3.28 Nasal-Assimilation P-27.

The velar nasal /n^g/ normally assimilates to the point of articulation of a following stop; exceptions are largely limited to artificially slow speech. Thus /gulmun^g/ 'belly' has case forms like Locative /ama-gulmun-duj/ Pergressive /ama-gulmum-baj/ and Relative /ama-gulmun^y-jin^yun^g/ In the case of following /j/ we sometimes get /n/ instead of fully assimilated /n^y/ hence a variant Relative /ama-gulmun-jin^yun^g/.

A few prefixes and compound initials end in a nasal which behaves like /n^g/ Since it occurs only in environments where it acquires surface point of articulation by assimilation, we represent it conservatively as an archiphoneme /N/ though setting it up more concretely as //n^g// would produce the right outputs. An ex. is Inverse //-N-// a morpheme occurring within pronominal prefixes, as in ANA→1Sg_a prefix //n^ga-N-w₁u- // → /n^gan^ggu-/.

Our reluctance to set this up specifically as //n^g// is due to the fact that in some prefixes what appears to be final //n// also undergoes point-of-articulation assimilation. Thus what we call the B morpheme in pronominal prefixes, representable as /-w₂an-/ under a maximally abstract analysis, does show assimilation. For example, 1Sg_b //n^ga-w₂an-// appears as //n^gan-// and undergoes assimilation in combinations like /n^gam=bann^ga-na/ 'I will dance' and /n^gan^g-ga(u>l=n^gu-yi:/ 'I will drink'. (The evidence for transcribing the underlying form with /n/ rather than /n^g/ or archiphoneme /N/ is indirect but is related to the rule for Pronominal d-Insertion P-20.)

In other morphemes, final /n/ is stable on the surface, avoiding assimilation. Thus /man-/ 'group', a derivational prefix, occurs in combinations like /-man=bayama-/ '(group) to keep going', and /da:n/ 'guts' in its compound-initial form /-dan-/ retains its /n/ in /-dan-gara=gayi-'-/ 'to have a bellyache'. The same pattern is seen at the boundary between a noun and following case suffix, as in /a-mu:n-baj/ 'by foot'.

It appears, then, that the rule applies to /n^g/ and /n/ as a regular process, but to /n/ only in pronominal prefixes in which B morpheme //-w₂an-// is involved. The rule does not apply to /n^y/ which retains its laminoalveolar point of articulation with a following stop. The same is true of /n/ There is no evidence one way or the other for /m/ which does not occur in any relevant underlying combination.

(P-27)	Nasal-Assimilation
a.		ng		→ [αF] // ____-	Stop [αF]
		N
b.		n		→ [αF] // -w2a____-	Stop [αF]

Actually, the rule is slightly oversimplified as stated. We have mentioned that, in part (a) of the rule, we can get /nj/ as a variant of /n^yj/ when /j/ follows the nasal. In addition, when the stop is /dh/ the output is what we write as /ndh/ rather than /nhdh/ There is no opposition between /ndh/ and /nhdh/ since the latter does not occur, consistent with the general principle that interdentals do not occur syllable-finally, but the sound we hear in /ndh/ is /n/ rather than /nh/.