Preview of the results

19 in cases of clear cross-height harmony such as that in Akan and other 9V and 10V languages. In this thesis I follow Casali 2003; 2008 and others in assuming that both [+ATR] and [-ATR] dominance are attested, and as mentioned in §1.4.2 above, as well as §7.1 below, the existence of these types of dominance is largely predictable from the inventory. The existence of height harmony as opposed to tongue root harmony is not an issue which I directly address. I recognize that there are other types of systems e.g. Ikoma’s neighbors Kuria and Simbiti which might be best explained in terms of total height harmony. Therefore, I allow for the possibility that both height and tongue root features could be involved in harmony systems.

1.5 Preview of the results

Ikoma is in many ways a typical Bantu language, with a canonical noun class system, seven vowels, high lexical similarity with surrounding Bantu languages and a number of regular sound changes in Proto-Bantu cognates which can be quickly observed from a look at common Bantu lexical items. Nominal and verbal morphology, the consonant system, and nominal lexical tone are all relatively typical. There are, however, a number of very unusual patterns, especially in the vowel harmony system, which is the topic of this thesis. Keeping track of these unusual patterns and unexpected asymmetries will undoubtedly be difficult at times, so I begin here with an overview of some of the unique features of the system so that the presentation in later chapters is easier to follow. A crucial point to highlight in the introduction is that the data and analysis presented in this thesis is admittedly unusual and complex in a number of ways and thus might cause skepticism in the minds of some readers. Essentially, I argue that Ikoma has 20 the seven vowel system i e a o u and exhibits evidence of both [-ATR] and [+ATR] markedness. The existence of both [-ATR] and [+ATR] as seemingly active values in a single language is not very usual and calls for some explanation. Furthermore, neither type of apparent dominance in Ikoma is entirely typical. Seven-vowel Bantu languages often have vowel harmony in prefixes, and in this way Ikoma is not out of the ordinary. However, the nature of Ikoma’s prefix harmony is quite strange because, while the feature [-ATR] is the trigger for prefix alternations, the [-ATR] feature itself does not spread. Instead, [-ATR] vowels in the stem trigger a vowel height change in the prefix, from mid [e o] to high [i u], which I analyze as a process of dissimilation. I describe prefix vowel dissimilation in detail in Chapter 5. Despite evidence of [-ATR] vowels triggering prefix dissimilation, vowel harmony in verbal suffixes has patterns which, at least at first glance, indicate [+ATR] markedness. 13 [+ATR] appears to spread leftward from dominant [+ATR] suffixes, but there are peculiarities here as well. In verbs, the alternation is categorical for front vowels in which [ ] alternates to [e], but only gradient for back vowels in which [ ] is somewhat raised but does not neutralize with [o]. 14 Another strange detail is that one mid-vowel suffix the subjunctive is an undergoer and a trigger of vowel alternations. It is underlyingly [+ATR] [-e] and appears to spread [+ATR] leftward to root vowels, 13 Some of the difficulties are now becoming apparent concerning how to describe a language in which both values of the feature [ATR] seem to be marked. Note that the apparent spreading of a feature such as [+ATR] might also be analyzed as delinking of [-ATR], depending on one’s assumptions concerning feature geometry. For the purpose of description, without relying on any particular analysis, I use the terms “superficial” or “apparent” [+ATR] or [-ATR] spreading to describe patterns which, at face value, look like spreading 14 Note that the decision to consider this alternation to be categorical for the front vowels is still up for debate. Though there is strong evidence that the front-vowel assimilation is neutralizing, it is perhaps not quite as clear as we might hope. Nonetheless, for a number of reasons discussed in §6.4.2, I tentatively conclude that the assimilation is categorical. The back vowel patterns are more straightforward, since the assimilation is clearly only partial. See §6.4.2 for more discussion. 21 but the suffix is realized as [-ATR] [- ] following the back vowel in the root. Furthermore, there is also an underlyingly [-ATR] mid-vowel suffix, which is the only non-low [-ATR] affix in the language. The applicative suffix is [-ATR] [- ] but is realized as [+ATR] [-e ] after [e o] roots. Therefore, there is superficial spreading of both [+ATR] and [-ATR] from verb root to suffix. Vowel harmony in verb stems and suffixes is the topic of Chapter 6. There is also apparent [+ATR] spreading in noun stems. Underlying are realized as [e o] preceding high vowels in the stem. This is somewhat different than verbs, since in nouns both front and back vowels alternate, whereas in verbs only the front vowels are affected. However, another asymmetry becomes apparent in nouns, where we see that long and are resistant to raising, even though the corresponding short vowels are targets. Note also that vowel hamony in both noun and verb stems does not interact with prefix dissimilation. Stem vowel alternations do not affect prefix vowel alternations, possibly because prefix and stem harmony are of such different types. Apparent spreading of [+ATR] in stems and suffixes is very different than the [-ATR]-triggered mid-to-high vowel alternation in prefixes. Another repeated theme of the thesis is the use of both phonetic and phonological evidence to solidify the fact that Ikoma’s degree 2 vowels are e o, not see especially Chapter 3. Also, recall from §1.4 above that 7V languages with e o as the degree 2 vowels virtually always exhibit harmony patterns showing [-ATR] dominance Casali 2003. If this is indeed Ikoma’s inventory, then the dominant feature [-ATR] is consistent with the inventory-dependent harmony hypothesis. However, apparent [+ATR] spreading in stems and suffixes confuses matters, since we do expect dominance of both 22 [ATR] values in a single language, and we do not usually find [+ATR] dominance in a language with this vowel inventory. In Chapter 7, I propose that, even though it might first seem that [+ATR] markedness and dominance is the more regular pattern, there are some facts that would not be well-explained by such an analysis. Instead, an analysis in which [-ATR] is the more generally marked or dominant feature is perhaps tenable. If this is the case, the patterns exhibiting apparent [+ATR] dominance might be explained by other means, such as dominance reversal. Therefore, Ikoma might actually conform to Casali’s 2003 inventory-dependent dominance hypothesis. Note, however, that this hypothesis can only be confirmed by working through a comprehensive treatment of these patterns within a specific theoretical framework. Unfortunately, such a task is beyond the scope of this thesis, so my proposals must remain tentative. These vowel patterns are certainly unique, and my goal is to present the data as accurately as possible, supported by acoustic phonetic data, and to note the patterns which are typical and those which are not. For areas in which my analysis is tentative e.g. in the discussion of categorical versus gradient [ATR] spreading, I aim to present and discuss all of the available evidence, which allows the reader to decide which facts are most convincing.

1.6 Vowel formant analysis methodology