Graphic antipairs

« previous post | next post »

Currently on the internet in China, there is a flurry of discussion on characters that are mirror, flipped, reversed, or inverted images of each other.  Here are some of the examples that have been cited (except for the last two sets, which were added by me to illustrate other types of minimal differences):

chǎng 厂 ("factory") || yí, 乁, ancient form of yí 移 ("move; shift") or 及 ("and; reach to")

piàn 片 ("sheet; piece; slice") || pán 爿 ("half of a tree trunk")

yù 玉 ("jade") || sù 玊 ("jade with a blemish; a jade worker; a surname")

chì 翅 ("wing; fin") || chì 翄 ("wing; fin"), a variant of chì 翅 ("wing; fin")!!

chǎng 昶 ("bright; long day; expansive; surname") ||  ǎi 昹 ("name of a star")

zè 仄 ("narrow; oblique tones in prosody; a feeling of unease") || wáng 亾 ("death; destroyed; lost perished"), an early variant of wáng 亡; another early variant is 兦

bì 币 ("currency; money") || zā 帀 ("to go round; to make a circuit; to make a revolution; to turn round; encircle"), a variant of 匝

yǔn 陨 ("fall from the sky; falling / shooting star") || yún 郧 ("name of an ancient kingdom or county; surname")

rén 人 ("person; human") || rù 入 ("into; enter")

yǐ 已 ("already") || jǐ 己 ("self") || sì 巳 ("the hours from 9 to 11; 6th terrestrial branch in the sexagenary calendrical cycle

Since there are hundreds, if not thousands, of such graphic minimal pairs that differ only by orientation or by presence, absence, or placement of a single tiny feature, one can get a sense of how challenging the Chinese script must be for dyslexics.  I'm not dyslexic, but I find that keeping these tiny differences straight requires enormous powers of concentration and memorization.


Geoffrey Pullum, "The Awful Chinese Writing System", Lingua Franca, The Chronicle of Higher Education (1/20/16).

David Moser, "Why Chinese Is So Damn Hard" (html),, readings; (简体字:为什么中文这么TM难?) (繁體字:為什麼中文這麼TM難?); from Schriftfestschrift: Essays on Writing and Language in Honor of John DeFrancis on His Eightieth Birthday (Sino-Platonic Papers (pdf) No. 27, August 1991), edited by Victor H. Mair.

Victor H. Mair, "Aphantasia — absence of the mind's eye" (Language Log, 3/24/17)

Other Language Log posts

[Thanks to Zeyao Wu]


  1. Michael Watts said,

    April 23, 2018 @ 12:43 pm

    部 bù "part" (of something); 陪 péi "accompany".

  2. John Rohsenow said,

    April 23, 2018 @ 1:09 pm

    Victor mentions dyslexia. Professor Hill Gates suggests that in fact aphantasia, lack of visual memory, which apparently affects 2%`of the English speaking population, causes great difficulty in learning and remembering Chinese characters.

  3. cameron said,

    April 23, 2018 @ 3:56 pm

    Has anyone formulated a mirror-palindrome in Chinese? Is it possible?

  4. Michael Watts said,

    April 23, 2018 @ 4:22 pm

    I don't see how the 币 / 帀 class exemplifies a "mirroring". Those are the same character with one stroke leveled. There is no axis or level of composition by which you could reflect one onto the other.

    This class does have a pretty common pair in it, though, 千 "1000" and 干 "dry" / "do".

  5. unekdoud said,

    April 23, 2018 @ 5:47 pm

    Rather notoriously, 够 is considered a simplified form of 夠, even though both are common variants in most regions.

  6. Bathrobe said,

    April 23, 2018 @ 7:09 pm

    I've seen 玊 before (I think it was in a large park at Huairou) and at first glance totally missed the difference.

  7. liuyao said,

    April 24, 2018 @ 8:50 am

    Mirror images we are all good at distinguishing (bd pq), though kids seem to have some trouble with 6 and 9 in print, especially on blocks). The list has many rather obscure characters that one would never see, perhaps likened to old typography of the Roman alphabet, or to distinguish ß from beta. Context should help.

    日 vs 曰 would be a good example for common characters.

  8. Anthony said,

    April 24, 2018 @ 9:35 am

    I take it that ease of distinguishing mirror images means that we perceive them as quite different from normal images. Some astronomical telescopes are configured to present an upright but right/left reversed image, and beginning astronomers are warned that they will have trouble using a normal lunar map; mirror-image maps are recommended.

  9. liuyao said,

    April 24, 2018 @ 10:30 am

    Another notorious example: 戊戌戍戎 (this year is 戊戌, two sexagenary cycle -versary of the event that made this pair famous for modern Chinese, along with 甲午, 庚子, and 辛亥. Despite that they have been used for countless events in the past two millennia, from this point on they are frozen to the fall of Qing dynasty in ways that other pairs don't.)

