The Modern Japanese Writing System

Before proceeding further, however, it is important that there are no misunderstandings about technical terms. A careful description of how the current Japanese writing system works, much less how it worked in earlier times, is far beyond the scope of this book. For the sake of readers with little or no acquaintance with Japanese, I will dissect a single written sentence in detail here to illustrate the basic terminology and concepts involved; the going may be a little rough, but a brief authentic example is worth a much longer, indirect explanation. Immediately afterward, I will offer some linguistic analysis of key parts of the same sentence for which Chinese characters (kanji) appear. Even those at home in Japanese should read this section, for much of the opposition to script reform in Japan has turned upon the question of whether or not kanji are, in some sense, indispensable. Here is the sentence in the Hepburn romanization:

Kare wa nyūsha irai nakazu tobazu de pinto shinakatta ga, konkai no kiki de wa rīdāshippu o tori, kyūsho o umaku kirinuketa no de, uwayaku oyobi dōhai-tachi kara sukkari minaosareta. (Tagashira and Hoff 1986, 91)

First, I want to emphasize first that there is nothing exceptional or contrived about this sentence. It is a random example of expository written Japanese, selected from a reference work aimed at non-Japanese students of Japanese as a second language devoted to matters other than script.

Second, it is worth mentioning that the romanized version just presented serves quite well as a prompt or cue to Japanese pronunciation for a literate speaker of English, who is likely to know what a macron signifies, can guess that the vowels are supposed to come out roughly as in Italian, and has seen enough movies to imitate a Japanese voice for the consonants and overall delivery. Of course, like all practical writing systems, the Hepburn system of romanization (or rōmaji 'roman letters') greatly underspecifies what one would actually hear if a speaker of Japanese were to say the sentence aloud. It omits, for example, any indication of distinctive rises or falls in pitch (called, somewhat misleadingly, akusento 'accent'); it assumes that the reader is not confused by the fact that the <n> in <nakazu>, coming before a vowel, stands for a consonant whereas the <n> in <pinto>, coming before a consonant, stands for a vowel-like resonant, or that the <n> in <pinto> is articulated more or less like the <n> in English <ant> while the <n> in <konkai> (also a resonant) sounds more like <n> in English <ankle>. Such predictable variations in pronunciation are a problem if you do not know Japanese; if you do, then Hepburn romanization, like any satisfactory writing system, allows for both the transcription and the recovery of any utterance in the language it was designed to serve.

Here is Tagashira and Hoff's translation of the example sentence:

Since joining the company, he had not done anything special and had been rather mediocre. However, in the recent crisis he assumed leadership and got us skillfully through it, and as a result he has been re-evaluated by his colleagues and superiors.

Here is how the sentence looks in ordinary Japanese script:8


Three kinds of characters appear in the text: hiragana, katakana, and Chinese characters or kanji. Alphanumeric symbols (A-Z, a-z, 0-9) and a wide variety of punctuation marks are also found in modern Japanese texts.


The characters in the two kana syllabaries are used, with a few exceptions, to represent the syllables (more precisely, the morae, or syllable parts [J. onsetsu]) actually heard in pronunciation. Hiragana are used by default; katakana are used like italics in English when special marking is prescribed, such as setting off words, like rīdāshippu in the example, recently borrowed into Japanese from languages other than Chinese. Kanji have two basic uses: words borrowed from literary Chinese or coined in Japan on the model of literary Chinese (much as English speakers make up words using Latin and Greek roots) are written with the corresponding kanji; such words are called kango. Kanji are also used to stand in for one or more syllables of certain native Japanese words according to conventional rules that must be learned; words of this kind are called Yamato kotoba.

When used in the writing of kango, the kanji are said to take an on readings; when used in the writing of Yamato kotoba, they are is said to take kun readings. Most kanji have at least one on and one kun; frequently used kanji may have several of each variety. In the case of Yamato kotoba (e.g. dekiru 'be able, develop') and frequently used kango (e.g. taihen 'greatly'), the use of kanji is optional; hiragana may correctly be used in many cases at the writer's discretion. For example, it is virtually obligatory to write the words kiki and dōhai with the kanji strings 危機 and 同輩 , respectively, as seen in our sentence, but the words kirinuketa and minaosareta could just as well be written entirely in hiragana -- きりぬけた instead of 切り抜けた, みなおされた instead of 見直された.

