Classical Conditioning and PCT

[From Bruce Abbott (2011.02.19.1510 EST)]

BP: Bill Powers (2011.02.18.1455 MST)] –

BA: Bruce Abbott (2011.02.17.2250 EST)

I’m sorry that what appears below is a bit long-winded, but I thought it necessary to provide a detailed, point-by-point explanation or clarification of my position or rebuttal of yours, showing both where I agree and disagree, in order to be as clear as possible about my position.

BA: Certain environments will force a change in organization somewhere in the hierarchy of control systems. For example, if the old means of control are unavailable, one must find new ones or give up control over that perception.

BP: This is true in a broad sense, but not in detail. The environment by itself can’t force any organism to do any specific thing.

I’m not claiming that it does.

It has no means for doing so. Martin Taylor says, and I agree, that all the environment can do is give properties to any particular environmental feedback function, and apply disturbing forces to environmental variables.

Yes, of course.

It is the organism that determines what variables it will sense, and then how it will try to control them.

This statement raises a host of questions for me. How does the organism determine what variables it will sense? Can it decide not to sense them? How does it determine how it will try to control them? Where do its options come from? What does the determining?

Once that determination is made, a controlled variable has been defined. Now we can see environmental variables in terms of their effects on this new controlled variable; we can see them as disturbances of the CV, or of the connection between the organism’s action and consequent changes in the CV.

Yes.

If the organism reorganizes, the same environment will have different effects on an altered CV,

(if reorganization changes the input function)

and actions will change to oppose those different effects even though the environment is the same.

Or to put it differently, observed behavior will change.

If you don’t know what the CV is, you can’t tell what it is in the environment that matters.

Yes; without appropriate tests, you don’t know why exposure to this particular environment resulted in the observed changes in behavior.

BA: We are who we are, to a great extent, because of what happened to us when we interacted with parents, the kids on the block, our classmates, teachers, and many others. We learned to trust or suspect, to love or hate, to feel confident in our abilities or unsure of ourselves.

BP: I think we have to clean out this little pocket of SR thought. You seem to be describing personality characteristics that are the result of what happened to us, in a cause-effect way.

Well, yes, but not in an S-R way. The causal sequence runs through the reorganization system’s random change and selective retention process. The relatively stable changes that result from this process emerge from an interaction between the environmental inputs (disturbances to controlled variables, changed environmental feedback functions, etc.) and the current organization of the control hierarchy.

If that were true, all people would be affected the same way by the same events, and they aren’t.

That isn’t true, even from an S-R perspective. Everyone is different enough from everyone else that one cannot expect the same response to the same inputs.

You are actually describing ways in which people reorganize to prevent external events from having adverse effects on them, and to increase the occurance of beneficial effects.

Yes, and that comes as no surprise to me, because that is exactly what I intended to describe. Perhaps I wasn’t clear enough about that.

At the same time these influences are felt, we are reorganizing our behavior so as to change the influences and their effects on us.

Well, “we” aren’t reorganizing our behavior (as if there is some little person inside there doing these things), our nervous system is reorganizing ITSELF so as to change the influences and their effects on us.

What we retain is not the effects of those influences, but the organization we created and then kept changing to improve control or regain control where it was lost.

I’d replace the second “we” by “the reorganization process,” but otherwise find nothing inconsistent here with what I had said about this process previously.

BP: What we want from life is thus protected against such influences, the more so as we learn better methods of control. Others do not determine what our basic needs and desires are. Those come from our nature as human beings which has been changing and expanding since before we were human.

When you state these things in this context, it sounds like you are offering a rebuttal to something I asserted, which I never did.

BP earlier: It seems to me that learning is a very difficult achievement that requires a lot of organization to accomplish; without this organization, evolved over a few billion years, exposure to “certain things” would have no more systematic effect on an organism than it would have on a rock or a bowl of jello. Nothing happens just because of “exposure” to something. Saying that it does seems to me like a way to minimize the need to understand the inner mechanisms that organisms have developed over history for controlling the world around them and inside them.

BA: I agree. Does that surprise you? If predictable changes in behavior occur under certain well-defined conditions, one has demonstrated that such conditions are sufficient to bring about those changes.

BP: No one hasn’t. That shows only that the conditions include some feature that disturbs something which a particular organism is controlling. Another organism may show similar changes under entirely different circumstances. You are forgetting that both reorganization and natural environments have a large random component. Repeating the same condition will not bring about the same changes, and even if it did, normal changes elsewhere in the environment see to it that what is needed to repeat a desired effect is most likely to be a different behavior.

