Velar nasal /ng/ and nasal archiphoneme /N/ (which can be regarded as /ng/ are deleted before another nasal consonant. Pronominal prefixes ending in the B morpheme /-w2an-/ where the evidence for positing underlying alveolar //n// is somewhat indirect and tenuous, show the same treatment before a nasal. There is also reason to believe that these nasals are deleted before (surface) liquids, though not all relevant underlying combinations can be cited in the data.
Exx. involving /ng/ are /wulang/ 'blood' in case form /wula-miri/ 'by means of blood', and some compounds of /=na-/ 'to see' like /-dhalama=na-/ 'to see broken-off branch' with /-dhamalang-/ There are no consistent counterexx. at boundaries, though in artificially slow and careful speech the /ng/ is sometimes heard.
Inverse morpheme /-N-/ a component of many transitive pronominal prefixes, is deleted before other nasals; the available exx. involve following /m/ /ng/ /n/ Thus MANA→ 1Sga //nga-N-ma-// → /ngama-/ with no surface effect of the //-N-// morpheme, so this is homophonous with 1Sg→ MANAa //nga-ma-// → /ngama-/.
Pronominal prefixes ending in B morpheme //-w2an-// behave like morphemes with final //ng// or archiphoneme //N// in connectionwith Nasal-Assimilation P-27, and also behave like morphemes ending in these nasals in the present deletion rule. Thus //ngan=ngu-yi:// → /nga=ngu-yi:/ 'I will eat it', where //ngan-// itself is analysable as //nga-w2an-// with the B morpheme on a more abstract analysis. However, in other cases morpheme-final /n/ remains as such before another nasal, as in /ra:n-miri/ 'by means of a vine sp.' and in compound /-mun=ngalngala-/ 'to have foot shining'.
There are not a great many attested combinations of /ng/ /N/ or /n/ before a (surface) liquid. This is because a following /lh/ /r/ or (in most cases) /r/ will undergo Hardening P-18 after a nasal, so we end up with a nasal-stop sequence. However, there are a few combinations in which /r/ escapes Hardening, and there are also some combinations with following /1/ or /1/ the two liquids which have no Hardening counterpart. In the available combinations, the preceding /ng/ /N/ or /n/ is deleted, at least in the most common pronunciation. Indeed, deletion of the nasal in this environment may apply even more generally than before another nasal, since /-man-/ (a Pl derivational prefix or compound initial) normally keeps its /n/ before another nasal but seems to drop it in the usual pronunciation of compound /-ma(n)=lu:lha-/ 'to wade as a group'. An ex. with /ng/ is /-ya=lalaga-/ 'to make (it) louder' with compound initial related to /ya:ng/ 'voice, sound'.
As noted above, ordinarily morpheme-final /n/ is not deleted before another nasal. Likewise, /ny/ is often retained before another nasal; for a more restricted deletion rule in some preconsonantal environments see Palatal-Deletion P-26. Nasals /nh/ and /m/ do not occur in any relevant underlying combination, so we have no evidence for or against their deletability before another nasal. The remaining nasal, retroflexed /n/ is stable before other nasals.
The following is an attempt to summarise the facts.
| (P-30) | Nasal-Deletion | |||||
| a. | Nasal | → Ø // ___- Lateral | [i.e., /1 1 r/] | |||
| b. | | ng | | → Ø // ___- Nasal | ||
| N | ||||||
| c. | n | → Ø // -w2a___- Nasal | ||||
| where /-w2an-/ is the B morpheme in a pronominal prefix | ||||||
For some of the combinations, if not all, we could also consider an analysis in which the nasal is assimilated in all features to the following nasal or lateral, and then deleted by Geminate-Contraction P-31, to which we now turn.