string 勞賜時 respectfully conveys the meaning ‘when the Emperor contracts an illness’. It is likely that this usage of 賜 as an honorific expression was influenced by Korean writing. For example, one of the mokkan unearthed near the 7th-century capital area of Silla in Korea has the string of Chinese characters 牒垂賜, in which 賜 is an honorific word attached to the verb 垂 meaning ‘(for somebody in a higher position) to hand something over (to a person in a lower position)’, and 牒 is an object referring to this particular mokkan. Furthermore, as a whole, 牒垂賜 politely conveys the following sense: ‘the government hands over this imperial order’. These Chinese characters typically represent the Silla-language word order. In the 7th–8th century Japanese mokkan, however, 給 is used in most similar honorific expressions. As 給 and 賜 were occasionally interchangeable in the original Chinese usage, both Koreans and the Japanese could use 給 in the place of 賜. While in Korea this usage of 賜 was taken over to the 8th century possibly because the Chinese pronunciation of 賜 resembled that of Korean honorific infix /si/, Japanese public servants chose to use the character 給 in daily business because this character had fewer strokes in writing and the Japanese pronunciation of 賜 /tamapu/ was quite different from the Chinese one.
4.4 Writing systems employed for the transcription of Old Japanese poems The Man’yōshū records 4516 Old Japanese poems. According to the epigraphs added to each poem, the earliest poem was composed in the first quarter of the 7th century.
470
Takashi Inukai
Although a few of these poems are thought to have been composed in the 5th century, this is merely an oral tradition handed down from generation to generation. The problem of when and how these poems were transcribed by Chinese characters remains to be solved. The poems in the Man’yōshū are written with both the Japanese readings (primarily logographic writing, but also including phonographic kungana) and the Chinese readings (phonographic ongana) of Chinese characters. The majority of them have a combination of the two readings as in the following system: nouns and verbs are transcribed with the Japanese readings, and inflections and particles with the Chinese readings. Whereas some poems are written exclusively with the Chinese readings of Chinese characters in the system which assign each character to represent one Japanese syllable. A few poems are written exclusively with the Japanese readings without transcribing inflection and particle. The origin of these writing systems in the 7th century can be solved in the light of the following fact: In all of the mokkan written in verse in 7th century and early 8th century, that have been unearthed since 1989, the poems adopt exclusively the writing system in which each character represents one Japanese syllable. And in these mokkan written in verse, the Chinese characters are used mostly as phonographic ongana, occasionally as phonographic kungana, but never as logographic writing. The next question arises: Why was the writing system based mostly upon the Chinese readings adopted in spite of the fact that it was possible for people in the 7th century to write Japanese sentences with the Japanese readings of Chinese characters? Actually, on mokkan written in prose for daily business, the order of characters in logographic writing occasionally shows Japanese word order. The situation was the same in ancient Korea. For instance, on the Monument with Inscription of Hwarang’s Oath (壬申誓記石) written in 552 or 612 A.D., the order of characters shows clearly Korean word order. The answer is that Chinese readings were suitable for precisely transcribing pronunciations of Japanese poems. Japanese readings may have been used in daily communication because they were more easily understood. Nevertheless, just as ‘石’ was pronounced as /ipa/ at one time and /isi/ at another, so a given character could permit plural pronunciations depending on how it occurred in a word or phrase in those days. The present author supposes that, in the 7th–8th centuries, Japanese poems were recited with melody at various ceremonies. A few 7th–9th century mokkan written in verse have a special form apt for display in ceremony. They are actually twice as long and have larger characters than ordinary mokkan, in addition to being written vertically only on the front of the mokkan, never running to the reverse side. By contrast, only one or two mokkan written in verse in the mid-8th century adopt the same writing system as the majority of the poems in the Man’yōshū did, i.e., the system in which nouns and verbs are transcribed with the Japanese readings, and inflections and particles with the Chinese readings. This seems to have the implication that there was progress in the Japanese writing system in daily and colloquial aspects of
21 What mokkan (wooden documents) can tell us about ancient Japanese language
471
8th century Japanese language, just as the systematic similarity can be detected in present-day Japanese, where nouns and verbs are transcribed by means of Chinese characters and inflections and particles by the Japanese syllabic system kana. Thus, all of the discussion above prompts us to conclude that various writing systems employed in the Man’yōshū reflect the literary expressions in those days. Moreover, it may be the result of the work of the editor(s) after the second half of the 8th century. The editor(s) chose the writing systems most appropriate for each poem.
5 Wooden documents as universal media It is well-known that Roman records written in the 2nd century A.D. were unearthed in Vindolanda in England. These are documents written on wooden wax-filled tablets with metal pens or on wooden sheets in ink. The wooden sheets measure some 18cm long, 9cm wide, and 1.5–3mm thick, and have inscriptions in the Roman alphabet representing military reports and procedures, the amount of supplies, formal and private letters, and writing exercises, to name a few. The writers were soldiers and their family members. Old Roman records of these sorts have also been unearthed in Italy and Switzerland as well. In Novgorod in medieval Russia, white birch sheets were used for writing records, reports, letters, writing exercises, etc. Many of them are reports sent from farming areas to Novgorod. The oldest of them was written in the first half of the 11th century and the latest in mid-15th century. The inscriptions were engraved on the sheets in the Cyrillic alphabet with iron or bone pens. Their writers include aristocrats, merchants, priests, and farmers. Also, in medieval Norway, short slender wooden boards were utilized by merchants of the Hansabund (the Hanseatic League) to record business dealings, and on the boards runic letters were engraved with metal pens. As shown above, it is evident that using wooden short pieces for daily written communication and recordings was universal for people in ancient times, whether in Asia or in Europe. This means that the comparison of them is of great significance in studies of languages.
References Inukai, Takashi. 2005. Mokkan ni yoru nihongo shokishi [A history of Japanese writing based on wooden documents]. Tokyo: Kasama Shoin. (Revised in 2011.) Nara Bunkazai Kenkyūjo [Nara National Research Institute for Cultural Properties] (ed.). 2020. Mokkan: Kodai kara no tayori [Wooden documents: Voices from Ancient Japan]. Tokyo: Iwanami Shoten. Yoshimura, Takehiko, Shinji Yoshikawa, Akio Kawajiri (ed.). 2020. Moji to kotoba [Script and culture], (Shirīzu: Kodaishi o hiraku [Series: Opening up ancient Japanese history]). Tokyo: Iwanami Shoten.
472
Takashi Inukai
Supplemental materials Hamamatsu-shi Kyōiku Iinkai [Hamamatsu City Municipal Board of Education] (ed.). 2008. Iba iseki sōkatsu hen: Bungaku siryō, jidai betsu sōkatsu [Iba site, a comprehensive report: Written materials, summary by period]. Hamamatsu: Hamamatsu-shi Kyōiku Iinkai. Kizugawa-shi Kyōiku Iinkai [Kizugawa City Municipal Board of Education] (ed.). 2014. Kizugawa-shi maizō bunkazai chōsa hōkokusho dai 16 shū: Kamio-dera ato “Baba-minami iseki” hakkutsu chōsa hōkokusho [Kizugawa City cultural properties report, no. 16: Excavation Report, Kamio-dera temple site, i.e. Baba-minami site]. Kizugawa, Kyoto: Kizugawa-shi Kyōiku Iinkai. Kyōto-shi Maizō Bunkazai Kenkyūjo [Kyoto City Archaeological Research Institute] (ed.). 1997. Nagaoka sakyō shutsudo mokkan [Wooden documents excavated from the left capital (sakyō) of the Nagaoka capital site], vol. 1. Kyoto: Kyōto-shi Maizō Bunkazai Kenkyūjo. Kyūshū Rekishi Shiryōkan [Kyushu Historical Museum] (ed.). 2002. Dazaifu seichō ato [The main government office of Dazaifu]. Fukuoka: Kyūshū Rekishi Shiryōkan. Kyūshū Rekishi Shiryōkan [Kyushu Historical Museum] (ed.). 2014. Dazaifu seichō shūhen kanga-ato V: Fuchō chiku ibutsu hen 2 [Administrative structures surrounding Kyushu’s government headquarters in Dazaifu complex V: The Fuchō area 2]. Fukuoka: Kyūshū Rekishi Shiryōkan. Miyagi-ken Tagajōseki Chōsa Kenkyūsho [Miyagi Prefecture Research Institute of the Taga Castle Site]. 2011, 2013. Tagajōseki mokkan [Wooden documents from the Tagajō site]. https://www.thm.pref. miyagi.jp/kenkyusyo/pdf/tagajomottukan.pdf. https://www.thm.pref.miyagi.jp/kenkyusyo/pdf/ tagajomottukan2.pdf Mokkan gakkai (ed.). 2003. Nihon kodai mokkan shūsei [Compendium of Japanese ancient wooden documents]. Tokyo: University of Tokyo Press. Mukō-shi Kyōiku Iinkai [Mukō City Municipal Board of Education] (ed.). 1984, 1993. Nagaoka-kyō mokkan [Wooden documents from the Nagaoka capital site], vol. 1–2. Muko, Kyoto: Mukō-shi Kyōiku Iinkai. Nagano-ken Maizō Bunkazai Center [Archaeological Research Center of Nagano Prefecture] (ed.). 1996. Nagano-ken Yashiro iseki gun shutsudo mokkan [Wooden documents excavated at the Yashiro sites]. Nagano: Nagano Maizō Bunkazai Center. Nara Bunkazai Kenkyūjo [Nara National Research Institute for Cultural Properties] (ed.). 1969–2010. Heijō-kyū mokkan [Wooden documents from the Nara palace site], vol. 1–7. Nara: Nara Bunkazai Kenkyūjo. Nara Bunkazai Kenkyūjo [Nara National Research Institute for Cultural Properties] (ed.). 1995, 2001, 2006. Heijō-kyō mokkan [Wooden documents from the Nara capital site], vol. 1–3. Nara: Nara Bunkazai Kenkyūjo. Nara Bunkazai Kenkyūjo [Nara National Research Institute for Cultural Properties] (ed.). 1978, 1980, 2012, 2019. Fujiwara-kyū mokkan [Wooden documents from the Fujiwara palace site], vol. 1–4. Nara: Nara Bunkazai Kenkyūjo. Nara Bunkazai Kenkyūjo [Nara National Research Institute for Cultural Properties] (ed.). 2007, 2009. Asuka-Fujiwara-kyō mokkan [Wooden documents from the Asuka-Fujiwara capital sites], vol. 1–2. Nara: Nara Bunkazai Kenkyūjo. Tokushima-ken Maizō Bunkazai Center [Archaeological Research Center of Tokushima Prefecture] (ed.). 2011. Kenkyū kiyō “Shin-shu” [Bulletin of research “Shin-shu”], vol. 9. Tokushima: Tokushima-ken Maizō Bunkazai Center.
Shōju Ikeda
22 Early Japanese dictionaries 1 Introduction The oldest extant Japanese dictionary of Chinese characters, the Tenrei banshō meigi 篆隷万象名義, an early 9th century dictionary in which characters are arranged by their radicals, is an abridged version of the Chinese Yupian 玉篇. Two more dictionaries were compiled in the 10th century that included a large number of Japanese renderings (wakun 和訓) written in man’yōgana phonograms: one is the Shinsen jikyō 新撰字鏡, which lists the characters by radical, and the other is the Wamyō ruijushō 倭名類聚 抄, which arranges the characters by meaning. In the 11th century Buddhist glossaries (ongi 音義) took a step forward and began recording the Japanese meaning and the pronunciation (jion 字音) of Chinese terms in katakana. The 12th century saw the appearance of the Ruiju myōgishō 類聚名義抄, a radical-arranged dictionary with Japanese renderings in katakana, and the Iroha jiruishō 色葉字類抄, a dictionary arranged by meaning whose entries are grouped according to the iroha poem sequence. Overall, these Heian-period dictionaries are key sources for a better understanding of the Japanese language and its evolution over time.
2 Previous studies Yoshida (1971) paved the way for a series of studies providing a general overview of the history of Japanese dictionaries. This research has demonstrated the importance of early dictionaries as resources for the study of the history of the Japanese language. Nishizaki (1995)’s handbook, for example, is a quick and convenient reference tool that includes a description of the main dictionaries compiled in Japan as well as a detailed index of bibliographic references (including photographic reproductions, concordance indexes, and suggestions for further readings, among other things).1 In the case of Heian-period texts,
Acknowledgments: This work was supported by JSPS KAKENHI Grant Number 25370506, 16H03422. 1 Kawase (1955, 1977) conducted pioneering studies of old dictionaries from a bibliographic point of view infused with a cultural approach. A description of the main dictionaries and a list of relevant references can be found in Kokugo Gakkai (1980), Hida et al. (2007), and Okimori et al. (1996). Tsukishima (1953), Toshio Yamada (1978), Nakada (1983), and S. Ikeda (2011a) are solid introductory works on the subject. T. Takahashi and H. Takahashi (2006) focused on transcribing and annotating the introductions and colophons of old dictionaries. Yayoshi (1964) studied the diachronic history of encyclopedic dictionaries. The photographic reproductions and concordance indexes are too many to list, but they can be easily accessed with a simple query on such research engines as CiNii (http://ci.nii.ac.jp/ja). Bailey (1960) https://doi.org/10.1515/9781614512851-023
474
Shōju Ikeda
Tsukishima (1964) stands out as the basic reference for the understanding of dictionaries through the lens of vernacular reading practices, or kundoku. More recently, K. Yamada (1995) wrote about the approach and methodology with which these sources ought to be examined.2 The works of Tadao Yamada (1967), Sadakari (1998), and Miyazawa (2010) are equally important. I have written elsewhere on the research methods employed by Sadakari and Miyazawa (S. Ikeda 2008), which begin with the premise that dictionaries are linguistic sources and move on from there. The main topics of concern in dictionary studies include such issues as authorship, nature of the work, sources, organization, alternate versions, and influence on later works. These topics will be touched in this chapter, focusing on how early Japanese dictionaries have been used as linguistic sources. Dictionaries are classified depending on which element of a character – its form, pronunciation, or meaning – is used to arrange the entries. As a result we have jisho 字書, radical-based dictionaries that arrange the characters according to their form; insho 韻書, rhymers that list the characters based on pronunciation; and gisho 義書, which organize the characters according to their meaning.3 Other categories include Buddhist glossaries, allograph dictionaries (jiyō 字様 Ziyang character models), encyclopedic dictionaries (ruisho 類書), dictionaries of poetic words (kago jisho 歌語辞書), herbals (honzōsho 本草書), and Sanskrit dictionaries (bongo jisho 梵語辞書). Below, the description of these dictionaries will be unfolded in the order stated above.
3 A prehistory Historical records indicate that the earliest Japanese dictionary was the Niina 新字, a work that unfortunately no longer exists but whose existence is recorded. Arranged in 30 volumes, it was compiled in 682 by Sakaibe no muraji Iwatsumi 境部連石積 and other scholars. For the entries, the “modern” Tang dynasty characters (kinji 今字) were adopted. The consensus is that it would have included a comparison between kinji and “older” characters (koji 古字). Each character was accompanied by an explanation of its meaning (Kojima 1979). The Niina is also believed to be somehow related to the compilation of the Teiki
offers a good English-language overview of the history of Japanese lexicography from the beginning of the Heian period to the end of the Muromachi period (1333–1600). Ōshima (1997) is solid introductory to the history of Chinese lexicography. 2 Yoshioka (2006) also examined dictionaries through Japanese historical studies. 3 K. Ueda and Hashimoto (1916: 295–297) organize dictionaries according to the two principles of structure and function. Under “structure”, they include jikeibiki jisho 字形引辞書 (characters arranged by form), bunruitai jisho 分類体辞書 (characters arranged by meaning), and onbiki jisho 音引辞書 (characters arranged by pronunciation). Under “function”, they list dictionaries for reading and dictionaries for writing, where jikeibiki are dictionaries for reading, bunruitai for writing, and onbiki dictionaries that served both purposes.
22 Early Japanese dictionaries
475
帝紀 and of the Ritsuryō 律令 (Yoshimura 2004). The Kitaōtsu iseki shutsudo “ongi mokkan” 北大津遺跡出土「音義木簡」dates back to the second half of the 7th century, corresponding to the reign of Emperor Tenmu and to the compilation of the Niina. This wooden tablet fragment is considered to be the earliest proof of the existence of a vernacular reading of Chinese in Japan. It deploys, for the first time, the allograph of the character 誣 coupled with a string of phonograms that are different from the plain form of a verb; the verb is meant to be read in Japanese in the appropriate context (Inukai 2005). 阿佐ム (1) 「誈 加ム移母 」 → 阿佐ム加ム移母(アザムカムヤモ) azamuka-mu yamo deceive-conjec part ‘[one] would be going to deceive [someone else]. . .’
The Wamyō ruijushō 倭名類聚抄, which I discuss below, refers the Yōshi kangoshō 楊 氏漢語抄 and the Benshiki ryūjō 弁色立成 as 8th century dictionaries. Commentaries and glossaries provide further evidence for the adoption of Chinese dictionaries. The Yōshi kangoshō was compiled in ten sections during the Yōrō era (around 720), while the Benshiki ryūjō was edited in eighteen chapters. Quotations from both texts are very similar to the Rinji zōyōji 臨時雑要字 section of the Shinsen jikyō 新撰字鏡 (see Section 3 below) and, for this reason, they are both considered dictionaries organized by meaning, meant for practical-everyday use, in which the Chinese terms were explained using man’yōgana phonograms. Such dictionaries tend to be listed under the general category of kangoshōrui 漢語抄類 ‘Bilingual Chinese-Japanese dictionaries’.4 While kangoshōrui refers to dictionaries made in Japan, they do have Chinese counterparts that have been identified. For example, the Suwu yaominglin 俗務要名林 (S.617, P.2609, P.5001)5 found in Dunhuang features a similar structure, with Chinese terms arranged by meaning accompanied by a description of their meanings and Chinese character pronunciations (Zhang 2012). More often than not it appears that scholars in this period tended to use the Yupian 玉篇, a Chinese dictionary in 30 volumes, compiled by Gu Yewang 顧野王 of the Liang 梁 dynasty in 543 and arranged by radical.6 Another well-known text of the same period, 4 See chapters in Sadakari (1998: 110–143)’s Shinsen jikyō ‘Rinji zōyōji’ to Kangoshō, and Miyazawa (2010: 409–427)’s Wamyōshō to Kangoshō. 5 The vast majority of the Dunhuang manuscripts are available online on the International Dunhuang Project (http://idp.bl.uk/). “S.” designates the manuscripts collected by Aurel Stein (1862–1943) and held by The British Library, while “P.” indicates the manuscripts collected by Paul Pelliot (1878–1945) and preserved in the Bibliothèque nationale de France. The Pelliot collection is also available at Gallica: http:// gallica.bnf.fr/. 6 Extant volumes of the Yupian include numbers eight, nine, eighteen, nineteen, twenty-two, twenty-four, and twenty-seven. See Okai (1933), and T. Ueda (1970). Takata (1987, 1989) has examined the details of fragments S.6311 and Dx1399 (held by Institute of Oriental Manuscripts, St. Petersburg) from Dunhuang. The main studies on these fragments are those by Okai (1933), Mabuchi (1952), and T. Ueda (1985, 1986).
476
Shōju Ikeda
the five-volume rhymer Qieyun 切韻 edited by Lu Fayan 陸法言 of the Sui dynasty in 601, seems to have been less popular in Japan.7 The fifty-volume Ryō no shūge 令集解 of the Jōgan era (859–877) (of which only 35 volumes exist now) by legal scholar Koremune no Naomoto includes quotations from commentaries on the Nara-period Ritsuryō. Unsurprisingly, a closer investigation reveals that the main source for the koki 古記 ‘Ancient records’ section edited around 738 was the Yupian. In the ryōshaku 令釈 section compiled around 787–791, however, the Qieyun is cited more extensively than the Yupian (Hayashi 1976). This shift indicates that, starting in the second half of the Nara period, the Qieyun began to be used extensively within Japanese scholarly circles.8 The Buddhist commentaries by Zenju and Myōichi Konkōmyō saishōō kyō chūshaku 金光明最勝王経註釈 (late 8th century), by contrast, tend to refer to the Yupian more often than to the Qieyun.9 Especially noteworthy is the fact that the Japanese meanings of words were given in a number of different ways. Sometimes they were given by quoting from kangoshō, such as in the following lines from the koki section of Ryō no shūge: 古記云。輿無輪也。輦有輪也。漢語抄云。輿。母知許之。腰輿。多許之。跡云。輦者。 己之久留萬 ‘According to the ancient records, 輿 (yú, ‘sedan chair’) has no wheels. 輦niăn has wheels. 輿, motikosi. Kosi (vernacular reading of 腰) 輿. For 輦 they give kosikuruma ‘hand-cart’.’ (Vol. 5 Shokuinryō shudenryō no jō ‘the administrative code on officials in the palace, the section of the administration division’ 巻五職員令 主殿寮条).
Other times they were provided by inserting the readings in the interlinear spaces of the text: 於毛比 於知奈牟 (2) 惟美波 行墜土須流爾 於毛比美波 Omopi mi-ba Think see-cond ‘when I think’
7 For a survey of the extant volumes of Qieyun see T. Ueda (1973). T. Ueda (1984) offers a detailed analysis of the fragment collections. Niimi and Suzuki (1968) is a useful reference for the use of primers. 8 See Tōno (1987). Tōno points out that the presence of copies of the Yupian in Japan and of the Qieyun in Dunhuang seems to be related to the late reception of the latter in Japan. Equally important is the fact that copies of the Yupian survived thanks to the sutras that were written on the verso of the scrolls, whereas the Qieyun consists of bound books copied on both sides of the paper. 9 See Shirafuji (1968). On the issue of the reception of Chinese Classics in Zenju’s commentaries see Kōno (2007).
22 Early Japanese dictionaries
477
於知奈牟土須流爾 Oti-na-mu to suru ni fall-perf-conjec comp do when ‘When I am certain that that will fall’ as seen in Zenju’s Jōyuishikiron jukki joshaku 成唯識論述記序釈 (late 8th century) (Shirafuji 1969). Buddhist glossaries appeared in the late Nara period. They include the Shin’yaku Kegonkyō ongi shiki 新訳華厳経音義私記 (Ogawa Hiromi’s collection, two volumes, compiler unknown, late 8th century), the Daiji edition of the Shin Kegonkyō ongi 新 華厳経音義 (compiler unknown, late 8th century), and the three-volume Daihannyakyō ongi 大般若経音義 (late 8th century) by Gangōji’s 元興寺 Shingyō,10 all of which contain Japanese readings and rely not only on the Yupian but also on the Yiqiejing yinyi 一切経音義 (twenty-five volumes, 661) by Xuan Ying 玄応 of the Tang dynasty.11 The Shin’yaku Kegonkyō ongi shiki in the Ogawa collection is known for drawing on such primary sources as the Huayanjing yinyi 華厳経音義 (720) by Hui Yuan 慧苑 of the Tang dynasty and the Daiji edition of the Shin Kegonkyō ongi and for revising them and integrating them with the Yupian and with Xuan Ying’s Yiqiejing yinyi. It also features a large number of Japanese readings recorded in a variety of ways.12 These include: giving the meaning in man’yōgana phonograms preceded by the character 倭, as in 下音天寂反、 the following entry for ‘flute’ 七孔也、 簫笛上照、訓布潁、 (布延 → puye ‘flute’); providing 倭云布延 the appropriate reading based on context in the interlinear spaces of the text, as in 乃知久伊矣 (乃知久伊矣与保須奈 → noti kui oyobosu na ‘leave no regret thereaf後悔無及 与保須奈 ter’); or providing a Japanized fanqie 反切 as in 憐利爾 . Some of the Zetian wenzi 則天文字, 反 the so-called “characters of the empress Wu Zetian (690–705)” of the Zhou, such as 圀 for 国, 秊 for 年, and 曌 for 照 are also included in this work.
4 Dictionaries arranged by radicals Fragments and extant copies provide ample evidence of the fact that both lay and Buddhist scholars in the Japanese secular scholarly community made extensive use of
10 The Ishiyamadera edition 石山寺本 and the Raikōin edition 来迎院本 consist only of the second book in three volumes, while the Daihannyakyō yōshūshō 大般若経要集抄 is but an excerpt. 11 For the sources of Shin’yaku Kegonkyō ongi shiki see Inoguchi (1974). For the sources of the Shin Kegonkyō ongi (Daiji edition) and Shingyō’s Daihannyakyō ongi see S. Ikeda (1980). 12 Okada (1943) is a useful reference for an historical investigation of the Japanese readings. There are numerous studies on the Shin’yaku Kegonkyō ongi shiki as a source for Chinese readings. S. Ikeda (1986) analyzes the nature of this glossary in light of its commentaries, glossaries, and dictionaries.
478
Shōju Ikeda
the Yupian. Dictionaries arranged by radical, however, were compiled exclusively by Buddhist scholars.13 After a two-year period of studying abroad, Kūkai (774–835), the founder of the Shingon school, returned from Chang’an to Japan in 806 and compiled the six-volume Tenrei banshō meigi. He edited the first four books personally, while the remaining two were edited by someone else at a later time. Tenrei banshō meigi is an abridged version of the Yupian and is meant for practical use. It preserves the structure of listing the characters under 542 radical headings, the same as the Chinese original, but it also deletes and simplifies many of the otherwise overly detailed annotations of the original. These editorial choices notwithstanding, one cannot deny its value as a reliable key to access the original Yupian, which is now lost.14 The first four books mostly include approximately 1,000 characters in the seal script – tensho 篆書 (kenshintai 懸針体). Evidence also points to the use of Shuowen jiezi 説文解字 by Xu Shen 許慎 of the Later Han dynasty as a reference for the compilation of this work (Kanda 1966, Ieno 1981, and Takata 1995). The twelve-volume Shinsen jikyō 新撰字鏡 was completed by the Nara monk Shōjū around the Shōtai era (898–901). Extant copies are categorized in two groups: one includes the complete manuscript of 1124, which is held by the Archives and Mausoleum Department of the Imperial Household Agency 宮内庁書陵部, and the other consists of the various abridged editions. The characters are organized under 160 radicals starting with 天 ‘heaven’, 日 ‘sun’, 月 ‘moon’, 肉 ‘flesh’, and 雨 ‘rain’; the dictionary wraps up with a ‘miscellaneous characters provisional section,’ or Rinji zōyōji. For the most part it is arranged by radical, although it also features elements typical of dictionaries arranged by meaning.15 The preface describes the work as a revised edition of the Yiqiejing yinyi that was eventually expanded by collating it with the Yupian, Qieyun, and Shiki 私記. Also, by adding quotations from the Shōgakuhen 小学篇 and the Bencao 本草, it grew to reach a total of twelve volumes. Its sources are not always identified, but it is possible to extrapolate them by looking at the order in which they are cited in the individual entries and by comparing the citations with similar expressions featured
13 See S. Ikeda (2005, 2011b). Mochizuki’s (1999) original approach tries to shed light on the compilation of dictionaries using the religious world as social background. 14 The photographic reproductions, index, list of characters, and collation table between the extant volumes and fragments of Yupian and the revised Daguang yihui Yupian 大広益会玉篇 by Song dynasty scholar Chen Pengnian 陳彭年 in the Kōzanji Tenseki Monjo Sōgō Chōsadan (1977) are especially useful reference tools in this respect. Miyazawa (1998) warns against the tendency to give the Yupian fragments more relevance than the Tenrei banshō meigi. 15 The 天 ‘heaven’, 親族 ‘kinship’, 木 ‘tree’, 草 ‘grass’, and 鳥 ‘bird’ sections and the Rinji zōyōji are arranged according to meaning. Sakakura (1950) states that the entries under the radicals 天 ‘heaven’, 人事 ‘human matters’, and 自然界 ‘natural world’ are organized with subheadings indicating meaning, such as 衣食住 ‘food, clothing and shelter’, 交通 ‘transportation’, 器具 ‘implements’, 植物 ‘plants’, and 動物 ‘animals’.
22 Early Japanese dictionaries
479
in parts of the text where the sources are explicitly mentioned.16 Around 3,000 Japanese renderings are recorded in man’yōgana. The twelfth century saw the compilation of the outstanding Ruiju myōgishō 類聚名 義抄. The original edition – gensenbon 原撰本 – was compiled around 1100 by a monk of the Hossō school, and was followed, in the second half of the twelfth century, by the Kōeki edition 広益本, a revised and expanded version produced by a monk of the Shingon school. The Ruiju myōgishō organizes the characters in three sections – butsu 仏 (Buddha), hō 法 (Dharma, the Law), sō 僧 (Sangha, the Priesthood) – and by 120 radicals. The only extant version of the original edition is the Zushoryō manuscript 図 書寮本 (Archives and Mausoleum Department of the Imperial Household Agency), an incomplete text of the Insei period containing only the first part of the hō section. The complete set of the Kōeki edition is the Kanchiin manuscript 観智院本 (Tenri Central Library), a famous text from the early or mid-Kamakura period.17 Ruiju myōgishō is a dictionary of great value as it compiled Japanese renderings from the vernacular readings of Chinese Classics. Comparing the word forms in Japanese renderings between the Zushoryō and Kanchiin manuscripts, the following situation should be noted: in Japanese renderings in the Zushoryō manuscript, continuative forms are used mainly, which relied the original Chinese Classics text, but in the Kanchiin manuscript, dictionary forms are used for the most part (Tsukishima 1988). Both the Shinsen jikyō and the original edition of the Ruiju myōgishō rely on Buddhist glossaries as Yiqiejing yinyi, etc; both were later integrated and expanded with the Yupian and with such rhymers as the Qieyun. The original edition of the Ruiju myōgishō explicitly quotes its sources, thus revealing a scholarly approach that placed a great deal of authority on the Chinese classic texts of the Heian period through an exhaustive use of quotations from respected sources (see G. Ikeda 1969). By contrast, the Kōeki edition omits almost all the references to its sources, indicating a completely different approach. The Ruiju myōgishō also includes a large number of Japanese renderings derived from the vernacular readings of the Chinese Classics that became codified around the mid-Heian period among lay scholars’ families (Tsukishima 1964). Another dictionary worthy of mention is the Jikyō 字鏡. The only extant version – the so-called Sesonji edition 世尊寺本 – consists of two books from the mid-Kamakura 16 Sadakari (1998) analyzes the sections that include the Yiqiejing yinyi, and quotations from the Yupian, Qieyun, and other unidentified sources. Japanese renderings are especially numerous in the sections that include the Yiqiejing yinyi and unidentified sources. T. Ueda (1981) successfully demonstrates that the Qieyun version adopted here was the revised edition by Zhangsun Neyan 長孫訥言 and that annotations from the Yupian are incorporated in the the quotations from the Qieyun. Nagai (2009) also suggests that the quotations from Yupian were used when the Yiqiejing yinyi and the Qieyun did not provide any entry for a character. 17 See Okada (1944). Yoshida (2011) is a useful reference in this respect. For an analysis through the lens of vernacular reading practices, see Tsukishima (1959). For studies from the perspective of Chinese pronunciation and tone evolution, see Komatsu (1971) and Numoto (1982). For source-based research, see Miyazawa (1977), Yamamoto (1990), and S. Ikeda (1994).
