Major romanization change coming in Japan

« previous post | next post »

From Pinyin News (3/4/24):  

"Japan to switch official romanization from Kunrei-shiki to Hepburn"

Japanese newspapers are reporting that Japan will officially switch from Kunrei-shiki romanization to Hepburn romanization.

In a front-page column last week, the Asahi Shimbun said, “A draft report recently published by the Council of Cultural Affairs pointed out that the Hepburn system is more widely used than the Kunrei system, and it is expected that the notation will be adjusted to reflect this. It is surprising because the writing system has not changed for about 70 years, but if confusion can be avoided, the change is to be welcomed.”

Some examples of differences:

Kunrei Hepburn
Aiti Aichi
Atugi Atsugi
Gihu Gifu
Hukusima Fukushima
Sinzyuku Shinjuku
Titibu Chichibu
Tukizi Tsukiji


This will make standard what most people have been doing for decades.

Selected readings


  1. Jamie said,

    March 4, 2024 @ 8:44 am

    About **** time

  2. Neil Kubler said,

    March 4, 2024 @ 9:28 am

    Actually, from a linguistic and pedagogical point of view, this is a regrettable change that unnecessarily complicates the description and learning of Japanese grammar and is not in accord with the "psychological reality" of the Japanese sound system for native speakers (also not in accord with the Japanese writing system). Examples: In Hepburn, the stem of the verb "wait" is written with "ts" in MATSU "to wait", with "t" in MATTA "waited" (informal), but with "ch" in MACHIMASHITA "waited" (formal). So learners must remember mats-/mat-/mach-! But in Kunree-siki, the stem of the verb is the same for all ("t"), since these are written MATU, MATTA, and MATIMASITA. Makes the grammatical description a whole simpler!

  3. Peter Grubtal said,

    March 4, 2024 @ 10:12 am

    @Neil Kubler
    Your example sets out the issue very nicely. But making "the grammatical description a whole (lot?) simpler", begs a number of questions. One of them being "cui bono?": the Japanese themselves? academic linguists? language learners/teachers?, or transcription into English for non-Japanese speakers?
    If it's the latter, surely an orthography based on English phonology is better, i.e. Hepburn.

    We have the dreadful example of pinyin which ignores the latter criterion (and without diacritics doesn't even represent Chinese pronunciation correctly), and has been responsible for making Chinese culture even more inaccessible to everyone, apart from a coterie of academics.

  4. Jim Unger said,

    March 4, 2024 @ 10:30 am

    Neil makes good points, but, in my opinion, when it comes to official Japanese positions on romanization, it is more important that there be a single national standard than that it be the linguistically better one. Right now, the ISO standard is, more or less, Kunrei-siki, but it languishes because Hepburn is, for instance, preferred by the U.S. Library of Congress; the resulting confusion encourages some people to make up spellings that belong to no system at all, such as .

    What Japan needs is a single standard that it teaches diligently to school-children and makes legally valid in all contexts alongside traditional orthography. It was early school education that enabled Japan to switch quickly to the metric system after the Occupation (MacArthur had ordered Japan to use American units), and since abolition of the present writing system is impractical (and politically impossible), the best way to remove the obstacles that it poses for the increasingly large number of non-Japanese workers who support the domestic economy is to establish genuine digraphia legally. (I made these points in two lectures in Japan back in 2013.)

    In my opinion, the best national standard would be the original Nippon-siki system (as in seen in some pre-war physics textbooks at Todai), but politicians are as unlikely to adopt it as to do away with traditional script. So if they want Hepburn, that's fine with me: the question is, will they then take the next two crucial steps or continue to pretend that Japan has no writing system problem?

  5. Dwight Williams said,

    March 4, 2024 @ 10:45 am

    So I've grown up with Hepburn-style Romanization all this time, hm?

