Some irregular Spanish verbs with infinite in "-ir" seem to have an interesting pattern in their conjugation:

For some verbs with "o" as last vowel in the infinite stem (e.g. dormir, morir), the form of the last stem vowel in conjugated forms seem to follow the following pattern (examples are in italics):

stressed: ue (duermo, duerma)
unstressed, next vowel/diphthong i: o (dormimos, dormía)
unstressed, next vowel/diphthong ie/a/ió: u (durmieron, durmamos, durmió)

Similarly for some verbs with last vowel in the infinite stem "e" (e.g. mentir, sentir):

stressed: ie (siento, sienta)
unstressed, next vowel/diphthong i: e (sentimos, sentía)
unstressed, next vowel/diphthong ie/a/ió: i (sintieron, sintamos, sintió)

When looking as the unstressed vowel patterns in particular, this reminds me of vowel harmony.


  1. Does this qualify as vowel harmony at all?
  2. If so, where does this phenomenon come from in Spanish? I usually relate vowel harmony to Uralic/Turkic languages, but not IE/romance. Does a similar phenomenon appear in other IE/romance langauges?

5 Answers 5


Why yes, Romance languages do have vowel harmony, or what the Romance linguistics literature likes to call metaphony. The following information on Spanish is all from Alkire & Rosen 2010, which is my only source of knowledge about Spanish phonology.

Alkire & Rosen 2010 devote subchapter 5.2 to "yod effects in Spanish". The effect of raising of a stressed vowel under the influence of a following high vowel gesture (anticipation) seems to be especially prominent in the verb system, where it has been largely morphologized. But you find it in nouns as well, so you have:

VĬTREU   'glassy'    vidrio   'glass'

SĒPIA   'cuttlefish'   jibia

PLŬVIA   'rain'    lluvia

The stressed vowels in Latin should have given Romance high mids, which in Spanish should go on to give you mids: [e] and [o]. Insead, what you you have is [i] and [u], evidence for anticipatory raising. Alkire & Rosen give many more examples.

Yod effects can also be found in Italian, but they're more restricted.

PŬGNU   'fist'     pugno

FAMILĬA   'family'  famiglia

CONSĬLIU 'counsel' consiglio

Same story here as for Spanish, except it happens before [ɲ] and [ʎ]. There are other examples as well, including examples where the expected raising fails to happen.

French has a lot of yod-related effects too. See ch. 5.3 of Alikre & Rosen 2010.

Maybe someone else can fill us in on Portuguese. :)

Romanian has this stuff too, but it involves stressed low vowels alternating with mids before an [j] in the inflection (a-ə, e̯a-e, o̯a-o; note that the diphthongs are considered phonemic) and before an [e] in the plural marker (only e̯a-e). This correlates with gender in nouns and person in verbs, so it's also morphologized to a large extent. See Chitoran 2002 (abstract).


[mare]   'sea'     [mərj]  'seas'

[se̯arə]   'evening'   [serj]   'evenings'

[bo̯alə]   'disease'   [bolj]   'diseases'


[aratə]    'show.3rd.sg'   [arəʦj]   'show.2nd.sg'

[vise̯azə]   'dream.3rd.sg' [visezj]   'dream.2nd.sg'

[do̯arme]   'sleep.3rd.sg'   [dormj]   'sleep.2nd.sg'

  • 1
    Quite a detailed answer! :)
    – Alenanno
    Oct 25, 2012 at 10:23
  • 1
    Very cool stuff, thanks! I definitely need to spend more time to go through the whole answer, but just one thought. "Raising of a stressed vowel under the influence of a following high vowel" isn't what's happening in my examples, is it? It looks more like "raising of an unstressed vowel under the influence of a following mid/low vowel".
    – dainichi
    Oct 25, 2012 at 10:29
  • @dainichi I got so excited about vowel harmony that I completely failed to read your examples. Some of those stem alternations in Spanish verbs are regular changes from Latin. Stressed low mids diphthongized in Spanish, giving you siento and duermo, vs. sentir and dormir. But your third forms, the preterites, do have metaphonic raising of unstressed vowels from [j] of -io and -ieron (only in -ir verbs). So you're right that this is for unstressed vowels, but I think it's only in verbs. Other yod-induced raising seems to be under stress, but sb correct me if that's not the case.
    – lapropriu
    Oct 25, 2012 at 11:40
  • @lapropriu, actually, it's not only the preterites, but also the present subjunctive (sintamos, sintáis, durmamos, durmáis) where no [j] is following.
    – dainichi
    Oct 26, 2012 at 0:30
  • 2
    @dainichi Romance metaphony is not a living, purely phonological process, so you won't always clearly see the conditioning context. There is evidence that it was once active in the language, but it might not be active any more. Additionally, there's the question of where morphology fits into the life-span of a phonological process, with stuff like paradigm leveling, morphologization etc.
    – lapropriu
    Oct 26, 2012 at 17:37

While Lapropriu gives evidence for a sort of vowel harmony, diphthongizing "stem-changing" verbs are not an example of vowel harmony.

