Diacritics: Every Diacritical Mark, With Its Letters to Copy

What is the difference between a diacritic, an accent and a diacritical mark?

Diacritics are the marks added to letters, and a diacritic is any one of them. Oxford's learner's dictionary (opens in a new tab) defines it as a mark such as an accent, placed over, under or through a letter, to show that the letter is said differently from the same letter without it. The word comes from the Greek diakritikos, from diakrinein, to distinguish. Diacritical mark means the same thing, built on the adjective.

Accent is the looser word. In everyday English it covers any of these marks: Oxford's entry for accent (opens in a new tab) gives the sense a mark on a letter and points to the acute, the circumflex and the grave.

Unicode draws the line more tightly. Its names list (opens in a new tab) files the marks on this page in a block called Combining Diacritical Marks. Inside it, only the acute, grave and circumflex families have ACCENT in their names. The rest are just named: COMBINING DIAERESIS, COMBINING CEDILLA, COMBINING OGONEK.

So every accent is a diacritic, and the diaeresis on ü or the ogonek on ą is a diacritic that is not an accent. The names list also records the other names each mark goes by, and what some of them do in one language:

Unicode name	Code	Other names in the list	What the list notes
COMBINING ACUTE ACCENT	U+0301	stress mark; Greek oxia, tonos	marks the Mandarin second tone in pinyin
COMBINING GRAVE ACCENT	U+0300	Greek varia	marks the Mandarin fourth tone in pinyin
COMBINING CIRCUMFLEX ACCENT	U+0302	hat	none
COMBINING DIAERESIS	U+0308	double dot above, umlaut; Greek dialytika	none
COMBINING TILDE	U+0303	none	IPA nasalization; a Vietnamese tone mark
COMBINING MACRON	U+0304	long	marks the Mandarin first tone in pinyin
COMBINING BREVE	U+0306	short; Greek vrachy	none
COMBINING OGONEK	U+0328	nasal hook	Polish, Lithuanian
COMBINING HOOK ABOVE	U+0309	hoi	a Vietnamese tone mark
COMBINING DOT BELOW	U+0323	nang	a Vietnamese tone mark
COMBINING DOUBLE ACUTE ACCENT	U+030B	none	Hungarian, Chuvash

Umlaut, then, is not a second mark. Unicode treats it as another name for the diaeresis, one character for both jobs.

Most lists of the types of accent marks stop at the familiar ones: acute, grave, circumflex, diaeresis, tilde and cedilla. The accent marks list at the top of this page goes further. It holds every mark a letter in AccentPad's catalogue carries, ranked by how many languages write it, so the rare diacritics that no CLDR alphabet uses sit at the end instead of being left out.

Which diacritics make a letter of its own, language by language?

Diacritics do one of two jobs. A mark can make a new letter, with its own place in the alphabet. Or it can sit on a letter that stays the same letter, and only change how it sounds or where the stress falls. Sorting shows which job it does, because a dictionary files a new letter in its own place.

The Unicode Collation Algorithm (opens in a new tab) builds this in. Its first level sorts the base letters; accents only count at a later level, so the difference between o and ô is swamped by any difference in the letters themselves. Each language's rules in the Unicode CLDR then promote the marked letters that it treats as letters of their own to that first level.

That gives a test anyone can run: compare a marked letter with its plain letter by the language's own rules, with accents ignored. If the two still differ, the language sorts the marked letter as a letter.

Read across a row and the pattern is clear. Spanish promotes ñ alone; why that one letter matters is on the N with tilde page. Swedish promotes å, ä and ö but sorts é with e. Finnish promotes the same three and sorts š and ž with s and z. German sorts ä, ö and ü with their plain letters.

French promotes nothing. Its diacritics change sounds, not letters, and the OQLF rule for computer sorting (opens in a new tab) ignores diacritics except to order words that differ only by them, plain letter first, then acute, grave, circumflex and diaeresis.

Vietnamese splits one letter in two. Its breve, circumflex and horn make the letters ă, â, ê, ô, ơ and ư, which sort apart. Its tone marks do not, so ấ files with â, not as a letter of its own. Sorting is not the whole story, though. Czech lists á among its letters and still sorts it with a, as the Czech alphabet page explains.

How does Unicode store a diacritic: spacing or combining?

Text stores diacritics in two ways, and the same é can be either. It can be one precomposed code point, U+00E9. Or it can be a plain e followed by U+0301 COMBINING ACUTE ACCENT. Unicode's normalization FAQ (opens in a new tab) calls the two forms canonically equivalent, and says programs should always compare them as equal. NFC picks the composed form and NFD the decomposed one; UAX #15 (opens in a new tab) defines both.

A combining mark always follows its base. Unicode's FAQ on combining marks (opens in a new tab) describes a combining sequence as a base character followed by one or more combining characters. So a letter with no precomposed form still exists as text: a base and a mark, in that order. A letter can even carry two marks at once, which is why some letters appear on two cards above.

There is a third look-alike that is not the same text at all. Unicode also encodes spacing copies of the common marks, such as ´ U+00B4 ACUTE ACCENT and ˘ U+02D8 BREVE. Chapter 7 of the Unicode Standard (opens in a new tab) says they exist mainly for compatibility with older character sets, to show a mark on its own. They take up their own space and never attach.

What is typed	Code points	Is it é?
é, precomposed (NFC)	U+00E9	yes
e plus the combining acute (NFD)	U+0065 U+0301	yes, canonically equivalent
e plus the spacing acute	U+0065 U+00B4	no: two characters, the mark beside the e

The stroke is the exception on this page. Unicode does not decompose ł, ø or đ into a letter and a mark, so they have no combining form at all. The L with slash page covers what that means for ł.

Why does stripping diacritics break search?

Because the strip does not know which language it is reading. The usual way to make text free of diacritics is short: decompose it to NFD, then delete every combining mark. That merges Spanish ñ with n and Swedish å with a, the very letters the table above shows those languages sorting as letters of their own.

We ran that strip over the 252 small letters with a mark in AccentPad's catalogue when this page was built (Node.js v26.0.0, ICU 78.3, CLDR 48.0). 247 came out as a plain letter from a to z. 5 came out unchanged: đ, ħ, ł, ø and ŧ. Their mark is drawn into the letter, so there is nothing to strip.

For French it is closer to the rule, since French sorting ignores diacritics anyway; even there it throws away the order of words that differ only by an accent. And it leaves the stroke letters alone, so a name with ł or ø still fails a plain-ASCII match after it.

The fix is to compare text by the reader's language instead of deleting diacritics. A comparison with accents ignored, run on the language's own rules, is the test this page used above. It matches é with e for a French reader and keeps ñ apart from n for a Spanish one.

Is every mark under a letter a cedilla?

No, and the cedilla card above already shows why the old rule that it only goes under c fails: Latvian ģ and Turkish ş carry it too, and the names list notes both languages beside those letters.

The trouble with diacritics below the letter is that the name and the drawing do not always agree. Unicode's names list says this about Latvian ķ, ļ and ņ: despite their names, they should normally be displayed with a comma below. It adds that ģ has three major glyph variants.

Romanian goes the other way. Its ș and ț take U+0326 COMBINING COMMA BELOW, and the names list says the cedilla letters should not be used for Romanian; the S with accent page tells that story. Two other marks also sit under the letter and are easy to mistake for a cedilla: the ogonek, which Unicode also calls the nasal hook, and the dot below, which marks a tone in Vietnamese.