Robert Blust ("Patterns of sound change in the Austronesian languages". In: Philip Baldi [ed.]. 1990. _Linguistic Change and Reconstruction Methodology_. Berlin/NY: Mouton de Gruyter. 231-267) cites a number of interesting AN developments involving nasality (being a vast family with a relatively well-reconstructed history, AN is an excellent testing ground for would-be "diachronic universals"). Here are a handful of examples:
"In Palauan, a velar nasal appears before the reflex of all Proto-Malayo-Polynesian initial vowels (*anak > {ng}alek 'child', *ikan > {ng}ikel 'fish, *uRat > {ng}urd 'vein, artery'). The suggestion that these developments are due to the fossilization of old grammatical markers is superficially attractive, but since the epenthetic consonant is associated with all grammatical classes in both languages there appears to be little reason to consider it a product of anything but phonological change" (p. 242).
In a number of AN languages there is a merger of word-final *l and *n and/or of final voiced stops with the homorganic nasals (*b and *m, etc.), the result being always a nasal consonant (p. 247).
There is a curious phenomenon called "rhinoglottophilia" (Matisoff 1975), consisting in a correlation between glottal or pharyngeal activity and the position of the velum (note the characteristic nasalisation of uh-huh-like grunts). There are several examples of nasality conditioned by "laryngeals" in AN (e.g. in Rennellese). In Seimat (western Admiralty Islands) old *p and *d fell together as /h/, but may still be distinguished by the nasalisation o the following vowel, appearing only after /h/ < *d, e.g. *puaq > hua 'fruit', but *dua > hu~a 'two (in counting trees)' (p. 249). Presumably nasalisation appeared and became lexicalised at a time when the lenited reflex of *p was still /f/ or the like.
Cf. also:
----- Original Message -----
From: Piotr Gasiorowski
Sent: Tuesday, July 09, 2002 12:14 AM
Subject: Re: [phoNet] Nasals from Non-nasals

Spontaneous nasalisation and the development of nasals from non-nasals are perfectly possible, though not very common. I'll post some examples shortly.