DESCRIPTIVE LINGUISTICS IN JAPAN

HATTORI SHIRO

I. A SURVEY OF THE HISTORICAL DEVELOPMENT

Japanese descriptive linguistics, together with kokugaku (national learning), belongs to the branches of learning which showed an original and peculiar development in Japan in the . In 1695, the Buddhist monk Keichu published a book, Waji shdransho, the result of his study of the origins of orthography. He had discovered that although corrupted in later documents, correct kana orthography could be found in the literary sources of the eight and ninth centuries. His free spirit and exact method of study greatly influenced contemporary scholars, and kokugaku, the study of the culture of Ancient Japan, became much more scientific than ever before. There then appeared a growing number of descriptive studies pertaining to the language of the old literary sources. These came now to occupy the center of scholarly interest. Scholars like Motoori Norinaga and Fujitani Nariakira came to consider that the kana orthography discovered by Keichu reflected the pronunciation of Ancient (Early Heian) Japanese. The detailed study of kana orthography of Japanese of the Nara period (the eighth century) was completed by Ishizuka Tatsumaro in oku no yamamichi (written before 1798) and the foundation was established for the study of the phonological system and structure of Archaic Japanese of Nara, which were different even from those of the early (the ninth century) and very differ- ent from those of Modern Japanese. On the other hand, due to the fact that the subject matter was limited to definite periods of Old (Nara and Early Heian) Japanese, the grammatical studies became synchronic and there appeared without any foreign influence a number of excellent descriptive researches. Remarkable contributions in this field were the first attempts to classify the parts of speech in Japanese by Fujitani Nariakira (1738-79) in his works Kazashisho (written in 1767), Ayuisho (written in 1773, but published in 1778), etc. In the first of these works he studied the interjections, conjunctions, adverbs, pronouns and prefixes; in the second work he turns his attention to the enclitic particles, "auxiliary verbs", and suffixes. His Yosoisho, a study on the verbs and adjectives, was lost. The great scholar Motoori Norinaga (1730-1801), known for his Kojikiden (1764-98) also contributed greatly to studies of grammar in his Kotoba DESCRIPTIVE LINGUISTICS IN JAPAN 531 no tama no o (completed in 1779, published in 1785), an excellent syntactical study of some of the particles, whereas his Mikuni kotoba katsuyosho (started in about 1782, published in 1886) classified the conjugations of verbs and adjectives. After Fujitani and Motoori came many scholars who described with success the grammar of , such as Suzuki Akira (1764-1837), Motoori Haruniwa (1763-1828), Tojo Gimon (1786-1843), and others. One must note as a characteristic of the grammatical studies of this period the fact that their main object was Old Japanese, that is, the language of the eighth-ninth centuries; as a consequence, when they refer to the spoken language of their time, it is only to draw comparisons with the ancient forms. Based as it was on materials culled only from ancient literary sources, this kind of grammatical study followed principally a deductive process based on a detailed study of the textual function (i.e., the distribution) of words and signemes.1 As for the meaning of words and signemes, the scholars endeavored to grasp it by analogy with present-day Japanese, and by making deductions from the context or from old dictionaries and the use of Chinese characters. Because the object of their study was Ancient Japanese, the meanings of the texts could hardly be grasped in the same precision and accuracy as is possible in the study of the modern language. For that very reason, however, they minutely investigated the textual function of words and signemes, and consequently descriptive studies developed which are admirable even from the present-day point of view. Another characteristic of the studies of the ancient language of the period under review may be said to be the fact that scholars applied to their own written language the rules they had discovered in their analysis of the ancient language. The kana orthography discovered by the monk Keichu became the orthography of the written language used by these scholars, and this language developed into the gikobufi (the imitated ancient language) which made use of the grammar and the vocabulary of ancient texts. In the era (1868-1912) this language, mixed with the written languages of the traditional Sino-Japanese reading style and the epistolary style, evolved into what was called futsubun (the common written language). The distance, however, between this written language and the normal spoken language of the time had become so great that a movement started in the Meiji era to develop a written language based on the spoken idiom, which became known as the genbun itchiron, the "write as you speak" movement. Especially since 1895 the litterateurs and other writers who entered this movement created after a while a written language called kdgobun (the written language in the spoken style), which has by the present day almost superseded all other styles in writing Japanese. As a consequence, the beginning of the Meiji era saw the grammatical studies becoming necessarily oriented towards the study of the daily spoken language. It is true that the philologists of the Edo period often referred to the spoken language in their study of the classical language. A good example can be seen in Motoori Haruni- wa's work, Kotoba no yachimata (written in 1806, published in 1808), in which he 1 These are what L. Bloomfield called "morphemes"; see Hattori Shiro (1964a), fn. 1. 532 HATTORI SHIRO referred to the system of verbal conjugations of the spoken language. However, real descriptive study of the spoken language did not progress very far during that period, mainly because too great a respect was shown to the ancient language, called "elegant language", and one looked down on the spoken language as "vulgar speech". Even when as in the Meiji era, the necessity of the study of the spoken language was advocated, this study never progressed satisfactorily; among the many reasons for this, we wish to underline specially the two following factors: 1. Even in its incomplete shape, the grammatical study of the ancient language had already been worked out, and scholars tried to write the grammar of the spoken language in the framework of that of the ancient language. It was no easy task to eliminate this tendency which remained for a long time. Notwithstanding the evident relationship between the ancient language and the modern spoken language, more than a thousand years had elapsed between the two, resulting in great differences in structure and system. Therefore a descriptive study of the modern language could not be done in the framework of the grammar of the ancient language without distorting the description. 2. With the introduction of many Western scientific disciplines after the Meiji restoration, the garmmars of English and the other European languages also became known to our country. On the other hand, the study of Japanese by Westerners, already sporadically seen during the Edo period,2 led to the publication of Japanese grammars by J. J. Hoffman (1867) and W. G. Aston (1871). B. H. Chamberlain was commissioned by the Japanese Ministry of Education to write Nihon shoburiten [An elementary ], published in 1888. These works tried to describe the in the frame of the structures and systems of Western languages, especially English, and cannot be called very successful. On the contrary, grammatical notions concerning the European languages, introduced by these works had to some extent a distorting influence on the descriptive study of Japanese. As a representative work of this period, we may quote Otsuki Fumihiko's Detailed Japanese grammar (written in 1882, printed in 1897).3 It is a description of the grammar of of the ninth to eleventh centuries, embodying in its description of conjuga- tions the results of studies by Japanese philologists, but it was influenced by Western grammars, especially in its classification of the parts of speech. A short time later, however, an epoch-making work succeeded in casting off this foreign influence; Yamada Yoshio's A Japanese grammar (partly published in 1902, completed in 1908), by striving to grasp the structure and system of the Japanese language as such, put grammatical studies on their true course. Here again, however, the subject of the study was the "modern standard written language", namely, what we have called above the "common written language", which was very different from the daily speech of that time.

2 The oldest of these works is Joao Rodriguez, S. J., Arte da lingoa de Japam, Nagasaki, 1604—1608. 3 Hereafter all works recorded in the Bibliography will be referred to by their translated titles. For the Japanese titles and the renderings in English, see the Bibliography. DESCRIPTIVE LINGUISTICS IN JAPAN 533

But as a result of the "write as you speak" movement, alluded to above, the need was felt for descriptive grammars of the spoken language; articles and books treating this subject start to appear in succession around the year 1900. We have first works like the book Grammar of Colloquial Japanese (1901) by Matsushita Daisaburo in which the description of the daily spoken language was attempted. Meanwhile, the need for establishing a common standard language was stressed, and as "the written language of spoken style" became established, the tendency to describe this language became more and more pronounced. The Ministry of Education published in 1906 the results of a dialect investigation conducted by the Japanese Language Commission, called Report on a survey of the grammar of the spoken language; the same commission published in 1916 a Spoken language grammar which was in reality a normative grammar of the spoken standard. In the first years of the Showa period (after 1926) we see the grammatical investiga- tion of the spoken language proceeding with great rapidity. However, partly because the development of a phonological analysis of the spoken language trailed behind, there was great difficulty in throwing off the habit of relying on written language texts of the so-called "spoken style". Even when Hashimoto Shinkichi made a great step forward in tackling the spoken language directly by introducing the concept of bunsetsu, a prosodemic unit, because of his using &a«a-spelling instead of a phonemic notation, his analysis too went no further, in many respects, than that of the texts of the "written language of spoken style". After World War II, we began to see finally the descriptive study of the spoken language developing in the phonological as well as grammatical aspects, and the descriptive method was further extended to the fields of dialectology and of foreign languages.

II. SOME FUNDAMENTAL PROBLEMS

1. The Concept o/Synchronie

Even if one cannot say with accuracy that the scholars of the Edo period knew what synchronie meant, it remains true that their study of grammar was a synchronic description of the ancient language. Their awareness of the changes brought about in language by the passage of time was the very reason why they attached so much importance to the study of the ancient language, but they did not organize a system- atic study of the history of Japanese from antiquity up to their time. One can say, however, that the concept of synchronie was slowly born out of their findings that no rules could be clearly formulated unless one limited strictly one's study to an exactly limited period of time, even in such a field as that of kana orthography. In the Meiji era (1868-1912), especially around the turn of the twentieth century, 534 HATTORI SHIRÔ with the introduction of Western linguistic works4 Japanese scholars came to have a clear concept of what historical grammar meant, and consequently they came to their own concept of descriptive grammar distinct from historical grammar. For instance, we read in Yamada Yoshio's preface (1902) to his Japanese grammar the following statement, that his work is not a "historical grammar" which "depicts historical changes", but its aim is to "write descriptively only", and to describe "the standard written language of the present day" "limiting itself to one single period" (p. 3). In his book Introduction to linguistics (1922),5 the eminent linguistic theorist Jinbo Kaku says that on the basis of "the study of one language of one period" which "analyzes the linguistic facts by limiting itself to one language (or dialect) of one period" and which "systematizes the knowledge obtained through observation, analysis, and synthesis of all the facts", we can proceed to historical and comparative studies. These statements point naturally to the need of distinguishing synchronic and dia- chronic studies. However, some scholars mixed the two points of view. For instance, some did not distinguish between diachronic sound changes and synchronic sound alternations, calling both of them "phonological interchanges". The famous grammar- ian Matsushita Daisaburo in his Standard grammar of spoken Japanese (1930) shows originality in his synchronic analyses and descriptions but sometimes gives way to a diachronic point of view. Not withstanding these instances, one can say that diachronie and synchronie were in general not confused, and descriptive studies developed steadily.

2. Meaning and Introspection

Descriptive linguistics in Japan, having evolved from the study of the Japanese language by scholars whose mother tongue it was, never fell into such extreme anti- mentalism as to exclude the study of meaning through introspection by the researcher himself. The earlier scholars saw in the textual function (the distribution) of words and of signemes the key to an accurate grammatical analysis of the ancient language, which was the first object of their studies. Even when, later on, scientific work directed itself to the language then spoken and written, the traditional procedures continued to

4 For example, the following linguistic books were translated: A. Darmesteter, The life of words as the symbols of ideas, 1886, translated by Kanazawa Shozaburo, 1897; A. H. Sayce, Principles of comparative philology, 1874, translated by Ueda Kazutoshi and Kanazawa Shôzaburô, 1898; W. D. Whitney, Life and growth of language, 1875, translated by Hoshina Koichi, 1899; O. Jespersen, Pro- gress in language, 1894, extracts translated by Shinmura Izuru, 1901 ; and H. A. Strong, Introduction to the study of the history of language, 1891, an adaptation of H. Paul's Prinzipien, translated by Yasugi Sadatoshi, 1901. 5 This book by Jinbo was published after F. de Saussure's Cours de linguistique générale (1916), but was written independently from this latter work. The contents show this conclusively. DESCRIPTIVE LINGUISTICS IN JAPAN 535 influence them. Even a scholar like Yamada Yoshio, who pursued his analysis of meaning with the help of psychology and logic, utilized in a large measure the textual function of words. Pushing forward the traditional approach, Hashimoto Shinkichi put the greatest stress on shokuno, that is, on textual functions, in arriving at his word classes. He has the following to say concerning meaning :6 Language is not only composed of meanings but also and necessarily of accompanying forms. Classification according to meaning does not belong to linguistics unless it is supported by differentiations in form. And the distinction and identification of meanings occur mainly in the interior of man's mind, so that there is no way of direct observation from the outside; the only way of recognizing differences in meaning is the indirect way of conjecturing it through linguistic forms. In his grammatical studies Hashimoto put the main stress on textual functions of forms, and did not analyse meaning sufficiently. However, in his definition of "gram- mar" he does refer to meaning as follows:7 Grammar is an arrangement existing inside a certain language, a general rule con- cerning the structure of its meaningful units. According to Hashimoto, "phones" and "syllables" are units without meaning, while "words" and "sentences" are meaningful units.8 In this way, traditional grammatical studies in Japan, even when they exclusively stressed textual function, worked always in relationship with meaning, and this "meaning" was reached mainly through introspection. Even in the field of phonology, scholars like Arisaka Hideyo took a strongly mentalistic stance and relied in great measure on introspection.9 Jinbo Kaku, in his latest work, Linguistic theory (1961), considers both cases of perception and those of reproduction as the material of linguistic study (p. 152f., p. 190), but the latter are experienced only through introspection. Hattori Shiro has been claiming that linguistic analysis must endeavor to be as objective as possible, but at the same time it ought not to exclude the subjective part of direct experiences. In a recent paper,10 he says that the affirmation made by some structuralists that "sounds can be studied objectively, but not so the meanings", is erroneous, and that the study of meaning can be made much more objective, while both meaning and sound cannot be studied without resorting to the subjective direct experience of the speakers; also, that one must defend the completely scientific legitimacy of utilizing such subjective experiences in linguistic study.

6 See Hashimoto Shinkichi, 1934, 43-4; the book published in 1934 is based on the lectures given by Hashimoto at the Faculty of Letters of Tokyo Imperial University in 1929 on "Japanese grammar" and in 1932 on "Outline of the grammar of the national language". 7 Hashimoto Shinkichi, 1937, in Collected works, 1.334. One finds similar words in a 1935 paper, ibid., 209. 8 Hashimoto, 1937, ibid., 331. 9 Arisaka Hideyo, 1940. 10 Hattori Shiro, 1964b; the English version is found in the journal Foundations of language, vol. 2 (1965). 536 HATTORI SHIRO

3. Langue and Parole

Through independent observation and thinking and without having come into con- tact with F. de Saussure's work, Jinbo Kaku formulated concepts which were very near to those of langue and parole.11 These concepts are explained in his work Intro- duction to linguistics (1922). In this book, he first distinguishes between "concrete sounds" and "abstract sounds," as in the following passages: In the same Ohayo 'Good morning' said by different people, and even by the same man in different circumstances one shall find slight differences. From the slightly different perceptions of the utterances, one abstracts some common qualities, eliminating those qualities which are not common in them, and thus forming an abstract "sound- idea", by which is meant "the concept" of the sound. Every single utterance of Ohayo is a "concrete sound" embodying all the qualities referred to above. These single utterances contribute to the formation of an abstraction by their repetition. A friend of mine said Ohayo to me this morning and my family and others also greeted me with the same word yesterday and the day before. All these concrete examples of Ohayo, piling up, have left only a common essence in my memory. Everything besides this common essence has been rejected to shape the sound-idea, which I now have (p. 23 f.). I will call by the term "characteristic elements" the abstracted portions which are common in the concrete sounds uttered by various persons on various occasions, and by the term "casual elements" those elements which come to be casually added in each concrete instance (p. 26). Therefore the study of the vowel "a" in Japanese entails as one of its tasks the study of its remembered sound-idea, and as another task the study of the constant elements in each of the concrete instances of "a" pronounced by various persons. The same is true for each consonant (p. 31). In the same way, we find Jinbo differentiating between "concrete meanings" and "characteristic meanings" or "meaning-ideas"; these he identifies with Hermann Paul's "okkasionelle Bedeutung" and "usuelle Bedeutung", respectively (p. 60f., 67). What Jinbo calls a "speech-idea" (Sprachvorstellung) is a combination of a "sound- idea" and a "meaning-idea" (p. 65). The speech-ideas shaped by abstraction from concrete sounds and concrete meanings are accumulated in the brain of the speaker, forming complicated systems and structures: I will call by the name "speech action" (Sprechtatigkeit) the act of concretizing this speech-idea in a particular instance. This speech action is usually referred to in such expressions as "to use words", "to speak", "to say something", "to utter", "to exchange words", "to converse", etc. Speech action therefore is an intentional action in which concrete sounds are used with the aim of expressing and communicating concrete meanings. The speech-idea is the basis of a speech action, or, to use a metaphorical expression, speech-ideas are the materials of speech. The speaker has ready in his brain a large quantity of speech materials, and by taking out those necessary for each occasion,

