Almost pregnant


   RE: Rick Marken 960109

   I bet you will be surprised when Bruce delivers Killeen's PCT
   "baby" when no one (especially you and Killeen) knew he was
   pregnant. It will take some time and much work.

   I thank you for the nice comments on the PCT Research statement
   (apparently Shannon has a different opinion) but I must give
   credit where credit is due: that entire statement was written
   by BRUCE ABBOTT not me - I haven't changed one word of it from
   his original post (date ???). One of my reasons from posting
   it for the THIRD time is to get some comments on it so I can use
   it (as well as others) as a guide for doing PCT research. This
   question of what is PCT research comes up often on the net and
   rather than tell someone they are doing it improperly (Brian is
   one example that stands out in my mind!) we could tell them
   a proper way. It can also be used as a "template" to judge
   extant research (as you did in your post).

   I don't believe there is any way to resolve the differences in
   assessment that I use and you use when it comes to deciding what
   should be included as PCT. Your criteria are just far stricter
   than mine. On most aspects of others' work I agree with you but
   there are a few where we don't agree. Since you don't have to
   agree with me there is no harm done from my view. I hope that
   is the case from your view.

   Regards, Chuck

[From Rick Marken (960109.1200)]

CHUCK TUCKER (960108) --

If all living systems are "negative feedback control systems" then all
theories of human behavior are IMLPICITLY PCT/HPCT.

Living systems are perceptual control systems but I don't think all
theories of behavior are implicitly PCT. In fact, I don't think any are --
except, of course, PCT. The S-R bug program was not "implicitly" a
perceptual control system; it was _explicitly_ a perceptual control system.
The only one's who didn't know this were the people who wrote the model and
the people who read the paper (except me, of course).

His [Killeen's] math is wrong and his formulas are flawed... but his idea
is very close to PCT.

I read Killeen's paper and Bruce's description of Killeen's "mechanics of
behavior" model and I saw nothing that seemed close to PCT. Where was the
closeness? You say:

he introduced an "internal mechanism" which is analogous to a reference
condition and a process similiar to a perceptual signal; he come very close
to PCT.

In a control system, a perceptual signal is compared to a reference signal;
the amplified difference between reference and perceptual signal drives an
output that keeps the perceptual signal matching the reference; this
is control of perception. If Killeen's "internal mechanism" were actually
analogous to a reference signal, then it would be compared to the equivalent
of a perceptual signal (by subtraction or some other function) and it would
operate in the equivalent of a closed loop; I don't recall seeing anything
like this in Killeen's model.

There are many models in psychology that postulate "internal mechanisms"; but
none of these "internal mechanisms" (like motor programs, schemas, plans,
models, etc) are analogous to a reference signal; they are not specifications
for perceptual input; they are specifications for output.

In control theory, the reference signal is "internal" but, far more important
than where it is is what it _does_. In a control loop, what the the reference
signal does is tell the control system what to perceive (actually, what level
of a perceptual variable to experience), not what to do; the reference siganl
is a specification for input, not output.

Was there really something in Killeen's model that suggested that the
postulated "internal mechanisms" are specifications for input?

It's hard to judge how close an existing model is to PCT when the existing
model does not recognize 1) the existence of control (and, hence, controlled
variables) and 2) the fact that what is controlled by an organisms is
its perceptual input, not it response output.

Based on your judgment that Killeen's model is similar to PCT, you say:

All that is needed is what Bruce and Bill are doing; point out his errors
and show him that PCT (formulated properly) provides the theory he is

This sounds pretty simple, and it would be wonderful if it were true. But I'm
afraid the most likely result of pointing out Killeen's errors will be what
Bill Powers (950609.1545 MST) suggested:

You may expose Killeen's mistakes, but it will be a miracle if he doesn't
find some way to claim that they aren't mistakes.

It seems to me the idea of a theory being close to PCT is similar to the idea
that a person can be close to being pregnant. I suppose one person is closer
to being pregnant than another if they are a woman rather than a man, hetero
rather than homo, between the ages of 13 and 50 rather than between the ages
of 70 and 100, if they have had sex, etc. But if they're not pregnant, all
this "closeness" to pregnancy seems a little irrelevant to me. I suppose a
theory with internal mechanisms might be closer than another to PCT. But if a
theory is not PCT, all this "closeness" seems irrelevant to me.

By the way, I thought your description of the "Principles of PCT-Guided
Research" (CHUCK TUCKER 960107) was _excellent_. You stated the the goals of
PCT research very clearly. In particular, I like your statement of the main
goal of PCT research:

a. to discover what perceptual signals are under control in a given
specific task;

I don't think anyone would guess, based on Killeen's model, that this should
be the primary goal of behavioral research. I would judge Killeen's model
"close" to PCT if, based on his work with the model, Killeen recommended that
future operant research be aimed at the determination of the perceptual
signals that are being kept under control by the organisms in these studies.

Your description of "THE METHODS" for achieving the goals of PCT research was
also exceptionally clear. I was particularly impressed with the way you
succintly "dis-ed" conventional research methods:

3. The behavior of closed loop systems cannot be correctly analyzed using
    methods which assume unidirectional causality.

and clearly stated the proper way to study closed-loop systems:

4. The first step in understanding behavior is to identify the perceptual
    signal(s) that the behavior controls. Doing so requires applying "The
    Test," which involves applying disturbances to the putative perceptual
    signal(s) and observing whether (and how well) the disturbances are

Very nice work.