Full-width and half-width punctuation for Chinese and Japanese text. Which to use, why mixing them looks wrong, and how the quotation conventions differ.

Full-width and half-width

The single most useful thing to know about CJK punctuation is that most marks exist twice:

Full-width	Half-width	Name
`，`	`,`	comma
`。`	`.`	full stop
`；`	`;`	semicolon
`：`	`:`	colon
`？`	`?`	question mark
`！`	`!`	exclamation mark
`（　）`	`( )`	parentheses

A full-width mark occupies the same square as a Chinese character. A half-width mark is the ordinary ASCII one and occupies about half that. In a line of Chinese characters — each a perfect square on a fixed grid — an ASCII comma leaves a gap that is visibly too small, and the line stops looking typeset.

The rule in Chinese and Japanese text is simple: use the full-width mark inside CJK text, and the half-width mark inside Latin text, even when the two are in the same paragraph.

他說：「這樣可以嗎？」        ← full-width throughout
He said, "Is this all right?"  ← half-width throughout

Do not add a space after a full-width mark

A full-width mark already contains its own trailing space — that is what the extra width is. Typing ， followed by a space produces a gap of one and a half characters, which is the most common giveaway of text typed on a keyboard set to the wrong mode.

正確：今天很熱，我們待在室內。
錯誤：今天很熱， 我們待在室內。

The same applies to 。？！：；.

Quotation marks differ across regions

This is where the conventions genuinely diverge, and neither is a mistake:

	Primary	Nested
Taiwan, Hong Kong	`「」`	`『』`
Mainland China	`“ ”`	`‘ ’`
Japan	`「」`	`『』`

Traditional-Chinese text uses corner brackets; Simplified-Chinese text uses curly quotation marks. A page converted from one script to the other without converting its quotation marks reads as slightly foreign even when every character is correct.

Vertical text uses the rotated forms ﹁﹂, which is why the corner brackets exist at all — they work in both orientations, and curly quotes do not.

The marks with no Western equivalent

Four are worth knowing because there is no obvious substitute:

Mark	Name	Used for
`、`	頓號	separating items in a list — not a comma
`《》`	書名號	titles of books, films, works
`〈〉`	單書名號	titles of articles and chapters within a work
`‧`	間隔號	separating parts of a transliterated foreign name

The 頓號 is the one most often got wrong. In Chinese, list items are separated by 、 and clauses by ，, so 蘋果、香蕉和橘子 is correct and 蘋果，香蕉和橘子 is not.

Dashes and ellipses

Mark	Name	Convention
`—`	破折號	written doubled in Chinese: `——`
`…`	刪節號	written doubled in Chinese: `……`
`～`	全形波浪號	ranges, and a softening tone in casual writing

Both the dash and the ellipsis are conventionally two characters wide in Chinese typesetting, which is why they are typed twice. In English each is used singly.

Where the mode gets switched by accident

Most Chinese input methods toggle full-width and half-width punctuation with Shift+Space or a mode button, and it is easy to hit by mistake. The symptoms:

A document where some commas are wide and some are narrow.
Code that will not compile, because a full-width （ was typed inside a string or a bracket — the compiler sees a character it does not recognise as punctuation at all.
Filenames and URLs that look right and do not match, for the same reason.

Full-width and half-width marks are different characters with different code points. ， is U+FF0C and , is U+002C. Nothing will treat them as equal, so a find-and-replace has to target the exact one.

Punctuation marks