From Jason Gosnell (2005.06.07 15:25 CST)
[From Bill Powers (2005.06.07.0758(]
What this means is that we have to revise all the basic
definitions, particular those of “behavior” and
“reinforcement.” Observable behavior means either the actions the
organism is producing to control something else, or it means the something else
that is under control and not the actions. Reinforcement does not exist – that
is, things and events in the environment do not have special effects on the
organism that cause it to produce the right class
of behaviors. What a person means by “feeling reinforced” is only
that something has helped the person get something the person is trying to get.<<
Thanks for taking the time to churn this
out. I can see why I am losing my behaviorist understanding and why I can’t
seem to regain it-it feels wrong to me and when I try to step into it I
feel delusional somehow. I want to make sure that I give it it’s due so I
am going to see if I can translate it into the PCT model for myself. The main
issue for me is still the idea of external control in behaviorism and how that
really conflicts with my actual experience of living in the world. As a model,
I get behaviorism, in my experience, for me, it doesn’t hold up.
I really think that if one studies PCT,
one’s consciousness has a chance to actually shift in a deeper way. I
realized a similar understanding when I went to a Zen retreat-they sit still
in meditation facing the wall. The strictist form is that you just sit-you
don’t do anything but just sit there and notice what comes up in
awareness. Anyone can do this with practice. What I realized was that I had
brought the entire world with me-the world wasn’t exactly my
problem. So, I had already been making errors based on my view and orientation
to “the world” out there-blaming errors mostly. That so and
so is my problem is a very incomplete understanding. Normally, the mind is very
involved in these external causes of our actions/feelings/thoughts/desires/etc.
Only in reflecting inwardly, so to speak, does one discover other processes at
work and the larger truths behind them. Strictly speaking, no one else can
really be my problem. Or, you could say, that is a rather one-sided, distorted
view, but common amongst humans.
Regards…Jason
[mailto:powers_w@FRONTIER.NET]
AM
the Phenomenon Real?
[From Bill Powers (2005.06.07.0758(]
Hank Folson (2005.06.06.1800) –
Some very nice suggestions in your post. In fact it was an excellent piece of
writing, which has inspired me. I think you’ve almost got the answer, but as
Martin Taylor showed, the distinction isn’t quite right yet. Yes, people do see
ongoing processes as behavior as Martin says, but this doesn’t mean they
understand the PCT view. Pardon me while I emit an essay.
One step onward might be to think of named
behaviors. When a behavior gets a name, we expect to see it more than once. Oh,
there’s that pirouette again, there’s that bar-press again, there’s that
getting-an-A again. To name a behavior we have to have something in mind beside
a continuous flow of one configuration into another, one relationship into
another, one program into another. The Eastern philosophers, I think,
recognized this continuous flow of experience and being as the real nature of
the world, while it is we who insist on dividing the flow into “things”
and “events.” Of course the flow, too, is a product of perceptual
functions – it’s not that easy
to decide what is the reality and what is created by the observer.
So anyway, named behaviors imply repeated perceptions, and of course since all
perceptions can vary along a scale from 0 to max, it implies a specific level
for the recognized perception. Something has occurred “again,” which
means you’re remembering a previous occurrance, comparing it with the present
one, and noting that the two perceptions match; they are not only of the same
kind, but they occupy the same place on their scale of variation. This is not
comparison in the sense of a comparator, however: it’s perception of a
particular state of the relationship called “similarity.” The way you
can verify this is to set the reference for the similarity perception to some
other state. Do the piroutte “again”, but about half as fast. Now
you’re maintaining a specific difference between the present and past
perceptions, showing that exact duplication is only one point on a scale of
similarity (this is much like tracking: you can keep the cursor a fixed
distance away from the target, as well as on it).
