Ordinarily, /w2/ at the beginning of a stem resists deletion. However, there are two kin-term roots which show alternations of the type /-bV/ after a stop vs. /-V/ after a vowel or in word-initial position. We handle these by posit a basic form /-w2V/ with the usual /w2/ → /b/ after stop by Hardening P-18. This requires an idiosyncratic deletion of the /w2/ for these stems in non-Hardening environments.
The more common of the two roots is /-w2ibi/ 'Mo' (Table 5-1, Bii); the base form just shown is consistent with Pl forms /mij-bibi/ 'your mothers' and /mij-bibi-yung/ 'his/her/their mothers'. With FSg prefix /ngara-/ however, we find /ngar-ibi/ 'your mother' and /ngar-ibi-yung/ 'his/her/their mother'. The reduction of /ngara-/ to /ngar-/ is regular before a morpheme beginning in /i/ (but not, say, /yi/ or /wi/ , see V-Truncation P-46. We therefore need a rule converting /-w2ibi/ into /-ibi/ before V-Truncation applies.
With no noun-class prefix, /-w2ibi/ takes the form /yibi/ or /ribi/ 'your mother' (add suffix /-yung/ for 3rd person form). The variant /yibi/ is compatible with an intermediate form /ibi/ with the /y/ added as a low-level phenomenon by Initial Semivowel-Insertion P-5. However, /ribi/ represents an optional restructuring of the basic form of the root (presumably by resegmentation of /ngar-ibi/ as /nga-ribi/ influenced also by the 1Sg form /rigang/ 'my/our mother'). Speakers who consistently use /ribi/ instead of /yibi/ for 'your mother' (with noun-class prefix absent) require recognition of two distinct basic forms for '(your) mother', one /-w2ibi/ to account for /mij-bibi/ and another /ribi/ for the word-initial case; the simplest way to analyse /ngaribi/ (i.e., /ribi/ with noun-class prefix present) is as /nga-ribi/ (for /nga-/ surface allomorph of FSg /ngara-/ see discussion of Feminine Prefix-Truncation Ρ-34; this latter rule could be generalized to take care of /nga-ribi/ if needed).
A somewhat similar alternation occurs with /-w2awan-nyij/ 'father and child' (with dyadic suffix), as seen in /mij-bawan-nyij/ 'father and children' (Pl /mij-/ inducing Hardening P-18 of /w2/ to /b/ . Compare prefixless /awan-nyij/ and prefixed /na:-'wan-nyij/ 'father and child' (the prefix is M /na-/ , pointing to a form without the initial /w2/ (See Table 5-1> Aii).
We should caution that all of these alternations are rather frozen, and the derivations suggested highly opaque. In the case of 'father and child', there is an alternative representation with /w1/ instead of /w2/ hence Pl /mij-gawan-nyij/ though the other form is still /awan-nyij/ with no prefix. (There are a few other stems with fluctuating initial /w1/ or /w2/ ) For speakers with /w1/ no special rule (P-12, below) is needed, since //w1awan-nyij// will become /awan-nyij/ anyway by w1-Deletion P-9.
In any event, the special rule P-12 is needed for /w2ibi/ and for most speakers also for /w2awan-nyij/ It is difficult to identify the precise phonological context for the rule, since there are so few relevant forms. The exx. given above show that the deletion applies word-initially and after /a/ but does not apply after /j/ The compound /lhal-ibi-yung/ '(clan) country of his mother' shows deletion after /1/ Derivatives like /ma-bibi-yung/ 'its (MANA class) mother', base form //maG-w2ibi-yung’// show nondeletion after stop archiphoneme /G/ The evidence is incomplete, but it looks as though /w2/ is retained after stop or nasal, where Hardening P-18 will convert /w2/ to /b/ and it is deleted in non-Hardening environments. If we just order P-18 before the deletion rule, no special phonological restrictions are needed.
The rule applies to the only two kin terms beginning with /w2/ so it is reasonable to specify this in the rule.
| (P-12) | Stem w2-Deletion |
| w2 → Ø // -___ | |
| at the beginning of a kin-term noun |