480
Shōju Ikeda
period (collection of the Tōyō bunko 東洋文庫蔵), roughly corresponding to the last volume of the three-volume original. It was compiled by using the Shinsen jikyō, the Ruiju myōgishō (both the original and the Kōeki editions), and the Mumei jisho 無名字 書 as main sources. It is generally considered to fall along the line of the Kōeki edition of the Ruiju myōgishō because it does not indicate its sources (See related works included in Tadao Yamada 1967 and Sadakari 1998). It is also known as the Iwasaki edition 岩崎 本, after the name of its previous owner. The Gyokushō 玉抄 quoted in the original edition of the Ruiju myōgishō includes Japanese renderings in katakana. The overall number of volumes and the name of its compiler are unknown, but scholars agree that the text was likely edited in the late Heian period. Additionally, the Yupianchao 玉篇抄 (also Yuchao 玉抄) quoted in the Kusharon ongi 倶舎論音義 contains notes in Chinese on the meaning of some characters and it is believed to be the same Tang-period book in thirteen volumes that is listed in the Nihonkoku genzaisho mokuroku 日本国見在書目録 (891). The influence on the Japanese dictionaries from the Chinese allograph dictionaries could be observed in this period. The Zhengming yaolu 正名要録 (second half of S.388) by Lang Zhiben 郎知本 exerted a strong influence on the Shinsen jikyō; the Ruiju myōgishō (original edition) quotes from both the Ganlu zishu 干禄字書 (710–720) by Yan Yuansun 顔元孫 and the Xinding yiqiejin leiyin 新定一切経類音 by Guo Yi 郭迻 of the Tang dynasty; while the Kōeki edition is similar to the Longkan shoujing 龍龕手鏡 by Xing Jun 行均 of the Liao dynasty.18
5 Dictionaries arranged by pronunciation Rhymers related to the Qieyun were widely used by both lay and Buddhist scholars. These dictionaries, however, were compiled exclusively by men of letters and by scholars versed in Chinese studies. The Tōkyū setsuin 東宮切韻 (in twenty or twenty-three volumes, or twelve books, depending on the records) by Sugawara no Koreyoshi (812–880) is now lost. Extant fragments suggest that it was likely based upon the fourteen versions of the Qieyun (by Lu Fayan, Cao Xian 曹憲,19 Guo Zhixuan 郭知玄, Master Shi 釈氏, Zhangsun Neyan 長孫訥 言, Han Zhishi 韓知十, Wu Xuanzhi 武玄之, Xue Xun 薛峋, Ma Gao 麻杲, Wang Renxu 王仁煦, Zhu Shangqiu 祝尚丘, Sun Mian 孫愐, Sun Zhou 孫伷, and the monk Qingche
18 S. Ikeda (1995) focuses mostly on the Xinding yiqiejin leiyin but also includes broader considerations on this topic. A work entirely dedicated to the topic is Zheng (2010), which also includes an extensive description of the extant fragments of the Xinding yiqiejin leiyin. 19 The text by Cao Xian was the Guiyuan zhucongchao 桂苑珠叢抄.
22 Early Japanese dictionaries
481
沙門清徹) and complemented with annotations from the Yupian, which are presented as ‘contemporary proposals’ (今案).20 There are no Japanese renderings. The second and third oldest rhymers after the Tōkyū setsuin are, respectively, Miyoshi no Tameyasu’s Dōmō shōin 童蒙頌韻 (1109, collection of the Sonkeikaku bunko 尊経閣文庫, manuscript of 1556; two volumes) and Shien inshū 詩苑韻集21 (late-Heian period, collection of Tenri Central Library, Kamakura period manuscript; ten volumes). The Wakan nengō jishō 和漢年号字抄 in three volumes (the mid-Kamakura period, collection of the Sonkeikaku bunko, manuscript of 1479) is arranged by meaning; it contains a large number of unabridged quotations from the Tōkyū setsuin. From the Honchō shojaku mokuroku 本朝書籍目録 (late-Kamakura period) and other such indexes we know that other rhymers of the time included the Shōsetsuin 小切韻, the Shisei shōsetsuin 四声小切韻, the Wachū setsuin 倭注切韻 by Ōe no Asatsuna (886–957), the Wachū setsuin 和註切韻 by Fujiwara no Atsumitsu (1063–1144), the Suetsuna setsuin 季 綱切韻 by Fujiwara no Suetsuna, and the Kobun setsuin 古文切韻 and the Kōin 孝韻 by Fujiwara no Takanori (1158–1233). Although the Sakumon daitai 作文大体 (mid-Heian period, collections of Tenri Central Library, Kanchiin, etc.) is a one-volume primer for composing poetry, it opens with the preface to Ōe no Asatsuna’s Wachū setsuin.
6 Dictionaries arranged by meaning: Encyclopedias The Wamyō ruijushō, a dictionary in ten or twenty volumes22 compiled in 934 by Minamoto no Shitagō, is organized by meaning. It includes the Chinese words contained in kangoshōrui (Chinese and Japanese lexicons) with Japanese renderings given in man’yōgana. It also strives to offer thorough explanations for each term. For example, the entry for azumaya ‘arbor, bower’ reads as follows: 四阿 唐令云、 宮殿皆四阿弁色立成云四 (Senchū edition 箋注本, vol. 3, p.1 recto, “dwellings n. 6” 居処部 阿安都末夜 第六, “housings n. 27” 屋宅類廿七). The Shinsen jikyō (Tenji edition, vol. 12, p. 27 recto; 阿豆 – indicates that “Rinji zōyōji n.160” ‘housings’ 舍宅章) entry for azumaya – 四阿 万屋 texts of the kangoshō type included only the reading of a word in man’yōgana, without any further explanation. The Wamyō ruijushō, however, integrates the reading with the Tang ling 唐令 definition: 宮殿皆四阿 (‘palaces are all azumaya’).
20 See T. Ueda (1956). T. Ueda (1984) contains an exhaustive collection of Qieyun fragments. For pioneering studies on the subject, see Okada (1935a, 1935c, 1935d). 21 This text is also known as Heian injishū 平安韻字集 or Injishū 韻字集. 22 There are numerous studies on the Wamyō ruijushō. Recent research should move in the direction indicated by Miyazawa (2010). For an explanation of his research method and findings, see K. Yamada (2011). For the Japanese readings, see also Ōtsuki (2004).
482
Shōju Ikeda
The Hifuryaku 秘府略 by Shigeno no Sadanushi and others (1,000 volumes) was complied by imperial order in 831. The only extant volumes are n. 864, in the collection of the Seikidō bunko 成簣堂文庫, and n. 868, in the collection of the Sonkeikaku bunko. We know, however, that Hifuryaku was heavily based on the Hualin bianlue 華林遍略 and relied on other Chinese sources such as the Xiuwendian yulan 修文殿御覧, the Yiwen leiju 藝文類聚, the Hanyuan 翰苑, and the Chuxueji 初學記.23 This indicates, once again, a scholarly approach that placed a great deal of authority on the original text. These texts were followed by the appearance of another type of dictionary that was more encyclopedic in style and had a didactical purpose. The Heian period saw the compilation of the Kuchizusami 口遊 (one volume; preface by the author dated 970, manuscript dated 1263, held by Shinpukuji 真福寺)24 the Sezoku genbun 世俗諺 文 by Minamoto no Tamenori (three volumes, of which only the first has survived; preface by the author dated 1007, manuscript of the late Heian period in the repository of Tōji’s Kanchiin 東寺観智院), and of the Yōgaku shinanshō 幼学指南抄 (thirty volumes; dispersed in the collection of Taiwan National Palace Museum 國立故宮博物 院 and Hisahara bunko 久原文庫, Kyoto University). The early Kamakura period texts include Fujiwara no Yoshitsune’s 藤原良経 Gyokkan hishō 玉函秘抄 (three volumes; manuscript of the early Nanbokuchō period, collection of the Sonkeikaku bunko); Fujiwara no Takanori’s 藤原孝範 Meibunshō 明文抄 (five volumes; Edo-period manuscript, collection of the Jingū bunko 神宮文庫); the Nihon kotohajime 日本事始 (two volumes; manuscript of 1498, collection of the Sonkeikaku bunko); and Sugawara no Tamenaga’s Bunpōshō 文鳳抄 (ten volumes; compiled after 1215, manuscript of 1299, collection of the Sonkeikaku bunko).
7 Dictionaries arranged by meaning: Iroha sequence The Iroha jiruishō 色葉字類抄 by Tachibana no Tadakane, in two or three volumes, is organized in forty-seven sections named after the syllables of the iroha poem. Entries are arranged according to the syllables and are grouped in 21 categories based on meaning, from ‘celestial phenomena’ (tenshō 天象) and ‘geographical features’ (chigi 地儀) to ‘family names’ (seishi 姓氏) and ‘names’ (myōji 名字).25 The Iroha 23 Hifu 秘府 was the Court library. All the quotations included in this dictionary were thought to have been taken directly from the library collection of the Imperial family. However, as Iida (2000) has demonstrated, there are numerous quotations taken not from the original but from similar texts. 24 See Yōgaku no Kai (1997) for an annotated edition and a detailed photographic reproduction. 25 A detailed list of references can be found in Nakada and Minegishi (1977). For the most recent developments in this field, see H. Takahashi (2000).
22 Early Japanese dictionaries
483
jiruishō is a thoroughly innovative work in that it includes the Chinese characters for Japanese and Chinese words of the time as well as their readings in katakana. Characters with the same reading are prioritized by frequency of use. The two-volume edition was compiled between the Tenyō and Chōkan eras (1144–1165), while the three-volume edition is believed to have been edited in the Jishō era (1177–1181). The early Kamakura period manuscript version of the three-volume edition held by the Sonkeikaku bunko is the best known. The Setsuyō moji 節用文字 (Insei period manuscript, collection of the Seikidō bunko), in one book, and the Sezoku jiruishō 世俗字類抄 (the original text was probably compiled during the Insei period; a two-, a three-, and a seven-volume edition exist) belong to the same family of texts; so does the revised ten-volume Iroha jiruishō 伊呂波字類抄 of the early Kamakura period. Their relationships can be summarized as follows:26 7-volume Sezoku jiruishō Urtext (sohon
8 Dictionaries arranged by meaning: Poetic words The oldest dictionary of poetic words is the Kigyoshō 綺語抄, in three volumes, by Fujiwara no Nakazane (1057–1118). The Kigyoshō provides information on the meanings and use of difficult poetic words and expressions, for example amanohara ‘the heavens’ explained as oozora o iu ‘[it] refers to the sky’, listing them under 16 different meaning-based categories: ‘celestial phenomena’ (tenshō 天象), ‘seasons’ (jisetsu 時節), ‘the earth’ (kongi 坤儀), ‘water’ (mizu 水), ‘the sea’ (umi 海), ‘deities and sages’ (shinsen 神仙), ‘human affairs’ (jinrin 人倫), ‘public offices and ranks’ (kan’i 官位), ‘human conduct’ (jinkō 人行), ‘dicta’ (genshi 言詞), ‘dwellings’ (kyosho 居処), ‘boats and vehicles’ (shusha 舟車), ‘rare treasures’ (chinpō 珍宝), ‘fabrics’ (fuhaku 布帛), ‘animals’ (dōbutsu 動物 ), and ‘plants’ (shokubutsu 植物). 26 Among studies conducted by Yamada Yoshio, Okada Yoshio, Ishino Tsuruko, Kawase Kazuma, Wakasugi Tetsuo, Minegishi Akira, and Kōno Toshihiro, but it was Ogawa (1999) who clearly defined the relationship between the different editions through a detailed collation of the texts.
484
Shōju Ikeda
Equally well-known is Fujiwara no Norikane’s (1107–1165) Waka dōmōshō 和歌 童蒙抄 (1118–1127), in ten volumes, which arranges poems with explanatory notes under the following twenty-two categories: ‘heaven’ (ten 天), ‘seasons’ (jisetsu 時節), ‘the earth’ (chi 地), ‘man’ (hito 人), ‘the human body’ (jintai 人体), ‘dwellings’ (kyosho 居 処), ‘treasures’ (hōka 宝貨), ‘literary matters’ (bun 文), ‘military affairs’ (bu 武), ‘arts and crafts’ (gigei 伎芸), ‘food and drink’ (inshoku 飲食), ‘music’ (ongaku 音楽), ‘fishing and hunting’ (gyoryō 漁猟), ‘dresses and ornaments’ (fukushoku 服飾), ‘expenditures’ (shiyō 資用), ‘Buddhas and deities’ (busshin 仏神), ‘herbs’ (kusa 草), ‘trees’ (ki 木), ‘beasts’ (kemono 獣), ‘fish and shellfish’ (gyokai 魚貝), and ‘insects’ (mushi 虫). The tenth volume gives also an outline of the structure of Japanese poetry (katai 歌体), of recurrent rhetorical errors (kabyō 歌病), and of poetry contests (utaawase 歌合). Other texts that help in the understanding of Heian-period poetic words include the Waka shogakushō 和歌初学抄 (composed before the Nin’an era, 1166–1169), in one volume, by Fujiwara no Kiyosuke and the Ōgishō 奥義抄 (1135–1144), in three volumes.
9 Herbals The earliest texts on traditional Chinese herbal medicine (honzōsho 本草書) to be introduced to Japan were the Bencao jizhu 本草集注, in three volumes, by Tao Hongjing 陶 弘景 of the Liang dynasty, and the Xinxiu bencao 新修本草 in Tang dynasty by author Sujing 蘇敬 (Mayanagi 1993). The oldest Japanese herbal is the Honzō wamyō 本草和名, two volumes, composed by Fukane no sukehito during the Engi (901–923) era. This book includes man’yōgana readings of the various herb names and explanations based on a large number of Chinese sources. This text, whose oldest extant edition is a xylograph from 1796, is also listed among the sources of the Wamyō ruijushō (Tsukishima 1965). The Ishinpō 医心方 (982, Nakarai edition 半井本 edition and Ninnaji edition 仁和寺本) is a thirty-volume medical book by Tanba no Yasuyori; it provides a brief explanation in Chinese and then gives the Japanese reading for each entry in man’yōgana. The Yasuyori honzō 康頼本草, in one volume (included in the Zoku gushoruijū 続群書類従), is attributed to the same author; it was probably composed after the Insei period. The Kōjishō 香字抄, in two or three volumes, collects the teachings of the Tanba family in the late eleventh century. Quoting from the Song text Kaibao chongding bencao 開宝重訂本草, it describes some of the plants used during esoteric Buddhist rituals. The Kōjishō 香字 抄 was subsequently expanded by the Shingon monk Ken’i (1072–1158), who revised it by collating it with the Chongguang buzhu bencao 重広補注本草 by Chen cheng 陳承 of the Song dynasty and renamed it Kōyōshō 香要抄 (two volumes). Ken’i is also the author of the Yakushushō 薬種抄, Hōyōshō 宝要抄, and Kokuruishō 穀類抄. One of his disciples, Shinkaku (1117–1180), is the author of the one-volume Kōyakushō 香薬抄 (also known
22 Early Japanese dictionaries
485
as Shuntsūshō 俊通抄, manuscript of 1165). A different book by the same title was also compiled by Kōzen (1119–1203).27
10 Buddhist glossaries Buddhist “pronunciation glossaries” that chose difficult characters or words from specific texts and explained the correct reading and meaning with annotations and comments are alternatively known as ongi (‘pronunciation and meaning’), onkun 音訓 (‘Sino-Japanese and Japanese readings’), or shakumon 釈文 (‘textual commentaries’). They tended to organize the characters in the same order in which they appeared in the original texts (kan’ongi 巻音義), although some are arranged by radical (henritsu ongi 篇立音義) and others by Sino-Japanese reading (onbiki ongi 音引音義). During the Heian period in particular, numerous glossaries were adopted and compiled in the world of religious studies.28 These texts are frequently quoted in later dictionaries such as the Shinsen jikyō and Ruiju myōgishō; many of them are of great value for the study of historical linguistics.29 The most popular reference text in the world of Buddhist studies was Xuan Ying’s Yiqiejing yinyi. There is no evidence to suggest that Hui Lin’s 慧 琳 Yiqiejing yinyi 一切経音義, compiled in one-hundred volumes in 807, was also used in Japan. The exegesis of sutras translated after the age of Xuanzang 玄奘 (602–664), for example the Daihannyakyō 大般若経, the Konkōmyōsaishōō kyō 金光明最勝王経, and the Kujakukyō 孔雀経, required the compilation of new glossaries. The Hokkekyō 法華経, for example, was originally translated by Kumārajīva (344– 413, J. Kumarajū 鳩摩羅什). However, its popularity inspired the production of countless commentaries and glossaries, all of which were later collected and reorganized in Japan. On the other hand, the study of Chinese classics at Daigakuryō 大学寮, the imperial institution of higher education, was strictly regulated. Only the commentaries by Zheng Xuan 鄭玄 and He Yan 何晏, for example, were allowed for the Analects (Lunyu 論語). The thirty-volume Jingdian shiwen 経典釈文 by Lu Deming 陸徳明 of the Tang dynasty was the most frequently used reference to look up the meaning and pronunciation of words. The Shūekishō 周易抄 by Emperor Uda (867–931) is more of a Japanese pronunciation dictionary than a Chinese one, as it contains numerous Japanese renderings in cursive man’yōgana. The Nihon shoki shiki 日本書紀私記, on the other hand, can be considered an ongi of sorts, since it uses man’yōgana to record the 27 Establishing a clear lineage of the books related to herbs and medicine is not an easy task. Kondō (1980) provides a thorough survey of the main sources in the field. 28 For an index of Buddhist glossaries, see Mizutani (1949). See also Takata (1994) for Chinese Buddhist pronunciation glossaries. 29 Yoshida (1955) first hinted at the importance of pronunciation and meaning glossaries in the compilation of dictionaries. Yoshida’s argument provided a major impetus for studies on the historical evolution of old Japanese dictionaries.
486
Shōju Ikeda
vernacular reading of the Nihon shoki as it was given during a series of public lectures in the early Heian period. Also pertaining to the world of Buddhist studies is the Shibunritsu ongi 四分律音 義, in one volume (manuscript of the early Heian period, Archives and Mausoleum Department of the Imperial Household Agency), that combines the Sanskrit words from the fourteenth volume of Xuan Ying’s Yiqiejing yinyi (written in Chinese characters) with their Japanese renderings. This source in particular shows how pronunciation glossaries were not only frequently used but were also often subject to revision and integration When he returned to Japan from China in 806 Kūkai brought with him the esoteric Buddhist canon; he then compiled the Kongōchōkyō ichiji chōrin’ō giki ongi 金剛頂経一 字頂輪王儀軌音義 (Kōzanji edition 高山寺本 and others), in one book. As the example for the entry kasa ‘canopy’ indicates – 繖左尓反 – each character includes a Japanized 訓加左 fanqie as well as a Japanese rendering in man’yōgana. Glossaries related to esoteric Buddhism also include the Daijō rishuroku haramitsukyō shakumon 大乗理趣六波羅 蜜経釈文 (early Heian period, collection of Kanda Kiichirō), in one volume, and the Kujakukyō ongi 孔雀経音義 (956, attributed to Kansei, manuscript of 1111), repository of the Daigoji 醍醐寺), in three volumes. Mere reproductions of Chinese dictionaries, these texts preserve the format of the old dictionaries of the early Heian period, even though the former quotes from the Shochūyō 書中要, integrating the main text with Japanese renderings in man’yōgana. The Kujakukyō ongi 孔雀経音義 of the mid-Heian period (repository of Daigoji 醍醐 寺, one volume) explains the pronunciation of each character (fanqie and phonemic tone indication) based on Chinese sources adding entries about the character’s meaning. The Kujakukyō ongi is well-known because it contains the oldest extant gojūon table of the fifty sounds of the Japanese language. Also worthy of mention are such glossaries as the Kujakukyō tanji 孔雀経単字 (formerly held by Kōzanji 高山寺, now in the collection of Ogawa Hiromi), in one volume, and the Kujakukyō ongi 孔雀経音義 (manuscript of the Insei period, repository of Tōshōdaiji 唐招提寺).30 The Myōhō rengekyō shakumon 妙 法蓮華経釈文 (976, repository of Daigoji and Tenri Central Library), in three volumes, by Chūzan of the Hossō school, relies on the Hokkekyō onkun 法華経音訓 (included in the twenty-seventh volume of Hui Lin’s Yiqiejing yinyi), in one volume, by Master Cien Ji 慈恩大師基 (632–682) of the Tang dynasty. The work is also revised and integrated with the Yupian and the Tōkyū setsuin but does not contain any Japanese rendering. The Daihannyakyō onkun 大般若経音訓, in four volumes, by Shingō (934–1004), a disciple of Chūzan, is now lost. A study of its many extant fragments indicates that the book gave the Sino-Japanese reading (go’on) of each character in katakana as well as an explanation of the character’s meaning in the form of a “split annotation” (warichū 割注) in Chinese. A similar approach is found in Fujiwara no Kintō’s (966–1041) Daihannyakyō
30 See Hokkaidō Daigaku Bungakubu Kokugogaku Kōza (1988).
22 Early Japanese dictionaries
487
jishō 大般若経字抄 (manuscript of 1164, repository of Ishiyamadera 石山寺), a one-volume glossary in which each character includes a side annotation on its pronunciation and an explanation of its meaning either in Chinese or in katakana. This last example in particular indicates how, by the twelfth century, the Japanization of dictionary content was proceeding at a faster pace than the early Heian period, when all the annotations were written in Chinese and borrowed from continental sources. The one-volume Konkōmyō saishōō kyō ongi 金光明最勝王経音義 (manuscript of 1079, collection of the Daitokyū kinen bunko 大東急記念文庫) is known for containing the oldest text of the iroha poem. It is a kan’ongi-type glossary that includes indications on the Sino-Japanese readings by the means of fanqie as well as Japanese renderings in man’yōgana, as 志牟反 SIMU, okasu ‘attack’, 髻計音毛 KEI, motodori ‘hair rolled up these examples attest: 侵 止々利 乎加須 in a bun, topknot’. It also distinguishes between two types of nasal endings: /-ŋ/, indicated by レ, and /-n/, indicated by >. These marks were used exclusively by the Hossō and Shingon schools, which indicate that the compiler was a monk belonging to one of these. The Kusharon ongi 倶舎論音義 (Kyoto University, Kōyasan University 高野山大 学, Kanazawa bunko 金沢文庫), in three volumes, is based on the Guangyun 広韻 and on Xuan Ying’s Yiqiejing yinyi; it gives Japanese renderings in katakana as well as invaluable indications on early Kamakura period lexical pitch accent recorded by means of tone glosses (shōten) (Nishizaki 2010). With Genshin (942–1017) the Tendai school produced the Hokkekyō gidoku 法華 経義読 (also known as Eshin sōzu gidoku 恵心僧都義読, included in the 31st volume of the Dainihon bukkyō zenshō 大日本仏教全書). A comparison between the one-volume Hokkekyō tanji 法華経単字 (kan’ongi type) copied in 1136 by Minamoto no Sanetoshi and the Kujō edition 九条本 of the one-volume Insei period Hokkekyō on (onbiki type) reveals an idiosyncratic use of fanqie among Tendai scholars, as well as the existence of glossaries that were completely different from those used by the Nara and Shingon schools.
11 Sanskrit dictionaries According to the Kōyakushō 香薬抄, one of the oldest Sanskrit dictionaries is the lateNara period Bongoshū 梵語集, in three volumes, by Shingyō, now lost (Miho 1974). In the Heian period Shinkaku 心覚 (1117–1182) compiled the Tarayōki 多羅葉記 (also known as Tarayōshō 多羅葉鈔・多羅要鈔, manuscript of 1235, repository of Daigōji), in two or three volumes. It listed Sanskrit words in the iroha order. Other similar texts include the Gashushō 鵝珠抄 (also known as Shinmoku (shō) 心目 (抄)or Shinkakushō 心覚抄, included in the Shingonshū zensho 真言宗全書), a book on esoteric practices in three volumes, and the Bongoshū 梵語集 (probably an Insei period text, attested in a
488
Shōju Ikeda
manuscript of the early Kamakura period held by Kōyasan kongō sanmaiin 高野山金剛 三昧院), in one volume, by an anonymous compiler.31
12 Conclusions Studies on old dictionaries started with Kawase (1955), who conducted his research from a bibliographic viewpoint. A further step was from a linguistic viewpoint, which was summarized in Yoshida (1971). The whole achievements were later summarized and organized by Nishizaki (1995). In recent years, foreign researchers also have started to show a great deal of interest on this topic. With the first two decades of the 21st century already behind us, the time has come for a final reorganization of the results achieved up until this point. The present paper hopes to offer a small contribution to such task.
References Bailey, Don Clifford. 1960. Early Japanese lexicography. Monumenta Nipponica 16(1/2). 1–52. Hayashi, Noriaki. 1976. Ryō no shūge shoin gyokuhen hansetsu kō [On the fanqie of the original Yupian quoted in the Commentaries on the Administrative Code]. In Osaka Rekishi Gakkai (ed.), Kodai kokka no keisei to tenkai [Formation and development of the ancient Japanese state], 667–715. Tokyo: Yoshikawa Kōbunkan. Hida, Yoshifumi, Yoshihide Endō, Masanobu Katō, Takeyoshi Satō, Kiyoto Hachiya and Tomiyoshi Maeda (eds.). 2007. Nihongogaku kenkyū jiten [The research encyclopedia of Japanese linguistics]. Tokyo: Meiji Shoin. Hokkaidō Daigaku Bungakubu Kokugogaku Kōza (ed.). 1988. Hokudai kokugogaku kōza nijusshūnen kinen ronshū jisho-ongi [Studies in dictionaries and commentaries: Festschrift for the 20th anniversary of Japanese Linguistics Depertment, Faculty of Letters, Hokkaido University]. Tokyo: Kyūko Shoin. Ieno, Junko. 1981. Tenrei banshō meigi no tensho [On the seal script of Tenrei banshō meigi] (presentation abstract). Kokugogaku 126. 72–73. Iida, Mizuho. 2000. Kodai shiseki no kenkyū 2 [Studies on the ancient history books 2]. Tokyo: Yoshikawa Kōbunkan. Ikeda, Genta. 1969. Heian jidai ni okeru honbun o ken’i to suru gakumon keitai to yūsoku kojitsu [The relationship between scholastic organisation appealing from text authority and court practice in the Heian period of Japan]. In Kodaigaku Kyōkai (ed.), Engi tenryaku jidai no kenkyū [Studies in the Engi to the Tenryaku periods], 387–412. Tokyo: Yoshikawa Kōbunkan. Ikeda, Shōju. 1980. Jōdai butten ongi to Gennō issaikyō ongi: Daijibon Shinkegonkyō ongi to Shingyō Daihannyakyō ongi no baai [On the relations between Buddhist glossaries in early Japan and Xuanying’s Yiqiejing yinyi: The case of the Daiji edition of the Shin Kegonkyō ongi and Shingyō’s Daihannyakyō ongi]. Kokugo Kokubun Kenkyū 64. 64–77.
31 See Okada (1935b, 1941) and Mabuchi (1984). Sumiya (1971) has conducted a detailed investigation on the different manuscripts of Tarayōki.
22 Early Japanese dictionaries
489
Ikeda, Shōju. 1986. Shin’yaku Kegonkyō ongi shiki no seikaku [The character of Shin’yaku Kegonkyō ongi shiki]. Kokugo Kokubun Kenkyū 75. 1–16. Ikeda, Shōju. 1994. Ruiju myōgishō no shutten kenkyū no gendankai [Notes on the researches on the sources of Ruiju myōgishō]. Shinshū Daigaku Jinbungakubu Jinbunkagaku Ronshū 28. 23–31. Ikeda, Shōju. 1995. Zushoryō-bon Ruiju myōgishō to Ruionketsu [On the relations between Zushoryō-bon Ruiju myōgishō and Leiyinjue]. Kuntengo to Kunten Shiryō 96. 26–37. Ikeda, Shōju. 2005. Yupian he Riben de guzishu [Yupian and dictionaries in early Japan]. In Baosan Zhang and Rubin Yang (eds.), Riben hanxue yanjiu xutan: Sixiang wenhua pian [Studies on Japanese Sinology 2: Thoughts and Cultures] (East Asian Civilizations 39), 295–312. Taipei: National Taiwan University Press. Ikeda, Shōju. 2008. Nihon kojisho no kenkyūhō to jissai [Practice and methodology for studying ancient dictionaries in Japan]. Hanguk Munhwa 44. 297–318. Ikeda, Shōju. 2011a. Nihon ni okeru jisho hensan no rekishi [The Japanese dictionaries quoted in the second edition of the Nihon kokugo daijiten]. Nihon Gengo Bunka 18. 23–33. Ikeda, Shōju. 2011b. Japanization in the field of classical Chinese dictionaries. Journal of the Graduate School of Letters 6. 15–25. Inoguchi, Takashi. 1974. Shin’yaku kengonkyō ongi shiki no kunko: Genponkei Gyokuhen no riyō [On the interpretation learning of Chinese characters of Shin’yaku kegonkyō ongi shiki: On the use of the original Yupian]. Bungakushi Kenkyū 15. 62–73. Inukai, Takashi. 2005. Mokkan ni yoru Nihongo shokishi [A history of written Japanese in the ancient wooden tablets]. Tokyo: Kasama Shoin. Kanda, Kiichirō. 1966. Tenrei banshō meigi kaidai [Introduction to Tenrei banshō meigi]. In Mikkyō Bunka Kenkyūjo (ed.), Kōbō daishi zenshū 6 [Works of Kobo Daishi 6]. 1–14. Kōyachō: Mikkyō Bunka Kenkyūjo. Kawase, Kazuma. 1955. Kojisho no kenkyū [A study on Japanese ancient lexicography]. Tokyo: Dainihon Yūbenkai Kōdansha. Kawase, Kazuma. 1977. Kojisho gaisetsu [An introduction to Japanese ancient lexicography]. Tokyo: Yūshōdō Shoten. Kojima, Noriyuki. 1979. Moji no yure: Asukachō Niina no shūhen [Fluctuation of writing: Niina and the Asuka period Japan]. Bungaku 47(5). 1–20. Kokugo Gakkai (ed.). 1980. Kokugogaku daijiten [A grand dictionary of Japanese linguistics]. Tokyo: Tōkyōdō Shuppan. Komatsu, Hideo. 1971. Nihon seichōshi ronkō [Studies in the history of tones in Japan]. Tokyo: Kazama Shobō. Kondō, Yasuhiro. 1980. Kōzanjizō honzō kankei shiryō ni tsuite: Kōzanji-bon Kōyakushō o chūshinni [Herbological treatises in Kōzanji: With a special reference to Kōyakushō in the Kōzanji collection]. In Kōzanji Tenseki Monjo Sōgō Chōsadan (ed.), Kōzanji tenseki monjo no kenkyū, 587–614. Tokyo: Tōkyō Daigaku Shuppankai. Kōno, Kimiko. 2007. Nara-Heianki ni okeru kanseki juyō no ichi kōsatsu: Zenju Immyō ronsho myōtōshō o tegakari toshite [A refelction on acculturation of Chinese classics in the Nara to Heian period Japan: With a special reference to Zenju’s Immyō ronsho myōtōshō]. Kokubungaku Kenkyū 151. 11–21. Kōzanji Tenseki Monjo Sōgō Chōsadan (ed.). 1977. Kōzanji kojisho shiryō 1[Kōzanji’s ancient lexicographic collection 1] (Kōzanji shiryō sōsho 6). Tokyo: Tōkyō Daigaku Shuppankai. Mabuchi, Kazuo. 1952. Gyokuhen itsubun hosei [A revised list of the original Yupian quoted fragments in other books]. Tōkyō Bunrika Daigaku Kokugo Kokubun Gakkai Kiyō 3. 1–152. Mabuchi, Kazuo. 1984 [1965]. Nihon ingakushi no kenkyū 3 [Studies in Japanese Historical Phonology 3], 2nd edn. Kyoto: Rinsen Shoten. Mayanagi, Makoto. 1993. Chūgoku honzō to Nihon no juyō [Chinese herbology and Japanese acceptance]. In Peigen Xiao and Makoto Mayanagi (eds.), Chūgoku honzō zuroku 9 [Illustrated Chinese Herbology 9], 218–229. Tokyo: Chūō Kōronsha. Miho, Tadao. 1974. Gangōji Shingyō senjutsu no ongi [Shingyō of the Gangōji Temple’s commentaries]. Kokugo to Kokubungaku 51(6). 58–73.