That diversion was called for to separate a reference level for a repeated
perception from the source of the
reference level, which can sometimes
be the desire to create a particular degree of similarity between a past
perception and a present one. Once one has chosen the desired degree of
similarity, the next stage of the process is to adjust the reference level for
the present perception so as to bring that perception (by acting or imagining)
into the desired relationship with the remembered one: similarity, or a
specific degree of dissimilarity in some specific dimension.
Now back to the point. A named behavior or action implies that an observer sees
the same behavior or action often enough (and near enough the same) to merit a
name. And for that to happen, that behavior must be under control, because
normally the actions we see another organism performing are varied (not
controlled) as a means of controlling other perceptions. Only under special
circumstances will we see the action itself being controlled relative to the
same reference level, again and again.
What are those circumstances? Obviously, since behavior or action itself is of
no interest to the control system generating it, the circumstances must be such
as to require that a specific kind and degree of action be produced in order to
control some other perception, the one of interest to the control system. When
that is the case, the organism will end up generating that specific behavior,
as long as it (still) produces the result that the organism (still) wants to
experience.
What we see as named behaviors, therefore, must be the control of some specific
action that is required to produce an effect that the organism wants to
experience. Note the implication about the environment: the environment must be
such that repeating a given action will in fact produce the same effect every
time, or often enough. Not many environments have that property. Special
environments are required, in which nothing can alter the relationship between
the action and its effect. Sometimes such environments occur naturally, but in
the laboratory they are created mainly by protecting against disturbances.
So: Named behaviors are seen in environments where producing the same output
action is required to produce an effect that the organism is trying to control.
The implication is that the rest of the time, when (as is normal) many
different actions are required on different occasions to produce the result
that the organism wants, we do not see any named behaviors: we just see
“activities” of a nonspecific sort. This is what we see “between
behaviors.” Behavior or action, of course, never ceases while we are
awake: it just varies. We may not see anything happening when no disturbances
are varying, but the control systems are still turned on and still generating
whatever action is required to maintain all controlled variables at their
reference levels.
In my opening examples I included doing a pirouette in the same list with
getting an A (the top mark in American schools). But there is a subtle
difference here. We can see that the pirouette is a direct and immediate
consequence of what the dancer’s muscles are doing, so much so that it is no
surprise to learn that the pirouette is a controlled result of muscle action.
But getting an A is not quite as clear because the A is awarded by someone
else, and the actions needed to get one are extremely variable, often depending
on who is giving the grade. To see getting an A as a named behavior is
therefore not quite so straightforward as seeing performing a pirouette that
way.
Getting an A, in fact, is a very clear example of varying one’s behavior as the means of obtaining a desired
outcome. We don’t attribute the professor’s writing down of an ‘A’ to an effect
of the student’s muscles. It is an effect of the professor’s muscles. What the
student does is study, or perform lab work, or suck up to the professor, or
cheat. Those behaviors are not to be confused with the professor’s writing down
of a letter of the alphabet. There is a causal connection, presumably, but we
do not confuse the effect with its cause.
However, in other circumstances this confusion is common. We label the rat’s
behavior as “pressing the bar.” By that we mean (on reflection) that
by some action or other, the rat gets the bar to move far enough to close a
pair of electrical contacts. This is a well-known problem, which B. F. Skinner
neatly bypassed by naming the behavior of the rat in terms of its consequences.
He said that bar-pressing behavior is the class of all actions that depress the
bar (bar, lever, whatever). Furthermore, by that process of definition, a stimulus
is the class of all events that have the effect of producing that class of
actions, so we have a class causing a class rather than a physical process
causing a physical process. The result of this is that as long as the
experimenter can see that the right kind
of effect follows from the right kind
of cause, the same generalizations apply. If the lever goes down again, the rat
has emitted the “same behavior” again, even though it has actually
produced a completely different action, or even the exact opposite action in
terms of its immediate physical effects (moving right to press the lever as
opposed to moving left if the rat happens to be on the other side of the
lever).