I do agree with you here that I overstated the case – but not as much as you suggest. Remember, I was referring specifically to the process of classical conditioning. It’s a relatively primitive process that piggybacks onto previously established control systems, many of which are innately organized rather than having been created through reorganization (e.g., we don’t LEARN to salivate when some substance enters the mouth). Classical conditioning has been demonstrated in relatively primitive organisms such as sea slugs (aplysia) and nematodes (flat worms with a nervous system of fewer than 200 neurons), as well as complex ones such as ourselves. Pair a tone and food in the right way for a few dozen trials and every normal dog will begin to salivate to the sound of the tone. After finding that one cannot prevent the playground bully from beating him up, a normal child will undergo a relatively predictable set of physiological changes the next time he sees the bully approaching. There’s a lot more consistency in the changes that follow such experiences than you are willing to grant.

BA: Pavlov’s dogs always began to salivate to the sound of that metronome after that sound had been followed by the delivery of meat powder into the dog’s mouth. Whether such
an arrangement of the dog’s environment brings about such changes in a normal animal is not a matter for debate; the question to be addressed is how: what is the mechanism?

BP: The dog may salivate when there is no food powder blown into its mouth, and it may fail to salivate when the metronome ticks. “Always,” as you know, is an exaggeration or idealization of the truth.

As for the first part, that a dog may salivate for other reasons than food in the mouth is true but not relevant. I did not claim that the sound of the metronome would gain EXCLUSIVE access to salivation. As to the second part, that the dog may fail to salivate when the metronome ticks is true, because the internal state of the dog may be different from its normal state at that moment. So I grant you that “always” is an idealization. I didn’t mean to imply that every tick of the metronome would produce salivation, but that every normal dog exposed to this procedure would reach a point where the metronome’s sound (ticking at a given rate) would become a reliable elicitor of salivation. [There are known ways to block this result; here I’m discussing results with normal dogs who have not been exposed previously to these other conditions.]

In order to get what looks like predictable control, it is necessary to overpower an animal and prevent it from satisfying its own needs on its own schedule, and limit the environment so that one and only one action will have the desired result - a most rare and artificial circumstance in nature.

That’s certainly not the case for classical conditioning. We are not talking here about a process in which you disturb some controlled variable and then restrict the animal’s options for correcting the resulting error. We are talking about a process in which the animal learns about relationships among events and conditions in its environment, as the animal perceives them, and takes advantage of them to improve control over certain of its controlled perceptions.

As I think I remarked once in a Skype conversation with you, I can cause you to pick up a spoon by first serving you soup that I know you like at lunchtime when I know you will be hungry, then making sure there is no handy utensil but the spoon and preventing you by some means from bending down and slurping the soup out of the bowl or picking up the bowl and drinking from it. You must already have nearly complete physical control over an animal to make it do any arbitrary thing you want. That is why you see cages in experimental laboratories.

I don’t quite agree with you here, either (I don’t have to deprive a dog of anything, nor administer any form of punishment, to teach it to heal, sit, or fetch), but since we are discussing here a proposed explanation for classical conditioning (within the PCT framework, or so I believed), and not operant conditioning, the last few paragraphs of your post are simply irrelevant to this conversation.

In my reply to Boris Hartman, I tried to explain how, within the PCT framework, certain correlations among environmental events or conditions, as experienced by the individual, can lead to relatively predictable changes in behavior of the sort labeled as classical conditioning. My explanation was based on my understanding of your own proposal for explaining classical conditioning. I asserted that an implication of that proposal is that conditioning will occur under those conditions that foster it, whether the person wants it to happen or not.

Apparently you disagree with that conclusion. One line of attack was to deny that the process of classical conditioning is capable of producing relatively consistent outcomes, but this assertion must be weighed against overwhelming evidence to the contrary. Another was to assert that the explanation represents S-R thinking. But that assertion will not stand up to an examination of what I actually stated, which was phrased completely in terms of the PCT processes of control and reorganization. So, either you have misunderstood what I was trying to say, or I have missed something important. I don’t see where I’ve gone off the tracks, but I’m willing to be educated.

Bruce A.

[From Bill Powers (2011.02,.19,.1350 MST)]

BP earlier: …Martin Taylor says, and I agree, that all the environment
can do is give properties to any particular environmental feedback
function, and apply disturbing forces to environmental variables.

BA: Yes, of course.

BP earlier: It is the organism that determines what variables it will
sense, and then how it will try to control them.

BA: This statement raises a host of questions for me. How does the
organism determine what variables it will sense? Can it decide not to
sense them? How does it determine how it will try to control them? Where
do its options come from? What does the determining?