490
Shōju Ikeda
Miyazawa, Toshimasa. 1977. Zushoryō-bon Ruiju myōgishō to Myōhō rengekyō shakumon [The Zushoryō-bon Ruiju myōgishō and the Myōhō rengekyō shakumon]. In Matsumura Akira Kyōju Kanreki Kinenkai (ed.), Kokugogaku to kokugoshi: Matsumura Akira kyōju kanreki kinen [Japanese linguistics and the history of Japanese language: A Festschrift for the 60th birthday of Professor Matsumura Akira], 367–382. Tokyo: Meiji Shoin. Miyazawa, Toshimasa. 1998. Sūbun sōshoban Tenrei banshō meigi ni tsuite [On Sūbun sōsho’s printing of Tenrei banshō meigi]. Heisei kyū nendo Kōzanji tenseki monjo sōgō chōsadan kenkyū hōkoku ronshū. 41–55. Miyazawa, Toshimasa. 2010. Wamyō ruijushō shohon no kenkyū [A study on the texts of Wamyō ruijushō]. Tokyo: Bensei Shuppan. Mizutani, Shinjō. 1949. Butten ongi shomoku [Bibliography of commentaries for Buddhist scriptures]. Ōtani Gakuhō 28(2). 61–74. Mochizuki, Ikuko. 1999. Bukkyōkai ni jisho wa atta ka: Kojisho no shinkenkyū [Did Buddhist monks arrange “dictionaries”?: A fresh eye to the Japanese ancient lexicography]. Tokyo: Kasama Shoin. Nagai, Keiji. 2009. Shinsen jikyō no Gyokuhen in’yō ni kansuru shiron [An essay on the quotation from Yupian on Shinsen jikyō]. Nagoya Daigaku Jinbun Kagaku Kenkyū 38. 25–32. Nakada, Norio. 1983. Nihon no kojisho [Premodern dictionaries of the Japanese language]. In Norio Nakata, Toshimasa Wada and Yasuo Kitahara (eds.), Kogo Daijiten, 1808–1835. Tokyo: Shōgakukan. Nakada, Norio and Akira Minegishi. 1977. Iroha jiruishō: kenkyū narabi ni sōgō sakuin [Iroha jiruishō: Studies and Indexes]. Tokyo: Kazama Shobō. Niimi, Hiroshi and Ryūichi Suzuki (eds.). 1968. Honpō zanzon tenseki ni yoru shūitsu shiryō shūsei [Collection of Chinese lost works as found in Japanese preserved materials]. Kyoto: Kyōto Daigaku Jinbun Kagaku Kenkūjo. Nishizaki, Tōru (ed.). 1995. Nihon kojisho o manabu hito no tame ni [For the learners of Japanese ancient lexicography]. Kyoto: Sekai Shisōsha. Nishizaki, Tōru. 2010. Kusharon ongi no kenkyū [A study on Kusharon ongi]. Kyoto: Shibunkaku Shuppan. Numoto, Katsuaki. 1982. Heian Kamakura jidai ni okeru Nihon kanjion ni tsuite no kenkyū [The Sino-Japanese phonology from the Heian to Kamakura period]. Tokyo: Musashino Shoin. Ogawa, Tomoko. 1999. Setsuyō moji to jiruishō shohon no keifu [Lineage from Setsuyō moji to Jiruishō manuscripts]. Kokugo Kokubun Kenkyū 112. 23–35. Okada, Yoshio. 1935a. Tōkyū setsuin kō [Studies on Tōkyū setsuin]. Ritsumeikan Bungaku 2(5). 551–570. Okada, Yoshio. 1935b. Bongo jishoshi gaisetsu: (1) Heianchō makki made [An introduction to Sanskrit dictionaries: 1. Towards the Heian period]. Ritsumeikan Bungaku 2(8). 1041–1062. Okada, Yoshio. 1935c. Tōkyū setsuin itsubun kō [Studies on scattered texts of Tōkyū setsuin]. Ritsumeikan Bungaku 2(11). 1432–1455. Okada, Yoshio. 1935d. Wakan nengō jishō to Tōkyū setsuin itsubun [Wakan nengo jisho and scattered texts of Tōkyū setsuin]. In Tozō Koizumi (ed.), Ritsumeikan sanjūgoshūnen kinen ronbunshū, bungakuhen [Memorial for the 35th anniversary of Ritsumeikan University, the volume of humanities], 83–111. Kyoto: Ritsumeikan Shuppanbu. Okada, Yoshio. 1941. Bongo jishoshi gaisetsu: Shinkaku yori Edoki made [An introduction to the Sanscrit dictionaries: From Shinkaku to the Edo period]. In Kenjirō Takase (ed.), Ritsumeikan daigaku hōbungakubu bungakuka sōsetsu kinen ronbunshū [Memorial essays for the establishment of Ritsumeikan Daigaku Hōbungakubu Bungakuka], 161–226. Kyoto: Ritsumeikan Shuppanbu. Okada, Yoshio. 1943. Shin’yaku Kegonkyō ongi shiki wakunkō [The Japanese readings in Shin’yaku kegonkyō ongi shiki]. Kokugo Kokubun 11(3). 1–95. Okada, Yoshio. 1944. Ruiju myōgishō no kenkyū [A study on Ruiju myōgishō]. Kyoto: Ichijō Shobō. Okai, Shingo. 1933. Gyokuhen no kenkyū [A study on Yupian] (Tōyō bunko ronsō 19). Tokyo: Tōyō bunko. Okimori, Takuya, Tokihisa Kurashima, Tomoki Katō and Takenori Makino. 1996. Nihon jisho jiten [The encyclopaedia of dictionaries of Japan]. Tokyo: Ōfūsha.
22 Early Japanese dictionaries
491
Ōshima, Shōji. 1997. Jisho no hatsumei: Chūgoku gengogakushi nyūmon [An invention of dictionary: An introduction to the history of Chinese linguistics]. Tokyo: Taishūkan Shoten. Ōtsuki, Makoto. 2004. Wamyō ruijūsho no wakun: Wakun no nai kōmoku [The Japanese readings in Wamyō ruijushō: With special reference to items without Japanese readings]. Kokugo Kokubun 73(6). 1–24. Sakakura, Atsuyoshi. 1950. Jisho to bunrui: Shinsen jikyō ni tsuite [Dictionaries and their orders: With special reference to Shinsen jikyō]. Kokugo Kokubun 19(2). 46–59. Sadakari, Itoku. 1998. Shinsen jikyō no kenkyū [Studies of Shinsen jikyō]. Tokyo: Kyūko Shoin. Shirafuji, Noriyuki. 1968. Jōdai bunken ni mieru jionchū ni tsuite (1) [On the notes on Sino-Japanese readings in ancient works (1)]. Ibaraki Daigaku Jinbun Gakubu Kiyō Bungakuka Ronshū 2. 49–74. Shirafuji, Noriyuki. 1969. Jōdai gengo shiryō to shite no butten chūshakusho [The Buddhist commentaries as linguistic resources for Old Japanese]. Kokugo to Kokubungaku 46(10). 149–161. Sumiya, Yoshiyuki. 1971. Tarayōki shohonkō [Manuscripts of Tarayoki]. In Mabuchi Kazuo Hakase Taikan Kinen Kokugogaku Ronshū Kankōkai (ed.), Kokugogaku ronshū: Mabuchi Kazuo hakase taikan kinen [Studies of the Japanese linguistics: A Festschrift on the retirement of Dr Kazuo Mabuchi], 569–591. Tokyo: Taishūkan Shoten. Takata, Tokio. 1987. Gyokuhen no Tonkō-bon [On the Dunhuang version of the Yupian]. Jinbun 33. 53–64. Takata, Tokio. 1989. Gyokuhen no Tonkō-bon Hoi [Supplementary note to “the Dunhuang version of the Yupian”]. Jinbun 35. 162–172. Takata, Tokio (ed.). 1994. Chūgokugoshi no shiryō to hōhō [Materials and methods for the historical study of the Chinese language]. Kyōto: Kyoto Daigaku Jinbun Kagaku Kenkyūjo. Takata, Tokio. 1995. Kaisetsu Tenrei banshō meigi [An introduction to Tenrei banshō meigi]. In Mikkyō Bunka Kenkyūjo (ed.), Teihon Kōbō daishi zenshū 9 [Works of Kobo Daishi 9], 739–759. Kōyachō: Mikkyō Bunka Kenkyūjo. Takahashi, Tadahiko and Hisako Takahashi. 2006. Nihon no kojisho: Jobun-batsubun o yomu [Premodern dictionaries of the Japanese language: A close reading to prefaces and postscripts]. Tokyo: Taishūkan Shoten. Takahashi, Hisako. 2000. Iroha jiruishō no kachi [Research values of Iroha jiruishō]. Nihongogaku 19(11). 230–241. Tsukishima, Hiroshi. 1953. Kojisho nyūmon [The introduction of the dictionaries of Japanese old languages]. Kokugogaku 13(14). 102–109. Tsukishima, Hiroshi. 1959. Kundoku shijō no Zushoryō-bon Ruiju myōgishō [The importance of the Zushoryō-bon Ruiju myōgishō in the history of the development of vernacular readings]. Kokugogaku 37. 35–53. Tsukishima, Hiroshi. 1964. Chūko jishoshi shōkō [A study of the Heian period dictionaries]. Kokugo to Kokubungaku 41(10). 94–105. Tsukishima, Hiroshi. 1965. Honzō wamyō no wakun ni tsuite [On Japanese renderings in the Honzō wamyō]. Kokugogaku Kenkyū 5. 1–10. Tsukishima, Hiroshi. 1988. Kanbun-kundoku to kojisho no kokunten [Vernacular readings of Chinese texts and early glosses on earlier Japanese dictionaries]. Chūō Daigaku Kokubun 31(4). 1–13. Tōno, Haruyuki. 1987. Jōdai bungaku to Tonkō bunken: Dōkyō, jisho, Ōbonshi shishū o megutte [The Japanese ancient literature and Dunhuang materials: Taoist scriptures, dictionaries, and Poems of Wang Fanzhi]. Man’yōshū Kenkyū 15. 325–346. Ueda, Kazutoshi and Shinkichi Hashimoto. 1916. Kohon setsuyōshū no kenkyū [A study of earlier setsuyōshū]. Tokyo Teikoku Daigaku Bunka Daigaku Kiyō 2. Tokyo: Tōkyō Teikoku Daigaku. Ueda, Tadashi. 1956. Tōkyū setsuin ronkō [A note on Tōkyū setsuin]. Kokugogaku 24. 79–94. Ueda, Tadashi. 1970. Gyokuhen zankan ronkō [A study of the original Yupian fragments]. Kōbe Jogakuin Daigaku Ronshū 17(1). 21–37. Ueda, Tadashi. 1973. Setsuin zankan shohon hosei [Supplements to extant Qieyun fragments] (Tōyōgaku Bunken Sentā Sōkan 19). Tokyo: Tōkyō Daigaku Tōyō Bunka Kenkyūjo Fuzoku Tōyōgaku Bunken Sentā.
492
Shōju Ikeda
Ueda, Tadashi. 1981. Shinsen jikyō no setsuin bubun ni tsuite [On quotations from Qieyun in Shinsen jikyō]. Kokugogaku 127. 13–20. Ueda, Tadashi. 1984. Setsuin itsubun no kenkyū [Studies on extant texts of Qieyun]. Tokyo: Kyūko Shoin. Ueda, Tadashi. 1985. Gyokuhen itsubun ronkō [A study of extant texts of the original Yupian]. Kuntengo to Kuntenshiryō 73. 1–9. Ueda, Tadashi. 1986. Gyokuhen hansetsu sōran [A complete list of fanqie in the original Yupian]. Kobe: Ueda Tadashi. Yamamoto, Hideto. 1990. Zushoryō-bon Ruiju myōgishō ni okeru Shingō Daihannyakyō onkun no in’yōhō ni tsuite: Eizan bunkozō Sokushinshō shoin no Shingō Daihannya onkun to no hikaku yori [On citation of Shinko’s Daihannyakyo onkun in Zushoryō-bon Ruiju myōgishō: Comparison with the citation in Sokushinsho of Eizan Bunko]. Kutengo to Kunten Shiryō 85. 1–29. Yamada, Kenzō. 1995. Nara, Heian jidai no jisho [The Nara to Heian period lexicography]. In Tōru Nishizaki (ed.), Nihon kojisho o manabu hito no tameni [For the learners of Japanese ancient lexicography], 68–118. Kyoto: Sekai Shisōsha. Yamada, Kenzō. 2011. (Review) Miyazawa Toshimasa, Wamyō ruijushō shohon no kenkyū [Review: A study on the texts of Wamyō ruijushō]. Nihongo no Kenkyū 8(1). 122–129. Yamada, Tadao (ed.). 1967. Honpō jishoshi ronsō: Yamada Yoshio tsuioku [Studies in the history of Japanese dictionaries in memoriam Yamada Yoshio]. Tokyo: Sanseidō. Yamada, Toshio. 1978. Nihongo to jisho [The Japanese language and its dictionaries]. Tokyo: Chūōkōronsha. Yayoshi, Mitsunaga. 1964. Nihon no hyakka jiten [Japanese encyclopedias]. In Heibonsha (ed.), Hyakka jiten no rekishi [A history of encyclopedia], 40–70. Tokyo: Heibonsha. Yōgaku no Kai (ed.). 1997. Kuchizusami chūkai [Kuchizusami: A commentary]. Tokyo: Benseisha. Yoshioka, Masayuki. 2006. Kodai no jisho [Japanese ancient lexicography]. In Mahito Uehara, Taichirō Shiraishi and Masayuki Yoshioka (eds.), Gengo to moji [Languages and Writings] (Rettō no Kodaishi Hito, Mono, Koto 6), 239–270, Tokyo: Iwanami Shoten. Yoshida, Kanehiko. 1955. Kokugogaku ni okeru kojisho kenkyū no tachiba: Ongi to jishoshi [A standpoint of studying ancient dictionaries in Japanese linguistics]. Kokugogaku 23. 30–40. Yoshida, Kanehiko. 1971. Jisho no rekishi [A history of Japanese dictionaries]. In Atsuyoshi Sakakura (ed.), Goishi [A history of the Japanese lexicon] (Kōza Kokugoshi 3), 413–538. Tokyo: Taishūkan Shoten. Yoshida, Kanehiko. 2011. Kojisho to kokugo [The Japanese language and its earlier dictionaries]. Kyoto: Rinsen Shoten. Yoshimura, Takehiko. 2004. Kodai no bunka to shisō [On ancient cultures and thoughts]. In Rekishigaku Kenkyūkai and Nihonshi Kenkyūkai (eds.), Higashi Ajia ni okeru kokka no keisei [Formation of states in East Asia] (Nihonshi Kōza 1), 301–338. Tokyo: Tōkyō Daigaku Shuppankai. Zheng, Xian Zhang. 2010. Guoyi jingyin yanjiu [A study of Guoyi’s Jingyin]. Changsha: Hunan shifan daxue chubanshe. Zhang, Lei. 2012. Xinzhuan zijing yanjiu [Studies of Shinsen jikyō]. Beijing: China Social Science Press.
Masayuki Tsukimoto
23 Kunten texts of Buddhist provenance (butten 仏典): Their characteristics and actuality 1 Introduction The introduction of Chinese characters and of texts written in Chinese into Japan dates back to the end of the 4th century or the beginning of the 5th. In the beginning, documents in Chinese were read and understood as a foreign language. However, the Japanese soon developed a technique whereby a Chinese text could be immediately rendered in their native language by rearranging the word order of the original to reflect Japanese syntax. The birth of kundoku 訓読 is veiled in mystery, but it is generally believed to have occurred in the 6th century. The earliest extant proof of kundoku in Japan can be found in an inscribed wooden slip or mokkan dating from the second half of the 7th century. The text of the Kitaōtsu iseki shutsudo ongi mokkan 北大津遺跡出土音義木簡 (Ōtsu-shi, Shiga) deploys for the first time a man’yōgana phonogram gloss for a Chinese character (誈: 阿佐ム加ム移母 azamukamu ya mo ‘could (it) deceive (us)’), which is different from the mere plain form of a verb, showing clearly that the inflected verb is meant to be read as Japanese within an appropriate context. It was not until the end of the 8th century, however, that the results of the reading process would be inscribed within the original Chinese text and left permanently therein for the reader to see. The oldest document with this kind of notation is the fifth scroll of the Kegonkyō ryakuso kanjōki 華厳経略疏刊定記, a commentary on the Avataṃsaka sūtra preserved at the Daitōkyū kinenbunko 大東急記念文庫. In this document, marks are used for the punctuation and to indicate a native syntactic order different from that of the original text. These marks in white and vermillion are supposed to date back to 783 or 788. A further step in the evolution of these annotations in the interlinear space of the text came with the inscription of phonograms, including man’yōgana, katakana and hiragana to the side of Chinese character lemmata. An early example of this is the Ōgutsumarakyō 央掘魔羅経 (ca. 800, Shōsōin shōgozō 正倉院聖語蔵). In this document, Chinese characters whose Sino-Japanese reading is indicated by man’yōgana or katakana inserted on their right. It is noteworthy that the size of the phonograms is nearly the same as the size of the Chinese characters to which they refer. This is unusual from the standpoint of later kunten glossing practice. The size of the phonograms used for glosses in kunten materials produced during the first half of the 9th century was not standardized: for example, the Jōjitsuron 成実論 (Tattvasiddhi-śāstra) preserved in the Shōsōin repository has very small phonogram glosses compared to the size of the characters used for the main text. Such variations in size are https://doi.org/10.1515/9781614512851-024
494
Masayuki Tsukimoto
probably attributable to the fact that, at first, the readings written in by the anonymous glossator were meant to be private annotations not to be shared with the community of scholar monks for research purposes. The next stage in the evolution of glossing practice for religious documents is attested by the adoption of morphosyntactic glosses (okototen ヲコト点) and of phonographic glosses (kanaten) using katakana. Okototen consist of a complex system of dots and similar symbols that, placed around the perimeter of a Chinese character or within its area, indicated the addition of a postposition or a functional morpheme in the Japanese reading. The adoption of morphosyntactic glosses around the beginning of the 9th century is attested in such texts as the Myōhō rengekyō 妙法蓮華経 (800 ca.) preserved at Shin-Yakushiji. In other words, by the beginning of the 9th century the Japanese had developed all the structural features of kunten materials: inversion glosses (kaeriten 返 点), punctuation marks (kugiriten 句切点), phonogram glosses, and morphosyntactic glosses. With the exception of tone glosses (shōten 声点), which were introduced from China at the end of the 9th century, the existence of all other features indicates that a rapid evolution of the glossing practice took place within the religious world. We do not know for sure which field of learning served as the foundation for the practice of kundoku, but there are no doubts that the practice of inscribing the results of the reading process in the interlinear space of the text with the aid of characters and different marks started in the world of Buddhist studies. Various theories, as well as evidence from surviving documents, suggest that the glossing of religious texts in Japan was developed at Tōdaiji in Nara and centered around a group of scholarly monks devoted to the study of the Kegon or Avataṃsaka sūtra (Kegonkyō). Some Japanese scholars claim that the evolution of Japanese kunten materials can be directly linked to the Korean glossing tradition of Silla. This hypothesis cannot be dismissed, though it is my opinion that it should be examined with greater attention to detail.
2 The development of glossing of texts of Buddhist provenance Most kunten materials from the 9th century contain glosses in white. For example, the Konkōmyō saishōō kyō 金光明最勝王経1 (glossed ca. 830, Saidaiji) not only presents glosses in white, but also quotes excerpts from Chinese commentaries inserted in the interlinear space of the main text or in the margins. The explanatory glosses and the reading aids were necessary for a complete understanding of the text. KASUGA Masaji (1878–1962) highly praised this text for its value as a linguistic source (Kasuga 1942).
1 A translation of the Suvarṇa-prabhāsôttama-sūtra by Yijing (635–713), a Tang period (618–907) monk.
23 Kunten texts of Buddhist provenance (butten 仏典): Their characteristics and actuality
495
Japanese scholars tend to argue that white glosses were adopted so that the annotations could be erased by pouring water onto them. The text of the Konkōmyō saishōō kyō, for instance, contains a large number of glosses written on top of deleted notes. This supports the idea that 9th century kunten were temporary, erasable notations jotted down quickly in the interlinear space of the main text. Studies on the evolution and use of okototen have also revealed the existence of cultural exchanges between the Tendai and Shingon schools at the end of the 9th century. Such exchanges were centered on Ishiyamadera, along the line connecting Enryakuji and Ninnaji. They played a significant role in the transmission of kundoku to lay scholars by way of such key historical figures as emperor Uda (867–931). This particular topic is discussed in more detail by Kosukegawa (this volume). The 10th century brought about a revolution in the world of kunten materials. The emergence in Kyōto of new schools of Buddhism such as Shingon and Tendai boosted the production of glossed manuscripts that rivaled in number those annotated by the scholarly monks of the old Buddhist schools in the ancient capital of Nara. The Soshitsujikarakyō ryakusho 蘇悉地羯羅経略疏 (glosses ca. 950, Ishiyamadera) is an example of kunten material annotated by the Tendai school. The glosses are written in black ink and do not rely on any kind of colored pigment such as white or vermillion. This stage documents the transition from a temporary to a permanent form of notation. Moreover, the document itself is not a text borrowed from China or Korea but is a condensed commentary on the Susiddhi-kara-mahā-tantra-sādhanôpāyika-paṭala (frequently abbreviated as Soshitsujikyō 蘇悉地經 or Susiddhi-tantra) by Ennin (794–864), a Japanese monk who traveled to China to study esoteric rituals and doctrines. In other words, the Soshitsujikyō is a text written in Chinese by a Japanese monk and later annotated by an anonymous Japanese cleric to be read in Japanese. The adoption of indelible glosses in black ink is evidence to two things: first, that the glossed document was deemed worthy of transmission to future generations, and second, that there existed an indissoluble bond between the glosses and the main body of the text. The 11th century brought about new trends in glossing practice. These include, first and foremost, the coexistence within the same document of more than one system of glosses. An exemplary text in this respect is the Daibirushanakyō jōjūgiki 大毘盧遮那 経成就儀軌 preserved at Ishiyamadera. This document is annotated with kunten in four different colors. The oldest ones, dating back to 1052, are in vermillion, followed by white glosses transcribed in the second half of the eleventh century. Next are the annotations in black ink, probably datable to 1148, and last are those in green pigment, written around the 1150s. These four types of kunten represent four different systems of glosses and four distinct ways of reading the same text. The amount of time elapsed between the addition of the first set of glosses and the addition of the last one, approximately 100 years, makes it impossible for any single person to have added all four sets of annotation. This is not to say there are not single texts with kunten written by the same scholar. One such case is the Atabaku giki 阿吒薄倶儀軌 (1095, Ishiyamadera). This kunten text
496
Masayuki Tsukimoto
contains one type of vermillion gloss and two different kinds of black glosses, for a total of three different sets of glosses. The colophon at the end of the scroll mentions two other books used during the collation process (hoka ryōhon). This means that Keiga, the glossator, put down on paper two extra ways of reading the text that differed from the one he had initially supplied, for a total of three different systems of glosses. 嘉保二年五月廿二日書了 (朱書) 「以他両本引合点了 慶雅」 Copied on the 22nd day of the 5th month of the 2nd year of Kahō (1095) Collated with two other books – Keiga (in red)
This text belongs to the category of so-called kanatenbon, texts that do not rely on okototen but solely on kana to generate the kundoku reading. As the case of Ōgutsumarakyō illustrates, kunten texts without morphosyntactic glosses existed since the 9th century, but after the emergence of okototen in the 9th century, the vast majority of kunten texts began to rely on this practical technique of notation. And yet the number of texts annotated with okototen began to decline from the 12th century on, and by the 13th only a few documents were still using morphosyntactic glosses. I believe that there are two main reasons for the rapid decline in the use of okototen in the 12th and 13th centuries: First, okototen were invented in the 9th century in order to note down kundoku very quickly by means of single-stroke annotations (mostly dots and lines) when listening to lectures. However, by the 12th–13th centuries, instruction had to a large extent taken the place of lectures, allowing for a more relaxed pace of annotation, and so two- or three-stroke katakana came to be used instead. Second, there are so many different systems of okototen that it is unlikely that anyone would have been able to remember all of them. In contrast, the shapes of katakana had been unified in the 11th century and could be read and written by everyone in the world of learning (scholars of the nobility and Buddhist clergy), and so they were easier to use. In the 12th century, the majority of kunten materials of Buddhistic origin indicate multiple readings with parallel systems of glosses. These are typically distinguished by color, usually vermillion and black. White glosses become rarer. If the document makes use of morphosyntactic glosses, these are usually written in vermillion (with the rare exception of a small group of Tendai scholars who inscribed okototen in black). This gives the following two-way classification: (A) Texts with both okototen and kana written in vermillion (B) Texts with okototen in vermillion and kana in black ink Thus in the case of (B), the readings indicated by the vermillion and black glosses supplement one other. Figure 1 illustrates the Dainichikyōsho 大日経疏 in the collection of the University of Tokyo, a representative example of 12th century kunten materials. This text is an
23 Kunten texts of Buddhist provenance (butten 仏典): Their characteristics and actuality
497
instance of (A), with both okototen and kana written in vermillion. However, we find some kana glosses which indicate readings distinct from those of the vermillion glosses, showing that this manuscript indeed employed two systems of glossing.
Figure 1: Dainichikyōsho, Vol 4, The University of Tokyo.
3 Studies on kunten materials of Buddhist origins and the results thereof There is no questioning the significance of kunten materials of Buddhist origin for both synchronic and diachronic studies of the Japanese language. Modern studies in this field began in 1909 with the publication of Kanazukai oyobi kanajitai enkaku shiryō 仮名遣及仮名字体沿革史料 by ŌYA Tōru (1850–1928). An auxiliary member of the Kokugo chōsa iinkai, the Japanese Language Survey Committee, Ōya was the first to carry out a detailed examination of a large number of manuscripts, most of them held at the Shōsōin in Nara. His goal was to collect as many examples as
498
Masayuki Tsukimoto
possible of kana usage to aid in the project of standardization of the two syllabic writing systems, hiragana and katakana. His work resulted in the foundation of a completely new field of research and played an unanticipated role in unveiling the historical evolution of katakana. Another key moment in the development of kunten studies was the publication of KASUGA Masaji’s work on the Konkōmyō saishōō kyō (Kasuga 1942). It brought to light hitherto unknown aspects of 9th century Japanese, such as the fact that the adjective kibi-si, used with the meaning of ‘high in density’, belonged to the -ku conjugation (-ku katsuyō) and not, as previously believed, the -shiku conjugation (-shiku katsuyō). Other important findings included the differences in vocabulary between kunten and contemporary wabun texts as well as the identification of certain fixed patterns of grammatical co-occurrence related to the character 況, read as イハムヤ ipamuya (> iwamuya). Basing his research upon 9th century glossed documents, Kasuga demonstrated that ipamuya did not co-occur with the final postposition woya at the end of the same clause, as had been previously thought. Instead, ipamuya governed the choice of postposition in the preceding clause. Dative (ni), comitative (to), and accusative (wo) were always followed by the postposition pa (> wa) when ipamuya appeared in the following clause. The sole exception was the postposition pa itself, which resulted in the form paya and not the expected papa. HASHIMOTO Shinkichi (1882–1945) wrote only a few papers dealing specifically with the use of kunten materials as linguistic sources. Nonetheless, thanks to a gloss inserted next to the Chinese compound 偉大 on a copy of the Da Tang xiyu ji 大唐西域 記 (1163, Ishiyamadera) he was able to establish the meaning of the adjective toposirosi (‘great in extension and magnificent’) in the 8th century Man’yōshū (Hashimoto 1925). One of the most influential bodies of work in this field is TSUKISHIMA Hiroshi’s (1925–2011) research on the vocabulary of kunten materials. In Heian jidai no kanbun kundokugo ni tsukite no kenkyū, published in 1963, Tsukishima compares the lexicons of 11th and 12th century kunten materials of Buddhist origin with that of Genji monogatari. The comparison allowed him to identify two groups of words carrying the same meaning but pertaining exclusively to either a Sinicized or native written language form. For example, in lieu of pisoka-ni (> fisoka-ni > hisoka-ni), a word meaning ‘secretly’ and characteristic of kunten materials, a pure wabun text tends to use misoka-ni, sinobi-ni, or sinobiyaka-ni, words sharing the same semantic value as the former, but more contextualized within the native form of the written language. To (tagapi-ni >) tagawi-ni (> tagai-ni), a word often found in glossed materials with the meaning of ‘mutually’, Genji monogatari prefers katami-ni for the same meaning. These findings made it possible to measure the degree to which a given piece of writing conforms to kanbun-kundoku or wabun style. This measure in turn plays a key role in the study of so-called Sino-Japanese hybrid writing (wakan konkōbun). As we have seen, Japanese language studies, synchronic and diachronic alike, have achieved important results by using kunten materials of Buddhist origin as sources. However, in the following pages I will discuss aspects of these materials that previous
23 Kunten texts of Buddhist provenance (butten 仏典): Their characteristics and actuality
499
research has either overlooked or underestimated the importance of. Specifically, I will focus on issues related to the characteristics of kunten materials and their production. As we saw in Section 2, during the 9th century kunten were intended as temporary and personal notations which were easy to erase. Scholars have long postulated that early kunten materials were produced by disciples during lectures given by their master in the main Buddhist temple hall. Listening to the master’s discourse as he translated the original Chinese text into Japanese, the disciples, who had to commit the contents of the scriptures to memory, recorded along the side of the main text the proper way of interpreting it. Not coincidentally, the colophon of many a manuscript ends with the expression 聴了 “finished listening”; kana and okototen were developed as quick and compact forms of notation which enabled the disciples to keep up with the teacher’s exposition. However, one would be hard put to argue that all 9th century documents were produced under such circumstances. The aforementioned Konkōmyō saishōō kyō, for example, quotes excerpts from Chinese commentaries inserted in the interlinear space of the main text or written on the margins. Such detailed explanations could hardly have been jotted down in haste while listening to a lecture. The kana and the okototen of this document are also so carefully added that it is impossible to believe they were put down on paper on the spur of the moment. I contend we should group kunten materials of Buddhist origin produced in Japan during the 9th and 10th centuries in the following categories. α. Materials related to lectures A. Records of lectures (writings compiled by disciples while listening to the master’s explanations) B. Revised records of lectures (clean copies of A texts) C. Materials used in preparation for the lecture (Documents written by the master before the lecture) β. Materials unrelated to lectures D. Individual research results (texts not directly related to the master-disciple transmission) E. Texts for personal study (glosses transferred from C and D texts) In other words, not all kunten materials produced during the 9th and 10th centuries were mere transcripts of public lectures; in fact, they can be divided into specific categories depending on the circumstances of their production. This new taxonomy is a matrix that allows us to better understand the linguistic characteristics of each document. Existing studies tend to emphasize the borderline colloquial flavor of 9th century materials, pointing at the temporary nature of the glosses as evidence. This conclusion, however, can be seen as oversimplistic in light of the above-mentioned classification. Having said this, is also true that not all documents find a proper placement within this scheme. For example, the Konkōmyō saishōō kyō does not belong to group A; more research is needed to establish its assignment to one of the categories from B to E.
