[Martin Taylor 950630 14:00]
Rick Marken (950630.0815)
Bruce Abbott (950629.1730 EST) --
How else would it have any effect an aspect of responding but
through the organism?
By the way, this was a rhetorical question, right? Surely a real
PCT theorist like yourself knows the answer to this VERY basic
question.
This is an open post test, by the way. Both Bill and I gave the answer
yesterday.
It's not such a basic question, not in the colloquial sense. It's a basic
question in the sense of being fundamental, but leading to some very
interesting considerations.
Rick's answer of yesterday will serve (950629.1845):
Recall that the output, o, of a control system will be related to a
disturbance, d, to a controlled quantity, q, by the approximation
o = -1/h(d), where h is the function describing the environmental
connection between o and q via the environment. The researchers
you mentioned studied the relationship between responding, o, and
milk concentration, d, as though o = f(d), where f is the function
describing the organismic connection between o and d via the organism.
Fair enough... And yet, and yet...
PCT researchers ALSO want to determine what kinds of control systems are
inside the organism and how those internal structure are connected, etc., etc.
The approximation o = -1/h(d) applies to ANY reasonably functioning control
system, which means that if control is very good, it is not possible to
distinguish among different control structures. Only structures that would
not control well could be discarded.
There's another point here, too, and this relates to the "many means to the
same end" aspect of HPCT (as opposed to PCT). In the formula, "o" is the
output of the _Elementary_ control unit whose perceptual signal is being
controlled. But the observable _actions_ are the outputs of many peripheral
control systems, which together form part of a changeable signal path
that is subsumed within the h() function (or rather its inverse). If
the experimenter could observe the CEV corresponding to the controlled
perception, this wouldn't matter (much), but the experimenter cannot do
so exactly. Accordingly, the variability of h() becomes important in the
observations. Part of that variability occurs between the output "o" and
the muscular periphery, and this is an aspect of the organism. What the
experimenter observes may be highly correlated with the true CEV for the
perception being controlled, but part--the variable part--is side effect.
If one wants to look at the properties of a particular organism, then,
one must look not at the success of control, but at the detail of how control
fails. But there ia problem, because the observed failure of control may
be a true failure of control, or it may be a failure of observation,
representing only side effects. Either way, observations of real or
apparent control failure MAY tell something about the organism's structure,
whereas observation of control success tells only that the organism is
a successful controller--as Rick points out, an essentially uninteresting
discovery.
Sorry--no time to do this properly; hope it makes some sense.
Martin