How to write "write" in Chinese and Japanese

« previous post | next post »

The word for "write" in Modern Standard Mandarin (MSM) is xiě.   The traditional form of the Sinograph used to write this word is 寫, var. 冩 (can you see the difference?).  In Japanese that would be pronounced "sha" or "utsusu", but it is considered an uncommon character (hyōgaiji), and means not "write", but "transcribe; duplicate; reproduce; imitate; trace; describe​; to film; to picture; to photograph".

There are a number of words for "write" in modern Japanese (e.g., arawasu 著す, shirusu 記す), but the most common is kaku 書く.  Yes, that kanji means "book" in MSM, but it meant "write" in early Sinitic, whereas 寫 means "write" in MSM but meant "to place; to displace; to relocate; to carry; to relay; to express; to pour out [one's heart, troubles, etc.]; to copy; to transcribe; to follow; to describe; to depict; to draft; to create quotations; to draw; to sketch; to make a portrait; to sign; to formalize" in Literary Sinitic (LS) and Classical Sinitic (CS)   This is a good example of how Japanese often tends to retain older meanings of characters in the modern language, whereas in MSM characters have a propensity to take on new and quite different, unexpected meanings (e.g., zǒu 走 ["walk" in MSM] meant "run" in LS and CS).

The simplified form of 寫 in Chinese is and in Japanese is 写 (see here).  The problem is that, in many browsers, word processing systems, online applications, etc., it is hard to produce the former, and only the latter will come up, even for Chinese.

It's the same sort of problem we encountered with the simplified forms of mén 門 ("door; gate"), viz., 门 (Chinese) and 门 (Japanese).  Often, when I want one, I get the other.  That is especially the case when I'm trying to get the Chinese simplified form 门, but the Japanese simplified form 门 keeps intruding.  For a prime example of how these two forms can get all fouled up, see "Foul Meat-gate" (7/29/14).

People who do not have to deal with hanzi and kanji on a daily basis probably wouldn't even notice the difference between and 写, between 门 and 门, but for those who are trying to write the script of one country rather than that of another country, it can be very frustrating.

Now for a little quiz, only for those who have NOT studied Sinograms / Chinese characters (hanzi, kanji, hanja), and do not look up the answer somewhere:  how many strokes does it take to write 寫?  Please feel free to submit your count until we have two or three people giving the same amount.

[h.t. John Rohsenow]


  1. Ellen K. said,

    June 15, 2019 @ 12:36 pm


  2. Frans said,

    June 15, 2019 @ 12:47 pm

    I think I see:

    2-4 at the top
    7 in the middle
    7 at the bottom

    I'll go with 17.

  3. Victor Mair said,

    June 15, 2019 @ 12:49 pm

    Thank you, Ellen K. and Frans.

    Keep 'em comin'!

  4. Brian said,

    June 15, 2019 @ 1:02 pm

    I'll say 12.

  5. F said,

    June 15, 2019 @ 1:11 pm

    I may not count because I've studied Japanese (but am not familiar with this character nor all of its components) but my guess is 13.

  6. F said,

    June 15, 2019 @ 1:13 pm

    Wait, 14 — I can't add.

  7. Kyle B. said,

    June 15, 2019 @ 1:19 pm

    As with "F said", I've studied Japanese but am not familiar with 寫. I'd guess 14 strokes, as well. I'd guess 3 for the top, 5 for the middle, and 6 for the bottom.

  8. Jeremy said,

    June 15, 2019 @ 2:07 pm

    I think I can do it in 11.

  9. Victor Mair said,

    June 15, 2019 @ 2:26 pm

    Thank you very much, Jeremy.

  10. David Marjanović said,

    June 15, 2019 @ 2:32 pm

    I think I can do it in 11.

    The trick is that there are pretty strict rules about what counts as one stroke, and in which order and which directions the strokes of each character are to be written by hand. (Some of these rules are slightly different between China and Japan, and there's a very small number of characters whose stroke order is controversial or unclear in China – not sure about the stroke count.)