500
Masayuki Tsukimoto
Based on these premises, how can we classify kunten materials of religious origin produced from the 11th to the 12th centuries? Such texts show a marked decline in the number of documents from groups A to C and make it necessary to introduce a new category: γ. Materials related to the transmission of knowledge F. Texts previously copied for didactical purposes During the 11th and 12th centuries the transmission of knowledge from master to disciple played a pivotal role within the Tendai and Shingon schools. In the world of esoteric Buddhist studies, the master would first lend his own glossed text to the disciple and ask him to copy it in its entirety. Only when the disciple had transferred the glosses to his personal text was he ready to receive the master’s private teachings. The process of transferring glosses required a certain amount of time and differed sharply from the public form of learning exemplified by texts of group A during the 9th and 10th centuries. Such a radical change in the transmission of knowledge among religious scholars seems to be related to the emergence of esoteric Buddhism in the 11th and 12th centuries and to the importance placed on the direct transmission from master to disciple within these schools. Kunten materials produced during the Heian period, from the 9th to the 12th centuries, include a large variety of texts; therefore one should tread lightly in using such sources as evidence for the historical evolution of the Japanese language. The language of texts from groups A to D is, to a certain degree, a mirror of the language used at the time in which the glosses were composed. On the other hand, the language of texts from groups E and F should be analyzed with particular care. As we have seen, both texts for personal study and texts previously copied for didactic purposes relied mainly on transcriptions from older manuscripts. Therefore, the linguistic traits of these glosses could reflect changes occurring over a long span of time, ranging from some years to some decades. The use of the word iten 移点 to indicate the process of transferring glosses from a text to another dates back to at least the first half of the 11th century, when it was used in the colophon of the Dainichikyō 大日経 (1040), preserved now at the Gotō Museum. Much as they believed that all 9th century kunten materials were the result of transcriptions made during public lectures, up until twenty years ago scholars were also convinced that 11th and 12th century glossed manuscripts were exact copies of older texts. In 2007, however, MATSUMOTO Mitsutaka published Heian Kamakura jidai kanbun kundokugo shiryōron, a work in which he not only demonstrated that 12th century glosses were not in fact faithful re-transcriptions of ancient manuscripts but also showed how, in the same time period, a new way of reading was born. Building on this approach, it is possible to envision four different scenarios for the glossing of religious texts during the 11th and 12th centuries in Japan: 1. Faithful transcription of the original glosses 2. Transcription with some changes to the original glosses
23 Kunten texts of Buddhist provenance (butten 仏典): Their characteristics and actuality
3. 4.
501
Collation with various documents and transcription of the different glosses in a single text Creation of a totally new way of reading different from the existing tradition
The reason behind the creation of a new way of interpreting a document or behind the rejection of existing tradition probably lies in the dissatisfaction felt by some scholarly monks who were either uncomfortable with adhering to past models or aspired toward a more accurate exegesis of the Buddhist scriptures. Unlike 9th century monks, during the 11th and 12th centuries Buddhist religious scholars could no longer travel to China to resolve their doctrinal doubts. The only choice left to them was to collect as large a number of kunten materials related to a specific text as possible and carry out comparative research. In the meantime, a handful of gifted scholars also attempted to offer their own original interpretation. If we apply these taxonomies to the texts I discussed earlier, then the Soshitsujikarakyō ryakusho is likely a document pertaining to groups D or E, the Daibirushanakyō jōjūgiki is a group F text which was used on four different occasions and, according to the 11th and 12th century classification, a type 3 text. The Atabaku giki is an E group and a type 3 text while the Dainichikyō (Figure 1) is an F group document which was used on two different occasions. Knowing the characteristics of each kunten document and the circumstances behind its production are basic prerequisites if we want to use them as sources for linguistic research.
4 Conclusion This chapter has attempted to provide an overview of the differences in kunten texts of Buddhist provenance produced between the 9th and 12th centuries and of the circumstances under which they were created. I have made no mention of 13th century texts for two reasons: first, 13th century glossed materials are very similar to 12th century documents and second, not much research exists on 13th century kunten texts of Buddhist provenance. As a matter of fact, data concerning the texts produced between the 9th and the th 12 centuries is also fragmentary and relatively limited in breadth when compared to the large number of available texts. According to the nationwide survey carried out by TSUKISHIMA Hiroshi between 1945 and 1995 and presented in his Heian jidai kuntenbon ronkō – kenkyūhen (Tsukishima 1996), there exist 3,355 manuscripts dating from the 9th to the 12th centuries. The religious value of these texts is often a hindrance to their investigation; in the case of texts of secular Chinese provenance, their limited number is also a challenge. And then, of course, there is the sheer issue of numbers: even if it was possible to examine
502
Masayuki Tsukimoto
one document each week, it still would take about 65 years to look at all 3,355 texts. Such an enterprise would require a lifelong dedication to the research. And not even a scholar of Tsukishima’s stature could give a precise estimate for the number of surviving glossed manuscript in Japan (my own estimate of the number of extant kunten materials of religious provenance produced between the 9th and the 12th centuries is around 5,000 to 10,000). Almost all kunten sources are preserved in temples, museums, or by private organizations. Obtaining authorization to see a manuscript often takes a long time and more often than not permission to take photographs is not granted due to religious reasons. As a consequence, the only option left to the scholar is to follow the example of the monks who lived 1000 years ago and transcribe the glosses manually onto a printed copy of the text that was prepared in advance. The last work by Tsukishima, Kunten goi shūsei (2007–2009), in 9 volumes, collects over 200,000 examples of lexica from kunten materials he transcribed by hand from all the manuscripts he investigated during his lifetime. Considering such difficulties, it is not surprising that the number of researchers devoted to this field is actually in decline. There are only very few Japanese specialists in this field under the age of 60. Consequently, we have great expectations when it comes to the work of foreign scholars. In recent years, the interest of Korean scholars in Japanese kunten material has increased exponentially, leading to the creation of reading groups. Moreover, scholars from the U.S. and Europe have begun to show interest in this kind of material. It is thanks to their efforts that the contributions on kunten research in this volume have been included.
References Hashimoto, Shinkichi. 1925. ‘Tohosirosi’ kō [On tohoshiroshi]. Nara Bunka vol. 6: Man’yō kenkyūgō. 4–8. Nara: Kōgyokudō Shoten. Kasuga, Masaji. 1942. Saidaiji-bon Konkōmyō saishōō kyō koten no kokugogakuteki kenkyū [Linguistic research on the old glosses in the Saidaji manuscript of the Konkōmyō saishōō kyō]. Tokyo: Shidō Bunko. Matsumoto, Mitsutaka. 2007. Heian Kamakura jidai kanbun kundokugo shiryōron [On the materials on the language of kanbun-kundoku in the Heian and Kamakura periods]. Tokyo: Kyūko Shoin. Ōya, Tōru. 1909. Kanazukai oyobi kanajitai enkaku shiryō [Materials on changes in the use and shape of kana letters]. Tokyo: Bensei Shuppan. Tsukishima, Hiroshi. 1963. Heian jidai no kanbun-kundoku-go ni tsukite no kenkyū [Research on the language of kanbun-kundoku in the Heian period]. Tokyo: University of Tokyo Press. Tsukishima, Hiroshi. 1996. Heian jidai kuntenbon ronkō: Kenkyūhen [A study of Heian period kunten texts: Research volume]. Tokyo: Kyūko Shoin. Tsukishima, Hiroshi. 2007–2009. Kunten goi shūsei [Compendium of kunten vocabulary], vols. 1–9. Tokyo: Kyūko Shoin.
Teiji Kosukegawa
24 Kunten texts of secular Chinese origin (kanseki 漢籍) 1 Introduction This chapter offers an overview of kunten texts of secular Chinese origin, known in Japanese as kanseki 漢籍. In kunten research, the system of glossing is distinguished from kunten in Buddhist texts, or butten 佛典. I will focus on the following points: 1. Kunten glossing of Chinese classics was introduced one century after the introduction of glosses in Buddhist texts. 2. The kunten glosses in texts of secular Chinese origin mirror the contents of classic commentaries. 3. Kunten glosses on Chinese classics in Japan share some characteristics with glossed manuscripts found in Dunhuang.
2 The practice of glossing in the secular and religious worlds Based on extant kunten glossed manuscripts of secular Chinese origin (kanseki 漢籍), we know that the practice of annotating kanseki texts with morphosyntactic glosses (okototen ヲコト點) and simplified phonogram glosses (shōkaku kana 省畫假名) was inherited from butten glossing practice, specifically from the Tendai school during the first half of the 10th century.1 Scholars have various conflicting theories on the status of kunten glossing of secular Chinese texts in Japan before the 10th century. However, judging from extant materials, kunten glosses on secular Chinese texts appear over a century later than kunten glosses on Buddhist texts. This fact is also attested to by the large disparity in the quantity of materials. Only some 200 kanseki texts with kunten glosses are attested from the Heian and Kamakura periods (Kokugo Gakkai 1980: 251), compared to a grand total of more than 5,000 glossed documents from the Heian period (794–1185) alone (Tsukishima 1996: 26). These statistics show the vast preponderance of
1 According to the classification of the different systems of morphosyntactic (okototen) glosses proposed by NAKADA Norio in 1954, later revised by TSUKISHIMA Hiroshi in 1986, the okototen used in Chinese classical texts are included in group 5 – daigo gunten 第五群點. Group 5 okototen are characterized by a clockwise placement of the symbols that mark the postpostions te-ni-wo-ha in the quadrilateral space around a Chinese character and starting from the lower left corner. See Nakada (1954: 439), Tsukishima (1969: 98), and Yoshida et al. (eds.) (2001: 93). https://doi.org/10.1515/9781614512851-025
504
Teiji Kosukegawa
Buddhist texts in the corpus of kunten materials, and underscore the fact Buddhist texts provide the overwhelming bulk of kunten materials available for linguistic research.
3 Kanseki kunten glosses match the contents of classical Chinese commentaries In the Nara period and medieval Japan, the Chinese classics were handled and studied at the Daigakuryō 大學寮, the imperial institution of higher education created under the Ritsuryō system (Momo 1947). Thus, everything from the type of text handled to the selection of commentaries, from pedagogical methods to examination procedures, was specified in detail and strictly regulated. Students were required to study the version of a text as read according to the interpretation of a specific commentary. The main text and its commentary were thus treated as a single set, not as two separate units. Students who were conferred the status of reader (dokusha 讀者) were then admitted to the next level, lecturer, or kōsha 講者. Extant kanseki kunten materials reflect the didactic system of the Daigakuryō. While exceptions do exist,2 as a general rule the entire content of the commentary was transcribed in the form of meticulous annotations inserted within the main body of the text – i.e. warichū 割注 (lit. “splitting notes”). This technique for glossing a text differs greatly from that of Buddhist clerics, who saw a sacred text and its commentaries as entirely separate genres. The following examples illustrate the relation between individual glosses and their sources. The Ueno text of the Hanshu yangxiong zhuan 漢書楊雄傳 (Ueno-bon Kanjo yōyū den 上野本漢書楊雄傳) is a kanseki kunten text in which a Tang period (618–907) manuscript has been interpolated with glosses dating back to the second year of Tenryaku (948). The annotations within the main text come from the Hanshu jizhu 漢書 集注 (Kanjo shūchū), a commentary by the renowned Tang scholar Yan Shigu 顏師古 (581–645), while the interlinear notes (gyōkanchū 行間注) and the marginalia (rangaichū 欄外注) quote two other commentaries, the Hanshu xunzuan 漢書訓纂 (Kanjo kunsan) and the Hanshu gujin yiji 漢書古今集義 (Kanjo kokin shūgi). The crucial point here is that the kunten glosses directly quote the Chinese commentaries. For example, the character 捷 (Chinese jie, ModJ jō, shō ‘win, quick’) in line 29 has on its right a vermillion kana gloss reading oyoba(mu) ya オヨバ (ム) ヤ ‘Will it extend to?’ The gloss in black to the left of the same character gives the reading majipara(mu ya) マジハラ (ムヤ)‘Will it mix with?’ The first gloss is the Japanese ren-
2 Numerous texts written during the Six Dynasties (222–589) or the Tang period (618–907) such as the Wenxuan 文選 (Monzen) or the Youxianku 遊仙窟 (Yūsenkutsu) were not associated with a specific commentary.
24 Kunten texts of secular Chinese origin (kanseki 漢籍)
505
dition of the comment jie ji ye 捷及也 ‘Jie is “extend to”’ in the main body of the text in Hanshu jizhu, while the second gloss matches the interlinear note jie jie ye 捷接也 ‘Jie is “connect”’.3 Ueno-bon Kanjo yōyū den 上野本漢書楊雄傳, line 29–30 29: . . . 豈駕鵞之能捷《應劭曰蓬陼蓬萊之陼在海中晉灼曰 30: 捷及也師古曰鴐鵞鳥名也解在司馬相如傳鴐音加》. . . 《 》 ( : inserted notes) interlinear note on left side of “捷”: 訓如曰捷接也晉云々 In the same Ueno-bon Hanshu yangxiong zhuan, the reading for the character chao 超 (ModJ chō ‘exceed, surpass’) indicated by the black gloss in line 5 is topo-u site トホウ シテ ‘being far’. Once again this follows the commentary by Yan Shigu, who specifies that chao yuan ye 超遠也 ‘Chao is “far”’ in the Hanshu jizhu. Interestingly, it should be noted that MOROHASHI Tetsuji’s Daikanwa jiten 大漢和辭典 (1955–1960) as well as almost all modern Japanese-Chinese character dictionaries cite this example as the only evidence supporting the reading sumiyaka ‘fast, prompt’ for the character chao 超. The reading sumiyaka, however, is based on an error; as KANDA Kiichirō (1935; 各本遠作 速) points out, later editions of the Hanshu yangxiong zhuan confuse the character yuan 遠 ‘far’ used in the note from the Hanshu jizhu with the character su 速 ‘fast’, generating an erroneous reading that has persisted to the present day. Only printed copies of the Hanshu from more recent times include the incorrect phrase chao su ye 超速也 (Ishizuka 1992b: 4). This case illustrates how kunten glossed kanseki texts have a high critical value for modern lexicography. Ueno-bon Kanjo yōyū den 上野本漢書楊雄傳, line 5–6 5: . . . 淑周楚之豐烈兮超既離 6: 虖皇波《應劭曰 . . . 超遠也 . . .》 The Iwasaki-bon Guwen shangshu 古文尚書 (Iwasaki-bon Kobunshōsho 岩崎本古文尚 書) is a Tang period manuscript with kunten glosses from the 10th century. The annotations within the main text are based on commentary by Han Dynasty (206 BCE–220 CE) Confucian scholar Kong Anguo, but we find evidence throughout the text that the anonymous glossator also referred to the Tang period commentary Shangshu zhengyi 尚書正義 (Shōsho seigi). For example, in line 3 of the warichū, the character ju 距 is accompanied by a kana giving the reading itareri イタレリ ‘reached’ for this character. There is no internal explication of the character ju 距 in the entire text. However when we consult the Shangshu zhengyi, which fully re-annotated the Guwen shangshu in the Tang period, we find an explanatory note ju shi ye 距至也 ‘Ju is “reach”’ added to Kong Anguo’s commentary. This matches itareri, the 10th century Japanese gloss, and confirms that the Japanese glossator consulted this specific source (Kosukegawa 2005: 86).
3 The first scholar who treated the issue of the relation between the glosses of Hanshu yangxiong zhuan and its commentaries was MATSUMOTO Mitsutaka. See Matsumoto (1982: 28–40).
506
Teiji Kosukegawa
Iwasaki-bon Kobunshōsho 岩崎本古文尚書, line 2–3 2: . . .濟河惟沇州《東南據濟西北 3: 距河也》. . . Shangshu Zhengyi 尚書正義, vol.6 兖州○傳東南至距河○正義曰 . . . 據謂跨之距至也 These examples demonstrate how the kunten glosses on kanseki texts relied on a corpus of specific commentaries on the Chinese classics; the glosses were not, in other words, based on the glossator’s own interpretation or dictionaries and other references of his own choosing. This reliance on specific commentaries is the most salient characteristic of kanseki kunten materials.
4 The glossing of kanseki materials in Japan resembles the glossing of Dunhuang manuscripts As noted at the outset of this paper, extant kanseki kunten materials are the product of the transmission of a glossing tradition from Tendai Buddhism to lay scholars around the 10th century. However, the study of Chinese classics remained confined to the Daigakuryō, a public organization with a strict set of rules based on an ancient administrative code. According to these regulations, texts could only be interpreted on the basis of specific commentaries imported from the continent. Since both the commentary-based approach and the institutional model of the Daigakuryō were borrowed from China, one would naturally assume there existed similarities between Japanese kunten texts and the Chinese practice of glossing as seen in the Dunhuang collection.4 In 10th century Japan Chinese classics were annotated in remarkably different ways, especially insofar as the number of phonogram glosses (kanaten 假名點) is concerned. For example, the Ueno edition of the Hanshu yangxiong zhuan, which, as per its colophon, was transcribed in 948, displays numerous phonogram glosses, while the Shishuo xinshu 世説新書 (Sesetsushinsho), which lacks a clear colophon, has but a few. Even fewer glosses appear in the second volume of the Chunqiu jingchuan jijie 春秋經傳 集解 (Shunjū keiden shikkai), whose morphosyntactic glosses belong to group 1 (postpositions following the order te-wo-ni-ha) and not 5 (Kosukegawa 2008: 15–30). It is still unclear whether the presence of a large number of phonogram glosses is related to the time period in which the text was annotated. This is an issue for future research. In any case, comparisons between older and more recent editions of the same work seem to point in that direction. 4 About the tradition of glossing texts in ancient China see Ishizuka (1970: 2–38), Ishizuka (1995: 39–65), Ishizuka (1992a: 229–261), and Ishizuka (1993: 30–50).
24 Kunten texts of secular Chinese origin (kanseki 漢籍)
507
Moreover, 10th century Japanese kunten texts of secular origin share a number of features with their Dunhuang counterparts. These include paragraph marks (kadan 科 段), punctuation (kutōten 句讀點) and marks to indicate the use of a character according to its secondary or tertiary meaning (poyin 破音). I should also point out that a comparison between any given Japanese annotated text and its Dunhuang counterpart reveals the adoption of poyin marks around the same characters. For example, the character cheng 稱 (1) at the beginning of the Mushi 牧誓 section and the character zhang 長 (2) in the Pangeng 盤庚 chapter of the aforementioned Guwen shangshu are annotated in the same way in both the Japanese and Chinese versions. (1) Kanda-bon Kobunshōsho 神田本古文尚書牧誓篇, line 82–83 82: . . .虎 83: 賁三百人《勇士稱也. 若虎賁獸言其猛也. 皆. 百夫長》. . . 《 ( 》 : interlinear notes, □: poyin, □.□: punctuation) Dunhuang MS. Stein 799, line 37–38 37: . . .虎賁 38: 三百人《勇士稱. 若虎賁獸. 言其猛. 皆百夫長 已上》. . . (2) Iwasaki-bon Kobunshōsho 岩崎本古文尚書盤庚篇, line 135–137 135: . . .我乃劓殄滅之亡遺 136: 育. 亡卑 易 種于茲新邑《劓割. 育長也. 不吉之人當割絶滅之无遺長其類. 无使易 種於此 137: 新邑也》. . . Dunhuang MS. Pelliot Chinois 2516, line 26–27 26: 我乃劓殄滅之. 亡遺育. 亡卑易種于茲 27: 新邑《劓割. 育長也. 不吉之人. 當割絶滅之. 無遺長其類. 無使易種於此新 邑》. . . Dunhuang MS. Pelliot Chinois 2643, line 87–89 87: . . .我乃劓殄 88: 滅之. 亡遺育. 亡 卑易種于茲新邑《劓割. 89: 育長也不吉之人. 當割絶滅之. 無遺長其類. 無使易種於此新邑也》. . . These glosses were based on the entries cheng 稱:chizheng fan 尺證反(去聲)and zhang 長: dingzhang fan 丁丈反(上聲)as provided in the exegetical dictionary Jingdian Shiwen 經典釋文 (Keiten shakumon) by Tang scholar Lu Deming 陸徳明 (556–627). This demonstrates that both Chinese and Japanese scholars relied on the same sources to gloss Chinese classics. The existence of similarly annotated versions of the same document produced at different times and in different places suggests that a full understanding of the glossing tradition of secular texts in Japan cannot come exclusively from the study of the transfer of knowledge from the religious to the lay world at the localized level. As the
508
Teiji Kosukegawa
Dunhuang documents show, an external transmission route from China to Japan also existed. This is a point that, I believe, future research ought to consider.
References Ishizuka, Harumichi. 1970. Rōran Tonkō no katenbon [The diacritic manuscripts of Loulan and Dunhuang]. Bokubi 201. 2–38. Ishizuka, Harumichi. 1992a. Tonkō no katenbon [The diacritic manuscripts of Dunhuang]. In On Ikeda (ed.), Tonkō kanbun bunken (Kōza Tonkō 5) [Chinese documents from Dunhuang (Lecture series on Dunhuang 5)], 231–261. Tokyo: Daitō Shuppansha. Ishizuka, Harumichi. 1992b. Ueno-bon Kanjo yōyū den no kunchū to shuten [On the functional relationship between annotations and vermillion glosses of the Ueno text of The biography of Yang Xiong in the Book of Han]. Kuntengo to Kuntenshiryō 88. 1–5. Ishizuka, Harumichi. 1993. The Origins of the Ssŭ–shêng Marks. Acta Asiatica 65. 30–50. Ishizuka, Harumichi. 1995. Shōten no kigen [The origins of the Ssŭ–shêng marks]. In Hiroshi Tsukishima (ed.), Nihon kanjionshi ronshū [Collected papers on the history of Sino-Japanese kanji readings], 39–65. Tokyo: Kyūko Shoin. Kokugo Gakkai (eds.). 1980. Kokugogaku daijiten [Dictionary of Japanese linguistics]. Tokyo: Tōkyōdō Shuppan. Kosukegawa, Teiji. 2005. Shōsho seigi to no kankei kara mita kobun shōsho Heian chūkiten no mondai [The issue of the mid-Heian period glosses on The book of history in relation to the annotations by Kong-Yingda]. In Ishizuka Harumichi kyōju taishoku kinenkai (eds.), Nihongaku, tonkōgaku, kanbunkundoku no shin tenkai [New developments in Japanology, Dunhuangology and vernacular reading of Chinese texts], 79–94. Tokyo: Kyūko Shoin. Kosukegawa, Teiji. 2008. Kunten shiryō no tenkaishi ni okeru Yūrinkanzō Shunjū keiden shikkai kan daini no ichi [The historical significance of the Fujii Yūrinkan Museum text of volume 2 of the collected annotations on The book of spring and autumn annals for the development of kunten materials]. Nihongo no Kenyū 4(1). 15–30. Matsumoto, Mitsutaka. 1982. Kanjo yōyū den tenryaku ninenten ni okeru kundoku no hōhō [The method of vernacular reading in the Tenryaku 2 (948) glosses on The biography of Yang Xiong in the Book of Han]. Kokugogaku 128. 28–40. Momo, Hiroyuki. 1947. Jōdai gakusei no kenkyū [Research on the school system in the Nara period]. Tokyo: Meguro Shoten. Nakada, Norio. 1954. Kotenbon no kokugogakuteki kenkyū: Sōron hen [Linguistic studies on early ChineseJapanese glosses: Overview]. Tokyo: Kōdansha. Tsukishima, Hiroshi. 1969. Heian jidaigo shinron [A new account of Heian period Japanese]. Tokyo: Tōkyō Daigaku Shuppankai. Tsukishima, Hiroshi. 1986. Heian jidai kuntenbon ronkō: Okototenzu kanajitaihyō [Research on kunten materials of the Heian period: Okototen charts and tables of kana shapes]. Tokyo: Kyūko Shoin. Tsukishima, Hiroshi. 1996. Heian jidai kuntenbon ronkō: Kenkyūhen [Research on kunten glossed materials of the Heian period: General research]. Tokyo: Kyūko Shoin. Yoshida, Kanehiko, Hiroshi Tsukishima, Harumichi Ishizuka and Masayuki Tsukimoto (eds.). 2001. Kuntengo jiten [Dictionary of kunten language]. Tokyo: Tōkyōdō Shuppan.
Shingo Yamamoto
25 Japanized written Chinese: Its features and contribution to the history of the Japanese language 1 About waka kanbun 1.1 The term waka kanbun The term Waka kanbun (和化漢文) – literally, “Japanized written Chinese” – is used to designate kanbun texts written by native Japanese speakers entirely in kanji, in general without kana.Waka kanbun takes as its visual model orthodox Classical Chinese (known in Japanese as jun kanbun 純漢文 “pure kanbun” or seikaku kanbun 正格漢文 “proper kanbun”), but it may incorporate distinctively Japanese lexical and syntactic patterns referred to as washū 和習和臭 “Japanese practices” or “Japanese odor” respectively.
1.2 The taxonomies of waka kanbun Texts written in waka kanbun range over a wide variety of contents, with the degree of Japanese elements they incorporate depending on the type of text. Previous attempts at a general taxonomy include the following: Sakakibara (1878), Tsukishima (1963), Oyama (1982), Minegishi (1986), Miho (2004). Sakakibara’s pioneering study classifies documents in Chinese written by Japanese speakers based on content and genre. Tsukishima constructs a taxonomy based on the relation between writer and reader and applies it to try to organize the wide range of extant Heian period (794–1185) documents in this genre. Oyama introduces two new criteria for classifying waka kanbun texts: whether or not the text adopts a style based on the contrapuntal rhythmical couplets of four and six characters characteristic of Chinese Six-Dynasties prose style (pianwen 駢文), and presence or absence of phonogrammatic writing (kana). Building upon and responding to these approaches, Minegishi (1986) attempts to systematize previous scholarship. As a first step, Minegishi defines as “pure” Chinese (jun kanbun) all texts that aspire to be taken as written Chinese and that conform to the lexical, syntactic and orthographic conventions of continental prose. By contrast, he defines as hentai kanbun 変体漢文 (“deviant or non-orthodox written Chinese”) texts in classical Chinese written by native Japanese speakers. Within this category, Minegishi classifies texts which aspire to be jun kanbun but exhibit orthography, lexicon, or grammar not found in the language of the Chinese classics as waka kanbun. He classifies as kirokutai 記録体 kanbun texts with a strong practical or utilitarian orientation such as https://doi.org/10.1515/9781614512851-026
510
Shingo Yamamoto
private records, which display their own traits distinct from jun kanbun. He distinguishs these two types of hentai kanbun from what he calls manabon 真名本, materials which are kanbun texts but which were originally conceived of or independently attested as kana texts (kanabun 仮名文) or mixed kana/kanji texts (kanji kana majiribun). Miho (2004) concurs for the most part with Minegishi, but he further proposes two subdivisions of hentai kanbun: monjotai (monjogo) 文書体(文書語), chiefly used in letters and bureaucratic documents, and kirokutai (kirokugo) 記録体(記録語), found in practical and private writings. The taxonomies described above testify to the difficulties involved in the classification of this material. To avoid misunderstanding, I would like to emphasize that in this chapter I use the term waka kanbun to designate the same class of texts that Minegishi calls hentai kanbun. We see, then, that the label chosen to indicate forms of written classical Chinese with native Japanese features depends on the scholar. KASUGA Masaji (1878–1962) and KOBAYASHI Yoshinori use the term waka kanbun, HASHIMOTO Shinkichi (1882– 1945), TSUKISHIMA Hiroshi (1925–2011), and MINEGISHI Akira (1935–2012) prefer hentai kanbun, YAMAGUCHI Yoshinori calls it kanshiki wabun 漢式和文 “Chinese-style Japanese”, while KOMATSU Hideo (1929-2022), OKIMORI Takuya, and INUI Yoshihiko use giji kanbun 疑似漢文 “pseudo-Chinese” instead (cf. Tanaka 2019: 69-82). In this paper I avoid the term hentai, with its negative connotation of “deviation” from jun kanbun “pure written Chinese”. Instead I adopt the term waka kanbun “Japanized kanbun”. One reason for this, as pointed out by Minegishi (1986), is that it is not the case that native Japanese speakers writing kanbun were always attempting to emulate the model of jun kanbun. A second reason comes into play when we consider that Chinese texts written on the Korean peninsula in premodern times are referred to as hanhwa hanmun 韓化漢文 “Koreanized classical Chinese” or sok hanmun 俗漢文 “vulgar classical Chinese”. This points up the need for objective terms that distinguish the localized use of written Chinese as a means of expression in various countries of the Sinosphere. Waka kanbun is an appropriate choice for this purpose in the case of Japan.
1.3 Japanese features in waka kanbun The following have been pointed out as distinctively Japanese features introduced by Japanese writers of kanbun texts. These features have been taken as criterial in identifying certain texts as waka kanbun. 1) In some passages, word order follows Japanese syntax. 2) Some characters are used with functions that are different from jun kanbun. 3) Honorific auxiliary verbs and other elements usually added in kundoku ‘vernacular reading practice’ are rendered in the text by means of a Chinese character. 4) Native Japanese lexical items (wago) are written in Chinese characters. 5) Presence of Sino-Japanese vocabulary created in Japan – wasei kango.
25 Japanized written Chinese: Its features and contribution
511
6) Use of phonograms, such as man’yōgana, hiragana, and katakana to record material other than place names or proper names. Note further that the vocabulary and syntax of waka kanbun are mainly based on kanbun-kundoku ‘vernacular reading practice’. When one or more of these traits are present, the text can be classified as a form of Japanized written Chinese.
