> I take it from the above that there are, in fact,
> stages in IE languages with laryngeals intact, which initiate changes
> before they are deleted.
Yes. The clearest are in Sanskrit, but they exist in some of the other
languages as well. In Sanskrit a particularly interesting pattern is the
variation in the same root:
c ~ k ~ kh
which derives from *keH ~ *koH ~ kHe.
Some Sanskrit roots show direct evidence, by having an -i- vowel (<H) in
forms which don't normally have this -i- vowel.
Another interesting one is the strange Class 9 present formation, where
there is variation between:
strong -nâ, weak -ni (before a consonant) and -n (before a vowel)
With laryngeals these become perfectly regular present formations:
strong *-neH weak *-nH. (> -ni-C, nV)
There are quite a few other bits of evidence as well.
Peter