PCT Island

[From Bruce Abbott (971230.1100 EST)]

Bill Powers (971230.0308 MST) --

Thought 1. I don't want to build a bridge between PCT and behaviorism. More
below.

Well, that clears the field of action for me, doesn't it?

Thought 2. Some facts discovered by psychologists will probably be
retained, but when PCT principles are applied many of them will look
different. For some reason, I thought of "just-noticeable differences." If
we replicated the experiments that showed this phenomenon, we would still
see it. But if we did a control-system experiment with the same
perceptions, we would not see it. S.S. Stevens actually thought that
perceptions were quantized; a control-system experiment would show that
they are not. Signal-to-noise ratio is not the same thing as perceptions
that occur in discrete steps. This is the sort of thing that would happen,
I believe, with most psychological "facts." When you look at them with
control-system experiments, they will look different.

The idea of the just-noticeable difference originated over 100 years ago
with Fechner, who suggested that there is a threshold below which a given
difference between two stimulus intensities could not be detected. I don't
know what Stevens thought about this idea. Fechner used the JND to
mathematically derive the form of the relation between stimulus intensity
and the experienced intensity of the stimulus (logarithmic). Stevens used a
somewhat different test procedure (magnitude estimation) from Fechner's and
found a power relation rather than a logarithmic one. Neither of these
continuous functions requires the assumption that the underlying perceptions
vary in magnitude discretely.

The standardized hearing test used to assess absolute thresholds for tones
of various frequencies is a simple tracking task. The person being tested
is asked to push a hand-held button until a tone is heard, then release it
until the tone disappears, then press until heard again, and so on. Holding
the button down slowly increases the tone intensity, whereas not holding it
down allows the tone to slowly decrease in intensity. The person ends up
oscillating the tone intensity just above and below the threshold, and the
average is taken to be the threshold for that frequency.

A very different analysis was applied in the early 1960s as an application
of signal detection theory, which was developed at Bell Labs. This analysis
disposes with the notion of threshold. Instead, one derives measures of
"sensitivity" (called D' [D prime]) and "bias." D' is essentially a measure
of signal-to-noise ratio and depends on the signal intensity, noise
intensity, and the observer's sensory capability. Like the individual
control parameters in tracking tasks, it is stable in the individual over
time, except for those changes due to aging or trauma. PCT isn't going to
change any of this.

That statement, if made by somebody with clout, would raise a furor: it
suggests that all psychological facts need to be re-examined, and that if
they were re-examined using the methods of PCT, their status as facts
would, in some unknown array of cases, suffer. Even to say that _some_
facts would change, if somebody of importance were to say this, would imply
that _all_ facts must be re-examined, because we don't know _which_ facts
would change. If the idea that some facts might change were ever accepted,
then psychology would have to be rebuilt from the ground up because nobody
knows now which facts those are. The only way to find out would be to redo
all the basic experiments, but with the possibility of controlled variables
in mind.

I don't buy these assertions. Control theory makes certain rather definite
statements about what facts ought to be observed if a particular variable is
being controlled. It ought to be rather easy to spot which supposed facts
would change, which would not, and which are simply not spoken to, if PCT
methods were to be applied.

This is why there is absolute resistance to PCT. The moment one admits that
there is a basic phenomenon that has been overlooked by a science, the
entire structure of that science is called into question. The only defense
is to reject PCT, or the idea that it could possibly make any significant
difference to psychology as a science.

It would be more accurate to say that the phenomenon has been
mischaracterized than to say that it has been overlooked. But however you
see it, I don't see how control theory will change the basic empirical facts
of psychology. Perception will still be perception, memory will still be
memory, personality will still be personality. Some explanations will
change, in some areas. For this reason, I do not believe that there is any
"absolute resistance to PCT." Rather, I think that there has been a certain
complacency. Current views seem to be adequate to handle the phenomena and
have served to guide and organize research efforts in a way that has seemed
fruitful. Why change?

Psychooogy is a SYSTEM of thought.
There is no way to change just one thing in a system.

Ah, would that this were so. (I am referring to the first sentence.)

In behaviorism, the reference level cannot be an independent variable. It
must either be a fixed physical property of the organism, or it must
somehow be a function of environmental variables. To allow the reference
level to be an independent variable inside the organism would be to admit
that every behavior is determined in part by the organism and _only_ in
part by its surroundings. And this means that every previous conclusion
that involved expressing behavior as a function of external conditions must
be called into question. A major independent variable has been left out of
consideration, and this independent variable does not correspond to any
observable environmental condition.

