🖋 Language notes

Why can’t you read the “ki” on a soba shop’s sign?

13 August 2026 ・ Runa ・ from primary sources at NINJAL and Unicode

The sign of an old soba shop, or the wrapping paper of a Japanese sweet. Sometimes the “ki” in “kisoba” has a shape you are not used to. That character you feel you ought to be able to read, and cannot. …That is not a cursive flourish, and not a design: it is a kana that used to be in ordinary use. Its name is hentaigana. Today I am writing about how that character got a seat in the electronic world. At the end, a place appears where the numbers do not agree.

✍️One sound had several characters

In the old hiragana, “a” was not just one character.

The hiragana we use now has one character per sound. “Ka” is only “ka”. But this became so fairly recently.

Hiragana were originally born by simplifying kanji, so depending on which kanji they came from, the same sound could have different shapes. For “ka”, there is one from 加, one from 可, one from 閑, and so on (you can look at this correspondence character by character in the list from the Center for Open Data in the Humanities). These characters — the ones no longer in use — are what we call hentaigana.

🏫1900: out of the classroom

At the turn of the century, hiragana became “one character per sound”.

That date is written in a paper by Takada Tomokazu and colleagues at the National Institute for Japanese Language and Linguistics.

Read on its own, this looks like the story that “hentaigana were abolished in 1900”. …But the same paper writes, immediately afterwards, this.

And hentaigana did not vanish completely either. The same paper writes that “even in the present day, hentaigana are sometimes used in the shop names and trademarks of Japanese-style establishments, and so continue to carry a part of Japan’s written culture” (translated from the Japanese). …The soba shop’s sign is one of those survivors.

📇Even so, there were people in difficulty

Out of the classroom, but still in people’s names.

According to NINJAL’s explanation, before 1947 hentaigana could be used in personal names. So names in hentaigana are alive even in today’s family registers and resident records. To exchange administrative information, it is a problem if those characters cannot be handled by a computer.

In fact, 168 hentaigana are included in the Ministry of Justice’s unified family-register characters. …So this was not a story about a hobby of old documents, but also about characters in use at government offices right now.

There was another need, on the scholarly side. People researching the history of Japanese characters and orthography, and people transcribing old documents, were in trouble over “characters that do not exist in the electronic world”. So NINJAL and others gathered those whose shapes have been stable since letterpress printing, and made a character set called hentaigana for academic information interchange.

🌐Getting a seat in the electronic world

Two international meetings, and the seat was settled.

NINJAL’s page keeps a record of that journey.

• 📍 October 2015, Matsue (ISO/IEC JTC1/SC2/WG2 meeting) … proposal for encoding from Japan. The encoding model was decided
• 📍 September 2016, San Jose (same) … it was decided to include 286 hentaigana in the ballot draft of the standard
• 📍 20 June 2017 … 286 hentaigana encoded in Unicode 10.0.0

They went into two blocks, Kana Supplement and Kana Extended-A. And these 286 characters are, reportedly, the union of the unified family-register characters and the hentaigana for academic information interchange. All 168 of the family-register characters are in, and 147 of those are shared with the academic side.

…So the needs of the government office and the needs of research flowed together into one table. The character on a soba shop’s sign was put to a vote in an international standard — thinking of it that way left me feeling slightly strange.

🔍286 characters, and 285

When I counted, one was missing.

Since it bothered me, I counted the character database that Unicode publishes, myself. That list with character codes and official names side by side.

Characters whose name contains “HENTAIGANA LETTER” —

One is missing. NINJAL writes 286. When I count, 285. …Is one of them wrong?

The one character of difference turned up at once. U+1B001. The official name of the character at this code point is this.

It is named as a hiragana, not as a hentaigana. That is why counting “HENTAIGANA LETTER” did not catch it.

Going to look at Unicode’s names list (NamesList), this character carried these notes.

% marks an alias and * marks a note. So this character is recorded as “having the alias HENTAIGANA LETTER E-1” and “derived from 6C5F”. 6C5F is the kanji 江.

And there was one more thing I found out. Looking into when this character entered Unicode —

As of 2010, this character already had a seat, as “the hiragana ye of an old period”. When the 286 hentaigana were added in 2017, one of them was judged to be the same character as the seat that already existed. So no new name was given; instead it was written in as an alias.

…And having got that far on my own, I opened the Unicode standard itself. Chapter 18, the description of East Asian scripts. The answer was written there, just so.

First, just before it, this is written.

“U+1B001 was originally encoded to represent a long-obsolete syllable — the sound that would have come between “yu” and “yo”. That syllable merged with “e”, and is now represented by U+3048, hiragana “e”.”
— The history of the sounds of Japanese itself, that “ye” became “e”, was written into a character-encoding standard.

“The hentaigana that would have been named HENTAIGANA LETTER E-1 has been unified with the existing U+1B001, and is aliased accordingly. … The 285 remaining characters in these blocks are the other hentaigana.”

The number 285 that I counted was written in the standard, just like that. That it is registered as an alias also remains in Unicode’s alias list (NameAliases), as 1B001;HENTAIGANA LETTER E-1;correction.

NINJAL’s “286” is the size of the set of hentaigana that were encoded. My “285” is the number of characters given the name HENTAIGANA. One character is in that set while keeping a hiragana’s name.

🌙In closing

What quietly surprised me most while looking into this was that people are the ones deciding “whether it is the same character”.

Both were born by simplifying the same 江, and one came down as the character for “e”, the other for “ye”. Whether those two are the same character — people judged that at an international meeting in the 2010s, and decided whether to make the names one or keep them two.

And that judgement was properly left in writing. The course of events and the decision are written in the body of the standard, it is registered in the alias list under the kind correction, and the names list notes “derived from 江”. …The decision “we have decided to treat these as the same character” is placed where it can be read. That, to me, looked the most honest of all.

One last thing. The same standard also writes where hentaigana are used now. Japan’s family registers, and — business signage.

The “ki” on the soba shop’s sign is still hard to read. But that character has a number and a name. You can type it, and you can search for it. …A character that became unreadable was given a place to belong. That, I thought, is rather a good thing.

📚Sources

• NINJAL, “Hentaigana for academic information interchange” — the purpose of the character set, naming before 1947, 168 family-register characters (147 shared with the academic set; the two as a union), the Matsue and San Jose meetings, 286 encoded in Unicode 10.0.0 (20 June 2017), the blocks
• Takada Tomokazu, Yada Tsutomu, Saito Tatsuya, “Hentaigana: past and future”, Johokanri 58(6) (2015) — the 1900 Enforcement Regulations, the observation that it followed the decline in letterpress printing, survival in shop names and trademarks, naming before 1947
• Unicode, “UnicodeData.txt” — the number of characters named “HENTAIGANA LETTER” (I counted them myself), and the official name of U+1B001
• Unicode, “NamesList.txt” — the alias “HENTAIGANA LETTER E-1” for U+1B001 and the note “derived from 6C5F (江)”
• Unicode, “DerivedAge.txt” — U+1B001 is Unicode 6.0; the hentaigana (U+1B002–U+1B11E) were added in 10.0
• The Unicode Standard, chapter 18 “East Asia” — “HENTAIGANA LETTER E-1 has been unified with U+1B001”, “the 285 remaining characters”, and that hentaigana are still used in family registers and business signage
• Unicode, “NameAliases.txt” — 1B001;HENTAIGANA LETTER E-1;correction
• CODH, “List of Unicode hentaigana” — a list with source kanji, and examples of cursive forms
• MEXT, “Enforcement Regulations for the Elementary School Order” (extract) — what is posted is an extract and does not include the table of kana forms