11 See fn. 5. DESCRIPTIVE LINGUISTICS IN JAPAN 537 he uses them. If these speech-ideas would remain only in the brain, they would be no better than a buried treasure. In speech actions, these materials receive for the first time their true values. Linguistic study means therefore to address oneself to both the speech-ideas and the speech actions, and it would stay incomplete if it considered only one of the two. Furthermore, a separate study of these two would be incomplete unless it considered also their mutual relationship. The reason for this is of course that speech-ideas are born out of the reiteration and abstraction of speech actions, and that speech actions are made possible on the basis of speech- ideas (p. 68 f.). One has to acknowledge that Jinbo's concept of "speech-ideas" is very similar to that of F. de Saussure's langue. According to Hattori Shiro, we cannot observe at the present stage of science the "langue" latent in our brains; therefore, we need to take recourse to that reality of the spoken language which can be observed (or subjected to introspection), that is, the concrete act of speech (both active and passive) and the sounds produced thereby. However, in all these, one can find the recurrent features induced by social habits. The actual concrete "act of utterance, sounds of utterance, and act of understanding" are "individual substances" in which one can distinguish both individual and social habit features; the latter are a reflection of what de Saussure called langue, concludes Hattori. If the "act of utterance and sounds of utterance" correspond to what de Saussure called parole, Hattori claims that one can find the features of langue in parole itself. On the other hand, if by reproducing "speech-ideas" one can submit them to introspection, these subjective direct experiences are also "individual sub- stances" and therefore are expected to be different in two persons. It is supposed that the only thing which can be identical in two persons would be the social habit features, that is, the langue features found in these subjective experiences. It is exactly the same as when the sounds uttered by two persons are separate "individual substances" and never identical, the only thing which can be identical being the social habit features therein. Following this line of thought, Hattori defines an "utterance" as the concrete act of speaking (one or more sentences) along with the sounds produced thereby, and a "linguistic production" (which consists of one or more "sentences") as the social habit features found in the utterance. By making this distinction, he attempts to facilitate the analysis of utterances, eliminating from linguistic study the difficulty caused by mixing the concepts of utterance and "linguistic production" and by re- garding the "sentence" which is a constituent of the latter as an actual event.12 In a paper published in 1964, Uemura Yukio, on the basis of a minute observation of the sounds and situations of utterances, contributed a very refined and accurate analysis of the nature of the langue-like and paroleAike, aspects of the sounds of utterances and what these aspects can convey in communication. 12 See papers, especially those after 1949, in Methods in linguistics. Jifibo Kaku in his Introduction to linguistics, 59, considers as casual elements, that is, as parole features, the variations in intonation found at the ends of sentences, including the ascending intonation expressing the "meaning of a question". Hattori, however, considers the patterns of intonation as elements in the shape of "sen- tences", that is, as langue features. 538 HATTORI SHIRO

III. Phonology

From the days of Empress Suiko (reigned 592-628) through the Nara and Heian periods, the Japanese studied the Chinese language and were fairly well versed on the sounds of their own language. The fact that the Manyogana orthography of the Nara period precisely reflects the pronunciation of Japanese is largely dependent on the knowledge the Japanese had of the Chinese language. Examples of the Chinese way of transcribing tones applied to Japanese accents are seen in documents of the Nara period. On the other hand, Siddham had already been introduced to Japan in the same Nara period and many scholars studied it especially during the Heian period. Further, from the Kamakura period (1192-1333), the lines and columns of "the fifty syllabary sounds" were arranged in accordance with that of the Siddham and this system has continued down to the present time. Parallel to study of Siddham, research in Chinese phonology, undertaken in order to obtain knowledge of the Chinese sounds, has been conducted continuously throughout each period up to the Meiji era and has served to have the Japanese re-evaluate the sounds of the Japanese language. Despite the above studies, it is possible to say that the Japanese came to possess scientifically systematized knowledge even in the case of the Japanese sounds only when phonetics of the Western world was introduced to Japan in the Meiji era. Publications in phonetics started to appear from around 1900. However, significant among publications of this period is the Large dictionary of Japanese (1892-1893) by a litterateur, Yamada Bimyo. This dictionary included unique and original notations necessary and sufficient to pronounce each word entry in the Tokyo accent, transcrib- ing differently the sounds [g] and [rj], which cannot be differentiated by the kana. Phonetic studies in the sounds and accent of Japanese, centered around Jinbo Kaku and Sakuma Kanae, progressed during the Taisho era (1912-1926) and the first budding of phonology is seen in the former. Jinbo's theorizing is expressed in a passage previously quoted13 from his publication entitled Introduction to linguistics (1922), but we will further quote the following passages which have some connection with the "phoneme": Now, as the next step, to pursue a fact which may be referred to as "one sound", we take up and consider only one portion of it, disregarding all of the other properties of a concrete sound. As an example, let us start with the word [ohajol]. Let us accumulate the examples of [ohajol] spoken many times by one person and [ohajol] voiced by many different people. These utterances were all made to express the greeting "Good morning" and therefore can be accumulated and compared in terms of the meaning they express. If the peculiar and casual elements of each utterance are removed and only the common portions are extracted from the many accumulated examples of [ohajol], the sounds [o] [ha] and [jol] included in this word can be obtained. This [o] then is an abstract element. This [o] is the initial [o] included in the word [ohajol], but still cannot be defined as being the same sound, for example, as the [o] which is the initial syllable in such words as [ojajubi] 'thumb', [oqawa]

13 Chapter II, 3. Langue and Parole. DESCRIPTIVE LINGUISTICS IN JAPAN 539 'stream', or [omoni] 'heavy load'. Now, using the same procedure, accumulating and abstracting the word [ojajubi] uttered by many different people in various con- crete sounds, the sound [o] is again obtainable. This sound [o] then is the [o] in the word [ojajubi]. Now, are the [o] in [ohajol] and [o] in [ojajubi] the same? It is premature to define them as being the same. These two [o] may bear different peculi- arities. If so, these two sounds, and other examples of [o] such as those included in [oqawa] and [omoni] should be accumulated and only the common portion should be extracted. After that we can come to a general conclusion that the Japanese [o] is such and such a sound. Then, to what degree should [o] as one sound, for an example, be defined? If the special [o] in [ohajol] is classified as one sound, there should be countless other examples of [o] in the Japanese language. However, the fact that [o] is defined as one sound means that the extent of abstraction has advanced and that [o] has been abstracted from all the various words that include the sound [o] (p. 279 f.). It is remarkable that in constructing his theory Jinbo assumed the possible differences between [o] in various words, when people of the time assumed beyond doubt that all examples of the [o] were the same. As a matter of fact, the [o] in [ohajol], due to the influence of the succeeding [h], which is often voiced, is more or less breathy, while the [o] in [ojajubi], due to the succeeding [j], is somewhat palatalized. The [o] in [oqawa], due to the [q], is more or less nasalized, and the [o] in [omoni], because of the [m], is also nasalized and further the hps are somewhat narrowed. The naive opinion of stating that these various examples of [o] are the same sound uncon- sciously included the phonological viewpoint. Jinbo thus arrived at a concept similar to that of the "phoneme", but it is clear from his statements in other places14 that his concept included various entities without discriminating among them: "abstract sound", "phonemes", and "morphophonemes". Jinbo's theory, which belongs to the early days of phonology, influenced H. E. Palmer.15 As stated in the latter part of this chapter, Miyata Koichi in 1927 made public his phonological views with regard to the accent of the .16 The phonological significance of his views was, however, not adequately appreciated by either the people who disagreed or agreed with his theory. Arisaka Hideyo was a scholar who established a unique theory in phonology and greatly influenced academic circles in Japan. His papers started to appear in 1931,17 and later were compiled and published in one volume entitled Phonology (1940). One of the most significant characteristics in the synchronic portion of his theories is the attempt to group various similar sounds into one phoneme by a concept called the "aim-idea" or "phoneme-idea". Arisaka explains this point as follows: First, when observing my own pronunciation, ao 'blue' ordinarily is [ao], akai 'red' is [akai], and miyage 'present' is [mijaeqe]. The qualities of the sounds [a], [a], [a], and [ee] are all different. However, when these words are pronounced slowly and

14 Jinbo Kaku, 1922, 280f. 15 H. E. Palmer, The principles of romanization, Tokyo, 1930. " Miyata Koichi, 1927, 1928. 17 Arisaka Hideyo, 1931. 540 HATTORI SHIRO distinctly, ao becomes [ao], akai [akai], and miyage [mijarje]. In other words, the four different phones become the same vowel [a]. The reason for this is that the ideal in my mind, in other words, the aim-idea, is just one [a]. Accordingly, when one is careful and conscious, this ideal is fully actualized in the act of pronunciation. How- ever, when one is not attentive, the act of pronouncing is not fully conducted and it is influenced by various elements and thus takes the forms of [a], [a], or [ae], which are only partial actualizations. (However, even when a word is pronounced without adequate attention, if the circumstances happen to be favorable, it is, of course, possible that the actualization is similar to [a]). In other words, the above [a], [a], [a], and [ae] are physiologically and physically different sounds, but from the intentions of the speaker, they are all actualizations of the same "aim-idea" ((a)). In this sense, some connection of meaning exists among [a], [a], [a], and [ae]. In this case, we call the common aim-idea ((a)), which is the foundation of this connection the phoneme and differentiate it from the individual sounds [a], [a], [a], and [ae] which are actualizations of it (p. 13).

In terms of American phonemics, Arisaka's words may be paraphrased as follows, that in order to group the allophones [a], [a], [a], and [ae] into one phoneme /a/, an 'aim-idea', a mentalistic idea, is being applied. However, this theory should not be ignored as being merely mentalistic when we consider the following points. To take an example, the word [mijaeqe], when pronounced slowly and carefully, becomes [mijaqe], and if uttered extremely carefully and distinctly, becomes closer to [mijaije]. Thus, it can be said to fluctuate in a certain direction showing an unlimited number of variations. We could consider the above passage from Arisaka as grasping this objective actuality together with the subjective direct experience corresponding to it. Furthermore, the fact that a number of scholars, by applying this theory when ob- serving the pronunciations of various dialects and foreign languages, have started from careful and distinct pronunciations and have profited greatly in analytical studies, cannot be overlooked. The fact that Arisaka differs from Jinbo is made clear in the following perspicacious statement taken from Arisaka: The phoneme-idea, as stated, possesses only a portion of the many characteristics actual sounds have, but, on the other hand, it is not a mere abstract of the common characteristics of the various sounds which are the actualizations of it. For instance, if the phoneme-idea ((a)) is an abstract concept composed merely of the common characteristics of [a], [a], [a], and [ae], how would the fact that by pronouncing these sounds carefully they all become close to the ideal sound [a] be explained? The phoneme-idea ((a)), with respect to the movements of the tongue and lower jaw, apparently possesses the characteristics of the ideal and concrete sound [a]. The phoneme-idea generally disregards functions of the organs of speech which are not directly concerned with its actualization, but on the other hand possesses relatively substantial ideal and concrete contents with regard to the functions of the organs of speech important to its actualization. This fact can be readily understood when one considers how the phoneme-idea comes into existence (p. 14). Arisaka also explains the process of the establishment of the phoneme-idea, in a way different from Jinbo: DESCRIPTIVE LINGUISTICS IN JAPAN 541 The phoneme-idea is derived from the accumulation of countless impressions of sounds which one has heard from various people in the past. However, the individual sounds are various in their physical intensities, and may also differ greatly depending on the emotional values. Accordingly, the depth and continuity of the impression provided by a sound varies greatly. In other words, a sound uttered casually and without emphasis generally leaves only a weak impression and usually fades away in a short period of time. On the other hand, a sound uttered strongly and with emphasis is not only generally great in physical intensity, but also shows the strong attitude of the speaker himself in making the utterance, arouses the attention of the listener, and naturally serves to become a profound and strong impression which is preserved for a longer period of time and ultimately forms a phoneme-idea. There- fore, when we think of a phoneme or a word, the thought that usually comes to mind is the form of the phoneme or word when pronounced most carefully and distinctly. The form of the phoneme or word uttered carelessly usually does not remain in memory (p. 14 f.). Furthermore, Arisaka makes the following statement with regards to the preservation and reproduction of the phoneme-idea: Well, the phoneme-idea, which is established as stated, normally exists below the threshold of consciousness, but if attention is directed to it, it is reproduced in various forms whether clear or ambiguous. Of course, the more the degree of concentration of attention, the more complete the reproduction will be, and the less attentive, the more incomplete the reproduction will be. It could be illustrated in recalling the impression of an acquaintance. If you were cautious and attentive, your memory will vividly tell you the details of the person, but if you were lax in your observation, only an outline of the person will come to mind (p. 15). In disagreement with Arisaka's theory, it may be possible to say that a person not only clearly remembers sounds pronounced carefully, but to an extent also remembers inattentively pronounced sounds as a matter of social habit. Arisaka's description, however, seems to stand more sharply in relief and to come closer to a true picture than the plain, mechanical listings of, for example, [mijaeqe], [mijaqe], and [mijarje] as mere free variations. There are, however, points in Arisaka's theory with which it is difficult to concur. The following is one of the most basic and important points on which we can take issue: Now, the phoneme is basically an idea and therefore it does not always possess all of the characteristics of the sounds which are its actualization. For example, the pho- neme ((a)) includes only a portion of the various characteristics that the actual sound [a] possesses. The elements included in the phoneme-idea are limited to the actions of the major organs of speech, including, of course, the accompanying acoustic and tactual elements. The actions of the other organs are eliminated. For instance, the four [m]'s which appear in matsu 'pine' [mats], momi 'fir' [momi], and mugi 'wheat' [mur)i], are all actualizations of the same phoneme-idea ((m)), but the movement of the tongue when pronouncing them is different in each case. In the case of the previously mentioned [a], [a], [a], and [ae], the position of the tongue came closer to [a] when they were pronounced carefully. In the case of the [m]'s, however, no matter how carefully the words are pronounced, the position of the tongue remains different. The reason for this is that, in the phoneme-idea ((m)), the movement of the tongue is disregarded (p. 13 f.). 542 HATTORI SHIRO

In other words, the tongue, in the case of ((a)), is the "major organ of speech", but this is not so in the case of ((m)). Nevertheless, there is no mention of what criteria were used in classifying an organ of speech as a "major" one. A similar condition is seen in the case of [k]. Although it is transcribed uniformly in [ka], [ki], [ku], [ke], and [ko] in broad phonetic notation, these five [k]'s each vary in the point of articu- lation. On this point Arisaka makes the following statement: Basically, the question whether some particular sound is pronounced relatively carefully or in a careless manner, or whether some particular articulation (for ex- ample, the movement of the lips) is done with care or executed rather carelessly, is dependent on the prevailing general tendency according to the circumstances of the society in the particular locality and period of time. (This, in other words, is the reason for the occurrence of sound changes peculiar to a certain time and place in history). Accordingly, the range of fluctuation, when a certain phoneme is actualized, varies with the language of each locality and period. However, the phoneme of the k- group, when actualized, fluctuates over a wider range than does the phoneme of the t- group in almost every language of the world. This is probably basically due to the natural characteristics common to human beings throughout the world. First of all, the root of the tongue, compared with the tip of the tongue, is structurally less agile and accordingly minute adjustments cannot be made as quickly. The second reason may possibly have some connection with the size of the tactual space threshold at the place of articulation. (The tip of the tongue has one of the smallest space thresholds within the human body) (p. 18). As a reason for the k- group having a wider range of "fluctuation" than the t- group, the following should also be considered. The articulations of vowels and [k] both require the use of the upper surface of the tongue (the middle and rear of the tongue in the latter). Therefore, in order to let the [k] assimilate with a succeeding vowel, the easy way is to move the point of articulation. Be that as it may, the basic problem is the definition of the word "fluctuation". For example, the [k] in the Japanese [ki] is palatalized and the point of articulation is in the extreme front when compared with the [k] in [ka]. However, this does not mean that the former is palatalized by casual assimilation due to the influence of the succeeding [i] when aiming at the [k] of [ka]. Even in the most careful pronunciation of [ki], articulation is aimed primarily at the palatalized [k], and therefore the "phoneme-idea" must be a palatalized [k], different from the non-palatalized [k] of [ka]. Why then does Arisaka recognize these two as belonging to the same phoneme ((k)), saying that ((k)) "fluctuates" from a palatalized to a velar [k]?18 Similarly, why is the [m] in "miyage" [mijaeqe], which differs from the non-palatalized [m] in "matsu" [mats], considered to belong to the same phoneme ((m)), when the former is palatalized, and the "aim-idea" of the pro- nunciation is definitely a palatalized [m]?19 Hattori Shiro noticed these inconsistencies long ago and started developing a