Again, I disagree with your chain of reasoning. The only thing behaviorism
asserts is that all conclusions be founded on observables. If you can
measure it, you can admit it. Consider hunger. Presumably hunger is a
function of numerous variables, including the levels of certain nutrients in
the blood, fullness of the stomach, and levels of certain circulating
hormones such as those released by fat cells. At present we don't know all
the factors, or how they combine to produce the hunger being experienced.
However, one _can_ show that an animal's willingness to work for food
increases (up to a point) with the number of hours since last meal, and
decreases in a systematic way with the amount of food ingested. In PCT
these changes in willingness to earn food can be understood via models in
which certain quantities are affected in certain ways by time-since-meal,
amount of food consumed, and many other variables, which result in certain
reference values or perceptual values being produced. But these models to
not "call into question" the observations, they serve to organize and
explain them, in a way different from before.

This omission shows up in subtle forms. One of them is the repeated attempt
by behaviorists to express behavior as a function of reinforcement and
discriminative stimuli. The form of the function that is always proposed is
such that it can be reduced to a single equation, with behavior on one side
of the equation and the environmental conditions on the other. But with the
analysis that takes reference levels and the effect of output on input into
account, we see that _two independent_ equations are always needed, one to
describe the feedback connection and the other to describe the actual
input-output function of the organism. No manipulation of the equations
describing only the observable environment can yield a unique prediction of
behavior.

_No manipulation of the equations describing only the observable
environment can yield a unique prediction of behavior_. That is the problem
with behaviorism, with its basic philosophy of external determination. That
is the gulf that can't be bridged.

And that is why I don't want to bridge it. To do so would be to attempt to
find common ground for the statements "The observable environment is the
ultimate determinant of behavior" on the one hand, and "The observable
environment is NOT the ultimate determinant of behavior" on the other hand.
The conflict between the PCT view and the behaviorist view is clear and
sharp. There is no way to reconcile not-A with A. And however complicated
the rhetoric gets, that is what the difference between PCT and behaviorism
boils down to: a direct contradiction.

The central assertion of behaviorism is not that the environment causes all
behavior, but that all behavior must be explained in terms of observables.
Until recently, most of the observables have been external to the organism,
i.e., out there in the environment, where they can be measured. This
viewpoint developed at a time when too many "explanations" invoked
unobservable mental events, whose only claim to reality was the very data
they were invented to explain, or invoked "physiological" mechanisms having
the same status. (The practice of invoking hypothetical physiological
mechanisms having no basis in observation came to be called, pejoratively,
"physiologizing.") Presently there are few behaviorists who would insist on
a strictly environmental approach, but all would insist that any
explanations be couched in terms that can be measured, either directly or
indirectly. If PCT were to be rejected by behaviorists, it would not be due
to its specification of references or other aspects of internal
organization. (I say "were to be rejected" because I don't think it has yet
been seriously considered.)

So, Bill, you can make PCT an island if you want to. Be my guest -- lay
under the palms and enjoy the surf. But I've got my bridge-building
construction set right here under my arm, and I intend to use it to connect
PCT with EAB, with or without you.

Regards,

Bruce

[From Bill Powers (971230.1012 MST)]

Bruce Abbott (971230.1100 EST)--

Thought 1. I don't want to build a bridge between PCT and behaviorism. More
below.

Well, that clears the field of action for me, doesn't it?

Go to it.

Thought 2. Some facts discovered by psychologists will probably be
retained, but when PCT principles are applied many of them will look
different. For some reason, I thought of "just-noticeable differences." If
we replicated the experiments that showed this phenomenon, we would still
see it. But if we did a control-system experiment with the same
perceptions, we would not see it. S.S. Stevens actually thought that
perceptions were quantized; a control-system experiment would show that
they are not. Signal-to-noise ratio is not the same thing as perceptions
that occur in discrete steps. This is the sort of thing that would happen,
I believe, with most psychological "facts." When you look at them with
control-system experiments, they will look different.

The idea of the just-noticeable difference originated over 100 years ago
with Fechner, who suggested that there is a threshold below which a given
difference between two stimulus intensities could not be detected. I don't
know what Stevens thought about this idea. Fechner used the JND to
mathematically derive the form of the relation between stimulus intensity
and the experienced intensity of the stimulus (logarithmic). Stevens used a
somewhat different test procedure (magnitude estimation) from Fechner's and
found a power relation rather than a logarithmic one. Neither of these
continuous functions requires the assumption that the underlying perceptions
vary in magnitude discretely.

