Phonology: case studies

By Scott Myers and Megan Crowhurst
Department of Linguistics at the University of Texas

The interaction between vowel deletion and palatalization

Palatalization of a consonant depends on the vowel following that consonant - palatal stops occur only before front vowels. Vowel deletion, on the other hand, affects what vowel follows a consonant. After vowel deletion is applied to C V1 # V2, C will be followed by V2, not V1. These two processes necessarily interact. Will Palatalization depend on V1 or V2?

As it turns out, whether a dorsal stop in the configuration /C V1 # V2/ is palatal or velar depends on V1, not V2, even though it is V2 that follows the stop in surface pronunciation. This is illustrated in the following examples:

Utterance-final Before a C-initial word Before a V-initial word
umugi munini
"a big town"
umugi ukomeye
"a strong town"
ubwenge bwinshi
"a lot of intelligence"
ubwenge ashaka
"intelligence that he/she is looking for"
itariki nziza
"a good date"
itariki atifuza
"a date he/she doesn't want"
amaboko maremare
"long arms"
amaboko akomeye
"strong arms"
"water buffalo"
imbogo zicaye
"sitting water buffaloes"
imbogo iryamye
"a sleeping water buffalo"

Due to palatalization, we generally find palatal stops only before front vowels, and velar stops only elsewhere. But just in this case of underlying /C V1 # V2/, we find palatal stops before back vowels, as in the first three entries in the third column of the table. We also find velar stops before front vowels, as in the last two entries in that column.

Our statement of distribution for palatal and velar stops needs to be revised so that it refers to the alternation introduced by vowel deletion:

  • A palatal stop occurs only before a front vowel, or word-finally in a word that has other alternants with a front vowel following the stop.
  • A velar stop occurs only elsewhere, or word-finally in a word that has other alternants with a back vowel following the stop.

In a derivational analysis, the necessity of referring to one alternation in the statement of the distribution for another alternation is an indication that the two rules are crucially ordered. If Palatalization is applied before Vowel deletion, it yields the correct results.

Underlying representation    /umugí ukomê:/     /imbogo iɾɟâ:mɲé/  
Palatalization    /umuɟí ukomê:/   ------------ 
Vowel deletion    /umuɟ úkomê:/    /imbog íɾɟâ:mɲé/  
Surface representation    [umuɟúkomê:]     [imbogíɾɟâ:mɲé]   

Starting with the same underlying representations, and applying the rules in the opposite order, yields the wrong outputs:

Underlying representation    /umugí ukomê:/     /imbogo iɾɟâ:mɲé/  
Vowel deletion    /umug ukomê:/     /imbog iɾɟâ:mɲé/  
Palatalization  ----------    /imboɟ íɾɟâ:mɲé/  
Surface representation    *[umugukomê:]     *[imboɟíɾɟâ:mɲé]   

Applying the rules in this order yields a simpler pattern in which palatalization just depends on the vowel that follows it in surface representation, but that is not the pattern in Kinyarwanda.

The interaction between vowel deletion and palatalization in Kinyarwanda is opaque, which means that the two restrictions on distribution cannot be stated in terms of phonological context, but must refer to the alternation class of the sounds involved. In this case, the distribution of velar vs. palatal stops must refer not only to the following vowel in surface representation, but also the following vowel in other alternants. In a derivational account, the analysis of an opaque interaction always involves a rule ordering where one rule obscures the environment of the other. In this case, vowel deletion applying after palatalization obscures the distribution of velar and palatal stops by eliminating in some cases the vowel that conditioned palatalization. This is a counterfeeding and a counterbleeding order, depending on whether the deleted vowel is back or front.

In Optimality Theory there is no rule ordering. Thus opaque interactions among phonological patterns are a special challenge for OT. A variety of strategies have been proposed to address this challenge within OT, including output-output correspondence (Benua 1997), comparative markedness (McCarthy 2003), and candidate chain theory (McCarthy 2007).


Benua, L. (1997). Transderivational Identity: Phonological Relations Between Words. PhD dissertation, University of Massachusetts, Amherst.

McCarthy, J. (2003). Comparative markedness. Theoretical Linguistics 29. 1-51.

McCarthy, J. (2007). Hidden Generalizations: Phonological Opacity in Optimality Theory. Equinox, London.

  • Two rules A and B are in a counterfeeding relation if A precedes B, and B would feed A if it applied before A.
  • Vowel deletion is in a counterfeeding relation with Palatalization in any case in which an underlying velar stop is followed by a nonfront vowel - front vowel sequence. In this case vowel deletion would delete the nonfront vowel, and if vowel deletion applied before palatalization, the deletion would create a situation fro palatalization to apply.

  • Two rules A and B are in a counterbleeding relation if A precedes B and B would bleed A if it applied before A.
  • Vowel deletion is in a counterbleeding relation with Palatalization where an underlying velar is followed by a front vowel - nonfront vowel sequence. In this case vowel deletion would get rid of the front vowel following the velar stop, which would eliminate an opportunity for Palatalization to apply, if Palatalization applied after Vowel deletion.