Are there any diacritics not on the top or bottom of a letter?

Question

I don't know of any besides the horn on Ơ and Ư and the middle tilde on ᵯ and some other consonants I'm interested in particular in a diacritic precomposed with both "I" and "U"

Answer 1

Hebrew is one of those languages.

The dagesh is placed inside the letter. For example:

• Bet without dagesh: ב
• Bet with dagesh: בּ‎

The shuruk vowel point (nikkud) is placed to the left of the letter (afterwards, from the perspective of reading, since reading is done from the right to the left).

• Vav without pointing: ו
• Vav with shurruk: וּ

Note that these diacritics are considered optional in most contexts, and in fact are typically omitted from most printed material. The assumption is that if you know the language well enough, you will know how to pronounce the word without needing to check the diacritics. Diacritics are used in dictionaries, children's books, materials for learning Hebrew, and the Bible.

Answer 2

Some examples: d̵ which does not render well here (using the separate diacritic), but exists in Ð and đ, as well as e.u. o̵ (again, rendering problem), that is o-bar, and do on. The latter is official IPA, the former is orthography (đ in a number of languages including Saami, Vietnamese and S. Slavic). The rhotic hook ˞ is also not on the top or bottom, though it is top-ish. Slash is a conponent of a few symbols, e.g. Ⱥ in Saanich, ø in Norwegian, Danish, Faroese and S. Saami, Ł in various W. Slavic languages, ƛ in Lushootseed and other languages of the Pacific Northwest.

Answer 3

Comanche uses “U bar” <Ʉ, ʉ> in the official orthography for /ə/. Other languages that use this letter in their official orthographies include Kanakanabu (an Austronesian lanuage of Taiwan) and Koyukon for /ɞ/ (an Athabascan language spoken in Alaska).

<Ɨ, ɨ> is used in Mfumte (Nfumte), a Grassfields Bantu language of Cameroon for /ɯ/.

Apart from vowel letters, <Ŧ, ŧ> known as “T with a bar” or “T with a stroke sign” is used in Northern Sámi alphabet, where it represents the voiceless dental fricative [θ].

Answer 4

In the Japanese hiragana syllabary, the softening mark (dakuten) on the symbol て (te) is technically inside the letter, giving で (de).

Answer 5

In Modern Greek the accent (tonos — τόνος) is placed on the left of capital letters:

Ά Έ Ί Ό Ύ Ώ

Contrast that with the diairesis sign (dialytika — διαλυτικά), which is placed on top:

Ϊ Ϋ

In Ancient Greek the same holds true for the acute accent (oxeia — οξεία), the grave accent (vareia — βαρεία), the circumflex (perispomeni — περισπωμένη), the rough breathing (psilòn pneûma — ψιλὸν πνεῦμα), the smooth breathing (dasù pneûma — δασὺ πνεῦμα), and their combinations:

Ἀ Ἁ Ἂ Ἃ Ἄ Ἅ Ἆ Ἇ Ὰ Ά Ἐ Ἑ Ἒ Ἓ Ἔ Ἕ Ὲ Έ Ἠ Ἡ Ἢ Ἣ Ἤ Ἥ Ἦ Ἧ Ὴ Ή Ἱ Ἲ Ἳ Ἴ Ἵ Ἶ Ἷ Ὶ Ί Ὀ Ὁ Ὂ Ὃ Ὄ Ὅ Ὸ Ό Ὑ Ὓ Ὕ Ὗ Ὺ Ύ Ὠ Ὡ Ὢ Ὣ Ὤ Ὥ Ὦ Ὧ

Again, contrast this with lowercase letters, which have the corresponding signs placed on top in both Ancient Greek and Modern Greek:

ἀ ἁ ἂ ἃ ἄ ἅ ἆ ἇ ἐ ἑ ἒ ἓ ἔ ἕ ἠ ἡ ἢ ἣ ἤ ἥ ἦ ἧ ἰ ἱ ἲ ἳ ἴ ἵ ἶ ἷ ὀ ὁ ὂ ὃ ὄ ὅ ὐ ὑ ὒ ὓ ὔ ὕ ὖ ὗ ὠ ὡ ὢ ὣ ὤ ὥ ὦ ὧ ὰ ά ὲ έ ὴ ή ὶ ί ὸ ό ὺ ύ ὼ ώ ῒ ΐ ῖ ῗ ῢ ΰ ῦ ῧ ῶ

Answer 6

Polish has one letter with oblique strike in the middle and two letters with hooks in the right bottom corner. All these have upper and lowercase forms.

Ł, ł, Ą, ą, Ę, ę

Answer 7

In Czech, the caron on lowercase letters "t" and "d" is placed to the side rather than directly on top: ť, ď.

Answer 8

One approach to answering this type of question is to query the Unicode Character Database. One of the many useful features of Unicode that is it tracks various properties of each character, some relevant ones being: Is a Mark a Diacritic?, In What Position if any Does the Mark Combine with a Base Character?, What Block of Characters is a Character in?, What Characters does a Pre-Composed Character Decompose to?, and What is the Name of the Character?

Unicode has an online search utility, UnicodeSet, that allows us to specify one or more properties and will return all code points that match.

Let's start with a strict interpretation of the title of this question:

Are there any diacritics not on the top or bottom of a letter?

Now, there are actually several options for top and bottom ("above" left or right, "below" left or right, attached, etc), but those are still either "above" or "below" so they will be excluded. That leaves us with: left, middle / inside, and right.

The following search criteria returns 21 code points that are diacritics that combine on the left, right, or in the center (i.e. "overlay"):

[[:Diacritic=Yes:]&[:ccc=/^(1|21|224|226)$/:]&[:^Block=Musical_Symbols:]]

The ccc values (Canonical Combining Class) are:

1. 1 = Overlay (i.e. inside)
2. 21 = Hebrew Dagesh
3. 224 = Left
4. 226 = Right

I had to exclude 4 "Musical Symbol" characters that for some reason are marked as being diacritics (which might be an accurate categorization, but clearly not wanted in this case).

The question then goes on to mention:

... the horn on Ơ and Ư, and the middle tilde on ᵯ and some other consonants.

The "Horn" diacritic is categorized as "Attached Above Right" (ccc = 216), which is still "above" so it was initially excluded, but we can include it now along with "Attached Below Left" (ccc = 200), and "Kana Voicing" (ccc = 8). I originally excluded "Kana Voicing" class (the subject of @Jan's answer) due to the Unicode Standard (page 730 in version 12.0) describing those marks as follows (emphasis mine):

18.4 Hiragana and Katakana

Combining Marks. Hiragana and the related script Katakana use U+3099 combining katakana-hiragana voiced sound mark and U+309A combining katakana-hiragana semi-voiced sound mark to generate voiced and semivoiced syllables from the base syllables, respectively. All common precomposed combinations of base syllable forms using these marks are already encoded as characters, and use of these precomposed forms is the predominant JIS usage. These combining marks must follow the base character to which they apply. Because most implementations and JIS standards treat these marks as spacing characters, the Unicode Standard

And I wasn't sure if that really fit with a strict interpretation of the title. Either way, they are included now. The updated search criteria is:

[[:Diacritic=Yes:]&[:ccc=/^(1|21|224|226|200|216|8)$/:]&[:^Block=Musical_Symbols:]]

That returns 25 matching code points.