    The point of this is to keep the characters reproducible. Handwriting which merges several strokes into one is often hard to read, but not as hard as it could be, because the fixed order and directions don't leave many options for which strokes can merge.

  11. Joke Kalisvaart said,

    June 15, 2019 @ 2:46 pm


  12. David Marjanović said,

    June 15, 2019 @ 3:28 pm

    I forgot to mention that "run" turning into "walk" has also happened in something like the northern half of Germany (to laufen… the cognate of leap).

  13. PeterL said,

    June 15, 2019 @ 3:52 pm

    15 (14 for the variant)

  14. Jamie said,

    June 15, 2019 @ 4:19 pm

    I have also studied Japanese. I am not familiar with this character but I am going to say 14, too.

  15. Not a naive speaker said,

    June 15, 2019 @ 4:49 pm

    Did a copy and paste and serious enlargment of the glyph

    Looked at the "thick" ends

    counted 12 or 13

    Grabbed the Kanji Fast Finder; is not one of the 1945

  16. Thomas Rees said,

    June 15, 2019 @ 5:10 pm

    Almost the same as Kyle B., but I think the middle part needs 6 strokes: 3 for the F-shaped left part, 2 for the right part, and one for the long horizontal base. The upper-case gamma Γ shape needs two strokes; the reversed (⅂) only one. East Asian right-handed bias! 3 + 6 + 6 = 15

  17. Jeremy Fagan said,

    June 15, 2019 @ 5:14 pm

    I think 19, but I may be over-egging the pudding..

  18. Stephen Hart said,

    June 15, 2019 @ 5:50 pm

    I also count 14 strokes, assuming (perhaps ignorantly) that you can only change direction once in a single stroke.

    And like Not a naive speaker, I also had an easy time seeing the difference in the "hat" in 寫, once I set it to 144 point type. Then, once I knew what to look for, it's also evident at 12.5.

  19. Hamletmonkey said,

    June 15, 2019 @ 5:59 pm

    From another student of Japanese: 15 strokes.

  20. Scott P. said,

    June 15, 2019 @ 6:06 pm

    My guess is 12.

  21. Barclay said,

    June 15, 2019 @ 6:15 pm


  22. Allen Thrasher said,

    June 15, 2019 @ 6:21 pm

    This Indologist says 13.

  23. Jim said,

    June 15, 2019 @ 6:35 pm

    Studied Japanese 20 yrs ago. I think it's 15 strokes.

  24. Victor Mair said,

    June 15, 2019 @ 6:48 pm

    Thank you, Jeremy Fagan, Scott P., Barclay, and Allen Thrasher.

  25. Julian said,

    June 15, 2019 @ 6:57 pm

    3 top + ?6? middle + 6 bottom = 15?
    I admit that I've played with characters enough over the years to recognise about four of them and to know (IIRC) that some lines with corners count as a single stroke.

  26. Francis Bond said,

    June 15, 2019 @ 8:33 pm

    FWIW, the Japanese character is not reproducing correctly in my browser. I think they are unified code points, so you can only see the difference if you specify the fonts.

  27. JorgeHoracio said,

    June 15, 2019 @ 8:53 pm

    In my absolute ignorance, I venture 18

  28. PeterL said,

    June 15, 2019 @ 9:47 pm

    臼 in the middle seems to be 6 strokes:
    So, 3 for the cap (2 for the variant), 6 for the middle, 6 for the bottom = 15 (14 for the variant)
    (Following the Japanese rules for stroke counting, as I remember them)

  29. Victor Mair said,

    June 15, 2019 @ 11:03 pm

    Thank you, JorgeHoracio.

  30. unekdoud said,

    June 16, 2019 @ 12:43 am

    @Francis Bond,

    For each pair of 写 and 门, the Chinese one is marked up with the HTML lang attribute (lang="zh-CN") and the Japanese one is not. For me, the unmarked one displays identical to Japanese in Firefox and to Chinese in Chrome, which clearly selects fonts differently! (I believe it's also due to my OS settings.)

