Dear Stephen,
SH> I think of its function as a kind of
SH> link between perception and conception -- percepts go in and concepts come
SH> out.
In Theravada terms, 'phassa' comes in and 'vohaara' comes out.
http://www.accesstoinsight.org/canon/sutta/anguttara/an06-063.html
SH> In other words, visual vij~naana cognizes a red blob (the vastu "out there")
SH> which is them present in vij~naana as an aakaara, which is also "coloured"
SH> by a vedanaa reaction -- so it is no longer reallty a pure sensum.
As I understand, you relate the linearized model with the
vi~n~naa.na -> (phassa) -> vedanaa -> sa~n~naa
linkage.
I suppose that this linearized model stems from Madhupindika sutta
http://www.accesstoinsight.org/canon/sutta/majjhima/mn018.html
where Ven. Maha Kaccana explains the sequence of perception.
However in the suttas told by Buddha himself the model is ramified
"phu.t.tho, bhikkhave, vedeti, phu.t.tho ceteti, phu.t.tho
sañjaanaati."
http://dhamma.ru/lib/paticca.pdf
Though all models are a matter of conventionality, I find the ramified
model more exact. It also better agrees with the findings of modern
neurophysiology that the input signals are simultaneously processed by
several subsystems of neural systems. Thus the processing of the
signals by vedanaa and sa~n~naa is better described as not sequential
but parallel, though these processes are of course interrelated.
In the Theravadan model the main agent of influence on vi~n~naa.na and
phassa is craving - ta.nhaa .
http://www.accesstoinsight.org/canon/sutta/samyutta/sn12-064.html
SH> Sa.mj~naa then extracts / constructs (often called nimitta-udgraha) the
SH> salient feature -- the nimitta -- from that aakaara (sensum) and labels it
SH> as "red" (= inference).
The key function here is attention (manasikaara) which is directed on
the perceptual image (nimitta). Being not directly conditioned by the
contact (phassa), it can be relatively easily taken under volitional
control to attend to selected class of perceptual images (nimitta).
Best wishes,
Dimitry