Ikoma speakers’ perceptions of vowel quality

114 Finally, there is a loose correlation in Ikoma between voice quality and tongue root advancement. Many previous studies of African vowel systems e.g. Stewart 1967; Fulop et al. 1998; Kingston et al. 1997; Casali 2002b have noted a difference in voice quality between [+ATR] and [-ATR] vowels. [+ATR] vowels are often described as breathy, deep or hollow, whereas [-ATR] vowels are creaky or choked. This difference is sometimes evident in Ikoma, but perhaps not as clearly as in some other languages with a more typical cross-height harmony system. The [-ATR] vowels [ ] do often have a creakier quality, whereas [e o] are often more clear or breathy, but the difference is subtle and can differ from word to word and speaker to speaker. Though the height 2 vowels e o often do have a clearer quality, in some words they sound to me more like [ ], with a creakier quality. My conclusion on this point is that e o have a broad phonetic range. 40 The height three vowels tend to sound consistently creakier, but again, there are exceptions. Though there is not always a clear correlation between tongue root and voice quality in Ikoma vowels, the impressions described above do serve as additional clues to the nature of the vowel inventory. The degree 2 and 3 vowels are the pairs which are difficult to distinguish, which suggests that these vowels are at the same height. Furthermore, that the degree 1 and 2 vowels are never confused is suggestive that they are at different heights. This issue is discussed further in the following section.

