Control of and in imagination

[Martin Taylor 2010.07.12.10.17]

[From Rick Marken (2010.07.08.2210)]

  Martin Taylor (2010.07.05.00.05)--

In particular, since most perceptual pathways have much higher bandwidths than
the muscular part of the output pathway, if the perceptual bandwidths are
under study, it is often impractical to have feedback through the external
environment (which means that tracking methods cannot be used). The feedback
pathway must, in those cases, be internal (in imagination), and not visible
to the experimenter.

But if the feedback path exists only in imagination how can the
subject control the variable that the subject has been instructed to
control? In fact, in all experiments, there is always an overt
response (like a button press) that affects the state of the
controlled variable. The output may occur after a lot of imagination,
but it occurs and that's how the experimenter (and the subject!) can
see that the variable the subject was instructed to control is,
indeed, under control.

The button response does not, and cannot, affect the state of the controlled variable, if that variable is the relationship between which interval the subject perceived to contain the tone and the numerical label of that interval, at least not if the subject is a normal healthy person.

I've been puzzling as to how, after so many years, you and I can have such different views of control in imagination, and why your view now seems so different from your view some years ago (as I remember it) when you were talking about the control involved in finding (I think) the word to fit in a crossword -- or something similar. At that time, your approach was to vary, in imagination, the possible words until one fit the clue and the space. That is exactly what I have proposed and you dismiss for answering the question of which interval contained a signal. You dismiss it on the grounds that control in imagination allows any answer to be made to fit.

I think I have found the answer to my puzzle. It is that we have been verbally conflating two different things. Control IN imagination depends on, but is different from control OF imagination.

Control OF imagination permits one to imagine anything that will satisfy some reference value. It is free. You can imagine yourself lying on a tropical beach, or wandering among the moons of Jupiter, or that peace has descended upon the world. The outer environment provides no constraints, and there are no obvious degrees of freedom limitations. If you want a particular imagined perception, and you are able to generate it, you can have it.

Control IN imagination is different. When one controls IN imagination, one is using the ability to create arbitrary perceptions, and using the created perception as one of the inputs to the perceptual input function of some other control system -- call it the system that controls perception X. In order for the X control system to achieve its reference value, the imagined perception must take on a value that works with the other inputs to the X perceptual input function to produce the X reference value for the X perception.

Control IN imagination is constrained. If the other input(s) to the X perceptual input function come from the outer world, the imagined input to the X perceptual input function must take on some specific value (or belong to a set of specific values) if the X control unit is to have zero error. The X control system output varies the imagined perception to work with the other inputs to its perceptual input function to reduce its error value in the same way that a control system in a tracking study varies the movements of a joystick to work with the disturbance in reducing the tracking error. The tracking control system varies the reference value supplied to the top of a hierarchy of lower-level control systems that eventually result in muscular output. In the same way, the X control system varies the reference value to a control OF imagination system to produce varying imagined perceptions.

Let's take the example of the experiment in which a subject listens to four noise bursts, in one of which is embedded a 500 Hz tone. But let's not ask the subject to report to the experimenter which interval it was. Let's just ask the subject to satisfy herself that she would be able to report to the experimenter if the experimenter asks. To make this into an experiment, let's say the experimenter does ask, but only after a small proportion of the trials, and then only about the trial before the one just completed. So, On trials 1..4, say, the subject just keeps silent, but on trial 5 the experimenter asks "Which interval was it on trial 4?".

In that experiment, you (Rick) would presumably argue that the subject could never provide the answer, since on trial 4 the subject had no button to push, no vocalization to hear, no externally detectable sign at all as to which interval had the tone. I would argue that on each trial the subject controlled IN imagination by varying the imagined answer ["2","1", "3" -- yes, that's right, "3"], and remembering "3" in case the experimenter asked after the following trial. The generation of the answer is, to me, entirely distinct from the dialogue with the experimenter.

In this example, the imagination "3" is freely generated, without constraint. That's control OF imagination. But no imagined perception other than "3" matches the externally supplied input to the relationship control system. That control IN imagination is as constrained as is any physically observable tracking output. Only when something in the external environment, such as the dialogue with the experimenter, depends on that imagination does the constrained imagination become a reference value for some output hierarchy of control systems.

The output part of the network of influences would be exactly the same as it would have been had the experimenter asked "Think of a number between 1 and 4, and tell me what number you thought of". What the subject told the experimenter in that case would be the result of control OF imagination. What the subject tells the experimenter in the actual experiment is still the result of control OF imagination, but in the experiment, control OF imagination is constrained by the control IN imagination of the relationship perception. The difference is just the same as would be the difference between an experimenter saying "move the joystick to some place you want it" and "move the joystick until the cursor is level with the target". Both tasks use the same output hierarchy, but in one case the output value is unconstrained (from the experimenter's viewpoint), whereas in the other it is constrained by the subject's willingness to do the task and by the location of the target.

All of that is a longwinded way of saying that we have been in the habit of just saying "control in imagination" to represent two very different things, which I have here labelled "control OF imagination" and "control IN imagination".

Martin
PS. I'm away again for about a week. I may access e-mail in that time, but I thought it better to post this before I left.

[From Rick Marken (2010.07.12.1500)]

�Martin Taylor (2010.07.12.10.17)--

Rick Marken (2010.07.08.2210)--

But if the feedback path exists only in imagination how can the
subject control the variable that the subject has been instructed to
control?

The button response does not, and cannot, affect the state of the controlled
variable, if that variable is the relationship between which interval the
subject perceived to contain the tone and the numerical label of that
interval, at least not if the subject is a normal healthy person.

OK, then it's the numerical label response, not the button press,
that affects that controlled variable. If no response affects the
controlled variable then it's not a controlled variable; it's just a
variable.

I've been puzzling as to how, after so many years, you and I can have such
different views of control in imagination

We don't have different views of control in (or of) imagination. We
have different views of the controlling done in a psychophysical
experiment. I think subjects in such experiments are asked to (and do)
control perceptions, such as the relationship between tone interval
and numerical label, by making the appropriate response (saying,
writing of pressing a button to indicate the intended numerical label
for the tone interval) in the experiment.

There certainly may be imagination involved in this control process;
that's a theory that can be tested against the data. But it's pretty
clear that subjects are controlling a perception that they and the
experimenter can perceive (not just something the subject controls in
imagination); it's the perception the subject is are asked to control
in such experiments, like a match of tone interval to numerical label.
This can be seen by the fact that these perceptual variables stay in
(or near) their reference states despite disturbance. Thus, the
subject in a forced choice keeps the tone interval matching the
numerical label on about 75% to 100% of the trials (depending on the
intensity of the tone) despite the disturbance of the changing
interval in which the tone occurs (such experiments are typically set
up so that control is neither perfect nor terrible; so the tone
intensity would be set so that the subject controls at the 75% correct
level -- or some d' equivalent).

Best

Rick

···

--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

[From Bill Powers (2010.07.12.1232 MDT)]

Martin Taylor 2010.07.12.10.17 --

MMT: The button response does not, and cannot, affect the state of the controlled variable, if that variable is the relationship between which interval the subject perceived to contain the tone and the numerical label of that interval, at least not if the subject is a normal healthy person.

BP: You seem to be referring to the publicly observable relationship, not the subject's perception of a relationship. Otherwise I don't understand how you could say that.

If the controlled perception is a relationship between two lower-order perceptions, then the perception can be controlled if at least one of those lower-order perceptions is affected by motor outputs (or in imagination).

In a tracking experiment, a relationship between target and cursor is controlled -- for example, "cursor some fixed distance above the target." The position of the target is set by a pattern generator, so the subject can't affect it. However, the subject can still control the relationship perception by varying the other input: the position of the cursor.

If the subject controls by imagining a position of the cursor, an observer will not see anything being controlled because there will be no action that other people can see to move the real cursor.

It is possible to get approximate control, under special circumstances, by controlling the perception using an imagined action causing an imagined input to the relationship perceiving input function. This can be done by sending a copy of the imagined reference action simultaneously to the lower-order system that can actually produce that action. Even though the person does not perceive the effect of the publicly observable action on the publicly observable relationship, the action can produce a real result somewhat similar to the imagined result. This is the basis of the "modern control theory" or "model-based control theory" approach to explaining behavior. If the action is quantized, as in the pressing of the button, the difference between the imagined outcome and the real one is minimized. But that is a special case and not the general organization we can expect to find. The "reafference copy" idea is just a mistaken concept of how control works.