    While it's best to actually specify the font if you're concerned about things like stroke count and stroke crossings, the language tag seems to be doing rather well at concisely solving the 门 problem.

  31. DBMG said,

    June 16, 2019 @ 1:41 am

    I'd write that in 11 or 12 strokes (top 2, middle 3-4, bottom 6) but I bet it's in flagrant violation of that brush-handling protocol I'm vaguely aware of.

  32. raempftl said,

    June 16, 2019 @ 2:26 am

    Japanese learner: 15

  33. DEMAY said,

    June 16, 2019 @ 2:52 am

    wenlin /
    stroke 寫 (15 strokes)

  34. Vilinthril said,

    June 16, 2019 @ 4:14 am

    I also studied Japanese for a year ages ago; I think it should be 15, maybe 14.

  35. VanT said,

    June 16, 2019 @ 4:43 am

    Japanese beginner here (but I have never seen this character before this article): I count 15.

    I think when you haven't had contact with these characters before, the "tricky" parts could be recognising that 臼 part has 6 strokes and the 宀 part has 3 strokes, the bottom part should be straightforward either way.

  36. Victor Mair said,

    June 16, 2019 @ 5:45 am

    Thanks to everyone who participated in this little survey. We have a range from 11 to 19, and everything in between, but curiously not 16. Those who had studied kanji before, but did not necessarily know this particular character, clustered around 13-15 strokes.

    All right, time to reveal that the "correct" number of strokes for 寫 is 15. If you scroll down through this zdic entry, first of all in the top left you'll see a large rendition of the character in a box with the strokes written very clearly. The trickiest part is the beginning of the middle part of the character just beneath the roof, such that even people who have studied kanji before but don't know this one can't decide whether the top left corner of that section is one stroke or two: it should be two, unlike the top right corner of the middle section across the way, which looks similar but is one curving stroke.

    **You can see the strokes written out in sequence, brush-like, here.**

    For a truly dramatic rendition of the third tone of xiĕ, press the speaker button.

  37. Philip Taylor said,

    June 16, 2019 @ 6:40 am

    I got the stroke count right (viewed at 72pt made it reasonably easy) but am intrigued by the "truly dramatic rendition of the third tone of xiĕ". The speaker has what my three teachers of MSM would have called "a modern, affected, pronunciation, typically used by young women" in which the "xi" group is sounded more like /tsj/ than /t͡ɕ/, the latter being what they taught us was the more traditional, "correct", pronunciation.

  38. Victor Mair said,

    June 16, 2019 @ 9:28 am

    "affected… pronunciation"

  39. Keith said,

    June 16, 2019 @ 3:17 pm

    I get 寫 in 12 strokes.

  40. Jim Breen said,

    June 16, 2019 @ 9:15 pm

    I held off, because I knew the official (Japanese) stroke-count for 寫 was 15, as can be seen at:

    The issues around which glyphs you see for hanzi/kanji such as 写 are non-trivial, as they have identical code-points. The visual representation relies on the display system, e.g. an app or WWW browser, having access to meta-data such as the language, and also a repertoire of either separate font files for the languages or unified font files with embedded language-specific information. It all gets worse when one tries to display the Chinese and Japanese glyphs in the one document.

    Surprisingly this is an area which Google managed to stuff up in its Android systems for many years. Phones in Japan had Japanese fonts and phones elsewhere had Chinese fonts, and it was very difficult to get around this. It took a long time to get the people working on Android to recognize the problem, and several years and version releases to get it fixed properly.

  41. John Swindle said,

    June 16, 2019 @ 10:37 pm

    I was surprised when I learned that while you could write with a typewriter in English (the name even includes "writer"), in Chinese you could only type with it (打字 da3zi4, literally "strike characters"). To use a typewriter to write (写字 xie3zi4, literally "write characters"), I was told, would entail dipping it in ink and maneuvering it above the paper. Is the same true for computers? I suppose so.