3.6.2 Ikoma speakers’ perceptions of vowel quality

Concerning the question of which vowel inventory Ikoma has, helpful evidence comes from Ikoma speakers’ perception of these vowel qualities, especially when compared to 40 As discussed in §7.3, it is likely that Ikoma’s 7VM inventory developed from a historical 7VH inventory. We might speculate that a recent e o sound change could be the cause of a “broad phonetic range” of the degree 2 vowels. 115 the intuitions of mother tongue speakers of other Mara languages which have different vowel inventories. At this point, background information is needed concerning the experimental participatory method which was used for initial research into a number of Bantu languages in the Mara Region. This method, closely following the work of Kutsch Lojenga 1996, was implemented in workshops held in Mara, Tanzania during the years of 2006-2008. During that time, a series of orthography workshops involved a number of linguists from SIL Int’l and approximately seven mother tongue speakers from each of the following languages: Ikoma, Kuria, Zanaki, Simbiti, Ikizu, and Ngoreme, as well as the five-vowel languages Jita, Kwaya and Kabwa. See §7.3.1 for a brief description of the vowel systems of some of these 7V languages. The relevant part of the research method went as follows, exemplified using the process as it took place in Ikoma, since I was a first-hand participant. 41 First, approximately 1700 words were collected using a Swahili wordlist. 42 Mother tongue speakers of each language who were also fluent and literate in Swahili gave equivalent terms in their language, and they themselves wrote the words as they thought best using Swahili’s five-vowel alphabet. 43 Since Ikoma did not have an orthography at that point, writing the words based on Swahili’s orthography was a good way to get their 41 I am grateful for the help and expertise of Oliver Stegen and Danny Foster who were also involved in guiding the Ikoma team through the process of discovering the vowel phonemes. Oliver Stegen was also instrumental in working out an initial hypothesis concerning the nature of Ikoma’s vowel harmony. 42 This wordlist was a Swahili translation of the SIL Comparative African Wordlist SILCAW developed by Snider and Roberts 2004. 43 Note that Swahili is the language of wider communication in Tanzania, and it has the five-vowel system i a u, written orthographically as i e a o u. It is the language of instruction in primary schools, and many Tanzanians are literate in Swahili. Very few, however, are literate in their first language, since most of the local languages in the country do not have an established orthography or any written materials. At this point, speakers of the 7V languages were not at all aware that their language had more than five vowels. 116 impressions concerning their language based on a writing system with which they were already familiar. Because we anticipated that many of the Mara languages had seven vowels, the next step of the process was to modify the initial five-vowel transcriptions into seven- vowel transcriptions. In order to accomplish this split in a consistent manner which also directly involved Ikoma speakers, the 1700 words were printed on cards and sorted based on grammatical category, stem shape and stem vowels. For example, all nouns with a CVCV stem shape were sorted into smaller piles depending on the vowels in the stem. Therefore, there were piles of cards for CaCa, CiCi, CeCe, CiCa, CeCo, etc. Each of these piles was then read through by a group of Ikoma speakers. One person read the word on the card while the others were instructed to listen specifically for the stem vowels, comparing the vowels in each word with the others in the pile. The task was simply to decide whether or not the vowels in one word were the same or different than those in the next word. They began with the easiest pile the CaCa stems, which was not expected to split, and then progressed to the high-vowel and mid-vowel stems, since we were anticipating that some of those piles would split into two, revealing the underdifferentiated phonemes. As Ikoma speakers read through the pile written with orthographic o i.e. CoCo, speakers and researches worked together, learning to hear the difference between [CoCo] stems and [C C ] stems. The same is true for CeCe, which was split into piles of [CeCe] and [C C ]. Note that for Ikoma, the o pile was easier to split that is, the difference in vowel quality was easier to hear, both for researchers and Ikoma speakers alike than the e pile. After the initial split, the piles of 117 cards were read through several more times to check for accuracy. Later, other piles involving different vowel combinations were split as well e.g. CeCa split into [CeCa] and [C Ca]. 44 The interesting and relevant point to note is that in the different language groups which took part in this research method, different vowel piles split, and those differences seem to correlate with the language’s vowel inventory type. In languages with the Ikoma- type inventory with contrast in the mid-vowels e.g. Ikoma, Kuria, Simbiti, Ikizu the piles of e, o split into two groups each, whereas in Zanaki, which has [ATR] contrast in the high vowels, the i, u piles split. Essentially, this means that speakers of the first group of languages identify both the height 2 and height 3 vowels with the Swahili graphemes e, o, whereas Zanaki speakers regularly identify height 1 and height 2 vowels with the Swahili graphemes i, u. Furthermore, Zanaki follows predictions made by Casali’s 2003 inventory-dependent vowel harmony hypothesis, which predicts that systems with the inventory i a u will have [+ATR] dominance, as is clearly the case in Zanaki. On the other hand, in Ikoma, in which the e o piles split that is, in which native speakers naturally lumped the height 2 and height 3 vowels together as e o, the system exhibits some form of [-ATR] dominance. 45 These results are also consistent with Kutsch Lojenga’s experience with the 9V language Ngiti. As she describes in her 1996 paper on this participatory research method, Ngiti speakers also initially wrote their language using Swahili’s five-vowel orthography. She explains 44 The wordlists which resulted from these workshops were the starting point for the data used in this thesis. Many corrections, adjustments and additions have been made at various points since. 45 As is discussed in many places throughout this thesis, Ikoma exhibits evidence of both [-ATR] and [+ATR] dominance. I suggest that [-ATR] dominance could be a more recent innovation which followed an [e o] sound change. See §7.3 for more details. 118 Having learned to read in Swahili, the people had written their nine-vowel system against the grid of the Swahili five-vowel system in such a way that the [-ATR] and its [+ATR] counterpart were systematically represented by the same symbol. Therefore, [i] and [ ] had been grouped together and written with the letter i. Similarly, [u] and [ ] were both written as u. For the mid vowels, [e] and [ ] were written as e, while both [o] and [ ] were written as o. We see then that Ngiti and the Mara languages show the same tendency for Swahili-literate speakers to group [ATR] contrasts at the same height together under a single vowel symbol. For Ikoma, this is strong evidence that the height 2 and 3 vowels, which were both written as the Swahili e, o, are mid vowels. I have taken the time to outline these procedures, results and correlations for a number of reasons. First, it is helpful to document this less-than-typical research procedure, especially as it relates to understanding mother tongue speaker intuitions about their vowels systems, particularly in the Tanzanian context in which seven-vowel languages are spoken within reach of the strong influence of Swahili’s five-vowel system. Secondly, and more importantly, in this trial involving the Mara languages, as well as in Ngiti, there is a clear correlation between a language’s vowel inventory type and the vowels which speakers tend to group together. I conclude that this record of native speaker intuition is additional evidence that the height 2 vowels in Ikoma are the mid vowels e o, not the high vowels as in neighboring Zanaki.

3.7 Summary of evidence of a 7VM inventory