18 Arisaka, Phonology, 76, writes kaki 'persimmons' as ((kaki)). 19 It is a perfect palatalized [m] in Hattori's pronunciation, but Arisaka makes the following statement: "((m)) in ((nami)), as a result of the influence of the succeeding ((i)), is usually actualized by raising the front of the tongue to some extent". (Phonology, 18). DESCRIPTIVE LINGUISTICS IN JAPAN 543 concept of "opposition" and "distribution".20 After the war, however, due to the influence of American phonemics, he developed a theory of his own.21 The following are the major elements of his theory: Principle of Systematic and Symmetrical Distribution (abbreviated to "Principle of Distribution") Principle of Simple and Symmetrical Structure (abbr. "Principle of Structure") Prinziple of Economy Principle of Environment Assimilation (abbr. "Principle of Assimilation"). Among these working principles he attaches the most importance to the last. The final decision as to the grouping of phones into phonemes is dependent on the Principle of Assimilation. The other principles provide only useful suggestions in the progress of research.22 The Principle of Assimilation means that: Phones, which have a complementary distribution, can be recognized as belonging to the same phoneme only when they can be explained from the viewpoint of pho- netics to be different forms of the same sound assimilated by different environments. According to this principle, the sound [h] and [q] in English are to be recognized as belonging to different phonemes, although they exist in complementary distribution. To prevent [h] and [q] from being regarded as belonging to the same phoneme, several linguists have stipulated the condition "phonetic similarity". However, the degree of similarity usually has not been clearly stated. By working with these principles we reach the following conclusion: the above mentioned [k] in [ki] and [k] in [ka], [m] in [mi] and [m] in [ma], exist in com- plementary distribution and do not oppose each other in the same environment; it can also be explained that they are forms of the same sound assimilated by their various environments; thus they can be recognized as belonging to the same phonemes /k/ and /m/, respectively. However, by persistently adhering to the principle of assimilation, there sometimes occur cases in which it must be recognized that some phones belong to different phonemes, even when many scholars consider them as belonging to the same phoneme. For example, in the Japanese language (in the Tokyo dialect) the [t] in [ta] and the [ts] in [tsu], are each considered to belong, according to this principle of assimilation, to different phonemes, /t/ and /c/ respectively. Similarly, [g] in [gakkol] 'school' and [q] in [Jolqakkol] 'primary school' are regarded as belonging to /g/ and /r)/. These two points concur with Arisaka's theory, but disagreements also occur. For example, Arisaka defines the Japanese [sa], [Ji], [/a], as ((sa)), ((Ji)), ((Ja)) contrary to Hattori's /sa/, /si/, /sja/, and Arisaka gives the Japanese [ta], [tji], [tsu], as ((ta)), ((tji)), ((tsu)), while Hattori interprets them as /ta/, /ci/, and /cu/, respectively. 20 Hattori Shiro, 1939-40 (Methods in linguistics, 220-39). 21 Hattori Shiro, 1951 and other papers published between 1954 and 1959 (in Methods in linguistics). 22 In addition to these, a further 'Principle of the Phonemic System' (ibid., 296) has been advanced as a working hypothesis. 544 HATTORI SHIRO

By applying the Principle of Distribution and the Principle of Structure Hattori interprets the Japanese syllables which "start with a vowel" as starting with the consonant phoneme /'/ (a voiced counterpart of the voiceless consonant phoneme /h/). He is also trying to apply this /'/ to other languages.23 By applying the Principle of Structure, and by assuming for example, that 'bed', 'tent', 'end', 'take', 'pound', and 'strike' in English have the same syllable structure /CV"C/, and also that a cluster of phonemes can stand in the position of both /C/ and /V/, he phonemically transcribes the shape of these words as /bed/, /tent/, /'eiicl/, /tejk/, /pawnd/, and /strajk/ (where ¡^ / denotes the cluster of phonemes). Furthermore, by setting up the concepts of a voiced laryngeal phoneme /,/ and of the cluster of phonemes, he claims that it is possible to abolish the juncture phoneme /+/> which is recognized by some American phonemicists.24 The following are examples: A common American transcription Hattori's transcription /naytreyt/ nitrate /nijtrejt/ = CtCVC /nayt+reyt/ night-rate /riijtrejt/ = CtfCC^C /nay+treyt/ Nye trait /ri^'trejt/ = Ct"CCVC /an+eym/ an aim /'an'ejm/ = CVCC^C /aneym/ a name /'anejm/ = CVCVC Accent in EngHsh is usually described as a type of phoneme ("secondary phoneme" or "suprasegmental phoneme") called stress which is superimposed upon a particular vowel phoneme. Accent in Japanese, however, since the days of Jinbo and Sakuma, has ordinarily been described as something superimposed upon a sequence of syllables (upon the syllable in the case of a monosyllabic autonomous word). For example, according to Sakuma,26 the sounds of the three words, hashi 'edge', hashi 'chopstick', and hashi 'bridge', as far as the phones are concerned, are the same [haji], but they are distinguished by the addition of the following different accents, "low-mid pattern", "high-mid pattern" and "low-high pattern". Similarly, Sakuma holds that the words sakura 'cherry blossom', hibachi 'brazier', otama 'dipper', and atama 'head', possess the following accent patterns respectively, "low- mid-mid", "high-mid-mid", "low-high-mid", and "low-high-high". In this way, the accent pattern is described as the sequence of "high", "mid", or "low" pitch on each syllable (strictly speaking "mora"). In opposition to the above, Miyata Koichi directed attention to the points in which the voice sharply ascends or descends. Noticing that the voice was bound to ascend between the first and second syllable if the first syllable (first mora) was "low", he concluded that the descent of the voice was the nucleus of the accent phenomenon in Japanese, and termed this the "accent kernel", proposing the following transcription:26 23 Hattori Shiro, 1961 (in English). 24 Compare further with Hattori Shiro, 1961 (in English). 26 Sakuma Kanae, 1919. 26 Miyata Koichi, 1927. DESCRIPTIVE LINGUISTICS IN JAPAN 545

Sakuma's terminology Miyata's transcriptions low-mid low-high high-mid low-mid-mid low-high-high low-high-mid high-mid-mid low-mid-mid-mid low-high-high-high low-high-high-mid low-high-mid-mid high-mid-mid-mid Miyata claims that the "low-mid" and "low-high" patterns, "low-mid-mid" and "low-high-high" patterns, and "low-mid-mid-mid" and "low-high-high-high" pat- terns are not distinguishable when words are pronounced in isolation. Miyata's transcriptions, which attempt to denote only the phonologically distinc- tive features, were very noteworthy, but Japanese academic circles of that time were still not well versed in phonological thinking and his theory was not accepted. Arisaka Hideyo distinguishes between the latent pattern and actual pattern. Has hi 'edge' and has hi 'bridge' are different as latent patterns, but as actual patterns they have the same "low-high" pattern when pronounced in isolation. When used with a particle, for example ga "nominative", the former has the "low-high-high" pattern, and the latter the "low-high-low" pattern. Thus, he advocates the two-level theory as opposed to the three-level theory advanced by Sakuma and others.27 Hattori Shiro attempted to introduce phonological considerations into the field of accent research and proposed the concept of "prosodeme".28 The prosodeme, like the "accent pattern", is something superimposed upon a sequence of syllables (or a syllable), but differs from it in the following respects: 1) The concept of accent pattern was derived from a consideration of pitch alone, whereas the concept of prosodeme encompasses not only pitch but all of the phonetic features, such as intensity, articulation, etc. The prosodeme is assumed to have the force to consolidate the sequence of syllables it covers. At the juncture of two proso- demes, there appears, in varying degrees, a change in one or more of the phonetic features mentioned above. For example, even if the pitch continues level, the phones in the initial position of a prosodeme often tend to be pronounced more strongly and articulated more clearly. 2) Accent patterns do not go beyond the description of the pitch level of each mora, whereas the prosodeme differentiates between distinctive and non-distinctive features.

27 Arisaka Hideyo, 1941. 28 Hattori Shiro, 1954 {Methods in linguistics, 240-72). 546 SHIRO HATTORI

For example, in the prosodemes of the Tokyo dialect, if we call the high (and intense) portion which immediately precedes a low (and weak) portion (or which has the power to lower and weaken the following portion) an "accent kernel", the distinctive features are whether there is an accent kernel or not, and if there is, where is it located. Accordingly, Sakuma's "low", "low-mid", "low-mid-mid" patterns, etc. should be called "kernelless prosodemes", while his "high", "high-mid", "high-mid-mid" patterns, etc., should be classified as "prosodemes with the kernel in the first mora", his "low-high", "low-high-mid", "low-high-mid-mid" patterns, etc., as "prosodemes with the kernel in the second mora", and the "low-high-high" "low-high-high-mid" patterns, etc., as "prosodemes with the kernel in the third mora". Usually, when the prosodeme with the kernel in the final mora is uttered in isolation, it is of the same tone as the "kernelless" prosodeme. The introduction of the concept of the prosodeme made possible the description of even such dialects as those of the southern part of the Tohoku region (North- eastern Japan), which had been called "dialects without accent" as having prosodemes between which there is no opposition. Hattori has also attempted to apply the concept of the prosodeme to foreign languages. He holds, for example, that the English words "differ" and "defer" are distinguished by the prosodeme only, that is, one of the prosodemes /CVCVC/ or /CVCVC/ (/'/ indicates the kernel of the prosodeme) is superimposed upon the same sequence of phonemes /difar/. In Russian the words napa 'pair' and napa 'time' are distinguished by superimposing the prosodeme /CV'CV/ or /CVCV/ upon the same sequence of phonemes /para/. He also considers French, Polish, Czech, etc., to possess prosodemes between which there is no opposition. Thus, there is no language in the world that does not have phonemes, and, accord- ing to Hattori every language is expected to possess prosodemes as well. Furthermore, Hattori now considers it necessary to posit the concept of a "second- ary prosodeme" (or "synnomous prosodeme").29 For example, formerly the words hanawa 'floral ring', hana-wa 'the nose (topic),' and hana-mo 'the nose also' in the Tokyo dialect have been described as having the same 'low-mid-mid' accent pattern. It is his opinion that while hanawa is to be described as possessing one three-mora kernelless prosodeme /ooo/, hana-wa and hana-mo should be described as having a two-mora kernelless prosodeme /oo/, suffixed by a one-mora kernelless secondary prosodeme /-o/. Similarly, the form kakuna 'don't wrie!' is to be described as having one prosodeme /obo/ (/7 indicates the kernel), while kaku-na 'you write, he writes', etc., which is similar in meaning to kaku-ne, should be described as having the sequence of prosodemes /o^-o/. Hattori derived the concept of the secondary prosodeme through study of his native dialect (that of Kameyama, Mie Prefecture), but feels it can be applied to other dialects and to foreign languages as well.30

29 First mention of relative phonemena was made in Hattori's 1947 paper (Methods in linguistics, 415ff.). 30 Hattori Shiro, 1961. DESCRIPTIVE LINGUISTICS IN JAPAN 547

Besides the above mentioned, there are other social habit features, that is, langue features, such as patterns of intonation, emphasis, etc. in the sounds of utterances, but these will not be discussed at this time. It is worth noting, however, that these features express a certain meaning whereas the phonological units explained in this chapter do not principally express meanings.

IV. GRAMMAR

1. Introduction

Our linguistic activities, we assume, are governed by a langue which lies latent in our neural systems; thus, particular features appear, repeated over and over again. How- ever, this kind of langue cannot at present be examined from the outside, and intro- spective examination too is virtually impossible; only when one part of the langue is reproduced can we conduct this introspective examination. Hence, in any study of the spoken language we begin our investigation by looking into an example of actual speech activity and whatever is brought about by that activity; that is, the psycholo- gical and physiological actions of producing an utterance and of understanding it, the sounds of the utterance and its "situation" (defined later) along with tape recordings and films, if these have been made. The investigator must, of course, draw as much information as possible from the speaker. Since any speech activity is a momentary happening, there is no time to think concurrently of a detailed analysis. Especially in grammatical study, it is customary first to transcribe the sounds of an utterance phonologically (this may be done in Japanese by means of kana), and then to proceed to an analytical study. In this case, the analysis is not very much different from the study of the written language. Not only is this so, but when the "situation" in which a particular utterance has been made is unclear, the analysis proceeds at greater disadvantage than when the written language is being studied. Since written material is produced from the beginning without expecting to rely on the "situation", so long as the writer and the reader possess the same langue, it is possible for the latter to understand a text simply by reading it. But a natural conversation usually takes place with the speakers dependent on the' 'situation''; hence, when one studies, independently of the situation, a text which is a transcription of a recorded utterance, one is at a clear disadvantage. Part of the difficulty in studying the spoken language resides in these points. We assume that the sounds of an utterance (more strictly, the langue features found in them) are made up of the phonological shapes of various words and signemes, along with intonation patterns etc., arranged in a particular order and unified. How- ever, we find various restrictions on the arrangements of these elements. In the study of grammar we examine one part of the rules relating to these restrictions. We call the functions of words and signemes appearing in certain restricted positions in a sentence 548 HATTORI SHIRO

their "textual functions", and we say that these words and signemes possess "situa- tional functions" as well. By the situational function of a word or a signeme is meant its function of being capable of indicating (or expressing) the relevant characteristics of the things and events (or of the content of the speaker's consciousness) in the situation of the utterance. 'Situation' has a broad meaning here; it includes not only the world external to the speaker, but what might be called his internal world, at the moment of utterance. The writer considers the textual and situational functions of each word or signeme to be two different aspects of the same thing; it is thus expected that there will be a one-to-one correspondence between them. Hence, in grammar and sememics, in contrast with phonology, it becomes necessary to study the textual and situational functions of words and signemes, that is, of units which are combinations of phonological shapes and meanings. In grammar it is the textual function which assumes the greater importance while in sememics it is the situational, but in order to achieve completeness we must study both in grammar and in sememics these two functions in their mutual relationships. However, in the case of a transcribed text of an utterance (this is also true of course of a text of the written language), we are dealing with something that has been separated from its situation, so that if we were to rely on analysis of the text alone, we could clarify only the textual functions of the linguistic units. Moreover, when we deal with limited texts, we can clarify only one part of their textual functions. On the other hand, linguistic units exist which have the same phonological shapes but different meanings—homonyms are the best examples; hence, if we do not understand the meanings we cannot even identify the linguistic units. Thus, when we are given only a limited text of an unknown foreign language, we are unable to decipher it, but in the case of our own native language we can easily understand and study it. This is a direct result of the fact that in the latter case, a langue exists within the student himself, so that when he reads the text there occurs within him a consequent sub- jective direct experience called the meaning. This advantageous phenomenon, which is found in the study of one's own language, should not be dismissed as mentalistic, but should be utilized to the fullest degree possible. Uncritical utilization, however, is dangerous in that it tends to make the research overly subjective. In order to increase the objectivity of the results it is necessary to make use of this phenomenon, adequately, consciously, and with great caution. As has already been explained, descriptive grammar in Japan, in the first stage of its development in the Edo period, took for its materials the texts of the old classical language. By means of a tightly controlled methodology scholars examined the textual functions of words and of signemes with excellent results. At the same time they made use of all possible methods and materials in an effort to grasp (nacherleben) the original meanings intended by the authors of the texts. This also contributed to the success of these studies. During the Meiji period scholars began to study their own spoken language, which was much easier for them to analyze because the meanings the texts represented were DESCRIPTIVE LINGUISTICS IN JAPAN 549 obvious to them, and there was less need for painstaking research. At the same time the study naturally tended to be less careful, and minute examination of the textual functions of words and signemes was often neglected. Generally speaking, Japanese scholars never went to such an extreme anti-mentalism as to reject the introspective examination of meaning; but at the same time, owing to the traditional tendencies existing since the Edo period, they never became so mentalistic as to disregard the textual functions of words and signemes. Scholars differ to some extent on these points; for example, Yamada Yoshio greatly promoted grammatical study by his analyses of meaning, and he also gave considerable attention to textual functions, but did not attain sufficiently precise results. Hashimoto Shinkichi con- ducted a very precise study of textual functions of words and bunsetsu (syntagma), but did not work in the field of minute analysis of meaning. Tokieda Motoki, by reviving the theory of Suzuki Akira (1764-1837), has tried to give a new direction to the study of meaning, but he does not consider in detail the textual functions of words. On the other hand, just as it is important in the study of phonemes and prosodemes to clarify the systems they form, it is also important in grammatical research to study the systems which the units in question form and the relationships between the units. Studies of conjugation and types of sentences have been made from this point of view but more precision is necessary. This is especially true in the case of the study of the interrelations between taxemes (and sentence-patterns)—including the interrelations between major and minor sentences—where these interrelations have been revealed to some extent but not thoroughly enough. In this sense, the recent attempts to apply American generative grammar to the Japanese language should be welcomed. How- sver, in order to insure that they are incorporated into the traditional grammatical etudy of Japan, they should be conducted in close connection with the precise study of meaning.

