In response to Miguel's question, here's an
outline of my personal theory about the origin of upper decad names in
Gothic.
First, let's take stock of the most
important forms of "ten" found in Germanic. PIE *dek^m > PGmc. *texun
> *texu, with *tixun ~ *texun restored dialectally from the i-inflected
variant *texuni.
The form *texu became declinable as a noun and was
reinterpreted as a u-stem (neuter *téxu- or masculine *texú-s > *tegu- ~
*tigu- ~ *tugu-, cf. OIcel. tigr 'set of ten'). West Germanic tended to
generalise invariant *-teg ~ *-tig ~ *-tug in decad names, but Gothic (as well
as Old Scandinavian) retained inflected forms in the numerals 20-60: Goth. Acc.
tiguns, Gen. tigiwe, Dat. tigum. The Gothic nominative is unattested but can be
reconstructed as [+ Verner] *tigjus < *tigiwiz (cf. OIcel.
tiger).
The big question is how to interpret the -tê-
element in the Gothic names of the upper decads:
70 sibuntêhund
80 ahtautêhund
90 niuntêhund
100 taíhuntêhund/taíhuntaíhund
Szemerényi's hypothesis (which assumes the
generalisation of *penkWe-dk^omt > *penkWe:k^omt > *fimfe:xund, hence
analogical *seftune:xund > metathesised *sefunte:xund > Goth. sibuntêhund,
and so forth) is paper linguistics: *fimfe:xund is not attested at all (nor is
*seftun) and the chain of analogical changes needed to extract <-têhund>
involves ad hoc solutions. Hypotheses assuming têhund < *de:k^mt- fare little
better, since there is no independent evidence of such a vriddhied form. The
interpretation of <-tê-> as a preposition ("seven to a hundred") hardly
merits discussion.
It should be emphasised that the problematic forms
are characteristically Gothic rather than common Germanic -- cf. such
straightforward formations as OIcel. siau tiger, átta tiger, níu tiger, tío
tiger (cf. fimm tiger ~ fimtigi '50', etc.), OHG sibunzug, ahtozug ~ ahzech,
niunzug, zëhanzug ~ zênzech (= hunt). A similar use of <hund> in decad
names is found in West Germanic (most consistently in Old English), but there
<hund> comes first:
70 hund-seofontig
80 hund-eahtatig
90 hund-nigontig
100 hund-têontig (= hund)
110 hund-aendlaeftig
120 hund-twelftig
The correspondence <sibuntêhund> =
<hund-seofontig> suggests that the correct division of <sibuntêhund>
is <sibuntê-hund>, where <-tê> corresponds to <-tig>. Since
<-tê> can hardly derive from Goth. *-tigjus, we can hypothesise that the
more archaic neuter variant *-texu < *-téxu: survives here:
*sibuntexuxunda- > *sibunte:xunda (haplology
combined with compensatory lengthening)
I don't think this dialectal use of <hund>
in upper decad names has anything to do with the original function of PIE
*dk^mtóm as a Gen.pl. form ("of sets of ten"). It's quite evident that
<hund> in these constructions means simply "a hundred (or something of
that order anyway)", and that the decad names in the range 70-120 denoted such
"approximate hundreds" in an explicit manner, with some redundancy.
(A curiosity: Crimean Gothic had <sada>
'100', an Iranian loan.)
Comments welcome.
Piotr