This is the Achilles heel of behaviorism. If we lift this intellectual patch to
see what it covers, we find that physical causation is seriously violated by
the concept of “classes” of behaviors. We find totally different,
even opposite, physical causes having the same physical effect. What we find,
in fact, is that somehow the organism is able to vary its behavior, the
physical causes it produces, so that when they are combined with other physical
processes and influences that have changed, the sum still comes out the same.
As far as any premises of behaviorism are concerned, this can happen only by
magic. What we find, in fact, is that the actual, physical, behavior that the
organism produces does NOT repeat
when it is reinforced.
Of course that is precisely the situation that PCT explains. Finally we get to
the idea of reinforcement. The idea of reinforcement is that through slow
evolutionary processes organisms have come to respond in a particular way to
certain environmental things and events, as if those things and events have
special properties. The “particular way” is to increase the
probability of the behavior that produced one of those special things or
events, called therefore “reinforcers.” “The behavior”, of
course, means (according to the intellectual patch) not the actual physical
motor behavior that is produced, but the effect
that is produced by whatever physical action occurs. The effect, of course has
to be that particular effect which the experimenter has arranged to cause the
reinforcing thing or event to appear in the organism’s environment.
When we lift the patch. what we see is that in this situation the action that
is made more probable is any action that
produces the specific effect that produces the reinforcer. In other
words, what is reinforced can be muscle tensions that move the rat to the left
just as often as the opposite tensions that move it to the right. How the
reinforcer chooses which of these
opposing members of the same class of behavior to encourage is not explained,
particularly since the choice might have to change with every reinforcement.
Having created this problem, behaviorists then found another patch to cover it.
It is the idea of the “discriminative stimulus.” If the environment
changes in such a way that a different action is required to produce the
required effect, the depression of the bar, then the compensating changes in
the organism’s actions must have been caused by stimuli from those same changes
in the environment. This idea was picked up by many psychologists, even by the
cyberneticist W. Ross Ashby who proposed it as a model of behavior superior to
the control-system model.
But this just creates another problem, which is how the disturbances of the
outcome of behavior manage to create exactly those stimuli that will act on the
organism’s nervous system in exactly, quantitatively, the right way to keep the
outcome the same. It’s not enough that the stimulus caused by being to the left
of the lever should cause the rat to move to the right; it must cause a
movement to the right by the right distance,
so the rat’s paw does not come down before it gets to the lever or after it has
passed it. And what of the case in which the change in the environment does not
produce any stimuli to the rat – for example, when the strength of the spring
under the lever is changed, so a greater effort is required to make the
contacts close? All one can do then is stubbornly insist that there must have
been some stimulus of the
required kind – otherwise, how could the correct outcome have been produced?
Of course that answer begs the question – that is, it assumes that the premise
that discriminative stimuli explain the result is correct, in order to conclude
that some discriminate stimulus must explain the result. If that premise is
wrong, then of course insisting that a discriminative stimulus must have occurred
is also wrong.
The premise is wrong, as we can
easily prove with experiments in which the environmental change, inside a
computer running the experiment, is guaranteed not to generate any stimulus to
the participant. People do not need to sense the cause of a disturbance to
compensate almost perfectly for it. What is going on is something entirely
different from what the behaviorists imagined. It’s what we in PCT call
control.
What this means is that we have to revise all the basic definitions, particular
those of “behavior” and “reinforcement.” Observable
behavior means either the actions the organism is producing to control
something else, or it means the something else that is under control and not
the actions. Reinforcement does not exist – that is, things and events in the
environment do not have special effects on the organism that cause it to
produce the right class of
behaviors. What a person means by “feeling reinforced” is only that
something has helped the person get something the person is trying to get.
You can’t believe in reinforcement, discriminative stimulu, operant (or any
other flavor of) conditioning, or classes of responses and at the same time
understand behavior as PCT shows it to us. The theory of behavior implied by
those old terms simply doesn’t hold water any more.
My good and valued friends who still think they are behaviorists, I’m sorry.
