On 2006-04-25 12:40, tgpedersen wrote:
>>>> Nouns of
>>>> the type *po:d-s 'foot', *nokWt-s 'night', *wo:kW-s 'voice', *do:m
>>>> 'house', etc., are free-occurring lexemes, not bound forms.
>>> Of course they are. I said they originated in bound forms.
>> And why should anyone believe that?
>>
>
> I have thought of setting up a system of bribes for that to occur. In
> the meanwhile I find solace in Grassmann's biography.
My question was serious. If you think that nouns like *po:d-s developed
from second members of verbal compounds with a governed first member,
then you're clearly wrong for the following reasons:
(1) I'm not aware of _any_ examples of verbal compounds with an o-grade
root noun as the second element. If a true agent root noun occurs in
such a context (and it was observed already by Schindler that o-grade
root nouns were _not_ ordinary nomina agentis) it is accented and
e-grade in the strong cases (as in Skt. nr.-hán- 'man-killing' [note the
palatalisation of *gWH], Lat. -spex, -ceps) and belongs to the
hysterokinetic accentual type (dat. nr.-gHn-é).
(2) If you say that *h3re:g^-s has a long *e: because *-h3ro:g^-s
'fixer' was rarer than a free-occurring e-grade root noun, this can be
taken to imply that the attested o-grade root nouns represent those
second members of compounds that were more frequent and therefore more
successful. Do you mean that compounds with FOOT as the second member
were more frequent than FOOT itself?
Non-verbal exocentric compounds like *h2rg^i-po:d-s 'swift-footed' are
of course possible, but don't ask me to believe that they were
overwhelmingly frequent when compared with *po:d-s (I could consider a
bribe, but it would have to be adequately high). Besides, such compounds
are specifically non-verbal, so neither *-h3ro:g^-s 'fixer' nor anything
of that sort could occur in them as an agent noun governing an object
(as in Eng. problem-fixer)
(3) In o-grade root nouns the underlying verb is very often
intransitive, which means that such a root noun could not occur in
verbal governing compounds at all.
Finally, the original question was not so much about the vowel grade of
*h3re:g^- but about its anomalous long vocalism across the paradigm.
Even in a language like Latin, which levels out numerous allomorphic
features, we have <pe:s> but <pedis> (unlike <re:gis>), and this unusual
behaviour of h3re:g^- as opposed to other root nouns, whether e-grade or
o-grade, recurs in branch after branch.