This is not restricted to -ir-conjugation verbs; other common ones are sentar (me siento 'I sit'; nos sentamos 'we sit') and perder (pierdo 'I lose'; perdemos 'we lose'). And there's nothing special about verbs; the same phenomenon occurs with stress alternation in nouns and adjectives. Example: /be.ne.ˈθwe.la/ Venezuela, /be.ne.θo.ˈla.no/ venezolano. (Aside: This process is no longer productive; more recent forms do not alternate: Puerto Rico /pwer.to.ˈrˑi.ko/ corresponds to puertorriqueño /pwer.to.rˑi.ˈke.ɲo/, not */porto/-.)

Generally, the alternating diphthongizations are derived from Latin short /e/ and /o/. In Spanish, these vowels regularly become /je/ and /we/ in stressed syllables, and /e/ and /o/ in unstressed syllables. This accounts for the alternation between most such verb forms, which differ in stress. It occurs regardless of the other vowels in the word, so it is not vowel harmony. (Non-alternating /e/ and /o/ come from Latin /eː/ or /i/, resp. /oː/ or /u/.)

Demonstration: /e/ in contar diphthongizes in stressed syllables, when followed by a syllable starting with any of the possible vowels for the -ar conjugation: * /e/: present subjunctive 3sg -> cuente * /a/: present indicative 2sg (tú) -> cuentas * /o/: present indicative 1sg -> cuento

but never diphthongizes in unstressed syllables: * /e/: past perfective indicative 1sg -> conté * /a/: present indicative 2sg (vos) -> contás * /o/: past perfective indicative 3sg -> contó

More information: historical sound changes; conjugations.

You could argue that the raising mutations in ir-conjugation verbs are like vowel harmony, since they only occur in that conjugation, and depend somewhat on the vowel in the following syllable. But there doesn't seem to be any tendency to regularly place the same vowels in multiple classes, which would be needed for it to be vowel harmony.

  • Thanks for the writeup, but the diphthongization is not the part I'm curious about as much as the change in unstressed vowels. I will make that clearer.
    – dainichi
    Oct 27, 2012 at 12:45
  • Do you have any support for your statement that it has to be valid for multiple classes to be vowel harmony? By classes, what exactly do you mean? Even languages with very strong vowel harmony have some exceptions AFAIK, so I don't see how a language usually without vowel harmony can't have vowel harmony in some specific cases.
    – dainichi
    Oct 27, 2012 at 13:04

1. No, this is not vowel harmony, but vowel raising.

Vowel raising

Vowel raising appears only in verbs of the third conjugation (-ir verbs), and in this group it affects dormir, morir, podrir (alternative of the more common pudrir) and nearly all verbs which have -e- as their last stem vowel (e.g. sentir, repetir); exceptions include cernir, discernir and concernir (all three diphthongizing).

vowel raising... affect[s] -e- or -o- in the last (or only) syllable of a verb stem... Vowel raising changes the mid vowels -e- and -o- to the corresponding high vowels: -i- and -u- respectively.

The reason cernir (and the verbs based on it: discernir, concernir) and hendir are exceptions to this is that they are alternative forms of the -er verbs cerner and hender, and as such still act (in this regard) like their original forms when conjugated.

2. As for where this phenomenon came from, in the Vulgar Latin which gave rise to modern Spanish, this tended to occur in certain vowels that occurred near a semi-vowel [j]:

Note that it is the [j] in this sequence that accounts for the raising of the root vowel throughout the imperfect subjunctive in verbs like dormir and pedir.

Vowel Raising

Under the influence of a nearby [j] – which has a very high tongue position – the vowels

/a, ε, ɔ, e, o/

were in many cases raised in stressed syllables in Vulgar Latin to

/e, e, o, i, u/


This series of sound changes can be regarded as an instance of metaphony, a process that leads to the approximation of one vowel’s quality to that of another.

e.g. in dormir the affected forms are those where a neighbouring -i- represents the semi-vowel [j] (as part of a diphthong). In the unaffected forms (if they have a neighbouring -i-) this -i- is a pure vowel [i].

