On 2006-01-20 16:27, tgpedersen wrote:
> Hm. How do we know that? Don't the occurences in the daughter
> languages match?
They do -- as regards the archaic compounds preserving the original
stem. Such first members have some special properties (e.g. zero grade
even if stressed: *dus-, *h1su-, *n.-, *mg^&2-, *[kW]twr.-, *tri-,
*h2ju-, *dru-, *p&2tr.-, etc.), and here again we have a good deal of
cross-branch agreement. The thematised type is quite obviously late and
analogical. For example, the "linking -o-" is extremely common in
Classical Greek compounds, but not in Mycenaean!
Piotr