Familiar cases of metathesis involve segments changing places, but metathesis can also operate at the subsegmental level, affecting individual features. I'm specifically interested in metathesis of POA features, like the following examples:

  • German finster "dark, gloomy" < * θimster (edit: this is doubtful as a case of feature metathesis, see jknappen's comments)

  • Colloquial Hebrew maʃaz "massage" < masaʒ


  1. What are some other examples of this process? It's difficult to find data as this type of metathesis seems to be rare and not much studied.
  2. Are there any studies that deal specifically with POA feature metathesis -- e.g. examining its typology, which features do and don't undergo metathesis under which kinds of circumstances, etc.?
  3. As a bonus question, are there any attested cases of voicing metathesis (e.g. tab > dap)? I'm aware that this shows up in child language acquisition errors, but beyond that?
  • Interesting. The Hebrew example reminds me of something I read about restrictions on Arabic triliteral roots: roots like sh-m-s are permitted, but roots like s-m-sh do not exist. Dec 6 '15 at 19:50
  • 2
    I think the German example can explained without invoking feature metathesis. Dec 7 '15 at 9:42
  • @jknappen How so?
    – TKR
    Dec 7 '15 at 17:26
  • 1
    @TKR: The long explanation can be found in Grimms Wörterbuch (in German) here: woerterbuchnetz.de/DWB/…. Summarized: The alternation th/f is not unprecedented and not conditioned by a following m. The ms is assimilated to ns. Both are natural processes. Dec 7 '15 at 19:56
  • 1
    Here's a similar example in Arabic: linguistics.stackexchange.com/questions/15354/… Dec 22 '15 at 3:23

In Greek declensions there are a couple examples of metathesis quantitatis which is simply vowel length swapping.

Within the second declension we have the Attic declension which consists of nouns like νεώς or λεώς. Originally these forms had long stem vowel and a typical short thematic vowel. Due to a regular sound change the lengths of the vowels changed places.

νηός > νεώς

ληός > λεώς

This is seen throughout the whole paradigm. (https://en.wikipedia.org/wiki/Ancient_Greek_grammar_(tables)#Attic_second_declension)

There is quantitative metathesis happening also in the third declension, in the genetive of nouns like βασιλεύς or πόλις.

The standard Attic genetive is βασιλέως and πόλεως whilst the Homeric Greek yields βασιλῆος and πόληος.



A Greek Grammar for Colleges, Herbert Weir Smyth

Gramatyka grecka, Marian Auerbach, Marian Golias

  • 1
    Greek quantitative metathesis is an interesting (and to my knowledge unparalleled) phenomenon, but I'd say it's quite different from consonantal POA feature metathesis.
    – TKR
    Dec 7 '15 at 17:28
  • If I am not mistaken a kind of quantitative metathesis happened also during the developments in Common Slavic period. It was connected to the loss of the Slavic "semivowels" (I am not sure what is the English term, in Polish, in Slavic lingustics we use term "jer"; the symbols used are <ь> and <ъ>)
    – czypsu
    Dec 7 '15 at 17:33
  • That's interesting, as I've been looking for parallels to Greek QM for a while. Do you know anywhere I could read more about it?
    – TKR
    Dec 7 '15 at 17:46
  • 1
    It turns out I was mistaken. The case I was talking about is merely a case of compensatory lengthening which accompanies the loss of aforementiond jer vowels, not a QM.
    – czypsu
    Dec 7 '15 at 19:57

I suggest Ultan's typological study for an overview, and the metathesis database for a few examples. I'm extremely skeptical about the prospects for finding a general case where just place features are moved but not voicing and manner, but place is certainly relevant as a conditioning factor.

  • Thanks, I'd looked through Ultan's dissertation without finding examples of place feature metathesis; I didn't know about the metathesis database, but will check that out.
    – TKR
    Dec 7 '15 at 17:29
  • On page 19, something similar to the Hebrew example is discussed: in some French dialects and creoles in Dominica, /s/ followed by /ʃ/ in the same syllable is subject to metathesis, resulting in /ʃ...s/ (example: French /sɛʃ/ "sèche" corresponds to Agde /ʃɛs/). Of course, it is not possible to distinguish this from ordinary metathesis, since the voicing and manner are identical between the two consonants. Dec 7 '15 at 21:36

This isn't direct evidence, but this paper mentions several proposed etymologies that make use of feature metathesis: Feature metathesis and the change of PIE *du̯ to Classical Armenian -rk (Diachronica 2013), by Jessica DeLisi

The only example of possible place feature metathesis mentioned is on page 481, and the evidence is ambiguous between that and a simple segmental metathesis.

However, there are some interesting examples in it of "aspiration anticipation," which could be regarded as a kind of metathesis, where an aspirated consonant or the segment /h/ is lost in the middle of a word, but the start of the word receives compensatory aspiration. Here's one of them:

PIE *h1eu̯s “to burn” became Greek heúō “to burn, scorch” with unexpected rough breathing compared to Sanskrit oṣati “burns”. Aspiration anticipation is quite common in Greek in words with internal *-s- > Pre-Homeric *-h-.


Maybe the following two examples from German count ...

  • Sybille /zy:bɪlə/, a given name coming from the classical name Sibylle /zi:bʏlə/. The rounding is swapped between the first and second vowel while other features (like length and tenseness) are kept.
  • Popular mispronounciation of Libyen as "Lübien". Again a vowel rounding swap.
  • I think these are spelling pronunciations, resulting from a purely orthographic fluctuation between y and i.
    – fdb
    Feb 12 '16 at 14:49
  • @fdb: I think they aren't. An unstressed ü is that uncommon in German that the speakers unconsciously want to correct "that mistake". Feb 12 '16 at 14:54
  • I don't understand your transcriptions -- what do I and Y stand for?
    – TKR
    Feb 12 '16 at 16:53
  • @TKR: It is kind of ASCII-IPA. I is small capital I (as in the English word fit) and Y is small capital Y (as in the German word Glück) Feb 12 '16 at 17:01
  • Thanks, this is interesting. What do you mean by length being retained, though?
    – TKR
    Feb 12 '16 at 23:59

One good example is what happened to PIE labiovelars in Greek. At some point, Greek lost its labiovelar consonants as such. Classical Greek has three consonant series

  • Greek labials Π, Β, Φ (from PIE *p, *b, *bʰ, respectively)
  • Greek dentals Τ, Δ, Θ (from PIE *t, *d, *dʰ)
  • Greek velars Κ, Γ, Χ (from PIE *k, *ɡ, *ɡʰ)

But PIE had a labiovelar series as well.

  • PIE labiovelars *kʷ, *ɡʷ, *ɡʰʷ

These labiovelar consonants wound up in Greek smeared all across the mouth as labials, dentals, or velars. Occasionally, as with the reflexes of *kʷel- 'wheel', one can find all three:

  • πολος 'axis', τελος 'end of cycle', and κυκλος 'circle' all come from PIE *kʷel-.
  • 2
    But there's no metathesis there, just different conditioned reflexes based on the adjacent vowels.
    – TKR
    Dec 7 '15 at 1:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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