This model of approximate control has obvious disadvanges, because it is based on the imagined outcome of an imagined action, not on present-time inputs from the sensors that report the consequences of the actual actions. Unexpected disturbances can't be handled, nor can changes in the subject's own output system. Small errors in the output action will not result in corrections.

I have suggested that in the light-and-button experiment, the controlled variable is the relationship between light and button press that was described in the instructions, as understood by the subject. The controlled perception is light and button OR no light and no button, with the reference condition being "true". An error is caused by light occurring with no button, or button occurring with no light. The first error condition can be corrected by pressing the button, which looks to the experimenter simply like a response to the light. The second can be corrected by releasing the button.

MMT (to Rick Marken): I've been puzzling as to how, after so many years, you and I can have such different views of control in imagination, and why your view now seems so different from your view some years ago (as I remember it) when you were talking about the control involved in finding (I think) the word to fit in a crossword -- or something similar. At that time, your approach was to vary, in imagination, the possible words until one fit the clue and the space. That is exactly what I have proposed and you dismiss for answering the question of which interval contained a signal. You dismiss it on the grounds that control in imagination allows any answer to be made to fit.

BP: You misunderstand. Both Rick and I reject your proposal because it does not entail a reality-check via the external feedback loop. It is not that imagination allows any answer, it's that imagination allows providing the appearance (to the subject) of having given the correct answer even if the subject hasn't actually given any answer at all that the experimenter can see.

MMT: I think I have found the answer to my puzzle. It is that we have been verbally conflating two different things. Control IN imagination depends on, but is different from control OF imagination.
....
Control IN imagination is different. When one controls IN imagination, one is using the ability to create arbitrary perceptions, and using the created perception as one of the inputs to the perceptual input function of some other control system -- call it the system that controls perception X. In order for the X control system to achieve its reference value, the imagined perception must take on a value that works with the other inputs to the X perceptual input function to produce the X reference value for the X perception.

BP: Yes, that is exactly what Rick and I have been talking about. X is the perceived relationship between the light and the button press. One of the inputs to the X-perceiver is the state of the light. The other, which you say is the imagined input, is the state of the button. A copy of the imagined button press can be sent to the lower button-press system as a reference signal (as in your diagrams), but without perceptual feedback from the actual button press there is no way for the higher system to know if that resulted in an actual button-press that the experimenter could observe. The higher system is satisfied by the imagined press even if no press really happened.

I contend that if we did some experimenting, we would find that subjects in the light-and-button experiment are perceiving the state of the button when they actually emit the answer, not imagining it. Such experiments are the only way to settle this issue. What you propose is possible, and once in a while we probably observe this sort of thing happening. But the type of control you propose is too unreliable to be the normal case.

MMT: Let's take the example of the experiment in which a subject listens to four noise bursts, in one of which is embedded a 500 Hz tone. But let's not ask the subject to report to the experimenter which interval it was. Let's just ask the subject to satisfy herself that she would be able to report to the experimenter if the experimenter asks. To make this into an experiment, let's say the experimenter does ask, but only after a small proportion of the trials, and then only about the trial before the one just completed. So, On trials 1..4, say, the subject just keeps silent, but on trial 5 the experimenter asks "Which interval was it on trial 4?".

In that experiment, you (Rick) would presumably argue that the subject could never provide the answer, since on trial 4 the subject had no button to push, no vocalization to hear, no externally detectable sign at all as to which interval had the tone.

BP: No, we would not. You completely misunderstand our position. Certainly the subject can, in imagination, search through various possible answers and pick the one that seems right. We don't argue with that. But the subject also has to perceive the answer when it is given, so as to be able to say "4 -- no wait, I meant to say 3." Or "4 -- OK, I'll say it again, 4" when the experimenter's cell phone rings and distracts his attention. The subject doesn't just say "4" in isolation from everything else; saying it completes the perceived relationship called "telling the experimenter which interval it was in."

I would argue that on each trial the subject controlled IN imagination by varying the imagined answer ["2","1", "3" -- yes, that's right, "3"], and remembering "3" in case the experimenter asked after the following trial. The generation of the answer is, to me, entirely distinct from the dialogue with the experimenter.

All of that is a longwinded way of saying that we have been in the habit of just saying "control in imagination" to represent two very different things, which I have here labelled "control OF imagination" and "control IN imagination".

That doesn't grab me -- it's just playing with words. "On" and "of" obviously have some special meanings to you, but they aren't mine. Control "OF" imagination means to me choosing to imagine or not to imagine. My main impression from this exposition is that you very much want the button press or the actual answer to be emitted without local feedback, so it can be treated as an open-loop response.

There's no point in going on with this argument without pausing to redo the experiments and introduce various disturbances to see if the loop is really closed or open when the actual behavior occurs. I will defer to those results, but not to anything else.

Best,

Bill P.

[Martin Taylor 2010.07.12.22.44]

In reply to

[From Bill Powers (2010.07.12.1232 MDT)]

and

[From Rick Marken (2010.07.12.1500)]

I suggest both of you reread the message to which you are replying [Martin Taylor 2010.07.12.10.17] . None of the comments by either of are relevant to the points I made, and I don't think I was THAT obscure, at all. Maybe I provided too much detail for you to take it in, but I'm usually accused of being too elliptical, so I may have erred in the other direction.

Let me try to say it much more briefly. Treat this as an abstract, and then go back to the original message for the detail. I think the main point is quite important.

I hope the following can serve as a terse guide that will make my earlier message easier to understand.

1. When the perceptual input function of a system controlling perception P has two inputs A and B, and A is varied by some external influence, a particular value of the perceptual signal P can be produced only by adjusting B to a restricted set of possible values (often only one possible value).

2. B may be ultimately derived from sensory data, from imagination or from both together, but wherever it may be derived from, P can take on its reference value only if the value of B is from a restricted set of values (usually only one possible value).

3. Input signals of the Perceptual Input function of a control system ordinarily are the perceptual signals of lower-level control systems. Some control systems derive their perceptual signals by feeding back their reference values directly, short-circuiting the normal feedback pathways through lower levels of the hierarchy. They are said to provide imagined perceptions.

4. If the control system is to control P, its output signal must influence B. If B is derived from sensory data, its output signal must act in the external environment. If B is an imagined perception, the output signal must influence the value of the imagined perception. There is no difference in principle between the situations.

5. An output signal affects muscular behaviour by means of altering the reference value for some control system whose output eventually results in muscular action. If the B value is derived from sensory data, the P control system's output varies the reference value for a control system in the output hierarchy until the B value is such as to make the P value equal its reference value.

5. Likewise an output signal can set a reference value for some control system that produces an imaginary perception by the "classical" means of feeding the reference value for its perception back to the perceptual input. If the B value is derived from imagination, the P control system's output varies the reference value for a control of imagination system until the B value is such as to make the P value equal its reference value.

6. When a control system feeds its reference value back as a perceptual value, that is control OF imagination. Control OF imagination can produce an unlimited range of perceptual values.

7. Using control OF imagination to provide the "B" signal in 1-3 and 5 above is control IN imagination. Control IN imagination requires the imagined "B" signal to take on a very specific value (or one from a choice of selected values).

8. Mixing up Control OF and control IN imagination has led to unfortunate confusion in several threads, including the threads on psychophysical experimentation.

9. The example in my earlier message, in which a person is not overtly responding to a question such as "which interval had the 500 Hz tone" but is answering it to her own satisfaction or remembering in case the experimenter later asks about it illustrates control IN imagination.

I hope the foregoing can serve as a terse guide that will make my earlier message easier to understand.

Martin

[From Bill Powers (2010.07.12.0705 MDT)]

Martin Taylor 2010.07.12.22.44 –

MMT: I hope the following can
serve as a terse guide that will make my earlier message easier to
understand.

  1. When the perceptual input function of a system controlling perception
    P has two inputs A and B, and A is varied by some external influence, a
    particular value of the perceptual signal P can be produced only by
    adjusting B to a restricted set of possible values (often only one
    possible value).

