>The problem with reconstruction is that many non-IE word surely was adopted
by PIE >language, and passed through daughter-languages. Cf. Latin Caballus,
that was not >a Latin word, but substituted Latin Equus in all Neolatin
languages (Portuguese >Cavalo, French Chevau, Italian Caballo).
But Caballus was borrowed from Celtic (more specifically, Gaulish
Caballos) - therefore it is likely still a PIE word (perhaps from PIE
*kaph-o "hoof." We might imagine a PIE *kap-lo "hoofed one" leading to a
Gaulish *cablo- which was extended out to caballo-). I think that quite a
few of the alleged pre-IE words in Germanic need to be checked for possible
early Celtic origins first before being assigned Pre-PIE status.
-C. Gwinn