Behaviorism and PCT simply cannot coexist.
Best,
Bill P.
Jason Gosnell [06.04.05 1540CST]
Well, it has finally happened to me…I no longer understand behaviorism.
Deep in the CSGnet archive is a post I wrote in 1998 related to this subject:
BEHAVIOR: THE PHLOGISTON OF PERCEPTUAL CONTROL THEORY (PCT)
Phlogiston was the word used in an early combustion theory to describe a
substance believed to create heat when it escaped from a burning
material. When the modern theory of combustion was being developed in the
1700’s, the word phlogiston was not redefined, it was abandoned. Why
wasn’t it simply redefined? I suggest because in the new oxygen theory of
combustion, there was no part of the process that was comparable to the
original concept of phlogiston as something escaping unaltered from a
burning material. Thus the word phlogiston was simply abandoned. The word
phlogiston already had a commonly and deeply accepted definition for
everyone. Its continued use, however redefined, would only lead to
confusion and errors.
The word ‘behavior’ is the phlogiston of Perceptual Control Theory. It
should be abandoned.
A generic definition of behavior, compatible with most pre-PCT theories,
is this: Behavior is the end result of a process.
Behavior is the end result of a process. If this is a reasonable
generalization, the word behavior has intrinsic to any use of the word,
the idea that behavior relates to the completion of a process.
If behavior is generally defined as being the end result of a process,
can the word have any place in perceptual control theory? No, it cannot.
What is traditionally described as ‘behavior’ is an interpretation of
observable human activity. This is a reasonable approach only if one
has no awareness of negative feedback systems. According to PCT,
observable human activity is our only means to do what living control
systems do, which is to control our perceptions. Any observed activity,
then, is not the end result of a process, and so it is not “behavior”
as
the word is generally defined. Observed activity is a part of the
functioning of a control loop, specifically, the output. There is never
an ‘end result’ for control loops. Whenever there is no perceived error
signal, the control system does not turn off, it simply stops producing
outputs, as there is no perception that needs to be brought to a
reference level. Any attempt to use the word behavior in describing this
ongoing, unending loop process can only create ongoing, unending
confusion.
If we really believe in PCT, we cannot redefine the word ‘behavior’ for
use in PCT. Under PCT, we hold that whatever definition people have for
the word ‘behavior’, that definition is firmly established, and any
attempt to redefine the word will either be resisted as a disturbance, or
the word will still be understood as originally defined, interfering with
everyone’s controlling of perceptions, (viz. Fred Nickols (980802.0520)).
Think about this: How often will “behaviors” seem to match
internal
goals? Often enough to lead many to believe that ‘behavior’ can be the
basis of a psychology?
So what will happen when we say there is no such thing as behavior? For
sure, you get people’s attention. Big error signal! Then you have to say
that “behavior”, as non-PCTers know it is an illusion. An illusion
that
has to do with the way living control systems operate. And PCT holds that
people are living control systems. So where does the illusion come in?
The curse of PCT is that the functioning of healthy control loops can
look like a cause-effect response system. All of us were raised to
believe that what we observe in humans and lab rats is cause-effect
related. It seems so fundamental that we are observing a cause, and then
seeing an effect in response to the cause. “It is just common sense”,
we
say. What could be more obvious?
When we deny the existence of behavior in our perceptual control theory,
we force the listener to look for something different. What do we do
next? I suggest that we get the listener to agree that what they call
“behavior” is the end of a process. Now they have acknowledged, and
are
reminded of, what they believe. Then we can try and introduce the concept
of the endless loops of control systems. No beginning, no end, just
continual adjustment of controlled variables, or no activity if nothing
has been disturbed.
One thing I would hope will happen will be an awareness that the polarity
between PCT and all other psychologies is absolute. Only one can be right
(Okay, both can be wrong.)
