And Rosta wabbe:

> Sorry. Suppose (for example) a language has A E I O U in stressed
> but only A I U  in unstressed sylls. A standard phonemeic analysis
> recognize 5 phonemes /a e i o u/ and state phonotactic/prosodic
> constraints that exclude /e o/ from unstressed sylls.
> But this misses the fact that there are 2 different sets of
> one for stressed sylls and one for unstressed sylls, and there is no
> a priori reason to identify the "/a/" of stressed sylls (which
> with 4 other vowels) with the "/a/" of unstressed sylls (which
> with 2 other vowels). Accordingly, the following 8 phonemes should
> recognized [in a move that radically departs from the practice of
> phonemic theory]: /'a 'e 'i 'o 'u a i u/.
> Not, of course, that I think the 8 phoneme analysis is satisfactory.
> But it is better than the orthodox 5 phoneme analysis.
> This is of course just one example. But it's not exotic. --In
> the contrasts in stressable and unstressable syllables are
> and the contrasts in onsets and codas are different.
> FWIW, I would analyse the above system along the following lines:
> * Primitives of segmental content are A, I, U.
> * E and O are made by simultaneous A+I and A+U.
> * The ability of A to combine with I/U is a property only of
>   stressed syllables. ["Tier separation"]

Well, as a non-linguist I could analize this as
  * Primitives: /a/, /e/, /i/, /o/, /u/
  * In unstressed syllables there are neutralization and /i/ and /e/
will both be realized as [i] and /o/ and /u/ as [u].  Wouldn't this be
a correct approximation.

In my conlags I have different contrast for stressed and unstressed
vowels.  In Chleweyish both /a/ and /e/ become [@] in unstressed
position... but this is something that is mostly seen from orthography
(except for monosyllabics) so it can be discuted if /@/ is another
phonem that is written <e> in verbs and <a> in nouns...

In Biwa I had described the phonology in these lines (SAMPA):
  vowels in open syllables:   i  e  9Y A  @\ ow }
  vowels in closed syllables: 1  E  y  V  9  O  U
  unstressed vowels:  I  @  M

Well, unstressed vowels follow vowel harmony patterns: if closest
stressed vowell is open or mid-open the unstressed vowel will be /@/,
and so.

In Biwa syllables could become de-stressed due to composition.  In
Chleweyish some monosyllabics can be stressed or unstressed and the
/A:/ <-> /@/ and /e:/ <-> /@/ changes can be appreciated.

Probably looking in your hipotetical language what happen when /'e/
and /'o/ lose stress it can be seen as a reduction: /'e/ <-> /i/ and
/'i/ <-> /i/, or a combination: /'e/ <-> either /a/ or /i/ or /aj/ or
/ja/ according to some rules.

About consonants contrast in codas or onsets, it is posible to look
what happens when a consonant changes from one position to another.

In Spanish, flapped /r/ and trilled /rr/ only contrast in intervocalic
positions not begining a word.  In codas the consonant is always
trilled.  In the begining of a word is trilled as well.  In consonant
clusters it is flapped.  When deriving, word inicial /rr/ remind /rr/:
real / -> irreal /, but [rr] in codas becomes a
flapped onset when suffix adds a vowel: amor /a.morr/ -> amores

-- Carlos Th