Kun readings were originally glosses based on character usage in literary Chinese; in hindsight, they are attempts to pair up the kanji associated with a word of literary Chinese and a Japanese word that share -- or once shared -- a common meaning. This heuristic is of limited reliability, however, because the use of kanji has evolved over at least a dozen centuries, during which both the Japanese and Chinese languages have undergone enormous changes and erudite Japanese writers have played games, committed scribal errors, and created new usages in the name of art. As a result, there are today many cases in which several different kanji can be used to represent all or part of the same native word; moreover, when part of word is written with a kanji and part with hiragana, historical practice on dividing the word orthographically is not always consistent, leaving the modern writer with a certain amount of latitude. For example, in other contexts, the words 鳴かず and 飛ばずin our sentence could be written 泣かず and 跳ばず, respectively, and in a premodern text 鳴ず or 泣ず (without か) would have been tolerated. (In fact, the situation is even more complicated because of unusual kanji usages known as ateji and jukujikun.9) In short, from a strictly synchronic perspective, the function of kanji in modern Japanese orthography is simply to replace strings of hiragana according to a vast tradition of accumulated conventional rules.

Nevertheless, few reference works characterize the function of kanji in this way. Kanji are instead said to stand for meanings or for words; i.e., they are described, respectively, as ideograms or logograms. Such accounts of kanji usage are implicitly diachronic. They start with the earliest use of kanji in Japanese texts; some even go back to the earliest forms of Chinese characters, suggesting a direct line of continuous development from ancient China to modern Japan. The problem with this doggedly historical approach is that imputes certain psycholinguistic processes and kinds of knowledge to modern literate Japanese without justification. It is certainly worth knowing that literary Chinese was used as a written language for various purposes in pre-modern Japan, that this practice led to the development of many varieties of written Japanese, and that these, in turn, underlie many of the conventional rules that now specify which kanji should stand for syllables in which words. (For details, the ambitious reader can consult Seeley 1991 and Twine 1991.) From the perspective of a Japanese youngster learning to become literate today, however, the etymological stories that explain why this or that kanji takes a certain reading in this or that word are just ex post facto rationalizations for rules that are essentially arbitrary, much like the explanations of anachronistic English spellings that teachers give British and American students as they expand their literate vocabularies. Our example sentence offers many instances of how such superficial "semantic" accounts of kanji usage fail.

Let's begin our linguistic analysis with nyūsha 入社 and irai 以来 .The latter is very old, a genuine borrowing from literary Chinese into Japanese; the former, which shares the same outward appearance, is a new coinage, a compound of the Sino-Japanese verb nyū 'enter' 入 with a putative Sino-Japanese noun sha 社 to form a new noun meaning 'entering a company'. The verb-first order here is strictly Chinese -- the Japanese grammatical order would be goal followed by verb10; moreover, sha is not a free noun but only the second syllable of kaisha 会社 the word Fukuzawa Yukichi devised as a translation-equivalent for English 'corporation' in the 19th century. Fukuzawa pressed the Sino-Japanese sha designated by 社 into service in his new word, retroactively augmenting its original sense in literary Chinese (a deity associated with a place) and its extended sense in pre-modern Japanese usage (in which it was associated with words denoting Shinto shrines).

The phrase nakazu tobazu de 鳴かず飛ばずで literally refers to a bird that neither sings nor flies, hence proverbially to a person of undistinguished achievements. As in countless other instances, there would be nothing wrong if this phrase were written entirely in hiragana (なかずとばずで) since it contains no Sino-Japanese words (kango). But of greater interest is the fact that in modern Japanese the verb naku 'cry, weep', of which nakazu is an obsolescent negative, is more commonly written with the kanji-hiragana combination 泣く. Prescriptive rules dictate that the kanji shown the example sentence should be used because the understood subject is 'bird', not 'person', for which 泣 is supposed to be correct. The difference goes back to the difference in meaning of the two Chinese words associated 鳴 and 泣 , respectively; in Japanese, however, there is only one word, naku, which admits human or non-human animate subjects indifferently. The situation is somewhat analogous to the rule in English that prescribes writing "It was Father who rose first from the table" but "his father is a lawyer": in speech, there is no difference in pronunciation; the syntax of the sentence (no article or possessive pronoun before "Father") makes it unmistakable that 'my/our father' is meant. The use of capitalization in the written version is thus strictly redundant.

