Bridges

[Myrexsw]

i.kurtzer (980104.2330)

I will likely call it quits after these quips.
Any more would be simply repetition.
So this is simply a clarification.

[Martin Taylor 980104 12:20]

Rick Marken (980103.2350)]

[Martin Taylor 980104 12:20]

Rick Marken (980103.2350)]

What got you into trouble with isaac (and me -- and Bill for that
matter) for was your claim that:

Psychophysical experiments _are_ control system experiments,
though not in the sense you intend.

I think your current position, however, is that psychophysical
experiments _are_ control system experiments in the sense
that they are experiments carried out on control systems.

now Martin:

That was always my position.

This is exactly what i am disagreeing with. I repeat, i disagree with this
statement.
This is confusing theory A--PCT--with methodology not-A.
To say that the experiments are control sysytems experiments because what they
were _really_ dealing with were control systems, is as much of a leap as
calling Democritus' inquiries the scientific method.

People act in a psychophysical experiment because
they are controlling some perception, just as they do in any other
situation (including other experiments).

And i agree. But to then don any experiment a control-system experiment is
revisionism..it annoits all others methods as if they were geared for our
questions--- in the sense that we are all scientists this is true; but in the
sense of what _methodology_ means it is false. Methods are crafted for close
theoretical comfort. And in the life sciences, I-O methods are cut from the
same cloth as I-O theories. They did/do not characterize the status of
perceptual variables as controllable input analogues, but as discrimators for
output generation. Whether we can appreciate what was really going on is
another thing altogther. Methodology is defined by how its practitions
practice. And if anything else they were/are not practicing control theory.

However, I can see how the three of you misconstrued my statement as
coming from the other side and therefore misread the whole of the
rest of what I had to say, which was a pity. Perhaps if you were to
take such statements in the context of all my other writings, you would
be able to understand better what I was getting at.

did i miscontrue you, or do we just disagree?

i.

[Martin_Taylor4]

[Martin Taylor 980109 0:40]

i.kurtzer (980104.2330)

To say that the experiments are control sysytems experiments because what they
were _really_ dealing with were control systems, is as much of a leap as
calling Democritus' inquiries the scientific method.

People act in a psychophysical experiment because
they are controlling some perception, just as they do in any other
situation (including other experiments).

And i agree. But to then don any experiment a control-system experiment is
revisionism..it annoits all others methods as if they were geared for our
questions--- in the sense that we are all scientists this is true; but in the
sense of what _methodology_ means it is false.

I get the sense that you dropped into this without the prior context. My
reason for making the statement to which you object was not to assert that
the methodology of psychophysics is a control-system methodology. It was
to counter the idea that such an experiment would be invalid because the
subjects were not controlling anything.

People (including me) have postulated other reasons why some psychophysical
experiments might be invalid, but the above is not a valid reason:-)

Bill answered properly (Bill Powers (980104.1041 MST))

1. In standard HPCT, is it assumed that one can control the values of
a perception more finely than those values can be discriminated within
the perceptual signal?

No. But it can control less finely;

There are lots of reasons why a system (control or other) might work
sub-optimally. If you know that one of the limiting conditions is not
the constraint, you can look elsewhere for the important variables. If
a control system is controlling as precisely as it could with a given
resolution of perceptual input function, then there's no point looking
elsewhere for any improvement. Conversely, if you know the precision of
perception by whatever method gives the best performance, you can then
say something about how well a person could control any perception that
uses the tested variable as a component. If the test that shows the best
precision happens to be a control-system test, fine. If it is an open-loop
test, the question arises as to whether as much precision could be available
when that perception is being controlled. But at least you know that
there exists a condition under which that precision can be achieved,
which tells you how well the variable might be controlled if (to use my
range-finder example) you get the output gear ratio right.

did i miscontrue you, or do we just disagree?

Hard to tell. I agree with your philosophical comments, or at least I think
I do. I don't think those comments are relevant to the usefulness of
psychophysical data to PCT. Perhaps we disagree on this last point.

