Many languages disallow vowel-vowel sequences in a word or phrase, instead inserting an extra consonant between them to keep them apart. Some versions of English do this, like when Kennedy would say

Cuba-r-and America

inserting an epenthetic /r/ between vowels of adjacent words.

Which consonants are most commonly used for this purpose across languages?

  • 1
    Do you specifically mean consonants inserted between word? I'd think so, but then a better question would be "what are all the consonants, since inter-word epenthesis of consonants in not very common.
    – user6726
    Mar 11, 2017 at 22:14

3 Answers 3


I haven't done a survey of languages or anything, so this isn't rigorous, just a few examples of epenthetic consonants that I have commonly seen in languages I am familar with.

  • glottal stop. This occurs in some cases in English, actually, for example "uh-oh". It's productive in some contexts in some accents of German.

  • glides. English accents with epenthetic /r/ actually only use it after low or mid monophthongs (and in some accents, centering diphthongs). After rising-diphthong vowels or high vowels, transitions similar to the glides [j] (after front vowels) and [w] (after back vowels) are used instead. I know that French may also have a [j]-like transition after /i/ when it comes before another vowel phoneme. French also has /j/ derived from historical diphthongs that breaks up what would otherwise be hiatus in the inflected forms of some verbs, e.g. je vois /vwa/, vous voyez /vwaje/. This occurs after /wa/ (spelled "oi~oy") and /ɛ/ (spelled "ai~ay"). I don't know how productive it is but it exists.

  • Some other relatively common, not-very-marked consonant. E.g. "t" in French: while French actually does tolerate the existence of sequential vowels in hiatus (as in créer "create"), there is morphologically conditioned epenthetic /t/ in a number of derived verbs such as "siroter" < "sirop". Eau qui dort gives some more examples in the comments and points out that /n/ is used after a nasal vowel. Of course French also has /t/ and /n/ fairly often as "liaison" consonants inserted between words (the phenomenon of liaison is also conditioned by morphological and syntactic factors, not just phonology); another somewhat common liaison consonant is /z/. The most common liaison consonants in French are all coronals, and the coronal place of articulation is generally considered to be one of the least marked (I think the main possible competitors for title of "least marked POA" are velar and glottal, but velars don't act very much like the least-marked POA in French as far as I know, and French doesn't have any glottal phonemes.)

  • 1
    [j]'s also found in Ponapean and in Swati. :) Mar 10, 2017 at 15:27
  • Reg. French: /j/ in voyez is quite inevitable due to the preceding closed diphthong /ei/, which evolved from pretty much every proto-romance closed /e/, so the /j/ is not epenthetic between /e/ and /e/ (videtis > vedets > *veets), let alone /a/ and /e/. With -ai-/-ay-, this is actually /j/ disappearing before consonant or word end, not appearing between vowels because majority of these /j/ come from Latin /g/ (payer < pagare). T in siroter is again analogy - word final /ot/ and /op/ were both reduced to /o/, but with very few word ending with /op/, so the new derivations went with /t/.
    – Eleshar
    Mar 11, 2017 at 18:47
  • Reg. still the /j/: take e.g. the Latin volére - this would give you proto-romance /volerε/, old French /voleir/ > /vuloir/ > /vuloe**(r)/ or /vulwe**(r)/ up until final /vulwa:R/. But when this segment was followed by a vowel, the rising diphthong would obviously be perceived as ending in /j/, which would be preserved in this context even when the quality of the preceding vowels evolved dramatically.
    – Eleshar
    Mar 11, 2017 at 18:52
  • The coronal you mention at the end of verb stems is highly productive (I'm reacting mostly to your wording of non-automatic): any new verb stem that's introduced to the language has to end in a consonant (obviously inherited vocabulary lacks that) and if the original word lacks one, one is added. Either /t/ if the source word ended in an oral vowel, or /n/ (plus denasalition of the vowel) if it ended in an nasal vowel. Examples include afonner /afɔne/ from à fond /afɔ̃/, crapoter from crapaud, and a wealth of verbalised onomatopœia: gagatiser, from gaga; Mar 12, 2017 at 17:18
  • 1
    and a wealth of verbalised onomatopœia: gagatiser, from gaga; froufrouter from froufrou; cancaner /kɑ̃kane/ from cancan /kɑ̃kɑ̃/; ronronner /ʁõʁɔne/ from ronron /ʁõʁõ/, etc. Mar 12, 2017 at 17:22

This would be highly language specific.

AFAIK in English, the epenthetic R is a pollution from the syllable final R disappearing and the re-emerging in certain social strata, and appearing elsewhere first by hyper-correction, then simply by confusion (non-rhotic English would probably pronounce "Cuba" the same as "Cuber") until it evolved into full-fledged epenthetic.

In French, you have quite frequent "epenthetic" T, again for historical reasons - in reality it is T that was preserved in intervocalic environment:

il s'appelle - ille se appela(t)

s'appelle-t-il - se appelat ille

Again this may spread by analogy elsewhere.

But in general, it should be semi-consonant /j/ or /w/ that may be at first falsely perceived on the transition between a higher vowel and lower vowel:

/ae/ > /aje/ /oe/ > /oje/ or /owe/ (depending on the articulatory persistency of roundedness).


From a book titled Vernacular Universals and Language Contacts: Evidence from Varieties of English and Beyond, apparently citing this work about epenthesis:

Lombardi (2002), reviewing examples of epenthesis from a wide range of languages, has argued that glottal stops, as pharyngeals, have the least marked place of articulation for a consonant and hence are to be expected as epenthetic consonants (e.g. to break hiatus) “all things being equal” (2002: 246-7). Coronals, she argues, are “the next best choice” of epenthetic consonant (Lombardi 2002: 223) and so will appear when “for some reason constraint conflict results in glottal stop being impossible” (2002: 223). She continues: “glottal stop has the least marked place, but conflicting requirements may force the choice of slightly more marked, but still relatively unmarrked Coronal … careful analysis shows that interacting facts about position, inventory, etc. in a given language can explain why the least marked … is not chosen” (2002: 246-7). English English, as we have seen, provides some evidence of coronals being used to resolve hiatus—/r/ after non-high vowels, [n] after the indefinite article, but both contexts are now keenly adopting the glottal stop to replace these historical artefact coronals, in our London data, in Bedford, as well as further afield in South Africa and Singapore, for example. Language and sociocultural contact are wheat seem to unite the locations undergoing a shift to glottal epenthesis to break hiatus.

The use of /n/ to break hiatus appears, that I know of, not only in the indefinite article in English, but also (and coincidentally) in the Greek negative prefix a-, changing to an- when the main word begins with a vowel (hence aerobicanaerobic).

There are many examples of epenthesis in another work, though many of them do not refer to hiatus breaking. Notably, it says that /r/ is used to break hiatus in several German dialects in Bavaria.

In some Northern dialects of Italian an epenthetic consonant, either /v/ or /g/, is inserted (irregularly) before empty onsets, usually breaking a hiatus (which in turn often arised from the deletion of a medial consonant somewhere between Vulgar Latin and Romance): L. blada(m) → /bjeva/; L. pauore(m) → /pagyra/; L. uua(m) → /yga/. The medial /v/ in the names of the cities of Padova (Padua) and Genova (Genoa) is another example of this.

  • The Greek negative prefix comes from PIE syllabic /n/, which became /an/ in Proto-Greek; the /n/ then fell away before most consonants. (Cf Latin /in/ and Germanic /un/.)
    – Draconis
    Mar 12, 2017 at 3:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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