Nakazu, tobazu, and minaosareta at the end of the sentence are also interesting in that the roots of these verbs are, respectively, nak-, tob-, and minaos-. No words end in consonant in spoken Japanese (in final position, <n> is a resonant), but these consonant-ending forms are the longest needed to predict every inflected form of these and the majority of other verbs in the language by rule. What is interesting is that the parts of these words written with kanji are, respectively, just na 鳴, to 飛, and minao 見直 ; the remainders are written in hiragana (called, when used for this purpose, okurigana). In terms of linguistic analysis, this means that the kanji stand for less than whole morphemes (the shortest units of speech associated with definite meanings). In other cases (e.g. the ateji mentioned above), a kanji may represent more than a whole morpheme.

The word uwayaku is remarkable because uwa- is a bound form of the native word ue 'top, above'; it only occurs in compounds. Here, it is prefixed to yaku '[an] official', a loan from Chinese. If you know Japanese, then you know this word; you might find it odd that it was written上役 , with two kanji, until someone explained that ue was pronounced uwe at an earlier stage of the language, but you would be unlikely to make the mistake, common among non-Japanese learners of Japanese as a second language, of misreading this word at jōyaku, applying a common but here incorrect on reading to the first kanji. Many non-Japanese believe that, because they can guess the meaning of a word like uwayaku by knowing the meanings of other words written with the same kanji, "knowing a kanji" in this ad hoc sense is sufficient for written communication; for them, the correct reading is a mere detail of little consequence. While it is easy to see why they should think this way, they are mistaken -- as Chinese college students who think they can coast through Japanese texts by looking only at the kanji.

As already pointed out, the "meanings" associated with many kanji do not remain fixed over time. The 役 that stands for yaku in uwayaku, for example, now has multiple "meanings" in modern Japanese: it shows up in yakusha '[stage] actor' 役者, yaku ni tatsu 'serve a purpose, be useful' 役に立つ , and other expressions that no longer have much to do with each other, if they ever did. The opportunities for going astray are as numerous as the chances for making a lucky guess. Even more important, if and when native speakers do guess at words written with kanji, what they are trying to guess is generally a reading, not a "meaning"; unlike the non-Japanese learner, they can have native-speaker intuitions or entertain etymological speculations in Japanese itself.

This applies to Yamato kotoba as well as to kango, as shown by the verbs kirinuketa切り抜けたand minaosareta見直された, two especially rich words in our example. Morphologically, these are compounds of native verbs: kiru 'cut', nukeru 'come off, escape, be omitted', miru 'see, look at', and naosu 'adjust, repair'. But historically, the compounding took place so long ago that each verb is now just a single lexical item: kirinukeru means roughly 'struggle through [a difficult situation]'; minaosu means roughly 'think better of someone'. Therefore, it is anachronistic to use kanji in writing these words so as to imply that they are "still" compounds of kiru and nukeru, miru and naosu, all of which are in productive use as separate verbs in the modern language; the prescriptive writing rule is contradicted by the lexical structure of the modern language. In fact, naosu can now be used after most verbal stems to form a compound meaning 'do X again, redo X'; e.g. mō ichido minaoshite kudasai 'please look at it again'. It would not be wrong to use the same kanji seen in our example to write the verb in this sentence as well, yet they are clearly not forms of the same word as far as the modern language is concerned.

As even this cursory analysis should make clear, to ascribe meaning directly to kanji is to confuse historical development with linguistic competence. From the standpoint of a young native speaker of Japanese, learning how to read and write with kanji is not a process of learning an a priori theory of "meanings" linked to kanji and juggling them to make up representations for previously unknown words, but rather a process of learning conventional rules that specify when and how to use kanji to write known words and developing an a posteriori theory to keep track of the rules. Anyone who has observed Japanese schoolchildren knows that they have vast and rapidly growing active vocabularies that far exceed what they are able to read or write with the prescribed kanji.

I hasten to add at this point that tens of thousands of Japanese regularly succeed in coping with kanji despite the onerousness of the task. The purpose of the foregoing discussion, apart from introducing some important terminology, is to begin undermining the notion that kanji fulfill some sort of mystical role in the Japanese writing system that makes them indispensable. In Chapters 1 and 3, we will look at facts about language and script in general and about alternative methods of writing Japanese that complete this process. The aim here is not to suggest that the writing system is so difficult that it is dysfunctional. That is a claim that script reformers have made, and we shall consider it separately, in chapter 2, on the basis of demographic data collected in Japan. It is only to establish that the rules for using kanji in modern Japanese writing, in and of themselves, are not of the same kind as the "rules" of Japanese phonology, syntax, or semantics. The former, like the notorious "rules" of English spelling, must be consciously learned; while not without some historical justification, they are synchronically arbitrary. The latter may not even be rules or programmatic instructions in the brain at all; we cook up theories of "rules" to summarize the reality that native speakers acquire naturally by growing up among other native speakers during the crucial period of early childhood that precedes schooling and without which schooling and literacy would be quite impossible.

