Perceptual motivation Explanation of height-changing alternations

171 As the above examples show, the same set of constraints predicts both the Bantu C pattern and the Ikoma pattern. And the same ranking of constraints predicts the correct height alternations for all four types of stem vowels high, mid [+ATR], mid [-ATR] and low. A summary of three different rankings of these constraints and their outcomes is in 143 below. 143 Typological predictions for underlying form ko- k Crucial Constraint Rankings Outcome Example a. D EP -ATR, A GREE -HI [ATR], M AX - STEM -ATR [-ATR] M AX -ATR, A GREE [ATR], M AX -H I prefix height alternation Ikoma ku- k b. A GREE -HI [ATR], M AX -H I , M AX - STEM -ATR M AX -ATR, [-ATR], A GREE [ATR], D EP -ATR prefix ATR alternation some Bantu C languages k - k c. M AX -H I , M AX -ATR, M AX - STEM -ATR, D EP -ATR A GREE -HI [ATR], [-ATR], A GREE [ATR] no harmony some Bantu C languages ko- k Ranking a works for Ikoma, and b for the Bantu C pattern. Note that ranking c was not discussed above. It results in a language with no prefix harmony at all, which is actually the case for a number of other Bantu C languages Leitch 1997.

5.4.2 Perceptual motivation

Finally, I return to the perceptual motivation. If [-ATR] spreading not dissimilation is a more cross-linguisticly common repair strategy, we might look to other factors to inform the question of why the height alternation is preferred in this case. Concerning the perceptual question, Stewart 1967 195-6 refers to a quote from Westermann and Ward 1933:129 who posit a possible perceptual explanation for harmony in Igbo: “the vowels are extremely difficult to differentiate from each other, i.e. i and , u and and o are respectively very near to each other acoustically and the alternating prefixes help to 172 distinguish the root vowels.” 68 Perhaps Ikoma is simply a more extreme version of the use of prefix vowel alternations to aide in distinguishing acoustically close root vowels. The prefix alternations from mid to high vowels function to signal the quality of the stem vowels which, as discussed in §3.6, can also be surprisingly difficult to distinguish, for both linguists and mother tongue speakers alike. A prefix alternation between [o] and [u] is a more significant jump from the more-expected alternation from [o] to [ ]. As briefly mentioned in §5.1.1 above, the mid-vowel prefixes can sometimes sound quite high. During fieldwork on Ikoma it was not uncommon for me to ask for clarification on the quality of a prefix vowel, since I occassionally heard [e o] as [i u], especially when the prefixes were pronounced very quickly. Native speakers, however, are quick to distinguish the prefixes. They associate the prefixes with Swahili’s graphemes i e o u which represent Swahili’s vowels [i u] respectively. In fact, when writing Ikoma with a Swahili-type orthography, every native Ikoma speaker I have worked with has instinctively and consistently written i u for the high prefix vowels and e o for the mid prefix vowels. These same speakers usually have a difficult time learning to distinguish the mid-vowel contrasts, but distinguishing prefix vowel alternations are very intuitive. Since speakers are much more aware of the difference between the mid and high prefix vowels than they are between the [+ATR] and [-ATR] mid vowels, perhaps this is the motivation for the alternations. If it is difficult for speakers to hear the [ATR] distinction between e and o , the prefix alternations based on height would assist listeners in maintaining contrast between the mid vowels. Of course it is difficult to be sure about the motivation for such an interesting dissimilation pattern, but it is possible 68 Kaun 1995 also suggests that vowel harmony is motivated by perceptual factors. 173 that the perceptual motivation has some role to play. It is possible that perceptual and theoretical motivations such as the OT analysis presented above have conspired together to cause the height-alternating prefixes. 5.5 Summary and implications for [ATR] markedness