> If we're talking about true monovocalism, phonemic or phonetic, it
> doesn't exist. However when you analyse Sanskrit such that, say, /e/
> really 'represents' a _surface_ /ai/, you're talking about something
> that is neither phonetic nor phonemic. It is completely abstract.
No, in Classical Sanskrit it is phonemic. /e/ actually is /a/ + /i/, and
the same is true of the other vowel analyses (I won't bore you with a list).
This is true both historically and synchronically; historical */ai/ gives
/e/, and synchronically, whenever /a/ and /i/ stand side by side, they
produce /e/. And to prove it, /a/ + /e/ gives /a:i/. In classical
Sanskrit, you can think of written <e> as simply another way of writing
<ai>. Nothing abstract about that. The same is true of /a:/ and /o/.
Abstraction comes in when we analyse written <y> and <i> as the same phoneme
(and mutatis mutandis <v> and <u>, <r.> and <r> and so on ). But since
these can be analysed (theoretically) as merely phonetic variants, they are
allophones - so there is no more abstraction than allophony. Perfectly
reasonable. (Or it would be if it worked.)
Peter