In brief, one must recognize that kanji are not necessary for the writing of Japanese. This is not at all the same as calling for their abolition from Japanese writing. It has been my unfortunate experience that those emotionally attached to the indispensability of kanji insist that anyone who thinks otherwise must be a kanji abolitionist. This is of course an utter non sequitur. Still, to preclude any possibility of misunderstanding, let me point out the practical obstacles to doing away entirely with kanji in Japan, for they are so great as to make such a development virtually impossible at this time.

There are basically three kinds of obstacles. First, too much material exists in traditional written form; it can only be replaced gradually, and some of it must be preserved for as long as possible because it possesses intrinsic artistic or historical value. Second, many tens of thousands of Japanese have a personal stake in the maintenance of the orthographic status quo; their replacement in the labor force, let alone other areas of daily life, would require at least a generation. Third, there is, at present, no consensus on a standard with which to replace the existing writing system; there are many workable alternatives besides romanization, and, even with romanization, there are different systems that could be adopted.

Finally, apart from these material, educational, and orthographic obstacles, there is no evidence of the political will needed to abandon kanji; therefore, the most that could conceivably happen would be a gradual phasing out of kanji over a several decades. Even if such a phase-out were to begin, it might never reach completion: the result would more likely be an indefinite coexistence between traditional script and a new orthography free of Chinese characters, a state of digraphia in which each form of writing would occupy largely complementary niches in Japanese life. It is even possible that such a state of affairs might arise in the absence of direction from government: even today, the vast majority of those who use Japanese script on computers input data in romanization; to that extent, even though they may refuse to read data in romanized form, they already, in a psychologically fundamental way, make use of an alphabetic representation of Japanese words and phrases.

Having taken note of the political and practical forces that mitigate against the abolition of kanji, let me recall the lesson of the analysis of our example sentence: the function of kanji in the current Japanese writing system is to replace strings of kana. The historically accumulated heuristics underlying the conventional rules that prescribe how and when such replacements should be made are neither a part of the Japanese language nor a necessary part of the knowledge that enables literate Japanese to read and write. Thus, there are no linguistic reasons why kanji could not be abandoned. To those who insist that holding this view necessarily entails a personal desire to see kanji eliminated from the Japanese writing system, all I can say is that linguistics is an empirical science11 and has yet to find any evidence that writing in general, much less the details of a particular writing system, influence speech.12 As far as I can see, the doctrine of indispensability of kanji stands in relation to the science of linguistics in much the same way as so-called scientific creationism stands in relation to the science of biology. I do not know of any biologists intent on curtailing religious freedom, but there are many creationists who would turn the teaching of biology to their own ends. Likewise, I think it only fair to point out that, whatever the political and social foibles of the Japanese script reformers, they were not proposing anything that was -- or is -- impossible in principle.


