En réponse à Nik Taylor : >But even without minimal pairs, if you can find two sounds in the same >*environment* that would be enough to call them phonemes. E.g., if I >understand the rules for [O]-[o] correctly, [o] is used in open >syllables, [O] in closed. (My apologies if I'm mistaken) As I've said in an earlier post (but I sent it two minutes ago, so you couldn't know), I know of at least one minimal pair: "Maud(e)"/"mode". So you're indeed mistaken. Same for /2/-/9/. Although  appears normally only in open syllables and  in closed ones, I know a few examples of  in closed syllables, ruining the claims of those who tried to convince me for years that  and  were a single phoneme ;)))) . That's what I mean when I say that the native speaker's intuition should be valued more than it is now. I've never met a speaker of any language recognise a distinction which is not truly phonemic in his language, unless pointed specifically at it. If the speaker is able to hear the distinction without having been pointed at it, then it'll be extremely probably a phonemic distinction. > If you could >find a word with  in an open syllable, or [o] in a closed syllable, >that would proof of phonemicity. I have :) . >[h]-[N] is a non-argument because they're radically different sounds. I thought that this was no argument when we were talking about allophones :))) . >Personally, I don't see any reason not to have more than one definition >of phoneme, such that under one definition, [o] and [O] would be >analyzed as one phoneme in French, and under another as two. But if you cannot get a unitary definition of the phoneme, what's the point of even trying to define it? Christophe Grandsire. http://rainbow.conlang.free.fr You need a straight mind to invent a twisted conlang.