1.4 Waka kanbun, a brief history The oldest extant document in waka kanbun is the Bosatsuhankazōmei (Tokyo National Museum) – an inscription engraved on the lower frame of the pedestal of a bronze statue portraying a seated nyorai, probably dating to 606 (for a full set of images of the statue and the inscription see https://emuseum.nich.go.jp/detail?langId=ja&web View=&content_base_id=100730&content_part_id=0). (1) 歳次丙寅年正月生十八日記高屋大夫爲分韓婦夫人名阿麻古願南无頂礼作奏也 歳の丙寅に次る年、正月生十八日に記す。高屋大夫、分れし韓の婦夫人、名は阿麻 古が為に願ひ、南无頂礼して作り奏す “This inscription is written on the 18th day of the first month of the year of Hinoetora (606). Lord Takaya gives a prayer for his deceased wife from Korea, Amako, and erects this statue in her name.” The expression 作奏 used at the end of the text includes the character 奏 expressing the humble auxiliary verb mawosu (mōsu) after the verb 作. This is a Japanese construction which does not exist in Chinese. Through the end of the Nara period, official public documents were written in kanbun in such a way that they could be read as Chinese texts, in the Chinese of the time. Formal historical records, volumes on the legal system, and kanshi poetry collections were written relying on their respective stylistic models imported from the continent. Sometimes, however, native elements were added that incorporated distinctly Japanese features within a relatively Sinicized context. Such documents vary in range from official histories such as the Nihon shoki (720) to poetry collections such as the Kaifūsō (751); from epigraphic inscriptions on metal or stone (kinsekibun) to imperial edicts, administrative documents, and Buddhist commentaries such as the Hokkegisho (615). Among Heian period texts, the Shoku Nihongi (797), Ruijusandaikyaku (11th century), and Ryō no gige (833) are examples of historical works. Records of court ceremonial or ancient family protocols include the Seikyūki (mid-10th century) and Gōkeshidai (11th century), while kanshi poetry collections include the Kankebunsō (900), Honchōmonzui (mid-11th century), and Honchōzokumonzui (1140). This kind of textual
512
Shingo Yamamoto
production continued unabated through the medieval period and beyond; with time, Japanese features become more and more noticeable. At the same time there exists another group of texts which were not written as completely based on continental models, but were still used by native Japanese speakers as a means of recording facts and events in kanbun. These include the Kojiki (712; with the exception of the introduction), the various Fudoki, and the titles and interlinear notes of the Man’yōshū. At the beginning of the Heian period this form of waka kanbun was adopted in such diverse writings as public and private records, the diaries of noblemen, such as the Midōkanpakuki, the diary of Fujiwara no Michinaga (966–1027), or Gonijō Moromichiki, the diary of Fujiwara no Moromichi (1062–1099), as well as collections of Buddhist tales such as Nihon ryōiki (9th century) or Gōdanshō (11th century), fictional biographies such as Shōmonki, and even classical textbooks for letter writing (Ōraimono), such as the Kōzanji-bon koōrai or Kasen’ōrai. A form of waka kanbun was adopted in the Kamakura period (1185–1333) chronicle Azuma Kagami (13th century) and came to be referred to as Azuma kagamitai. Texts in waka kanbun were produced well into the 1880s and the modern era. The emergence of the hiragana and katakana syllabaries in the Heian period led to the creation of a new form of writing that mixed logographic and phonographic scripts (kanji kana majiribun); nevertheless, waka kanbun was never abandoned. The century-long success of this type of written language can be attributed not only to the prestige enjoyed by Chinese culture within the Sinosphere, but also to the large amount of information that this kind of logographic script made available to the reader in a few lines, a crucial issue when it comes to writings meant to have a practical use.
2 Main issues in research on waka kanbun As discussed in 1.2, research on waka kanbun in Japan divides into two main subject areas: texts that aspire to the model jun kanbun as represented by classical Chinese literature, and texts that do not aspire to this model, but rather were produced as a form of intercommunication and documentation among native speakers of Japanese; the focus of research up to the present has been placed on the latter, so-called kirokutai, texts with a strong practical orientation such as private records. In 1992, MINEGISHI Akira, the author of the most detailed taxonomy of waka kanbun (Minegishi 1986), published a paper entitled “Honchōmonzui no bunshō ni tsuite – Nihon kanbunbuntaihantei no kijun o motomete” (About the text of the Honchōmonzui – Looking for a standard evaluative criterion for Chinese texts produced in Japan) in the journal Kokugo to Kokubungaku. Honchōmonzui is a collection of prose and poetry in Chinese compiled in the middle of the 11th century by Fujiwara no Akihira (989–1066). Minegishi points out that the prose segments included in the anthology were conceived from the beginning as orthodox Chinese texts and, as such, differ greatly from the variety
25 Japanized written Chinese: Its features and contribution
513
– kirokutai – used in the diaries of Heian period noblemen. He contends that it is possible to draw a clear line between the two groups by analyzing the degree of respect for the basic constituent order. As evidence, he points to a highly refined use of characters differentiated by context, adoption of character compounds with specific classical Chinese precedents, widespread use of obscure kango expressions, and the use throughout of highly complex syntactic structures. But is it truly possible to distinguish two subtypes of waka kanbun so sharply? My view on this point is skeptical. In the following pages I take up the topic of a relatively neglected category of documents, kanbun texts written by Japanese Buddhist clerics for public and private use. Within this category, I consider how specific linguistic features vary according to genre across texts written by a single individual.
3 The Kajūji administrator of monastic affairs, Kanjin, and his writings 3.1 The life and writings of Kanjin Kanjin (1084–1153, sometimes read Kanshin) was a greatly revered monk of the Shingon school also known by the names Kajūjiryūso – i.e. founder of the Kajūji school, and Kajūjihōmu, Kajūji administrator of monastic affairs. Kanjin was the son of the Associate Counselor, Administrator of the Right, and Minister of the Treasury (sangiudaiben ōkurakyō) Fujiwara no Tamefusa (1049–1115) and an eighth-generation descendant of the Chancellor (daijōdaijin) Fujiwara no Takafuji (838–900), after whose posthumous title the temple was named. In the 2nd year of Hōan (1121), Kanjin was introduced to the esoteric doctrines of the Ono-ryū sect by Genkaku (1056–1121) and became the seventh Head Administrator (chōri) of the Kajūji, a position at the top of the temple hierarchy. The bonds within Tamefusa’s family were especially strong; using Kajūji as a spiritual link between members, all the heads of this lineage were appointed to the highest ranks in the court bureaucracy after having served as heads of the controlling board (benkan) or as imperial secretaries (kurōdo). Some of them were close advisers of emperors Gosanjō (1034–1073) and Shirakawa (1053–1129). At the root of their successful political careers was the clever use of the enormous archive of records and documents the family held – a veritable intellectual property, so vast that they were also ironically called “the household of diaries”. Kanjin was no exception. He was supported by his family in pursuing his ecclesiastic career and devoted his talent to the world of religious studies. He wrote several treatises on Buddhist services; his large body of work enabled him to rise as the founder of a new school of thought. His activities span the political world of the lay nobility as well as the religious world of Buddhist law, two different but complementary sectors of
514
Shingo Yamamoto
Insei period (1086–1185) society. Therefore, his work sheds light on the role played by the so-called kiroku in the Japanese society of the late 11thand 12th centuries. Although Kanjin’s works are as varied as his many interests, by and large they all showcase an elegant prose meant to accompany Buddhist rituals. These include illustrative sermons (hyōhaku), praises of virtues (tandoku), replies (hentō), and teachings and warnings (kyōkai). These texts are collected in the Kōzanji-bon hyōhakushū, the Daigoji-bon hyōhakushū and in the Zappitsushū preserved at the Nara National Museum (see Yamamoto 2006). Kanjin is well known for the manuals (jisōsho) he complied on the program and etiquette of rituals of the esoteric school. Among them, the Ruihishō is especially noted for its detailed descriptions of rituals and ceremonies combined with excerpts from various sutras, for its beautiful illustrations, and for outing into writing the secrets of the oral tradition passed down by esteemed monks of the past. (Yamamoto 2007: 19–29). Kanjin also left a number of diaries. The Eijininen Shingon’inmishuhōki published in the Zokugunshoruijū and the Kanjinhōmugoshichinikki held by the Kanazawa Library have been published in modern editions. Less well known is the Chōshōgannen kanjōnikki, a manuscript copy in one volume dating back to the first half of the 14th century and preserved at Tōji. It is to this manuscript that I wish to turn in the next section. Before I discuss the Chōshōgannen kanjōnikki, let me add that kunten materials allegedly based on Kanjin’s glosses have been found in the repository of Kōzanji and in other monasteries in the temple precincts (see Yamamoto 2001, 2003, 2004). The majority of these documents feature a specific type of okototen ヲコト点 commonly known among scholars as Tōdaijiten 東大寺点. It is believed that the Tōdaijiten were transmitted directly from Tōdaiji to Kajūji and that eventually they were passed on to Kanjin by Genkaku, who inherited the tradition of the Ono-ryū from his master Hanshun (1038– 1112). In this way, it is believed, they ended up becoming the traditional form of kunten notation for Kajūji. Kanjin was also a prolific waka poet and some of his works are included in the anthology Shokugosen wakashū (1251). There are records of poetic exchanges between him and Fujiwara no Mototoshi (1060–1142), who rivaled Minamoto no Toshiyori (1055–1129) as one of the representative poets of the Insei period. Kanjin’s linguistic praxis is thus notable for its variety as well as for its the significance for research on the linguistic history of the Insei period (cf. Yamamoto 2011: 235–247).
3.2 The linguistic features of Kanjin’s diary Going back to the question I raised earlier, do we find differences in specific linguistic features when a single individual writes across different genres within the general category of kanbun? In the following paragraphs I compare excerpts from Kanjin’s diary and from drafts of his sermons in an effort to establish whether there is any similarity between these texts, all composed in waka kanbun despite belonging to two different genres.
25 Japanized written Chinese: Its features and contribution
515
The diary is the aforementioned one-volume Chōshōgannenkanjōnikki held at Tōji (ref.: box 285, n. 6). According to its colophon the manuscript was copied by a monk named Sonchū on the 29th day of the 9th month of the 2nd year of Karyaku (1327). The text transcribed by Sonchū was a second manuscript copy of Kanjin’s original by Kōzen (1121–1204, also Kōnen), one of his direct disciples, in the 4th year of Ninpyō (1154). In the first year of Chōshō (1132) Kanjin, who held the offices of Administrator for Monastic Affairs and Gon-Preceptor (gon no risshi), was nominated Great Preceptor (daiajari) and administered the consecration (kanjō) of his disciples Nenpan and Gyōkai (1109–1181). The first half of the diary includes the chronicle of the consecration ceremony, during which the precepts and the mystic teachings were transmitted. The ceremony lasted three days, from the 16th to the 18th of the 10th month of 1132. The second half of the diary includes praises of virtues by Ingen and Myōkai as well as replies by Nenpan and Gyōkai, all written in the elaborate Six-Dynasties prose style. Moreover, the record of one of Kanjin’s dreams is also included as an appendix to the diary. 年四十九先師
長承元年十月十六日癸卯、天晴る、 . . .予年に之を授けらる 始めて以て之を行ず、二人 〔=念範◾行海〕並びて受くること頗る穏便に非ず、而を去ぬる比に両人の択ぶ所の 日次に院の修法に依りて各以て延引す “It is the 16th day of the 10th month of the first year of Chōjō (1132). There are clear skies today. I (who am 49 years old; the same age as my late master when he performed this ceremony) perform this ceremony for the first time. It is not ideal for two monks to undergo the ceremony at the same time; However, the dates that each monk chose for their ceremony overlapped with ceremonies for the emperor and thus had to be postponed.” (3) 又来廿四日上皇令始熊野御精神給、可奉仕御導師之由、俄被仰下 又来る廿四日、上皇は熊野御精神(進) を始めしめたまふ、御導師に奉仕すべき由、 俄に仰せ下さる “And, on the coming of the 24th day of the month, the Emperor Emeritus began a ceremony in Kumano. He promptly ordered that the head priest serve him.” (4) 十七日甲辰土曜并宿天晴. . .以件水令入五瓶香水等并用堂中雑事 十七日甲辰土曜并宿、天晴る、件の水を以て五瓶に入れしめ、香水等と并せて堂中 の雑事に用いる “It is the 17th day of the month. There are clear skies today. I poured the sacred water in question into five bottles and used it together with perfume to clean the temple.”
516
Shingo Yamamoto
(5) 十八日乙巳日曜鬼宿天晴. . .余着平袈裟自後戸偸向東対簾中為窺嘆徳儀也. . .又 受者所持三衣箱同置其前持幡二人擎幡須欄下左右也、而其處有潺〔左訓にはセ ン〕後無便宜仍不令立之 十八日乙巳日曜鬼宿天晴. . .余、平袈裟を着て後戸より偸に東の対の簾中に向か ふ、嘆徳の儀を窺はむが為なり. . .又受者は三衣の箱を所持し同じく其の前に置く、 幡を持てる二人は幡を擎ぐるに欄下の左右に立つべきなり、而を其の處、潺有りて 後に便宜無し、仍て之を立てしめず “It is the 18th day of the month. There are clear skies today. Today I wore a plain robe and walked quietly from the back door to the curtain in the east room in order to observe a ceremony commemorating virtuous deeds. The participant of this ceremony brings a box with three robes and places them in from of himself. Two flag bearers are supposed to stand on the right and left side, holding flags, but cannot do so, since there is a stream running through the garden. Therefore, no flag bearers are used at the ceremony.” Examples (2–5) share the same linguistic features displayed in the diaries of Heian period noblemen. For example, the adversative use of the character 而 (read as sikaru o) at the beginning of a sentence in (2) and (5), the use of formal nouns such as yosi 由 in (3), the use of adnouns such as sannuru 去 in (2) and kudan no 件 in (4), are all distinctive traits of kirokutai. Moreover, honorific expressions such as the construction sime. . .tamau 令. . . 給 in (3) are peculiar to waka kanbun (they are found neither in kanbunkundoku nor in wabun, which uses se {sase} tamau in this function). Furthermore, we see in example (3) that the object Midōshi ni hōshisu beki yoshi 可奉仕御導師之由 ‘that the head priest serve him’ precedes the verb ōsekudasaru 仰下 ‘order (respectful)’, indicating Japanese word order. Nevertheless, not unexpectedly, most passages are written according to orthodox Chinese word order. We can also point out the use of difficult characters not in common use in contemporary Japanese practice, such hisoka 偸 ‘quietly’ and yarimizu 潺 ‘garden stream’ (a notation to the left of this character suggests the reading sen) in example (4). The text also employs specific numeral classifiers for objects of used in Buddhist ceremonies and rituals such as kyaku 脚 for high seats (kōza 高座) and small desks (wakizukue 脇机), tō 桶 for units of pure or holy water (jōsui 浄水), and shi支 for toothpicks (shimoku 歯木) and aspersoria (sanjō 散杖). Such vocabulary is distinctive to a diary of Buddhist ritual.
3.3 The linguistic features of Kanjin’s sermon drafts These writings are clearly intended to be pure classical Chinese (jun kanbun) based on the model of contrapuntal rhythmical couplets of four and six characters that flourished in China during the Six-Dynasties.
25 Japanized written Chinese: Its features and contribution
517
Among such sermons is Tōjimieikudōshi hyōhaku written in the 5th year of Hōen (1139) and included in the Kōzanji-bon hyōhakushū. (6) 覚月光陰、雖送三百之星霜、法水流深、遙及千万之支流 覚月の光は陰くして、三百の星霜を送ると雖も、法水の流れは深くして、遙かに千万 の支流に及ぶ “Even if a great amount of time passes, the doctrine of Buddha is profound and reaches all corners of the world.” (7) 事隆大師之報恩、勤給諸徳之真影 事るに、大師の報恩を隆くして、諸徳の真影に勤め給ふ “When the head priest begins a ceremony, he thinks of the immeasurable grace of Grand Master Kōbō (translator’s note: Kūkai) and performs a ceremony in front of an image of the founders of the sect.” (8) 故一宗官長、数口定額、毎今排灌頂院、逐年供祖師之像 故に、一宗の官長、数口の定額、今毎に灌頂院を排き、年を逐ひて祖師の像を供す “Therefore, the leaders of the sect and several selected priests now open the door to Kanjōin and present an image of the founders of the sect each year.” (9) 依之、一宗弥興隆、 四海悉安穏 之に依りて、一宗は弥よ興隆し、 四海は悉く安穏たらむ “Therefore, this sect will finally prosper and there will be peace all over the world.” The passage cited in example (6) showcases a typical pienwen construction of two contrapuntal rhythmical couplets. Here we have a juxtaposition of two lines of four and seven characters, a stylistic variant of the classical pattern of four and six characters, known as zōkakku, common to the classical Chinese tradition. This may seem to indicate that the document was written strictly according to orthodox Chinese rules. However, but a few lines later, in example (7), the character 給 is used as the honorific auxiliary verb tamau and it is positioned immediately after the verb to which it is affixed. The presence of Japanese features in the text does not stop at the use of honorifics, but also involves the use of classifiers typically employed in private records. For example, kō 口, used for counting monks, appears in example (8). Other lexical choices seem to be designed to aid in the quick comprehension of the text. In example (9), the discourse connective koreniyorite is expressed by 依之, a simpler and more intuitive expression than the refined but convoluted 因茲.
518
Shingo Yamamoto
These examples demonstrate that it is indeed possible to find linguistic features typical of waka kanbun in texts produced by a single author, even when those texts are originally intended to be “pure” Chinese (such as, in this case, sermons). A passage from another sermon by Kanjin, the Tōjikanjōdaiajarisanmayakai hyōhaku, reads: (10)
法皇御宇、更貽勧賞於後代、然則、禁闕恭平、玉躰无恙、射山静謐 法皇の御宇に、更に勧賞を後代に貽す、然れば則ち、禁闕は恭平にして、玉躰は恙 无く、射山は静謐たり “In times when the Emperor Emeritus ruled, his distinguished deeds were praised and remembered for ages to come. Therefore, the Imperial Palace was at peace, the Emperor was healthy, and the Palace of the Emperor Emeritus was quiet and calm.”
The word yasan 射山 in the last phrase indicates the residence of the retired Emperor. This particular use of the character compound is limited to Japan (where this form of government existed in the Insei period) and is an example of a waka kango or Japanized character compound. We are thus able to identify in the text Japanese-like features spanning writing and lexicon as well.
4 Conclusions 4.1 Motivations for the inclusion of Japanese features In previous research, the criterion for dividing Japanized kanbun into two subtypes was whether or not it was written with the objective of emulating orthodox written Chinese (jun kanbun). Material written with this objective has been seen as aiming for a high literary standard and a consciousness of a public audience. The second subtype consists of works which did not aspire to be readable as orthodox written Chinese but rather were conceived from the beginning as Japanese texts, emphasizing practical utility, and more suited for private writings. This taxonomy in previous research can be schematized as follows (the types and features of waka kanbun): a) Aspiring to a Classical Chinese (jun kanbun) model – literary – public b) Aspiring to be read as Japanese – practical – private However, other than works clearly intended for artistic appreciation, such as the prefaces to poetic collections, we should conceive virtually any text as having at least a certain degree of practical applicability. It seems appropriate to consider all materials not intended for mere literary appreciation as texts with potential practical applica-
25 Japanized written Chinese: Its features and contribution
519
bility. This being the case, even documents written for formal public occasions such as bureaucratic communications, pleas to the emperor, letters of resignation or, in the case of Buddhist ritual texts, sermons and dedications or written vows (ganmon) should be seen as documents with such practical features. It should come as no surprise, then, to find various Japanized features in such texts. As we saw in the comparison of Kanjin’s diary and his sermons in section 3, it is possible to identify Japanese linguistic features in both types of texts produced by the same author. Thus, while granting the existence of codified genres introduced from China and granting the existence of texts that attempted to emulate them, the conclusion is justified that even such documents, produced as they were in Japan, aspired from the beginning to be Japanese language texts. For this reason, it seems unjustified to place them in a category separate from the category of diaries or classical textbooks for letter writing. In other words, there is no need to rigidly separate texts heretofore conceived as aspiring to the model of jun kanbun, such as bureaucratic communications, pleas to the emperor, letters of resignation, sermons, and invocations, from the language of diaries, private letters, and textbooks; both should instead be seen as parts of one linguistic continuum. With this in mind, what sort of yardstick should be adopted for evaluating the linguistic features of waka kanbun? The scope of this chapter does not permit a concrete discussion, but in what follows I point out a few possible directions. Previous research has not placed much emphasis on the distribution of the contrapuntal rhythmical couplet form (benbun, Chinese pianwen), the standard introduced by Oyama (1982). In my opinion, the time has come to refocus on Oyama’s proposal. The presence of contrapuntal couplets can be used as a yardstick to assess how literary a given text is; at the same time it increases the restrictions on the introduction of Japanese elements into the text. Structuring a text around contrapuntal couplets not only limits the number of characters to be used in a single phrase but also forces the writer to adhere strictly to orthodox Chinese syntax. This is because the need to adhere to head-initial Chinese word order and the constraint that the number of characters in paired couplets must match prevents the insertion of such Japanese elements as honorific auxiliary verbs. It is not the case that adherence to Chinese word order alone determines whether a text is written in “pure” Chinese or not. Rather, I believe that a more effective yardstick is the extent to which a text uses matched couplets in the pianwen style. Once this yardstick is adopted, we find that works such as diaries and textbooks, hitherto considered typical examples of practical writing, in fact also use contrapuntal prose; the yardstick thus allows us to estimate the literary element in these works as well. The prediction is that in the relevant portions of such texts Chinese word order will be strictly maintained, and it will be difficult to identify distinctively Japanese elements.
520
Shingo Yamamoto
4.2 Texts for future study In this paper I presented the results of research based on selected texts by Kanjin, a twelfth century Buddhist monk. Comparing the Japanized written Chinese in one of his diaries and in several of his sermons, I was able to identify features shared by both genres, demonstrating a linguistic continuity within waka kanbun. Kanjin is not the only author of texts belonging to different genres in Japanized written Chinese. For example, the compiler of Honchōmonzui, Fujiwara no Akihira (989–1066), also authored the biographical chronicle Shinsarugakuki and the Unshūōrai. Shukaku hōshinnō (1150–1202), the prince-monk son of Emperor Go-Shirakawa (1127–1192), wrote a large number of sermons at Ninnaji as well as three diaries, Uki, Saki, and Gyoki (cf. Gomi 2003: 205–224, Uki Kenkyūkai 2011). In future research I hope to incorporate these works into an expanded overall picture of Japanized Chinese from the Kamakura period.
References Gomi, Fumihiko. 2003. Shomotsu no chūseishi [Medieval history of books]. Tokyo: Misuzu Shobo. Miho, Tadao. 2004. Komonjo no kokugogakutekikenkyū [Japanese linguistic research on old handwritten documents]. Tokyo: Yoshikawa Kōbunkan. Minegishi, Akira. 1986. Heian jidai kokiroku no kokugogakutekikenkyū [Japanese linguistic research on old records and diaries from the Heian period]. Tokyo: Tōkyō Daigaku Shuppankai. Minegishi, Akira. 1992. Honchōmonzui no bunshō ni tsuite: Nihon kanbun buntai hantei no kijun o motomete [About the text of the Honchōmonzui: Looking for a standard evaluative criterion for Chinese texts produced in Japan]. Kokugo to Kokubungaku 69(11). 42–55. Oyama, Norihisa. 1982. Hentai kanbun no buntaishi [A stylistic history of hentai kanbun]. In Morioka Kenji, Yutaka Miyaji, Hideo Teramura and Yoshiaki Kawabata (eds.), Kōza Nihongogaku 7: Buntaishi 1 [Studies in Japanese linguistics, vol. 7: Stylistic history 1], 149–187. Tokyo: Meiji Shoin. Sakakibara, Yoshino. 1878. Bungeiruisan: Monshijō [Compendium of the liberal arts: The history of writing styles]. Tokyo: Monbushō. Tanaka, Sōta. 2019. Heian jidai ni okeru hentai kanbun no kenkyū [Research on hentai kanbun in the Heian period]. Tokyo: Bensei Shuppan. Tsukishima, Hiroshi. 1963. Heian jidai no kanbun-kundoku-go ni tsukite no kenkyū [A study of the lexicon used in glosses from the Heian period]. Tokyo: Tōkyō Daigaku Shuppankai. Uki Kenkyūkai. 2011. Den Shukaku hōshinnō cho Uki no shahon ni tsuite [On the manuscript of Uki handed down as the work of Cloistered Imperial Prince Shukaku]. In Bukkyō Bunka Kenkyū Ronshū vol. 13/14, 35–69. Yamamoto, Shingo. 2001. Kōzanjikyōzō Kajūjihōmu Kanjin kankei bunken mokurokukō 1 [Catolog of literature related to Kanjin of Kajūji held in the scripture house of Kōzanji, part 1]. In Kōzanji tensekimonjo sōgōchōsadan kenkyū hōkoku ronshū [Research reports of the Kōzanji temple documents survey team], 67–74. Tokyo: Kōzanji tensekimonjo sōgō chōsadan. Yamamoto, Shingo. 2003. Kōzanjikyōzō Kajūjihōmu Kanjin kankei bunken mokurokukō 2 [Catolog of literature related to Kanjin of Kajūji held in the scripture house of Kōzanji, part 2]. In Kōzanji tensekimonjo sōgōchōsadan kenkyū hōkoku ronshū [Research reports of the Kōzanji temple documents survey team], 79–89. Tokyo: Kōzanji tensekimonjo sōgō chōsadan.
25 Japanized written Chinese: Its features and contribution
521
Yamamoto, Shingo. 2004. Kōzanjikyōzō Kajūjihōmu Kanjin kankei bunken mokurokukō 3 [Catolog of literature related to Kanjin of Kajūji held in the scripture house of Kōzanji, part 3]. In Kōzanji tensekimonjo sōgōchōsadan kenkyū hōkoku ronshū [Research reports of the Kōzanji temple documents survey team], 49–62. Tokyo: Kōzanji tensekimonjo sōgō chōsadan. Yamamoto, Shingo. 2006. Heian-Kamakura-jidai ni okeru hyōhaku, ganmon no buntai no kenkyū [Research on the style of hyōhaku and ganmon texts from the Heian and Kamakura period]. Tokyo: Kyūko Shoin. Yamamoto, Shingo. 2007. Kamakura jidai ni okeru jiinkyōzō to sono shoshakatsudō [Scripture houses and sutra transcription in the Kamakura period]. Gunki to Katarimono [War chronicles and narratives] 34. 19–29. Yamamoto, Shingo. 2011. Kajūjihōmu Kanjin no gengoseikatsu ni tsuite [On the language life of Kanjin of Kajūji]. In Gengohenka no bunseki to riron [Analysis and theory in language change], 235–247. Tokyo: Ōfū.
Fumitoshi Saito
26 Early modern kanbun and kanbun-kundoku 1 Introduction Kanbun-kundoku is a kind of a translation system in which, by marking up the original foreign language classical Chinese text with symbols like kaeriten (marks showing the order in which words and phrases are to be read) and kana, the text is converted into Japanese. In contrast to the general case of translation in which no trace of the original text remains after the translation is complete, the case of kanbun-kundoku has the special characteristic that the original text remains as is even after the translation is complete. Kanbun-kundoku has been conducted since long ago (the end of the Nara Period [710–794 CE]), but the methods of kanbun-kundoku in the early modern (Edo [1603–1868]) period, while inheriting much from the earlier tradition, had changed greatly historically and also showed a number of peculiar characteristics influenced by systems of Confucianism. The sentence style (kanbun-kundoku-bun) and the word usage/grammar (kanbun-kundoku-gohō) deriving from this later kanbun-kundoku influenced not only the early modern language but also modern literary style. Furthermore, the Dutch Studies scholars (rangakusha) of the early modern period and the Western Studies scholars (yōgakusha) (also, English Studies scholars [eigakusha]), continued earlier approach to language study and used the methods of kanbun-kundoku in the study of the Dutch and English languages. For this reason, among the Dutch Studies and English Studies materials are found many patterns that incorporate the structures and methods of kanbun-kundoku unmodified and which, as structures characteristic to translationese and direct translations from European languages, have had a large influence on the structure of modern Japanese. Sections 2–4 of this chapter overview the changes that occurred in kanbun-kundoku throughout the early modern period. Section 5 provides examples of causative constructions to demonstrate some of these changes concretely.
2 Kanbun-kundoku and translation Kanbun-kundoku up to the beginning of the early modern period was a kind of translation process in which the original foreign, classical Chinese text was glossed with symbols like kaeriten and kana so it could be interpreted as Japanese. However, in modern times it is difficult to interpret as Japanese the meaning of the kanbun-kundoku-bun, which is meant to be a translation resulting from the reading of the classical Chinese text, and so in many present-day kanbun texts a translation into modern Japanese is provided in addition to the original sentence and the annotated kanbun-kundoku-bun, as shown https://doi.org/10.1515/9781614512851-027
524
Fumitoshi Saito
in the following examples from the opening to The Analects drawn from two currently well-read paper-back editions. The first one is from Kaneya (1963): Original: 子曰、学而時習之、不亦説乎。有朋自遠方来、不亦楽乎。 Kundoku-bun: 子 の 曰わく、 学びて 時 に これ を Si no notamawaku, manabite toki ni kore wo master gen speak studying time LOC this ACC 習う、 亦 説ばしからず や。 朋 あり 遠方 narau, mata yorokobasikarazu ya. Tomo ari, enpoo learn again pleasing.not SFP friend exist far.away より 来たる 亦 楽しからず や。 yori kitaru, mata tanosikarazu ya. from come.PRF again enjoyable.not SFP Modern Sensei ga iwareta. “manande wa tekitoo-na Japanese: master NOM say.HON.PST studying TOP suitable-ADN ziki ni o-sarai-suru, ikanimo kokoro-uresii koto time LOC HUM-review-do ever.so.much heart-glad thing da ne. (Sono tabi ni rikai ga COP SFP that (repeated).time LOC understanding NOM hukamatte wa koozyoo site iku no da kara.) deepen TOP improve do.get go that COP since Dareka tomodati ga tooi tokoro kara mo someone friend NOM far place from also tazunete kuru, ikanimo tanosii koto da ne. visit.GER come ever.so.much pleasant thing COP SFP (Onazi miti nituite katari-aeru kara.) same path regarding talk-RECP since The Master spoke. Studying and reviewing at a suitable time is such a pleasant thing, isn’t it? (Because each time one’s understanding deepens and improves). Having some friend come to visit from a far-away place is such an enjoyable thing, isn’t it? (Because one is able to have a discussion about studies in common)
The second one is from Kaji (2009): Original: 子曰、学而時習之、不亦説乎。有朋自遠方来、不亦楽乎。 Kundoku-bun: 子 曰く、 学びて 時 に 之 を si iwaku manabite tune ni kore wo Master says studying always LOC this ACC 習う。 亦 説(悦) ばしからず や。 朋 遠方 narau mata yorokobasikarazu ya tomo enpoo learn again pleasing.not SFP friend far.away
26 Early modern kanbun and kanbun-kundoku
Modern Japanese:
525
自り 来たる あり。 亦 楽しからず や。 yori kitaru ari mata tanosikarazu ya from come.PRF exist again enjoyable.not SFP Roo-sensei wa, bannen ni sinkyoo wo old-master TOP late.in.life LOC mental.state ACC koo arawasareta. (tatoe huguu na this.way express.HON.PST even misfortune COP toki deatte-mo) manabu koto wo tuzuke, time COP.GER-even study act ACC continue (itudemo sore wo katuyoo dekiru yooni) at.any.time that ACC make.use.of can so.that tune ni hukusyuu suru. Sono yoo ni always LOC review do that way DAT site zibun no mi ni tuite-iru no do.GER self GEN body LOC attached-be thing wa nanto yukai dewanai ka. Totuzen, TOP how pleasant is.it.not SFP suddenly yuuzin ga tooi tooi tokoro kara (watakusi wo friend NOM far far place from me ACC wasurenaide) tazunete kite kureru. forget.GER.not visit come give.me natsukasikute kokoro ga atatakaku naru dewanai ka. nostalgic.GER heart NOM warm become is.it.not SFP
The old Master expressed his feelings in his old age in this way: Continue studying (even in times of misfortune) and constantly review (so that you may put your learning into practice at any time). Isn’t what you gain for yourself in this way truly pleasant? A friend suddenly comes to visit from far, far away (not having forgotten about you). Don’t you feel nostalgic and doesn’t your heart grow warmer?
