Front vowels

Table 10 exemplifies sound correspondences involving front oral vowels in the three languages.¹⁵ Note that there is a notable degree of overlap between the correspondences, where the vowels tend to be more closed in Tîrî than in Ajië and Xârâcùù.

Table 10: Correspondences between front oral vowels

Starting from the top with set i₁₇, note that a close vowel /i/ is found in all three languages in a comparatively large set of cognates, which must illustrate a regular development from *i in the proto-language. Below in set e₃, /i/ is also found in Tîrî, whereas /e/ is found in the other two languages. Here Tîrî appears to have merged of two distinctions in the proto-language, which are distinctly retained in the other languages. This suggests the vowel reflected by set e₃ was likely *e in the proto-language as well, which must have been regularly raised to /i/ in Tîrî.

Based on this reconstruction, the two intermediary sets e₁₂ and i₃ are ambiguous, where the correspondences in Ajië suggest *e, while the correspondences in Xârâcùù suggest *i.¹⁶ Based on the reconstruction proposed above, the correspondences in Tîrî could suggest either *i or *e, which

15 Note in the following table that the set marks the identifier assigned by EDICTOR and not the reconstructed form.

For each correspondence set, the correspondence pattern is exemplified followed by how many times it occurred in the dataset. Each correspondence is exemplified with one matching cognate set in the three languages.

16 The nasality in this word is secondary, probably resulted from the loss of the initial consonant (see 4.1.2.2).

offers no further clarity about the identify of the proto-sound. The correspondence sets e₁₂ and i₃ are only find in word-final position in multisyllabic words, and may therefore be subject to a

conditioned sound change. It is therefore possible that these sets reflect a parallel raising of *e in Xârâcùù. It is on the other hand not surprising to find diverging correspondences like this in unstressed position. For now, these correspondences remain ambiguous.

Moving on to the correspondences below in set ɛ₅, close /e/ is found in Tîrî, while mid-open /ɛ/ is found in the other two languages. This pattern is likewise widely attested in the dataset, and undoubtedly illustrates a regular development from the proto-language. Following the majority principle as explained in 3.2, a mid-open vowel *ɛ can be reconstructed in reference to this set, which must have been regularly raised to /e/ in Tîrî. As a result of this change, *ɛ took the place of former *e, which had shifted to /i/.

Based on this reconstruction, another ambiguous pattern can be identified in the neighboring set e₆, in which Ajië suggests reconstructing *e, while Xârâcùù and Tîrî suggest reconstructing *ɛ.

This correspondence is comparatively quite small, and therefore likely irregular. Because /e/ and /ɛ/

are neutralized in certain environments in Ajië (see 2.3.1.1), I suspect this could have driven a mix-up of the two in some words in the language. If so, this set likely reflects *ɛ as well, on account of the regular reflexes in Xârâcùù and Tîrî, where the vowel has been sporadically raised to /e/ in Ajië.

There is therefore solid evidence that the non-close vowels were regularly raised in Tîrî, rather than independently lowered in the other two languages. At the same time, the mid vowels *e and *ɛ have evidently escaped raising in a handful of words in Tîrî, as exemplified by sets e₂ and ɛ₇ respectively. There is no clear conditioning for these marginal correspondences in Tîrî, and as such they must be deemed sporadic. This is also the only attestation of /ɛ/ in Tîrî in the dataset. It is therefore likely that this irregular reflex is representative of borrowed material in Tîrî, introduced in loans from either Ajië or Xârâcùù, which have neither undergone a raising of the same vowel. The rarity of this sound in Tîrî further supports the hypothesis that the distinction was regularly lost in the language, after which it may have been reintroduced through borrowings from the other

languages, which regularly preserve the distinction. Thus, the following chain shift can be regularly reconstructed to Tîrî, as illustrated below, in reference to sets i₁₇, e₃, and ɛ₅ respectively.

Protoform Tîrî

*i /i/

*e /e/

*ɛ

With the reconstruction of the front oral vowels in mind, we may precede to correspondences involving front nasal vowels, as exemplified in table 11 below.

Table 11: Correspondences between front nasal vowels

Starting with sets ĩ₄, ĩ₅, and ĩ₆, note that there is a close vowel in all three languages, with differing nasality. These status of the nasality on the vowels in the proto-language in these sets is highly varies from cognate to cognate. Where no nasal consonant is present from which nasality could have hypothetically spread, it must at this point be concluded that a nasal vowel was also present in the proto-language. This suggests the presence of the close nasal vowel *ĩ in the proto-language, which formed the nasal counterpart to former *i.

Moving on to the correspondences in set ɛ̃₅, all three languages show a mid vowel, but exemplify a minor difference in degree of closeness. The majority principle suggests reconstructing a mid-open vowel *ɛ̃ here, which must have been raised in Ajië. There is also a degree of overlap between the former two as exemplified in sets ĩ₁ and ẽ₁₄, where a mid-close vowel is found in Ajië, but a close vowel in the other languages. On account of the raising of *e in Tîrî, this could suggest reconstructing a mid-close vowel *ẽ instead. However, these correspondences are notably marginal, which makes it more likely that these exemplify irregular developments from either *ĩ or *ɛ̃.

It is also possible to identify a degree of overlap between mid /ẽ/, /ɛ̃/ and open /ã/, as

exemplified by sets ã₁₁ and ɛ̃₃. In most of the words in question, it is not clear if the vowel was *ɛ̃ or

*ã in the proto-language. Reconstructing a variation between the two is also not out of question, as the respective vowel qualities are quite similar, to which nasality further reduces the distinction. On a related note, de La Fontinelle (1976) also described a nasal vowel [ɛ̃] in the speech of older Ajië speakers, which younger speakers had replaced with [ã] at the time of her study.¹⁷ It is not clear if

17 This vowel is found in one word included in the dataset, /kãi/ ‘eat (starches)’, for which she recorded the archaic form /kɛ̃i/. de La Fontinelle (1976) analyzes this word as containing the transitive suffix /-i/. Thus, the fronting of the vowel in this word could be the result of contact assimilation as well.

this vowel was related to the proto-vowel I reconstruct here, or if this vowel evolved as a variant of /ã/ at a later date. For this reason, most of these cases are best regarded as ambiguous for now.