We need to offer an alternative to the word ‘behavior’ that has meaning
in PCT, and does not have an inappropriate definition in the non-PCT
world. I suggest simply using the word “activity”. The PCT-unaware
can
see ‘activity’ as a partial description of ‘behavior’, as they know it.
The word, ‘activity’ does not include the concept of completion in either
PCT or non-PCT psychologies. Thus the definitions of the words ‘activity’
and ‘behavior’ are easy to keep separate. Of course, we should look for
other possible replacement words, in hope of finding something even
better. (I haven’t thought about whether your comments Fred,
(980802.0520) may disqualify ‘activity’.)
···
-----Original Message-----
From: Bill Powers
Sent: Tuesday, June 07, 2005 11:31
To: CSGNET@listserv.uiuc.edu
Subject: Re: reinforcement - Is
"Response to Stimulus: The Manipulation of Fanciful Interpretations of
Reality" Does this sound familiar? It is "Behavior: The Control of
Perception", restated as many people understand these words.
My purpose isn’t to play with the title of Bill Power’s book, but to move
on to other words that have problems in PCT. I think that when Bill wrote
the book, the word perception was generally used as Bill used it. Today,
I see many examples of people defining perception as a choice, a personal
- even arbitrary - interpretation. The tendency to define control as
manipulation has been discussed on CSGnet.
I close by suggesting that it is just as important to look at other words
that conflict with PCT. The circular arguments we see too often on CSGnet
are examples of the problems that arise. I have always wondered why
whenever someone offers an equation like a=b+c, each term is always
defined. But when we use words, there is seldom a definition offered,
even after it is clear to the participants that they are talking past
each other. I must mention that clear definitions will help lead to
resolution of conflicts only if that is what the participants are
controlling for. This is another curse of PCT!
Sincerely, Hank Folson
“Behavior” can exist in passive stimulus-response systems.
“Behavior”, as traditionally defined, can not exist in PCT. Many
PCTers do use the term, redefined in a way compatible with PCT. The problem for
all is that the traditional definition has been in their, and their listeners,
heads for decades. Interestingly, PCT says that the longer you hold an idea,
the more you will resist any attempts to change it. My original post was a big
“disturbance” -PCT term- and was resisted by some of those who are
living control systems…
“Behavior” is not a real phenomenon under PCT. Is
“reinforcement” another phenomenon that does not exist in Living
Control Systems? It makes sense for stimulus-response systems. For purposive
living control systems, it would have to fit into an existing hierarchical
system of control loops.
If the reinforcement is something the subject is controlling for, PCT says the
person (or lab rat) will perform whatever actions it has to in order to get to
its reference level for the specific reinforcement being offered (sometimes
limited when some other control loops develop higher priority error signals
than the original loops). The phenomenon is there, but it is a coincidence.
In the classic definition of “reinforcement” that runs through my
mind, the person offering the reinforcement has chosen the reinforcing item
because it is something they think will produce the “behavior” they
are looking for from the recipient. Successful reinforcement occurs only if the
reinforcement offered helps the recipient to get closer to some goal. If this
is so, “reinforcement” is a term that has little real-world use for
living control systems. Try putting 2 carat diamonds in the lab rat’s feed
chute, and see how well reinforcement works. Be careful, though. You may find
out what some lab assistants are controlling for. 
Sincerely,
Hank Folson
All electronic mail communications originating from or transmitted to Bridgeway Center, Inc. are subject to monitoring. This message and the information contained in it, which may consist of electronic data attachments, are the confidential and proprietary communications of Bridgeway Center, Inc. and are intended to be received only by the individual or individuals to whom the message has been addressed If the reader of this message is not the intended recipient, please take notice that any use, copying, printing, forwarding or distribution of this message, in any form, is strictly prohibited. If you have received this message in error, please immediately notify the Bridgeway Center, Inc. Privacy Officer at (850) 833-7540 and/or forward the message to hipaa@bridgeway.org and delete or destroy all copies of this message.
