> In IE, the feminine suffix *-ih2 is obviously the
> unthematized counterpart of diminutive *-ik-o-.
That (pre-PIE?) rule (-k# >-h2#) seems to be standard now?
That would lead me to believe that the roots *pak- (with "sideform"
*pag-, suspiciously called "perfectly natural" by root
editors) "pole;peace;country;border" (whew!) and *pah2- "let graze,
guard, herd". Anyone one who accepts the jumble of senses of the
former will not be able with a good conscience to criticize that
merger on sematic grounds. But that merged root has has suspiciously
similar cousins in Proto-Austronesian (cf
http://groups.yahoo.com/group/cybalist/message/37244
), so this root is no good candidate for proving the PIE nativeness of
neither plain *k nor *a.
Torsten