Are there writing systems where there are cases of written form of words not preserving the order of speech, i.e. text(A) precedes text(B) in the written form, but speech(B) precedes speech(A)?

Only words written in the native script(s) should be considered: semiotics such as "$10" which is read as "10 dollars" are excluded explicitly.

And of course the distinction between LTR/RTL doesn't matter here, as the ordering here would be defined to be the respsective direction of each writing system.

  • 1
    Are you really excluding numerals? Are these not part of the "native scripts"?
    – fdb
    Commented Jul 28, 2021 at 15:12
  • 20
    Numbers are actually the best example for this. So, in German you write 46, but you read it as sechsundvierzig..
    – fdb
    Commented Jul 28, 2021 at 16:34
  • 2
    Devanagari long and short /i/ vowels violate the order.
    – jlawler
    Commented Jul 28, 2021 at 16:49
  • 4
    @JoL, that's how it is in German, the tens and ones are "swapped". Of course also in tens of thousands etc., so 1 234 567 reads like "one million, two hundred four-and-thirty thousand, five hundred seven-and-sixty", except with less spacing of course. I can't remember exactly but it seems they just write it all in a row as one word, shudder.
    – ilkkachu
    Commented Jul 29, 2021 at 10:45
  • 3
    @JoL rather like dates in American MDY format. And "four and twenty" was formerly common in English too
    – Chris H
    Commented Jul 29, 2021 at 11:10

11 Answers 11


Hieroglyphic Egyptian has a feature called "honorific transposition", where certain nouns can be written at the beginning of a noun phrase (regardless of their actual syntactic position) if they're more important than all the other nouns.

For example, one type of priest is called a ħm-nṯr, "servant of the god". Grammatically, in this construction, the "owner" noun (nomen regens) needs to come second, and its Coptic descendant is hont. But since "god" is clearly more important than "servant", it's written orthographically as nṯr-ħm.

This mostly only happens with words relating to deities ("god", "pharaoh", names of gods, etc), but during certain periods it was also common to transpose a father's name in front of a child's name. The scribe who wrote the best-surviving copy of the Shipwrecked Sailor, for example, recorded his name as jmny zꜣ jmnꜥꜣ, literally "Amoni, son of Amona'a" ("owner" noun coming second as above). But depending on the time period, this could instead be read as jmnꜥꜣ zꜣ jmny, "Amona'a, son of Amoni". Google says the papyrus is from the 12th dynasty, so I think it's the latter; transposing the father's name was common in that era.

There were also sometimes "aesthetic transpositions", where the graphemes were rearranged for purely aesthetic reasons. Variant spellings of the word for "in front of" demonstrate that it should be pronounced ḫft. However, the glyphs for and t are small, and f is long and flat, so arranging them in this order tends not to look good. Instead, it's most commonly written ḫtf, so that the two small signs can be arranged over the flat one.


There is a whole class of languages that use Abugida writing system which shows this phenomenon on a regular basis. Most prominent examples are Hindi and Thai.

Using Thai writing for examples below.

No final (CV):

  • เ◌ + ม = เม /e+m/ = /me/
  • โ◌ + ม = โม /o+m/ = /mo/
  • ไ◌ + ม = ไม /ai+m/ = /mai/

With Final (CVC):

  • เ◌ + ม + ม = เมม /e+m+m/ = /mem/
  • โ◌ + ม + ม = โมม /o+m+m/ = /mom/

Di- and triphthongs are pronounced in order different to the order of character appearance:

  • เ◌ + ต + ◌ี + ย + ง = เตียง (bed) /e+t+ii+j+ng/ = /tiang/

A more complex example demonstrating epenthesis (svarabhakti):

  • เ◌ + ส + ม + อ = เสมอ (forever) /e+s+m+_/ = /sa-mer/
  • เ◌ + ฉ + ล + ◌ี + ย + ง = เฉลียง (corridor; hall) /e+ch+l+ii+j+ng/ = /cha-liang/ (compare to the เตียง example above).
  • How come <e+tᵃ+ii+jᵃ+ng> gives ti*e*ng, but in the seemingly entirely parallel (leaving out the <ch>) case of <e+lᵃ+ii+jᵃ+ng> the result is li*a*ng? Commented Jul 29, 2021 at 9:18
  • 1
    the vowel is usually considered a diacritic here, rather than an independent glyph, so this is a kind-of borderline case as to whether it counts
    – Tristan
    Commented Jul 29, 2021 at 11:01
  • @JanusBahsJacquet, thanks, corrected. Commented Jul 29, 2021 at 14:06
  • 1
    I thought the OP was asking about word order, not grapheme order. Conceptually, /ka/ and /ke/ are different letters, even though they're clearly related and there are rules about how to derive one from the other. The vowel strokes are not thought of as independent from the consonant they attach to, much like the dot of an "i" or "j" is not thought of as a separate part. If you're going with letter order, there are English words like "what". When the /h/ is pronounced, it is "hwat", and that's how it was originally written, such as the first word of the epic poem "Beowulf".
    – CJ Dennis
    Commented Jul 30, 2021 at 2:00
  • 1
    @CJDennis It can probably be argued that what contains /hw/, but it’s more economical to treat it as a single phoneme /ʍ/, written with a digraph, parallel to /θ/ <th>. /h/ does not otherwise appear syllable-initially before consonants, the only exception being the sequence /ju/, which behaves differently to sequences of /w/ + vowel – as indicated also by the fact that /ʍ/ can also appear before /ju/, as in whew /ʍjuː/, while a sequence /hwj/ would be entirely unparallelled. Commented Jul 30, 2021 at 11:15

German numbers.

