> Are we not heading to the point that, for a nearest neighbour
> approach, we have to allow for auditory closeness as well as
> articulatory closeness?
Definitely. An important and often crucial part of any sound change is
the speaker-to-hearer link (how else could anything be
re-phonemicised?). It's naive to imagine that sound change takes place
in the mouth of the idealised "average speaker" and so to insist that it
ought to be articulatorily continuous.
Of course. There has to be some correlation between perception and articulation.
refers both to the perceptual and articulatory properties but if they are
correlated then "no problem".
It is because they are correlated that linguists can freely give both articulatory
descriptions of phonemes with the implicit assumption that one can be derived
from the other
to a good degree of approximation.