> Personally, I'm inclined to view *kap- as a loanword. Surely it's at
> least *possible* for IE to have acquired loanwords before it broke
> apart into dialects.
In fact, if the pre-PIE /a/ (out of an inventory /i/, /a/, /u/) _is_
the ablaut vowel and -> (/e/, /o/, zero) then 'pure' or 'inherited' PIE
PIE roots can't have had /a/, and if they seem to have it anyway, they
must have been loaned into PIE. Cf. roots like 'spa' in English, which
is otherwise /a/-less (since Middle English /a/ -> /รค/, /eI/), apart
from a few sociolect- or register-dependent words like /fa:D&r/ and and
a small periphery of words it has attracted (Engl. /ga:D&R etc).
And that's where plain /k/'s association with /a/ makes it suspect of
being a loaned phoneme itself. But I'm repeating myself.
Torsten