Looking first at the changes in kanbun-kundoku that took place in the early modern period, I would like to pursue the reasons that kanbun-kundoku sentences had to end up becoming something other than translations.
3 Changes in kanbun-kundoku in the early modern period The kundoku of the Chinese canon changed in the early modern period from the kundoku system of the Hakase-ke (a family of scholars who held a hereditary professor (hakase) position teaching at the imperial university continuously since the Heian period) because of new commentaries by Shushi (Zhu Xi [1130–1200]). Furthermore,
526
Fumitoshi Saito
as shown in the following passage, a variety of kundoku systems (named for the person who added the marks showing word order and readings, such as Dōshun-ten, Gotō-ten, Issai-ten) arose at this time. The Satsuma priest, Nanpa Bunshi, read the Four Books and Hayashi Razan read the Four Books and Five Classics and since then there have been tens of scholars emulating them and their various books have spread into the world. (DAZAI Shundai [1660–1747] Wadoku yōryō [Japanese reading essentials]. Volume 1, leaf 1, reverse. 1728)
Among the newly proposed systems, there was already a proposal espoused in the early modern period by OGYŪ Sorai and his disciple DAZAI Shundai that advocated abandoning kanbun-kundoku itself and, since kanbun was a foreign language, advocated reading it as a foreign language without going through the half-way measure of kanbunkundoku. In his Kun’yaku jimō [Japanese reading suggestions], published in 1738, OGYŪ Sorai [1666–1728] argues: If scholars now think to learn translated works, they should entirely get rid of the Japanese readings and reordering of words that have been learned (and used) in Japan since early times. (Volume 1, leaf 3, reverse)
Saying, in other words that they should not read the characters in their Japanese readings but should read them in their Chinese readings and should stop changing the order of words to fit Japanese word ordering. Furthermore, the opening of the first volume (Daigen jissoku [Epigraph of ten rules]) of his Yakubun sentei [Essentials of translated works] (established 1711) is famous as a work criticizing kanbun-kundoku. In it he says, Here [in Japan] we naturally have our own language; in China they naturally have their own language. They are fundamentally different in their substance and they are not in any way coincident. Taking [the Chinese text] and rearranging it to read as Japanese may seem to make [the text] intelligible, but this is really distorting [the text]. (. . .) Therefore, a scholar’s first efforts need to be directed only to learning the original quality with regard to the language of the Chinese. (Imanaka and Naramoto 1974: 547) Thus, I have determined a method for learning for students of [the works of] Mencius. First, use the study [methods] of Nagasaki: use colloquial language in teaching, use Chinese readings when reciting, use the vernacular of this country [Japan] for translating. Absolutely do not create a rearranged-into-Japanese-order reading. (idem, page 555)
Arguing for abandoning kanbun-kundoku style “rearranged-into-Japanese-order reading” and directly understanding the kanbun text as Chinese. This insistence was further taken up by his disciple DAZAI Shundai who argued that kanbun-kundoku was simply a device of convenience, saying, Although, generally speaking, when reading a Chinese text, it is better to use Chinese pronunciation and to read in order from beginning to end [top to bottom], in this country, because it is difficult to learn the Chinese pronunciation, out of necessity [people] read in the Japanese pronunciation. (Volume 1, leaf 2, obverse)
26 Early modern kanbun and kanbun-kundoku
527
The insistence that kanbun, used to write a foreign language, should be read unmodified as Chinese thus became one motivating force causing the kanbun-kundoku of the early modern period to change greatly. The direction of the change was, to put briefly, one of “simplification”, and points that can be adduced as concrete characteristics of this simplification include the fact that the number of hodokugo (words added to complete a reading) decreased and that use of ondoku (Chinese pronunciations) increased. Regarding the influence of the kanbun-kundoku of the early modern period on the Japanese language, in the modern period ŌTSUKI Fumihiko [1847–1928] wrote the following in his Kō-Nihon bunten: Bekki (Ōtsuki 1897: 19–20). Probably the level of language has never been as informal and incoherent as it is now. When seeking the cause of this state of affairs, although it is true that some derives from changes in the language over many years and in the lack of school education, the greatest cause is absolutely the reading marks for kanbun and the reason that it has become a source of misfortune lies in the systems for reading marks that have been churned out in great numbers over the last hundred years. In the Four Books and Five Classics, although the Dōshun marking system is not without distortion in places, still it has continued the traditions of the marking system of the Kan (Sugawara) and Gō (Ōe) clans and it preserves the use of sutegana (kana written small) and furigana (kana written alongside a kanji to show intended reading) to show lexical qualities like transitive, intransitive, potential, passive, past, present, and future. However, from the time of the “Three Kansei Teachers [KOGA Seiri (1750–1817), BITŌ Jishū (1745–1814), SHIBANO Ritsuzan (1736–1807); some sources cite OKADA Kansen (1740–1816) in place of Koga]”, the furigana of the old marking systems have been abandoned and everything has been read in the Sino-Japanese pronunciation. As a result, by simply learning the many words that, though they have the same readings in Japanese are written with different characters and have different meanings by chanting in rhythm, insufficient attention comes to be paid to the fact that they have different characters and, when reading Chinese works, words with different meanings are confused [understood in a mixed way]; when writing Chinese, characters get reversed and Japanese-esque usages arise. I hear that it is quite a job to prevent these problems and to correct them. Well then, since the time of these “Three Teachers” kanbun-kundoku has greatly progressed and revisions of its substance are striking. When setting out to become a specialist in kanbun, that is how it ought to be, (reading in ondoku only, interpreting Chinese text, even though Chinese texts cannot be created), However, doing things in this way, the old marking methods have become lost. Once the opening to the breakdown had been breached, things got to the point that the systems of So-and-so of subsequent Confusion scholars proliferated and the furigana and sutegana of the older methods decreased drastically and they variously were revised to become things that they were not (. . .) [and] finally texts that were neither rules nor quality came to be created – and the most drastic of these was the Issai-ten. Here lies the true culprit responsible for the dissolution of the quality of language.
In other words, among the various kundoku marking systems that appeared in the early modern period, in contrast to the early period Dōshun system (the method of HAYASHI Razan [1583–1657]), which continued the methods of the Hakase-ke marking system and “preserves the use of sutegana (kana written small) and furigana (kana written
528
Fumitoshi Saito
alongside a kanji to show intended reading) to show lexical qualities like transitive, intransitive, potential, passive, past, present, and future”, which was well regarded even in modern times, the Issai system (the method of SATŌ Issai [1772–1859]) from the latter period was regarded as “the true culprit responsible for the dissolution of the quality of language” and was evaluated very poorly. In other words, the kanbun-kundoku of the early modern period was steadily simplified and the Japanese (the kanbun-kundoku sentences) that resulted from the Issai system was far removed from natural Japanese. The kanbun-kundoku text was quite odd from the point of view of Japanese grammar, but, conversely, it held one great advantage for learners of kanbun. Because it was mechanical, it was easy to get from the kanbun-kundoku sentence back to the original Chinese (kanbun). Ōyō kanbun-gaku [Applied Chinese studies] by OWADA Takeki [1857–1910], published in 1893, was a book that was primarily critical of the Issai system but, on the other hand, it pointed out that the Issai system was a kundoku method that was intended to “provide a model for composition [in Chinese] while reciting aloud the translated reading due to fearing to stray too far from the convenience of learning kanbun”. The “reciting aloud the translated reading” here refers to the particular early modern method, also called “sodoku (reading a text aloud without trying to understand it)”, of learning kanbun which consisted of first memorizing and reciting aloud from memory the kanbun-kundoku sentence alone and only later teaching the meaning separately. In summary, at the beginning of the early modern period, kanbun-kundoku had the function of “translation” to a certain extent, but in the later period the Japanese that the kanbun-kundoku gave rise to became even more unnatural, and, in the modern period became a target for criticism. The Issai system was the archetypical exemplification of this. The kanbun-kundoku sentences constructed in this way had a quality very different from the language of daily use and it was here that the “pattern” of kanbun-kundoku sentences was constructed. One sort of literary style that had incorporated this “pattern” came to be favored as the “kanbun-kundoku style” and spread even more widely through the practice of “sodoku”. But in the present times, it has become difficult to understand the meaning of the text from the kanbun-kundoku sentences alone and, as described earlier, many currently published kanbun texts have come to present each sentence as a set consisting of the original kanbun (Chinese) sentence, the kanbun-kundoku version, and a translation into modern colloquial Japanese.
4 Kanbun-kundoku as colloquial translation in the early modern period In this way kanbun-kundoku was rejected by OGYŪ Sorai insisting, “use Chinese readings when reciting, use the vernacular of this country [Japan] for translating. Absolutely do not create a rearranged-into-Japanese-order reading”. In the world of kanbun-kundoku as well, there were reading (kundoku) systems that aimed to produce kanbun-kundoku
26 Early modern kanbun and kanbun-kundoku
529
as a colloquial translation. One of these was the Sanpei system of UNO Meika [1698– 1745]. As we can see in the following citation, in the Sanpei system, one word may be read as several depending on the context, aiming for a kundoku that was a colloquial translation, but the Sanpei system was criticized in kanbun circles as being “vulgar”, with MIURA Baien [1723–1789] making the following admonition in his Baien dokuhō (‘The Baien reading method’), published in 1773. Lately the system of the teacher UNO Meika has appeared and it is strange. He apparently translates a character using a Japanese word without drawing on the ancient translations but uses vulgar words that serve to translate exactly. He does not read 且 as katu ‘well, moreover’ or 暫 as sibaraku ‘some time’ but as maa ‘well’ and tyotto ‘just a bit’. However, if one can distinguish the elegant from the vulgar, then one should grasp the translated meaning following the elegant. (. . .) DAZAI Shundai says we should avoid falling into the vulgar and that Sanpei should quit academia and go write haiku. (Baienkai 1912: 327)
There were also works that appeared in the early modern period criticizing the Issai system, which had become extremely mechanical. An archetypical example of this is Kunten fukko [Restoration of kunten], published in 1835, by HIO Keizan [1789–1859], who wrote: There are those knowledgeable about things of old who devote themselves to Neo-Confucian studies. Watching them teach the reciting aloud of the commentaries to the Four Books, they have formed the bad habit of discarding the marking systems of Seika and Razan and recite them without translation. They try to form their own schools. In this way, while studying Zhu Xi, they turn their backs on Zhu Xi. I call it disrespect. I call it rudeness. (. . .) One who, observing these works, recognizes the lacks in the current system, will immediately return to the system of old.
From the above description we can see that in calling for a restoration of kundoku, Keizan looks to the kundoku systems of FUJIWARA Seika and HAYASHI Razan as his model, but what HIO Keizan is using as the standard for his criticism is “listening from the other side of a wall and seeing”, emphasizing the unnaturalness as Japanese. However, the kundoku system advocated by HIO Keizan (the Hio system) remains within the framework of kanbun-kundoku and it cannot be said that his kundoku renditions were understandable to Japanese speakers at the time.
5 Characteristics of the kanbun-kundoku of the early modern period (using the example of causative constructions) The concrete characteristics of the kundoku methods of the early modern period, which had become simplified, can be summed up, as described earlier, in the two points: 1) the number of hodokugo (words added to complete a reading) decreased, and 2) use of ondoku (Chinese pronunciations) increased
530
Fumitoshi Saito
Below I would like to take causative constructions as an example and show what 1) (“the number of hodokugo (words added to complete a reading) decreased”) consists of. Causative expressions are found not only in modern kanbun style sentences but also influenced Dutch Studies materials and English Studies materials and, through these, we can look at the connection with direct translation method for European materials. Examining the patterns used in the early modern kundoku versions of the Analects to express the causative construction – X causes Y to Z – we find the following three patterns, I, II and III. (The following conventions are used in transcribing examples below. 1. Vertical writing has been made horizontal. 2. Kana originally to the right of kanji are transcribed above and kaeriten originally to the left are transcribed below. 3. The following lines contain a reading of each kanji, a kundoku rendition, and an English translation, respectively.) site wo 使 Y Z simu, 2, B 1 A Read: X, Y wosite Z-simu. AGT -cause.to.do
I
X
II
X
mu wosite 使 Y Z 2 1 Read: X, Y wosite Z-simu. AGT -cause.to.do
mu ni 使 Y 2 Read: X, Y ni AGT
III X
Z 1
Z-simu. -cause.to.do
The pattern in (I), where the causative auxiliary 使 (other causative auxiliaries include: 令, 遣, 教, and 俾) is read twice originated in the Hakase-ke system. In Keian Oshōkahō waten [Keian Oshō school method marking system] established in 1501 by KEIAN Genju [1427–1508] it says, “Insert ‘site’ below the character and add ‘simu’ to its right side. If you mark it this way [viz., as in pattern I] the reading order markings overlap”. For this reason, that the markings showing the reading order (2, 1, B, A in the above example) overlap, causing confusion, he rejects this pattern. However, this pattern was used in texts of the early modern period that continued the Hakase-ke system tradition (for example, the Dōshun system of the early Edo period) and KAIBARA Ekiken [1630–1714]
26 Early modern kanbun and kanbun-kundoku
531
made much the same observation as KEIAN Genju in his work Tenrei [Marking system examples] published in 1703. Examples of marking with the character 使 論語
曰
Rongo
iwaku
wosite 使 門人 為 2 simu monzin ta
「子路 Siro
ra 1-reverse
Rongo iwaku: Siro monzin wosite sin Analects said Zhong You, disciple agt servant ‘The Analects said, “Zhong You made the disciples act as attendants.”’ 又
曰
mata
iwaku
「能 yoku
reru
使
2 simu
wosite 枉
者
magu mono
ra
直
1 nao
臣」 sin tarasimu. cause.to.act
」
Mata iwaku: Yoku magureru mono wosite naorasimu. ‘It further said, “It is possible to cause one that is bent to recover.”’ The wosite should be read placed later [below]. If written in the following way, the markings showing reading order multiply and become confusing. site 使
wo ra 門人 為 臣 2, B 1 A-reverse simu monzin ta sin monzin wosite sin tarasimu disciple AGT servant cause.to.act ‘made the disciples act as attendants’ This is not the intent in writing the characters. The reading order for the character 令 should is the same. (Volume 1, leaf 31, reverse)
As a result of such criticisms, in the latter part of the early modern period, pattern II, which is generally used in the present times, became predominant. Although the change to pattern II avoids the confusion of reading a single character (e.g. 使) twice, there is no change in the use of wosite for marking the target of the causation (the causee). However, in the Issai system, which used an extremely simplified kundoku method, the causee is marked with ni, an even more simplified form than wosite. Furthermore, in the Issai system, there is a tendency to drastically reduce the number of hodokugo (words added to complete a reading), even omitting the formal noun koto ‘thing, that’ and perfective auxiliaries (See Saitō 2011, Chapter 2). The causative auxiliary verb primarily used in kanbun-kundoku is simu and for the causee marking, according to Ōtsubo 1981: 683, “wosite is commonly used but ni and wo are also sometimes used”. Also, according to Oda 2003: 24, who searched the Konjaku monogatarishū, wo, womotte, wosite, ni, and notameni are all found. Thus,
532
Fumitoshi Saito
while there are examples using ni in the pattern Y-ni Z-simu, as in the Issai system, in the kanbun-kundoku systems of the early modern period, expressions using wosite in the pattern Y-wosite Z-simu have become fixed patterns. Thus in his Kunten fukko (1835) criticizing the Issai system, HIO Keizan gives the following description. 子
su 疾病
、子路
Si
sippei
Siro
使 2 simu
ni 門人
為
monzin
ta
ra 1-reverse
Si sippei su Siro monzin ni sin Master illness do Zhong You disciples agt servant ‘The master fell ill. Zhong You made the disciples act as attendants.’
臣 sin tarasimu cause.to.act
Marking the text this way, it has the reading, si sin ni monzin ‘cause the servants to be in/among the disciples’. Even though, of course, a Chinese person’s text would not be read in this way, given the influence of this marking system, it certainly could be this way. (. . .) Thus, if one were not confronting the text itself but were listening from the shadows, one might hear it and interpret it as saying that Confucius became ill so Zhong You made Confucius become a servant to the disciples. Because we have learned from sight and from hearing that 子 si ‘Master’ refers to Confucius, there is no mistake, but to someone who didn’t know anything, how would they hear and interpret Tarobei ga byooki de atta. ‘Tarobei was ill.’ Ziroobei monzin ni simobe to naraseta ‘Jirobei made (someone) become servants to the disciples.’ OR ‘Jirobei made the disciples become servants (to someone).’ It is because it is read with monzin ni that this sort of confusing ambiguity is heard. Even though it would be unambiguous with monzin wo, this does not meet with the rules for the marking system. However, if we read it as monzin wosite as in the ancient system, wouldn’t it fit well. (Volume 2, leaf 20, reverse)
The phrase “not confronting the text itself but were listening from the shadows” points to the fact that the kundoku texts of the Issai system were unnatural as Japanese. HIO Keizan’s claim that monzin wo is better than monzin ni is also worth noting. According to Oda 2003: 24, in the Konjaku monogatarishū, a late-Heian setsuwa, or short story, collection written in a mixed Japanese-Chinese style, the principle is recognized “that causatives of intransitive verbs are ‘wo causative sentences’ and causatives of transitive verbs are “ni causative sentences”, and he claims that “although in modern Japanese (it is claimed that) there are both ‘ni causative sentences’ and ‘wo causative sentences’ for the causatives of intransitive verbs, in classical Japanese the ‘ni causative’ was almost never seen” (Oda 2003: 30). Concerning the patterns for causatives, pattern I with the double reading of a single character continued to be used into the early Meiji period but, with the publication of “Survey regarding the teaching of kanbun” (Kanpō [Bureau reports 1912.3.29]), a standard was presented for kundoku. This report organized the kundoku methods, which had been entirely unconstrained until then, and, through provision of a fixed policy, became the yardstick for future kanbun-kundoku systems. Concerning the causative, the report gave the following pattern and since that time pattern II has become established for kanbun-kundoku, including school education.
26 Early modern kanbun and kanbun-kundoku
533
However, 使 and 教 should conform to the example below. 能
使
yoku
si
mu 2
wosite 枉者
直
magurerumono
1 naora
yoku magureru mono wosite naora-simu. possible bent person agt recover.cause ‘It is possible to cause one that is bent to recover.’ How, then, were these causative expressions used in modern period kanbun-kundoku style texts? I would like to examine them concentrating on how the causee is marked. First, we find the following description in Chamberlain (1886: 80): In causative constructions the name of the person who is made to perform the action is marked by the postposition wo shite (very rarely ni shite or wo);
That Y-wosite Z-simu is the basic pattern is thus confirmed. Below are some examples from the magazine Taiyō (NINJAL 2005). Makotoni zi bakarini kaisyaku suru kara truly character only interpretion do since sono kaisyaku ga matigatte that interpretation nom be.mistaken.ger Koosi wosite kusaba no kage ni Confucius agt leaves.of.grass gen shadow loc nakasimuru yoo na koto ni itaru. cause.to.cry kind.of adn that dat reach ‘Really, it’s because they only interpret the characters that the interpretation is mistaken – It’s enough to make Confucius cry in the shadows of the grass [turn over in his grave].’ (Tani Kanjō, Hei o tarashi, shoku o tarashi, tami wa kore o shin ni su 1895.03)
534
Fumitoshi Saito
Wagahai wa waga bunsyoogo wosite ima issoo we top our literary.language agt now more katugengo ni tikazukasimuru no hituyoo wo living.language to cause.to.near gen necessity acc kanzite yamazaru nari. Bunsyoogo wosite feel.ger not.stop be literary.language agt katugengo ni tikayorasimu to iwan living.language to cause.to.draw.near quot call.for yorimo, musiro katsugengo wosite more.than rather living.language agt bunsyoogo ni tikayorasimu to iwan literary.language to cause.to.draw.near quot call.for hoo, sinbungaku no ate ni toru beki alternative new.literature gen goal dat take ought hoosin naru besi. policy be ought ‘We ceaselessly feel a need to make our literary language these days ever closer to the spoken language. Rather than calling for the written language to draw closer to the spoken language, it would be a better policy to take as the goal of our new literature to call for the spoken language to draw closer to the literary language.’ (Ōnishi Hajime, Bungakujō no Shinjigyō. 1895.05) Sansen mondai wa kookei naikaku wosite kekkoo participate.in.war question top succeeding cabinet agt carry.out sesimuru mune wo motte su to. cause.do purport acc by do quot ‘(decided) to have the next cabinet deal with the question of going to war’ (1917.08 Nisshi [diary]) Sikasite koosyoo seikaku, ziri meiseki, hiroku investigation meticulous facts clear broadly saguri tooku tugi, tyoosya wosite sono yooryoo probe far.off succeed listener that essentials sirasimetari. cause.to.learn.pst ‘With meticulous investigation, clear facts, probing broadly, and receiving information from far off, we inform the listener of the essentials. (Matsumura Kaiseki, Meiji no Shinreikai [The World of the Meiji Spirit]. 1895.05)
26 Early modern kanbun and kanbun-kundoku
535
Ware ga yoo yori kanzi wo manabisi ga I nom childhood from kanji acc learn.pst gen gotokuni, hito mo yoo yori kanzi wo same.as people also childhood from kanji acc manabi-oreba, toosei no hito ni kanzi-haisi wo learn-be.prov today gen people agt kanji-abolition acc kekkoo sesimuru no muzukasiki wa, onozukara ni carry.out cause.do gen difficulty top itself dat zyunzite sassu beki. following realize ought ‘If people have been learning kanji as I learned kanji from childhood, the difficulty of causing the abolition of kanji to be carried out should be self-evident.’ (Miyake Setsurei, Kanji Ridōsetsu [An Argument for Making Kanji Advantageous]. 1895.08) Sunawati sono issetu wo yaya yooini namely that one.verse acc somewhat easily dokusya ni rikai sesimen gatameni reader agt understand cause.do in.order.to mazu korera no kensan ga nasaretaru first these gen diligent.study nom has.been.done rikkyakuten wo dokusya ni sirasimete standpoint acc reader dat cause.know.ger ‘In order to get the reader to understand that one verse relatively easily, first of all inform the reader of the standpoint on which these arduous studies have been performed.’ (S.T. Sei, Shoppenhawā-shi no Shina Shūkyō-ron [Schopenhauer’s China Religion Theory]. 1895.11) The examples in Taiyō are overwhelmingly of the Y-wosite Z-simu pattern, as shown above, but, though few, there are also some examples of the Y-ni Z-simu pattern of the Issai system with the causee marked with ni. Finally, I would like to look at how the causative expressions from kanbun-kundoku were used in Dutch Studies and in English Studies. Kaitai shinsho [Dissection new book], published in 1774, was a work that translated Dutch into kanbun and in Volume 4 Chapter 28 on muscles quite a few examples of causative expressions appear in the pattern (X-kin wa) Y-wosite Z-sesimu [(X-muscle TOP) Y-AGT Z-do.cause] ‘(X-muscle) makes Y Z.’ as in the following examples (Volume 4, leaf 18, Obverse).
536
Fumitoshi Saito
site wo 使 下顎 mu B,2 1 kagaku
上
ra
A aga
kagaku wosite agarasimu ‘causes the lower jaw to rise’ site
使 mu B,2
下顎
wo
1 kagaku
下
ra
A saga
kagaku wosite sagarasimu ‘causes the lower jaw to lower’ Regarding the causative expressions used in Kaitai shinsho, one peculiarity is that nearly all the subjects are non-animate. Concerning causative expressions with inanimate subjects, such as Noogyoo ga kuni wo hatten saseru [agriculture NOM nation ACC develop cause.to.do] ‘Agriculture causes a nation to develop.’ or Nani ga kanozyo wo soo saseta ka [What NOM her ACC thus cause.to.do Q] ‘What made her do that?’, according to Nihon bunpō daijiten (Matsumura 1971), “since they were not originally in Japanese, they came to be used under the influence of translations from Western languages”. As Kaitai shinsho is a medical text describing how various parts of the body function, it was probably inevitable that there would be many sentences with inanimate subjects, but, on the other hand, the question remains whether the structure of the original Dutch text sentences may have had an influence when translating (composing kanbun). Next, Yaku Oranda bungo [Translated Dutch written language] published in 1855 by ŌBA Sessai [1805–1873], influenced by the trend of the time for learning Dutch, writes a Japanese translation beside each word of a Dutch sentence and then, by rearranging the order of the Japanese glosses, interprets it as Japanese, following the kanbunkundoku formula (and further providing a colloquial translation in parentheses following the sentence). As shown in the following examples, sentences with ‘laten (laat)’ in the sentence use wosite as in warera wosite [we AGT] to mark the causee (Volume 2–3, leaf 4, Obverse). simeyo laat
warerawosite ons
kamiwo god
(warera wa yurusi no tameni we top grace gen for ‘Let us pray to God for his grace.’
tameni yurusino negawa om vergiffenis smeeken kami god
ni dat
negawau let.us.ask
yo) sfp
26 Early modern kanbun and kanbun-kundoku
537
There are also examples with wo, like the following example (Volume 2–2, leaf 10, Reverse). simeyo laat
karewo hem
ika gaan
(kare wo him acc ‘Let him go.’
ikasime cause.to.go
yo.) sfp
This method of studying Dutch, following the formula of kanbun-kundoku, was also carried on in English Studies. Uwiruson-shi Daini Riidoru Hitori Annai [Mr. Willson’s second little guide to independent reading] (Baba Eikyū, Hosoi Kikichi (trans). 1885) was a work that used the kanbun-kundoku formula, adding a Japanese gloss beside each English word, numbering the words, and showing the Japanese order with the numbers. Note that ‘make’ and ‘let’ are read doubly with site and simu, as in the following examples. What makes naniga site∠simuruka (1) (3) (8)
you think it is a boy? nanziwo kangae sorega aruto danzide (2) (7) (4) (6) (5)
nani ga nanzi wosite sore ga danzi dearu to what nom you agt that nom boy cop quot kangae-simuru ka think-cause q (186–6) He karewa (1) and sikasite (7)
let the boy ride him, site∠sime danziwo nora kareni (3) (6) (2) (5) (4) he did not run karega nasa zarisi hasiri (8) (10) (11) (9)
kare wa danzi wosite kare ni nora-sime, sikasite he top boy agt he on ride-cause and kare ga hasiri-nasa-zarisi. he nom run-do-not.pst (105–3) As observed above with causative expressions, the kanbun-kundoku methods changed in the early modern period and those usages were not only inherited by the kanbun-kundoku style sentences from the modern period on, but were also used in Dutch and English Studies materials.
538
Fumitoshi Saito
References Baba, Eikyū and Kikichi Hosoi (translators). 1885. Uwiruson-shi daini riidoru hitori annai [Mr. Willson’s second little guide to independent reading]. Tokyo: Zuiji Shobo. https://dl.ndl.go.jp/info:ndljp/pid/870670 (accessed 5 July 2021) Baienkai (eds). 1912. Baien zenshū [The complete works of Baien], vol. 2. Tokyo: Kodokan. Chamberlain, Basil Hall. 1886. A simplified grammar of the Japanese language: Modern written style. London: Trübner & Co. / Yokohama: Kelly and Walsh. Hio, Keizan. 1835. Kunten fukko [Restoration of kunten]. Publisher unknown. Imanaka, Kanji and Tatsuya Naramoto (eds.). 1974. Ogyū Sorai zenshū, dainikan [The complete works of Ogyū Sorai, vol. 2]. Tokyo: Kawade Shobō Shinsha. Kaneya, Osamu (annotator). 1963. Rongo [The Analects]. Tokyo: Iwanami Bunko. Kaji, Nobuyuki (translator and annotator). 2009. Rongo: Zōhoban [The Analects: Augmented edition]. Tokyo: Kodansha Gakujutsu Bunko. Matsumura, Akira (ed.). 1971. Nihon bunpō daijiten [Dictionary of Japanese Grammar]. Tokyo: Meiji Shoin. NINJAL (ed.). 2005. Taiyō corpus: Zasshi “Taiyō” Nihongo deetabeesu [Japanese language database from the magazine Taiyō]. Tokyo: Hakubunkan Shinsha Publishers. Ōba Sessai. 1855. Yaku Oranda bungo [Translated Dutch written language]. Publisher unknown. Oda, Masaru. 2003. Kotenbun ni okeru shiekibun, ukemibun no kakuhyōji: Konjaku monogatarishū o shiryō toshite [Case marking in causative and passive sentences in classical Japanese: Materials from the Konjaku monogatarishū]. Gifu Seitoku Gakuin Daigaku Kiyō, Gaikokugo Gakubu Hen 43. 23–32. Ōtsubo, Heiji. 1981. Heian jidai ni okeru kuntengo no bunpō [The grammar of kunten language in the Heian period]. Tokyo: Kazama Shobo. Ōtsuki, Fumihiko. 1897. Kō Nihon bunten: Bekki [Extended Japanese grammar: Addendum]. Tokyo: Yoshikawa Kōbunkan. Ōwada, Takeki. 1893. Ōyō kanbun-gaku [Applied Chinese studies]. Tokyo: Hakubunkan. https://dl.ndl.go.jp/ info:ndljp/pid/868698 (accessed 5 July 2021) Saitō, Fumitoshi. 2011. Kanbun-kundoku to kinsei Nihongo no keisei [Kanbun-kundoku and the development of early modern Japanese]. Tokyo: Bensei Shuppan.
Valerio Luigi Alberizzi
27 The influence of kanbun-kundoku vocabulary on the Japanese language 1 Vernacular readings and the Japanese language Kanbun-kundoku 漢文訓読 played an important role in the incorporation of Sinitic elements into written and ultimately into spoken Japanese. In kanbun-kundoku small phonogram glosses, or kanaten 仮名点, were added to the right or left side of the Chinese characters in the original text to suggest its Japanese reading. The technique of glossing the inflectional morphemes associated with Japanese verbs and adjectives allowed reconciliation of the differences between Chinese, an isolating language, and Japanese, an agglutinative one. Dots, keys, and other reading marks, commonly known as kunten 訓点, were added to the lower side of a character to indicate the different syntactic order that resulted from the conversion from an SVO to an SOV language. In addition to black ink marks, the writers often used colored pigments and even inscribed the surface of the manuscript with a stylus to make the rendition into Japanese as clear and unmistakable as possible. The colors of these glosses changed over time, enabling scholars to date the annotation of the texts. Glosses tended to be white in the 9th and 10th centuries; light vermilion between the 9th and the 10th century; deep vermillion from the 10th to the 12th century; and black from the 10th century onward. Stylus glossing appears in a variety of texts ranging from the 9th to the 13th century. Other colors were used between the 10th and 12th centuries in conjunction with vermillion glosses. Other types of glosses that offer hints about the time when a manuscript was annotated include (w)okototen ヲコト点, kanaten and kaeriten 返点.