BP: This may sound a bit more acceptable if I say “processes
intrinsic to the organism, not in the external world, determine what
variables the organism will sense…” and so on. I don’t mean to
imply that we are conscious of all such processes, though we do
consciously experience some effects related to intrinsic error.

I have been thinking for a long time about how the organism decides what
variables to create at one level given variables at the level below. That
boils down to the question, “Why perceive anything?” The
generic answer under reorganization theory is, “To minimize
intrinsic error.” But that answer is too vague and general.

In the reorganization demo 7-2, all the reorganization is in the output
function. The input weghts are set at random, and then the output weights
are reorganized to minimize overall error. That criterion works fine for
reorganizing the output weights. It works, apparently, for any random
selection of input weights (proving that was the point, originally, of
this demo). This shows that however the inputs are organized, a control
system can reorganize to achieve good control, even very close to a
condition of conflict.

When the input weights are selected to be close to a conflict situation
(determinant near zero), we find that that output excursions needed to
maintain control get very large – a lot of energy is being consumed and
wasted and control is relatively poor (though still quite good). This
suggests that we might use the average absolute output magnitude as a
criterion for reorganizing the input weights. If the output magnitude is
minimized, we will get input weights with the least amount of inherent or
incipient conflict. The perceptions will become relatively independent of
each other, which is tantamount to saying that they come to represent
invariants of the input space. That might well be how perceptual input
functions are formed. We have a lot of work left to do in that direction.

BP earlier: Once that
determination is made, a controlled variable has been defined. Now we can
see environmental variables in terms of their effects on this new
controlled variable; we can see them as disturbances of the CV, or of the
connection between the organism’s action and consequent changes in the
CV.

BA: Yes.

BP earlier:If the organism
reorganizes, the same environment will have different effects on an
altered CV,

BA: (if reorganization changes the input function)

BP earlier: and actions will change to oppose those different effects
even though the environment is the same.

BA: Or to put it differently, observed behavior will
change.

BP: It’s not always easy to observe what has changed. In the first demo
of LCS3, we deliberately arranged for the behavior needed to control the
three aspects of the red ball to be the same regardless of what was to be
controlled. The position, the shape, and the orientation are all affected
at the same time by moving a slider on the screen. All that is different
about the observed behavior when the object of control is changed is
which pattern of disturbance is symmetrically related to the slider
movements. There is no principle requiring any given disturbance to be
applied to a given aspect (the same pattern might be used in all three
cases, or the patterns could be switched around from time to time). So
just watching a person’s behavior can’t tell us what the person is
doing.

BP earlier: If you don’t know
what the CV is, you can’t tell what it is in the environment that
matters.

BA: Yes; without appropriate tests, you don’t know why exposure to this
particular environment resulted in the observed changes in
behavior.

BP: But exposure to this particular environment didn’t “result
in” (cause) the behavior. Exposure is only exposure, not an
influence. You can’t tell, just by looking at the environment, which of
the changing variables you can see is relevant and which is irrelevant,
if you haven’t figured out the organism’s purpose in generating the
behavior, or which effect of the behavior was intended and which other
effects were accidental.

BA: We are who we are, to a
great extent, because of what happened to us when we interacted with
parents, the kids on the block, our classmates, teachers, and many
others. We learned to trust or suspect, to love or hate, to feel
confident in our abilities or unsure of ourselves.

BP earlier: I think we have to clean out this little pocket of SR
thought. You seem to be describing personality characteristics that are
the result of what happened to us, in a cause-effect way.

BA: Well, yes, but not in an S-R way. The causal sequence runs through
the reorganization system’s random change and selective retention
process. The relatively stable changes that result from this process
emerge from an interaction between the environmental inputs (disturbances
to controlled variables, changed environmental feedback functions, etc.)
and the current organization of the control hierarchy.

BP: But the random changes remove any direct causal connection between
events in the environment and the changes in behavior. The changes are
indeed random. The only systematic relationship is between the behaviors
we see changing and disturbances of the consequences of the
behavior, and then only those consequences that the organism is
controlling. The behavior changes so as to keep and improve control of
the consequences (and not, as Skinner proposed, the other way around –
the consequences do not change so as to keep the behavior the
same).

BP earlier: If that were true,
all people would be affected the same way by the same events, and they
aren’t.

BA: That isn’t true, even from an S-R perspective. Everyone is different
enough from everyone else that one cannot expect the same response to the
same inputs.

