[From Rick Marken (920125.1500)]
Avery Andrews (930123.1350) --
An important point is that PCT does not challenge the existence of CPGs,
rather, it simply claims that they will normally produce reference levels
for perceptions (and will therefore in general be able to produces
error signals and drive behavior when the afferent pathways or cut, although
considerable retuning will be necessary to get passably effective behavior).
While re-reading this nice little comment by Avery (I think a CPG is a
Central Program Generator?) on an article by Schmidt, I had a bit on an "aha"
experience. Perhaps (I aha'ed) one reason for the glaring mistakes about
PCT in the motor control literature is the difficulty of understanding how
control of a scaler (perceptual) variable can be responsible for "control" of
the
vector (output) variable that we see as behavior. I often see diagrams of
systems
that control sequences of responses or several simultaneous responses which
look like an upside down bush (not the ex-president ) with the trunk at the
top
and the branches diverging toward the bottom. This, I think, is the most
"natural" way to think of the process of producing these outputs. "Lifting an
arm", for example, requires setting the appropriate values for the tensions in
5 or so muscles. The "lifing an arm" response is thought of as a vector with
five components that must be set appropriately.
When people think of control theory, they think of set points for scalars --
so
if control theory is applied to arm lifting, the tendency is to imagine that
one
control system should be assigned to each scalar component of the vector
output (each muscle). But it is also possible to have a control system
controlling
a scalar quantity that is a function of ALL components of the output vector at
once
-- a single scalar number, p, can vary as the sum of 5 muscles tensions, for
example. The scaler perception now represents a complex of tensions -- a
sensation in the BCP hierarchy. This perception will now be controlled by
varying
all the muscle tensions appropriately (not very interesting control if there
is no
disturbance to the indivdual tensions -- but very interesting when there are).
The "aha" is just that I think it is hard for people to understand that scaler
perceptual variables (which are, indeed, all that a control system controls)
can
represent complex aspects of the environment, each aspect influenced by MANY
of the systems own outputs. The model of perception in BCP looks the same
as the model of output in conventional models except that the top of the
perception
bush is the controlled variable,p. There is also an output bush in PCT; the
top of
this bush is an error signal.
There are plenty of difficult problems for PCT; but they are the inverse (so
to
speak) of the problems of conventional models. The problems of conventional
models involve finding ways to turn a command signal into all the "right"
outputs
that achieve the command. The problems for PCT models involve finding ways to
represent the consequences of outputs as a scalar perception. Conventional
approaches
want to know how to generate outputs; PCT wants to know how to generate per-
ceptions. Both problems are equally difficult (probably) -- but the PCT
problem has
one advantage -- it is soluble. The presense of unpredictable and undetectable
disturbances in the environment make the conventional goal impossible to
achieve
in the real world.
Best
Rick