Metathesis of *tk^ - how old

From: grzegorj2000
Message: 40147
Date: 2005-09-19

The most known example of *tk^ is:

1) Sanskr. r°ks.a- "bear" = Avest. arEs^a- = Greek arktos, (dial.)
arkos = Arm. arj = Albanian arí = Lat. ursus = Welsh arth = Hitt.
hartagga = Aquitanian harsus < *Hrtk^o-

The other example is less clear, though:

2) Sanskr. taks.an- "carpenter" = Avest. tas^an- = Greek tekto:n,
and also Lat. texo: "I weave", Slavic tesati "to hew, to chop into
shape", Old High German dehsa(la) "axe", Old Icel. thexla, possibly
also German Dachs "badger". The Hittite form is interesting: takS-,
takkeS- "take, assemble", and it can suggest a more complex
consonantal group here: -*ktk-. Cf. also Slavic tUkati "to weave" <
*t°k-a:-, hence tesati < *tetk^-a:- < *tek-t°k-a:-
(reduplication, vowel reduction and simplification). Latin,
Sanskrit, German and Hittite forms could even be derived from *tek-s-
.

Now the problem. It is: was a very early metathesis of *tk^ in IE
languages, except Anatolian and Hittite? Which is the evidence for
such an early metathesis?

In fact, the 2nd example is not convinceable. Are there more
examples for IE *tk^?

Let's check for details:
1) Latin -s- (-x- doubtful, see above) does not prove the
metathesis: *-tk- > *-tsk- > *-ts- > *-s-;
2) Celtic -t- doesn't, either(*-tk- > *-tsk- > *-ts- > *-t-);
3) Indian - we do not know, maybe *-tk- > *-tsk- > *-tks- > *-ks.-
(*-tsk- > *-kst- > *-ks.- less problable!),
4) the same in Iranian: (*-thxs^- > *-s^- - how to explain the
reduction of -x-, but -xš- possible in Avestian!)
5) Armenian *-tk^- > *-j- [dz] also cannot prove the metathesis,
6) In other groups the development is unclear.

Resuming, the metathesis is attested only in Greek, and even not in
all dialects (Arcadian arkos not arktos!)

The development of *dhg'h- (like in Greek khtho:n "earth, terrain,
ground" but khama'i "on the ground" and Old English guma "man", cf.
arktos ~ arkos!), *dhgWh (like in Sanskr. ks.iti- "destruction",
Avestan xs^yo- "disaapearence", Greek phthisis), possibly also *tp
(Old Greek ptolis "town, city", next only polis), seems to be
similar - a metathesis in Greek, no metathesis elsewhere.

There is a view that the evidence of the early metathesis is the
development of -k + tlo-/tro- which is as if the same as the
development of -tk-, like in Latin ve:lum "sail" and Slavic
veslo "oar" < *wek^tlo- "a drive of a transportation meaning" <
*weg^H-tlo-.

However, this example is not convinceable because of the diminutive
vexillum "standard, flag" which suggests *vexlom, not *veslom
(compare with ursus, not *urxus!). It seems that the development *kt
> *ks > *s > NULL (lengthening of the previous vowel) was only
before -l- while the original -tk- > -s- not before -l-.

So, are there Latin words with unchanged original -rkt- (rct in
spelling) which would prove the rightness of my point? Or, are there
words with -rs- < *rkt (not *rtk) which would be evidence for early
metathesis?

Notice also *tt > ss (or s after a long vowel) in Latin (seems like
*tt > *tst > *ts > *(s)s analogical to the postulated *-tk- > *-tsk-
> *-ts- > *-s- without the early metathesis)

And what about the word for "thousand"? Its protoform seems to be
*tu:-t-k^mti- (Slavic *tyse.tj- or *tyso.tj-, Lith. tu:kstantis,
Lett. tukstuotis, Gmc. thu:sandi). Despite of many sources, it seems
to me that *tu:s- was improbable here because of Baltic facts which
point to *tu:-t-, not *tu:-s-. And, if it really was *tu:-t-k^mti-,
this would be another piece of evidence for the late and limited (to
Greek) metathesis.

Are there more evidence for or against?