And, finally, in constructing such a theory, does one 'extrapolate?' That is, can one extrapolate backwards in time. Or, actually, can one extrapolate *in time* at all?

Insel, something bothered me about this too; but, I got so carried away by Nostratic that the post slipped my mind.

One can extend the inferences of biological evolution, to the somewhat similar process of reconstructing the evolution of language and extrapolate therefore. What is probably nagging you, is the "exrapolating backwards" bit. Extrapolation is neither backward or forward, but is certainly a comparative/relative process