BP: Yes, that is what I said, too. If the relationship being controlled
is the difference between A and B, or P = (B - A), and the reference
level for the relationship is r, then if A is given, the steady-state
value of B will be

B = r-A

That is the value to which the control system will bring B. I discussed a
logical relationship rather than a numerical one, but the same principle
applies.

MMT: 2. B may be ultimately
derived from sensory data, from imagination or from both together, but
wherever it may be derived from, P can take on its reference value only
if the value of B is from a restricted set of values (usually only one
possible value).

BP: Yes, it’s that value such that f(A,B) = r

MMT: 3. Input signals of the
Perceptual Input function of a control system ordinarily are the
perceptual signals of lower-level control systems. Some control systems
derive their perceptual signals by feeding back their reference values
directly, short-circuiting the normal feedback pathways through lower
levels of the hierarchy. They are said to provide imagined
perceptions.

BP: Yes, so far nothing new.

MMT: 4. If the control system is
to control P, its output signal must influence B. If B is derived from
sensory data, its output signal must act in the external environment. If
B is an imagined perception, the output signal must influence the value
of the imagined perception. There is no difference in principle between
the situations.

BP: Yes, except for the external effects of controlling B. If B is
controlled only via imagination, there are no external effects.

MMT: 5. An output signal affects
muscular behaviour by means of altering the reference value for some
control system whose output eventually results in muscular action. If the
B value is derived from sensory data, the P control system’s output
varies the reference value for a control system in the output hierarchy
until the B value is such as to make the P value equal its reference
value.

BP: In that case, controlling the relationship P = f(A,B) changes the
externally observable value of B.

MMT: 5. Likewise an output
signal can set a reference value for some control system that produces an
imaginary perception by the “classical” means of feeding the
reference value for its perception back to the perceptual input. If the B
value is derived from imagination, the P control system’s output varies
the reference value for a control of imagination system until the B value
is such as to make the P value equal its reference
value.

BP: Agreed. In this case there is no change corresponding to a change in
B that an external observer could see.

MMT: 6. When a control system
feeds its reference value back as a perceptual value, that is control OF
imagination. Control OF imagination can produce an unlimited range of
perceptual values.

BP: I don’t understand this. If the reference value is fed back to the
same system, it enters where perceptual signals usually enter, and is
interpreted as if the resulting higher-order perception had assumed the
value of the reference signal. So there is still only one value of
perception that will match the reference signal. You’ll have to explain
the situation you’re imagining in more detail.

MMT: 7. Using control OF
imagination to provide the “B” signal in 1-3 and 5 above is
control IN imagination. Control IN imagination requires the imagined
“B” signal to take on a very specific value (or one from a
choice of selected values).

BP: Yes, it’s the value that causes P to match its reference level. I
don’t see any difference between 7 and 6, except that in 6 you seem to be
assuming that any of a wide variety of values of B will cause P to match
the reference signal, and as I understand the model, that’s just
false.

MMT: 8. Mixing up Control OF and
control IN imagination has led to unfortunate confusion in several
threads, including the threads on psychophysical
experimentation.

BP: It appears to me as if the confusion is all yours. I don’t think the
case you describe as “control OF” exists. There is no case in
which any old value of B will cause the perception to match the reference
signal. You haven’t stated any conditions that would suggest any way in
which the OF case can happen.

MMT: 9. The example in my
earlier message, in which a person is not overtly responding to a
question such as “which interval had the 500 Hz tone” but is
answering it to her own satisfaction or remembering in case the
experimenter later asks about it illustrates control IN
imagination.

BP: Yes, I agree. Still nothing new about that aspect of your
proposal.

MMT: I hope the foregoing can
serve as a terse guide that will make my earlier message easier to
understand.

BP: The problem is that you don’t understand why I’m disagreeing with
you, and haven’t made much of an attempt to find out why. I think I have
managed to narrow the area of disagreement to your statement about
“control OF” imagination. As I understand what you’re trying to
say, you’re describing a nonexistent case. We seem to agree about
everything else you’re saying – it’s the same thing I have been
saying.
The main place where we seem to disconnect is the point where, to use
your term, the person does “overtly respond” to the
task-specific question or stimulus. At that point, I claim, the input to
the relationship-control system becomes the lower-order perception of the
person’s actual response (instead of the imagined one), which is
how the person knows if the response actually relates correctly to the
question or stimulus when he hears what he is saying or sees what he is
doing. He needs to monitor at the relationship level to see how he can
correct an error if there is some kind of disturbance that affects the
message or the result of the action. You seem to treat the “overt
response” as occurring open-loop with respect to the relationship
under control (though not at the lower level where the response is
actually carried out at the motor level). I disagree with that, although
that is just a factual matter that can be supported or refuted by
experimental evidence about how people actually do such things. If there
is a disturbance that affects the relationship during the overt response,
will the person notice it and try to make a correction? If not, I will
have to give your idea much more credence.

By the way, you will notice that the “switches” I proposed in
B:CP are necessary in your diagram. As shown, the higher-level output
signal is shown as being connected both to the imagination path and to
the reference path going to a lower level. If the person mulls over
possible choices for a response, the diagram you have drawn would cause
those choices to be overtly generated as they are changed. What is needed
is a switch that at least disconnects the path to the lower systems while
the correct response is being found. Preferably, it should also
disconnect the inputs to the relationship control system from the lower
system, because you can’t both imagine and actually perceive the same
variable.

A working model would help clarify a lot of details.

Best,

Bill P.

[From Rick Marken (2010.07.13.0740)]

�Martin Taylor (2010.07.12.22.44)--

I suggest both of you reread the message to which you are replying �[Martin
Taylor 2010.07.12.10.17] . None of the comments by either of are relevant to
the points I made

They still look pretty relevant to me. I think the problem is that you
are trying to sell an imagination-based model of the behavior in an
experiment (such as a reaction time of forced choice detection
experiment) as a fact when it is really a theory. Bill and I are
proposing an alternative theory of these experiments -- a theory that
is organized around the control of the perception of a relationship
between stimulus and response. As Bill noted in his last post (I
believe) it's pretty easy to test these theories but in order to do
this properly I think we need an actual working version of your
theory, implemented by you. I'm afraid that if we implemented our
understanding of your theory you would complain that we got it wrong
if it fails our test (which would involve applying a disturbances to
the variables that are supposedly uncontrolled). But as it stands now
I think we understand your imagination-based model of these
experiments, we just don't believe that it's the correct model. But
maybe yours is the correct model (that would certainly be good news
for conventional psychologists). So how about a working version that
we can test in a simple experimental situation.

Best

Rick

and I don't think I was THAT obscure, at all. Maybe I

···

provided too much detail for you to take it in, but I'm usually accused of
being too elliptical, so I may have erred in the other direction.

Let me try to say it much more briefly. Treat this as an abstract, and then
go back to the original message for the detail. I think the main point is
quite important.

I hope the following can serve as a terse guide that will make my earlier
message easier to understand.

1. When the perceptual input function of a system controlling perception P
has two inputs A and B, and A is varied by some external influence, a
particular value of the perceptual signal P can be produced only by
adjusting B to a restricted set of possible values (often only one possible
value).

2. B may be ultimately derived from sensory data, from imagination or from
both together, but wherever it may be derived from, P can take on its
reference value only if the value of B is from a restricted set of values
(usually only one possible value).

3. Input signals of the Perceptual Input function of a control system
ordinarily are the perceptual signals of lower-level control systems. Some
control systems derive their perceptual signals by feeding back their
reference values directly, short-circuiting the normal feedback pathways
through lower levels of the hierarchy. They are said to provide imagined
perceptions.

4. If the control system is to control P, its output signal must influence
B. If B is derived from sensory data, its output signal must act in the
external environment. If B is an imagined perception, the output signal must
influence the value of the imagined perception. There is no difference in
principle between the situations.

5. An output signal affects muscular behaviour by means of altering the
reference value for some control system whose output eventually results in
muscular action. If the B value is derived from sensory data, the P control
system's output varies the reference value for a control system in the
output hierarchy until the B value is such as to make the P value equal its
reference value.

5. Likewise an output signal can set a reference value for some control
system that produces an imaginary perception by the "classical" means of
feeding the reference value for its perception back to the perceptual input.
If the B value is derived from imagination, the P control system's output
varies the reference value for a control of imagination system until the B
value is such as to make the P value equal its reference value.