42 is read as "Zweiundvierzig", literally "Two and Forty".
This order swap only happens for numbers between 20-100.

  • 1
    It starts at 13 = dreizehn = "three ten"
    – knabar
    Commented Jul 30, 2021 at 7:36
  • 3
    @knabar Actually, many languages do the same thing with "teens" -> Thirteen, not "Ten-Three". Which is pretty odd, when you look at it, but explainable as simply extending the first ten into second ten (assuming people rarely had reason to count past that)
    – Thomas
    Commented Jul 30, 2021 at 10:32
  • @RonJohn French numbers are pronounced in the same order as English.
    – asac
    Commented Jul 30, 2021 at 14:04
  • @asac you're right. I was thinking of "80", which is "four 20" in French. Odd, but a different rule.
    – RonJohn
    Commented Jul 30, 2021 at 14:23
  • 1
    @md2perpe the Danish number system is not used in Norwegian (it's in fact one of the things that Norwegians like to ridicule as utterly incomprehensible about the Danes). The standard Norwegian system is now basically the same as in English (77 = syttisju ≈ seventy·seven), though many people still use the German system (77 = syvogsytti ≈ seven·and·seventy), which was standard in Riksmål writing – Riksmål being indeed mostly Danish, but I doubt the number order came from Danish. Commented Jul 30, 2021 at 22:51

Sumerian cuneiform has certain fossilized transpositions, where a sequence of signs got re-analyzed as a single logogram and then did not change when the syntax of the language changed.

For example, it seems that adjectives used to precede the nouns they modify in Sumerian, but this order became archaic and eventually died out by the Ur III period: earlier gal-niŋir "chief herald", later niŋir-gal. But a few of these early compounds fossilized in writing, and later lú-gal "king" remained written GAL.LÚ. (Likewise ušum-gal "dragon" remains written GAL.UŠUM.) We know that the pronunciations were lú-gal and ušum-gal based on Akkadian glosses and phonological properties ("of the king" is written GAL.LÚ-la, for example, showing that the stem ends with an L), but the "inverted" spellings survived into Akkadian and Hittite, where the native words for "king" are still written with the logogram GAL.LÚ.


Japanese was once written as Chinese, with markers to indicate how the characters should be reorganised to Japanese grammar. This style of writing is called kanbun. The Chinese characters are written as they would be in China, with numerals and other markers written between them to indicate the order in which they should be read.

  • 3
    @jogloran James K is referring to Kanbun, a system of marking Classical Chinese text so that it can be read aloud in Japanese, which tends to put verbs last. See Wikipedia's 楚人有鬻盾與矛者 example.
    – yawnoc
    Commented Jul 29, 2021 at 13:29
  • 1
    @Janus: if you check the edit history, the responder has just completely changed the topic of their post, which is a little rude. It used contain an inaccurate claim about Cantonese.
    – jogloran
    Commented Jul 29, 2021 at 16:43
  • 1
    @jogloran Ah, I see! I wondered how Cantonese entered into this at all, but didn’t notice the edit that deleted that part of the answer. Commented Jul 29, 2021 at 16:44
  • Actually, I realised that this answer is still incorrect — is it really true that "Japanese was once written as Chinese"? I would say no. Kanbun is a completely separate style to Japanese prose. The source text in Kanbun is Chinese, through and through.
    – jogloran
    Commented Jul 29, 2021 at 17:05
  • 1
    The source is Japanese, as there are indications on reading which indicate that a Japanse reading is intended for some characters. The order of the characters is Chinese, and doesn't match the reading order, so markers are added to indicate the reading order. When read, you read according to the indicated order, not the page order. The underlying language is Japanese.
    – James K
    Commented Jul 29, 2021 at 17:15

English (in particular British English) does this more or less systematically, when a final e is preceded by certain consonant combinations. The e appears in pronunciation before the consonant or changes the pronunciation of vowels before the consonant:

