The subject of this article is the rendering of Japanese into roman letters, mostly as a guide to correct pronunciation but also as an aid for looking up information and inputting Japanese text by a computer keyboard.

Because of its simple set of sounds, Japanese can easily be written with 23 roman letters (abcdefghijkmnoprstuvwyz, that is, the 26 letters minus l, q, and x) and two special symbols (^ or ¯, and ’). The ^ (circumflex) or ¯ (macron) is written over a vowel to represent a lengthened vowel, and the ’ (apostrophe) is used between the end-of-syllable n and a following vowel or y.

The Japanese term for romanization is ローマ字綴り (also written ローマ字つづり) Rōma-ji tsuzuri (literally, Rome-letter spelling).

The worldwide de-facto standard for writing Japanese words in roman letters is Hepburn romanization.

Both “transcription” and “transliteration” refer to transforming the characters of one language into the characters of another language. But with transliteration, this transformation is always reversible; that is, given a transliteration of a text, one can unambiguously reconstruct the exact spelling of the original text. For example, if two different characters like お and を are represented by the same letter o, (or え and the particle へ are represented by the same letter e), we are dealing with a transcription. But if お and を are represented differently by o and wo (or え and へ are represented differently by e and he), we are dealing with a transliteration.

“Romanization” is a transformation into roman letters (AaBb...Zz) of the characters of a language that is not written with roman letters (Japanese, Korean, Russian, Arabic, Thai, Georgian, Hindi, etc.). A romanization could be either a transcription or a transliteration.

Since a 1954 cabinet order (kunrei), three systems of romanization have been recognized in Japan: Kunrei, Nippon, and – alternative to these and primarily used when dealing with foreign countries – Hepburn. The text of this cabinet order (in translation) is given in the Official romanization rules, and a comprehensive overview of how specific syllables are romanized according to the Hepburn and Kunrei romanizations is presented in the Kana romanization tables.

All three systems of romanization use the same 23 letters and two special symbols (given above) for rendering Japanese in roman letters. The differences between them concern only some of the syllables, but how the syllables are represented is decisive for approximating their correct pronunciation and for searching for words in databanks and traditional reference works.

The officially favored Kunrei romanization is taught to children in their fourth year of school. On the other hand, special regulations for passports, road signs, and maps prescribe spellings that are based on Hepburn romanization. There is no unified national system of romanization. And various institutions and associations have their own often very detailed rules for romanization. In some cases such detailed rules have been developed to replace the traditional Japanese writing system.

Outside of Japan there have been and are national romanization systems for spelling Japanese words. An example in the United States is standard ANSI Z39.11-1972 of the American National Standards Institute, which is titled “American National Standard System for the romanization of Japanese”. This standard, which is based on Hepburn romanization, was abandoned in 1994 in favor of the ISO standards (Kunrei and Nippon), which had been introduced five years before. A similar standard, BS 4812:1972, also introduced in 1972 and likewise based on Hepburn, was adopted in Britain under the title “(British Standard) Specification for the romanization of Japanese”.

In addition, national scientific organizations like libraries and library associations have been working to develop romanization standards, because searching the literature is only possible with a unified way to represent foreign text in roman letters. One such standard is set forth in “Manual of romanization, capitalization, punctuation, and word division for Chinese, Japanese, and Korean” (1959) of the American Library Association and the Library of Congress in Washington. This, like the Japan collection (books from and about Japan) in the Staatsbibliothek in Berlin, is based on Hepburn romanization.

The 1989 standard ISO 3602:1989 for the “Romanization of Japanese (kana script)” of the International Organization for Standardization is based on Kunrei romanization but also allows Nippon romanization for a “stringent transliteration”. This marks a failure of the effort to establish a universally accepted international standard. A proved international de-facto standard (Hepburn) cannot simply be replaced by a newly proclaimed standard (Kunrei and Nippon) that is neither practical nor is in wide use but is merely being imposed on the international community by nationalistic Japanese interest groups. Even Japanese government agencies use Hepburn romanization within Japan and abroad. Another obstacle hampering the further spread of the ISO standard is that these two official documents are available only in English and French, and can be ordered only in printed form or as a PDF file for a price. Finally, the ISO rules are riddled with inconsistencies and do not even fulfill their purpose of making it possible to deduce the original kana spelling from the romanization. However, a revision of the standard has been announced. A version of the ISO standard in Japanese is found here ( 988 KB).