Stevens got the idea that JNDs were additive, so he did experiments that
were designed to show steps in estimates of perceptual magnitudes. I don't
think anyone was able to replicate his results.

A JND is a matter of signal-to-noise ratio. If the amount of noise
increases as the signal increases, the size of a JND will appear to
increase even if the perceptual signal is perfectly linear with the
stimulus. The interpretation of the results as a logarithmic or power law
assumes that the slope of the perceptual input function is measured by the
JND, which is not true. There may be other reasons for deducing a nonlinear
input function, but the variation of JND with signal magnitude is not a
valid one.

The standardized hearing test used to assess absolute thresholds for tones
of various frequencies is a simple tracking task. The person being tested
is asked to push a hand-held button until a tone is heard, then release it
until the tone disappears, then press until heard again, and so on. Holding
the button down slowly increases the tone intensity, whereas not holding it
down allows the tone to slowly decrease in intensity. The person ends up
oscillating the tone intensity just above and below the threshold, and the
average is taken to be the threshold for that frequency.

This gives a measure of the lower threshold, but not the slope of the input
function. The slope makes as much difference to hearing sensitivity as the
threshold does, and perhaps more. Nevertheless, this is a good
control-system experiment (with an integrating environmental feedback
function), so its results would probably be retained. My impression is that
experiments in which the subject is given control of the stimulus will be
most likely to involve true controlled variables, and to survive any
re-evaluation.

A very different analysis was applied in the early 1960s as an application
of signal detection theory, which was developed at Bell Labs. This analysis
disposes with the notion of threshold. Instead, one derives measures of
"sensitivity" (called D' [D prime]) and "bias." D' is essentially a measure
of signal-to-noise ratio and depends on the signal intensity, noise
intensity, and the observer's sensory capability. Like the individual
control parameters in tracking tasks, it is stable in the individual over
time, except for those changes due to aging or trauma. PCT isn't going to
change any of this.

D' is probably what I mean by the slope of the input function, and "bias"
is the intercept, or threshold. I presume that these studies, too, involved
the "method of adjustment," which is likely to involve true controlled
variables. Of course a measure of slope, like all measures of subjective
perceptions, can be done only relative to some other measure of another
perception, barring actual measurement of neural spike trains. If all input
functions are nonlinear in a similar way, there is no way to detect that
common nonlinearity through behavioral experiments.

I don't buy these assertions. Control theory makes certain rather definite
statements about what facts ought to be observed if a particular variable is
being controlled. It ought to be rather easy to spot which supposed facts
would change, which would not, and which are simply not spoken to, if PCT
methods were to be applied.

That would be true if the original experimental reports specified whether
they were measuring disturbances or controlled input variables. But since
this distinction is not made in traditional experiments, most of the time
we can't tell. Also, don't forget that the PCT approach is NOT to assume
that there is always a controlled variable. The possibility always remains
open that we are observing a stimulus-response system. The only way to
discount that possibility is to look for controlled variables and test
them, and prove that they exist. If no controlled variable can be found, we
don't just assert that a control system is being observed. If controlled
variables are not mentioned, the only way to see if they were present would
be to do the experiment over.

The original experiments must always be done over if we can't determine
from the reports whether a controlled variable was actually being measured.
In method-of-adjustment experiments, we can make a pretty good guess as to
the controlled variable. But in all others, that's not so easy to do
without setting up the experiment and trying to replicate it.

This is why there is absolute resistance to PCT. The moment one admits that
there is a basic phenomenon that has been overlooked by a science, the
entire structure of that science is called into question. The only defense
is to reject PCT, or the idea that it could possibly make any significant
difference to psychology as a science.

It would be more accurate to say that the phenomenon has been
mischaracterized than to say that it has been overlooked.

How have controlled variables been "mischaracterized?" That implies that
they have actually been measured but have been called something else. I
don't think that's true. A controlled variable, after all, will show little
correlation with either the "stimulus" or the "response." Even if it
happens to be measured, it will be rejected by any ANOVA.

I don't see how traditional experiments could reveal controlled variables.
It takes a special technique to do this, and I haven't seen that technique
used in any of the traditional papers I've read.

But however you
see it, I don't see how control theory will change the basic empirical facts
of psychology. Perception will still be perception, memory will still be
memory, personality will still be personality. Some explanations will
change, in some areas. For this reason, I do not believe that there is any
"absolute resistance to PCT." Rather, I think that there has been a certain
complacency. Current views seem to be adequate to handle the phenomena and
have served to guide and organize research efforts in a way that has seemed
fruitful. Why change?