Ah, but the question goes on to say:

I'm interested in particular in a diacritic precomposed with both "I" and "U"

Fortunately we check decompositions. The following search criteria uses the non-Canonical decompositions to match any single character (i.e. the .) followed by a character having one of the combining classes noted in the previous search (excluding "Attached Below Left", ccc = 200, as that was breaking the search and not returning any value anyway):

[[:toNFD=/(?i).(\p{ccc=1}|\p{ccc=21}|\p{ccc=224}|\p{ccc=226}|\p{ccc=8}|\p{ccc=216})/:]&[:Alphabetic=yes:]]

That returns the following 106 code points (I had to filter out non-alphabetic characters as the search would otherwise return 164 code points, some of which are music symbols, some are math, etc):

ゞ ヾ ơƠớỚờỜỡỠởỞợỢ ưƯứỨừỪữỮửỬựỰ אּ-זּ טּ יּ כּךּ לּ מּ נּ סּ פּףּ צּ-שּשּׂשּׁ תּ ゔヴ がガ ぎギ ぐグ げゲ ごゴ ざザ じジ ずズ ぜゼ ぞゾ だダ ぢヂ づヅ でデ どド ばバぱパ びビぴピ ぶブぷプ べベぺペ ぼボぽポ ヷ-ヺ

Please note that no characters with a middle tilde are returned ("ᵯ" == U+1D6F). This is due to none of them having a stated decomposition, which is noted in the Unicode Standard (page 298 in version 12.0):

Other Phonetic Extensions. The remaining characters in the phonetics extension range U+1D6C..U+1DBF are derived from a wide variety of sources, including many technical orthographies developed by SIL linguists, as well as older historic sources.

All attested phonetic characters showing struckthrough tildes, struckthrough bars, and retroflex or palatal hooks attached to the basic letter have been separately encoded here. Although separate combining marks exist in the Unicode Standard for overstruck diacritics and attached retroflex or palatal hooks, earlier encoded IPA letters such as U+0268 latin small letter i with stroke and U+026D latin small letter l with retroflex hook have never been given decomposition mappings in the standard. For consistency, all newly encoded characters are handled analogously to the existing, more common characters of this type and are not given decomposition mappings. Because these characters do not have decompositions, they require special handling in some circumstances. See the discussion of single-script confusables in Unicode Technical Standard #39, “Unicode Security Mechanisms.”

NOTES

1. The O.P. also commented about focusing on Latin characters. With that in mind, in that final search we could have used [:Script=Latin:] instead of [:Alphabetic=yes:]. That would have returned just the 24 code points involving "O" and "U", as you can see here:
   [[:toNFD=/(?i).(\p{ccc=1}|\p{ccc=21}|\p{ccc=224}|\p{ccc=226}|\p{ccc=8}|\p{ccc=216})/:]&[:Script=Latin:]]

2. In that final search criteria, the (?i) is to set the search pattern to be case-insensitive. This isn't necessary when using the . to mean "any" character, but I left it in the search criteria to make it easier for anyone to change the . into a particular letter and still have it return upper-case and lower-case results.

3. Just FYI: there are even some diacritics that span two characters, either "above" or "below". For example, a double-tilde (code point U+0360):

   y͠o

Answer 9

(Others have already mentioned Hebrew, Serbo-Croatian đ and the Polish letters.)

A lot of the diacritics originated as a letter.

The dots in an umlaut represent an e. You can find examples with the little e inside the the a.

ь was added to л and н to yield Serbo-Croatian љ and њ (lj and nj).

I assume something similar for Russian ы.

But at least in the prescriptivist traditions of their respective homelands, they're not considered diacritics.

Answer 10

On the case of ð, Ð, ø and Ø discussed above, these are not diacritics. These are separate characters which there exists no "unmodified" form of.

Also, in Icelandic, Á, É, Í, Ó, Ú, Ý, Ö, á, é, í, ó, ú, ý, ö (right-leaning accent - NOT vertical) are all considered separate characters and have their place in the alphabet (and ö/Ö in Swedish and Norwegian) - and should therefore not be considered as alphabet characters using modifiers/diacritics - even though you do (for traditional, not practical reasons) need to push a modifier (mute-acute) to create all of these except ö/Ö which has a separate keyboard key.

I didn't see anyone mentioning the interpunct from Catalan and Occitan ( l·l / s·h )