  • durmieron duɾˈmjeɾõn; durm duɾˈmjo
  • dormimos doɾˈmimos; dormía doɾˈmia

Note: This does not readily explain the vowel raising that occurs in forms such as durmamos.

  • What definition of vowel harmony are you working with where raising harmony is not harmony?
    – user6726
    Dec 18, 2017 at 15:59
  • Vowel Raising can be described as a type of metaphony. Vowel harmony is sometimes used synonymously with metaphony. Usually, however, "vowel harmony" refers specifically to a synchronic process operating in a particular language, normally requiring all vowels in a word to agree in a particular feature (e.g. vowel height or vowel backness).
    – iacobo
    Dec 18, 2017 at 17:52
  • From the perspective of use in the linguistic literature, that's plainly inaccurate: "all vowels agree" is based on a misunderstanding of Turkish, Finnish and Hungarian.
    – user6726
    Dec 18, 2017 at 18:30
  • The part about "requiring all vowels in a word to agree in a particular feature" simply does not jibe with actual use in the scientific linguistic literature.
    – user6726
    Dec 18, 2017 at 21:24
  • If that's not your rationale for excluding this as a case of VH, I don;tstand why you say it's not VH.
    – user6726
    Dec 18, 2017 at 21:50

There's also a completely separate vowel harmony process in some Spanish dialects that have an alternation between high and low vowel allophones. An open allophone at the end of a word triggers using the open allophone in preceding syllables' vowels.

Wikipedia gives some examples:

In Eastern Andalusian and Murcian Spanish, word-final /s/, /θ/ and /x/ (phonetically [h]) regularly weaken, and the preceding vowel is lowered and lengthened:

/is/[i̞ː] e.g. mis[mi̞ː] ('my' pl)
/es/[ɛː] e.g. mes [mɛː] ('month')
/as/[æ̞ː] e.g. más [mæ̞ː] ('plus')
/os/[ɔː] e.g. tos   [tɔː] ('cough')
/us/[u̞ː] e.g. tus   [tu̞ː] ('your' pl)

A subsequent process of vowel harmony takes place so that

  • lejos ('far') is [ˈlɛxɔ],
  • tenéis ('you all have') is [tɛˈnɛi] and
  • tréboles ('clovers') is [ˈtɾɛβɔlɛ] or [ˈtɾɛβo̞lɛ].

They give a citation:

  • Lloret, Maria-Rosa (2007), "On the Nature of Vowel Harmony: Spreading with a Purpose", in Bisetto, Antonietta; Barbieri, Francesco, Proceedings of the XXXIII Incontro di Grammatica Generativa, pp. 15–35.
  • Why another answer?
    – Alenanno
    Jan 25, 2013 at 11:13
  • 1
    @Alenanno: Because it's a separate unrelated process. Feb 6, 2013 at 7:47
  • Oh I see, but there was no need to link me that. :P
    – Alenanno
    Feb 6, 2013 at 10:19

If you are looking for vowel harmony, as someone mentioned, it is in Andalusian varieties where you can find it, even with morphological content (for instance, the contrast between horroroso "awful" in singular, with closed o vowels and "horroroso" in plurar with open harmonized vowels. Since final -s is deleted in those varieties, harmony has become, in the case of o and e, the way to mark plurality.

If you are looking for metaphony (in synchronic productive terms, not as a factor in the history of the Romance languages), you should then go to Asturian (which is not a dialect of Spanish, but a system of its own, one of the primary Latin dialects in terms of Menéndez Pidal, which, by the way, suffers discrimination in Spain, as Aragonese does, when compared with Galician (but, again, all fascist rulers came from Galicia and built a mafia-system that pervades the whole economy of the Spanish language instruction, inside and outside the country).

In Asturian, instead of perro for dog you have, after a medial stage perru > pirru.

The only other Romance languages which something similar are some Italian varieties, I think Apulian was one of them, but I am talking by heart.......

By the way, what the User Mechanical snail wrote is accurate and should suffice as basic explanation for the e / ie and o / ue alternations, which are stress-dependent in the history of the Romance languages......Nothing to do with vowel harmony.

  • Would you please change "all" for "many" or whatnot?
    – 25254
    Sep 11, 2019 at 23:50

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.