  6. Peter Grubtal said,

    March 4, 2024 @ 10:58 am

    @Neil Kubler
    And of course no system is perfect, and in fact "Matimasita" doesn't represent the pronunciation correctly, because of the devoiced vowel: but it's easier it do it in Hepburn as "machimash'ta", whereas "matimas'ta" looks odd or misleading.

  7. KeithB said,

    March 4, 2024 @ 11:12 am

    It must be nice having a language Pope that can actually decide things like this.

  8. eli said,

    March 4, 2024 @ 11:13 am

    Hepburn makes most sense from an L1 English background, but I really do like kunrei-shiki and think it has a lot of benefits, and I'll miss it… though I think it'll take a while for everything to get updated, much less for how people actually conceptualize romanizing the language to change, which will take at least a generation.

    Eleanor Harz Jorden describes it best in her 1987 book:

    "Various systems of romanization—representation of the Japanese language by letters of the Roman alphabet—are in use in Japan today. The system used in this book is an adaptation of the Shin-kunrei-shiki 'New Official System' and will be designated as JSL Romanization. Other common romanizations are Hepburn (Heboň-shiki, also called Hyōjun-shiki 'Standard System') and Nippon-shiki 'Japanese System.' The differences among them are slight and can be learned with little difficulty. For example, the word for 'romanization' is variously represented as follows:

    JSL: roomazi
    Shin-kunrei-shiki: rōmazi
    Hepburn: rōmaji
    Nippon-shiki: rōmadi

    Hepburn romanization is the system most familiar to Westerners; but there are three cogent reasons for not using it in a Japanese textbook.
    1. JSL, Shin-kunrei-shiki, and Nippon-shiki bear a direct relation to Japanese structure, whereas Hepburn has no such connection. Thus, in describing Japanese inflection, many statements become unnecessarily complicated and parallelism is obscured if Hepburn romanization is used. For example, compare the following:

    To form the stem of consonant verbals:
    Using JSL, Shin-kunrei-shiki, or Nippon-shiki: change final -u to -i
    Using Hepburn romanization: change final -u to i, but if -u is preceded by ts, change the ts to ch, and if -u is preceded by s, add h after the s.

    The complexity of the second statement results not from special cases in Japanese verbal structure, but only from the fact that Hepburn romanization is based on languages of the West (its vowels have values roughly as in the Romance languages, its consonants as in English) rather than on the Japanese language."

  9. Michèle Sharik Pituley said,

    March 4, 2024 @ 11:26 am

    I’ve been to Gifu, Fukushima, & Shinjuku, & IIRC, that is how the romaji is spelled on signs there.

    I do remember seeing a sign for Paruru Hall (in Chiba) spelled as Palulu Hall, which, for English speakers at least, is closer to the actual sound. (Who decided to use an R, anyway?)

  10. eli said,

    March 4, 2024 @ 11:40 am

    @Michèle It depends on what signage you're talking about, e.g.

    "In Japan itself, there are some variants officially mandated for various uses: Railway Standard (鉄道掲示基準規程, Tetsudō Keiji Kijun Kitei) […] mostly follows Modified Hepburn, except syllabic n is rendered as in Traditional. Japan Railways and other major railways use it for station names."

    But all of my (Japanese) coworkers use kunrei-shiki when typing on their computers.

  11. Coby said,

    March 4, 2024 @ 12:24 pm

    Are there still topolects in which (Hepburn) "chi" is pronounced as [ti]? I remember seeing (a long time ago) a Japanese movie or TV show in which an older man pronounced 友達 as tomodati.

  12. Ebenezer Scrooge said,

    March 4, 2024 @ 12:53 pm

    There is a very small corner of US law that relies on accurate Romanization: the name of foreign companies who want to file a financing statement. If the name isn't completely accurate, the financing statement is ineffective. (This is a courtesy to those who subsequently search the financing statement–they shouldn't be expected to anticipate every possible spelling error.)

    Fortunately, the Japanese have seemed to anticipate this problem–Japanese corporate charters contain the company's name in Romaji. This name is easy for a non-Japanese reader to pick out of the charter, since it is the only Romaji in the charter.