Martin

[Bruce_Gregory9]

[From Bruce Gregory (980109.1208 EST)]

Rick Marken (980109.0840)

to Martin

Perhaps you could describe, in detail, a psychophysical experiment
that has already been done -- one that provides (according to you)
results that are valid and useful from a PCT perspective. The
presentation of such an experiment would be a strong argument
against my claim (and isaac's and Bill's too, I think) that all
conventional behavioral research results must be reevaluated
using PCT methodology to see what was actually going on.

Can anyone play? An experiment in which a subject indicates
whether or not she can hear a tone at some frequency. If she
repeatedly fails to detect the tone, we can be sure that a tone
of this frequency cannot be a controlled variable in some other
situation.

Bruce

[R_Marken]

[From Rick Marken (980109.0840)]

Martin Taylor (980109 0:40) to i.kurtzer (980104.2330)

My reason for making the statement to which you object was not to
assert that the methodology of psychophysics is a control-system
methodology. It was to counter the idea that such an experiment
would be invalid because the subjects were not controlling anything.

You were, then, countering an idea that was neither isaac's not
Bill's nore mine. No one said that psychophysical experiments are
invalid because the subjects are not controlling; we said that
psychophysical experiments are invalid because they are (with
the possible exception of "method of adjustment" experiments) not
control system experiments; they are done with no awareness that
variables are being controlled and with no systematic efforts made
to determine what these variables might be.

Perhaps you could describe, in detail, a psychophysical experiment
that has already been done -- one that provides (according to you)
results that are valid and useful from a PCT perspective. The
presentation of such an experiment would be a strong argument
against my claim (and isaac's and Bill's too, I think) that all
conventional behavioral research results must be reevaluated
using PCT methodology to see what was actually going on.

Best

Rick

···

--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken

[Bruce_Gregory9]

[From Bruce Gregory 9980109,1420 EST)]

Bill Powers (980109.1143 MST)]

Bruce Gregory (980109.1208 EST)--

>Can anyone play? An experiment in which a subject indicates
>whether or not she can hear a tone at some frequency. If she
>repeatedly fails to detect the tone, we can be sure that a tone
>of this frequency cannot be a controlled variable in some other
>situation.

Let me play stupid.

Hmmm. Why do I sense a trap? Beware of Bill playing stupid is
good advice to follow.

How do you know that the perceptions being controlled
in this situation are the same ones being controlled when you let the
subject affect the amplitude of the tone without having to say anything
about it?

You're right. I don't. It's just a place to start.

I believe that some experiments have been done in which people say they
can't detect some sensation (to their conscious knowledge), yet still
behave as if they're controlling some perception related to it ("blind
sight").

Right again.

And last, detecting that there is a tone just above threshold tells us
nothing about how that perception behaves in the region between threshold
and maximum tolerable loudness.

True.

Is there any disadvantage to allowing the subject to control the perception?

None that I can think of.

Is the argument here that we should revisit older experiments only when we
can't think of a logical reason not to do so?

We should revisit older experiments when we have some reason to
think that the data they provide are misleading in some regard
or we think we can improve on their results. At least those are
the reasons experiments are usually redone.

Bruce

[Bill_Powers1]

[From Bill Powers (980109.1143 MST)]

Bruce Gregory (980109.1208 EST)--

Can anyone play? An experiment in which a subject indicates
whether or not she can hear a tone at some frequency. If she
repeatedly fails to detect the tone, we can be sure that a tone
of this frequency cannot be a controlled variable in some other
situation.

Let me play stupid. How do you know that the perceptions being controlled
in this situation are the same ones being controlled when you let the
subject affect the amplitude of the tone without having to say anything
about it?

I believe that some experiments have been done in which people say they
can't detect some sensation (to their conscious knowledge), yet still
behave as if they're controlling some perception related to it ("blind
sight").

And last, detecting that there is a tone just above threshold tells us
nothing about how that perception behaves in the region between threshold
and maximum tolerable loudness.

Is there any disadvantage to allowing the subject to control the perception?
Is the argument here that we should revisit older experiments only when we
can't think of a logical reason not to do so?