6. When a control system feeds its reference value back as a perceptual
value, that is control OF imagination. Control OF imagination can produce an
unlimited range of perceptual values.

7. Using control OF imagination to provide the "B" signal in 1-3 and 5 above
is control IN imagination. Control IN imagination requires the imagined "B"
signal to take on a very specific value (or one from a choice of selected
values).

8. Mixing up Control OF and control IN imagination has led to unfortunate
confusion in several threads, including the threads on psychophysical
experimentation.

9. The example in my earlier message, in which a person is not overtly
responding to a question such as "which interval had the 500 Hz tone" but is
answering it to her own satisfaction or remembering in case the experimenter
later asks about it illustrates control IN imagination.

I hope the foregoing can serve as a terse guide that will make my earlier
message easier to understand.

Martin

--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

[Martin Taylor 2010.07.14.17.05]

[From Bill Powers (2010.07.12.0705 MDT)]

  Martin Taylor 2010.07.12.22.44 --

  MMT: 6. When a control system

feeds its reference value back as a perceptual value, that is
control OF
imagination. Control OF imagination can produce an unlimited range
of
perceptual values.

  BP: I don't understand this. If the reference value is fed back to

the
same system, it enters where perceptual signals usually enter, and
is
interpreted as if the resulting higher-order perception had
assumed the
value of the reference signal. So there is still only one value of
perception that will match the reference signal. You’ll have to
explain
the situation you’re imagining in more detail.

This is the issue, isn't it? Rick said that the problem with control

in imagination is that imagination can produce ANY value, and that
this meant it would be impossible for a person to determine, say,
which interval of four contained a 500 Hz tone unless the person
made an overt response.

What Rick says (that you can imagine anything you want) is true of

what I call “control OF imagination” – it’s your imagination loop.
One can freely control imagination to produce any perceptual value
one wants. It is not true of what I call control IN imagination, as
you seem to agree in the parts of your message I am not quoting. You
may be able to control the imagined perception to produce an
arbitrary value, but only one value of the imagined perception will
bring your perception to its reference value if there are other
inputs to the control system’s PIF.

    MMT: 7. Using control

OF
imagination to provide the “B” signal in 1-3 and 5 above is
control IN imagination. Control IN imagination requires the
imagined
“B” signal to take on a very specific value (or one from a
choice of selected values).

  BP: Yes, it's the value that causes P to match its reference

level. I
don’t see any difference between 7 and 6, except that in 6 you
seem to be
assuming that any of a wide variety of values of B will cause P to
match
the reference signal, and as I understand the model, that’s just
false.
That’s not the point of 6. The point of 6 is that you can set the
reference value to anything, and get that perception. In 7, you
can’t. The reference value has to be right.

Here's a set of diagrams that might make clearer what I am talking

about.

![control_imagination.jpg|777x695](upload://16nfyCDK7I9nADdJHPEyZlZqGEW.jpeg)
    MMT: I hope the

foregoing can
serve as a terse guide that will make my earlier message easier
to
understand.

  BP: The problem is that you don't understand why I'm disagreeing

with
you, and haven’t made much of an attempt to find out why. I think
I have
managed to narrow the area of disagreement to your statement about
“control OF” imagination. As I understand what you’re trying to
say, you’re describing a nonexistent case. We seem to agree about
everything else you’re saying – it’s the same thing I have been
saying.

Now do you understand why I said your comments (and Rick's) were

irrelevant?

  The main place where we seem to disconnect is the point where, to

use
your term, the person does “overtly respond” to the
task-specific question or stimulus. At that point, I claim, the
input to
the relationship-control system becomes the lower-order perception
of the
person’s actual response (instead of the imagined one),
which is
how the person knows if the response actually relates correctly to
the
question or stimulus when he hears what he is saying or sees what
he is
doing. He needs to monitor at the relationship level to see how he
can
correct an error if there is some kind of disturbance that affects
the
message or the result of the action.

Error correction is impossible once the overt response has been

made. It’s over. The button has been pushed. The response is the
response. The parrot is dead. Of course, the subject might try to
tell the experimenter “I think I pushed the wrong button” and the
experimenter might mark that response as a discard, but whether
that’s possible depends on the experimental setup, and it’s most
unusual to allow that kind of thing. Normally, all that happens is
that the next stimulus is presented. (I did do it in part of my
thesis study, though),

  You seem to treat the "overt

response" as occurring open-loop with respect to the relationship
under control (though not at the lower level where the response is
actually carried out at the motor level).

I really don't know what you mean by this observation. I've let it

slide in the past, but to deal with that issue explicitly, I did
start a thread called simply “Feedback reciprocity” in which I used
the car-in-lane example to suggest that in many (I happen to believe
most) cases of control in which a higher-level system output drives
the reference value for a lower-level perception as part of its
feedback pathway, the lower-level perception thus controlled does
NOT form any part of the higher-level perception. That’s the
situation I drew in figure 4 above, with the two pathways from the
lowest level output, one to its own controlled variable, the other
(a side-effect from its point of view), influencing something that
contributes to the higher-level perception.

In a canonical tracking study, for example, that lower-level system

might perceive hand orientation, or at an even lower level, muscle
tensions, but a side-effect of the controlled hand orientation and
muscle tensions is that the mouse moves, and the physical
environment is such that when the mouse moves, so does a cursor on
the screen. The visually sensed position of the cursor on the screen
contributes to a relationship perception betwee the cursor position
and the target position. The hand orientation and the muscle
tensions contribute nothing (nada, as Rick likes to say) to the
perception of where the cursor is on the screen. But you have to say
they necessarily contribute to the relationship perception between
the cursor position and the target position, wouldn’t you? – or you
would if you want to be consistent.

I suppose you would say that I therefore claim that the output side

of most control systems is “open-loop”. I would refute that by
saying that the loop is closed by the effect of the output on the
controlled variable, which in the tracking example is the
relationship between target position and cursor position on the
screen, and in the 500Hz tone detection experiment is the subject’s
attempting to perform the prescribed task (and at a yet higher
level, possibly, the subject’s intent to please the experimenter.

Since the presentation of the stimulus is a disturbance to the

relationship perception, the control of the relationship perception
is done by the subject forming an opinion as to what the answer
should be, NOT by the subject communicating that opinion with the
experimenter by means of a vocalization, a button push, or wagging
the appropriate finger!

Consider now the example I offered earlier:

---------[Martin Taylor 2010.07.12.10.17]
···

On 2010/07/13 10:37 AM, Bill Powers wrote:

Let's take the example of the experiment in which a subject listens

to four noise bursts, in one of which is embedded a 500 Hz tone. But
let’s not ask the subject to report to the experimenter which
interval it was. Let’s just ask the subject to satisfy herself that
she would be able to report to the experimenter if the experimenter
asks. To make this into an experiment, let’s say the experimenter
does ask, but only after a small proportion of the trials, and then
only about the trial before the one just completed. So, On trials
1…4, say, the subject just keeps silent, but on trial 5 the
experimenter asks “Which interval was it on trial 4?”.

In that experiment, you (Rick) would presumably argue that the

subject could never provide the answer, since on trial 4 the subject
had no button to push, no vocalization to hear, no externally
detectable sign at all as to which interval had the tone. I would
argue that on each trial the subject controlled IN imagination by
varying the imagined answer [“2”,“1”, “3” – yes, that’s right,
“3”], and remembering “3” in case the experimenter [happened to ask]
after the following trial. The generation of the answer is, to me,
entirely distinct from the dialogue with the experimenter.

In this example, the imagination "3" is freely generated, without

constraint. That’s control OF imagination. But no imagined
perception other than “3” matches the externally supplied input to
the relationship control system. That control IN imagination is as
constrained as is any physically observable tracking output. Only
when something in the external environment, such as the dialogue
with the experimenter, depends on that imagination does the
constrained imagination become a reference value for some output
hierarchy of control systems.

------------

You commented [From Bill Powers (2010.07.12.1232 MDT)]

: "You completely misunderstand our position. Certainly the subject
can, in imagination, search through various possible answers and
pick the one that seems right. We don’t argue with that. But the
subject also has to perceive the answer when it is given, so as to
be able to say “4 – no wait, I meant to say 3.” Or “4 – OK, I’ll
say it again, 4” when the experimenter’s cell phone rings and
distracts his attention. The subject doesn’t just say “4” in
isolation from everything else; saying it completes the perceived
relationship called “telling the experimenter which interval it was
in.”