  10. Chas Belov said,

    April 24, 2018 @ 12:00 pm

    This would be a good time for me to reiterate my disappointment that ping pong in Chinese is not fully mirrored: 乒乓not乒乒

  11. Chas Belov said,

    April 24, 2018 @ 12:03 pm

    Oops, didn't work. I was trying the following code for the last character, but apparently span with a style attribute is not permitted in contents:

    [span style="-moz-transform: scale(-1, 1);-webkit-transform: scale(-1, 1);-o-transform: scale(-1, 1);-ms-transform: scale(-1, 1);transform: scale(-1, 1);"] 乒[/span]

    (brackets substituted for greater than/less than so you can see the code)

  12. Chas Belov said,

    April 24, 2018 @ 12:03 pm

    Also, would be interested to see this used as a technique to get around censorship.

  13. Neil Dolinger said,

    April 24, 2018 @ 12:16 pm

    yǐ 已 ("already") || jǐ 己 ("self") tough to avoid tripping on for sure!

    liuyao's examples reminded me of another pair of commonly used characters that while not quite mirrored, bedevil me to this day:

    我 Wǒ I
    找 Zhǎo find

  14. Neil Dolinger said,

    April 24, 2018 @ 12:22 pm

    "日 vs 曰 would be a good example for common characters."

    It's even worse with this pair, as both are used as radicals in dictionaries. I can't count how many times I guessed incorrectly which one to pick when trying to look up a character in the dictionary.

  15. Andreas Johansson said,

    April 25, 2018 @ 12:17 am

    Speaking of dyslexia, way back when (late '90s or early '00s) I was told that it just didn't exist in the hanzi-using world, the implication being that this was an advantage of hanzi over alphabetic scripts (rather than, say, superior teaching methods in use in China or something genetic). Anyone actually know anything about the relationship, if any, between script type and the incidence of dyslexia?

  16. Jerry Friedman said,

    April 25, 2018 @ 1:43 pm

    liuyao: Mirror images we are all good at distinguishing (bd pq)

    Some people with dyslexia have trouble with those.

    I teach physics, which uses the Greek letter λ, so a lot of my students find themselves writing it for the first time. A sizable minority write the mirror image instead.

  17. ~flow said,

    April 26, 2018 @ 10:05 am

    @Andreas Johansson: As far as I remember the notion that spelling-challenged Westerners could learn Chinese characters without troubles goes back to one very limited experiment by one American schoolteacher; upon finding that students could reproduce some basic characters without missing or swapping strokes, an article got published in the local paper from where it was subsequently picked up by the media and blown up to fantastic proportions. In other news, they have lots of words for snow in Greenland.

  18. ~flow said,

    April 26, 2018 @ 10:28 am

    FWIW, 財前謙 (ざいぜんけん) discusses characters that are identical except for their componential arrangements in his 2010 book, 字体のはなし . He notes that they've recently become a topic among specialists for CJK writing who call them 動用字 ('mobile' characters?); the term refers especially to cases like 峯 v 峰, 群 v 羣 where the different arrangements are considered alternative spellings, not different characters.

  19. Andreas Johansson said,

    April 27, 2018 @ 2:12 am

    @~flow: That spelling-challenged westerners could learn hanzi without trouble wasn't quite what I was told – namely that dyslexia doesn't exist among "native" hanzi-users – but I could see what I was told originating in that study via a process of Chinese(!) whispers.

  20. Giodisseo said,

    May 2, 2018 @ 6:02 pm

    Yet another example of typical scriptal whimsy from Mainland China. Fascinating! Any links to examples in the wild? My googlefu is failing me on the quest.

    I ask because I'm curious to see whether this is more about obscure characters than minimal pairs. The ones provided by Victor Mair (and Michael Watts) are only meant to be classic illustrations of the latter, as clarified in the original post, so would we all agree they wouldn't fully qualify for this particular exercise?

    If I'm allowed to extend Professor Mair's analysis a bit, I'd say that, based on the original list, the criteria for inclusion seem to be that characters:
    1. Be "mirror, flipped, reversed, or inverted images" of the ones intended, as explained in the post; but possibly also
    2. Be infrequently used, archaic, or obsolete.
    The fun seems to be all in getting to read and write "different" characters – to see, as it were, old familiar words in shiny new clothes (in what, I suspect, might be otherwise mundane contexts). In a very small way, it's actually quite rewarding to almost feel your brain automatically flip obscure characters for you, as if you were suddenly in possession of some magical powers of decryption.

    Swapping 人, 已 or 部 for such equally high-frequency characters as 入, 己 and 陪 wouldn't be particularly clever or amusing, just a bit obtuse and confusing as they're all usual fixtures of running text. I suspect they might be perceived as "uneducated" typos or, at best, disarmingly lame jokes – not at all in keeping with the clever-clever spirit of writing/reading such gems as 乁,爿or 玊.

RSS feed for comments on this post