Best,

Bill P.

[R_Marken]

[From Rick Marken (980109.1300)]

Bruce Gregory (9980109,1420 EST) --

We should revisit older experiments when we have some reason to
think that the data they provide are misleading in some regard
or we think we can improve on their results. At least those are
the reasons experiments are usually redone.

And PCT provides a _very_ good reason to think that the data
provided by older experiments are misleading. That's what the
_behavioral illusion_
(http://home.earthlink.net/~rmarken/ControlDemo/Illusion.html) is all
about; that's what the constant talk of
controlled variables is all about; that's what the push for
behavioral research that involves a Test for Controlled Variables
is all about.

PCT suggests that behavior is organized around the control of
perceptual inputs: controlled variables. "Older" (conventional)
experiments in psychology (and this includes nearly all psychophysical
experiments) are done with no awareness of the possible existence of
controlled variables, and the input-output-based methodology used in
conventional psychology provides no way to tell what variables were
being controlled in these experiments. That is why the data from
"older" (non-PCT) behavioral experiments is of no use to PCT; that
is why we must _revisit_ the older experiments (if we think they
might be important) -- and repeat them using PCT-based methodology.

I think it is unfortunate that there are many people (like Bruce
Abbott and Martin Taylor) who understand the PCT model but simply
will not accept this important implication of PCT -- that
conventional research methods cannot provide data that is useful
to PCT. It is unfortunate because these people are not going
to go out and do PCT research (witness, for example, Bruce's VI
study).

There is more to PCT than just understanding the mechanics of the
PCT model; PCT also involves understanding how the model relates to
the behavior you see -- how the model maps to _reality_ (well, to
our _perceptions_ of behavior). I can't see how someone who really
understands the possibility that visible behavior is the means by
which an organsim controls its own perceptions can also think
that conventional psychological research data tells them anything
about that behavior.

Best

Rick

···

--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken

[Fred_Nickols3]

[From Fred Nickols (980128.1750 EST)]

Bill Powers (980127.2208 MST) --

<snip>

That's not quite the idea. It's more like "if you teach a person how to
move his arms and legs in a certain way, he may end up pulling a fish out
of the water. But if he understands that the purpose of the movements is to
put a baited hook in the water where a fish is, you don't need to tell him
how to move his arms and legs, and he'll probably catch the fish a lot
sooner."

<snip>

I believe I mentioned once before that, in the early 1970s, the Navy's
programmed instruction writers' course was redesigned based on a belief
that, if the students could correctly evaluate programmed instructional
materials, they would require very little in the way of instruction
regarding how to write them. This is, of course, very closely akin to
Bill's comment above.

In behaviorist circles, this approach is known as "discrimination" training,
that is, developing the ability to recognize the end product when one sees
it. It strikes me that two very different schools of thought have arrived at
a very similar prescription. Is there a bridge to be built here or is this
nothing more than a coincidence?

Regards,

Fred Nickols
The Distance Consulting Company
nickols@worldnet.att.net
http://home.att.net/~nickols/distance.htm

[R_Marken]

[From Rick Marken (970907.1100)]

Bruce Abbott (970907.1015 EST) --

It seems to me that several people have completely forgotten what
this exercise was all about. As some (evidently not you three)
may recall, it represents an attempt to build a bridge between
PCT and those who work in the field pioneered by B. F. Skinner,
called "the experimental analysis of behavior" (EAB). Now a
bridge is designed to span a gulf between two shores, but to
serve its function, each end of the bridge must be firmly anchored
on its respective bank.

Bridges are used_to get from one bank to another. An extremely
strong, well-anchored bridge has been built from the banks of
EAB land to those of PCT land. The bridge started with Bill's
1971 "Rat Experiment" article and continued with the 1973 Science
article (and exchange of letters between Bill and some of the
priests of EAB land); the bridge was further strengthened by
B:CP and many subsequent papers. I have made modest contributions
to this bridge, as has Tom Bourbon (see his paper on reinforcement
theory at http://www.edford.com/rtpvrft.html).