The relationship, if you want to call it a relationship --I

wouldn’t–, “telling the experimenter” is a different control system
from the relationship “perception of stimulus related to possible
answer”. “Telling the experimenter” is a part of the dialogue loop
with the experimenter. (By the way, if “our” refers to you and Rick
both, why does Rick insist that the subject cannot determine the
correct answer until the button is pressed? )

Fairly low in the output pathway of "telling the experimenter" is a

button-control loop in which the perception is which button is
actually pressed, and if the pressed button differs from the
reference value that was provided it by the subject having
determined what the answer should be (and on the experimenter’s
request having switched that answer into the path that generates
overt output), then the button-control loop will (if allowed)
continue to act by changing the overt response until it matches its
reference value (the subject’s opinion as to the correct answer). So
what else is new?

  I disagree with that, although

that is just a factual matter that can be supported or refuted by
experimental evidence about how people actually do such things. If
there
is a disturbance that affects the relationship during the overt
response,
will the person notice it and try to make a correction? If not, I
will
have to give your idea much more credence.

A disturbance to what? The only way I can see disturbing the

relationship variable (perceived stimulus versus chosen answer) is
by altering the stimulus, and since the stimulus may have been
finished half an hour before the subject is asked to answer “Which
interval was it back then” (it’s certainly finished when the fourth
interval has completed, but half an hour is more dramatic), that
pathway is long gone. If you want to push the subject’s finger onto
a different button, that’s a disturbance to the button-pushing
control system, not to the relationship control system. Failure of
the button-pushing control system will affect the subject’s dialogue
with the experimenter, but not the subject’s opinion as to which
interval contained the tone.

  By the way, you will notice that the "switches" I proposed in

B:CP are necessary in your diagram.

Yes, of course. They just complicate the diagram and there is no

value in drawing them unless they are the topic of interest.

There might well be some value in a thread on how those switches are

controlled, but that’s a very different thread. In the context of
this exchange, manipulating such a switch might well be part of the
output pathway of a “telling the experimenter what interval held the
tone” control unit.

Martin

[From Rick Marken (2010.07.14.1812)]

Martin Taylor (2010.07.14.17.05)–

This is the issue, isn't it? Rick said that the problem with control

in imagination is that imagination can produce ANY value, and that
this meant it would be impossible for a person to determine, say,
which interval of four contained a 500 Hz tone unless the person
made an overt response.

What? Boy are we talking past one another. A person can do the forced choice detection task in imagination just fine. But if the subject does the experiment entirely in imagination it will be impossible for the experimenter to determine which of the intervals the subject selected as the one containing the tone.

What Rick says (that you can imagine anything you want) is true of

what I call “control OF imagination” – it’s your imagination loop.
One can freely control imagination to produce any perceptual value
one wants. It is not true of what I call control IN imagination, as
you seem to agree in the parts of your message I am not quoting. You
may be able to control the imagined perception to produce an
arbitrary value, but only one value of the imagined perception will
bring your perception to its reference value if there are other
inputs to the control system’s PIF.

What Rick actually says is this: only one value of the imagined “response” component of the controlled perception will bring the controlled perceptoin to it’s reference. So if the controlled perception is a match of perceived tone interval with response interval, then if the subject hears the tone in interval 3 then he can keep this perception under control (in imagination) only by imagining himself responding “interval 3”.

Error correction is impossible once the overt response has been

made.

Yes, further error correction is impossible after the reponse has been made. But error correction (control theory style) has occurred when the response is made.

  You seem to treat the "overt

response" as occurring open-loop with respect to the relationship
under control (though not at the lower level where the response is
actually carried out at the motor level).

I really don't know what you mean by this observation. I've let it

slide in the past, but to deal with that issue explicitly, I did
start a thread called simply “Feedback reciprocity” in which I used
the car-in-lane example to suggest that in many (I happen to believe
most) cases of control in which a higher-level system output drives
the reference value for a lower-level perception as part of its
feedback pathway, the lower-level perception thus controlled does
NOT form any part of the higher-level perception. That’s the
situation I drew in figure 4 above, with the two pathways from the
lowest level output, one to its own controlled variable, the other
(a side-effect from its point of view), influencing something that
contributes to the higher-level perception.

I think this is why we need to see a working model of your proposal. We have to know what exactly you are proposing. I think I do know what your model is; and if I’m right the response in your model is not open loop; the response itself is controlled. So if the response is “press” then that reponse control system gets the reference “press” and acts, via muscle outputs, to brings the perception of the response to that reference, compensating for disturbances, such as variations in the spring constant of keypress device, in order to do this. What seems to be “open loop” in your model is the connecton from this “response” to the state of the controlled variable. If this is your model – and it’s certainly a plausible model; it’s been used in psychology for years – then we can test it pretty easily.

In a canonical tracking study, for example, that lower-level system

might perceive hand orientation, or at an even lower level, muscle
tensions, but a side-effect of the controlled hand orientation and
muscle tensions is that the mouse moves, and the physical
environment is such that when the mouse moves, so does a cursor on
the screen. The visually sensed position of the cursor on the screen
contributes to a relationship perception betwee the cursor position
and the target position. The hand orientation and the muscle
tensions contribute nothing (nada, as Rick likes to say) to the
perception of where the cursor is on the screen. But you have to say
they necessarily contribute to the relationship perception between
the cursor position and the target position, wouldn’t you? – or you
would if you want to be consistent.

OK, so you are saying that the observed response in a psychophysical experiment is equivalent to the subject’s hand position on the mouse in a tracking task. If so, then this is, as I said above, pretty easy to test. But we have to agree on a working model of the situation so that we can see what your model predicts will occur given certain experimnetal manipulatons.

I suppose you would say that I therefore claim that the output side

of most control systems is “open-loop”. I would refute that by
saying that the loop is closed by the effect of the output on the
controlled variable, which in the tracking example is the
relationship between target position and cursor position on the
screen, and in the 500Hz tone detection experiment is the subject’s
attempting to perform the prescribed task (and at a yet higher
level, possibly, the subject’s intent to please the experimenter.

I dont think that analogy is quite correct. I think the subject is controlling for trying to “please” (I would say “cooperate with”) the experimenter in both cases. The subject does this in the tracking experiment by controlling the distance between target and cursor; the subject does it in the 500 Hz tone detection experiment by controlling the relatoinship between perceived tone interval and the tone interval indicated by the response. But let’s get away from this verbal bantering. Let’s see your working model of the 500 Hz tone detection experiment. I’ll show you mine if you show me yours;-)

Best

Rick

···


Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

[Martin Taylor 2010.07.14.22.22]

[From Rick Marken (2010.07.14.1812)]

        Martin Taylor

(2010.07.14.17.05)–

        This is the issue, isn't it? Rick said that the problem with

control in imagination is that imagination can produce ANY
value, and that this meant it would be impossible for a
person to determine, say, which interval of four contained a
500 Hz tone unless the person made an overt response.

      What? Boy are we talking past one another. A person can do the

forced choice detection task in imagination just fine. But if
the subject does the experiment entirely in imagination it
will be impossible for the experimenter to determine which
of the intervals the subject selected as the one containing
the tone.

Well that's an advance! You seem to be converging towards my

position, which is always nice. The issue of the button press now
isn’t its contribution to the selection of the answer, but its part
in the dialogue with the experimenter. Progress.

        What Rick says (that you can imagine anything you want) is

true of what I call “control OF imagination” – it’s your
imagination loop. One can freely control imagination to
produce any perceptual value one wants. It is not true of
what I call control IN imagination, as you seem to agree in
the parts of your message I am not quoting. You may be able
to control the imagined perception to produce an arbitrary
value, but only one value of the imagined perception will
bring your perception to its reference value if there are
other inputs to the control system’s PIF.

      What Rick actually says is this: only one value of the

imagined “response” component of the controlled perception
will bring the controlled perceptoin to it’s reference. So if
the controlled perception is a match of perceived tone
interval with response interval, then if the subject hears the
tone in interval 3 then he can keep this perception under
control (in imagination) only by imagining himself responding
“interval 3”.

Good, good!! You are getting there.
        Error correction is

impossible once the overt response has been made.