  13. Michèle Sharik Pituley said,

    March 4, 2024 @ 12:54 pm

    @eli – the signage that is on the station maps, directional signs, menus, and other places around the cities, etc.

    I was there 8 times for concert tours. I mostly saw signs in train stations, concert halls, restaurants, hotels, street signs, markets/malls, etc.

    When your quote says “syllabic n is rendered as in Traditional” does that mean with an m, as in “Aka Tombo”?

  14. Jonathan Smith said,

    March 4, 2024 @ 12:59 pm

    fully endorse Jim Unger's standard >>> linguistically superior (whatever that means), "[w]hat Japan needs is a single standard that it teaches diligently to school-children and makes legally valid," etc. — here's hoping that at some point this message is heard in Korea, Taiwan, Thailand, etc…

  15. David Marjanović said,

    March 4, 2024 @ 2:10 pm

    We have the dreadful example of pinyin which ignores the latter criterion (and without diacritics doesn't even represent Chinese pronunciation correctly), and has been responsible for making Chinese culture even more inaccessible to everyone, apart from a coterie of academics.

    Even more inaccessible than what – surely not Wade/Giles, which is worse in numerous respects?

    Concerning diacritics, I've never (for what little that's worth) seen a romanization of Japanese that spelled the so-called pitch accent out. It's not as big a deal as tone in Mandarin, but there are minimal pairs nonetheless.

  16. J.W. Brewer said,

    March 4, 2024 @ 2:38 pm

    David M.: the so-called pitch accent isn't spelled out (or other explicitly denoted) in kana, is it? I think of romaji (in any system) as providing a predictable and reversible (okay for that you may need to remember the macrons in your romaji input …) system for transliterating back and forth between kana and roman glyphs. Representing actual Japanese pronunciation more accurately than kana does isn't necessarily one of the goals, and the Japanese already agree among themselves how to spell any of their lexemes (including those conventionally written with kanji) in kana.

  17. Chris Button said,

    March 4, 2024 @ 2:58 pm

    Matsu … matta
    Asobu … asonda

    Kunrei can't put the inflectionary logic back into the latter any better than Hepurn in the former.

  18. Jim Breen said,

    March 4, 2024 @ 5:05 pm

    Re: "Japan to switch official romanization from Kunrei-shiki to Hepburn"

    I find it slightly amusing that the Pinyin News used ヘボン式 (Hepburn shiki) to transcribe 訓令式 (kunrei-siki).

  19. eli said,

    March 4, 2024 @ 5:59 pm

    @Michèle Yep, e.g. Namba, Nippombashi, Shimbashi, and so on. This is largely true for governmental signage across Japan, at least; individuals can (and do) of course do it however they please, with inconsistencies, which creates some fun variation!

  20. Michèle Sharik Pituley said,

    March 4, 2024 @ 6:46 pm

    @JW: “Representing actual Japanese pronunciation more accurately than kana does isn't necessarily one of the goals”

    When I was taking Japanese language classes (at Soko Gakuin), we were told it was easy to learn to pronounce Japanese because the romaji represented the sounds. Just sound out the romaji. Easy peasy. Learning the the Hiragana “spelling variations” (such as “ei”, “ou”, little “tsu”, etc) was no big deal because being English speakers/readers/writers, we’re used to weird spelling variations.

    So it seems to me that using a romaji system which doesn’t accurately (as possible) represent the pronunciation makes it harder. YMMV.

  21. I-C said,

    March 4, 2024 @ 8:03 pm

    I do hope, however, that IMEs continue to support both for a long time. I completely see the sense behind this decision given Hepburn's popularity, but Kunrei is a lot more natural to type on a qwerty keyboard.