  • apple → /æpəl/
  • subtle → /sʌtəl/
  • centre → /sentəɹ/
  • gape → /ɡeɪp/ (contrast with gap)
  • 7
    I don't think gape is an example of this, as it's part of a pattern of a silent e after a consonant elongating the vowel before it (which, due to the great vowel shift, is no longer a long /a/ but a diphthong /ei/).
    – Hearth
    Commented Jul 29, 2021 at 15:51
  • 3
    @Hearth: Hmm, half a sentence got lost there. Please see my edit. Anyway, the e has a backwards effect on pronunciation, whether it appears itself or not. I will leave it to the asker whether it is what they are looking for. Commented Jul 29, 2021 at 16:15
  • 2
    That e is silent. pl, tl, etc. are in fact final syllables with no vowels. If you're inserting a schwa between the letters you're pronouncing it wrong.
    – Joshua
    Commented Jul 30, 2021 at 16:43
  • 1
    @wjandrea: Indeed they are syllabic consonants. I am aware of that concept but hadn't known the name before. If you find me a dialect that pronounces that e but doesn't have a keep-vowel artifact (that is, where a vowel keeps on being pronounced between words all the time) I'll gladly retract my "wrong" but I haven't heard one yet.
    – Joshua
    Commented Jul 30, 2021 at 21:07
  • 2
    @Joshua I pronounce syllabic /l̩/ vocalically ([ˈʔæpʰu], [ˈsɐtˢu] etc.), but other syllabic consonants have a clear schwa: button [ˈbɐɾən], prism [ˈpʰɻʷʏzəm].
    – bradrn
    Commented Jul 31, 2021 at 2:25

Hebrew consists of syllables, with a vowel (which may be silent) after each consonant (which may also be silent). A reversal happens in the case of a pasach (approximately as English far) when combined with one of the guttural consonants, aleph (which is silent), mappiq (pronounced) hey, ches (as in Scottish loch), or ʿayin (⟨ʕ⟩), at the very end of a word. There, the consonant-vowel combination is pronounced in reverse.

For example is the Biblical name Noah which more correctly would be spelled Noach since the final consonant is ches. The Hebrew literally spells out No-cha, but the pronunciation of the ches-with-pasach cluster is reversed to ach.

  • 1
    Note that Adam's transliteration is in the Ashkenazi tradition, which matches neither that of European Hebraicists (who would write paṯaḥ) or Modern Hebrew speakers, who write pataḥ or patach, and pronounce the middle consonant as /t/. Similarly for ches.
    – Colin Fine
    Commented Jul 30, 2021 at 18:04
  • 1
    But also not that the specific pronunciation is irrelevant for the described phenomenon; everyone agrees about this reversal.
    – Adám
    Commented Jul 31, 2021 at 22:36

As you have discovered from the answers, writing systems do not exactly submit to a strict ordering of elements. In Devanagari writing, elements can overlap: there are "ligatures", many of which are illustrated on this page. For example, the consonant elements of the ligature kva, kla and in general rCa for any consonant overlap in left-to-right order. Vowel in Devanagari are mixed, but none of them strictly precede or strictly follow, they all overlap the consonant to some extent (pi=पि, pi:=पी). A few "mostly" follow or "mostly" precede (when they have a "stick"), but the general pattern is that vowels overlap the consonant mark (pu=पु). The "exception" is that certain long vowels have a stick that hosts the vowel marker, so that the mark for "a:" entirely follows the consonant (pa:=पा) since short a doesn't have a vowel mark. Thus in pe:=पे there is (mostly) overlap because the mark for "e" doesn't sit on a stick, but in po:=पो, there is just "some overlap" with p, because the mark (basically the same thing as the mark for "e") is on a stick.

Korean Hangul likewise does not easily submit to a strict left-right ordering analysis: instead, each syllable is a "box" and you position the consonants and vowels depending on whether the element is onset, nucleus or coda. The coda is "below", so it overlaps the onset and nucleus.

  • 3
    It's not as simple as LTR or such, but there is a specific, predictable ordering of the jamo within each hangeul square, like there is for the glyphs within an Egyptian quadrat.
    – Draconis
    Commented Jul 28, 2021 at 19:34

Some numbers in English (and some numbers of other European languages)

13 is Thirteen, literally meaning Three and Ten. Though the example is not as nice as @Thomas' example for German.

This is not unique to English. In Italian we have:

13 is tredici, literally meaning Three (tre) and Ten (dieci), as well.


In the Devanagari script, the sign for a short i precedes the consonant it follow in the spoken language, the syllable ki is written कि, while the syllable kī with long i is written की. Of course you can interpret this in different ways (e.g., as a diacritic) but the fact remains that at least on a Devanagi typewriter, the short i is produced before the consonant.

  1. Tones (e.g. in Chinese) are often indicated by a symbol after the syllable, but the tone apples to the preceding nucleus. This is especially visible in Wade–Giles, where the tone is a superscript number (syllable-final consonants are rare in Mandarin, but there are some), e.g. Wei¹ Chai² Shih⁴ P'in¹-yin¹.
  2. Many languages change the pitch to indicate questions, this is marked in writing by a question mark placed after the clause (Spanish and Armenian being two major exceptions).

Not the answer you're looking for? Browse other questions tagged or ask your own question.