      Yes, further error correction is impossible _after_ the

reponse has been made. But error correction (control theory
style) has occurred when the response is made.

"Control theory style?" To my understanding, there are several

“control theory style” levels of error correction going on. One is
as you described quite nicely in the preceding paragraph starting
“What Rick actually says” (where I take “actually” in it’s French
sense of “at this moment”). Another is the control of the button
selection perception to match the reference value provided by the
one noted in the above paragraph; and yet another is control of the
perception of the button being pushed. Going up levels instead of
down, the subject is presumably controlling a perception of
performing the task correctly, and above that a perception of being
cooperative (acting to perceive the experimenter as being pleased).
All of these involve error correction (control theory style).

Why did you think it necessary to add that bracketed qualification

“(control theory style)”?

            You seem to treat the "overt

response" as occurring open-loop with respect to the
relationship under control (though not at the lower
level where the response is actually carried out at the
motor level).

        I really don't know what you mean by this observation. I've

let it slide in the past, but to deal with that issue
explicitly, I did start a thread called simply “Feedback
reciprocity” in which I used the car-in-lane example to
suggest that in many (I happen to believe most) cases of
control in which a higher-level system output drives the
reference value for a lower-level perception as part of its
feedback pathway, the lower-level perception thus controlled
does NOT form any part of the higher-level perception.
That’s the situation I drew in figure 4 above, with the two
pathways from the lowest level output, one to its own
controlled variable, the other (a side-effect from its point
of view), influencing something that contributes to the
higher-level perception.

      I think this is why we need to see a working model of your

proposal. We have to know what exactly you are proposing. I
think I do know what your model is; and if I’m right the
response in your model is not open loop; the response itself
is controlled. So if the response is “press” then that reponse
control system gets the reference “press” and acts, via muscle
outputs, to brings the perception of the response to that
reference, compensating for disturbances, such as variations
in the spring constant of keypress device, in order to do
this. What seems to be “open loop” in your model is the
connecton from this “response” to the state of the controlled
variable. If this is your model – and it’s certainly a
plausible model; it’s been used in psychology for years –
then we can test it pretty easily.

Remember, I'm arguing that psychophysicists have been doing mostly

the right thing for many years – inadvertently, perhaps, but still
the right thing. And producing perfectly valid results, usable in
any control-theoretic analysis where those results might be germane.

You said: "*      What seems to be "open loop" in your model is the

connecton from this “response” to the state of the controlled
variable.*"

Which of the many possible

“controlled variables” did you have in mind? Given that the
“response” is part of a dialogue with the experimenter and not part
of selecting the right answer, it has to be a controlled variable
that involves some perception of the experimenter? Which did you
intend? Trace that loop around, and you will find that it is most
probably closed. (When I say “most probably” I have in mind the
possibility of a non-responsive experimenter, perhaps robotic, who
gives no feedback that would allow the subject to determine whether
he is or is not doing the experiment as intended; such an
experimenter does not use what I described earlier as “good
technique”).

The connection between the response and the state of the controlled

variable is “open loop” in the same sense that the path between any
two components in a control loop is “open loop”. The path between
sensor and perceptual input function is open loop. The path between
a high-level control unit’s output function and the muscular output
to the world is open loop. Yes, the path between the output of the
relationship control unit’s output function and the button press is
open loop. The control loops in which these paths participate are
not open loop.

        In a canonical tracking study, for example, that lower-level

system might perceive hand orientation, or at an even lower
level, muscle tensions, but a side-effect of the controlled
hand orientation and muscle tensions is that the mouse
moves, and the physical environment is such that when the
mouse moves, so does a cursor on the screen. The visually
sensed position of the cursor on the screen contributes to a
relationship perception betwee the cursor position and the
target position. The hand orientation and the muscle
tensions contribute nothing (nada, as Rick likes to say) to
the perception of where the cursor is on the screen. But you
have to say they necessarily contribute to the relationship
perception between the cursor position and the target
position, wouldn’t you? – or you would if you want to be
consistent.

      OK, so you are saying that the observed response in a

psychophysical experiment is equivalent to the subject’s hand
position on the mouse in a tracking task.

I suppose it serves your polemical purpose to write nonsense, but it

doesn’t contribute to the discussion.

I made the point that the hand orientation does NOT contribute to

the perception of cursor position, in order to make the point that
the press of a button does NOT contribute to the perception of which
interval contained a tone.

It's exactly the claim that the button push must necessarily

contribute to the relationship perception that has been at the core
of this discussion since February. It was in “Bill’s model” when the
diagrams for the two models were first displayed together back then.
Today, for the very first time, you are seeming (I say “seeming”
advisedly) to accept that this claim is not true, and that in fact
the subject CAN control the relationship perception in imagination,
and CAN provide the resulting “answer” value as a reference to
whatever hierarchy of control systems eventually results in a button
push observable by the experimenter. At that point, you are
exceedingly close to what I have been saying all along.

      If so, then this is, as I said above, pretty easy to

test. But we have to agree on a working model of the situation
so that we can see what your model predicts will occur given
certain experimnetal manipulatons.

        I suppose you would say that I therefore claim that the

output side of most control systems is “open-loop”. I would
refute that by saying that the loop is closed by the effect
of the output on the controlled variable, which in the
tracking example is the relationship between target position
and cursor position on the screen, and in the 500Hz tone
detection experiment is the subject’s attempting to perform
the prescribed task (and at a yet higher level, possibly,
the subject’s intent to please the experimenter.

      I dont think that analogy is quite correct. I think the

subject is controlling for trying to “please” (I would say
“cooperate with”) the experimenter in both cases. The subject
does this in the tracking experiment by controlling the
distance between target and cursor; the subject does it in the
500 Hz tone detection experiment by controlling the
relatoinship between perceived tone interval and the tone
interval indicated by the response.

Yes. That's what I have been trying to get across since the very

first message back in (?)Fenbruary, when I drew a diagram that
showed precisely what you say. The physical response is part of the
loops that involve the experimenter, not part of the loop in which
the correct answer is determined. I’m glad you now get it.

      But let's get away from this verbal bantering. Let's see

your working model of the 500 Hz tone detection experiment.
I’ll show you mine if you show me yours;-)

When you finally complete your almost completed convergence to

acceptance of what I have been trying to explain, the two models
will be seen to be identical, so if you then show me yours, you can
have a look at it, because it will be mine.

Martin

[From Rick Marken (2010.07.14.2200)]

Martin Taylor (2010.07.14.22.22)–

Rick Marken (2010.07.14.1812)]

      But let's get away from this verbal bantering. Let's see

your working model of the 500 Hz tone detection experiment.
I’ll show you mine if you show me yours;-)

When you finally complete your almost completed convergence to

acceptance of what I have been trying to explain, the two models
will be seen to be identical, so if you then show me yours, you can
have a look at it, because it will be mine.

It doesn’t work that way in my world. In my world you propose and model and then you test it. You don’t present you model by trying to convince someone that their model is actually the same as yours. At every point in your post where you see me agreeing with something you said, I see continual misunderstanding on your part.

I happen to see no convergence at all between our models. If you think there is such a convergence then the only way to convince me of that is to show me your model. Otherwise, I will have to assume that I know what your model is from your verbal descriptions of it. And so far, I conclude from your descriptions that your model is simply the open-loop causal model of conventional psychology, an S-R model of the behavior in an experiment, with the extra added attraction of a control system at the end of the chain to control R; a control system that is simply a part of the causal chain running from S to R. It’s a model of behavior which we have tested and rejected over and over again.

Best

Rick

···


Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

[From Bill Powers (2010.07.15. 0430 MDT)]

Martin Taylor 2010.07.14.17.05 –

MMT: This is the issue, isn’t
it? Rick said that the problem with control in imagination is that
imagination can produce ANY value, and that this meant it would be
impossible for a person to determine, say, which interval of four
contained a 500 Hz tone unless the person made an overt
response.

BP: This is simply a misunderstanding – of what I have said, and
probably of what Rick has said. Of course you can adjust an imagined
perception to any of the values it can have; that is how you can make the
higher-order perception to which it contributes match the higher-order
reference signal. The imagination connection can’t, however, change one
kind of perception into a different kind. Once a perceptual input
function has become organized, only the magnitude of the perception can
be manipulated. When you are controlling the relationship between the
state of a light and the state of a button-press, you can imagine the
button in either of its two states, pressed or not pressed. That is what
those signals entering that relationship-control system mean; the meaning
is set by the nature of the receiving input function. The meaning is not
carried in the signal.