BP: Behaviorists may say that, but they don’t mean it. A
reinforcing event is assumed to increase the probability of the behavior
that produced it. The whole idea of stimulus control of behavior, whether
you mean reinforcing stimuli or discriminative stimuli, is predicated on
getting the same result from presentation of the same stimulus. This is
much like the way other psychologists keep saying “CORRELATION DOES
NOT IMPLY CAUSATION”, and then (as most of them do) blithely
proceeding to reason as if it does.

BP earlier: You are actually
describing ways in which people reorganize to prevent external events
from having adverse effects on them, and to increase the occurance of
beneficial effects.

BA: Yes, and that comes as no surprise to me, because that is exactly
what I intended to describe. Perhaps I wasn’t clear enough about
that.

I know you’e not surprised. You understand control theory very well. But
there are some logic-tight compartments in there which are in need of
being breached. I am trying to make a noise like an iceberg.

At the same time these
influences are felt, we are reorganizing our behavior so as to change the
influences and their effects on us.

Well, “we” aren’t reorganizing our behavior (as if there is some little
person inside there doing these things), our nervous system is
reorganizing ITSELF so as to change the influences and their effects on
us.

BA: But there is a person inside there doing these things. It is shaped
like a brain and it has experiences exactly like the ones we have. This
person is a composite of many systems, even many selves, but the point
here is that these systems are producing the control and the
reorganization. The environment has no guiding or determining effect on
these processes at all.

BP earlier: What we retain is
not the effects of those influences, but the organization we created and
then kept changing to improve control or regain control where it was
lost.

BA: I’d replace the second “we” by “the reorganization process,” but
otherwise find nothing inconsistent here with what I had said about this
process previously.

BP: Good. So it’s just the language you sometimes use that is
inconsistent with what I’m saying. The reorganizing process is, of
course, one aspect of the organization I call myself. This brain which I
occupy, or am, reorganizes itself.

BP earlier: What we want from
life is thus protected against such influences, the more so as we learn
better methods of control. Others do not determine what our basic needs
and desires are. Those come from our nature as human beings which has
been changing and expanding since before we were human.

BA: When you state these things in this context, it sounds like you are
offering a rebuttal to something I asserted, which I never
did.

BP: Then just take it as a reminder. “The reorganizing process”
is part of ourselves, not some mechanical device through which the
environment operates us. I reorganize; the environment does not
reorganize me. That is the important distinction here.

BP earlier: It seems to me that
learning is a very difficult achievement that requires a lot of
organization to accomplish; without this organization, evolved over a few
billion years, exposure to “certain things” would have no
more systematic effect on an organism than it would have on a rock
or a bowl of jello. Nothing happens just because of
“exposure” to something. Saying that it does seems to me like
a way to minimize the need to understand the inner mechanisms that
organisms have developed over history for controlling the world around
them and inside them.
BA: I agree. Does that surprise you? If predictable changes in behavior
occur under certain well-defined conditions, one has demonstrated that
such conditions are sufficient to bring about those changes.
BP: No, one hasn’t. That shows only that the conditions include some
feature that disturbs something which a particular organism is
controlling. Another organism may show similar changes under entirely
different circumstances. You are forgetting that both reorganization and
natural environments have a large random component. Repeating the same
condition will not bring about the same changes, and even if it did,
normal changes elsewhere in the environment see to it that what is needed
to repeat a desired effect is most likely to be a different
behavior.

BA: I do agree with you here that I overstated the case – but not as much
as you suggest. Remember, I was referring specifically to the process of
classical conditioning. It’s a relatively primitive process that
piggybacks onto previously established control systems, many of which are
innately organized rather than having been created through reorganization
(e.g., we don’t LEARN to salivate when some substance enters the mouth).

BP: I think that all control systems are created through reorganization,
either in a single lifetime or over many of them.

BA: Classical conditioning has
been demonstrated in relatively primitive organisms such as sea slugs
(aplysia) and nematodes (flat worms with a nervous system of fewer than
200 neurons), as well as complex ones such as ourselves. Pair a tone and
food in the right way for a few dozen trials and every normal dog will
begin to salivate to the sound of the tone. After finding that one cannot
prevent the playground bully from beating him up, a normal child will
undergo a relatively predictable set of physiological changes the next
time he sees the bully approaching. There’s a lot more consistency in the
changes that follow such experiences than you are willing to
grant.

