[From Rick Marken (930929.1100)]
Gary Cziko (930928.2220 GMT): I had sent some of the discussion
on manual control to Gavan Lintern of UIUC's Institute of Aviation
It's great that you're doing this, Gary. Please forward my
reply to Gavin, if you like. But I would suggest that, if
people like Wickens and Lintern really want to discuss the
relationship between PCT and conventional applications of
control theory in psychology, they would take the trouble
to subsrcibe to CSG-L. Then we could have a real dialog --
instead of these "'tis so, 'taint so" sniping matches
(pardon the violent analogy, Hal; peace, brother).
Gavan Lintern (920927)--
A dominant thrust in Psychology is to understand the nature of the
If this is true, then that understanding has not progressed
very far, as we can see in the next sentence:
Thus we control action to control perception in a
The only sense in which this could be correct is in terms of hier-
archical control; we control the PERCEPTION of some aspect of an
action in order to control a higher level perceptual variable. Thus,
we might control the perception of velocity in order to control the
perception of position. But control is ALWAYS the control of
perception; that is how a "controlling mechanism", ie. a control
system, works. I don' think the people who are studying the control
of action are aware of the fact that what is controlled about the
action is PERCPETION; at least, I have never seen any test done to
determine that perceptual aspect of "action" that is under control.
many psychologists focus at that level [action], and it is not a
trivial exercise to dismiss their work.
You're telling me. But I'll keep trying. Control systems do not
control actions, a fact that is demonstrable analytically,
experimentally and computationally. We (Powers, Marken, Bourbon)
have demonstrated this fact in all three ways -- over and over
and over again -- to (predictably) no avail. I am now convinced
that you cannot change system concepts by disturbing lower level
variables (which is all we are doing with these demonstrations).
Does this mean that changing people's minds about PCT is
Basically, the demos don't mean a thing if the observers are not
doing that reorganization thing.
Powers' model is instructionist specifically because of the
way set points are used in the model. The general notion is that lower level
set points are instructed from higher levels.
I'm not sure what "instructionist" means but it seems like it's
probably the wrong way to describe the way lower level reference
signals are set by higher level systems. The higher level
systems (like all control systems) simply generate whatever
output is required to keep their perceptual input NEAR their
own reference input. The outputs of these higher level systems
(which are typically continuously varying) INFLUENCE, but
do not DETERMINE, the reference inputs to lower level systems,
mainly because each lower level reference is usually
influenced by the outputs of SEVERAL higher level systems.
The origin of set points is a problem.
What's the problem? Set points of lower level systems have
their origin in (are varied by) the outputs of higher level
systems; there is no "problem" with the origin of "set points";
we have built working models of hierarchical control systems
to prove that no such problems exist. Ask Gary for a demo of
my spreadsheet hierarchy, for example.
To note that they come from higher levels does not resolve
the problem unless it is possible to specify the origin at
the highest level.
Again, what's the problem? The highest level is the highest level;
it exists (according to PCT) as a layer of neurons in the brain.
Are you claiming that the origin of the reference for these
highest level systems is a problem? The current hypothesis is
that the highest level references are the result of reorganization
(which happens to be a "selection" process).
contrast, ecological psychology, and dynamical systems have the set points
(Gibson's affordances, dynamical systems attractors) coming from lower levels.
Many problems here. First, if you mean by "lower level" what we mean by
"lower level" than you are saying that, in the above models, the set
points for controlled variables (like temperature) come from the sensory
surface; so I interprete this statement to mean that set points are
determined, ultimately, by the environment. So the environment determines
the goals of the organism. I would like to see the evidence for this.
Second, I have never seen versions of ecological or dynamical systems
models that actually behave in a realitic environment (one with
continuously varying but undetectable disturbances). So it really
doesn't matter where the set points in these models come from --
these models don't do what the PCT model does, they don't behave
in a realistic world (the same can be said, by the way, of the
trendy "situated" models -- they are not really models, from a
there is persuasive evidence in a few areas of human behavior regarding
the factors that enter into determination of an affordance (e.g., stair
climbing, sitting, phase transitions in limb organizations).
Could you desribe ONE piece of evidence regarding the factors that
determine an affordance; even better if it's persuasive.
One of the best clarifying examples of 3. above is Peter Kugler's account
of nest construction by African termites. By his account, there is no plan
set point regarding the patterns that are achieved.
That's the evidence? What Peter Kugler says?
At one time on CSGnet,
someone (Powers I think) argued in relation to this sort of thing that the
feedback loop had to be viewed at the level of individual termites as they
track the pheromone gradients. However, as an organizational scheme, that
the set points coming from below, and undercuts his primary thesis.
The set points come from below what? The collective result is a side
effect of individual control. There are no connections between
termites that allow the collective to function as a control system
in itself; so the goals of the individual termites are not "below"
those of a "higher" collective control system becuase there IS
NO COLLECTIVE CONTROL SYSTEM; there is no collective perceptual
signal; there is no collective reference input. If there is, then
please point out where it is.
Only if we believe the system is nonlinear do we need anything more
Someone (I think Marken) has spoken disparagingly about the value
of attractors as an explanatory concept for human behavior.
That's me. I "disparge them" only as models of CONTROL. They are
perfectly OK (as far as I know) as models of nonlinear (and non-
controlling) dynamic processes.
Nevertheless, if we believe that behavior is nonlinear, the only
currently known approach is through nonlinear dynamics.
Non-linearity is not an issue. PCT does not assume linearity in
any functions -- and even monotonicity can be mildly violated.
The issue is CONTROL. If behavior is CONTROL, the only currently
known approach is through PCT. The non-linear dynamics people
are not modelling control; they are not testing to determine
whether control is involved in the processes they do model
The POINT of PCT is CONTROL. The non-liear dynamics folks
(like the cognitive folks, the behaviorist folks, the situated
action folks, the ecological folks, the manual control folks,
etc) are missing that point -- and, it seems to me, they are
NOT missing it by accident. This is purposeful "missing the point".
The phenomenon of control and the PCT explanation of it are
DISTURBANCES to some fundemental assmptions about how people
WORK -- and there is ACTIVE resistence to PCT. If people were
not controlling for a variable that is disturbed by PCT (and if,
as Greg Williams claims, they were really doing stuff that is
very much like PCT) then why is virtually NO ONE in conventional
psychology doing PCT-like research or modelling. Wyy are there
absolutely no tests for controlled variables? I can think of two
possibile reasons for this; either 1) PCT is the most obviously
idiotic and/or valueless idea ever proposed in psychology
or 2) there is active resistence to PCT for reasons that have
nothing to do with its value.
One way to deal with a disturbance is to push back; the other is
to avoid it. Most psychologists have taken the second tack. It
would be great if you would take the first -- PUSH BACK. Show
us what's wrong with PCT; let's see if some truth can emerge
from an honest conflict; nothing comes of nothing.