1.1 (W)okototen (W)okototen are diacritic marks placed around the perimeter of a character or within its area to indicate postpositions or functional words. They developed in the 8th Century when, in the absence of an adequate pure phonographic system of writing, they offered a quick and small-sized form of notation. NAKADA Norio (1915–2010) has proposed a subdivision of these annotation systems into eight groups based on the position of the verbal gerund suffix -te, of the dative/ locative postposition ni, of the accusative postposition wo, and of the topic marking Acknowledgments: This paper is the result of a broader research project conducted under a JSPS KAKENHI Grant-in-Aid for Scientific Research (C) 26370552. https://doi.org/10.1515/9781614512851-028
540
Valerio Luigi Alberizzi
postposition wa, see Figure 1. These are usually placed in the corners of the box (tsubo 壺・ 坪 or ‘jar’) that ideally marks the perimeter of a Chinese character (Nakada 1954). The glosses belonging to groups one to four appeared in documents since the 9th century. The ones from the remaining groups gradually appeared after the 10th century. By the 11th century, however, only about a dozen of these systems were still in use.
Figure 1: The eight (w)okototen main groups (from Tsukishima 1991: 32).
Contrary to the belief that they consist only of one or two boxes, these glossing schemes could be remarkably complex. For example, the ni-shi-ha-ka glosses (Nishi hakaten 西墓 点) of the late 9th century-early 10th century showcase as many as 31 boxes and 230 types of glosses in the form of dots and other marks.
1.2 Kanaten As mentioned above, kanaten are phonogram glosses placed next to a character to indicate its Japanese reading. In the early 9th century, Chinese characters were used as phonograms without any kind of standardization. A case in point is the Yāngjuémóluó jīng 央掘魔羅経 (Japanese: Ōkutsu marakyō) manuscript of the Kōnin era (810–824). Early kana glosses appeared in the middle and late 9th century; they were created from the cursivization of Chinese characters or parts thereof. Rather than being treated as real phonograms, in these early stages these glosses were likely adopted by individual readers and their small circles for personal use. The lack of a common standard is evidenced by several variations in the orthography and the tendency not to rely on previous stylistic choices. A survey of ninth-century texts also indicates the use, within the same document, of two or more graphic variants of the same syllable, further attesting to such a formational stage. In texts annotated after the turn of the 10th century, the graphemes representing the sounds of Japanese tend to match despite the fact that glossators and historical periods
27 The influence of kanbun-kundoku vocabulary on the Japanese language
541
differ. This demonstrates a turn away from the ninth-century tradition of individual choices and a trend toward standardization likely made possible by the transmission from teacher to student within the same school. This trend was completed by the 11th century, when the same glosses came to be used in all documents. See Figure 2.
Figure 2: Phonogram glosses from 9th to 11th century (from Kobayashi 1979: 102–112).
1.3 Kaeriten Kaeriten, or inversion glosses, were used to indicate a change in the word order when rendering a Chinese text into Japanese. In modern kanbun-kundoku there is now a two-way relationship between the reading of a text and inversion glosses recorded accurately at each passage where their presence is necessary. This is very different from what one sees in manuscripts, where the drafting of glosses appears to be arbitrary. Inversion glosses can be divided into five categories according to their shape: symbols, cardinal numbers, punctuation marks, Chinese characters, and ‘wild goose’ marks (Ōtsubo 1992; Kobayashi 1974), see Figure 3. Symbols are one of the oldest forms of inversion glosses. Widely used during the 9th century in documents such as the Tōdaiji fujumonkō 東大寺諷誦文稿 (Manuscript of the Todaiji temple prayers, early 9th c.), they were later abandoned probably because they stained the text.
542
Valerio Luigi Alberizzi
Cardinal numbers as inversion glosses appeared since the late 8th century. They were adopted in the Xù huáyánjīnglüè shū kān dìng jì 続華厳経略疏刊定記 (Japanese: Zokukegonkyō ryakuso kanjōki), a commentary to the Avataṃsaka sūtra (Japanese: Kegonkyō 華厳経) and the oldest extant glossed manuscript in Japan. Therein, they mark syntactic inversions in about thirty passages (Tsukimoto 2000). They were also commonly used in the annotation of Chinese classics during the 10th–12th centuries in lieu of punctuation marks. Punctuation marks were used in Buddhist manuscripts from the 9th century. In the glossing of Chinese classics, they played a dual function: if placed on the lower left corner, they functioned as inversion glosses and markers of the verbal gerund suffix –te; if placed on the lower right corner, as inversion glosses and markers of the conditional suffix -ba. Chinese characters were used in the interpretation of long and complex sentences frequently with the support of cardinal numbers. In the 9th century only the characters jō 上 (up) and ge 下 (down) were used. The character chū 中 (middle) was adopted only in the late 10th century. The other two series kō-otsu-hei 甲乙丙 (A-B-C) and ten-chi-nin 天地人 (Heaven-earth-man) date back to the 11th century. ‘Wild goose’ marks – named after their shape resembling a flying goose – are the ancestors of the re ten レ点, the sign adopted in modern kanbun-kundoku to indicate an inversion in the order of two consecutive characters. They appeared in the second half of the 12th century and were positioned centrally between the two characters to which they referred. The gloss used in 12th century manuscripts is similar to modern the katakana re: the first stroke is vertical, the second, varying in length, is traced diagonally from left to right. In the 13th century this gloss maintained its central position but changed shape, now resembling spread wings. In the 14th century the sign was gradually moved to the left and its first stroke once again arranged vertically and reduced in size.
Figure 3: The five categories of inversion glosses (from Kobayashi 1974: 89–91).
27 The influence of kanbun-kundoku vocabulary on the Japanese language
543
In the pages that follow, I will a) provide a review of existing scholarship on the vocabulary of glossed texts (kunten goi) and on their taxonomies in order to expand our understanding of the field and to stimulate future inquiries (§ 2); b) outline the main features of kunten glossed-text vocabulary and its influence on the Japanese language (§ 3–4); c) survey textual evidence to show the role played by vernacular readings of Chinese texts in defining the formation of a hybrid form of Sino-Japanese writing (§ 5).
2 Studies on kanbun-kundoku vocabulary: a brief history The origins of modern studies on kanbun-kundoku can be traced back to 1903 and the activities of the Kokugo Chōsa Iinkai, the Japanese Language Survey Committee. ŌYA Tōru (1850–1928), an auxiliary member of the Committee, was the first to carry on a detailed examination of a large number of manuscripts, mostly held by the Shōsōin – the treasure house of the Tōdaiji temple in Nara. His goal was to collect a vast number of kana examples to be used for the standardization of the hiragana and katakana scripts. Ōya’s pioneer works, chief among which one must cite Kanazukai oyobi kana jitai enkaku shiryō (1909), shed light on a new field of research and revealed its potentialities. Kanazukai oyobi kana jitai enkaku shiryō offers a description of the kana adopted in about fifty manuscripts from the 8th to the 15th centuries. Each chart provides a short bibliographical explanation, a table collecting the various kana shapes and one about their uses with examples taken from the glosses at the margin of the texts. A quick skim through the pages of the book thus offers the reader a clear picture of the changes kana underwent through time. Before Ōya’s death, all the data and materials collected during his studies were turned over to KASUGA Masaji (1878–1962). In 1942 Kasuga published Saidaiji-bon Konkōmyō saishōō kyō koten no kokugogakuteki kenkyū, a monumental work in two volumes. The first includes the photographic reproduction of the ten-scroll manuscript with the Japanese rendition of the original Chinese text as displayed by the glosses. The second volume presents a detailed linguistic survey of the language as suggested in the glosses, from phonetics to grammar. Modern studies on kanbun-kundoku language and vocabulary can be traced back to the 1935 publication of Kanbun no kundoku ni yorite tsutaeraretaru gohō by YAMADA Yoshio (1873–1958). While examining the native literary prose and poetry of works such as Genji monogatari or Kokin wakashū Yamada noticed numerous examples of words and expressions that were not normally used in written standard Middle Japanese, or wabun 和文.
544
Valerio Luigi Alberizzi
Kanbun no kundoku ni yorite tsutaeraretaru gohō consist largely of a list of lexical and grammatical items – in modern Japanese – that can be directly linked to the practice of kundoku: (1) gotosi ‘as, like’, kaku no gotosi ‘such, in this way’, iwaku ‘according to, (what one) says. . .’, omoeraku ‘what one considered/thought’, negawaku wa ‘one hopes/beg that. . ., please. . .’, osimuraku wa ‘one regrets that. . .’, iwayuru ‘the so-called. . .’, nan nan to su ‘to be on the verge of doing/becoming something’, kaennan (iza) ‘return, go back’, nan surezo ‘why’, nakariseba/nakasseba ‘even if it’s not. . .’, sikari ‘in this/ that way, yes’, sikaredomo ‘however, still’, sikōsite ‘then, after that’, (w)o site. . . simu (causative), besi ‘must do, should do’, bekarazu ‘must not do’, subekaraku ‘(it is proper that one) should do, ought to do’, ani ‘why, alas’, imada ‘as yet, so far’, katsu ‘besides, at the same time, for a (little) while’, katute ‘once, not at all, not. . .yet’, kedasi ‘probably, if. . .’, sude ni ‘already’, sunawati ‘namely, at once’, musiro ‘rather (than)’, mosi ‘by any chance’, mosikuwa ‘or, otherwise’, ai- ‘mutually’, aete ‘not always. . .’, itarite ‘very’, kiwamete ‘extremely’, subete ‘completely, roughly’, hatasite ‘as was expected’, yorite ‘therefore, because of’, saiwai ni ‘luckily’, sikiri ni ‘repeatedly’, midari ni ‘at random’, aruiwa ‘(either. . .) or, probably’, oyobi ‘and, then’, narabi ni ‘and, as well as’, (ni) oite ‘at, in, on, during’, tame ni ‘for, because of, as a result of’, yue ni/yuen ‘for, due to’, motte ‘by means of, for, as’, tokoro ‘place’ (adnominal), iedomo ‘although, however’, hossu ‘to wish, be about to’, gaen sezu ‘to consent’, atawazu ‘be impossible’, nomi ‘only, nothing but’, iwanya ‘let alone, not to mention’, kore ari ‘there is’, kore nasi ‘there is not’, kore ‘this’ (reflexive). Unfortunately, Yamada used glossed editions of Chinese classics – kanseki – printed in the Edo period as primary sources and did not take into account manuscripts from the 10th to the 12th centuries, a time when kundoku was at its peak. While his work demonstrates that some vernacular reading features normally identified with the Heian period were in all likelihood carried over across the centuries, it still lacks empirical evidence based on direct textual investigation. Diachronic studies on the vocabulary that originated from Early Middle Japanese kunten glossed manuscripts began in the mid-20th century with the 1952 publication of Kunten shiryō to kuntengo no kenkyū by ENDŌ Yoshimoto (1905–1992). Endō (1952) was the first to draw a comparison between words deployed in the native written language and words found in glossed materials. He concluded that discrepancies were attributable to the gendered differences between a language conveyed by a female-centered register – i.e. wabun – and a male-centered one – i.e. kanbun-kundoku. Following in Endō’s footsteps, in 1963 TSUKISHIMA Hiroshi (1925–2011), one of the leading scholars of the language of kunten materials, identified two groups of words specific to the written language form and organized them into a systematic taxonomy (Tsukishima 1963). Tsukishima conducted a detailed survey of Genji monogatari and of Dà Cíēnsì Sānzàng fǎshī zhuàn 大慈恩寺三蔵法師伝 (Japanese: Daijionji San-
27 The influence of kanbun-kundoku vocabulary on the Japanese language
545
zōhōshiden), two works that epitomize native and kanbun-kundoku texts, respectively. The glosses in Daijionji Sanzōhōshiden date back to the 11th and 12th centuries. By comparing the lexicons of these two works, Tsukishima concluded that the vocabulary of kunten materials and the vocabulary of Early Middle Japanese documents can each be categorized into two sub-groups. The first consists of ‘characteristic language’, or tokuyūgo, which indicates expressions belonging exclusively to either a Sinicized or a native written form. The second subgroup, ‘deployable language’, or yōgo, includes expressions that are usually associated with one of the two varieties but can also be deployed in different contexts. Tsukishima’s taxonomy is more reliable than Endo’s not only because it is based on a larger amount of data gathered from the comparison of two full-length works written around the same time, but also because it posits that differences in vocabulary and grammar were determined not by gender but by the choice of a specific form of expression.1 Hence, lexical and syntactical choices depended on the field, purpose, and tenor of the message.
3 Kundoku vocabulary in context: the case of Ki no Tsurayuki Ki no Tsurayuki’s (868–945) preface to the Kokin wakashū (905) provides a case in point. Entirely written in hiragana, this text is often referred to as the first complete document in pure Early Middle Japanese, the one that marks the beginning of Japanese poetic criticism in the literary circles of the time as distinct from the otherwise prominent Chinese poetic forms. However, Japanese scholars have demonstrated that the preface is not an original composition but is in fact an adaptation inspired by the foreword to the Classic of songs – Shijing 詩経 (Japanese: Shikyō). A closer analysis does indeed reveal the presence of a vocabulary belonging to Tsukishima’s first subgroup. (2) こゝに、 いにしへのことをも、哥のこゝろをも、 しれる人、 わづかにひとり、 ふたり也き。 し かあれど、 これかれ、 えたるところ、 えぬところ、 たがひになむある。 koko ni inisipe no koto wo mo uta no kokoro wo mo sireru pito waduka ni pitori putari nariki sika aredo korekare etaru tokoro enu tokoro tagapi ni namu aru ‘After that there were one or two poets who knew the ancient songs and understood the heart of poetry. However, each had strengths and weaknesses.’ (Saeki 1958: 99; translation by Rodd 1996: 43)
1 Despite KOBAYASHI Yoshinori’s skepticism on the reliability of Daijionji Sanzōhōshiden for a comparison with Genji monogatari Tsukishima’s work is still considered the foundational reference work and an invaluable source for researchers. See Kobayashi 1966.
546
Valerio Luigi Alberizzi
(3) そのほかに、 ちかき世に、 その名きこえたる人は、 すなはち僧正遍昭は、哥のさまはえ たれども、 まことすくなし。 たとへばゑにかけるをむなを見ていたづらに心をうごかす がごとし。 sono poka ni tikaki yo ni sono na kikoetaru pito pa sunapati sauzyau Penzyau pa uta no sama pa etaredomo makoto sukunasi tatopeba we ni kakeru omuna wo mite itadura ni kokoro wo ugokasu ga gotosi. ‘Among the others, one of the best known of recent times was Archibishop Henjō, whose style is good but who lacks sincerity. His poetry is like a painting of a woman which stirs one’s heart in vain.’ (Saeki 1958: 100; translation by Rodd 1996: 43) (4) 大伴のくろぬしは、 そのさまいやし。 いはゞたきゞおへる山人の花のかげにやすめるが ごとし。 Opotomo no Kuronusi pa sono sama iyasi. Ipaba takigi operu yama-pito no pana no kage ni yasumeru ga gotosi. ‘Ōtomo no Kuronushi’s songs are rustic in form; they are like a mountaineer with a bundle of firewood on his back resting in the shade of the blossoms.’ (Saeki 1958: 101; translation by Rodd 1996: 46) (5) たとひときうつりことさり、 たのしびかなしびゆきかふともこのうたのもじあるをや。 tatopi toki uturi koto sari tanosibi kanasibi yuki-kapu tomo kono uta no mozi aru wo ya. ‘Times may change, joy and sorrow come and go, but the words of these poems are eternal.’ (Saeki 1958: 103; translation by Rodd 1996: 47) Adverbs, connectives, and auxiliaries such as tagapi ni, tatopeba, ga gotosi, and tatopi are generally absent from court literature of the Heian period, but are often used in the glossed readings of Chinese texts. While it is difficult to prove that their presence in Tsurayuki’s preface is directly linked to vernacular readings of The Classic of Songs, one may argue that the author was likely familiar with such a vocabulary, for as a highly educated man he was versed in the Chinese Classics and knew how to render them into Japanese. The same introduction features other unique traits; for instance, vocabulary from the kundoku stratum that reveal the familiarity of the author with official documents that used a variety uncommon in everyday parlance. Tsurayuki is also the author of the Tosa nikki (935), a pioneering work written in a genuine native style that uses mainly hiragana. Even this work, however, is in many ways different from the great classics of court literature and shows a strong influence of Sinicized varieties both in vocabulary and in contents.
27 The influence of kanbun-kundoku vocabulary on the Japanese language
547
おのれしさけをくらひつれば、 はやく (6) といふあひだにかぢとりもののあはれもしらで、 いなんとて、 「しほみちぬ。 かぜもふきぬべし。」 とさわげば、 ふねにのりなんとす。 to ipu apida ni kaditori mono no apare mo shirade onore sisake wo kurapitureba payaku inan tote “sipo mitinu kaze mo pukinu besi” to sawageba pune ni norinan to su. ‘. . . while he was reciting the poem, the captain of the boat, a rude man who did not know the aware of things, having being paid enough to drink as much as he wanted, was anxious to leave immediately. “The tide has risen! The wind is blowing!”, he shouted, going on board.’ (Suzuki 1957: 30) (7) こゝろざしあるにゝたり。 kokorozasi aru ni nitari. ‘He seemed to be a well-mannered and kind man.’ (Suzuki 1957: 31) (8) あるひとのこのわらはなる、 ひそかにいふ aru pito no ko no warapa naru pisoka ni ipu. ‘The child of one of the passengers bashfully said. . .’ (Suzuki 1957: 33) (9) そもそもいかゞよんだるといぶかしがりてとふ。 somosomo ikaga yondaru to ibukasigarite topu. ‘“Tell me, then, what was your poem going to be?”, asked a person eager to hear [the child’s poem].’ (Suzuki 1957: 34) Much in the same way as the preface to the Kokinshū, we see here a number of adverbs and connectives that were not generally used in 10th- and 11th-century Japanese literary prose. Instead of tagapi ni or pisoka ni, a text in Japanese would use such expressions as katami ni, sinobiyaka ni, or sinobite. While all these expressions have the same semantic value, the last three are better contextualized within a native written form. The choice cannot be justified by citing the author’s familiarity with orthodox Chinese. At the same time, Tosa nikki features syntactic patterns that pertain to the native variety of Early Middle Japanese, for example kakari-musubi – a construction where a particle (kakari) marks a constituent and binds the main predicate to an adnominal or exclamatory form (musubi). A comparison with the diaries of noblemen and court officials written in waka kanbun 和化漢文 (Japanized written Chinese) has led some scholars to postulate that Tsurayuki attempted to realize a kana version of this written style, a variety with which he must have been well acquainted (Tsukishima 1981). This theory would explain the presence of elements of such different origins in the same text. After being properly polished, this prose paved the way for the refined, “pure” native written language of later ages.
548
Valerio Luigi Alberizzi
4 The features of kanbun-kundoku vocabulary Early research argued that the language represented in kunten glosses was more colloquial than the language of contemporary literary texts. Unfortunately, it is difficult, within the text, to distinguish these elements from those pertaining to the written language. While it is undeniable that kanbun-kundoku was created and developed as an auxiliary interpretative method for a script and that its main linguistic features therefore pertain to written Japanese, its function as a form of impromptu notetaking seems likely to have introduced colloquial elements. Kunten glosses have a dual nature. First, they are words used to explain, in Japanese, an array of complex philosophical ideas, like those expressed in sutras or in Chinese classics. Second, they are also part of a refined lexicon used by a group of skilled scholars, lay or religious, to overcome the difficulties imposed by a different word order and by a lexical interpretation of Chinese characters and compounds. In this respect, kundoku vocabulary is different from the oral idiom of everyday life. The frequent inclusion of colophons is also an important feature of kunten glossed texts. Placed at the end of the volume or scroll, they often indicate the date and period when the manuscripts were copied. The language deployed in kunten materials can be roughly divided into two periods, namely, from the 9th to the 10th century and from the 11th to the 12th. The first phase is characterized by the presence of putative Old Japanese elements, such as the controversial nominative particle i. These elements did not carry over into the next generation of texts. By contrast, during the second phase kundoku underwent a renovation and a consolidation process. The lexical and grammatical choices of the texts became orthodoxy, were transmitted as such, and were arranged into a systematic code of glossing practices in the Edo period. Hence, no changes were made to the vocabulary of glossed texts after the end of the Kamakura period. For example, the desiderative auxiliary tasi appeared after the Kamakura period, but evidence shows that it was never adopted as a reading in kunten glossed texts. Kuntengo elude a systematic taxonomy; one’s best bet is to divide them by grammatical categories. Word formation is also an issue when it comes to kanbun-kundoku vocabulary. Its main features can be summarized as follows: 1) Nominalization of the infinitive verb form The use of the verbal infinitive as a noun was widespread in Old Japanese. However, the glosses of Early and Late Middle Japanese period manuscripts include a significant number of words that were hardly ever used in the native literary prose, words such as itapari ‘suffering’, itami ‘pain’, itupari ‘falsehood’, ipi ‘what has been said’, utukusibi ‘love, affection’, uruopi ‘moisture, water’, osori / osore ‘fear’, omopi ‘thought, feelings’, kiki ‘what has been heard’, konomi ‘preference’, satori ‘understanding, enlightenment’, samatage ‘hindrance’, seme ‘blame’, sosiri ‘calumny, blame’, sonape ‘provision’, nayami ‘trouble, anguish’, negapi ‘wish, desire’, pedate ‘gap, distance’, and mitibiki ‘guidance’.
27 The influence of kanbun-kundoku vocabulary on the Japanese language
549
The infinitive of passives in -(a)re- is also used in this function to derive result nominals such as torapare ‘captive’, ipare ‘what has been said’. 2) Compound verbs In classical Chinese, a language with remarkable isolating characteristics, words take on different grammatical functions (verb, noun, etc.) according to their position within the sentence. Most of the characters that the Japanese read as nouns were in fact verbs in classical Chinese. To offset such discrepancy the Japanese had to develop original strategies for the glossed reading of a text. The most common strategy was the verbalization of nouns: aka ‘dirt’ → akaduku ‘to get dirty’, ase ‘sweat’ → aseduku, aseayu ‘to sweat’, kata ‘model, form’ → katadoru ‘to model, to imitate’, kata ‘shoulder’ → katanugu ‘to undress’, siri ‘the behind’ → sirizoku ‘to step backward’, kusa ‘grass’ → kusagiru, kusadoru ‘to weed’, yubi ‘finger’ → yubisasu ‘to point’, muti ‘whip’ → mutiutu ‘to whip’, waki ‘the side’ → wakipasamu ‘to hold under one’s side’. Most of these verbs, first attested in a kunten gloss, gained currency over the centuries and are now firmly rooted in contemporary Japanese.
Compound verbs were also formed by agglutinating the infinitive of a verb with the conclusive form of another: utiketu (utu ‘to hit’ + ketu ‘erase’) ‘to extinguish, wipe out’, pikiwiru (piku ‘to pull’ + wiru ‘to bring with’) ‘to lead’, puminiziru (pumu ‘to step on’ + niziru ‘to walk slowly’) ‘to crush something under one’s feet’, motiwiru (motu ‘to have’ + wiru ‘to bring with’) ‘to make use of, to employ’, moteasobu (motu ‘to have’ + asobu ‘to play with’) ‘to amuse oneself with’, etc.
550
Valerio Luigi Alberizzi
3) Uses of su One of the features of kundoku is a frequent use of the verb su – ‘to do’ – preceded by a noun or an element with similar nominal features, not at all dissimilar from the verbs formed by a noun or an adjective followed by suru in the modern language. The following classification is the one Tsukishima proposed in 1963 (Tsukishima 1963: 563) and later revised in 2007 (Tsukishima 2007: 33–37). Only the expressions described under (g.) are found exclusively in kunten glossed texts. Uses from (a.) to (f.) are also attested in Old Japanese native texts (see Frellesvig 2013) but present a lexicon not easily found in Old and Middle Japanese native texts. None of the examples presented in (a.) occurs in Old and Middle Japanese native literature as a verbal form. a. noun + su asita su ‘to get light’, kutubami su ‘to put a bit (in a horse’s mouth)’, kumi su ‘to group’, tumi su ‘to commit a sin’, nukiasi su ‘to walk stealthily’, pideri su ‘to dry up (weather)’, makura su ‘to rest one’s head on something’, marone su ‘to sleep with one’s clothes on’, moda su ‘to hold one’s tongue’, monogatari su ‘to tell a tale’, yamapi su ‘to be struck down by illness’, yopapi su ‘to live long’, etc. b. verb infinitive + su pori su ‘to wish, to desire’, later transcribed in manuscripts of the Late Middle Japanese as its geminate variant possu or also po su and pon su, kakage su ‘to fly (a flag)’, sasurape su ‘to wander’, etc. c. adjective infinitive + su Often annotated with the euphonic variant: -ku su → -u su (u-onbin ウ音便). Iyasiu su ‘to despise’, kuwasiu su ‘to investigate, to go into’, tadasiku su ‘to correct’, pitosiku su ‘to equalize’, yoku su ‘to improve’, etc. d. noun or phrase + to su koto to su ‘to do something in particular, to dedicate to’, sono to su ‘to make one’s space’, nuki to su ‘to make a pillar of something’, nan-namu to su (euphonic variant of nari-namu to su) ‘to be on the verge of becoming something, or to be on the verge of doing something’ e. noun or restrictive particle + ni su Ni su is often recorded by means of its euphonic variant -n zu: tame ni su ‘to have a reason (a purpose) to do something’, tumabiraka ni su ‘to ascertain, to clarify’,
27 The influence of kanbun-kundoku vocabulary on the Japanese language
f. g.
551
tubusa ni su ‘to give in detail’, naigasiro ni su ‘to make light of, to neglect’, tomo ni su ‘to share, to partake’, pitotsu ni su ‘to bring together’, mopara ni su ‘to devote oneself wholly to, to monopolize’, posiki mama ni su ‘to have one’s own way, to do as much as one likes’, etc. instrumental / conjunctional particle te + su The most frequent expression is (ni) oite su ‘to act / behave according to something’. adjectival root + suffix mi + su The use of an adjectival root followed by the nominalization suffix mi is documented in Old Japanese prose and poetry. In the Early Middle period, however, this use disappeared completely from wabun prose and was only scarcely deployed in waka poetry. The same structure occasionally appears in documents with glosses at the end of the apodosis to convey a factual conditional sentence. From the Early Heian period, however, the glossators of kunten materials began affixing the verb su after mi, creating a pattern hitherto unseen in native Old Japanese or Early Middle Japanese prose and poetry. While its function is nearly identical to the one of the infinitive of an adjective followed by su, it appears as if this particular pattern was used only in association with a limited number of adjectives such as sa-si ‘narrow, small’, na-si ‘not’, and yo-si ‘good’ when they were meant to be read as verbs in the Chinese original. When the adjectival root was monosyllabic, as in the aforementioned cases, euphonic changes usually did not occur. When the root consisted of two or more syllables, however, sound changes were properly annotated in the glosses with the nasal /n/ transcribed as mu: ama-si ‘sweet’ → ama + mi + su → amami su → amanzu ‘to sweeten’, uto-si ‘distant, unfamiliar with’ → uto + mi + su → utomi su → utonzu ‘to be unfamiliar with’, omo-si ‘heavy, important’ → omo + mi + su → omomi su → omonzu ‘to weight, to be important’, karo-si ‘light, not important’ → karo + mi + su → karomi su → karonzu ‘to be light, ‘to be unimportant’, etc.2
4) Noun + napu Some nouns were verbalized by adding the suffix napu or its euphonic variant nau: ubenapu ‘to submit, to obey’, uranapu ‘to divine the future’, tuminapu ‘to put someone to death’, tomonapu ‘to accompany’, etc. 5) Adjectival root + kemu The suffix kemu likely originated from the agglutination of the Old Japanese adjectival morpheme kye with the conjectural auxiliary -(a)m-. In kunten materials conjecture was usually expressed by using the suffix karamu with an adjectival root. However, kemu was often used with a limited number of adjectives such as ayapu-si ‘dangerous’
2 According to Kasuga’s study of the Konkōmyō saishōō kyō, during the Old Japanese period euphonic changes did not occur even in case of a root with more than one syllable. See Kasuga 1985.
552
Valerio Luigi Alberizzi
→ ayapukemu, topo-si ‘far’ → topokemu, na-si ‘not’ → nakemu, yo-si ‘good’ → yokemu, yasu-si ‘calm’ → yasukemu, and the auxiliary be-si → bekemu. Bekemu is the only form transmitted through the centuries; over time it established a fixed correspondence with the adverbs ani and idukunzo according to the emphatic pattern ani / idukunzo . . . bekemu ya. 6) Kari declension of adjectives (kari katsuyō) Kunten materials make extensive use of the so-called kari declension of adjectives for the negative imperative form, which is frequently used as a reading associated with the characters 無, 莫 and 勿: naku are → nakare. Other kari declension forms such as the infinitive and the conjunctive are used occasionally; no examples have been found yet for the conclusive and provisional forms. 7) Noun + dukara The adverbialization of nouns by means of the suffix dukara is another characteristic of vernacular readings which still exists in the modern language: onodukara ‘naturally, by itself’, midukara ‘personally, for oneself’, tedukara ‘in person, personally’, medukara ‘with one’s eyes’.
4.1 Old Japanese uses in kundoku vocabulary The birth of kundoku in Japan is generally believed to have occurred in the 6th century. It was not until the end of the 8th century, however, that the results of the reading process would be inscribed within the original Chinese text and left permanently therein for the reader to see. The limited number of sources makes it difficult to establish with any degree of certainty the characteristics of vernacular readings and the degree to which they influenced Old Japanese. However, as Yamada (1935), Tsukishima (1963) and others have pointed out, it is possible to single out the presence of Old Japanese in kundoku vocabulary that was later transmitted to Heian period texts (see also Frellesvig 2010: 270–274). 1) Ku nominalization (ku gohō) In the 8th century the ending ku was commonly used to nominalize inflective parts of speech such as verbs and adjectives. During the Heian period this nominal form disappeared from native prose and poetry, although it was still deployed in the vernacular reading of Chinese texts. As such, it had two different functions: a. Introducing a quotation or direct speech This is arguably one of the most widely studied cases pertaining to the syntactic influence of Classical Chinese onto Japanese through kanbun-kundoku. It is epitomized by the form ipaku, which was generally deployed to introduce a quotation or direct speech. Also well known are the cases of the adnominal form of verbs such as ipu, miru, or the infinitive form of adjectives such as yo-si agglutinated
27 The influence of kanbun-kundoku vocabulary on the Japanese language
b.