  1. Sampson (1985) prefers the term semasiogram but does not himself classify Chinese writing as semasiographic. Nevertheless, he does claim that semasiographic writing systems exist; he specifically cites the so-called picture writing of the Yukaghir of Siberia. For a refutation of Sampson's claims about specimens of Yukaghir "writing" (which turn out to be by-products of a game played by young women, not actual writing at all), see DeFrancis 1989.
  2. The linguist Roy Andrew Miller has proposed a completely different, even more mean-spirited interpretation. "Mori Arinori and the élite that he represented found their long-cherished Japanese sociolinguistic preserve more and more threatened by invasion from the rest of Japanese society. What better way to halt the rising tide of the great unwashed than to declare the entire Japanese language incompetent and inadequate, and to urge instead the wholesale adaptation of a foreign language solely available -- and then, only under circumstances of enormous difficulty and expense -- to a limited number of the élite?" (Miller 1986, 97). Not only is there no documentary evidence supporting this cynical view, it totally ignores elementary facts about the nature of spoken and written language in Meiji-period Japan explained in the second part of this chapter.
  3. It is perhaps worth mentioning that Nitobe and Fukuzawa are now pictured on the \5,000 and \10,000 bill respectively.
  4. "Saitō Hidekatsu, Editor in Chief of the Moji to Gengo ('Writing and Language') and the Latinigo (an Esperanto magazine), was arrested in the autumn of 1938 and died in prison the following year while still awaiting trial, charged with advocating language reform. Hirai Masao, author of Kokugo Kokuji Mondai ('Problems of the National Language'), and a member of the Nippon Romaji Kai (Japanese Romanization Society) was arrested on 5 June 1939, held without trial or formal charge in a police station till the following June, then transferred to a prison and held till December 1940 when he was released on bail. In January 1941 he was tried by Chief Judge Iizuka Toshio, and sentenced to two years' penal servitude for advocating romanization of the writing system. He was given a suspended sentence for three years" (Hall 1949, 19).
  5. This unfortunate case of scholarly abuse was first documented in Unger 1991.
  6. Their preoccupation with anticommunism and desire to punish former ultranationalists no doubt also biased their judgments.
  7. Working independently, Mr. Banno Yūichirō, recently retired from the National Diet Library, and I located one of the key "missing" documents in early 1991; we met for the first time and both reported our findings at the 42nd All-Japan Conference on Romanization Education Research.
  8. In the text, the suffix tachi is omitted in the Japanese script version of the sentence. I've taken the liberty of fixing this oversight.
  9. Ateji are kanji used in an ad hoc way to write a Japanese word. There is no etymological relationship between the readings of the kanji and the portions of the word to which they are made to correspond. Writing medetai 'happy, auspicious' as 目出度い (where 目stands for me, 出 for de, and 度 for ta) is an example of ateji. Even the kanji used to write common words borrowed from Chinese can lose their etymological value due to a change in word-meaning in Japanese, as has occurred in the case of kyōdai 'sibling(s)' < 'male siblings; elder and younger brother(s)' (see Kaneko 1986). Jukujikun are a special case of ateji in which a string of two or more kanji is associated with a whole native word yet none of them represents a specific portion of the word. Writing asita 'tomorrow' as 明日 is an example of jukujikun. Note also that In numerous kanji-plus-okurigana combinations, the part represented by the kanji has no independent grammatical existence. In all such cases, whatever logographic information is present is conveyed not by individual kanji but by their linear combination with each other and/or adjacent kana. The situation is much like English <write> and <right>, in which the individual letters, though only crudely phonographic, are certainly not logographic; the distinctive sequence of the letters in each string adds morphological and syntactic information about the syllable /rayt/ that both strings represent.
  10. I.e., in Japanese clauses, verbs come at the end. In some compound nouns, a form of a verb may precede a noun, as in tabemono 'food' < tabe 'eating' + mono 'thing'; but in such cases, the basic modifier-modified order of Japanese determines the meaning (tabemono = 'that which one eats', not *'[the act of] eating things'). In compound nouns that are nominalized clause, the verb comes last; e.g. monogatari 'tale, story' < mono 'thing' + katari 'telling'.
  11. There is an unfortunate tendency today, especially among computer scientists, to regard linguistics as "soft"; perhaps this is because of the pretensions to "hard" science of the particular school of linguistics originating with Noam Chomsky, as Ellis (1993) suggests. Whether or not one accepts Chomsky's claims about innate linguistic capacity and his reductionist approach to meaning, the fact remains that linguistics is the scientific study of language and its use just as biology is the scientific study of living organisms and their interaction with the environment. As Harris (1993, 10-11) observes, "Ultimately, the matter of linguistics' fit to the category of science . . . is a pretty trivial one. . . . Its methods, goals, and resilient results come from a long tradition of treating language as a natural object -- sometimes a social object, sometimes a mental object, sometimes both, but always as something which could be observed, like the stars and the rocks, and sometimes poked, like the animals and the plants."
  12. See, for instance, Aronoff 1994, a review of a collection of papers "centered around the idea that literacy promotes linguistic awareness and hence linguistic analysis." Some articles in this volume "exploit this idea admirably, showing how literacy affects ideas about aspects of language ranging from phonology to syntax. . . . Other articles go beyond this connection between literacy and awareness or analysis to a much bolder claim: that written language has an effect on spoken language. Linguists have long rejected the possibility of such influence, and the argument provided here should not change that" (619a).