> We may be dealing with one of the IE verb roots in which the
> velar colours the e-grade vowel to *a (my other putative examples
> include *kan- 'sing', *kap- 'grasp' and *tag- 'touch'). If so, the
> phonetic realisation of the ablaut variants is simply *bHag-/*bHog-.
There may be another explanation for at least *kap- and *tag-
The PIE restrictions on roots of the type *TeT- are as follows:
+ghedh-
+ghed-
-ghet-
+gedh-
-ged-
+get-
-kedh-
+ked-
+ket
Let's translate the notation into something newer. These notations
all have in common that the traditionally voiced aspirated and the
voiced are separated only by voicedness and that voiced unaspirated
are marked separately, in my idea of it with prenasalization, which I
note as a preceding N (for superscript)
+ged-
+geNd-
-get-
+Nged-
-NgeNd-
+Nget-
-ked-
+keNd-
+ket
The restrictions are now easily formulated: The consonants may not
differ in voicedness (the prenasalized are considered neutral) and
they may not both be prenasalized.
Why?
Because within paradigms, the root vowel may assume zero grade, so
that we get
+gd-
+gNd-
-gt-
+Ngd-
-NgNd-
+Ngt-
-kd-
+kNd-
+kt
That means that roots of two of the forbidden shapes, *get- (trad.
ghet-) and *ked- (trad. *kedh-) would have undergone assimilation one
way or the other.
Let's assume there was a fictive root *keb- (trad. notation *kebh-)
meaning "take" or "give".
In the 3rd pl. pres it would be *kb-énti -> (by assimilation) ->
*gb-énti (trad. *ghbh-énti).
Here Latin would insert an /a/ to break up the cluster, as Jens has
shown -> *gab-énti (trad. *ghabh-énti) -> habent.
Germanic, which did away with the weak-stem forms of the present,
might have introduced an -e- by analogy from the strong forms:
*geb-énti (trad. *ghebh-énti) -> PGerm. *gébant "give"
In the ppp. it would be *kb-tó- -> (by assimilation) *kp-tó (trad. id.)
Latin, as above, inserts an /a/ -> *kap-tó-.
Germanic might have too: Gothic haft-s
The unmodified root would have survived in Germanic in the causative:
*kob-yé/ó- -> *habj- "have"
That's how the traditional roots *ghebh- and *kap- might be related.
Note that giban etc belongs to the Germanic Class V roots which insert
an /e/ in the ppp. This (goth. gibans) would automatically separate
any 'victims' of the zero grade (like haft-s) from the rest of the
paradigm.
On the other hand, Hans Kuhn has a long list of substrate words and
toponyms with the in Germanic impossible sequence *kVp- meaning
"grasp; top; cut".
As for *tag- "touch", let's pretend it was once *teNg- (trad. *teg-);
/Ng/ because of the nasal 'infix'
pres.ind.
*teNgom
*te:-
*te:-
*tNgomé
*tNgeté
*tNgént -> *dNg-ént(?) (trad. *dhg-ént)
ppp *tNg-tó -> *tak-tó
Perhaps the /d/ (trad. /dh/) infected the whole present?
*gdom
*de:-
*de:-
*gdomé
*deté
*gdént
Now it suddenly begins to look like *dhe-!
But I seriously suspect this root is wanderword having to do with the
new technology of forging:
http://www.angelfire.com/rant/tgpedersen/tw.html
which was fit into the existing paradigms in the various IE (and
other) branches.
As for *kan- cf. PGerm. hana "rooster", Estonian kana "hen". Both FU
and IE are intruders at the Baltic, among peoples who had had chicken
for a long time. Substrate should be suspected.
Torsten