m/anusvAra & s/visarga

[Dominique.Thillaud]

At 20:11 +0100 4/11/97, Jacob Baltuch wrote:

>To be pedantic it is not strictly, stricly true (cf gRhNAti, azvAbhyAm,
>where h is followed by N or bh by y) but it looks like your point still
>stands that I will always be able to restaure visarga from the context.
>The algorithm will be a bit more complicated though.

        Yes, I wrote too fastly. I have just the intuition we can obtain a
context-sensitive partition between the chars h and visarga if represented
by the same char (and duhkha, azvah are quite clear).

        What do you think about? :
R1: following an occlusive, h is always a diacritic for aspiration,
otherwise it is the code for the metaphonem H.
R2: H has two context-sensitive realisations: voiced is h, voiceless is
visarga; the choice between them result from backward assimilation (as
usual, silent is voiceless).

NOTE: I don't say nothing about H. This meta don't had any linguistical
reality, h and visarga have very different origins and meta-char (used just
for a convenient roman transcription of the devanagari script) would be
perhaps a better word.

        You can give two sorts of counter-examples:
1) exceptions to this rules: we'll try to change the rules.
2) overlapping of h and visarga contexts: I'll resign!

        Regards,
Dominique



Dominique THILLAUD
Universite' de Nice Sophia-Antipolis, France