3f96de.jpg

3f9855.jpg

···

======================================================

OK, I had skipped the diagrams for the moment, and now that I look at
them a little more closely, the whole problem is cleared up. Your
diagrams do not show the imagination connection as I defined it. You have
the reference signal from the higher system connecting to the input of
the perceptual input function of the system at the lower level. That is
not my proposal. In my diagrams I show a reference signal coming down
into a lower-order system, but it loops around and goes back up without
ever serving as a reference signal, or as any kind of input, for the
lower system. Instead, it takes the place of the perceptual signal that
the lower system would normally send to the higher one. Here is your
diagram that shows the misconception of my imagination model:

Emacs!

The diagram as I proposed it is below:

Emacs!\

Now the system you have labeled “Control OF imagination” is in
the imagination mode, its output becoming an input to its own input
function, just as if a still-lower system (not shown) had received the
output as a reference signal and instantly matched its perceptual signal
to that reference. The system below the lower level, however, never
received any reference signal and so did nothing. I’ve shown the
correction in red.

You could have put the relationship-perceiving system (the upper one)
into the

imagination mode instead. That would entail breaking the connection from
the upper system’s output function to the lower system’s comparator
(small circle), and routing it instead into the upper system’s perceptual
input function in place of the right-hand signal now shown entering it.
This is where those switches I proposed would have to go (your diagram
needs one where the X is, to keep lower systems from being activated
while the upper ones are imagining). The lower system below the X would
then receive no reference signal and would maintain whatever state of its
own perception corresponds to zero reference signal.

The diagram was, indeed, a help in figuring out what was going wrong. My
model of imagination is not organized as you proposed.

Best,

Bill P.

[From Bill Powers (2010.07.15.0535 MDT)]

Martin Taylor 2010.07.14.22.22 –

MMT (to Marken): Remember, I’m
arguing that psychophysicists have been doing mostly the right thing for
many years – inadvertently, perhaps, but still the right thing. And
producing perfectly valid results, usable in any control-theoretic
analysis where those results might be germane.

I think this is a reference signal, not a perception. Psychophysicists
may have done the right thing occasionally, but since they have had no
understanding of control theory at all, they never discovered the fact of
control, and so must have made many mistakes.

I have just been reading von Holst’s original paper on reafferance
(courtesy of Henry Yin), and actually may have read a translation or a
review of it in the 1950s while working to develop “feedback
theory.” I may have been misled by the way he draw the reafference
diagrams. From his descriptions in a translation of the original
manuscript, he was clearly looking at some hierarchical control phenomena
but misunderstanding what he (and Mittelstadt) were seeing. For example,
he spoke of vestibular signals as reafferent signals, but they are
clearly not simply copies of efferent signals. I find no indication in
the article that he actually found bifurcations in the efferent signals
that routed copies back to the afferent channels.

At one point, von Holst indicates that he is familiar – somewhat – with
the cybernetic concept of feedback, but he states emphatically, even
somewhat belligerently, that feedback has nothing to do with reafferance.
I detect a reviewer in the background who did understand feedback, which
von Holst clearly did NOT understand.

I think he mistook many controlled perceptual signals from lower levels
for copies of the downgoing reference signals simply because changes in
the reference signals produced immediate changes in the upgoing
perceptual signals at the same level. He saw the corresponding changes
but I doubt that he actually traced the circuits. We, of course,
understand how that happens, and it’s not “reafference” but
control. In fact, reafference as I understood von Holst’s proposal
through diagrams and third-party descriptions is simply a control system
in the imagination mode, which was probably why I didn’t follow up by
reading his own words. A mistake. However, even if I had followed up, I
might not have had the nerve to contradict such a famous person.

I think that being loyal to the old psychophysicists is inappropriate.
They did their best with the tools they had, as well as we could have
done, but they were missing the most important tool: control theory. How
can you explain how the CNS works without understanding its basic
principle of organization? Incorrectly, that’s how.

So we can acknowledge the attempts by our predecessors to understand how
the brain works, and admire their skills, ingenuity, and persistence, but
that does not mean they were right. Some of them came close, but wandered
right past what look now like the correct answers and into blind alleys
where they spent the rest of their careers. As we, ourselves, are no
doubt doing. Some little residue of progress may survive, but even PCT is
not the last scientific revolution. Ancestor worship is not what we
need.

Best,

Bill P.

Best,

Bill P.

[From Rick Marken (2010.07.15.0910)]

Martin Taylor (2010.07.14.22.22)

Rick Marken (2010.07.14.1812)]

        MT: Error correction is

impossible once the overt response has been made.

      RM: Yes, further error correction is impossible _after_ the

reponse has been made. But error correction (control theory
style) has occurred when the response is made.

MT: "Control theory style?" ...Why did you think it necessary to add that bracketed qualification

“(control theory style)”?

        I think this deserves a brief explanation. When you say "Error correction is

impossible once the overt response has been made" you imply that that the overt response is somewhat unimportant in the control process. Perhaps that is true in your model of an experiment. But in my model of an experiment the response is what corrects the error. For example, in my model of a reaction time experiment the subject is controlling relationship between S and R, such as the XOR relationship: light on (and) press V light off (and) no press. A disturbance to this variable, such as the light coming on, will produce an error in the system controlling this relationship; this error is corrected by the response, pressing the button.

        So my point was that in a control system controlling a relationship between S and R the error correction occurs _when_ R occurs. I said it is "control theory style" error that is corrected because it seemed to me that your comment about "Error correction being

impossible once the overt response has been made" suggested that you were thinking of “error” from the observer’s perspective. These experiments are typically set up so that, if an error occurs from the experimenter’s point of view, the subject can’t correct it. From the subject’s perspective, however, the error (control theory style) is corrected by the response, even if the response is an “error” from the experimenter’s point of view.

Best

Rick

···


Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

[Martin Taylor 2010.07.19.13.03]

[From Bill Powers (2010.07.15.0535 MDT)]

Martin Taylor 2010.07.14.22.22 --

MMT (to Marken): Remember, I'm arguing that psychophysicists have been doing mostly the right thing for many years -- inadvertently, perhaps, but still the right thing. And producing perfectly valid results, usable in any control-theoretic analysis where those results might be germane.

I think this is a reference signal, not a perception.

Funny! When I wrote the reasons why the above was true, you commended it and suggested I shoulld make my message into a publication. A few days later, that commendation has morphed into condemnation. Odd.

Martin

[From Bill Powers (2010.07.19.1158 MDT)]

Martin Taylor 2010.07.19.13.03 --

MMT (to Marken): Remember, I'm arguing that psychophysicists have been doing mostly the right thing for many years -- inadvertently, perhaps, but still the right thing. And producing perfectly valid results, usable in any control-theoretic analysis where those results might be germane.

BP: I think this is a reference signal, not a perception.

MMT: Funny! When I wrote the reasons why the above was true, you commended it and suggested I shoulld make my message into a publication. A few days later, that commendation has morphed into condemnation. Odd.

BP: Good Lord, is my memory really getting that bad? I remember recommending that you write an article on the particular subjects you discussed, but didn't realize you took this as an endorsement of past work in psychophysics. Psychophysicists do take more precautions than I had realized before reading your post, but they do so not knowing the implications, and anyway don't really understand all the precautions that are necessary. Because they most often set up their experiments to preclude effective and continuous feedback (as when they use button-presses to record responses that actually begin a considerable length of time before the contacts actually close), even their raw results can be questionable. Their interpretations, because of being uninformed by control theory, are probably incorrect most of the time -- for example, every time they describe an action intended to change a perception or defend it against disturbance as a "response." Organisms do not "respond." It's OK to use that term as shorthand for an intentional action, but only if you know it's shorthand and doesn't imply an output that is simply caused by an input. Psychophysicists, aside from you, don't know that even a simple time delay is made up of many time delays in different parts of the intervening circuitry concerned with variables as different as sensations and relationships. There's no such thing as "the" reaction time, and any conclusions based on such a concept are simply wrong.

