On Thu, 10 Jun 2010 07:55:24 -0700, Jesse Bangs <[log in to unmask]> wrote:

>Does anybody know of a language whose orthography consistently
>distinguishes between two non-contrastive allophones? A hypothetical
>example would be a world in which English orthography was designed by
>colonizers from India, who assigned different Devanagari symbols to
>aspirated and unaspirated voiceless consonants in English, despite the
>fact that those sounds don't contrast and have a completely
>complimentary distribution.

Well, Devanagari (for Sanskrit) itself has some examples, no?  AFAIK the
nasals [J] and [N] didn't have phonemic standing, and visarga and anusvara
don't either.  

The Mongolian script writes the front-vocalic [k g] and back-vocalic [q G]
allophones of /k g/ differently -- in fact it collapses the former two, so
the characters are [k g], [q], [G]!

In conlangs, I remember having this discussion with David P before.  I seem
to remember that some of his scripts do this... yes, for instance, Zhyler does:
where [k x] (of /k/) and [g G] (of /g/) and several nasals (of /n/) get
distinct representations.  
(Most -- all? -- of David's romanisations do this too.  But I don't know if
that counts.)