Four lines, four jobs: hyphen, en dash, em dash, minus
A keyboard has one horizontal line. Typeset text has four, and they are not interchangeable. What each one is for, and how the engine tells them apart.
A keyboard has one horizontal line, the hyphen. Typeset text has four: the hyphen -, the en dash –, the em dash — and the minus sign −. They differ in width, in vertical position, and in meaning. A text that uses the hyphen for all four reads fine until it does not: pages 5-9 looks like a compound word, -5 °C has a sign too short and too low for its digits.
Who does what
| Character | Job | Typed |
|---|---|---|
Hyphen - |
joins words: well-set, self-hosted |
stays as typed |
En dash – |
a range: 21–26 September, 10:00–12:00 |
21-26 becomes 21–26 |
Em dash — |
a break in a sentence — like this | word - word becomes word — word |
Minus − |
negative numbers: −5 |
-5 becomes −5 |
Three of the engine’s rules draw the lines. dash.range turns a hyphen between two numbers into an unspaced en dash, even when the writer spaced it (5 - 9 becomes 5–9). dash.sentence turns a spaced hyphen, dash or double hyphen into the sentence dash. symbol.minus turns a hyphen before a digit into a true minus. A hyphen inside a word is never touched: joining words is the hyphen’s actual job.
The sentence dash depends on the language
English display text gets a spaced em dash. German gets a spaced en dash — the em dash is simply not German practice. British literary text gets a spaced en dash too, which is why the engine’s literary register changes the dash along with the quotation marks.
The space before the sentence dash is a no-break space, so a line never begins with a dash that belongs to the sentence before it.
Honesty requires a note: the spaced em dash is a choice, not a law. Butterick prefers the unspaced em (word—word), Chicago agrees, and plenty of fine books use the spaced en. The engine’s sources registry records the decision instead of pretending there is consensus; disable: ['dash.sentence'] opts out entirely.
The dialogue dash
Languages that write dialogue with a dash each do it their own way, and the engine’s dash.dialogue rule follows the language pack: Ukrainian opens a turn with an em dash and a no-break space (— Привіт), Spanish attaches the raya to the first word (—¿Esperamos?), Finnish uses an en dash. A paragraph where later lines also open with a dash is read as a list and left alone. The quotation marks article tells the same story for quotes: the character depends on the language, not just the rule.
Try it
Paste It was -5 - maybe -7 - on 21-26 September into the playground and watch the Changes tab assign each line its character: minus, em dash, en dash, each from its own rule.
Something wrong or unclear here? Report an issue.