Yet it remains hard to understand how political interest groups in a single country, for whose people roman letters are something foreign, could impose on the people of many other countries who have grown up with roman letters, as a binding standard, a version of romanization that is impractical and foreign to their language. The absurdity of such outside interference is illustrated by imagining the United States, Australia, or Great Britain trying to prescribe for Japanese, via an ISO standard, how English words and proper names shall be transcribed in Japanese.

Especially in instructional materials, one repeatedly finds individual solutions that try to adapt the representation of words to the linguistic facts of the targeted market. Thus one may find r transcribed as l (the IPA sound ɽ lies between the two), or (in German materials) the voiced consonant z represented by s.

Such individual transcriptions in the widest sense include the sometimes humorous practice of many Japanese to spell their given or family name in an Anglicized or Germanized way that is intended to make it easier for foreigners to pronounce and remember. Thus the given name Jōji is occasionally spelled George, and the manga artist Naitō Yasuhiro writes his surname as Nightow.

Regardless of the type of publication (mass media, technical journals, instructional materials, dictionaries, data collections), Japanese words and proper names are spelled in a relatively uniform way worldwide, based mostly on Hepburn romanization. Such uniformity of transcription is not always found in other languages: consider the alternate spellings Peking/Beijing, el Kaida/al-Qaeda, Tschaikovsky/Çajkovskij. To be sure, in some countries there are deviations from the standard romanization of Japanese words, either to indicate a closer pronunciation to the original (Noh for nō in English) or to hark back to an older spelling convention (Tokio for Tōkyō in German). In the mass media the circumflex or macron (over long vowels) and the apostrophe (between the syllable n and a following vowel or y) are regularly dispensed with, so that the correct Hepburn transcription of the given name Jun’ichirō is regularly spelled Junichiro in the press. On the Internet, the long vowels that are marked with a circumflex or macron in Hepburn romanization are usually indicated, in a way analagous to kana spellings, by an appended u (after o), an appended i (after e), or a doubling of the vowel (for a, i, u, and the katakana e).

As has been mentioned, Japanese schoolchildren learn the letters of the ABC alphabet first by the Kunrei romanization of Japanese words. But when they later begin to learn English, they are confronted with sometimes completely different pronunciations for some of the letters and letter-combinations. After learning Kunrei romanization in school, they will hardly ever encounter it again, for Hepburn romanization is nearly ubiquitous in advertising and in the English words one runs into at home and on the street. Some of the problems that Japanese have with Hepburn romanization and English pronunciation trace back to their earlier misleading experience with Kunrei instruction.

Japanese who want the foreigners they deal with to pronounce their name correctly spell it in Hepburn romanization – for example, Fujii Jun’ichirō or, without diacritical marks, Fujii Junichiro – rather than the mispronunciation-inviting Kunrei romanization Huzii Zyunitiro. And to make sure that the names of their citizens are pronounced correctly abroad, Japanese authorities go one step further and transcribe the long vowel ō in travel papers with an appended h, as seen in the most common Japanese family name Satoh (instead of Satō, Sato, or – a popular variant spelling – Satow).

Because of its widespread use in Japan and internationally, the de-facto standard for writing Japanese words in roman letters is Hepburn romanization (in Japanese, Hebon-shiki Rōma-ji, literally, ‘Hepburn-type Rome letters’, also known as Hyōjun-shiki Rōma-ji, literally, ‘standard-type Rome letters’). This is the only one of the three officially recognized romanization systems that allows foreigners, with no prior knowledge, to pronounce Japanese words relatively correctly and intelligibly. Consonants are pronounced approximately as in English, and vowels approximately as in Italian. To be sure, the syllable n is transcribed unchanged as n before the consonants b, p, and m, where it is pronounced m, but even here it is sometimes transcribed as m.

Critics of this romanization, especially Japanese and overwhelmingly devotees of the Kunrei and Nippon romanizations, argue that Hepburn romanization is unsuitable internationally because it is so closely dependent on English. This ignores the fact that English is by far the most widespread language that is written with roman letters. Besides the world’s native speakers of English, there are over a billion Indians and the population of further countries in which English is an official language or is taught as the primary foreign language. And finally, in almost all languages, even non-European ones like Turkish, most roman-letter consonants are pronounced as in English.

How the individual kana characters are transcribed according to Hepburn can be inferred from the Kana romanization tables; the differences in the pronunciation of the roman letters in the three romanization systems are presented in the following section “Kunrei romanization” and in A comparison of the romanization systems.