2. Sentence, Composition, and Utterance

Grammar is ordinarily expected to deal with the structure of "sentences", but not with that of a text consisting of two or more sentences. Yamada Yoshio, for in- stance, states in his Japanese grammar (1908): The upper limit of syntax is the description concerning the structure and use of simple and complex sentences. With respect to a text expressing one thought by means of a combination of simple and/or complex sentences, grammar studies each sentence, so far as the sentences are formally independent of each other, but it does not deal with the text as a whole (p. 1213). Tokieda Motoki, on the other hand, advocates as a branch of grammar a concept bunshorori, (study of compositions) in opposition to goron (study of words) and bunron (study of sentences).31 By bunsho (composition) he means a unified product (mainly 31 Tokieda Motoki, 1950. Mio Isago, in a 1948 book, points out the fact that not only two or more sentences in one utterance, but also short utterances exchanged between speakers, are connected in one situational context. 550 HATTORI SHIRO in the written language) consisting of one or more sentences (thus, even such lengthy works as essays, novels, etc., would be considered burisho). A field of study that deals with the structure of such linguistic products would certainly be useful, but it is not usually considered as belonging to grammar and will not be touched upon here. It is true, however, that some conjunctions link sentences; further, two or more sentences may succeed each other in one context, in terms of grammatical concord and abbrevi- ation of constituent elements. Therefore, grammar cannot be defined as a field which never deals with the textual relation between sentences. Now, what is a "sentence"? In Japan, as in the West, numerous scholars have tried to answer this difficult question. European grammars seem to have given an impetus to Japanese scholars to consider the problem of the definition of a "sentence". For example, Otsuki Fumihiko defines it as follows in his Detailed Japanese grammar (1897): A bun or bunsho (a sentence) is a sequence of words written in letters or characters and expressing a complete thought. If the expression is not complete, it is called a ku (a phrase) (p. 251). That which has a subject and a predicate is bun. A bun must have a subject and a predicate (p. 252). Criticizing this kind of definition Yamada Yoshio writes (Japanese grammar, p. 1182): In summary, the definition of a sentence as something that includes a subject and a predicate is obviously wrong, because a subject and a predicate presuppose a sentence and they are the result of the analysis of a sentence. To use the concepts of subject and predicate before the sentence is defined is tantamount to using the conclusion of the explanation of a thing in the explanation of the thing. Moreover, this definition fails to account for the existence of sentences without a subject or predicate. Referring to what is called "a word sentence" by H. Sweet and "Satzaquivalent" by W. Wundt, Yamada writes (ibid., p. 1183f.): When a word has reached the status of a sentence or a sentence equivalent, it is no longer a mere word, but has taken on qualities which activate it internally. If we see only its external shape, it is nothing more than a word and we must call it simply a "word". Seen from the outside, even what is called a complete sentence is nothing but a mere accumulation of words.32 Hence it is evident that in what is called a sentence there must be some internal activating factor. This internal element of a sentence is the "thought". Now, let us make a study of this thought. We have defined a sentence as a complete expression of a thought; a simple sen- tence is an expression of a simple thought. What we call "thought" is an active state of the consciousness of man, consisting in the unification of various ideas. There must be only one point of unification, because the focus of consciousness is one. Hence, in one thought there is one unifying action. We call this tokaku-sayo (apper- ception). 32 Concerning the point that the "shape" of a "sentence" is something more than the sum of the shapes of its constituent words, see p. 558 of this paper. Yamada gives us the impression that he thinks only of the sequence of printed words in written texts. DESCRIPTIVE LINGUISTICS IN JAPAN 551

Yamada gives the following definition of a sentence (ibid., p. 902): it is an expression of apperceived thought by means of a form called language. On the other hand, he defines his term ku (clause) as follows (ibid., p. 917): A clause is a linguistic expression of thought which has been organized by one apperceptive act. Thus, Yamada distinguishes bun and ku as follows (p. 1170): German English bun Aufsatz sentence ku Satz clause Elsewhere (p. 1174), he writes: Here we say that the basic elements of sentences are clauses; all sentences are com- posed of clauses. Analysis of a sentence should not go beyond analysis of clauses. Thus, we call that which consists of one clause a simple sentence and that which consists of two or more clauses a complex or a compound sentence. Furthermore, he broadly divides clauses into two classes (p. 1217f.): We have divided clauses into two categories: clauses of predicative form and clauses of exclamatory form... The difference between them is that the latter have a kokaku (a vocative word) as their kernel, while the kernel of the former is a jukkaku (a pre- dicative word). According to Yamada, clauses of predicative form "include declarative, interrogative, and imperative sentences and exclamatory sentences of declarative form" and clauses of exclamatory form include "the other exclamatory sentences" (ibid., p. 1198).33 He writes of clauses of exclamatory form: As for their form, they always consist of one taigen (nominal) as the kernel, with or without the modifier(s). That is, they do not have the structure, subject(s) plus pre- dicate^), but have only one nominal in the vocative case (ibid., p. 1218). As examples, he gives the following (ibid., p. 1219 f.): Shiritaru hito mo gana! 'acquainted people if I had' = 'I wish I had some acquaintances!' Uruwashiki tsuki kana\ 'beautiful moon what a' = 'What a beautiful moon it is!' Tsuki no kage no sayakesa! 'moon s 'light's clearness' He does not place the following sentences in this class, although they are usually called optative or exclamatory sentences: Tori mo nakanam! 'birds just I wish ... would sing' = 'I just wish birds would sing'! Ware wa atatakaki koyoi no yume ni iram kana\ 'I (topic) warm tonight 's dream into I wish to enter' = 'I wish I could fall into a warm dream tonight!' 33 Concerning the various classifications of sentences, see Mio Isago, 1948. 552 HATTORI SHIRO

In Yamada's classification, these sentences belong to the category of clauses of predicative form because they have the structure "subject plus predicate" (ibid., p. 1222, p. 1228). Introducing the concept of chiñjutsu (predication, statement), Yamada coined the term chiñjuísugo (predicational) for yogeñ (conjugationals) and gaineñgo (concep- tional) for taigeñ (nomináis), and defined the former as "a word with the force of chiñjutsu". Yogeñ include verbs and adjectives (among others) while taigeñ include nouns, pronouns, and numerals (ibid., p. 156). However, he does not seem to distinguish clearly between chiñjutsu and the tokaku-sayo (apperception) mentioned above. For instance:

Both [verbs and adjectives in Japanese] have the force of chiñjutsu, and for this reason they belong to the category yogeñ (ibid., p. 163). Yogeñ express the properties of the taigeñ which precede them and at the same time express the toitsu-sayo (unifying action) of human thought (ibid., p. 233). We have mentioned that yogeñ express both hiñi (the predicate) and the copula (tokaku-sayo = apperception) at the same time (ibid., p. 497).

If chiñjutsu is equivalent to tokaku-sayo as would seem to be the case if we judge from the statements quoted above, then we are forced to the conclusion that Yamada's clauses of exclamatory form should comprise chiñjutsu too. However, his following statement contradicts this:

The best examples of clauses of predicative form are declarative sentences. In such sentences, the separation of the subject from the predicate necessarily occurs as a result of the thought process. These two elements are again unified by tokaku-sayd. Such a sentence differs structurally from a sentence of exclamatory form, which is an indivisible whole (ibid., p. 1211). He seems to mean here that clauses of exclamatory form do not include any tokaku- sayo, that is, chiñjutsu. A long series of discussions has taken place concerning chiñjutsu, and scholars have come to use this term in various meanings. We shall discuss this point at greater length later. It is noteworthy, however, that Yamada Yoshio, the coiner of the term chiñjutsu, is himself inconsistent, using it ambiguously as was explained above. However, Yamada's analysis of the meaning of yogeñ has contributed in many ways to the study of sentence structure, and it includes some points which can be accepted even today.34 Therefore, some passages relating to his theory of chiñjutsu will be quoted here (1922a, p. 677f.):

To conduct chiñjutsu, viewed from the aspect of thought, means to reveal the rela- tion between the two elements, subject and predicate, and to express by means of speech the internal operation of thought which has determined whether or not the subject and the predicate are to be united. The linguistic form which expresses only chiñjutsu is called the "copula" in logic The only yogeñ of Japanese which are

34 See fn. 55. DESCRIPTIVE LINGUISTICS IN JAPAN 553 copulas are soñzaishi (words of being), namely, nari, tari, de aru, da, desu, etc. and verbs, which we shall call jisshitsu-yógeñ (substantial conjugationals), comprise in an indivisible form the two elements: property (the substantial element) and copula (the formal element). Usually when grammarians say that yogeñ conduct chiñjutsu concerning the subject, they note only the fact that the property which is the substantial aspect of yogeñ is directed toward the subject. ... Now, it is evident that this internal aspect of the copula, that is, the force of chiñjutsu, is directed to- ward the contrast of the subject-idea and the predicate-idea, not toward the subject- idea alone. This relationship is illustrated in the following diagram:

subject -,

•copula -predicate-

As was explained previously, the relationship between Yamada's chinjutsu and the end of a sentence is not sufficiently clear, but his statements gave scholars the im- pression that chinjutsu had some relation to the termination of a sentence. For this reason grammarians argued about chinjutsu for many years.35 As it is impossible to refer to all of the papers on this subject, mention will be made only of those which have contributed to the analysis of the structure of the final part of the sentence. Watanabe Minoru, in a 1953 paper, considers only the underlined part of the following sentences as jutsugo (predicate):

Kyonen no konogoro ga natsukashii ne. 'last year's this time (nominative) longed-for (exclam.)' = 'How happy I was at this time last year!' Ashita kara tabako o yameru yo. 'tomorrow after tobacco (accus.) I will stop (I'll tell you)' = 'I'm going to quit smoking tomorrow'. Aitsu wa sugaku no tensai da wai. 'He (topic) mathematics 's genius is (exclam.)' = 'He's really a mathematical genius!'

Watanabe uses the term jutsugo bunsetsu (predicate syntagma) for the forms natsu- kashii-ne, yameru-yo, and tensai-da-wai, which are terminated by the shujoshi (final particles) ne, yo, and wai. He points out the difference in function between jutsugo and shujoshi:

By jojutsu (description) we mean the speaker's activity in describing the content of thoughts or things and events (p. 20). [The role of jutsugo ] is to finalize the action of jojutsu and to give completeness to its content; these are the positive characteristics of jutsugo (p. 21).

35 Mio Isago, 1939, Oiwa Masanaka, 1949, Fukuda Yoshisuke, 1959, Okubo Tadatoshi, 1964, Kawabata Yoshiaki, 1965. 554 HATTORISHIRO

Concerning shujoshi, he writes: The positive characteristics of shujoshi are that they are particles which are used when the speaker especially wants his listener's attention, or when he is attempting to persuade or convince himself of something (p. 20). Shujoshi are particles which neither increase nor decrease the content of jojutsu (description), and which have nothing to do with the action of jojutsu (p. 20). By chinjutsu as distinct from jojutsu, I mean the subjective activity expressed by shujoshi which is directed toward the listener or the speaker himself (p. 27). Saying that shujoshi possess the function of "completing a sentence" (p. 26), he writes: When chinjutsu is carried out by shujoshi, which governs the objective material of jojutsu, the sentence is completed (p. 33). Haga Yasushi, in a 1954 paper, further developed Watanabe's theory as follows: I define chinjutsu as "the activity of the speaker in synthesizing and completing a sentence" (p. 253). I consider chinjutsu to be located generally only at "the end of a sentence" (p. 254). The only words with chinjutsu are "the words which stand always at the end of a sentence" (p. 254). In these points Haga's opinion is similar to Watanabe's, but Haga deviates from the latter in distinguishing two kinds of chinjutsu. Shibai demo hajimaru ka net 'play or something begins (doubt) (exclam.). = 'Is there going to be a play or something?' In this sentence, shibai demo hajimaru represents jojutsu (description), while ne expresses "the appeal to the hearer". But the ka between the two forms belongs to neither category, but expresses the speaker's subjective doubt {modus) which is directed to the material (dictum) indicated by shibai demo hajimaru. Thus he recognizes two distinct types of chinjutsu: one which is carried out by joshi (particles) and jodoshi, expressing the speaker's subjective modus, and another which is carried out by shujoshi (p. 252): [1 ] The first kind of chinjutsu is a categorical expression of the speaker's attitude [affirmation, conjecture, doubt, decision, admiration, exclamation, etc.] towards the content of the matter represented objectively by means of the forms immediately preceding it. (However, this objective representation is lacking in the case of a sentence consisting of one interjection). ... I would call it jutteiteki chinjutsu or juttei (expression). [2] The second kind of chinjutsu comprises linguistic expressions which serve to bring up and to communicate the content of a matter or the speaker's attitude to the hearer (or sometimes to the speaker himself). Examples of this type of chinjutsu are notices, requests for a response, invitations, commands, responses, expressions used in addressing someone, etc. We may call this type dentatsuteki chinjutsu or simply dentatsu (communication). DESCRIPTIVE LINGUISTICS IN JAPAN 555

By showing such examples as : Ame ga furu +0 + yo. 'rain (nominai.) falls (affirmation) (notice)' = 'Say, it's going to rain', Ame ga furu darò ne. (conjecture) (address) = 'It looks like rain, doesn't it?' he says juttei forms can be followed by dentatsu forms, but that the reverse is im- possible; and he concludes: juttei always precede dentatsu (p. 253). In this way, the concept of chinjutsu, which was originally rather ambiguous, has been analyzed and clarified by succeeding scholars. Nevertheless, it is impossible to conclude that chinjutsu necessarily signal the end of a sentence. For instance :

Ame ga furu darò ne to itta. that he said = '(He) said, "It looks like rain, doesn't it?" '

In the above sentence, ne, which expresses chinjutsu (Haga's dentatsuteki chinjutsu), is not located at the end of the sentence. Therefore, we must seek the sign of the end of the sentence in something else. Otsuki Fumihiko and Yamada Yoshio analyzed texts of the written language and consequently did not consider phonetic characteristics when they defined the sentence. But there were some scholars who did pay attention to the phonetic aspect. For instance, Kusano Kiyotami gives the following definition in his Japanese grammar (1901):

A sentence begins when one starts to utter a word and ends when both the intonation and the meaning terminate, (p. 249). He considers both aspects, sound and meaning. Yasuda Kiyomon also notes the continuation and cessation of sound in a sentence in his book published in 1928:

Seen in terms of its constituents, that is, words, a sentence must always stop.... A sentence may consist of only one word, but in most cases it is complete when sound ceases after a continuation of several words (p. 303). Miyake Takeo, in his 1934 book, shows six basic patterns of kotoba-chòshi (intona- tion), that is, goki (tone of voice) : (1) affirmation, (2) negation, (3) interrogation, (4) doubt, (5) exclamation, address (a), (6) address (b), and says goki performs a very important function in the expression of chinjutsu (p. 4). This is a noteworthy opinion, because it points to the fact that intonation expresses various meanings. It also suggests that intonation signals the end of a sentence. However, Miyake admits that particles used at the end of a sentence such as ka, yo, ne, sa, and wa, verbs, adjectives 556 HATTORISHIRÓ and the copulas da and desu also express chiñjutsu. Hence, he does not seem to think that intonation is the only sign of the end of a sentence. Hashimoto Shiñkichi's definition of a sentence in his 1934 book clearly includes the phonetic aspect:

A sentence, seen from the aspect of content or meaning, is an independent [and complete] expression of a matter and a complete and coherent unity (p. 4). He gives the following as the external characteristics of a sentence (p. 6) : 1. A sentence is a sequence of sound. 2. There are always pauses before and after a sentence. 3. A particular intonation is given to the end of a sentence.

This marks a great advance in the definition of the sentence. Mio Isago, in a 1939 article, divided the formerly vague concept of chiñjutsu activity (= toitsu-sayo 'unifying activity') into two concepts: toitsu-sayo (unifying activity) and dañtei-sayo (predicative activity); toitsu-sayo precedes dañtei-sayó, he claimed (p. 76). In a 1948 book, he tried to define the sentence from a new viewpoint. He claims that the sentence may be considered from three structural viewpoints: (1) logical structure, (2) morphological structure, and (3) dynamic structure. Concerning the third he writes: In a sentence such a force, which is a direct factor in changing the course of speech, is working continuously but with constant fluctuation. The sentence ends where this force is cut off. The area where there is a continuation of this force is a sentence (p. 182). Characterizing the sentence as "a dynamic structure where such forces are working upon each other and preserve an equilibrium", Mio gives examples of these forces: When watashi-wa 'l-(topic)' is uttered, it has a persistent power which continues until the sentence is made complete by an answer to such questions as nani da 'am ...' or do shitañ da 'have done ..., 'etc. (p. 182). "Kesshite '(n)ever' ... nai 'not'", "Naze 'why' ... ka (particle of doubt)", "Moshi 'if' ... ba 'if'", "Yoshi 'even if' ... temo 'even'", etc. are examples of words of concord, which have great dynamic force. The force of the beginning word never ceases until the final word appears (p. 183f.).