BP: I think that we, as PCT theorists and experimenters, have a right to
our own definition of consistency and good demonstrations of facts. You
say, “every normal dog will begin to salivate at the sound of the
tone.” But you mean “Every normal dog that is extremely hungry
and incapable of escaping from the apparatus in which its behavior is
being observed”, which means it is not a normal dog. And of course
if the dog does not salivate, it is not normal in the opinion of someone
who believes that all normal dogs do so. It could still be a normal
dog.
BA: Pavlov’s dogs always began to salivate to the sound of that metronome
after that sound had been followed by the delivery of meat powder into
the dog’s mouth. Whether such an arrangement of the dog’s environment
brings about such changes in a normal animal is not a matter for debate;
the question to be addressed is how: what is the mechanism?
BP: I disagree: what we need to debate is whether an arrangement of the
dog’s environment “brings about such changes” at all. It is
more probable, in my opionion, that changes in the environment
never bring about any changes in a normal animal unless through
direct mechanical or chemical effects. Of course the term “bring
about” may be interpreted to mean “cause,” or merely
“set the occasion in which” a process is often observed. But
falling back on such ambiguity does not make for an interesting
discussion, since whichever way the fickle wind of proof blows, one can
always say he intended the other meaning. If “bring about” is
not meant to imply causation, then causation is admitted to lie
elsewhere; if it is meant to imply causation, then the assertion is
subject to easy refutation.

BP: The dog may salivate when
there is no food powder blown into its mouth, and it may fail to salivate
when the metronome ticks. “Always,” as you know, is an
exaggeration or idealization of the truth.

BA: As for the first part, that a dog may salivate for other reasons than
food in the mouth is true but not relevant. I did not claim that the
sound of the metronome would gain EXCLUSIVE access to
salivation.

BP: I would not say that the sound of the metronome “gains
access” to anything at all. That is a figure of speech left over
from the days when behavior was considered to be under the control of the
the environment, with the environment often being given properties close
to sentience and purpose in order to avoid giving those properties to
organisms.

BA: As to the second part, that
the dog may fail to salivate when the metronome ticks is true, because
the internal state of the dog may be different from its normal state at
that moment.

BP: It’s easy to imagine such an internal state, but not easy to verify
it. And it’s all too easy to believe such explanations for failure of the
hypothesis, if one is convinced that the hypothesis simply has to be
correct.

BA: So I grant you that “always”
is an idealization. I didn’t mean to imply that every tick of the
metronome would produce salivation, but that every normal dog exposed to
this procedure would reach a point where the metronome’s sound (ticking
at a given rate) would become a reliable elicitor of salivation. [There
are known ways to block this result; here I’m discussing results with
normal dogs who have not been exposed previously to these other
conditions.]

BP earlier: In order to get what
looks like predictable control, it is necessary to overpower an animal
and prevent it from satisfying its own needs on its own schedule, and
limit the environment so that one and only one action will have the
desired result - a most rare and artificial circumstance in nature.

BP: That’s certainly not the case for classical conditioning. We are not
talking here about a process in which you disturb some controlled
variable and then restrict the animal’s options for correcting the
resulting error. We are talking about a process in which the animal
learns about relationships among events and conditions in its
environment, as the animal perceives them, and takes advantage of them to
improve control over certain of its controlled
perceptions.

BP: You most certainly do have to restrict the animal’s options for
correcting error. How do you get an unrestrained animal to patiently wait
while it is repeatedly disturbed enough for it to try (unsuccessfully) to
make a corrective move? How can you expect to condition its behavior
successfully if its reaction to the US succeeds immediately in correcting
the error, so there is never enough error to call for reorganization? Or
if there never was an error to begin with?

BA: Apparently you disagree with
that conclusion. One line of attack was to deny that the process of
classical conditioning is capable of producing relatively consistent
outcomes, but this assertion must be weighed against overwhelming
evidence to the contrary. Another was to assert that the explanation
represents S-R thinking. But that assertion will not stand up to an
examination of what I actually stated, which was phrased completely in
terms of the PCT processes of control and reorganization. So, either you
have misunderstood what I was trying to say, or I have missed something
important. I don’t see where I’ve gone off the tracks, but I’m willing to
be educated.

BP: It’s all in the language. My position is that any description that
gives control to an environment that is not a control system is wrongly
stated. To say that classical conditioning produces anything,
consistently or not, is to imply it is a causative agent. My point
is that classical conditioning is neither an agency nor a causal process;
it is a way of describing an interaction so as to minimize the organism’s
part in it, and imply the greatest causal control by the environment.
This is the legacy of behaviorism.

I believe the correct description of behavior leaves the environment in
the position of the stage upon which the actor acts, perhaps full of
sound and fury in the acting but signifying, nevertheless, something
never understood before.

Best,

Bill P.