Kunrei romanization (in Japanese, kunrei-shiki Rōma-ji, literally, ‘cabinet order-type Rome letters’) transcribes kana partly as they are pronounced and partly schematically according to the position of the kana within the Fifty sounds table.

Vowels are transcribed as in Italian and consonants as in English, with the following exceptions: the initial sound of a syllable in a row of the Fifty-Sounds Table that is named for a consonant is predominantly spelled with that consonant, even when its pronunciation deviates from the pronunciation of that consonant. This applies to the following syllables (whose pronunciation/transcription is given in parentheses according to Hepburn romanization):

An exception are the row-d syllables ぢ zi (ji) and づ zu (zu), which are consistently transcribed as di and du only in the Nippon system.

As can be seen from the above listing, the small end-of-syllable sounds ゃ, ゅ, and ょ are consistently spelled ya, yu, and yo.

The Kunrei system is taught in the fourth year of elementary school, before any instruction in English (which generally begins in the seventh year), so that before schoolchildren learn how roman letters are pronounced in English, they learn how they are pronounced in Kunrei, which is sometimes a very different pronunciation. The trend toward beginning English instruction earlier, possibly at the same time as or even before Kunrei instruction, throws pupils into even more confusion, confusion that is constantly lamented by Japanese commentators. One gets the feeling with Kunrei of having to learn an idiosyncratic, useless romanization that moreover makes it harder to learn English and is contrary to the everyday experiences imprinted through Hepburn romanization (in advertising, on train and subway station names, etc.) and English (which is ubiquitous). The introduction of Kunrei as an ISO standard in 1989 makes an originally domestic Japanese problem into an international problem.

Nippon romanization (in Japanese, Nippon-shiki Rōma-ji, literally, “Japan-type Rome letters”) differs from Kunrei only in the following infrequently occurring syllables (K stands for Kunrei, H stands for Hepburn):

ぢ di (K: zi, H: ji), づ du (K: zu, H: zu), ぢゃ dya (K: zya, H: ja), ぢゅ dyu (K: zyu, H: ju), ぢょ dyo (K: zyo, H: jo).

The advantage of romanizing the identically pronounced but different kana じ and ぢ with zi und di, and ず und づ with zu und du, is that one can infer the original character from the romanized spelling. This facilitates the reconstruction (back-transliteration) of the original kana and the writing of the correct kana via a computer keyboard. Most input systems allow the user to differentiate between two different kana that have the same pronunciation, although at the price of having a few kana transliterated in a way that misrepresents how they are pronounced. Another characteristic of the Nippon romanization is that the same letter is pronounced differently (such as z in za and zi, or d in da and di), and sometimes the same sound is spelled differently (such as the IPA sound ʤ in zi and di).

The following summary lists kana and kana-combinations that are romanized differently in the three official romanization systems, along with examples of words and the pronunciation of the syllable in the International Phonetic Alphabet (IPA) of the International Phonetic Association.

Transcription as an aid for pronunciation

As shown by the above examples, Hepburn is the only one of the three romanization systems that provides a “pronunciation-true” transcription, meaning true to the way the letters are usually pronounced in most European languages, as documented for example in the IPA transcription. Kunrei and Nippon require that some letters be pronounced in a way that is not done in any world language. Another problem with both Kunrei and Nippon is that they require the same consonant to be pronounced in two different ways: s as s or sh, t as t or ts, etc. Why should people whose language is written with roman letters have to learn a romanization system in which these letters have a pronunciation that is completely different from the pronunciation they are used to? The slight advantages that Kunrei and Nippon offer over Hepburn do not compensate for the drawbacks of a less pronunciation-true transcription.

When it is important to come closer to the correct Japanese pronunciation than Hepburn romanization allows, there is always the possibility of resorting to a purely phonetic transcription, such as the IPA alphabet. But such phonetic transcriptions are not suitable for everyday use, because they require too many special symbols and diacritical marks for fluent reading and writing.

Transcription as an aid for looking up information

Knowledge of Hepburn romanization is indispensable for information retrieval, because almost all dictionaries, glossaries, databanks, and other reference works spell Japanese terms with Hepburn romanization.