Concerning the "function of buñsetsu (syntagma)", Mio discriminates between two kinds of buñsetsu :36 a) those which "have the power of requiring some other buñsetsu to follow them or of terminating the sentence and b) those which "have no such power", namely, taigeñ (nomináis). This theory has some resemblance to Hashimoto Shiñkichi's "buñsetsu sequence" theory, but was apparently brought forward indep- endently.37

36 Concerning buñsetsu, see IV, 3 of this paper. 37 Hashimoto Shiñkichi's lecture "On the structure of the sentence in terms of buñsetsu" was delivered on September 3, 1944, and printed in Kokugogaku Nos. 13/14 (1953). DESCRIPTIVE LINGUISTICS IN JAPAN 557

As mentioned in Chapter II, 3, Hattori Shiro has proposed a definition of "the utterance" as a unique speech act and the sound produced thereby. He defines "the linguistic product" as the social habit features (that is, the langue features) found in the act and the sound. He also says that a linguistic product consists of one or more sentences. Thus his terminology includes: utterance sentence sound meaning utterer His definition of the sentence follows:38 Considered from the aspect of shape, its end is characterized by an intonation pattern. Considered from the grammatical aspect, it is an independent unity in which all parts are finally unified by a sentence pattern (which is a kind of taxeme). Considered from the aspect of sense, it is complete. ... When the utterances uttered by various utterers correspond to the same sentence, we may posit an "ex- presser of the sentence" who is independent of the utterers. Theexpresser of a sentence is usually the utterer, but this is not always the case ... The expresser remains in the background until the end of the sentence, at which time he always comes forward. In the utterance which corresponds to the sentence quoted previously, Ame-ga furu dará ne, to itta 'He said "It looks like rain, doesn't it?"dard-ne has an intonation different from that which marks the end of a sentence, so that it is clear that the sentence does not terminate with this form, even though there may be a short pause after it. From the viewpoint of sense, also, the expresser of the sentence does not come forward at the end of daro-ne but makes his appearance only at the end of to itta '(He) said that ...'. Considered from the viewpoint of syntactic unification, ame- ga furu daro-ne is united with to 'that' and this unified phrase is in turn united with itta 'said', thus constituting a complete independent unity. In the case of lies or sarcasm, the utterer of the utterance is distinct from the expresser of the corresponding sentence. For instance, an utterer who knows that something is not there may utter the sentence Asuko ni aru 'It is there' in order to deceive the hearer. In this case, Hattori claims, the utterer is saying one thing while the expresser is saying another. According to Hattori, the utterance is an indivisible continuum both in sound and in meaning, whereas both the shape and sense of the sentence can be analyzed into discrete elements. However, a sentence is not a mere sequence of words; its constituent elements are united by various taxemes and eventually by a sentence pattern; and finally, an intonation is superimposed upon it. For example, 'this book-(topic) very is interesting' Kono hoñ-wa taiheñ omoshiroi.

= 'This book is very interesting'. 38 A 1957 paper (Methods in linguistics, 197f.). 558 HATTORI SHIRO

In the above sentence, kono and hon are united, then kono hon and wa are united; next taihen and omoshiroi are united, and the process is completed by the unification of kono hon-wa and taihen omoshiroi. The intonation is superimposed upon the whole form. (Moreover, some particular part(s) of the sentence may receive special emphasis). All of these constituent elements, that is, words, taxemes, intonations (and emphases) have both a shape and a sense. The shape of a taxeme manifests itself in the various phonetic characteristics of the sequence of sound when the unified forms are pronounced together in the proper way. The opposition between the following two sentences (1) and (2), both of which have the same constituents other than the taxeme, clearly illustrates the fact that a taxeme has a sense.

(1) Boku-wa unagi-da. 'I-(topic) eel am' = 'I am an eel' (personification of an eel, as in a children's story); (2) Boku-wa unagi-da = 'I am going to eat some eel'. Inasmuch as the constituents (words and intonation) of these sentences are the same, the difference in sense must be attributed to the difference in the sense of the taxemes (in this case, the sentence patterns). These two taxemes, however, are called homo- phonous, because they have the same shape39. Kawakami Shin has been conducting analytical research in the intonation patterns of the Tokyo dialect (1961, 1963), and the reports of the National Language Research Institute, Nos. 8 (1955), 18 (1960), and 23 (1963) are devoted to the study of the intonation and sentence patterns of spoken Japanese.

3. The Word and Buñsetsu (Syntagma)

Yamada Yoshio was the first Japanese scholar to try to define explicitly "word". In A Japanese grammar (1908), he writes.40 A word is an indivisible minimal unit of thought in language, which independently represents some thought (p. 76). He holds that shiroshi '(to be) white' consists of two units from the psychological and logical viewpoint, because it denotes the concept of the optical property of a thing

39 The English versions of the relevant papers are: 1956, 1957 (both in Methods in linguistics), and 1961. See also Proceedings of the Ninth International Congress of Linguists (1964), 278f. Mio Isago's "dynamic force" has some resemblance to Hattori's "taxeme" but differs from it. For Mio, shiroi 'white' in yuki ga shiroi 'snow is white' has the dynamic force to finish the sentence, whereas shiroi in shiroi yuki 'white snow' has the dynamic force to continue to the following form. For Hattori the two shiroi are the same form which has no force to finish or to continue, but in the former sentence it is in the position of the predicate of the taxeme ((subject + predicate)), whereas in the latter it is in the position of the modifier of the taxeme ((modifier + modified)). 40 On the same page Yamada quotes H. Sweet's definition of a word, which influenced his own definition. DESCRIPTIVE LINGUISTICS IN JAPAN 559 and the apperception of the human mind, but it is one unit "linguistically". On the other hand, ume no hana 'plum blossom' is one concept from the viewpoint of thought, but it consists of three words ume 'plum', no 'of', and hana 'flower', (p. 76f.) Yamada calls the above "an analytical explanation" of words, but he continues:

When we consider the meaning of words from a synthetic viewpoint ... words depend upon one another as the direct constituents of a sentence and must be the entities from which the sentence is constructed, even if their functions vary (p. 77).

Stating that "roots" and "affixes" are not words, he writes:

Roots denote some thought, but do not directly become the elements for constructing a sentence. They are the elements for constructing a word and indirectly become the elements for a sentence (p. 77 f.). Affixes are not words, but adjuncts to them. They are similar to roots, but differ from them in that they are attached to a word and construct another word, adding to it some kind of meaning (p. 78).

Yamada's examples of words agree to a great extent with what Hattori considers "words". (This will be discussed at greater length later). But Yamada's definition of a word is abstract and not sufficiently clear, lacking explanation concerning the phono- logical shape. Hashimoto Shinkichi proposed a concept bunsetsu, in an attempt to find also in phonological shape a clue to the definition of "word". Bunsetsu corresponds to Jinbo Kaku's kuu and Matsushita Daisaburo's shi (strictly tanshi).42 According to Hashimoto (1934), there can be no more than the seven pauses indicated by vertical lines when the following sentence is actually spoken. Watashi-wa | kind | yujin-to | futari-de | Maruzen-e | hon-o | kai-ni j ikimashita. 'l-(topic) yesterday friend-with two persons-together Maruzen-to book-(accus.) to buy went' = 'I went to Maruzen with a friend to buy a book yesterday'. "The smallest parts of a sentence which can be uttered separately from each other in actual speech" are termed bunsetsu by Hashimoto (p. 6). He says that "every bunsetsu has a certain fixed shape and meaning" (p. 7), and enumerates the following four points as the characteristics of the shape of bunsetsu (p. 8f.):

1. It is a sequence of syllables in a certain fixed order, which is always uttered un- interruptedly (that is, with no pause between its parts). 2. The relative pitch of every constituent syllable of the bunsetsu (that is, its accent) is fixed. 3. In actual speech it is possible to pause before and after a bunsetsu. 4. There are definite rules governing the various positions (initial, medial or final) in which sounds may occur within a bunsetsu.

41 Jinbo Kaku, 1922, 83-88. 42 Matsushita Daisaburo, 1928, 19. 560 HATTORI SHIRO

It is a definite advance to try to define linguistic units also from the aspect of shape. Having thus obtained the concept bunsetsu, which does not coincide with the usual concept of "word", Hashimoto further tries to define "word":

Bunsetsu may be further analyzed into linguistic units which have a meaning, that is, bunsetsu consist of words. Indivisible bunsetsu consist of one word (p. 11). His examples of bunsetsu follow: yama, kawa, yuku, shiroi, omoku, a (one word bunsetsu) 'mountain' 'river' 'go' '(be) white' 'heavily' 'oh' yama-no, kawa-o, yuku-to, omot-ta, yuka-nai, omoku-te 'mountain's' 'river-(accus.)' 'go-when' 'think-(perf.)' 'go-not' 'heavy-and' (two word bunsetsu) yama-ni-wa, kawa-sae-mo, yuko-o-keredomo, omowa-se-tai 'mountain-on-(topic) 'river-even -too' 'go-will-but' 'think-(caus.)-want' (three word bunsetsu) omowa-re-mase-n-desho-o-yo (seven word bunsetsu) 'think(honor.)-polite-not-be(polite)-may-(particle)'

= 'I don't feel that he can thinks so'. Thus, Hashimoto divides "words" into two classes (p. 12ff.): 1. words which can form one bunsetsu by themselves: yama, kawa, yuku, shiroi, a, etc. 2. words which cannot form one bunsetsu by themselves, but are always attached to words of the first class to form a bunsetsu-. -no, -o, -to, -ta, -nai, etc. He says that the characteristics common to these two classes of words are (p. 14f.): 1. The word is a unit which constitutes a bunsetsu. 2. Every word has its own fixed meaning. 3. The syllables and accent which constitute the shape of a word are usually constant, although they may change to some extent when combining with other words to form a bunsetsu. 4. A word is always uttered uninterruptedly. 5. When a word combines with other word(s) to form a bunsetsu the initial or final sounds of that word are not regulated by the rules for the initial or final sounds [of bunsetsu].

Compounds such as sakadaru 'wine barrel', amagasa 'umbrella', etc. are, he says, to be considered "words", but the components saka, daru, ama, gasa do not have the independence of such autonomous43 words as sake 'wine', taru 'barrel', ame 'rain', and kasa 'umbrella'. Hashimoto gives four reasons for his conclusion that these compounds "are only fractions of new words" (p. 16 f.). Affixes such as o-, -tachi, -meku in otera 'temple (polite)' (tera 'temple'), watashi- tachi 'we' (watashi 'I'), harumeku 'become springlike' (haru 'spring') are simila r to

43 Words or forms which can be uttered alone. DESCRIPTIVE LINGUISTICS IN JAPAN 561 the words of the second class in "that they are synnomous (dependent) units and form autonomous units by attaching themselves to an autonomous word"; he claims, however, that affixes differ from the words in class 2 in the following points:

Units which are formed by the addition of an affix or affixes are not to be regarded as bunsetsu but as "words", because they form various bunsetsu in the same way as "words" do. On the other hand, units which are formed by the addition of class 2 words, such as yama-ga 'mountain-(nominative)', yama-wa '-(topic)', yama-kara '-from', yama-ka '-(inter.)' never form bunsetsu by the addition of various [synnomous] words in the way [autonomous] words do. Hence, they are to be regarded as bunsetsu, not as words which can form bunsetsu (p. 21).

However, forms such as yuka-nai 'go-not', yuki-masu 'go-(polite)', it-ta 'go-(past)', yuko-o 'go-will', yama-da 'mountain-is' which have a suffixed jodoshi ("auxiliary verb") function in much the same way as the words yuku 'go' or shiroi 'white', so that: these units should be regarded not as bunsetsu but rather as words. Consequently, these jodoshi... should be considered suffixes, not words. From this point of view, it is reasonable that Yamada Yoshio regarded jodoshi as endings of yogen (verbs, adjectives, etc.) and yogen suffixed by jodoshi as one word (p. 22).

Nevertheless, Hashimoto made a concession to the tradition of regarding jodoshi as words for the following reasons:

The number of words to which a given affix may attach itself is rather limited. But class 2 words, that is, joshi (particles = enclitic words and suffixes) and jodoshi, can usually attach themselves to any one of a large group of words. ... In general they may be added to any word of a given class, that is, to taigen (nominals) or yogen (verbs, adjectives, etc.) (p. 23). Hashimoto terms class 1 words shi and class 2 words ji and considers joshi and jodoshi as belonging to ji.

Hattori Shiro has noted the fact that bunsetsu are not always uttered uninter- ruptedly; some enclitic words, which are the constituents of bunsetsu, are sometimes uttered separately from the forms to which they are suffixed. Taking into consider- ation examples from various languages he has proposed the following three criteria for distinguishing synnomous words (that is, free forms) from affixes (that is, bound forms):44

I. Those forms which are attached to various autonomou43 forms having different shokuno (grammatical textual functions) and/or different inflections are free forms, that is, synnomous words. II. If another word or other words can be freely interposed between the two connected forms, both are free forms, and accordingly the form in question is a synnomous word. III. If the two connected forms can be transposed, both are free forms.

44 A 1950 paper (Methods in linguistics, 461 f.). 562 HATTORI SHIRO

(Principle II may still require refinement in some points). Principle I can be applied, for example, as follows. The joddshi da (copula) is a synnomous word, because it can be suffixed to various autonomous words which have different shokuno, as in hon-da Qion 'book'), yomu-no-da (yomu 'read', no (nominalizer), shiroi-kara-da 'it is because it is white' (shiroi 'be white', kara 'because'), shizuka-da (shizuka 'quiet'), etc., whereas -nai, -masu, -ta, and -o, which are usually called joddshi are not synnomous words but suffixes because they can each be suffixed only to a particular verb con- jugational form which is bound and not autonomous. Hattori does not believe, however, that non-autonomous forms of this kind can be placed in two distinct groups, namely synnomous words and affixes, for he thinks that there are various degrees of synnomousness or boundness.45 It is noteworthy that forms which are "words" in terms of Hattori's criteria pretty nearly coincide with Yamada Yoshio's "words". Recently, it has become clear that most of the forms which are considered synnomous words have their own distinct synnomous prosodemes (that is, secondary prosodemes). For instance, the bunsetsu ame-made 'even rain' of the Tokyo dialect has usually been described as having a prosodeme with the kernel on its first mora a, but it has become evident that its shape has a combination of a prosodeme with the kernel on a and a synnomous prosodeme with the kernel on ma.46

4. Classification of Words

Although the scholars of the Edo period did not give clear definitions of such concepts as "word" or bunsetsu, they isolated some units (which correspond to what are today called autonomous and synnomous words, affixes, etc.) through intuitive analysis, and made a general classification of them in terms of their textual functions and meanings. For example, Fujitani Nariakira (1738-79) divided these units into four classes: na (nouns), yosoi (verbs and adjectives), ayui (suffixes and enclitics), and kazashi (prefixes and proclitics); and Suzuki Akira (1764-1837) set up four different classes: tai no kotoba (nominals), te-ni-o-ha (particles), arikata no kotoba (adjectives), and shiwaza no kotoba (verbs). Tsurumine Shigenobu's Gogaku shinsho published in 1833, in the late Edo period, tried to apply the word classification of Dutch grammar to the Japanese language. The grammars of the early Meiji period too are under the strong influence of European grammar. The classification of Otsuki Fumihiko's A detailed Japanese grammar (1893) still shows the same influence. In Yamada Yoshio's A Japanese grammar (1908) we find the first attempt at a detailed classification of words in terms of the structure of Japanese itself. An outline of Yamada's classification (ibid., pp. 156, 159) and Otsuki's follows:

45 Methods in linguistics, 487 ff. 46 Cf. chapter III, 546 of this paper. DESCRIPTIVE LINGUISTICS IN JAPAN 563

Yamada Otsuki

frelation-words enclitic particles te-ni-o-ha words ^dependent w. -fukushi adverbs, conjunctions, and interjections /concept-w.—taigen nouns

independent w.

chinjutsu-w.—yd gen adjectives, verbs, and "auxiliary verbs"

It is interesting to note that Yamada's classification is very close to that of Fujitani Nariakira. Yamada attaches more importance to meaning as a criterion, but he also devoted considerable attention to the textual function, so that his subclassifications of his four classes are full of excellent ideas. Yamada's classification was the result of his analytical study of the written language, but he gives the same classification in A lecture on the grammar of spoken Japanese (1922). Matsushita Daisaburo proposed a very different classification in a 1924 book. The following is an outline of the classification in his A standard grammar of spoken Japanese (1930): A danku (sentence) consists of shi (word(s) with or without suffixed particles). A sh is either a tanshi (simple shi) or renshi. A renshi consists of (two or more) shi. A tanshi consists of genji (simple word(s) or signeme(s)). A gehji is either a tanji (simple genji) or renji. A renji consists of (two or more) genji (p. 15).

