Anusvaara for word-final 'n'?

[Madhav Deshpande]

John Smith writes:

>Whitney, for instance, notes that the normal MS usages are "trIMlokAn" or
>"trI&llokAn", and suggests that a "better" usage would be "trI&l lokAn".
>(Grammar, 206a.) He also states that, "according to the Hindu
>grammarians", final *m* before l- becomes a nasalised "l", in exactly the
>same way as happens to final -n (71, 206, 213d) -- so that the use of
>anusvara for -n is "just as reasonabl[e]" as for -m. Nasalisation is a
>notoriously difficult area, and I'd be interested to know what "the Hindu
>grammarians" do actually say here, since Whitney is not the most reliable
>witness; but there is no reason to mistrust his comments on MS usage.

        As far as the usage of the MS goes, I have noticed the same thing
that a word-final 'n' is often turned into an Anusvaara, before 'l' and
before other consonants as well.
        However, the only Paninian authority for turning a word-final 'n'
into an Anusvara is in sandhis like devaan+ca > devaa.mzca.  Here, Panini
offers an option of an Anusvaara / Anunaasika vowel.  The only other
context where an 'n' can change to an Anusvaara in Panini is in
word-internal derivational contexts:  kuru+anti > kuru+a.mti > kuru+anti.
To a non-Paninian, such a derivational process may seem strange.  However,
this process prevents this 'n' from changing to '.n', because the rule of
retroflexion of 'n' after r/.s does not recognize this 'n', it recognizes
the Anusvaara, which cannot be changed to a retroflex '.n'.
        Anyway, apart from cases like devaa.mzca, there is no Paninian
sanction for turning a word-final 'n' into an Anusvaara.
        Best,
                                Madhav Deshpande