Japan's new kanji forms, 新字体 (shinjitai), are the same characters as the mainland's in about a third of cases. We compared the 302 pairs in Japan's 2010 list of everyday kanji where the old and new forms are separate Unicode characters, and the mainland's GB 2312 character set holds Japan's new form in 97; in the rest the mainland kept the old shape or has a third one. The old forms, 旧字体 (kyūjitai), are the Kangxi dictionary shapes Japanese print used before 1949. The 2010 list still prints one in brackets beside 362 of its 2,136 kanji.
The 1949 print forms copied shapes people already wrote by hand
A 1946 Cabinet notice, the 当用漢字表 (Tōyō kanji table), limited general use to 1,850 characters, and a 1949 table fixed their shapes. The Agency for Cultural Affairs' explanatory column says the new shapes were mainly simple forms people used when writing by hand.
The 1981 常用漢字表 (Jōyō kanji table) of 1,945 characters replaced that list and simplified some of its 95 additions to match. The 2010 list, proposed in the Council for Cultural Affairs' report, added 196 characters and removed five, and the report says it chose not to simplify the additions. The Cabinet notice of 30 November 2010 abolished the 1981 table.
Bracketed Kangxi forms shadow a sixth of the everyday list
Explanatory note 6 of the 2010 table calls its round-bracketed forms "so-called Kangxi dictionary forms" (いわゆる康熙字典体), added for reference and says forms with no marked difference were left out. The bracket count is what the table chose to print, a smaller set than every shape that changed.
In our count, 362 of the 2,136 kanji (16.9%) carry a bracketed form, with 364 old forms in all because 弁 stands for three: 辨, 瓣 and 辯. English Wikipedia's Kyūjitai article gives the same 364 pairs. The table's preface says it does not deny the kanji used in past writing, so it ranks neither form above the other.
Mainland standard forms mirror the Japanese new form in 97 pairs
Of the 364 old forms, 302 are separately encoded characters. We sorted them by what GB 2312 holds, using Unicode's Unihan simplified-variant field kept to GB 2312, or the old form itself where GB 2312 has it. The field alone names Japan's form in 86 pairs: it misses 12, such as 眞 and 真, where GB 2312 simply holds Japan's form, and counts 奬 to 奨 as a match although the mainland set has 奖.
| What GB 2312 holds | Pairs (of 302) | Share | Examples (old, new, mainland) |
|---|---|---|---|
| Japan's new form | 97 | 32% | 國 国, 會 会, 眞 真 |
| The old form | 47 | 16% | 乘 乗 乘; 佛 仏 佛; 卷 巻 卷 |
| A third form, or neither | 158 | 52% | 廣 広 广; 龜 亀 龟; 圓 円 圆 |
These count membership in a character set rather than usage, and the Unihan field is Unicode's data rather than the mainland's own standard list. Wikipedia's article on differences between shinjitai and simplified characters already sorts the school kanji by which side changed; our count covers the bracketed pairs instead.
Japan's new form looks like a traditional character set's entry in seventeen pairs
Japan's old forms are often assumed to be the traditional Chinese ones, yet in 17 of the 302 pairs Big5, the character set long used in Taiwan and Hong Kong, holds Japan's new form and lacks the old: 為 without 爲, 真 without 眞. Another 184 of the 302 new forms, 広 and 円 among them, sit outside both GB 2312 and Big5, though Unihan records them in wider Chinese sets. Big5 does hold the old form in 277 pairs, which explains the assumption.
Membership does not settle what anyone writes. Hong Kong's supplementary character set encodes 7 of the 17 old forms, and Hong Kong's Department of Health writes its own name 衞生署, with the old form, on its Chinese homepage.
Compatibility old forms pass for separate characters only before normalisation
The remaining 62 old forms are a known problem. Wikipedia's Kyūjitai article lists the 62 Jōyō kanji whose old forms "may not be displayed correctly" and says normalisation merges them with the new forms, and a 2022 Anki forums thread records a user whose exported deck had quietly changed. They are CJK compatibility ideographs, kept for round-trip with other standards: 55 in the U+FAxx range and 7 in U+F9xx, those 7 first encoded for a Korean standard.
Under NFC, the normalisation form many systems apply, each becomes a new form, and in our check all 62 land on exactly the character the table pairs them with. The old form of 侮 is 侮 U+FA30, identical in most fonts; in a browser console, "\uFA30".normalize("NFC") returns "\u4FAE". What we add: 52 of the 62 are on the names list, and the 2010 PDF's own text layer holds 66 compatibility code points.
Several new forms echo older characters with their own dictionary entries
By Unihan's Kangxi field, 82 of the 302 pairs have a new form with its own Kangxi dictionary entry. That is an upper bound on new forms that were already characters, since most are variants of one word, like 乱 and 亂. Some are not: 欠 and 缺, 芸 and 藝, 余 and 餘 have separate entries, and GB 2312 still encodes both forms as different characters in five pairs, among them 予 and 豫, and 缶 and 罐. The table gives 欠 one bracketed form without saying every 欠 was once 缺. Our guide to simplified and traditional conversion covers the same trap in Chinese.
Until 14 October 2026 our kyūjitai and shinjitai converter turned every new form with a single bracketed partner into that partner, so 欠伸 (a yawn) came out as 缺伸 and 芸香 as 藝香, and it marked the 62 compatibility outputs as converted though most fonts showed no change. That day's revision leaves 予, 余, 欠, 芸 and 缶 as written, shows the old form alongside as it did for 弁, and dot-underlines compatibility outputs. Its 364 pairs and its 204 name-list forms both matched our own derivations.
Registrable given names take after the old forms in 204 cases
Article 60 of the 戸籍法施行規則 (Family Register Act enforcement regulations, e-Gov text) sets what a given name in a Japanese family register (koseki) may use. Clause 1 allows Jōyō kanji, "limited to those outside the brackets" (括弧の外のものに限る). Clause 2 allows the 人名用漢字 (jinmeiyō kanji) names table, now 別表第三 (Appended Table 3), and clause 3 katakana and hiragana, excluding variant kana. The Ministry of Justice names page says kanji on either list can be used in a child's name.
So a bracketed old form is usable only if the names table lists it separately. In our count, 204 of the 364 bracketed old forms are on that table, 櫻 and 廣 among them, and the other 160 cannot be used in a newly registered given name under these rules. Wikipedia's 212 traditional characters on the names list match Unihan 17.0's names-list variants of Jōyō kanji; the 8 beyond our 204 are forms the 2010 table does not bracket. Our guide to Japan's 2020 school kanji table covers the names list, and whether a given name is accepted rests with the municipal office that receives the registration.