According to Matsushita, Japanese grammarians had already developed the concept of genji but no concept of shi; therefore their classifications were only of genji. Furth- ermore, they did not understand that the classification of parts of speech in Euro- pean grammar was based on the concept of shi. This misunderstanding caused them to apply European classifications to Japanese genji with absurd results. Matsushita says: "Therefore we must exclude joshi (enclitic particles and suffixes) and joddshi ("auxiliary verbs"), which are not shi, from the category "parts of speech" and re- classify the remaining seven parts of speech" (p. 18f.). His classification follows (p. 59):

[things] - •nouns, including pronouns etc. [actions] -verbs, including adjectives ,conccpt-5/i/ •adverbs, including conjunctions shi [modifiers] fukutaishi •subjective-.?/!/ -interjections 564 HATTORISHIRÔ

Matsushita's classification differs from Yamada's in that he excludes joshi and sets up a dichotomy between interjections and other shi. Matsushita also gives his system of classifying genji (p. 68 if.). Hashimoto Shifikichi classifies words purely in terms of textual functions and pays less attention to meaning. The classification which appears in his 1934 book follows: conjugationals which can— -yögen (verbs, become predicates adjectives) shi which can- -taigeri (nouns, become subjects pronouns, numerals)

conjugation-less can modify— -fukuyögen words the following (adverbs, conjunc- which cannot word tions, etc.) words become subjects cannot modify- -interjections the following word conjugationals -jodöshi P f ("auxiliary verbs") 'conjugation-less words -joshi (enclitics and some suffixes) We could say that there are as many classifications as grammarians.47 The following points may be the main causes of this great diversity: First, opinions differ as to the units to be classified. For instance, Hashimoto includes both joshi and jodoshi, Yamada includes only joshi, while Matsushita ex- cludes both. The writer believes that shokuno (grammatical textual function) and the meaning (grammatical situational function) are two aspects of one and the same thing, and stand in a one-to-one correspondence with each other. Hence they should be studied in terms of their mutual relationship rather than separately. However, grammarians tend to attach more importance to one or the other. Moreover, sufficiently precise research has yet to be carried out on either. Even when scholars agree in attaching importance to shokuno, they differ in select- ing the specific shokuno as the criterion for classification. For instance, if more importance is attached to the shokuno in the interior of bunsetsu, then the first division will be that of Yamada and Hashimoto, but if the shokuno in sentences is regarded as most important, interjections will be first separated from the other wordclasses, as in Matsushita's classification.

47 Sugiyama Eiichi, in his 1943 book, carried Hashimoto's procedure for classification as far as he could. Mikami Akira, in a 1953 book, clarified many points concerning the classification of words. DESCRIPTIVE LINGUISTICS IN JAPAN 565

As it is impossible to make a clear-cut distinction between the grammatical function and the lexico-semantic function, differences of opinion arise as to the range of the grammatical function, resulting in various forms of subclassification. It is necessary to study the difference between the two kinds of function.48 5. Syntax A word or a sequence of words consisting of one or more signemes and/or a bunsetsu or a sequence of bunsetsu consisting of one or more words, together with a super- imposed intonation, make a sentence. The study of the ordering, relationship, and structure in which these words or bunsetsu appear is called here "syntax". Japanese bunsetsu consist of one autonomous word with or without one or more synnomous words. The structure of bunsetsu is the same, even if jodoshi which are suffixed to yogefi are regarded as synnomous words. The author of Ayuisho (1778), Fujitani Nariakira, fully realized this point and gave the name ayui (something put on the feet) to the enclitic or suffixing forms which are now called joshi, jodoshi, and setsubiji (suffixes). Two or more of these non-autonomous forms are suffixed in a definite order, which have been gradually clarified by scholars. Recently, Watanabe Minora (1957), Miyaji Yutaka (1958), and Kabashima Tadao (1965) have tried to show the rules for order of suffixation by means of diagrams. Some bunsetsu (including those which can terminate a sentence) do not require another bunsetsu to follow them; there is another type, however, which does require a following bunsetsu which must be in concord with the first in meaning. The re- lationship of concord is fixed by social habit; some of the rules governing it belong to grammar, others to semantics. Grammatical rules of this type were noticed long ago by scholars of the Edo period. Motoori Norinaga, in a book published in 1785, made it clear that in Classical Japanese bunsetsu which end in (1) -wa, -mo, (2) -zo, -no, -ya, and (3) -koso require certain conjugational forms to terminate the sentence (for instance, in the case of the verbal ending -nu, bunsetsu which end in (1), (2), and (3) require the forms in -nu, -nuru, -nure respectively). Such rules have been gradually clarified by many scholars. As stated previously, Hashimoto Shinkichi proposed the concept of bunsetsu in 1934, but in 1944 he gave a lecture49 in which he proposed a new concept, refibunsetsu (a sequence of bunsetsu), and made it clear that bunsetsu are not juxtaposed on one level, but unified in varying depths of structure to form a sentence. For example: 'It-(topic) Japan-in two-(particle) not be fine tower is' Sore-wa Nihon-ni futatsu-to nai rippa-na td-de aru.

= 'It is by far the finest tower in Japan'. 48 See fn. 59. 40 Hashimoto's lecture on "the construction of the sentence in terms of bunsetsu" was given on September 3, 1944, and was published in Kokugogaku, Nos. 13/14 (1953). The manuscript was printed in his Kokubuhpo taikeiron (1959). 566 HATTORI SHIRO

Some scholars have tried similar analyses of sentences,80 but Hashimoto's theory is unique in many points. He holds that there are two types of bunsetsu in one sentence, "the terminating" and "the continuing", the latter of which is "taken up" by another bunetsu and is in concord with it. He tried to clarify the structure of sentences in terms of this relationship of concord. In his analysis of the sentence he always started with bunsetsu. Another characteristic of his approach is the fact that he did not go deeply into the analysis of meaning. Hashimoto's theory has been accepted by some scholars, and Morioka Kenji has had considerable success51 in clarifying the structure of the sentence by means of diagrams illustrating the relationship between bunsetsu. However, the attempt to describe sentence structure in terms of bunsetsu as its constituent units has been criticised from various angles. As stated previously, Yamada Yoshio referred to words as "the direct constituents of a sentence". Tokieda Motoki, advocating his own "process theory of language", proposes the following analysis :52

Ume-no hana-ga saita. 'plum's flower-(nominat.) bloom-(perf.)' = 'The plum blossoms are in bloom".

Hashimoto would analyze this sentence in the following way:

Ume-no hana-ga saita

By carefully listening to radio broadcasts, Hattori Shiro found that bunsetsu are sometimes uttered with an intermediate pause.53 For example ( | shows the pause):

honto-no-hatake-o-tsukuru \ to | iu-koto-wa ... 'real-'s-field-(accus.)-cultivate that say-thing-(topic) = 'the job of cultivating a real field ...'

50 Motoori Haruniwa gives similar diagrams in his Kotoba no kayoiji (preface of 1828). Matsushita Daisaburo's "renshi" and Saeki Umetomo's "renbunso" are concepts similar to "renbuhsetsu" (Kokugogakukai: Kokugogaku jiten, 962). 61 Morioka Kenji, 1964, 1965. 52 Tokieda Motoki, 1941; 1950, 250. 53 Hattori Shiro: Methods in linguistics, 416 (1947). DESCRIPTIVE LINGUISTICS IN JAPAN 567

suzu | namari \ akagane | no \ gdkih 'tin lead copper 's alloy' = 'an alloy of tin, lead, and copper' In the above examples, tsukuru-to and akagane-no are bunsetsu, which should be uttered uninterruptedly according to Hashimoto. Hattori stated84 that the "bunsetsu" kensetsu-no 'construction-(genit.)' for instance, may be uttered with a pause between kensetsu and no, when no and the other portion of the phrase are the imme- diate constituents in a sequence of bunsetsu, as in tsusan unyu kensetsu-no. He heard the following utterance in a radio speech: tsusan | unyu \ kensetsu \ no-kaku-daijin-ga ... 'ITI transportation construction 's-each-minister-(nominat.) (ITI = International Trade and Industry) = 'The I.T.I., Transportation, and Construction Ministers each ...' The opinions mentioned above are all to the effect that the juncture between immediate constituents can occur between the words which constitute one bunsetsu. However, Kuroda Sige-Yuki pointed out the fact that this juncture can appear within a form which everyone considers a "word". For instance, he first analyses (except for the intonation) the sentence Kaze-ga fuku as follows :54a Kaze ga fuk-\u 'wind (nominat.) blows' In the same way, Kaze ga fuk [ imasu '(polite)' Kaze ga fuk \ eba 'if' Kaze ga fui | ta '(perf.)' In his view all these forms should be analyzed as shown; and when this is done, we find that these analyses support each other. Likewise, he advocates the following analysis: Kare ga shacho da 'he (nominat.) president is' desu 'is (polite)' nara 'if ... is' 64 Ibid., 454 (1949). 64a Kuroda Sige-Yuki 1960. 29. Tamura Suzuko has proposed a similar analysis independently, and has tried to formulate a principle for analysis (not yet published). In the above book, Kuroda gives a new method of analysis and description of Japanese conjugational forms which is very different from traditional systems. 568 HATTORISHIRO

Broadly speaking, this analysis agrees with and, consequently, supports that of Yamada Yoshio which was done long ago with respect to meaning.56 This is important when viewed from the standpoint that analysis in terms of textual function and structure should agree with analysis in terms of meaning. Furthermore, the concord between forms should be studied in relation not only to the forms in question but also to the structure of the whole sentence. Minami Fujio has done an excellent study along these lines.56 The outline is as follows: The setsuzoku-joshi (conjunctive particles) diifer greatly in their function in constructing sentences; they can be roughly classified into three groups: A, B, and C. As examples of each group, the following three forms will be given here: (A) -nagara '(continuation of action)', (B) -node 'because', and (C) -ga 'but'. The particle -nagara is suffixed only to the rentaikei (a conjugational verb form), as in hanashi-nagara 'while speaking'. The form in -nagara can "take up" (= follow) bunsetsu which end in ni '(dative)', or o '(accusative)', or 'jotai-fukushi' (adverbs of state) like kuwashiku 'minutely', as in lKare-ni sono koto-o kuwashiku hanashinagara' 'Telling him in detail about the matter,...' (kare 'he', sono 'that', koto 'matter'). However, it cannot "take up" bunsetsu in -ga or -wa (particles which denote "the subject") or adverbs other than the above 'jotai-fukushV. The particle -node is suffixed to finite forms of yogen in suffixless form '(present- future)' or with the suffixes -masu '(polite)', -nai 'not', and -ta '(perfect)'. The form in -node can "take up", in addition to all the forms which the form in -nagara can take up, "the subject" in -ga, adverbial words of time such as kyo 'today' and kino 'yesterday', adverbs like kesshite '(n)ever' and rokuni 'well' which require a following negative, and the forms of group A. The particle -ga can be suffixed to all the forms to which -node can be suffixed, and, moreover, to the forms in -o '(conjecture)', -mai '(negative conjecture)', -daro '(sur- mise)'. The form in -ga can 'take up' all the forms which -node can take up, and, in addition, the topic in -wa, adverbs like tabun 'perhaps', masaka 'surely not', and the forms of group A and B. According to Minami, the following forms belong to these groups: A: -nagara '-ing', some "-/e" forms, -tsutsu 'ing', reduplication of "renyokei", e.g. hanashi-hanashi 'speaking' B: -node 'because', -tara 'if', -temo 'even if', -to 'when', -nara 'if', -noni 'though', -ba 'if', -te '(reason, cause)', -nagara 'although', etc. C: -ga 'but', -kara 'because', -ke(re)do(mo) 'but', -shi 'what with', etc. Minami thinks it necessary to set up another level, D, (or possibly even an E level), where 'imperative forms' such as nome 'drink!', nomuna 'don't drink', and nominasai 'drink', 'forms of doubt' like nomu-ka, 'affirmative forms' like nomu-wa, nomu-zo, and nomu-tomo, and affirmative forms suffixed by interjectional particles, e.g. nomu-

65 See fn. 34. 66 The Report of the Kokuritsu Kokugo Kenkyujo, No. 23 (1963), Minami Fujio, 1964 a and 1964b. DESCRIPTIVE LINGUISTICS IN JAPAN 569 ka-ne, nomu-wa-ne, nomu-tomo-sa, etc. can occur, and where they can be preceded by vocative forms, conjunctions, interjections, forms of interrogative meaning such as nani 'what', dare 'who', itsu 'when', doko 'where', and naze 'why', and such adverbs as zehi 'by all means', dozo 'please', etc. As Minami himself recognizes, this study needs further refinement, but it is evident that research on the structure of sentences and the study of concord between forms have been greatly promoted by his study. In the study of language, it is necessary in general, to investigate both the structure and the system not only in the aspect of sound, but also in all other aspects. It is necessary also to carry Minami's study further by trying to discover in what systems forms on the same level (that is, of the same group) stand in opposition to each other. The same thing can be said of the study of taxemes and sentence patterns. More- over, it is necessary to study the relationship between taxemes. In this sense, the recent trial applications of American generative grammar to the description of Japanese are to be welcomed. They will contribute greatly to the study of this aspect of the language if careful research is conducted, paying sufficient attention to the intuition of the speakers. It is to be noted, however, that Japanese grammarians also have had a concept similar to "transformation" for some time. For instance, what Mio Isago calls teni (transposition) in his 1948 book is similar to transformation (p. 68f.; p. 98f.): Ojiisan-wa yama-e shibakari-ni ikimashita. 'old-man-(topic) mountain-to firewood-gathering-for went' = 'The old man went to the mountain to gather firewood'. This sentence can be 'transposed' into: Yama-e shibakari-ni itta-no-wa ojiisan da.57 'he who went (topic) old man is' = 'The person who went to the mountain to gather firewood is the old man, This sentence can be further "transposed" into: Ojiisan-ga yama-e shibakari-ni itta-no-da.57 ' old-man-(nominat.)' = 'The old man went to the mountain to gather firewood'. = 'It was the old man who went to the mountain to gather firewood'. The latter transposition is possible, Mio says, only in the case of the "identity judg- ment" of Wundt.58 The study of the mutual relationship between taxemes (and sentence patterns) and the study of the possible "abbreviation" of some portion of them is one of the greatest tasks for the future. 67 It is better to use desu 'is (polite)' instead of da 'is', because the original sentence ends in -ma- shita '(polite)-(perfect)\ 58 Mikami Akira also has succeeded in the analysis of the sentecne utilizing this concept in his 1953 book, 101 ff. 570 HATTORI SHIRO

V. Sememics The study of meaning, even when restricted to the study of the sense of the sentence, covers a remarkably wide field. As the sense of intonation patterns and taxemes (and sentence-patterns) has already been touched upon, we shall here deal mainly with the sense of the word, which Hattori calls the "sememe". Compare, for instance, the sememes of the following four words: yomu 'reads' kaku 'writes' (A) yometread!' kake 'write!' (B)

