The causative and applicative suffixes are alternating morphemes, as illustrated in the following table:
(make someone V)
(V for someone)
(V each other)
|"to look after"|
|"to look at"|
|"to work with wood"|
|"to look after"|
Comparing the causative forms in the second column to the basic forms in the first column, we see that they are identical in form except that the causative forms have -i:ʃ- or -e:ʃ-, immediately before the final -a suffix. The strings -i:ʃ- and -e:ʃ- must therefore mark the causative, since that is the difference in meaning between the two columns. There are two forms of this suffix, which we indicate with a tilde between them: -i:ʃ- ~ -e:ʃ-. This means that these are two alternants- two different forms of the same morpheme.
The two alternants are distinguished by the vowel: i: ~ e: . These are the alternating sounds, and from the fact that they are alternating we can infer that there is a restriction on the distribution of these two vowels relative to each other. There must be a context where i: can occur and e: cannot, or vice versa.
Comparing the third column to the others, we find that it differs from the other columns in having the string -iɾ- or -eɾ- immediately before the final -a suffix. This is how the words in that column differ in form from the other columns, and in meaning those words differ from those in the other columns by having the meaning "applicative (Verb for someone)". Thus -iɾ- ~ -eɾ- must be an alternating item with the meaning "applicative (Verb for someone)". The alternating sounds are i ~ e , i.e. these are the sounds that distinguish the two alternants. There must be a restriction on the distribution of these two sounds relative to each other. There must be a context where one of the two sounds can occur and the other can't.
So we have two pairs of alternating sounds i: ~ e: and i ~ e , both of which are alternations between a high front vowel and a corresponding mid front vowel. Every time the causative has the form -i:ʃ- the corresponding applicative has the form -iɾ-, and every time the causative has the form e:ʃ- the corresponding applicative has the form -eɾ- . This makes it likely that it is the same alternation in the two cases.
To determine the restriction on distribution that leads to these alternations, we first look at the contexts where each of the two morpheme alternants occur. The conditioning environment for the alternations cannot lie to the right of the alternating morphemes, since that is -a in all cases. So we look at the segment immediately to the left of the alternating morpheme, and seek a generalization that distinguishes the contexts for the i/i: alternant from the contexts for the e/e: alternant. But this leads nowhere, since the immediately preceding consonant is often the same before -i:ʃ-/-iɾ- as we find before -e:ʃ-/-eɾ-, such as tap [ɾ] in [kumiɾiɾḁ] and in [kuɾeɾeɾah] . It can't be the immediately preceding consonant that conditions the alternation.
But if we look further to the left, at the closest preceding vowel, we do find a generalization. The alternant with the high vowel-i:ʃ-/-iɾ- occurs only when the closest preceding vowel is high [i, u] or low [a] , and the alternant with the mid vowel -e:ʃ-/-eɾ- occurs only when the closest preceding vowel is mid [e, o] .
That is the restriction on the distribution of the morpheme alternants, but that must be due to a more general restriction on the distribution of the alternating sounds. It is not a complementary distribution, like that for palatalization, since it is not the case the mid vowels occur only where the closest preceding vowel is mid. The vowels [i] and [e] are in contrastive distribution, as illustrated by the following minimal pairs:
Here we see [i] occurring in exactly the same context as [e], so the two sounds cannot be in complementary distribution.
The following statement of distribution applies:
This is a neutralization distribution. Two classes of sounds are in neutralization distribution if they are in contrastive distribution, but there is a context where one class can occur (the restricted class) and the other can't (the unrestricted class). In other words, there is a contrast between the two classes, but it is neutralized in that particular context.
This pattern of distribution is also one of vowel harmony, which is any kind of restriction of distribution that requires vowels in neighboring syllables to agree in some feature or features. In this case, nonlow vowels in neighboring syllables are required to agree in the feature [high]. This pattern of height harmony is widespread among the Bantu languages. Vowel harmony is a very common class of neutralization patterns in the world, and can involve any of the features distinguishing vowels: [high], [low], [back], [round], or [ATR] (Clements 1976, 1985; Clements and Sezer 1982). There is an example of vowel harmony involving [back] and [round] in Turkish, discussed below
Vowel harmony presumably arises from the coarticulatory influence of one vowel on one in a neighboring syllable (Öhman 1967, Zsiga 1998, Majors 1997).
Clements, G. (1976). Vowel Harmony in Nonlinear Generative Phonology. Indiana University Linguistics Club, Bloomington.
Clements, G. (1985). Akan vowel harmony: A nonlinear analysis. In D. Goyvaerts (ed.) African Linguistics. John Benjamins, Amsterdam. 55-98.
Clements, G. and E. Sezer (1982). Vowel and consonant disharmony in Turkish. In H. van der Hulst and N. Smith (eds.) The Structure of Phonological Representations (Part 2). Foris, Dordrecht. 213-255.
Majors, T. (1998). The parallel role of stress in vowel harmony and V-V coarticulation. In A. Doran, T. Majors, C. Mauk, and N. Merchant Goss (eds.) Exploring the Boundaries Between Phonetics and Phonology. Texas Linguistic Forum 41. Department of Linguistcs, University of Texas, Austin. 103-116.
Öhman, S. (1966). Coarticulation in VCV utterances: Spectrographic measurements. Journal of the Acoustical Society of America 39. 151-168.
Zsiga, E. (1997) Features, gestures, and Igbo vowels: An approach to the phonology-phonetics interface. Language 73. 227-274.