[Avery Andrews 931030.1130]
Emphasizing the role of reorganization on the output side might help in
making the distinctive nature of pct clearer. Suppose you have a
three level control system like Rick's spreadsheet model, with fixed
structure. Then you *can say* that there are rules for producing
outcomes (if the hand is X-much too far to the left, move it to the
right at rate Y, ...). It's just that these rules are differential
in formulation, so you have to take dynamics into account to understand
how they work (and, if you don't know what the high-level reference
settings are, you'll never really understand what is going on).
But now suppose that the model is made more flexible, so that the means
used to acheive the reference levels can be varied (gain factors switch
signs when the error signal starts to soar, as in some of Bill's and
Rick's trackign experiments). Now the organism no longer has any set
rules for attaining its outcomes - at any given moment it has rules, but
if they stop working, it finds some others that do. (if you have to
write holding a pencil with your teeth, it takes time to learn how to do
it, but you do get better).
Now the perceptual targets become clearly distinct from the rules used
to attain them, and clearly play a more central role in explaining
what's going on.
Maybe there's an opening for pct an animal psychology - there are
currently various people noting that animal behavior is well outside
the range of what can be explained by S-R models, and therefore
attributing to them human-like insight and reasoning powers. Pct
provides a wide range of intermediate hypotheses, and so might prove
attractive.
Avery.Andrews@anu.edu.au