Japanese Romanization: they still haven't decided

« previous post | next post »

All Japanese individuals who have attended elementary school since WW II have been taught to read and write romanized Japanese, and romanization is widely used for computer inputting and for other specialized purposes, particularly for those involving foreigners who do not know kana and kanji, but by no means for everyday reading and writing by Japanese citizens.  There are numerous different schemes for the romanization of Japanese, but the three main ones are:  HepburnKunrei-shiki, and Nihon-shiki.  More about each of them below, but first a rough comparison of the two leading systems:

From Momoko Jingu, "Cultural agency now weighing romanization of Japanese words", The Asahi Shimbun (10/1/22).

“Aichi” or “Aiti,” which is the preferred spelling?

That is one of the questions the Council for Cultural Affairs is now considering as it ponders ways to reduce confusion over the use of two different romanization systems for a language deemed to be one of the world’s most difficult to master.

A Cabinet notification issued in 1954 called for using the “kunrei”  [VHM:  "instruction"] style of romanization when writing Japanese words in English. But the notice also allowed for exceptions with the Hepburn romanization style if circumstances made it difficult to change established writing habits.

The kunrei style is mainly learned during Japanese language classes in elementary school. But the Hepburn style is more often used in the romanization of Japanese names in passports as well as on road signs.

“People use the kunrei style unconsciously and there are various opinions about which style is better,” said an official with the Cultural Affairs Agency. “We plan to hold discussions about how romanization should be approached after further studies about what works best.”

A few interesting notes about each of the three main systems:

Hepburn romanization [VHM:  devised by a Western missionary in 1867] generally follows English phonology with Romance vowels. It is an intuitive method of showing Anglophones the pronunciation of a word in Japanese. It was standardized in the United States as American National Standard System for the Romanization of Japanese (Modified Hepburn), but that status was abolished on October 6, 1994. Hepburn is the most common romanization system in use today, especially in the English-speaking world.

Nihon-shiki romanization [VHM:  devised by a Japanese physicist in 1885] was originally invented as a method for Japanese to write their own language in Latin characters, rather than to transcribe it for Westerners as Hepburn was. It follows the Japanese syllabary very strictly, with no adjustments for changes in pronunciation….  Also known as Nippon-shiki, rendered in the Nihon-shiki style of romanization the name is either Nihon-siki or Nippon-siki.

Kunrei-shiki romanization [VHM:  promulgated in 1937 by the Japanese cabinet, whence it is often referred to as "Cabinet romanization"] is a slightly modified version of Nihon-shiki which eliminates differences between the kana syllabary and modern pronunciation. For example, the characters  and  are pronounced identically in modern Japanese, and thus Kunrei-shiki and Hepburn ignore the difference in kana and represent the sound in the same way (zu). Nihon-shiki, on the other hand, will romanize  as du, but  as zu. Similarly for the pair  and , they are both zi in Kunrei-shiki and ji in Hepburn, but are zi and di respectively in Nihon-shiki….


It is fascinating to contemplate that the most ardent advocates of the Nihon-shiki romanization have been members of the Oomoto sect:

Oomoto (大本Ōmoto, Great Source, or Great Origin),[1] also known as Oomoto-kyo (大本教Ōmoto-kyō), is a religion founded in 1892 by Deguchi Nao (1836–1918), often categorised as a new Japanese religion originated from Shinto. The spiritual leaders of the movement have always been women within the Deguchi family….


This fits well with the history of phonetic script of kana as being closely associated with women — onnamoji 女文字 ("women's writing / script") or onnade 女手 ("women's hand") that developed during the medieval perios.  Cf. Chinese nǚshū 女書 ("women's writing / script") in late imperial and early 20th-century southern China.

Most intriguing of all, romanization was seriously considered as a replacement for the Japanese wriring system:

In the Meiji era (1868–1912), some Japanese scholars advocated abolishing the Japanese writing system entirely and using rōmaji instead. The Nihon-shiki romanization was an outgrowth of that movement. Several Japanese texts were published entirely in rōmaji during this period, but it failed to catch on. Later, in the early 20th century, some scholars devised syllabary systems with characters derived from Latin (rather like the Cherokee syllabary) that were even less popular since they were not based on any historical use of the Latin script.

Today, the use of Nihon-shiki for writing Japanese is advocated by the Oomoto sect and some independent organizations. During the Allied occupation of Japan, the government of the Supreme Commander for the Allied Powers (SCAP) made it official policy to romanize Japanese. However, that policy failed and a more moderate attempt at Japanese script reform followed.


From my own informal survey of script usage in Japan since the early 70s, the proportion of kana to kanji in general use has definitely increased, while the proportion of romaji to kana has also increased.  It will be interesting to see how this all plays out in the coming decades, especially with developments in electronic and digital software and hardware continuing to proliferate in unexpected ways.