(a) (b) The sememic features which the (A) forms have in common differ from those which the (B) forms have in common. The same statement applies to (a) and (b) forms. The sememic features common to the (A) forms or the (B) forms are called the "grammatical meaning" (strictly speaking, "sememic features" instead of "meaning"), while those common to the (a) forms or the (b) forms are called the "lexico-semantic meaning". It is usually thought that no '(genit.)' and da 'is' in 'Kodomo-no hon-da' 'It's a children's book' have grammatical meanings, while kodomo 'child, children' and hon 'book' have lexico-semantic meanings, Strictly speaking, however, all the nouns kodomo, hon, jidd 'children (lit.)', shomotsu 'book (lit.)', tsukue 'desk', etc. have a common grammatical meaning; in the case of kodomo, hon, and tsukue the only features which differ are their lexico-semantic meanings.59 Furthermore, we shall define kodomo and jidd, hon and shomotsu, respectively, as differing from one another in stylistic meanings. We shall give a brief survey of the studies mainly of the grammatical meaning here. The description of the lexico-semantic meanings in dictionaries has been very inac- curate and is lacking in scientific value especially in respect to basic vocabulary. The study of grammatical meanings had been conducted in the form of the study of te-ni-o-ha (particles) since very ancient times, but it became scientific only when Motoori Norinaga and Fujitani Nariakira began their research. It is a noteworthy fact that Suzuki Akira, in his Gengyo shishuron (1824), classified parts of speech into four classes: tai no kotoba (nouns), arikata no kotoba (adjectives), shiwaza no kotoba (verbs), and te-ni-o-ha (particles). He said that members of the first three classes "indicate something", that "they are words" and that "they become words by indicating things and events", whereas te-ni-o-ha "indicate nothing", are merely "voice", "voices of the mind attached to words". Suzuki's thinking greatly influenced Tokieda Motoki. As has already been mentioned, Yamada Yoshio in A Japanese grammar (1908) developed an excellent classification of words and promoted the study of grammatical 59 Scholars differ as to where to place the boundary between grammatical meaning and lexico- semantic meaning. Chomsky's "grammaticalness" includes what are usually called "lexico-semantic" features. Hattori Shiro, in his 1964a paper (§ 3), has tried to find exact criteria for distinguishing the two types of meaning. DESCRIPTIVE LINGUISTICS IN JAPAN 571 meaning, especially by subclassifying verbal endings, adverbs, and enclitic words in terms of textual function and meaning. The study of the grammatical meanings of the enclitic words and endings of spoken Japanese also became more and more detailed, and the books of Matsushita Daisa- buro (1930), Sakuma Kanae (1936, 1940, 1951, and 1959), Mikami Akira (1953), and the Report of Kokuritsu Kokugo Kenkyujo, No. 3 (1951), made great contributions in this field. Kindaichi Haruhiko, in a 1953 paper, published his conclusions on a detailed study of the meanings of some verbal endings, and Hattori Shiro came to a similar conclusion from a different viewpoint.60 It is necessary to compare these results with those of Haga Yasushi and Minami Fujio61 and to conduct more tho- rough and accurate research. The grammatical meaning has to do not only with the internal world of the speaker called "consciousness"—"the internal situation"—but also with the outer world, that is, "the external situation". Especially in the study of the meaning of pronouns and honorific forms, it is necessary to observe the relationship between the speaker and the external situation, including the hearer(s). Jinbo Kaku in his Introduction to linguistics (1922) did some work in this field and has refined it further in his Linguistic theory (1961). Sakuma Kanae is also one of the pioneers in this field. In a 1936 book and others, he has classified a group of words, including demonstrative pronouns, into a system which he terms the ko-so-a-do system. The name comes from a juxta- position of the first syllables of kore 'this', sore 'that', are 'that (over there)', and dore 'which'. Tokieda Motoki, in a 1941 book, has contributed to the study of the "situation". Mio Isago,ina 1948 book, has given an excellent classification of sentences in relation to the situation and has contributed to the study of the relationship be- tween the sentence and "the context of situation". As for the descriptive study of "honorific expressions", books by the following writers are worth mentioning: Yamada Yoshio (1924), Tokieda Motoki (1941), Egoyama Tsuneaki (1943), Mizutani Shizuo (1955), Mio Isago(1958), and the Report of Kokuritsu Kokugo Kenkyujo, No. 11 (1957). As stated in Chapter II, Jinbo Kaku long ago stressed the distinction between "the concrete meaning" and "the characteristic meaning". However, it is impossible to say that scholars since his time have been sufficiently aware of this distinction, although they have in general tended to describe the 'characteristic meaning' more or less unintentionally. The meaning of foreign languages, old classical Japanese, or non-native dialects has in the past been inaccurately described through the frame of the native language or dialect. In order to correct this shortcoming Hattori Shiro emphasized the necessity for establishing a branch of the descriptive study of meaning, which would correspond to phonology in the field of sound.62 As mentioned previously, he distinguishes be-

60 Hattori Shiro 1957 (Methods in linguistics, 209). Part of the discussion is in English (ibid., 795ff.). 61 Chapter IV, 2, p. 554f., 5 p. 568f. 62 Hattori Shiro 1953 (Methods in linguistics, 371 ff.). Later articles have also been reprinted in the same book. 572 HATTORI SHIRO tween the utterance on the one hand and the linguistic product and the sentence on the other, and by analyzing the sentence further, he reaches his concept of "word", which is the combination of shape and sememe. As the phoneme is the minimal phonological unit, so is the sememe the minimal semological unit. As the phoneme can be described as a bundle of distinctive (and other) features, so can the sememe as a bundle of grammatical and lexico-semantic (and stylistic) features, he claims.63 Kunihiro Tetsuya, in a 1962 paper, has attempted the analysis of the sememes of the Japanese case-particles: ga, no, o, ni, e, to, de, kara, yori, and made. In the case of the particle ni for instance, Kunihiro posits only one sememe: "indication of the object of close contact" + "fukushiteki igishitsu" (adverbial sememic quality) (p. 86), whereas the Report of Kokuritsu Kokugo Kenkyujo, No. 3 (1951), establishes nine categories of meaning for ni, the fifth of which is subdivided into three groups, while the eighth is broken down into seven. It is especially noteworthy that Kunihiro analyses the sememe into two elements as shown above, and that he analyzes the sememe of nouns into "igiso-jisshitsu" (sememic substance) and "meishiteki igishitsu'' (nominal sememic quality) (p. 91, note 7), and the sememe of verbs into "sememic substance", "verbal sememic quality" and "igitai" (sememic aspect) (note 11 of his paper). Roughly speaking, Kunihiro's sememic substance of the sememes of nouns and verbs corresponds to Hattori's lexico-semantic sememic features, and the other elements in Kunihiro's system to Hattori's grammatical sememic features. However, Hattori holds that the sememe of 'joshV (enclitic words) consists only of grammatical features. Kunihiro posits two sememes for the enclitic word to, showing that a form may have plural sememes. It is necessary to conduct more exhaustive research in order to determine whether or not Kunihiro's description has posited too small a number of sememes. In order to render as objective as possible the analysis of the sememe, which is usually conducted intuitively, Hattori has proposed the following working principles for the study of grammatical sememic features :M

63 "The grammatical feature" (strictly "the grammatical sememic feature") corresponds to what he once called "igishitsu" (sememic quality) in discrimination from the sememe, considering it as some portion of the sememe (cf. the 1957 paper, Methods in linguistics, 204). At the Ninth Inter- national Congress of Linguists which was held in August, 1962, Hattori used the terms grammatical feature and lexico-semantic feature (Proceedings, 280). On page 570 of this chapter, the author has stated that the "grammatical sememic features" are what are common to the sememes of yomu /'jo1mu/ '(to) read (something)', and kaku /ka?ku/ '(to) write (something)'. These features, however, are not expressed only by the signeme /u/ which is common to /'jonmu/ and /kanku/. Some of them, for instance the "transitive verb sememic feature" which is common to the two verbs, must be expressed by the signemes /'jonm/ and /ka1k/. This feature, which must be considered "grammatical", is lacking in the sememes of intransitive verbs such as /'jamu/ '(to) stop' and /'aku/ '(to) open'. Inasmuch as the verbs /'jamu/ and /'aku/ have the signeme /u/ which is common also to the in- transitive verbs /'jo1mu/, /kanku/, /'jaku/ '(to) burn (something)', etc., it is impossible to think that this /uI expresses the "transitive verb sememic feature". 64 Hattori Shir6 1964a. The analysis of meaning in Hattori's papers of 1957 and 1958 (Methods in linguistics, 198 ff., 793 if., 494ff., 401 if.) was conducted in terms of these principles, although not explicitly. DESCRIPTIVE LINGUISTICS IN JAPAN 573 I. Forms which can stand in the same grammatical position have the same grammat- ical feature in common (The Principle of the Same Position). II. Forms (including an intonation pattern) which can be mutually united by means of the same taxeme (or sentence pattern) possess concordant grammatical features (The Principle of Concordance). By "grammatical position" is meant the position in a taxeme (or a sentence pattern) or the position where the form is united with a synnomous word or intonation pattern. For example, forms which can stand in the position of the subject in the taxeme ((subject + predicate)) have a common grammatical sememic feature; and forms which can be suffixed by the case particle o '(accusative)' have another common grammatical sememic feature. In the latter case, the shared common feature is con- cordant with some feature of the sememe of o. Accordingly, in terms of these princi- ples, the sememes of the nouns mentioned earlier, kodomo 'child, children', hon 'book', jido 'children', shomotsu 'book', and tsukue 'desk', may be described as having in common the same grammatical sememic features which are concordant with the case particles ga, o, ni, no, etc.65 In a 1965 paper, Mio Isago mentions the "kodo-taisei" (action system) of "nani-ka-ga '(nominat.)' nani-ka-o '(accus.)' doko- kara 'from', doko-e 'to', otosu 'let fall'" 'someone lets something fall from some- where to somewhere', and that of "nani-ka-ga doko-kara doko-e ochiru 'fall'" 'some- thing falls from somewhere to somewhere'. In terms of Hattori's second principle, the sememe of otosu 'let fall' may be described as having grammatical features which are concordant with (in this case, "take up") ga, o, kara, and e, while the sememe of ochiru 'fall' may be described as having grammatical features which are concordant with ga, kara, and e. Hattori proposed the following working principles in a lecture on July 15, 1964, at International Christian University: I. Autonomous words which can be united with another autonomous word by means of the same taxeme (or sentence pattern) have the same lexico-semantic sememic feature in common.66 II. Two autonomous words which can be united possess concordant lexico-semantic sememic features.67 Hattori also stated that "an autonomous form consisting of an autonomous word and one or more synnomous words" can replace the "autonomous word", referred to in the above principles, but autonomous forms consisting of two or more autono- 65 Hattori Shir6, in his 1964a paper, has tried to represent sememic features with symbols, and to describe "textual features unable to manifest themselves at the same time" as coexisting in one sememe but as being "neutralized". 86 For example, autonomous words (nouns) such as ringo 'apple', pan 'bread', sakana 'fish', niku 'meat', etc. which can be modified by an autonomous word (adjective) such as oishii 'delicious' have in common a lexico-semantic sememic feature (which is concordant with one of the lexico- semantic sememic features of oishii). 87 For instance, two autonomous words, ringo 'apple' and oishii 'delicious', which can be united as in oishii ringo 'a delicious apple', possess concordant lexico-semantic sememic features. In other words, one of the lexico-semantic sememic features of ringo 'apple' is concordant with one of those of oishii 'delicious'. 574 HATTORI SHIRO mous words cannot replace it, because they generally have too many parole-like features. However, zo-wa 'elephant-(topic)' can be united with hana-ga nagai 'the nose is long' as in Zo-wa hana-ga nagai 'Elephants have long noses', and it is very probable that we have to describe the sememe of the noun zo 'elephant' as having, as one of its lexico-semantic features, a feature concordant with one of the sememic features of hana-ga nagai, because the habit of uniting zo and hana-ga nagai has become socialized, that is, has become langue-like.68 However, from the example Shinju-wa Nihon-ga yasui (pearl-(topic), Japan-(nom- inat.), cheap = 'Pearls are cheap in Japan') we must not conclude that Nihofi 'Japan' and yasui 'cheap' possess concordant lexico-semantic features, because this sentence is a transform derived from Nihon-wa shinju-ga yasui (Japan-(topic), pearl- (nominat.), cheap = 'Pearls are cheap in Japan') and Nihofi cannot be united with yasui except in such transformed sentences. As mentioned in the introduction to Chapter IV (p. 547 f.), the study of the sememe should be conducted through the observation of both the textual and the situational functions of words. Especially in the case of the study of lexico-semantic sememic features we have to pay special attention to the situational function. Moreover, we must not fail to study the systemic function in so far as it is possible—especially in the case of the study of the grammatical sememic features. In a 1965 paper, Kunihiro Tetsuya has made an excellent analytical study of the sememes of the Japanese adjectives tsumetai 'cold', samui 'cold', suzushii 'cool', nurui 'lukewarm', atatakai 'warm', atsui 'hot (fire)', and atsui 'hot (day)', and has tried to set up the system of these sememes. He also conducted studies of the cor- responding English adjectives. Although it cannot be expected that such a beautiful system of sememes as Kuni- hiro's will be found in many portions of the whole vocabulary, it is probable that improvements in the analytical approach will throw light on many heretofore unclear points. The study of the classification of vocabulary probably should proceed by starting with basic vocabulary and by discriminating various stylistic levels—for instance, by distinguishing kodomo 'child, children' and hon 'book' from jidd 'children (lit.)' and shomotsu 'book (lit.)'. We also have to deal separately with non-socio- habitual creative combinations of words aimed at attaining special effects in literary works, especially in poetry. Of course, idioms need special study, apart from their individual constituent words. In general, the classification of sememes should be done purely in linguistic terms, not in terms of other sciences or logic. Kuraishi Takeshiro's Iwanami dictionary of the Chinese language (1963), an original dictionary of the spoken Peking dialect, has greatly contributed to the description

68 It is almost impossible to divide clearly this type of habit into two: social and non-social. How- ever, some words, such as Japanese sakura 'cherry blossom', possess connotations which are different from English cherry blossom and these connotations should be described as belonging to the sememe of sakura. This comes from the social habits, ways of thinking, emotional reactions, etc., which the Japanese people have in relation to cherry blossoms, and to the stereotyped phrases which they are accustomed to utter very frequently concerning them. DESCRIPTIVE LINGUISTICS IN JAPAN 575 of meaning. Although the meaning of words is described by translation into Japanese, every word is assigned one of the numbers 5, 4, 3, 2, 1, 0, -1, -2, -3, -4, or -5 to show stylistic levels. In the appendixed "Index in terms of meaning" he gives an original classification of words in terms of meaning. The Kokuritsu Kokugo Kenkyujo's Word list by semantic principles (1964), which is based upon statistical studies of vocabulary conducted over a long period, is a list of 32,600 words, divided into a number of categories, in which the most frequent 7000 words are asterisked; an appendix contains an alphabetical index of words. The author (Hayashi Oki) writes that he has classified words "without paying much attention to precedents established in other dictionaries, although classified telephone directories, catalogues of merchandise, and illustrated zoological or botanical texts were sometimes consulted; it is original, but it may be dogmatic, and is not perfect" (p. 4). This work of Hayashi's is to be welcomed, however, as a trial effort. The Kokuritsu Kokugo Kenkyujo's Report No. 28, Japanese synonymy and its problems (1965) is an interesting and valuable book which is the result of investigation of the semantic differences of synonyms by means of questionnaires and experimental tests.

VI. ADDITIONAL REMARKS

The necessity of the study of "life through language" has been advocated especially since World War II, and the Kokuritsu Kokugo Kenkyujo has published as Reports No. 2 (1951) and No. 5 (1953) the results of research using the sociological and statistical approach. Shibata Takeshi has done a trial study of "An investigation of the language behavior of an individual during 24 hours" (1951).69 The Kokuritsu Kokugo Kenkyujo has published many reports on vocabulary based on statistical procedures: No. 4 (1953), No. 12 (1957), No. 13 (1958), No. 21 (1962), No. 22 (1963), and No. 25 (1964). In 1957, the Mathematical Linguistics Society of Japan was established. Some of these statistical studies contribute to descriptive linguistics. Also, studies in fields such as "stylistics", "linguistic aesthetics", etc.70 have clarified some points which are interesting from the viewpoint of descriptive linguistics. As has often been mentioned in this paper, Japan has been strongly influenced by Western sciences since the Meiji period. Linguistics is by no means an exception; Japanese scholars have become familiar with a great many books and articles either in the original or in translation. However, it is often difficult to confirm the extent of this influence. As one example, it is dangerous to affirm that Hashimoto Shinkichi is greatly influenced by de Saussure. Although he sometimes refers to de Saussure in his later works, it is obvious that his 1927 lecture on the history of the sounds of Japanese at Tokyo Imperial University is not yet under Saussurian influence.71 Nevertheless,

69 Cf. W. A. Grootaers: "Language behavior of an individual during one day", Orbis, 1:1, 1952. 70 Papers and bibliographies by Hatano , Kobayashi Hideo, and Yasumoto Yoshinori appear in Lectures on the modem language, Meiji Shoin, November 1963. 71 Kobayashi Hideo's Japanese translation of F. de Saussure's Cours de linguistique générale was published in 1928. 576 HATTORISHIRO after describing the synchronic structure and system of Japanese for the three periods, the eighth and sixteenth centuries and the present age, he discussed in this lecture the diachronic changes in the language. This marked the development and resystematization of traditional studies by Japanese scholars.72 Although Yamada Yoshio quotes in his Japanese grammar (1908) various passages from Western scholars, he does so for purposes of criticism or in order to utilize them in his own research. He does not swallow their ideas as a whole, nor does he merely parrot them. These scholars differ from mere translators or adapters in that they conduct their own linguistic research and solve difficult problems by means of their own observations and inquiries. There- fore, it is quite natural that they develop original ideas. Even if we find strong simi- larities to Western theories, we must not draw the hasty conclusion that these simi- larities are the result of adaptations of Western theories. Their statements are system- atic, consistent, and integrated, whereas those of adapters are conglomerations of fragmentary and often self-contradictory ideas. Rather than provide a detailed list of books and papers, the writer has endeavored to trace the development of theory and analytical procedure in Japanese descriptive linguistics. Also, he has had no time to deal with all the problems exhaustively, nor to introduce with comments various divergent opinions. In order to make up for these deficiencies, several works which are not referred to in this paper are given in the bibliography, which also includes a list of books with exhaustive or good bibliog- raphies.

