Dear John

Sorry for my late reply. Recent thunderstorms have seriously disturbed
mobile networks in Sri Lanka, and it is still not yet better to date.

You wrote:

> My understanding is that "sara.nanti" is simply a sandhi by
> assimilation of "sara.na.m" and "ti" (for "iti"). Thus sarana.m is
> still accusative but is set in direct quotes, and not instrumental at
> all.

I won't deny its being sandhi, for, sandhi is none other than
assimilation of adjacent sounds. However, in "pure" sandhi, there are no
semantic relations involved. If there weren't any sort of relation
between "sara.na.m" and "iti", it would have been pure sandhi. The
question is whether there is such a relation or not.

Now let's assume sarana.m is in accusative. If it were really the case,
"gacchaami" would have become a verb requiring a double object. Then
"iti" wouldn't have been necessary. Cf:

Puriso assa.m gaama.m nayati = The man brings the horse to the village
(Here "nayati" is a verb of double object)

On the other hand, suppose "iti" is used here to indicate a direct
quote. A quote is a self-contained set of words without any relation to
any word outside the quote. Then, in this case, we cannot claim that it
must be of the same case, i.e., accusative, as "buddha.m". On the
contrary, we must take the context into account.

Now let's go to the commentary of Khuddakapaa.tha.

. . . Ta.msama'ngii satto ta.m sara.na.m gacchati, vuttappakaarena
cittuppaadena “esa me sara.na.m, esa me paraaya.nan”
ti evameta.m upetiiti attho. (Pj1 - 16)

Trs.: . . . A being endowed with such a mind relies on him (i.e., the
Buddha) as a refuge. The meaning of it is " (He) approaches him (the
Buddha) with the aforementioned kind of mind as 'This is my refuge, this
is my shelter' "

Depending on this commentary we can interpret "buddha.m sara.na.m
gacchaami" in two ways.

The first is taking "sara.nanti" as a quotation, and translating it as
"I rely on the Buddha (thinking) 'This is my refuge' " In this method,
"sara.na.m" is in nominative case.

The second is taking "sara.nanti" as an indeclinable compound. Then the
translation would be: " I rely on the Buddha as refuge".

> Also, I didn't comprehend your statement that "sara.nanti" can be
> viewed as an indeclinable compound having the instrumental case. By
> definition an indeclinable doesn't have any cases, is it not?

In classic grammars, in fact, indeclinables are nouns but with
case-endings elided, and subsequently, without changing their forms in
various contexts. This view has practical reasons. For instance:

mahanto puriso (= a great man)
mahantena purisena (= with a great man)

You may notice in the examples above that adjectives take the cases and
numbers of the nouns they modify. Then what about "purisoti etena",
which should be translated as " with this word 'puriso'"? This is a very
common form in the commentarial literature.

Here the grammarians take "purisoti" as modified by "etena". Since they
must be of the same case, "purisoti" is viewed as an indeclianble
compound with the instrumental case elided.

with metta

Ven. Pandita