553
respectively to raku and ku: ipu + raku → ipuraku → ipaku ‘what one said’, miru + raku → miruraku → miraku ‘what one saw’, and yosi → yoke + ku → yokeku ‘what is good’. In the Heian period ku and raku became fossilized forms and started being treated as pure suffixes for verbs and adjectives. Some manuscripts show traces of the adnominal form of a verb + raku transcribed without any indication of a sound change: ipu + raku → ipuraku, kataru + raku → kataruraku. Around the mid-Heian period, expressions ending in ku lost their nominal function once and for all and became specialized markers to introduce direct speech: ipamaku ‘what one is trying to saying’, iperaku ‘what one said’, omopaku ‘what one thought’, omoperaku ‘what one considered’, etc. When followed by wa (< pa), forms ending in ku were used to convey a desiderative or dubitative meaning; they involved the adoption of a fixed grammatical pattern that established a binding agreement with the predicate of the sentence which they introduced – usually an imperative or prohibitive form. These constructions included: utagapuraku pa ‘one doubts (that)’, negapaku pa ‘one hopes (that)’, nikumuraku pa ‘one hates (that)’, nozomaku pa ‘one wishes (that)’, etc. When the same expressions were used to introduce a quotation, the conclusive form of the verb was placed at the beginning of the sentence without the ku ending (Ōtsubo 1981).3 Wrapping up an empathized statement in association with the adverb nomi Once expressions ending in ku lost their nominalizing use, they were also used with the restrictive particle nomi, which means ‘only’ or ‘just’ and is the rendition of the Chinese 耳 and 而已, to wrap up a sentence with a strong empathic nuance: siru ‘to know’ → siruraku nomi, taru ‘to be sufficient’ → taruraku nomi, etc. (Kobayashi 1952).
2) Old Japanese passives yu / rayu The passive auxiliaries yu and rayu likely originated in or before the early 8th century. During the Nara period they were already used in tandem with the couplet ru / raru, a couplet which would end up replacing them at the turn of the Heian period. Various examples dating back to the 9th century can be found in kunten glossed texts, but by and large their use as auxiliaries was discontinued in the early 9th century. Traces of their presence remain in two adnominals that survived in Modern Japanese: ara (infinitive of the verb ari) + yuru (adnominal of yu) → arayuru ‘all, everything’, vernacular reading associated to the Chinese 所有, ipa (stem of the verb ipu) + yuru (adnominal of yu) → ipayuru (ModJ iwayuru) ‘the so-called. . .’ for rendering the Chinese 所謂 or 所為. The glosses annotated next to 所有 and 所謂 in some 11th century manuscripts demonstrate that both arayuru and ipayuru were perceived more as single words than as composite adnominals formed by the association of a passive auxiliary with a verbal stem, as suggested by the Chinese syntactical structure. In a limited number of cases 所 is read
3 The most frequently used verbs with this function were negau ‘to hope’ and kou ‘to ask, to request’.
554
Valerio Luigi Alberizzi
separately as yuru, but by and large the two characters are either read as a single word or the entire reading is associated only to the second character of the couplet – i.e. 謂 or 為. 3) Adverbs The language of kanbun-kundoku is explicit and expository in character. Since the introduction of Chinese to Japan, the Japanese have had to come to terms with the interpretation of Chinese function words that featured a fixed concordance with various sentence-ending particles. This led to the extensive use, in vernacular readings, of adverbs placed at the beginning of a sentence that usually involved the adoption of a fixed grammatical pattern. For example, ani 豈 introduced a rhetorical question and required the interrogative particle ya at the end of the sentence (Kasuga 1985: 288–289) – a pattern likely modeled on the Chinese construction 豈. . .哉 qi. . .zai, which has the same meaning. Another example is kedasi 蓋 ‘probably, if. . .’, which is frequently associated with conjectural expressions. 4) Modification of adjectives ending in kesi into adjectival nouns Another distinctive feature of kunten materials in the transition from the 8th to the 9th centuries is the modification of certain adjectives ending in ke-si (for example isasake-si ‘slight, insignificant’, tsubabirake-si ‘clear, detailed’, and sumuyake-si ‘quick, swift’) into adjectival nouns: isasaka nari, tsumabiraka nari, sumiyaka nari. This use continued into the centuries that followed.
4.2 Adaptation of classical Chinese structures The strategies of kundoku in Japan concerned all aspects of textual analysis, and grammar was no exception. In this respect, the Japanese faced two main difficulties: the penchant of Chinese for multifunctionality and the heterogeneity of classical Chinese documents. Traditional classifications organize classical Chinese words into two groups, shici 実詞 “full words” and xuci 虚詞 “empty words,” namely lexical and grammatical morphemes. The presence of elements that were devoid of a proper semantic value must have posed a challenge for the ancient Japanese who wished to develop a functional strategy. They would have had two options. One would have been to assign to a character the reading that corresponded to the Japanese postposition with the same syntactic function and, if necessary, to rearrange its place in the text. The other would have been to ignore it altogether during the reading process, marking only the syntactic function with the aid of a gloss placed before the lexical morpheme to which it referred. Throughout the better part of the Heian period the second option was preferred; scholars referred to those characters as futokuji 不読字, or “unread characters,” a term which recent kunten scholarship adopted in lieu of okiji 置字 – “discarded characters”.
27 The influence of kanbun-kundoku vocabulary on the Japanese language
555
Semantic matters aside, Japanese manuscripts dating as far back as the Nara period show little consistency in the amount and interpretation of these characters, a fact one cannot solely blame on the peculiarity of available sources. For example, in his study of the Konkōmyō saishōō kyō, a document produced around 762 CE and annotated in 830 CE, Kasuga demonstrates how, in most cases, characters such as er 而 (clausal conjunction), yi 以 (preposition / complementizer), yu 於 (temporal preposition), yu 于 (dative preposition), zhi 之 (genitive marker), zhe 者 (nominalizer), and yu 與 (NP conjunction) were ignored during the reading process. Other characters which are frequently used in Chinese documents, namely yi 矣 (S-final modal particle), yan 焉 (indefinite pronoun), and ye 也 (S-final modal particle), are entirely absent from the text (Kasuga 1985). The inconsistency in number and type of characters in any one text may be the result of stylistic variations in the Chinese original and/or of omissions resulting from negligence on the part of the transcriber. Documents copied at about the same time, such as the Dàshèng dàjí dìzàng shílún jīng 大乘大集地藏十輪經 (Japanese: Daijō daishū Jizō jūrinkyō), annotated around 883 CE, or the Dàtáng Xuánzàng sānzàng fǎshī biǎoqǐ 大唐玄奘三蔵法師表啓 (Japanese: Daitō Genjō sanzō hōshi hyōkei), with glosses dating back to 860 CE, seem to support this theory. The former has more or less the same characters as Konkōmyō saishōō kyō, while the latter displays a great variety of grammatical morphemes. From a linguistic point of view the focus is on the fluctuation in the number of unread characters adopted by the Japanese and on the possible reasons behind changes in their interpretation. As a matter of fact, many of the elements that did not have a proper reading in the 10th century became associated to the Japanese lexicon in the first half of the 11th century and became grammatical morphemes with their own reading thereafter. In other cases, the lack of an established relation between a character and a Japanese function-word resulted in multiple readings. As a result of such ambiguity the characters were unread once again. Such was the case for yi 矣, sometimes associated with the copula nari or with the verb su ‘to do’ but frequently ignored in the reading process altogether (Kobayashi 1954). This inconsistent pattern, with characters shifting from being unread to being read and vice versa, makes it difficult to trust the dating of most manuscripts. Consequently, any reliable and systematic study of unread characters must first consider the syntactic function they played in Chinese and only then move on to examine their grammatical role in Japanese manuscripts through an evaluation of the glosses. “Unread characters” prevent a quick and clear interpretation of Chinese texts. Other aspects of the continental language, however, were easily adapted and became commonly used. Negatives are a case in point. Expression such as 不能, 不得, 不敢, 不 價, 不暇, 不勝, 未曾, 不肯 were given a vernacular reading and in many cases are still present in the modern language: 不能 → (ni / koto) atawazu ‘not be able to. . .’, 不得 → (koto) ezu ‘not succeed into. . .’, 不敢 → aete. . .zu ‘not. . .in the least’, 不價 → (ni) atai sezu ‘not be worthy of. . .’, 不暇 → (ni) itoma arazu ‘not having time for. . .’, 不勝 → agete. . .zu
556
Valerio Luigi Alberizzi
‘not. . .at all’, 未曾 → katute. . .zu / (imada) mukasi ni mo. . .zu ‘never. . .before’, 不肯 → (infinitive of a verb + ) gaenizu ‘not accept to. . .’, later misinterpreted as ‘to consent’.
5 The process of Sino-Japanese contamination As mentioned above (§3), works like Tsurayuki’s skillfully incorporated the conventional vocabulary of kunten materials into a text that followed the Japanese syntax and was rich in native expressions. In the case of written Japanese, a main frame consisting of the different registers of the native language intersects at various levels with a foreign one – i.e. classical Chinese. These varieties are different, but they all belong to a common lexico-syntactic repertoire whose use is determined by context. Hence, language variations within the same group were determined by the language’s social purpose as defined by use, not by user. For example, the refined native prose of the monogatari from the 10th and 11th centuries would have deliberately avoided words related to body parts, diseases, weapons, etc. At the same time such lexicon would have been essential in other fields (for example, medicine) or in writing historical records. Kunten glossed text are a treasure trove for such words: the Nakarai edition of Ishinpō 医心方 (Essential medical prescriptions) – a 12th century manuscript copy of the 10th century medical book by Tanba no Yasuyori (912–995) – is a virtual catalog not only for medical lexicon but also for the names of plants and animals; Buddhist didactical texts like Genshin’s Ōjōyōshū 往生要集 (The Essentials of rebirth in the pure land) are rich in words and expressions describing evil deeds, the punishments sinners have to suffer in hell, and the types of tools and weapons used for torturing them. Some words date back to the 8th century but were never used in Genji monogatari or in other works of court literature. The only exceptions appear in passages from the Utsuho monogatari (The Tale of the Hollow Tree), the first full-length novel of Japan, which was likely compiled during the late 10th century. However, Utsuho is a fragmented and unpolished work, a precursor to the golden age of monogatari literature. Especially relevant examples of this lexicon are: 肺 pukupukusi ‘the lungs’, 腎 murato ‘the kidney’, 脾 yokosi ‘the spleen’, 腸 parawata ‘the intestines’, 大腸 putowata ‘the large intestine’, 小腸 posowata ‘the small intestine’, 胃 kusobukuro ‘the stomach’, 膀 胱 yubaripukuro ‘the bladder’, 趺 anaura ‘the foot sole’, 齗 pasisi ‘the gums’, 唾 tuwaki ‘saliva’, 咽 nomuto ‘the throat’, 膿 umisiru ‘pus’, 商人 akipito ‘merchant’, 漁師 iwotori ‘fisherman’, 笞 simoto ‘cane’, 兜 kabuto ‘helmet’, 楯 tate ‘shield’, 弓 yumi ‘bow’, 矢 ya ‘arrow’, 太刀 tati ‘sword’, 鉾 poko ‘pike’, 鎧 yoropi ‘armor’, 刀 katana ‘sword’.
27 The influence of kanbun-kundoku vocabulary on the Japanese language
557
5.1 Contrastive dimorphic expressions The presence of native and Sinicized words in the same context may attest to a mixed lexical inventory, but it does not tell us anything about the intentions of the author. It would be interesting to know whether the combination of native and Sinicized words was completely random or whether the choice of words followed a certain logic. A comparison between expressions that share the same semantic area despite being unique to one of these two written varieties would effectively bring to light the differences between a Sinicized and a Japanese context. The presence of such expressions within the same text would not only attest to the existence of a Sino-Japanese hybrid writing but would also provide insights as to the intent of the text itself. Once again Tsukishima’s research (Tsukishima 1963: 350−352) provides us with valuable data. As part of his survey, Tsukishima identified one hundred and twenty-two expressions, all with the same meaning, which are characteristic of Sinicized or native varieties. He named them ‘contrastive dimorphic expressions’ – nikei tairitsu hyōgen. Both Tsukishima (in his later works) and other scholars have since cast doubt on the reliability of some of these elements. Nevertheless, this group of words is still considered an important starting point for researchers. Adverbs are a grammatical category that well epitomizes the use of contrastive dimorphic expressions in context. These words provide information about the manner, degree, place, or circumstances of the activity indicated by the element with predicative function. There exists a remarkable difference in the use of adverbs of state (jōtai fukushi) such as sumiyaka ni (a word typical of kunten glossed texts), used to indicate swiftness, and the use of the native pair to-si and paya-si. The same goes for degree adverbs (teido fukushi) such as the Sinicized masumasu and iyoiyo vis-a-vis the Japanese itodo all of which denote a rapidly accelerating process. Let us take a closer look at the latter as used in the epic ‘war tale’ of the Heike monogatari (The Tale of the Heike). In the entire text of the Heike monogatari masumasu is used only once. As such, it does not play a relevant role in the work. The correspondence of iyoiyo and itodo, on the other hand, appears twenty-seven and twenty-four times respectively, offering an excellent case study for a comparison. A perusal of the sentences containing iyoiyo reveals that this adverb is used to express a lively intensification of emotions, feelings, or phenomena. (10)
みめかたちいよ々うつくしく、 あたりもてりかゝやくばかりなり。 mime katati iyoiyo utukusiku, atari mo teri-kakayaku bakari nari. ‘. . . a youth whose ever-increasing beauty of face and figure seemed to suffuse the surroundings with radiance.’ (Kindaichi et al. 1960: 413; translation by McCullough 1988: 420)
558
Valerio Luigi Alberizzi
(11)
三位中将まづおさなき人々の御文を御らんじてこそいよ々せんかたなげにはみえら れけれ samuwi tyuuzyau madu osanaki pitobito no o-pumi o goranzite koso iyoiyo senkatanage ni wa mierarekere. ‘Koremori seemed moved beyond endurance by the sight of the children’s letters.’ (Kindaichi et al. 1960: 241; translation by McCullough 1988: 327)
(12)
いよ々思ひやまさり給ひけん。 iyoiyo omoi ya masari-tamaiken. ‘Perhaps the sight of it had quickened his longing.’ (Kindaichi et al. 1960: 246; translation by McCullough 1988: 330)
By contrast, the native adverb itodo is associated with expressions of sadness, misery, despair, or related to expressions that indicate anxiety, deep nostalgia, and passive or negative moods. (13)
あけがたの月の光にはえさせ給ひて、玉躰もいとゞうつくしうぞみえさせをはしまし ける。 akegata no tuki no pikari ni paesase-tamaite, gyokutai mo itodo utukusiu zo miesase o wa si-masikeru. ‘[The new Retired Emperor] looked even handsomer than usual in the light of the early morning moon.’ (Kindaichi et al. 1959: 273; translation by McCullough 1988: 132)
(14) うき世のきづなとおぼえて三位中将いとゞせんかたなげには見えられける。 ukiyo no kiduna to oboete samuwi tyuuzyau itodo senkatanage ni wa mierarekeru. ‘“Fetters that bind to a world of sorrow!” he thought. He hesitated, paralyzed by grief.’ (Kindaichi et al. 1960: 99; translation by McCullough 1988: 244) (15)
中々御文を御覧じてこそいとゞ御思ひはまさらせ給て候しか。 nakanaka opumi o goranzite koso itodo on’omoi wa masarase-tamaite sauraisika. ‘He felt even more unhappy after he read your letter.’ (Kindaichi et al. 1959: 239; translation by McCullough 1988: 114)
In the aforementioned examples, iyoiyo and itodo are used as synonyms to modify the adjectives in adverbial use utukusi ‘handsome’ and senkatanage nari ‘be at a loss’ and the verb omoimasaru ‘feel (increasingly) unhappy’, respectively. As for the context, iyoiyo utukusiku is used to portray the character of Rokudai – i.e. Taira no Takakiyo (1173–1199), while itodo utukusiku describes a depressed Retired Emperor Takakura (1161–1181) as he is paying a visit to the Toba mansion. Iyoiyo senkatanage ni expresses Koremori’s (1160–1184) burning desire to be reunited with his wife and children, who are in the capital. Itodo senkatanage ni indicates the gloom of
27 The influence of kanbun-kundoku vocabulary on the Japanese language
559
Fujiwara no Naritsune (1156–1202), deeply concerned for his pregnant wife and for his weeping children during Koremori’s flight from the capital. The context reveals that iyoiyo or itodo were chosen to indicate, respectively, the possibility of future hope or a situation of total despair. Finally, iyoiyo omoi ya masari-tamaikemu emphasizes the deep affection that Taira no Shigehira (1158–1185) feels toward his wife of many years upon receiving her letter; itodo on-omoi wa masarase-tamaite sauraisika is used to describe the excruciating pain of Shunkan (1143–1179), exiled at Kikaigashima, at the sight of a letter from his daughter, who is alone in the capital. In other words, in the Heike monogatari iyoiyo indicates a visible external progression detectable by the eye of the beholder; it is used to emphasize development or progress. On the contrary, itodo relates to a passive psychological introspection and is used in doom-and-gloom situations defined by grief, sorrow, and sadness (Hara 1977).
5.2 Sino-Japanese syntax The coexistence of native and Sinicized vocabulary within the same document does not, in and of itself, indicate the completion of a process of Sino-Japanization of the written language. As mentioned earlier, merging a Sinicized context with a Japanese one is an elaborate process that involves different degrees of complexity and whose rules cannot always be identified with certainty. Once again, research on kunten materials is of great help in this respect. In attempting to identify specific syntactical traits one can follow two paths of inquiry: one would look at the coexistence, within the same text, of syntactical structures characteristic of both varieties; the other would look for new structures born out of the natural merger of the two separate ways of expression, structures never before detected in the so-called “pure” forms of the written languages. Examples of the latter first appear in literary works such as Uchigikishū (Collection of Tales Heard and Recorded) and other types of anecdotal literature produced around the early 12th century. The identification of these structures was made possible by research conducted on “propositional attitude adverbs” (chinjutsu fukushi). This adverbial category, related to the modal characteristic of the utterance, is used to establish a strong, unchangeable agreement with the predicate of the sentence to which it is connected and involves the adoption of a fixed grammatical pattern – bunkei. If a native text features a certain grammatical pattern [A . . . B] and a Sinicized text is defined by a fixed correspondence [C . . . D], then a document in Sino-Japanese should feature a concordance between [A . . . D] and [C . . . B]. When these crossed patterns are present in the same text they testify to the high level of completion reached by this written form (see Figure 4).
560
Valerio Luigi Alberizzi
Figure 4: Sino-Japanese grammatical patterns.
For example, ipamuya (> iwamuya > iwan’ya) is a typical kuntengo. Placed at the beginning of a sentence, it indicates contrast with respect to the clause that immediately precedes it. Such clause is usually introduced by the adverbial particle (fukujoshi) sura according to the pattern . . . sura . . . . ipamuya . . . which is typical of the vernacular readings of a Chinese text. However, Early Middle Japanese texts such as Genji monogatari adopt a completely different pattern to convey the same meaning, a pattern based on the link between the adverbial particle dani in the first clause and the adverb masite at the beginning of the second. As a result, hybrid grammatical patterns of wakan konkōbun 和漢混淆文 are defined by the crossed concordance of the native adverbial particle dani with ipamuya, and by sura with masite. In the Heike monogatari, only dani is used with the function of adverbial particle. In four cases it is related to masite and in eight with ipamuya, which indicates the prevalence of the hybrid pattern over the native one. (16)
勅勘の者は月日の光にだにもあたらずとこそ申せ。何況や、 いそぎ都の内を追出さる べしと、院宣・宣旨の成たるに、 しばしもやすらふべからず。 tyokukamu no mono wa tuki-pi no pikari ni dani mo atarazu to koso mause. nanzo iwamuya, isogi miyako no uti o oi-idasaru besi to, winzen senzi no naritaru ni, sibasi mo yasurau bekarazu. ‘“It is said that a man punished by a sovereign is denied the very light of the sun and moon”, he said. “How could I hesitate even briefly, now that the Retired Emperor and the Emperor have both ordered my immediate expulsion from the city?”’ (Kindaichi et al. 1959: 146; translation by McCullough 1988: 60).
27 The influence of kanbun-kundoku vocabulary on the Japanese language
561
(17)
人の親の子をおもふならひはをろかなるが、 先立だにもかなしきぞかし。 いはむや是 は當家の棟梁、當世の賢人にておはしければ、恩愛の別、家の衰微、悲ても猶餘あ り。 pito no oya no ko wo omou narai wa orokanaru ga, sakidatu dani mo kanasiki zokasi. iwamuya kore wa tauke no tauryau, tausei no kenzin nite owasikereba, on’ai no wakare, ie no suibi, kanasimitemo nao amari ari. ‘A parent mourns the death of even the most unpromising child, but Shigemori had been the main stray of a family and the sage of a nation, and thus both the personal loss and the blow to the house of Taira were sources of endless sorrow.’ (Kindaichi et al. 1959: 245; translation by McCullough 1988: 117)
(18)
われと御位を儲の君にゆづりたてまつり、麻姑射の山のうちも閑になどおぼしめす さき々だにも、哀はおほき習ぞかし。況やこれは、御心ならずおしをろされさせ給ひ けんあはれさ、 申もなか々おろか也。 ware to mikurai o mauke no kimi ni yuduri-tatematuri, pakoya no yama no uti mo siduka ni nado obosimesu sakizaki dani mo, aware wa ooki narai zokasi. iwamuya kore wa, migokoro narazu osi-orosaresase-tamaiken awaresa, mausu mo nakanaka oroka nari. ‘Such an event must always call forth deep emotion, even when a sovereign acts of his own accord to enjoy the peace of retirement. But what of this Emperor, driven from the throne against his will! No words could do justice to his feelings.’ (Kindaichi et al. 1959: 270; translation by McCullough 1988: 131)
(19)
又安元の比おひ、御方違の行幸有しに、 さらでだに鷄人曉唱こゑ、明王の眠ををど ろかす程にもなりしかば、 いつも御ねざめがちにて、 つや々御寢もならざりけり。況 やさゆる霜夜のはげしきに、延喜の聖代、... mata Angen no koro oi, mikatatagai no gyaugau arisi ni, sarade dani keizin akatuki tonau koe, meiwau no neburi o odorokasu podo nimo narisikaba, itumo minezamegati nite, tuyatuya gyosin mo narazarikeri. iwamuya sayuru simoyo no pagesiki ni, Engi no seitai, . . . ‘Also, around the Angen era, the Emperor made a journey to avoid an unfavorable direction. Even under ordinary circumstances, he always started awake at the hour when “the time-keeper’s proclamation of approaching dawn invades the sovereign’s slumbers”, but he found it hard to sleep at all on that particular night, for the bitter, frosty cold made him recall, . . ..’ (Kindaichi et al. 1959: 390; translation by McCullough 1988: 199)
562
Valerio Luigi Alberizzi
(20)
帝都名利地、 鷄鳴て安き事なし。 おさまれる世だにもかくの如し。 況や乱たる世にを いてをや。 teito myauri no ti, niwatori naite yasuki koto nasi. osamareru yo dani mo kaku no gotosi. iwamuya midaretaru yo ni oite oya. ‘“The capital is a place where men compete for fame and fortune; after cockcrow, there is no rest”. If that is true even of a peaceful society, how much more must it apply to turbulent times!’ (Kindaichi et al. 1960: 94; translation by McCullough 1988: 242)
(21)
人はいづれの日、 いづれの時、必ず立歸るべしと、其期を定をくだにも久しきぞか し。況や是はけふを最後、 只今かぎりの別なれば、 ゆくもとゞまるも、 たがゐに袖をぞ ぬらしける。 pito wa idure no pi, idure no toki, kanarazu tatikaeru besi to, sono go o sadameoku dani mo pisasiki zokasi. iwamuya kore wa keu o saigo, tadaima kagiri no wakare nareba, yuku mo todomaru mo, tagai ni sode o zo nurasikeru. ‘A period of separation seems long enough when the day and hour of the traveler’s return are fixed, but theirs had been final goodbyes, eternal farewells; and both those who went and those who stayed had wept until their sleeves were drenched.’ (Kindaichi et al. 1960: 114; translation by McCullough 1988: 253)
(22)
たゞ大方の春だにも、 くれ行空は物うきに、況やけふをかぎりの事なればさこそは心 ぼそかりけめ。 tada ookata no paru dani mo, kure-yuku sora wa mono-uki ni, iwamuya keu o kagiri no koto nareba sakoso wa kokoro-bosokarikeme. ‘Even in an ordinary spring, it is sad to witness the celestial manifestations of the season’s departure: thus, we may imagine the distress felt by one who was never to greet another day.’ (Kindaichi et al. 1960: 281; translation by McCullough 1988: 348)
(23)
人の子はめのとなどのもとにをきて、時々見る事もあり。 それだにも恩愛はかなしき 習ぞかし。況や是はうみおとして後、 ひとひかたときも身をはなたず、。。。 pito no ko wa menoto nado no moto ni okite, tokidoki miru koto mo ari. sore dani mo on’ai wa kanasiki narai zokasi. iwamuya kore wa umi-otosite noti, pito-pi katatoki mo mi o panatazu. ‘Parents love their children – even when they send them to live with nurses and merely see them now and again, as some do. But Rokudai has never left my side from the moment of his birth.’ (Kindaichi et al. 1960: 397; translation by McCullough 1988: 411)
So far, researchers in Japan have identified the six patterns shown in Table 1. While this is an important step in the study of Sino-Japanized hybrid syntactical structures, the field is still young and will require more surveys in the future.
27 The influence of kanbun-kundoku vocabulary on the Japanese language
References Endō, Yoshimoto. 1952. Kunten shiryō to kuntengo no kenkyū [A study on kunten materials and their language]. Kyoto: Kyōto Daigaku Bungakubu Kokugogaku Kokubungaku Kenkyūshitsu Kokubungakkai. Frellesvig, Bjarke. 2010. A history of the Japanese language. Cambridge: Cambridge University Press. Frellesvig, Bjarke. 2013. Remarks on the verb suru in Old Japanese: A corpus based study. NINJAL Project Review 3(3). 152–177. Hara, Eiichi. 1977. Chūsei shotō ni okeru ‘itodo’ to ‘iyoiyo’ no taishōteki yōhō ni tsuite [On the contrastive use of itodo and iyoiyo at the beginning of medieval times]. Kanazawa Daigaku Kyōyōbu Ronshū – Jinbun Kagaku Hen 15. 1–10. Kasuga, Masaji. 1985. Saidaiji-bon Konkōmyō saishōō kyō koten no kokugogakuteki kenkyū [A linguistic study of the old glosses of the Saidaiji manuscript of the Konkōmyō saishōō kyō]. Tokyo: Benseisha. Kindaichi, Haruhiko, Kaoru Atsumi, Ichinosuke Takagi and Masao Ozawa (eds.). 1959. Heike monogatari – jō [Tale of Heike, vol. 1] (Nihon koten bungaku taikei [Compendium of Classical Japanese literature] 32). Tokyo: Iwanami Shoten. Kindaichi, Haruhiko, Kaoru Atsumi, Ichinosuke Takagi and Masao Ozawa (eds.). 1960. Heike monogatari – ge [Tale of Heike, vol. 2] (Nihon koten bungaku taikei 33). Tokyo: Iwanami Shoten. Kobayashi, Yoshinori. 1952. Kanbun-kundoku shijō no hitomondai [One problem in the history of Kanbunkundoku]. Kokugogaku 16. 47–64.
564
Valerio Luigi Alberizzi
Kobayashi, Yoshinori. 1954. Raku nomi maku nomi no genryūkō [Some reflections on the origin of raku nomi and maku nomi]. Bungaku Ronsō 8. 29–45. Kobayashi, Yoshinori. 1966. Kunten shiryō to shite mita ‘Kōfukuji-bon Daijionji Sanzōhōshiden kokunten no kokugogakuteki kenkyū – yakubun hen [The Kōfukuji manuscript of the Daijionji Sanzōhōshiden seen as a kunten material: the translated text]. Kokugogaku 65. 105–113. Kobayashi, Yoshinori. 1974. Kaeriten no enkaku [Origin and development of inversion glosses]. Kuntengo to Kunten Shiryō 54. 86–111. Kobayashi, Yoshinori. 1979. Kanbun-kundoku-tai [Kanbun-kundoku style]. In Susumu Ōno and Takesi Sibata (eds.), Iwanami kōza nihongo 10: Buntai [Iwanami series on Japanese language: written styles], 93–160. Tokyo: Iwanami Shoten. McCullough, Helen Craig. 1988. The tale of the Heike. Stanford: Stanford University Press. Nakada, Norio. 1954. Kotenbon no kokugogakuteki kenkyū, sōron hen [A linguistic study of the ancient glossed text: general discourse]. Tokyo: Kōdansha. Ōtsubo, Heiji. 1981. Heian jidai ni okeru kuntengo no bunpō [The grammar of Heian period kunten vocabulary]. Tokyo: Kazama Shobo. Ōtsubo, Heiji. 1992. Kaitei kuntengo no kenkyū, jō [Studies on the kunten language, 1, revised edition]. Tokyo: Kazama Shobo. Ōya, Tōru. 1909. Kanazukai oyobi kana jitai enkaku shiryō [Historical materials on the development of kana use and its orthography]. Tokyo: Kokutei Kyōkasho Kyōdō Hanbaijo. Rodd, Laurel Rasplica. 1996. Kokinshū: A collection of poems ancient and modern. Boston: Cheng & Tsui. Saeki, Umetomo (ed.). 1958. Kokin wakashū [Collection of Poems Ancient and Modern] (Nihon koten bungaku taikei 8). Tokyo: Iwanami Shoten. Suzuki, Tomotarō (ed.). 1957. Tosa nikki [The Tosa diary] (Nihon koten bungaku taikei 20). Tokyo: Iwanami Shoten. Tsukimoto, Masayuki. 2000. Daitōkyū kinen bunkozō Zokukegonkyō ryakuso kanjōki kan daigo no kunten ni tsuite [On the kunten of the fifth scroll of the Zokukegonkyō ryakuso kanjōki held by Daitōkyū kinen bunko]. In Kamakura jidaigo kenkyūkai (eds), Kamakura jidaigo kenkyū [Studies on the language of the Kamakura period] 23. 130–140. Tokyo, Musashino Shoin. Tsukishima, Hiroshi. 1963. Heian jidai no kanbun-kundoku-go ni tsukite no kenkyū [A study on the Heian period kanbun-kundoku language]. Tokyo: Tōkyō Daigaku Shuppankai. Tsukishima, Hiroshi. 1981. Tosa nikki to kanbun-kundoku [The Tosa diary and kanbun-kundoku]. In Hattori Shirō, Takashi Kamei and Hiroshi Tsukishima (eds.), Nihon no gengogaku 7: Gengoshi [Japanese Linguistics 7: History of language], 389–401. Tokyo: Taishūkan Shoten. Tsukishima, Hiroshi. 2007. Kunten goi shūsei [A collection of kunten vocabulary] 1. Tokyo: Kyūko Shoin. Yamada, Yoshio. 1935. Kanbun no kundoku ni yorite tsutaeraretaru gohō [On the lexical use conveyed in the vernacular readings of classical Chinese texts]. Tokyo: Hōbunkan.