The bridge is there, it's been there for nearly 30 years. Now all
we need is for the residents of EAB land to start crossing it
into PCT land. So far, there have been _very few_ immigrants from
EAB land. Instead, what we get is EABers crossing into PCT land
and trying to bring the language of PCT back to EAB land. This is
not what the bridge between EAB and PCT land is for; it's for one-
way travel.

The residents of PCT land see the world completely differently
than the residents of EAB land; the difference between the
two cultures is more than one of language. The residents of
PCT land see organisms as controllers of their own perceptions;
the residents of EAB land see organisms as controlled _by_ their
perceptions. The language difference between PCT and EAB results
from this difference in how these two cultures see the world.

We can translate between EAB and PCT, and we have: we have told
the EABers, for example, that what they call a "reinforcer" is
actually a "controlled variable". But the two cultures find it
difficult to use these terms interchangably because the words
connote quite different realities. A PCTer finds it impossible
to call a controlled varable a reinforcer because there is nothing
reinforcing about a controlled variable; and the word "reinforcer"
leaves out the important fact that the "reinforcer" is controlled
_by_ the organism. Similarly, the EABer has a tough time calling
a reinforcer a controlled variable because EABers find it
impossible to think of organisms as the ones doing the controlling;
when an EABer tries to say "controlled variable" it almost
always comes out "controlling variable". Our language reflects
what we think.

So the bridges are in place. They are solid and elegant. There
are signs all over the place showing why EABers should leave
the land of illusion and cross over the bridge into the land
of enlightenment. You could lead the way by pointing out the
signs and leading the EABers out of mental bondage. All you
have to do is stop trying to bring the language of PCT back to
EAB land and start trying to bring the EABers over to PCT land.

Now, off to the US Open.

Best

Rick

···

--

Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken/

[Bill_Powers1]

[From Bill Powers (971230.0308 MST)]

I woke up thinking about bridge-building, a kaleidoscope of thoughts. I've
only begun sorting them out. Just two thoughts for now.

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

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.

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.

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

Back to Thought 1.

This thought began with the realization that there is one obvious bridge
that could be built between PCT and behaviorism: every input-output
relationship contains an implicit reference level.

The formal definition of a reference level (remember that _level_ is used
to refer to an observable condition, whereas _signal_ refers to a model) is
_that level of input at which the output ceases either to increase or
decrease the input_. When the input is at its reference level, output
becomes zero with respect to its effects on the input. If there is no
effect of the output on the input, then the reference level of the input is
the amount of input at which the output just falls to zero.

This is quite independent of saying that there is a control system acting.
Even a pure stimulus-response system has a reference level for its input.
The reference level just means the amount of input that is _effectively_
zero with respect to producing output. If zero stimulus produces zero
output, then the reference level of the input is zero. But if there is a
certain level of input greater than zero that produces zero output, then
the reference level of the input is nonzero. Body temperature is an
example: the reference level for body temperature is that temperature at
which there is neither an increase in temperature-increasing behavior or an
increase in temperature-decreasing behavior. This applies to all forms of
thermoregulation.

Apply this idea to the use of food as a reinforcer for behavior. If the
input rate of reinforcement is less than some amount, behavior will tend
to increase it. If the reinforcement rate goes even higher, behavior will
change so as to tend to reduce the amount of reinforcement, if that is
possible. We can thus define a spectrum of behavior from food-acquiring
behavior to food-rejecting behavior as a function of input, and the
reference level for food input must be somewhere between the ends of this
spectrum.

The reference level is clearly a property of the organism, not of the
reinforcer. This could be shown by applying exactly the same spectrum of
reinforcements to different organisms and showing that the reference level
is different across organisms. Or it could be shown by applying exactly the
same reinforcement to a single organism over time, and showing that the
reference level changes over time. This would eliminate the reinforcer
itself as the determinant of the reference level.

If we start with an input-output system and add two factors, we end up with
a control system. The two factors are the reference level and the effect of
output on input. Both can be observed. When the reference level is
introduced to the analyis, it provides an independent variable that lies
inside the organism. The effect of output on input dictates the form of the
analysis, showing that two independent equations are needed to describe the
behavior of the system. With this beginning, PCT is inevitable.