  22. Jonathan Smith said,

    March 4, 2024 @ 8:43 pm

    To J.W. Brewer's thought about Romanization as transcription, I think it's incidental — i.e., the real desideratum is a single orthodox means of representing a given Japanese (or other)- language message in Roman script which will exist in parallel to the (ideally basically) single, orthodox native script representation of the same message. Of course it had better be possible to recover something pretty damn close to the intended message from either visual representation taken independently, but the two need not be straightforwardly mappable one to the other since they might involve different choices about level of representation (phonetic? phonological? morphemic?) and the like. Plus unavoidable further wrinkles since in reality, there will be (hopefully very minor) points of uncertainty/ambiguity in recovering said message from either script independently which need not be the same across the systems.

    But of course it should still be trivial for a human brain (or better AI) to convert one system to the other. E.g., the oft-heard claim that converting Chinese written in characters into Chinese written in pinyin is a computationally hard problem, and not just one that nobody wants/cares to solve, is a good chuckle.

  23. phspaelti said,

    March 4, 2024 @ 9:09 pm

    @eli: Correct for Nihon-shiki is rōmazi not rõmadi

    I am sceptical that "all" Japanese anywhere use kunreishik. The vast majority of Japanese freely mix Hepburn and Kunrei, even in the same word. So many use "si", "ti", and "tu", but "hu", "sya" and "zi" are decidely less popular.

    One point missing from all of this discussion is that Kunrei-shiki (and Nihon-shiki) have no real way of dealing with the difference between チ and ティ, as well as ツァ, ツォ, ファ, etc.

  24. Michael Victor Ryan said,

    March 4, 2024 @ 9:23 pm

    @Jim Unger
    Thirty-two year resident of Japan here.
    There is already something that is "legally valid" in all contexts. It is Katakana.
    All legal residents (both foreign born and native) can obtain the admittedly strangely named a "My Number Card" which is akin to a Social Security card in the US. On it is one's name in both Roman letters and Katakana. This card is used for official purposes in Japan, e.g. entering into contracts, identification at the bank. Actually I believe on contracts between two parties in Japan Roman letters are not allowed.
    There is no "problem" for people who immigrate here (legally of course!). They need to go on down to their ward office and get a My Number Card with their name in both Roman Letters and Katakana. (Although most Japanese will ignore the Roman letters.)

  25. Chris Button said,

    March 4, 2024 @ 10:35 pm

    A better prosodic comparison to the marking of tones in pinyin is the marking of vowel length in Japanese.

  26. J.W. Brewer said,

    March 5, 2024 @ 12:30 am

    Michèle Sharik Pituley: I guess what I was trying to say was more or less "as accurately as kana represents pronunciation, but no more so." I would agree that's an apparent advantage of Hepburn, but since Hepburn is how I myself was first taught to transliterate kana (as a 3d grader at the American School in Japan a half-century ago) I am not well positioned to distinguish between "Hepburn is good because it's objectively better" versus "Hepburn is good because it's what I'm most familiar and therefore comfortable with."

  27. J.W. Brewer said,

    March 5, 2024 @ 12:38 am

    @Jonathan Smith: But one of the particular quirks of the Japanese situation is the multiplicity of "native" scripts. For lexemes conventionally written in kanji there are conventional ways of representing them in kana and it makes sense for any romaji system to as it were transliterate the kana versions of the kanji words on the same basis as it transliterates words conventionally written in kana rather than via some different approach. Of course, many kanji are associated with multiple kana transliterations, but that's because there are multiple pairs (or larger sets) of Japanese words that are homographs in kanji but not in kana. (It's traditional to speak of the same kanji character having different "readings" but that's a bass-ackwards conceptualization of the situation.)

  28. Alyssa said,

    March 5, 2024 @ 10:53 am

    "Using Hepburn romanization: change final -u to i, but if -u is preceded by ts, change the ts to ch, and if -u is preceded by s, add h after the s."

    This is only a problem for Hepburn if you assume that learners will never need to deal with the spoken language – if they want to say words out loud, they'll have to learn these rules either way.

  29. Michèle Sharik Pituley said,

    March 5, 2024 @ 11:45 am

    @JW ‘I am not well positioned to distinguish between "Hepburn is good because it's objectively better" versus "Hepburn is good because it's what I'm most familiar and therefore comfortable with."’