In terms of broad categories, sure, these words will not change. But the
way we understand their meanings will change radically. Just consider the
word "behavior." If behavior is the _outcome_ of actions, rather than the
actions themselves, and if it is defined primarily by perceptions, how can
the meanings not change?

As long as psycholgy takes the apathetic view that your intepretation
suggests, why change indeed? I don't think that playing nicey-nice is going
to get the mule's attention.

Psychology is a SYSTEM of thought.
There is no way to change just one thing in a system.

Ah, would that this were so. (I am referring to the first sentence.)

But it is. This isn't visible, perhaps, to an insider, because all the
piddling little arguments between schools capture attention. But behind
these little arguments there are huge areas of common agreement. Even Allan
Newell, who could hardly be classed as a stimulus-response behaviorist,
said that in the final analysis, an organism is just one gigantic response
function. Zenon Pylyshyn, in proclaiming his computational approach to be
"materialist," says that that of course beneath all the cognitive
manifestations, all that is operating is normal cause and effect connecting
inputs to outputs. The general agreement is that in the final analysis,
what organisms do is caused by what happens to them. This fundamental
postulate has endless ramifications; calling it into doubt changes everything.

Again, I disagree with your chain of reasoning. The only thing behaviorism
asserts is that all conclusions be founded on observables. If you can
measure it, you can admit it.

But this is precisely my point, and the point where PCT radically departs
from traditional "scientific" psychology. Observation alone is not
sufficient; you must propose a model describing what you cannot observe.
Without that model, all accounts of behavior are incomplete and even
misleading.

Of course modeling must be done in a rigorous way; the old "intervening
variable" approach is an untenable method because there is no way to rule
out any proposal. Skinner hammered away at such approaches all his life,
not realizing that there is another way to do it that works much better. He
thought that the only alternative to the intervening variable approach was
to avoid models altogether.

But these models to
not "call into question" the observations, they serve to organize and
explain them, in a way different from before.

If you can reduce the observations to statements of low-order perceptions,
perhaps what you say is true. But observations are hardly ever reported
that way. Even to call an environmental event a "stimulus" and a subsequent
change in behavior a "response" is to mix theory with observation. Just to
pick an extreme example, do you think that the observed relationship of ESP
to precognition contains no theoretical assumptions? That these
"observations" will hold up to scientific scrutiny? More to the point, if
one has been observing the effects of stimuli on responses, and those
effects turn out to have involved controlled variables all along, we might
be able to reinterpret the raw data, but who publishes raw data? Or even
keeps it? Remember what happened when you tried to get raw data fromn Staddon?

The central assertion of behaviorism is not that the environment causes all
behavior, but that all behavior must be explained in terms of observables.

This is what I am saying is impossible to do. You _must_ postulate
unobservables. There is no other way to propose a testable model.
Observables explain nothing. They are what need an explanation.

Presently there are few behaviorists who would insist on
a strictly environmental approach, but all would insist that any
explanations be couched in terms that can be measured, either directly or
indirectly.

To say "indirectly" is to say that we must propose a model. What models
have behaviorists proposed?

If PCT were to be rejected by behaviorists, it would not be due
to its specification of references or other aspects of internal
organization. (I say "were to be rejected" because I don't think it has yet
been seriously considered.)

Well, we'll see what happens when they do seriously consider it. What is
going to happen when they realize that what an organism accomplishes by its
behavior is determined primarily by what the organism wants to experience?
That most of the behavior they observe is simply an organism trying to push
back against disturbances?

So, Bill, you can make PCT an island if you want to. Be my guest -- lay
under the palms and enjoy the surf. But I've got my bridge-building
construction set right here under my arm, and I intend to use it to connect
PCT with EAB, with or without you.

OK. Just make sure that in the process, PCT doesn't get distorted beyond
recognition.

Aloha,

Bill P.

[From Bruce Abbott (971231.1550 EST)]

Bill Powers (971230.1012 MST) --

Stevens got the idea that JNDs were additive, so he did experiments that
were designed to show steps in estimates of perceptual magnitudes. I don't
think anyone was able to replicate his results.

The idea that JNDs are additive was Fechner's.

A JND is a matter of signal-to-noise ratio. If the amount of noise
increases as the signal increases, the size of a JND will appear to
increase even if the perceptual signal is perfectly linear with the
stimulus. The interpretation of the results as a logarithmic or power law
assumes that the slope of the perceptual input function is measured by the
JND, which is not true. There may be other reasons for deducing a nonlinear
input function, but the variation of JND with signal magnitude is not a
valid one.

