Phonology: case studies
By Scott Myers and Megan Crowhurst
Department of Linguistics at the University of Texas
Vowel deletion
The following examples illustrate an alternation of a vowel with nothing.
| Utterance-final | Before a consonant-initial word | Before a vowel-initial word | |||
|---|---|---|---|---|---|
| kubona "to see" |
kubona neza "to see well" |
kubona umuntu "to see the person" |
|||
| umwana "child" |
umwana akunda "a child he/she likes" |
||||
| umuntu "person" |
umuntu munini "a big person" |
umuntu akunda "a person he/she likes" |
|||
| abantu banini "big people" |
abantu akunda "people he/she likes" |
||||
| umugabo "man" |
umugabo munini "big man" |
umugabo akunda "a man he/she likes" |
|||
| igitabo kinini "big book" |
igitabo akunda "a book he/she likes" |
||||
| intebe nshya "a new chair" |
intebe akunda "a chair he/she likes" |
||||
| ingurube nzima "a live pig" |
ingurube akunda "a pig he/she likes" |
In all these cases, there is an alternation between a vowel-final form of the word that occurs at the end of an utterance or before a consonant-initial word, and a consonant-final form of the word that occurs only before a vowel-initial word in the same phrase.
| Utterance-final | Before a consonant-initial word | Before a vowel-initial word |
|---|---|---|
The alternating sounds in this case are the word-final vowels, which are alternating with ø, i.e. nothing: i ~ ø, e ~ ø, a ~ ø, o ~ ø, u ~ ø.
The fact that these vowels are alternating indicates that there is a restriction on distribution involving word-final vowels. Either there is a context where a vowel cannot occur,
or there is a context where a vowel must occur, i.e. the string without the vowel is disallowed. In this case we can observe that if the vowel-final alternant occurred before a vowel-initial word,
there would be a sequence of vowels. We do not in fact see any such sequences in our data: a vowel never occurs before a vowel in another word. We can state this restriction succinctly as follows:
*V#V.
Here the asterix indicates that the configuration that follows is not allowed. The hash mark "#" indicates the edge of a word. This is not a complementary or a neutralization distribution. Those are distributions involving two classes of sounds. Alternations induced by complementary or neutralization distributions are alternations between two classes of sounds. One class occurs in a particular context and the other class doesn't. In an alternation with ø, on the other hand, we can't state the distribution of ø, since ø has no distribution. We can just state the sequence that does not occur.
The generalization we have made says that there is a context (before another vowel) where a vowel cannot occur. We should also consider the alternative: that there is a context where a vowel must occur. If the consonant-final alternant occurred at the end of an utterance, there would be a consonant-final utterance. This doesn't occur in our dataset. The problem with this line, though, is that there is a range of vowels that alternate with zero, and we can't predict from context which vowel belongs with which morpheme. Thus the true distributional statement that no utterance ends in a consonant does not lead to a viable analysis.
Derivational analysis
OT analysis