Shrug Emoticon
The shrug emoticon and the rest of the shrug family, ready to copy. The section below explains why a field that parses Markdown prints the face with one arm missing, and what to type instead.
Tap any tile to copy it. Nothing is sent anywhere — the copy happens in your browser.
The face is a Japanese syllable
The head in ¯\_(ツ)_/¯ is ツ, U+30C4 KATAKANA LETTER TU — the kana for the syllable romanised “tsu”. It is an ordinary letter, no more a smiley than a lowercase w is. Its three strokes happen to sit like two eyes above a curving mouth, and that resemblance is doing all the work.
Everything else is Latin-1 and ASCII:
| Character | Name | Job |
|---|---|---|
¯ |
U+00AF MACRON | An open, upturned palm |
\ |
U+005C REVERSE SOLIDUS | The left arm |
_ |
U+005F LOW LINE | A shoulder |
( |
U+0028 LEFT PARENTHESIS | The left side of the head |
) |
U+0029 RIGHT PARENTHESIS | The right side of the head |
ツ |
U+30C4 KATAKANA LETTER TU | The face |
/ |
U+002F SOLIDUS | The right arm |
Nine characters, thirteen bytes. Two of those characters sit outside ASCII, so the shrug is turned away by fields that accept nothing else; the faces that do fit in seven bits are on the ASCII emoticons page. Two near-identical kana are easy to mistake for ツ: シ (U+30B7, “shi”) has strokes that enter from the left and reads as a face tilted the other way, and ヅ (U+30C5, “du”) is ツ plus a voiced sound mark, which lands above the right eye like a wink.
Why one arm goes missing
Paste the shrug into a chat client that formats messages with Markdown and it arrives as ¯_(ツ)_/¯, left arm gone. That is not a font problem. In Markdown and the formatting dialects derived from it, a backslash before a punctuation character is an escape: it tells the parser to treat the next character literally, and then it is consumed and never printed. The shrug’s left arm is a backslash sitting immediately before an underscore, which is exactly the sequence an escape is made of, so the parser reads \_, prints _, and throws the arm away.
Doubling the backslash is only half a fix. Running each version through a CommonMark parser gives:
| What you type | What is printed |
|---|---|
¯\_(ツ)_/¯ |
¯_(ツ)_/¯ — arm eaten |
¯\\_(ツ)_/¯ |
arm restored, but the two underscores now pair up and italicise everything between them |
¯\\_(ツ)\_/¯ |
¯\_(ツ)_/¯ — correct |
So you have to escape twice: the backslash, and then the second underscore that would otherwise close an italic run. The alternative is to sidestep the parser entirely by wrapping the face in backticks, which puts it in a code span and prints it exactly as typed. CommonMark is one dialect among several and a client may implement another, so if a client mangles the face, test the escaped form in that client before trusting it. The same escape rules bite every face with a backslash or an underscore in it, and the copy-and-paste notes go through the other places a face arrives changed.
The same backslash causes trouble outside chat. In JSON it must be written \\, in regular-expression syntax it must be escaped to match, and in a shell that treats the backslash as an escape character it will consume whatever follows it unless the face is quoted. The kana-only shrugs — ヽ(ツ)ノ, ＼(－－)／ — avoid all of this, because they contain no backslash at all.
The Japanese shrugs are a different gesture
The ┐(´д`)┌ and ╮(╯_╰)╭ families are built the other way round: the arms are box-drawing corners, and they hang down from the shoulders rather than reaching up. The reading shifts with them. The Latin shrug is dismissive — not my problem, who knows. The box-drawing shrug is closer to being at a loss, and it is the one that pairs with a sweat mark. If that is the reading you are after, the confused faces carry it further, adding a question mark where the face expects nothing.
More faces built the same way are in the kaomoji collection; the other two single faces with a page to themselves here are the Lenny face and the table flip.