Vowel harmony and Umlaut are widely attested, whereby vowels are affected by other vowels in the sense of making them closer in features to other vowels in the same word. Is there a comparable phenomenon that leads to regular dissimilation of vowels?

This effect could work as an inverse form of vowel harmony, whereby there are two or more sets of vowels in the language, and no two (or more) vowels belonging in the same class can appear in the same word, or as a more restricted rule whereby, for example, similar vowels in consecutive syllables would tend to shift away from each other by exaggerating one feature in opposite directions.

I'm pretty sure this does not exist (the only thing remotely approaching this is the well-known tendency of onomatopoeic reduplication to alternate vowel features, such as in ding dong and tick tock), but I wanted to check.

3 Answers 3


In the realm of regular synchronic processes, vowel dissimilation is relatively uncommon (dissimilation itself is uncommon, and vowel-to-vowel dissimilation is most uncommon); however, it does exist. Examples are low-vowel dissimilation in Woleaian where /a/ becomes [e] before a in the next syllable (described in Sohn 1971), and reportedly the rule is found in Marshallese. Kera dissimilates /a/ to [ə] (or [ɨ]: sources differ). Wintu has a dissimilation where /e,o/ become [i,u] before /a/, though only lexically specified root vowels undergo this dissimilation. Finally, a fairly well-knownish (but under-documented) example of this type are dissimilative akan'e and jakan'e in SW Russian and NE Belorussian dialects: see Nesset 2002 for data and analysis.

One example of backness dissimilation is reported (Ito 1986). There is a stem-formation process in Ainu where various vowels may be added; within the subset where a high vowel is selected, the suffix is back after front vowels and front after front vowels. It is interesting that raising of low vowels is virtually the only kind of V-to-V dissimilation attested (lowering of high vowels, for example, is unknown).

There is a straightforward explanation for why harmony and disharmony are so different in their "totality". In the case of harmony, it is typically said that vowels must be the same for the spreading feature, so all vowels must be [+ATR] if one is. That is cashed out as a local requirement that a vowel must be [+ATR] if it is adjacent to a [+ATR] vowel – this iterates through the word, effecting complete spread of [+ATR] to all vowels (because each time the rule applies, a new environment is created). In the case of dissimilation, each time the rule applies, the environment is destroyed: so /pataka/ → [pateka], with no two adjacent low vowels. Hence dissimilations almost always create an alternating pattern, and you don't find a situation where "there can be only one low vowel in a word".


Sound shifts are not in general “regular”, but there are sound laws that work with a fair degree of predictability. One example is the phenomenon in Semitic known as Barth’s law. As originally formulated it dictates that in Hebrew and Aramaic

the prefix vowel of the prefix or imperfect(ive) conjugation in the base (i.e., G) stem is dependent on the theme vowel of the respective verbal base. When the theme vowel is /i/ or /u/, the prefix vowel is /a/, while when the theme vowel is /a/, the prefix vowel is /i/.


This law has been reformulated (and generalised) by Klaus Beyer to the effect that words which in proto-Semitic had the vowels a…a in the tonic and pre-tonic syllables dissimilate these to e…a in Aramaic. For example, Arabic ʼamara “he commanded” corresponds to Syriac ʼemar “he said”.

An interesting example of vowel dissimilation in French and English is discussed here: https://latin.stackexchange.com/questions/1972/does-noel-really-have-its-origin-in-latin


Yes. In Patricia Donegan's theory of vowels, contextual conditions on vowel fortitions are generally dissimilatory. (Vowel harmony and umlaut, which you mention, are lenitions rather than fortitions.)

  • Could you elaborate a bit on that, if it's not too much of a problem? (I am reading the whole thesis, but I'm not sure I'll get it.)
    – pablodf76
    Aug 28, 2017 at 13:55
  • 1
    No, I'd best not try that. I'm sure Patricia explains her theory much better than I could. However, there were some other papers of hers back in the 70s that covered subparts of her theory that might be helpful. Here's one: ling.hawaii.edu/faculty/donegan/Papers/… .
    – Greg Lee
    Aug 28, 2017 at 17:42

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.