A 'drift' from OV to VO?

[Frederick Newmeyer]

Matthew Dryer has shown that, once we correct for areal and genetic bias,
the 'preference' for OV order is greater than that for VO order in the
world's languages. But interestingly, I have seen it claimed in a variety
of places that attested (or uncontroversially reconstructed) word order
changes from OV to VO are far more common than those from VO to OV.

My first question is how widely accepted is such a claim among historical
linguists and typologists? Is there much support for such an idea and its
implication of an overall general 'drift' from OV to VO?

If this claim seems well motivated, the conjunction of the 'preference'
for OV and the 'drift' to VO is very curious, no? One might even conclude
that the OV preference is a remnant of a 'proto-world' OV (caused by
what?), which functional forces (but what functional forces?) are skewing
gradually to VO. And, indeed, linguists coming from a variety of
directions (Venneman, Givon, Bichakjian, and others) have concluded
something very much along those lines.

I'm curious what thoughts FUNKNET subscribers might have on this question.
I'll summarize if there is enough interest.

Fritz Newmeyer
fjn at u.washington.edu