Nenets is an example of a natlang where [N] and [h] are allophones:

peh-  'put'  >   penga  <subjunctive 3rd singular>
weh- 'dog'  >   wengd@ <two dogs> (nominative dual)

In fact, [n] and [h] are also allophones:

wh- 'tundra'   >  wnt@h  <dative singular> 

Essentially, you find [h] at the end of the word and [N] or [n] before a consonant. 

There's a pretty good sketch of Nenets online:


----- Original Message ----
From: Peter Bleackley <[log in to unmask]>
To: [log in to unmask]
Sent: Thursday, September 24, 2009 4:43:36 PM
Subject: [N] and [h]

The status of [N] and [h] in English is a good exmaple of why complementary distribution is not a sufficient condition to deduce allophony. However, it occured to me the other night that for a language with the right history, it would be possible for [N] and [h] to be allophones.

An ancestral [k] undergoes one of two sound changes, depending on its environment.

1) [k] >> [x] >> [h]
2) [k] >> [g] >> [N]

As long as environments 1 and 2 remain distinct, [h] and [N] will remain allophones of a single underlying phoneme. Bonus points if there are morphological processes that lead to such things as

suN <=> suha
Nom     Acc