I can see how both noise and signal increase with a system's amplification
(e.g., turning up the stereo), but in my own experience, noise in _my_
system (ears, brain) does not seem to rise with the loudness of the sound
I'm listening to. Does it in yours?

D' is probably what I mean by the slope of the input function, and "bias"
is the intercept, or threshold. I presume that these studies, too, involved
the "method of adjustment," which is likely to involve true controlled
variables.

D' is a relative measure of sensitivity, essentially the difference between
signal+noise and noise-alone distribution means as a proportion of the
variation about the mean -- or in other words, a measure of overlap of the
two distributions. "Bias" is the willingness of the participant to report
the signal as present when uncertain. Signal detection methodology allows
one to separate sensitivity and bias.

It would be more accurate to say that the phenomenon [of control] has been
mischaracterized than to say that it has been overlooked.

How have controlled variables been "mischaracterized?" That implies that
they have actually been measured but have been called something else. I
don't think that's true. A controlled variable, after all, will show little
correlation with either the "stimulus" or the "response." Even if it
happens to be measured, it will be rejected by any ANOVA.

Believe it or not, Bill, you are not the first person to notice that people
and other animals appear to have goals, intentions, purposes, wants, wishes,
etc. Lacking accurate knowledge about control systems and believing such
terms to imply either teleology or spooky causation (homonculus), others in
the past have proposed views in which such things have been characterized in
other ways, leading to the "appearance" of goals, intentions, purposes, etc.
Reinforcement theory is an example but there are others (e.g., expectancy
theory). Reinforcement theory says that you will engage in an act (the rat
presses the lever) not because you have a goal you are trying to achieve
(getting food), but because under these conditions in the past (the rat is
food-deprived), engaging in the act has been reinforced (the rat has
received food for pressing the lever). The apparently goal-directed
behavior is there and observed, but control of food-delivery is accounted
for by a non-control explanation. (This is why Baum, in that irritating
reply to your "Beyond Behaviorism" paper, insisted that "emphasis on purpose
. . . has been a hallmark of modern behaviorists' thinking."

I don't see how traditional experiments could reveal controlled variables.
It takes a special technique to do this, and I haven't seen that technique
used in any of the traditional papers I've read.

Any of the experiments on choice behavior put the organism in control of
what it will receive, within the limits set by the options provided. When
the organism shows a consistant preference for one option over others, it is
revealing that it is willing to vary its actions so as to obtain that
option. Sometimes a series of studies will be conducted to determine just
what it is about that option relative to the others that the organism "wants."

If PCT were to be rejected by behaviorists, it would not be due
to its specification of references or other aspects of internal
organization. (I say "were to be rejected" because I don't think it has yet
been seriously considered.)

Well, we'll see what happens when they do seriously consider it. What is
going to happen when they realize that what an organism accomplishes by its
behavior is determined primarily by what the organism wants to experience?
That most of the behavior they observe is simply an organism trying to push
back against disturbances?

I don't know what will happen, but it ought to be interesting. I suspect
that many psychologists will not find that to be a problem.

So, Bill, you can make PCT an island if you want to. Be my guest -- lay
under the palms and enjoy the surf. But I've got my bridge-building
construction set right here under my arm, and I intend to use it to connect
PCT with EAB, with or without you.

OK. Just make sure that in the process, PCT doesn't get distorted beyond
recognition.

You cut me to the quick, Cap'n! Do you think I'm likely to do this? If you
would prefer, I won't mention PCT at all. (:-< That way there will be no
danger of distortion.

Regards,

Bruce

[From Bill Powers (971231.1557 MST)]

Bruce Abbott (971231.1550 EST)--

The idea that JNDs are additive was Fechner's.

OK, but Stevens carried the idea further than that. He thought that we
should be able to observe steps in the psychophysical function with the
steps always being 1 JND apart, instead of a continuous relationship with
an uncertainty centered on every possible magnitude. He expected the steps
to be repeatable, with no judgments ever occurring between them.

I can see how both noise and signal increase with a system's amplification
(e.g., turning up the stereo), but in my own experience, noise in _my_
system (ears, brain) does not seem to rise with the loudness of the sound
I'm listening to. Does it in yours?

The noise in a neural signal (if Poisson distributed) goes about as the
square root of impulse rate. So while it increases with the signal, the
signal-to-noise _ratio_ actually decreases as the signal increases, because
the signal increases faster than the noise increases.

