>Is there some statistical proof which give us the significance of a
relationship? Say, if for >two languages A and B 1000 cognates are
identifidied, this is more significant if for the two >languages A and B
only, say, 250 cognates are identified.
One cognate is sufficient.
If it is shown to be a cognate, then the languages are related.
You mean, I presume, "similar words", which are not the same thing at all.
(a) they may be borrowings.
(b) they may be coincidence (this is surprisingly common)
(c) actual cognates may not look alike at all.
So it's the rules that matter, not the number.
Peter