The following are some excerpts I stumbled on in shuffling through Hopper & Traugott's Grammaticalization (1994):
... the focus on universals privileges the uniformity of rule types and reasoning types across languages and times... ... the linguistic forces that are evidenced today are in principle the same as those that operated in the past... There is no reason to believe that grammaticalization did not occur in languages spoken ten thousand years ago in much the same way as it does now. (38)
I think it's obvious how this can be encouraging to us Biblical Language nerds, in knowing that we can employ the tenets of grammaticalization to ancient languages.
There is no doubt that over time, meanings tend to become weakened during the process of grammaticalization. Nevertheless, all the evidence for early stages is that initially there is a redistribution or shift, not a loss, of meaning. (88)
This is a really interesting point. It's not that a lexeme's semantic punch gets stripped down at an equal rate through its ultimate "bleaching" process. At first, there's a little game of give-and-take among the various semantic flavors that are present in the lexeme's synchronic use.
As has been stated frequently in previous chapters, there is nothing deterministic about grammaticalization and unidirectionality. Changes do not have to occur. They do not have to go to completion, in other words, they do not have to move all the way along a cline. A particular grammaticalization process may be, and often is, arrested before it is fully "implemented," and the "outcome" of grammaticalization is quite often a ragged and incomplete subsystem that is not evidently moving in some identifiable direction. (95)
This was an important point for me to read. I just got done writing an article with a friend on the BH lexeme בלי and we were fortunate enough to have seen this little form in a number of different grammatical stages, in fact, we basically see this form go through a full cycle of the functional options available for any word within the corpus of the Hebrew Bible. It changes from a noun to various stages of an adposition and conjunction, and finally deteriorates into a negative marker. All that to say, I think I got mentally spoiled in seeing such a rich attestation to grammaticalization's identified stages of semantic and functional growth (and "death") that I needed this reminder that not all lexemes undergo such a rich and complete transformation as בלי.
... it is not language that changes, but rather the rules of grammar... (95)
This nuance is difficult to remember (and understand) when we so easily talk about "language change" rather than "rule change". I still don't fully grasp this concept...