BIBLIOGRAPHY

Arisaka Hideyo "Onsei no ninshiki ni tsuite" # ^ 0 M ft t- 1 X [On the cognition of sounds], Onsei no kenkyu if ^ ® W [The study of sounds], Vol. 4 (Tokyo, Onseigaku Kyokai [The Phonetic Society], Dec., 1931). , Onihron if fit [Phonology], (Tokyo, Sanseido, Dec., 1940). ,"Akusento no kata no honshitsuni tsuite" T ? >- h 0 © ® O ~C [On the nature of accent patterns], Gerigo kefikyu H no" W [Studies in language] Nos. 7/8, (Tokyo, The Linguistic Society of Japan, April, 1941). Egoyama Tsuneaki tX ULl ® Pfi, Keigoho fSC In [Honorific expressions] (Tokyo, Sanseido, Sept., 1948). Fujitani Nariakira # zt & Kazashisho ffi S2fi PP [A study of proclitic forms] (written in 1767). , Ayuisho PA f'p [A study of enclitic forms], written in 1773, published in 1778. Fukuda Yoshisuke ijig IH & fit, "Chinjutsu no kind" © fg [The function of chinjutsu], Zoku Nihon buripo koza [Lectures on Japanese grammar, second series] Vol. 1 (Tokyo, Meiji Shorn, June, 1959). 72 Hashimoto greatly influenced his students, and Arisaka and Hattori are among them. DESCRIPTIVE LINGUISTICS IN JAPAN 577

Haga Yasushi % U "'Chinjutsu' to wa nanimonoT' "[

Jinbo Kaku # ^ Gengogaku gairon a" i£u ^ Hr [Introduction to linguistics] (Tokyo, Iwanami Shoten, Nov., 1922). , Gengo riron lfs5L Hi Hr [Linguistic theory] (Tokyo, private edition [Iwanami Shoten], August, 1961). Kabashima Tadao W ir & zJi, "Gojun no riron to jissai" fg- Jig © M 1& h * ^ [Word order in theory and practice], Kogo bunpb koza • Wt 3C '/i M J® [Lect- ures on the grammar of the spoken language], Vol. 2 (Tokyo, Meiji Shoin, April, 1965). Kawabata Yoshiaki jl| ffi H P, "Bunron no hoho" % H © 77 & [Methods in syntax], Kogo bunpo koza • fo fH JH [Lectures on the grammar of the spoken language], Vol. 1 (Tokyo, Meiji Shoin, Feb., 1965). Kawakami Shin Jl| h M, "Kotoba no kireme to oncho" W M Bfj W) — JH ii I Jt [Bound forms in modern Japanese—uses and examples] (March, 1951). , Hokoku, No. 4, Gendaigo no goi chosa, fujin zasshi no yogo ft ifo © In jg: I® 4t — M A H © ffl [Studies of word frequencies in the modern language—a study of the vocabulary used in women's magazines] (March, 1953). DESCRIPTIVE LINGUISTICS IN JAPAN 579

, Hökoku, No. 5, Chiiki shakai no gengo seikatsu—Tsuruoka ni okeru jittai chösa i&M f± —ÄW t- tf £ * B 11 ® [Life through language in a regional community—A survey in Tsuruoka City] (March, 1953). , Hökoku, No. 8, Danwago no jittai Sic IS to © Ä fil [Research in colloquial Japanese] (March, 1955). , Hökoku, No. 11, Keigo to keigo ishiki 1$C I fjt Wr M He [Honorifics and consciousness of honorifics] (March, 1957). , Hökoku, No. 12, Gendaigo no goi chösa, sögö zasshi no yögo, zenpen 50, ft to ©l^ii ii, Ii H © ffl Wr, M Ü [A study of word frequencies in the modern language: a study of the vocabulary used in intellectual journals], Part I [Word frequency tables] (March, 1957). , Hökoku, No. 13, Gendaigo no goi chösa, sögö zasshi no yögo, köhen & Iii (as above, Part II: [Method and analysis] (Feb., 1958). , Hökoku, No. 18, Hanashi-kotoba no bunkei (1)—Taiwa shiryö niyoru kenkyü IS L Z l täCDjem (1) — Jt f£ W II X h [A study of sentence patterns in colloquial Japanese (1)—General conversation] (March, 1960). , Hökoku, No. 21, Gendai zasshi kyüjusshu no yögo yöji (1) Söki oyobi goihyö ^ ft It g£ + S © ^ ffl ^ (1) iff. f£ ¿i I [Vocabulary and Chinese characters used in ninety present-day magazines (1): General introduction and word frequency tables] (March, 1962). , Hökoku, No. 22, as above (2) Kanjihyö [Chinese character frequency tables], (March, 1963). , Hökoku, No. 23, Hanashi-kotoba no buiikei (2)— Dokuwa shiryö ni yoru kenkyü ILl l tä

Kuroda Sige-Yuki M E J&^ß, Gengo no kijutsu H fi % mt [Subject, general subject, topic word, and object word], Kögo bunpö köza • to Hü 1Ü [Lectures on the grammar of the spoken language], Vol. 2 (Tokyo, Meiji Shoin, April, 1965). Mitsuya Shigematsu H ^ Iii Kötö Nihon bunpö ¡gj ^ H ^ Ü [An advanced Japanese grammar] (Tokyo, Meiji Shoin, Dec., 1908). Miyaji Yutaka g "Bun to hyögenbun" £ l Ä 5PL 3C [Sentences and "ex- pression sentences"], Kokugo kokubun [Japanese language and literature], 27:5 (Kyoto, Kyoto University, May, 1958). Miyake Takeo H ^ Ä fi|3, "Onsei kögohö" Ür ^ • ii [A phonetic grammar of the spoken language], Kokugo kagaku köza HI In # ^ fH J® [Lectures on the science of the national language] (Tokyo, Meiji Shoin, Oct., 1934). DESCRIPTIVE LINGUISTICS IN JAPAN 581

Miyata Köichi ^ EB # —, "Atarashii akusentokan to akusento-hyökihö" if L ^ T^ty new theory of accent and transcription of accent], Ofisei no kenkyü ilr M © W [The study of sounds], Vol. 1 (Tokyo, Onseigaku Kyökai [The Phonetic Society], Sept., 1927). , "Nihongo no akusento ni kansuruwatakushi no kenkai" H & -fc v 1- Hfl h fi. © IL ffl [My view of accent in Japanese], Ofisei no kenkyü, Vol. 2 (Tokyo, Onseigaku Kyokai, Dec., 1928). Mizutani Shizuo tK & ^ Taigü-hyögen no kiso # ^ M © ^ ^ [Fundam- entals of honorific and humble expressions] (Tokyo, private edition, Oct., 1955). Monbusho, Kokugo Chosa Iinkai [The Japanese Language Commission of the Ministry of Education], Kögohö chösa hökokusho P IjS if fß ^ It [Report on a survey of the grammar of the spoken language], (Dec., 1906). , Kögohö • go ii [Spoken language grammar] (Dec., 1916). Morioka Kenji M |S| HÜ —, "Bun no kösei" £ © ® JsSc [The structure of the sentence], Kokugo to kokubungaku [Japanese language and literature], 41:7 (Tokyo, Shibundo, July, 1964). , "Fukubufi no közö" ^ 3t © fil is [The structure of complex sentences], Kögo bunpö köza P Wr3Cl(£.MM [Lectures on the grammar of the spoken language], Vol. 2 (Tokyo, Meiji Shoin, April, 1965). Morishige Satoshi ^ M Wi, "Nihon bunpö tsüron" H $ Ü i! ffr [An outline of Japanese grammar] (Tokyo, Kazama Shobo, Jan., 1959). Motoori Haruniwa if J® fit, Kotoba no yachimata jüö A Hf [Multiple crossings of words], written in 1806, published in 1808. , Kotoba no kayoiji JÜ [Word paths], 1828 preface, published in 1832(3?). Motoori Norinaga ^ Jgr EL M, Kotoba no tama no o BRJ 3£ [Beadstrings of words], completed in 1779, published in 1785. , Mikimi kotoba katsuyö shö i? HI |5] fli ffi [A study on conjugation in the national language], started about 1782, published in 1886. , Kojiki den if ^ ti If [Studies of the Kojiki], 1764-98. öiwa Masanaka ^ IE {"t1, "Bun no teigi" CD It [Definitions of the sentence] Kokugogaku HU fn [Studies in the Japanese language], No. 3 (Tokyo, Koku- gogakkai, Nov., 1949). ökubo Tadatoshi ^c >h % & f'J, "Chinjutsu to wa donna koto ka" ¿Ej l fi ¿f /L % Z. i [What is 'chinjutsu'l], Kokubungaku, kaishaku to kanshö S§ M ff t IE Jt [, its interpretation and appreciation], 29:11 (Tokyo, Shibundo, Oct., 1964). ötsuki Fumihiko jz M3C jl£, Kö Nihon bunten Hr H J)| [A detailed Japanese grammar] (Tokyo, private edition, Jan., 1897 [written in 1882]). , Kö Nihon bunten bekki Wk H if 3C Jl f£, [A detailed Japanese grammar, notes and explanations], Tokyo, private edition, Jan., 1897. 582 HATTORI SHIRÖ

Sakakura Atsuyoshi Si, Nihon bunpö no hanashi 0 ^ 3C Ü © IS [Lectures on Japanese grammar] (Osaka, Sögensha, April, 1952). Sakuma Kanae fe. RD Jfti, Kokugo no hatsuon to akusento Hjfp©ÜrÜr i T ? -t ^ }• [The pronunciation and accent of Japanese] (Tokyo, Döbunkan, Au- gust, 1919). , Gendai Nihongo no hyögen to gohö fä B fg- 0 ^ Jg, b Ig- Ü [The expression and grammar of modern Japanese] (Tokyo, Köseikaku, May, 1936). , Gendai Nihongohö no kenkyü ft 0 ^ In ii © W [A study of the grammar of modern Japanese] (Tokyo, Köseikaku, April, 1940). , Nihorigo no gengo-rironteki kenkyü [A theoretical linguistic study of Japanese] (Tokyo, Sanseido, August, 1943). , Nihongogaku B ^ to ^ [Japanese linguistics] (Tokyo, Asahi Shinbunsha, August, 1951). , Nihongo no gengo-riron B © W jig- iU ffe [Linguistic theory of Japa- nese] (Tokyo, Köseisha Köseikaku, Oct., 1959). Shibata Takeshi ^ IB "Gengo seikatsu no nijüyojikan chösa" H In & ffi © 24 fal IJ1 ft [An investigation of the language behavior of an individual during 24 hours], Gengo seikatsu Hi im ife '/¡§ [Life through language], No. 1 (Tokyo, Chikuma Shobö, Oct., 1951). , "Onsei—sono honshitsu to kinö" if — £ © if ¿ilfg [Sound- its nature and function], Kokugo kyöiku no tame no kokugo köza ¡JÜ go m w © fZ © PH In IS [Lectures for the teaching of the national language], Vol. 2 (Tokyo, Asakura Shoten, Oct., 1958). Sugiyama Eiichi & lil M Kokugohö hinshiron ® Wr pp fn] jir [A Japanese grammar, parts of speech] (Tokyo, Sanseidö, Dec., 1943). Suzuki Akira §v & fl§, Gengo shiskuron U Ig U9 iH jif [A theory of four parts of speech], 1824. Tokieda Motoki B# M Iß, Kokugogaku g67lf07l [HI ppj ^ [Fundamentals of Japanese linguistics] (Tokyo, Iwanami Shoten, Dec., 1941). , Nihon bunpö, kögohen El Ü > P In M [A grammar of spoken Japa- nese] (Tokyo, Iwanami Shoten, Sept., 1950). Tsurumine Shigenobu tl ^, Gogaku shiiisho fg ^ Wi # [A new book of language learning], 1833. Uemura Yukio # "Gengo-onsei wa nani o tsutaeru ka" H In Hr if? ti M •fc 1$ h fi \ [What do speech-sounds convey?], Gengo seikatsu Hi Wr ife ffi [Life through language], No. 151 (Tokyo, Chikuma Shobö, April, 1964). Yamada Yoshio lil BB # if, Nihon bunpöron B Ü l& [A Japanese grammar] (Tokyo, Höbunkan, Sept., 1908, partly in 1902). , Nihon bunpögaku gairon 0 Ü ¿P oft [An outHne of Japanese grammar] (Tokyo, Höbunkan, May, 1922 (a)). , Nihon kögohö kögi B ¿f- P In Ü M SS [A lecture on the grammar of spoken Japanese] (Tokyo, Höbunkan, Nov., 1922 (b)). DESCRIPTIVE LINGUISTICS IN JAPAN 583

, Keigohô no kenkyû fÎC to ii © W [A study of honorific expressions] (Tokyo, Hobunkan, June, 1924). Yasuda Kiyomon ^ H M ft P^, Kokugohô gaisetsu M fë M [An outline of the grammar of the national language] (Tokyo, Chukokan, March, 1928). Watanabe Minoru iâ St, "Jojutsu to chifijutsu" j® L [Description and chinjutsu], Kokugogaku HI ml [Studies in the Japanese language], Nos. 13, 14 (Tokyo, Kokugogakkai, Oct., 1953). , "Hinshiron no shomondai" bp IrJ Hfe © Si PnJ jH [Problems of parts of speech], Nihon bunpô kôza H iSC. M ¡Ê [Lectures on Japanese grammar], Vol. 1 (Tokyo, Meiji Shoin, Nov., 1957).

Lecture Series

Kokugo kagaku kôza H in JËS IS [Lectures on the science of the national language] (Tokyo, Meiji Shoin, 1943-1945), 79 books in 12 cases. Nihon bunpô kôza H ii M? fH [Lectures on Japanese grammar] (Tokyo, Meiji Shoin, Oct., 1957—July, 1958), 6 vols. Kôza, gendai kokugogaku jl|§ M? ft @ç| b§- ^ [Lectures on modern Japanese linguistics] (Tokyo, Chikuma Shobo, Nov. 1957 — July, 1958), 3 vols. Zoku Nihon bunpô kôza M H ^ iè lH ® [Lectures on Japanese grammar, second series] (Tokyo, Meiji Shoin, May, 1959—Oct., 1959), 4 vols. Kôza, gendaigo ÎH Ml ft fo [Lectures on modem language] (Tokyo, Meiji Shoin, Nov., 1963—April, 1964), 6 vols. Kôgo bunpô kôza P 3C '/i Ml [Lectures on the grammar of the spoken language] (Tokyo, Meiji Shoin, Nov., 1964—June, 1965), 6 vols.

Dictionaries

Nihon bungaku daijiten H ^C^ JzWi M [A large dictionary of Japanese litera- ture], edited by Fujimura Tsukuru || # ^ (Tokyo, Shinchosha, Vol. 1, June, 1932; Vol. 2, April, 1933; Vol. 3, June, 1934; Appendix, May, 1935). Kokugogaku jiten !H to ^WtM [A dictionary of Japanese linguistics] edited by Kokugogakkai SI ^ [The Society for the Study of the Japanese Language] (Tokyo, Tokyodo, August, 1955).

Books with bibliographies

Saito Seiei Hf M fit itr, ed., Kokugo kokubungaku ronburi somokuroku [H to HI jiC ^^fX® 0 ^ [A general bibliography of articles on the Japanese language and literature] (Aug., 1945—July, 1953), (Tokyo, Shibundo, Sept., 1954). 584 HATTORI SHIRÖ

Nippon Gakujutsu Kaigi Fl ¿^Ws it M [Science Council of Japan], ed., Bungaku, tetsugaku, shigaku, bunkenmokoroku VI: Kokugogaku-hen 3C fj Jfh Jfe 3ÖC B II VI: US in ^ [A bibliography of literature, philosophy and history, vol. VI: Studies of the Japanese language] (Aug., 1945—Dec., 1955), (Tokyo, Nihon Gakujutsu Kaigi, March, 1957). Kokuritsu Kokugo KenkyGjo IH JL ID »§• W IS fiff, Kokugo nehkan M s& 'St [The national language year-book] (the 1954 edition, Tokyo, Shüei Shuppan, May, 1954; the 1964 edition, Tokyo, Shüei Shuppan, July, 1964), 10 volumes have appeared so far. Saeki Umetomo 'ifi M, Nakada Norio «f IB ffiit and Hayashi Öki ^ 'X, ed., Kokugogaku M so ^ [Studies in the national language], Kokugo kokubun- gaku kerikyüshi taisei XV HI to IMJC^ffi [A complete history of studies of the national language and literature], Vol. 15 (Tokyo, Sanseido, Feb., 1961). Joseph K. Yamagiwa, ed., Japanese language studies in the Shöwaperiod = The Uni- versity of Michigan Center for Japanese Studies, Bibliographical Series, Number 9 (Ann Arbor, The University of Michigan Press, 1961).