Selected readings

[h.t. June Teufel Dreyer]


  1. Noel Hunt said,

    October 2, 2022 @ 4:56 pm

    It is not quite correct to say that [dz] づ and [z] ず are pronounced identically. Vance (An Introduction to Japanese Phonology, 1987) quotes Kawakami who states that [dz] always appears word-initially and after the mora nasal (ん [n], in Japanese), but [z] usually appears inter-vocalically, although not always. Vance adds that this accords with his own observations, but that [z] may appear word-initially when a pause does not precede. (Kawakami Shin, Nihongo Onsei Gaisetsu, 1977, 'An Outline of Japanese Sounds')

  2. Carl said,

    October 2, 2022 @ 5:32 pm

    The most common Japanese romanization is WaPuro-shiki which results in barbarisms like “jya” instead of either “ja” or “zya”.

  3. Jonathan Smith said,

    October 2, 2022 @ 8:05 pm

    Re: Noel Hunt's comment I thought the point was that most (?) varieties are neutralized such that the sound-symbol correspondence is mixed, that is, positional allophones [dz]~[z] wind up written (superficially) indifferently with "づ~ず"… cf. Spanish where "b"/"v" etc. write conditioned [b]~[β̞] etc. if I understand rightly.

  4. Jonathan Smith said,

    October 2, 2022 @ 8:09 pm

    Well, FWIW wikipedia Japanese phonology suggests that the above description is outdated as Maekawa 2010 finds that both variants were found in all positions, and that the time it takes to produce the consonant or consonant cluster […] was the most reliable predictor for affricate realization."

  5. Jim Breen said,

    October 2, 2022 @ 9:57 pm

    >> Hepburn romanization [VHM: devised by a Western missionary in 1867]

    Hepburn certainly used the romanization system now named after him, but there's no evidence he actually devised it. There was an active Romanization Society at the time, and an alternative view is that it was the source of the romanization system which Hepburn adopted.

  6. Russ said,

    October 2, 2022 @ 10:15 pm

    It's interesting to hear you say that the proportion of romaji to kana has increased over time. I'm sure that's correct for things like signs and instructions inside Japan. But in a native text of any length, like a newspaper article or longer, you rarely, if ever, encounter romaji in my experience.

  7. Andreas Johansson said,

    October 3, 2022 @ 6:41 am

    Tangentially, is the division of work between hiragana and katakana stable? From an outsider's point of view, using just one set of kana looks like a low-hanging fruit for simplification.

  8. Chris Button said,

    October 3, 2022 @ 6:52 am

    Something that would have been useful for Japanese kana when noting the go-on and kan-on Sino-Japanese pronunciations coming in, would have been some kind of “virama” to allow for a coda without a following vowel. Or they could have just coined something like -n since there are multiple options for onsets to use in that way (na, ni, nu, etc)

    All those go-on -ti (-chi) endings—later switched to kan-on -tu (-tsu)—would have been far easier to interpret rom the outset had they just been clearly noted as -t with no vocalic component in learned pronunciations. Instead it seems to be missionary transcriptions that made that clear.

    It would also have allowed some clarity around how to interpret -ki/ku, for which we fortunately have the palatalized and labialized velar codas in Pulleyblank’s reconstruction of Middle Chinese (although things are not one-to-one) based on Mantaro Hashimoto’s palatal proposals, as well as how to interpret -pu and obsolete -mu.

  9. J.W. Brewer said,

    October 3, 2022 @ 9:18 am

    While obviously no romanization system can be perfect (or all things to all people), surely the sensible pragmatic solution is for everyone to stick with the system that I myself learned as a boy almost a half-century ago, viz. Hepburn. That said, the chart suggesting that the ratio of the system used for various place names varies considerably is interesting. Since Hepburn is standard for signage in railway stations telling you where you've stopped I wondered if Go[sho/syo]gawara might be such a small place as to lack a train station, but wikipedia says it has one.

  10. Dara Connolly said,

    October 3, 2022 @ 4:31 pm

    There is a range of variation in the representation of long vowels in Romanised Japanese. They can be unmarked or represented by a doubling, a "h", or a macron. Even in the quoted text in the article we see:
    "Oomoto-kyo (大本教, Ōmoto-kyō…" where the long o is represented in two different ways in the same word.

  11. Nori said,

    October 3, 2022 @ 11:44 pm

    I think Nihon-shiki is pointless. I don't see any reason to use it. If Nihon-shiki were consistent, "Nippon-shiki" would be "Ni-tsu-pon-shiki". Before anyone tells me that small つ is different, I'll point out that Japanese itself, when unable to write the smaller version uses the larger one. My name on my bank book has a large one inserted in the middle of my name because the printer doesn't print the smaller one. Also, there's no reason to mix upper and lower case letters in Nihon-shiki.

    @Andreas- Yes, but for some reason, Japanese react quite strongly to the wrong script being used. They have an intuitive sense for when katakana vs hiragana should be used. There are general rules but they are broken sometimes because "katakana feels right in this situation" or whatever. This is usually on TV.

    Personally, I can't see why anyone wouldn't want to just use Hepburn or a modified Hepburn. There are rules I learned (like, "before B, turn N into M" えんぶ –> embu) that make Japanese native speakers crazy, but in general, they can use it just fine, and it is much, mich better for foreigners.

  12. Chris Button said,

    October 4, 2022 @ 11:36 am

    Regarding づ /du/ and ず /zu/, if /du/ was already palatalized as /dzu/ before prenasalization was lost, then a phonetic distinction between ⁿdz and ⁿz would need to be reliably maintained, and that is a very tall order!

RSS feed for comments on this post