Transcription as an aid for inputting Japanese text by computer keyboard

Usually when writing in Japanese on a computer keyboard, the text is input in roman transcription, optionally according to Hepburn, Kunrei, or Nippon romanization; the common Japanese word processing programs allow for all three. Long vowels are input according to how they are written in kana; for example, a long o is input as ou, instead of an an o with a circumflex or macron (ô or ō). As letters are keyed in, they are automatically converted, as specified, into either hiragana or katakana. And these kana phrases are in turn converted, as desired, into kanji. The Rōmaji-kana conversion tables show which letters and letter-combination to type in order to produce given kana and kana-combinations.

Transcription for special purposes

For linguistic purposes too, it is often indispensable to represent Japanese in romanization. The kana of Japanese form a syllabary, in which each symbol represents an entire syllable without specifying the phonemes that constitute the syllable. So the only way to specify the beginning consonant sounds of syllables like の or ざ would be with symbols like n or z – symbols that are readily available just by adapting a roman-letter phonetic alphabet, that is, by the simple expedient of romanization.

A true transliteration of Japanese would represent each Japanese symbol (hiragana, katakana, kanji) with a unique roman letter or letter-combination. It is possible to devise such a one-to-one transliteration for the kana symbols, allowing the original kana to be reconstructed from its romanization (e.g., from di we could deduce the original kana ぢ). But the kana represent less than one percent of all the different symbols (mostly kanji) that are used in written Japanese, so such a transliteration would have no practical utility.

Bottom line:

A comparison of the three romanization systems shows that Hepburn romanization is the only one that is suitable for all three purposes of a transcription system: as an aid for pronunciation, data retrieval, and keyboard input.

Portuguese missionaries were the first to reproduce Japanese texts in roman letters, in 1591, and soon thereafter they produced romanized bilingual dictionaries. The Portuguese were succeeded by Japanese who studied Westerns sciences from books written in Dutch as well as German and French scholars. These early romanizations of course differed from each other, because each followed the pronunciation of roman letters in its own non-Japanese language. And because each of these languages was spoken by a very limited number of people, none of these romanizations came into wide use. The spellings Tokio (for Tōkyō or Tokyo), Jokohama (for Yokohama), and Fudschi (for Fuji), spellings which are still seen in Germany today, are relics of a former romanization designed for Germans.

After Japan was forced open to international trade by the United States in 1854, commercial and cultural exchange with foreign countries suddenly assumed greater importance than ever before. In response to this need, a series of bilingual textbooks and dictionaries with romanized spellings of Japanese words were developed to help foreign merchants, diplomats, scientists, and others acquire a working knowledge of Japanese.

Among the many foreigners resident in Japan was the American physician and missionary James Curtis Hepburn (1815-1911), who produced “A Japanese and English Dictionary” in 1867. As a member of a commission consisting of Japanese and foreigners, he worked on developing a unified romanization. The transcription rules worked out by the commission first gained wide dissemination in the third edition (1886) of his dictionary, which is why this romanization now bears the name Hepburn. In Hepburn romanization, consonants are pronounced as in English, and vowels are pronounced as in Italian. After a few modifications, particularly those advocated by the Japanese “Society for the Dissemination of Romanization” (Rōmaji hirome kai), the version that is still used today was published in 1908 and came into wider use under the designation hyōjun-shiki (standard style) or shōsei Hebon-shiki (revised Hepburn style).

In 1885, in reaction to the foreign, primarily American, influence on the transcription of Japanese into roman letters, the Japanese physicist Tanakadate Aikitsu (1856-1952) published his Nippon-shiki Rōma-ji (Japan-style romanization), which in his opinion was better suited to the Japanese understanding of a systematic transcription. In his system, the beginning consonant sound is represented by the same letter for all five kana in each line of the Fifty sounds table, even though in some cases this leads to considerable deviations from the actual pronunciation. In 1937 a cabinet order (kunrei) gave this romanization, in modified form, official status as the only sanctioned form of transcription. In 1945, under the American occupation, the Kunrei system of romanization was replaced by the Hepburn system, until in 1954 the Japanese state, its sovereignty restored, again gave preference to the Kunrei system. Since the last third of the 19th century, the three romanization systems have been a political football pitting – to state the matter in an oversimplified way – Japanese nationalists (Nippon, Kunrei) against globalization (Hepburn).

International Standard ISO 3602. Documentation – Romanization of Japanese (kana script). First edition 1989-09-01. Genève: ISO 1989. V, 6 p.


October 2005, W.H. with Mark Spahn