The JND is determined by the absolute, not the relative, amount of noise in
the signal. If we define it as the amount of change that is reported
correctly just half of the time, we can see that the size of the JND would
increase about as the square root of signal magnitude. It would get larger
as the signal increases, but not in direct proportion. So the apparent
slope of the input function would be inversely proportional to something
less than the first power of the signal; in fact, the 0.5 power. This would
fit nicely with the so-called Power Law, if the actual relationship of
signal magnitude to stimulus magnitude were close to linear!

I'm not saying that's the case. But the JND approach can't show that
perception is anything but a linear function of stimulation.

It would be more accurate to say that the phenomenon [of control] has been
mischaracterized than to say that it has been overlooked.

How have controlled variables been "mischaracterized?" That implies that
they have actually been measured but have been called something else. I
don't think that's true. A controlled variable, after all, will show little
correlation with either the "stimulus" or the "response." Even if it
happens to be measured, it will be rejected by any ANOVA.

Believe it or not, Bill, you are not the first person to notice that people
and other animals appear to have goals, intentions, purposes, wants, wishes,
etc.

Oh, I believe it. I keep pointing out to people that simply recognizing the
goal-oriented nature of behavior does not amount to anticipating PCT. What
PCT does is to show that goals (etc) are internal, not external, and that
they are about not actions but perceptions. I think I'm probably among the
first to have realized those things (remembering not to omit William James).

Lacking accurate knowledge about control systems and believing such
terms to imply either teleology or spooky causation (homonculus), others in
the past have proposed views in which such things have been characterized in
other ways, leading to the "appearance" of goals, intentions, purposes, etc.

Right. The "recognition" of goals, intentions, purposes, and so on was
primarily in the form of showing that these words referred to objective
factors that controlled behavior. Remember that scientific psychologists
did not believe in purposive behavior in the sense it is described in PCT,
as something actually determined by the organism in advance of the
accomplishment. So these sort of things were recognized only to explain
them away.

Reinforcement theory is an example but there are others (e.g., expectancy
theory). Reinforcement theory says that you will engage in an act (the rat
presses the lever) not because you have a goal you are trying to achieve
(getting food), but because under these conditions in the past (the rat is
food-deprived), engaging in the act has been reinforced (the rat has
received food for pressing the lever). The apparently goal-directed
behavior is there and observed, but control of food-delivery is accounted
for by a non-control explanation. (This is why Baum, in that irritating
reply to your "Beyond Behaviorism" paper, insisted that "emphasis on purpose
. . . has been a hallmark of modern behaviorists' thinking."

Irritating is the word, all right. It's about as irritating as hearing
someone talk about Creation "Science."

I don't see how traditional experiments could reveal controlled variables.
It takes a special technique to do this, and I haven't seen that technique
used in any of the traditional papers I've read.

Any of the experiments on choice behavior put the organism in control of
what it will receive, within the limits set by the options provided. When
the organism shows a consistant preference for one option over others, it is
revealing that it is willing to vary its actions so as to obtain that
option. Sometimes a series of studies will be conducted to determine just
what it is about that option relative to the others that the organism

"wants."

That's fine, but it doesn't reveal controlled variables. Once you think you
have guessed what the organism wants (no scare quotes), you then have to
define it formally and apply disturbances to it that will show which
dimensions are kept constant and which are allowed to vary. Then you have
to establish that control of the variable is done by changes in the
behavior, and ceases when perception of the variable is made impossible (or
is distorted properly when perception is distorted).

Anyway, EAB experiments are only a small fraction of the total number of
experiments published by psychologists. I think EAB experiments are
probably the closest approach to PCT experiments in the world of
psychology, because they explicitly give the animal control over at least
some of its inputs. However, the fact that EAB experiments may retain some
value does not say that any other kinds of experiments will. And even EAB
experiments, as you and I know from recent experience, are not always to be
relied upon.

OK. Just make sure that in the process, PCT doesn't get distorted beyond
recognition.

You cut me to the quick, Cap'n! Do you think I'm likely to do this? If you
would prefer, I won't mention PCT at all. (:-< That way there will be no
danger of distortion.

I thought you got into some pretty fancy rationalizations in trying to show
that reinforcement is a reasonable idea. I know that you understand PCT
quite well when you're speaking within its frame of reference. What
concerns me is your evident desire to preserve the concepts of EAB or at
least show that they are and were always reasonable. To do that, I think
you would have to distort BOTH EAB and PCT.

Best,

Bill P.