Ordinarily, interdental /lh/ is a stable phoneme in the positions where it is permitted, except that it becomes /dh/ after stop or nasal by Hardening P-18. However, in one verb paradigm it appears to alternate with /y/ in some forms. This is /=bilya-/ 'to be tilted'. In the Past2, I recorded /=bilya-y/ and /=billhi-Ø/, where in addition to the alternation of /y/ and /lh/ we also have an alternation between /ay/ and /i/. Here it would appear that the /i/ variant involves optional application of V-Fronting P-50 to //iy//, then shortened to /i/ syllable- and word-finally by iy-Contraction P-15.
While the alternation of /ay/ with /i/ is not unusual, that of /y/ and /lh/ is. For speakers who have the forms as just listed, I incline to posit underlying //lh// becoming /y/ when followed (on the surface) by /a/, but kept as /lh/ when followed by /i/ (for the full paradigm see Table 11-5) NgA2 paradigm).
| (P-32) lh-Delateralisation |
| 1h → y // __a |
| in the stem /-billha-/ 'to be tilted' |
It is also possible to posit underlying //y// becoming /lh/ before /i/. In neither formulation would the rule be generalisable to other forms in the language. Elsewhere /llha/ and /lyi/ are acceptable sequences, as in /=w2allhi-Ø/, Nonpast3 of /=w2allha-/ 'to float', and /a-ngajal-(y)inyung/ 'of the (water) spring'.
It is reasonable to think that /=billha-/ has been affected historically by analogies from the two other verbs with identical or near-identical suffixal paradigms, /=lha-/ 'to stand' and /=yi-/ 'to sleep'. It is possible that /=billha-/ is historically a compound of one, but has been influenced by the other as well. For some speakers the stem is /=bil-yi-/ with exactly the same paradigm as /=yi-/ 'to sleep' and can thus be considered to be a synchronic auxiliary compound of it.