The role of analogy, alliteration and sandhi in counting

From: petusek
Message: 34683
Date: 2004-10-15

Vaclav Blazek (Numerals) offers a weird solution to some numerals in some of the World's languages.
 
He suggests that some of the strange phonetic irregularities in numerals may be due to an analogy, alliteration and sandhi that take place when numbers are counted in a set (e.g. from 1 to 10 or sim.) We have all encountered alliterative and analogical changes, of course. However, here is what Blazek assumes for IE (not that he claims it is the only correct explanation, in fact, he asks):
 
*oy- *dwoH1 *tri-*kWe *dur- *pen(gh?)- *kWe = "1", "2", "3" plus, "4", "5" plus >> *oy- dwoH1 *tri- *kWetur- *penkWe (???)
 
/t/ in *kWetur- would be due to analogy after *tri-
 
External parallels supporting the reconstruction *dur-, *dwor- etc. would be in Altaic *to:r-/*tu:r (Starostin).
 
Does anybody know of analogies in other language groups? Any supporting or disagreeing ideas?
 
Petusek