  42. PeterL said,

    June 17, 2019 @ 2:17 am

  43. Victor Mair said,

    June 17, 2019 @ 5:52 am

    In the section on character variants by country cited by PeterL from Wikipedia (see below for relevant quotation) in the previous comment, the discussion of 內 vs. 内 for the simple (4 strokes) phono-semantic Sinograph used to write the essential morpheme "nèi" ("inner; internal; inner") has particular poignancy for me, because I have always felt that the former form, with rù 入 ("enter") at the top center rather than rén 人 ("person; people"), both phonologically and semantically, made better sense (see this Wiktionary section on the glyph origin, etymology, and pronunciation of 內/内).

    And that is what some of my teachers taught me more than half a century ago when I first learned this character, though other teachers differed and told me that I must write 内, not 內, and that is current dogma on the character used to write morpheme "nèi" ("inner; internal; inner"). Yet it vexes me every time I have to write this character because it just doesn't seem right.

    As if learning thousands of characters weren't hard enough, those who wish to become literate in the Sinographic script also have to contend with countless maddening details of this sort. All in all, persistence in learning how to write the Chinese script is a task requiring enormous intestinal fortitude.


    In the twentieth century, East Asian countries made their own respective encoding standards. Within each standard, there coexisted variants with distinct code points, hence the distinct code points in Unicode for certain sets of variants. Taking Simplified Chinese as an example, the two character variants of 內 (U+5167) and 内 (U+5185) differ in exactly the same way as do the Korean and non-Korean variants of 全 (U+5168). Each respective variant of the first character has either 入 (U+5165) or 人 (U+4EBA). Each respective variant of the second character has either 入 (U+5165) or 人 (U+4EBA). Both variants of the first character got their own distinct code points. However, the two variants of the second character had to share the same code point.

    The justification Unicode gives is that the national standards body in the PRC made distinct code points for the two variations of the first character 內/内, whereas Korea never made separate code points for the different variants of 全. There is a reason for this that has nothing to do with how the domestic bodies view the characters themselves. China went through a process in the twentieth century that changed (if not simplified) several characters. During this transition, there was a need to be able to encode both variants within the same document. Korean has always used the variant of 全 with the 入 (U+5165) radical on top. Therefore, it had no reason to encode both variants. Korean language documents made in the twentieth century had little reason to represent both versions in the same document.

  44. Adrian Bailey said,

    June 17, 2019 @ 7:10 am

    I'd say 10 strokes, if a stroke ends when you take the pen off the paper. I'm also marvelling at the kaku character, which has 8 horizontal lines. Hard to imagine how either of these characters can always be legible.

  45. Antonio L. Banderas said,

    June 17, 2019 @ 9:59 am

    "China […] changed (if not simplified) several characters."

    This is what it has always seem to me, unfortunately.

  46. Chris C. said,

    June 17, 2019 @ 5:30 pm

    Looks like 14 strokes to me.

  47. PeterL said,

    June 17, 2019 @ 6:02 pm

    書 doesn't always mean "write" in Japanese. For example 書店 or "bookstore". Or (this example taken from wwwjdic):
    "When it comes to Chinese books that are overvalued worldwide I suppose it's Sun Tzu, isn't it?"

    Nelson has 寫 as entry 1338/F556 (3 strokes for the radical + 12) but doesn't have 冩.

  48. J. Goard said,

    June 18, 2019 @ 2:08 am

    My Hanja awareness is probably very good for a non-native Korean speaker, but doesn't extend much to proper writing. Looks like 13 to me.

  49. Victor Mair said,

    June 22, 2019 @ 3:31 pm

    Compare the audio pronunciations for xiě 寫 here, here, and here (you have to enter the character into the Chinese translation box to get the speaker button).

RSS feed for comments on this post