    That’s fair & might be the case with me, too.

  30. Carl Masthay said,

    March 5, 2024 @ 6:15 pm

    Hooray, I always thought that the phonemic kunrei-shiki was an abomination causing stumbling mispronunciations in many linguistic articles and sources. It may have been perfectly logical, but it should never have been promoted. One example is an entire thick dictionary in Ainu (by Suzuko Tamura) using spellings such as sik 'the eyes', actually /shik/ and caca 'old man', /chacha/. So, OK, one can get used to it. Dr. Hideki Narumi (now deceased) wrote such books involving Ainu. I was sad that he used the si spellings, and I wrote him so. So, hey, happy times with Hepburn!

  31. Chris Button said,

    March 6, 2024 @ 9:25 am

    @ Carl Masthay

    Isn't the "c" in "caca" representing the phoneme /ts/ though? While "ts" might be more familiar for an English speaker, that's hardly a controversial representation and not the same as "ts" being an allophone of /t/ before "u" in Japanese.

    (To be clear, I'm no Ainu language expert, but have done the old trick of googling the topic to then feign knowledge about it on a LLog post).

  32. AG said,

    March 7, 2024 @ 12:08 am

    I lived right next door to Hepburn's former residence in Yokohama for 5 years:

    … down the street was a grocer called "MEIDI-YA", which to my horror turns out to be the Nihon-shiki spelling of "Meiji":

  33. Michael Victor Ryan said,

    March 7, 2024 @ 9:04 pm

    @AG: Actually, the Japanese wiki page says that normally "Meiji" would be the spelling but the founder of the business felt that "Ji" would have been hard to foreigners to say and so changed it to "di" which he felt would be easier for them to pronounce. I don't think じ would ever normally be rendered "di".

  34. Philip Taylor said,

    March 8, 2024 @ 12:39 pm

    I am puzzled by the idea thay "Ji" might be had for foreigners to say — OK, a Briton would pronounce it one way (/dʒi/), a German another (/yi/) [I am uncertain about the vowel, so perhaps /yɪ/], but neither would have any problem pronouncing it. Of course, the majority of foreigners are neither British nor German, but I do wonder in what language(s) "Ji" would be problematic.

  35. Jim Unger said,

    March 11, 2024 @ 3:34 pm

    @Michael Victor Ryan
    The example you give is interesting, but I was thinking of the nurses from overseas trained in Japan who couldn't stay to do the work they were trained for because they couldn't pass the certifying exam, which was in kanji kanamajiribun. (I recall reading the story in the news about ten years ago.) The exam should have been available in romaji. The government underwrote their training in part in improve its image in Southeast Asia but also because an aging Japan needs more nurses than the domestic market can provide: it is absurd to train these people and then deny them certification just because they can't handle Japanese in traditional script. If they are willing to live and work in Japan, they should be given the opportunity to show how far their spoken language proficiency has taken them, not flunked because they haven't spent years learning kanji.

  36. Michael Victor Ryan said,

    March 11, 2024 @ 8:50 pm

    @ Jim Unger:
    Sir, thank you for responding. What you refer to as "traditional script" is used by everyone here on a daily basis at all socioeconomic levels. And that includes hospitals. It would be inconceivable to employ nurses who could not read regular Japanese as it would be employing nurses in the US who could not read English or nurses in France who could not read French. It would even be life threatening in a medical context. The real problem is that Japan over the years has done a poor job of promoting its language (and its admittedly cumbersome writing system) abroad, a la Goethe Institute and the British Council.
    Some make an argument for ditching Kanji for Romaji or just Hiragana. But such ideas are make-believe, just like the seemingly easier task of simplifying irregular English spelling is. Can you imagine actually converting English spelling to a more regular system and all that would entail at all levels of society? For the Japanese, jettisoning Kanji would be just as unrealistic.

RSS feed for comments on this post