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.

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.

This _is_ about PCT, by the way.

Best,

Bill P.

[Bruce_Abbott]

[From Bruce Abbott (971230.1310 EST)]

Rick Marken (971230.0840) --

Bruce Abbott (971230.1100 EST)

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.

I believe it's now over three years that you have been on CSGNet
and there is no sign of a bridge (in the form of a paper submitted
for publication) yet.

Well, perhaps I'm overly cautious, but I wanted to be sure that I (a)
understood control theory correctly, (b) could apply it in the context of
operant conditioning studies (i.e., develop and test appropriate computer
models), and (c) could offer supporting data. Parts a and b took a bit of
time, and I'm still working on c. (At the moment I'm 40+ sessions into a
study of VI schedule performance.)

I suspect that if you ever do get to building
your bridge it's going to turn out to be a lot like the original
Tacoma Narrows; I'll let you be the first to drive across it in
a storm of criticism;-)

Sounds like fun! I can hardly wait! (Seriously!)

By the way, are you ever going to tell me what _you_ think are
the major discoveries of PCT research over the past 50 years?
Does your silence mean that you think there are none?

No, and no. Now if you'll excuse me, I've got some rats (a.k.a. living
control systems) to test.

Regards,

Bruce

[Martin_Taylor4]

[Martin Taylor 971231 17:45]

Bill Powers (971230.0308 MST)]

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.

I think it is a few decades since the Fechnerian idea of "just noticeable
difference" has been theoretically respectable in psuchology. All the same,
it's a convenient fiction (just as is "cause").

S.S. Stevens actually thought that
perceptions were quantized; a control-system experiment would show that
they are not.

It's interesting to know that even if the perceptual input function is
quantized, a control system experiment would show that it is not. Why
bother with the experiment, if the result would be the same no matter
what the fact of the matter investigated?

Signal-to-noise ratio is not the same thing as perceptions
that occur in discrete steps.

No, indeed. That's why a d' (d-prime) measure is substituted for the notion
of jnd in respectable psychophysics. Nevertheless, this says nothing about
whether or not perception is quantized. That's a different issue.

(Parenthetically, d' is a measure of the separability of the signal state
from the non-signal state. It is more akin to the magnitude of the
perceptual signal than it is to "the slope of the input function"
(Bill Powers 971230.1012 MST). "Bias" (usually labelled "beta") is the
willingness of the person to assert that a signal is present. It is
not the threshold, though changes in beta are associated with changes
in what would be called a threshold in a casual "is it there or is it not"
study.)

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.

Psychophysical experiments _are_ control system experiments, though not in
the sense you intend. In a psychophysical experiment (other than one using
the method of adjustment) the subject is unable to influence what the
experimenter thinks of as the "stimulus," but is able to control an
important perception--the perception of the experimenter's satisfaction
with the subject's performance. The _mechanism_ for controlling this
perception is consistency of relation between stimulus and response. The
overt experiment is S-R. If the subject can hear the tone in an auditory
sensitivity experiment and randonly says "yes" or "no", the experimenter
will not be perceived as very satisfied.

There are real problems with a lot of psychophysical studies, but not being
control system experiments is not among them. One way of studying any
feedback structure is to break the loop somehwere and to look at the
input-output relations among the components of the loop. That's what
a psychophysical experiment (other than method of adjustment) does. The
loop component we call "environmental feedback function" is broken,
permitting the internal component to be studied in isolation. I think
that is useful, and the results help to establish parameters for the
functioning of completed control loops using the same perceptions.

The question arises as to whether the perceptual input functions operate
the same way when the resulting perception is being controlled as when it
isn't. This issue is not ordinarily considered within HPCT, since normally
the perceptual input function is taken to be whatever it is, and only the
magnitude of its output is controlled. But it is an issue, one that might
invalidate the uncritical use of the results of psychophysical studies
to assess the elements of related control loops.

Martin