The compound pahitatto appears problematic because of -atto, which is
clearly the nom. sg (masculine) - which in turn implies that the stem is
pahitatta. It is indeed that. One may ask, how can that be, since the nom.
sg of attta(n) is always attaa? The answer is that this reflects the
behaviour of a possessive (bahubbiihi) compound. The peculiar feature of a
bahubbiihi is that though it is made up of two (or more) nouns, it is
adjectival in function (though secondarily an adjective itself may fulfil
the role of a noun: a fact that we must disregard when first trying to
understand the nature of a bahubbiihi compound). If a word is an adjective,
it must be declinable in all all three genders, because in languages like
Pali, an adjective must agree with the word it qualifies in number , gender
and case. Because of this, a bahubbiihi must have a flexible form. This
is usually achieved by converting it to the declensional pattern of words
ending in -a. So when a bahubbiihi has atta(n), raaja(n) (etc.) as its
last member, it is delined as if that word was atta-/ raaja.. That is, it
does not follow the declensional pattern of words ending in -an, but
follows that of words ending in -a (like purisa, than like raajan).
You find brief reference to this in Steve Collins Grammar at p. 135 and in
Warder's at p. 137.
The same thing is found in Sanskrit, where the distinction is even more
striking.
Mahinda
[Non-text portions of this message have been removed]