When you say psychophysicists verify that the subject is controlling the same variable the experimenter wants them to control, I'm sure (or was sure) you don't mean that literally. I don't believe any psychophysists do an actual test for a controlled variable, which in its full technical form is just a test of a control-system model. I don't see any evidence that they have tested models at all.

When you couched your descriptions in control-theory terms, such as saying that experimenters make sure the subject is controlling against the correct disturbances, I thought you should write to say that this is what the experimenters were actually doing as opposed to what they thought they were doing. Most of the rest of that section was about such PCT reinterpretations (rather generous ones, but I thought this would be a good way to suggest more useful interpretations to these people). I didn't realize that you were trying to tell me that psychophysicists were already doing everything a PCT experimenter would do. If you meant that, I completely reject that idea. They were not. They missed the boat.

I've been reading von Holst's ("With" Mittelstadt) article on the "reafference principle," which I had never before read in the complete version (translated) published in 1950. It's amazing how close he came to inventing not only PCT but HPCT, just before I started working on those ideas. It's even more amazing when he reveals an acquaintance with cybernetics (through other people, not direct) and the feedback concept -- and that he firmly rejected both. I'm sure I read about his work in the 1950s, if not the work itself (which was initially in German), but dismissed it because he clearly didn't understand control theory and offered an unworkable model of how "reafference" actually happens. In a footnote, von H. notes that an acquaintance suggested that feedback was the explanation , and says firmly, with an exclamation mark, that reafference has nothing to do with feedback. He seemed to be saying that because he was trying to win an argument with someone who liked cybernetics, the latest fad. So there he was, caught in the act of dumping the baby with the bathwater. Heck, he didn't even understand the stretch reflex, which he evidently knew about only through another acquaintance. Very strange, considering all his work with neurological experimentation. A case of tunnel vision? All his discussions of reflexes are way, way off the mark.

The past is past. Why defend it as a matter of principle? I am frankly glad that I don't have to defend every idea I ever had or made public. Why should the fact that in my green years, I wrote a thesis on a particular subject be of any interest now, all these years later? If the ideas were right then, they're still right (however unlikely that may be), but if they were wrong, good riddance. I certainly don't want to be saddled with the burden of defending everything I ever said. Do you want to be in that position? Does loyalty to the past have anything to do with science?

Best,

Bill P.

[Martin Taylor 2010/07.29.16.52]

[From Rick Marken (2010.07.14.2200)]

        Martin Taylor

(2010.07.14.22.22)–

Rick Marken (2010.07.14.1812)]

                But let's get away from this verbal bantering.

Let’s see your working model of the 500 Hz tone
detection experiment. I’ll show you mine if you show
me yours;-)

        When you finally complete your almost completed convergence

to acceptance of what I have been trying to explain, the two
models will be seen to be identical, so if you then show me
yours, you can have a look at it, because it will be mine.

      It doesn't work that way in my world. In my world you propose

and model and then you test it. You don’t present you model by
trying to convince someone that their model is actually the
same as yours. At every point in your post where you see me
agreeing with something you said, I see continual
misunderstanding on your part.

Let's try to clear up these misunderstandings, then. I'll cite what

you wrote and relate it to my understanding. All references are to
[Martin Taylor 2010.07.14.22.22] in response to and with quotes from
[From Rick Marken (2010.07.14.1812)].

Point 1: The subject can use control in imagination to generate the

answer he intends to convey to the experimenter (using the example
of a four-alternative forced choice detection study).

[RM]  A person can do the forced choice detection task in

imagination just fine.

[MT Now] That sounds to me exactly like what I have been trying to

get across. How else should I interpret it than that the subject can
do the forced choice detection task in imagination just fine?

[RM] What Rick actually says is this: only one value of the imagined

“response” component of the controlled perception will bring the
controlled perception to it’s reference. So if the controlled
perception is a match of perceived tone interval with response
interval, then if the subject hears the tone in interval 3 then he
can keep this perception under control (in imagination) only by
imagining himself responding “interval 3”.

[MT Now] That seems to emphasise and make more precise the claim

that the subject generates the choice of answer in imagination. But
it contradicts [From Rick Marken (2010.07.15.0910)]: “in my model of
an experiment the response is what corrects the error.” and “in a
control system controlling a relationship between S and R the error
correction occurs when R occurs.”

Point 2: The overt response is part of the dialogue with the

experimenter, and is not a part of the control of the relationship
between “stimulus” and “choice of answer” in the 4AFC experiment.

[RM immediately following the first quote above] ....But if the

subject does the experiment entirely in imagination it will be
impossible for the experimenter to determine which of the
intervals the subject selected as the one containing the tone.

[MT Now] How else should I interpret this than as saying that only

when the subject makes the overt response can the experimenter know
which choice the subject made as to which interval contained the
tone? And how else should that be interpreted in control-theory
terms than that the overt response is part of the set of dialogue
loops that include both the subject and the experimenter?

Since these are the main points in the post on which I thought you

agreed with me, but about which you say: “At every point in your
post where you see me agreeing with something you said, I see
continual misunderstanding on your part.” would you mind explaining
in words of one syllable wherein my continual misunderstanding lies?

Martin
···
      I happen to see no convergence at all between our models. If

you think there is such a convergence then the only way to
convince me of that is to show me your model. Otherwise, I
will have to assume that I know what your model is from your
verbal descriptions of it. And so far, I conclude from your
descriptions that your model is simply the open-loop causal
model of conventional psychology, an S-R model of the behavior
in an experiment, with the extra added attraction of a control
system at the end of the chain to control R; a control system
that is simply a part of the causal chain running from S to R.
It’s a model of behavior which we have tested and rejected
over and over again.

      Best



      Rick

  Richard S. Marken PhD

  rsmarken@gmail.com

  [www.mindreadings.com](http://www.mindreadings.com)

[Martin Taylor 2010.07.19.17.16]

[From Bill Powers (2010.07.19.1158 MDT)]

Martin Taylor 2010.07.19.13.03 --

MMT (to Marken): Remember, I'm arguing that psychophysicists have been doing mostly the right thing for many years -- inadvertently, perhaps, but still the right thing. And producing perfectly valid results, usable in any control-theoretic analysis where those results might be germane.

BP: I think this is a reference signal, not a perception.

MMT: Funny! When I wrote the reasons why the above was true, you commended it and suggested I shoulld make my message into a publication. A few days later, that commendation has morphed into condemnation. Odd.

BP: Good Lord, is my memory really getting that bad? I remember recommending that you write an article on the particular subjects you discussed, but didn't realize you took this as an endorsement of past work in psychophysics.

I didn't, and don't, and I wouldn't have done even before I encountered PCT.

I said that good experimental practice in psychophysics actually resulted in the experimenters doing exactly what they should do if their practice had been designed by a PCT practitioner. Not all psychophysical experiments have followed good practice, and not all experiments labelled as psychophysics conform to the designs I described. So I don't make the extreme claim that you attribute to me.

Earlier [From Bill Powers (2010.07.15.0535 MDT)] you said:

[BP] Martin Taylor 2010.07.14.22.22 --

MMT (to Marken): Remember, I'm arguing that psychophysicists have been doing mostly the right thing for many years -- inadvertently, perhaps, but still the right thing. And producing perfectly valid results, usable in any control-theoretic analysis where those results might be germane.

[BP] I think this is a reference signal, not a perception.

Remember that in response to Marken's statement [From Rick Marken (2010.07.04.0840)]:"since all psychophysical research is based on a an open-loop causal model of the systems under study, those who study psychophysics (which used to include me;-) can't really be learning much about the "capabilities of the processing channels" of these systems using conventional methods.", I said [Martin Taylor 2010.07.04.17.02] "Fifteen or twenty years ago, when I was just getting my head into the PCT space, I might have agreed, but now that I understand PCT in a pretty deep sense, I don't. In fact, the more I think about it from a PCT perspective, the less I agree with your statement. "

I think this should demonstrate that it's a perception, and not a controlled one, let alone a reference value.

Psychophysicists do take more precautions than I had realized before reading your post, but they do so not knowing the implications, and anyway don't really understand all the precautions that are necessary.

Perhaps true. Not being able to see into their minds, I don't know the truth of that.

Because they most often set up their experiments to preclude effective and continuous feedback (as when they use button-presses to record responses that actually begin a considerable length of time before the contacts actually close), even their raw results can be questionable.

You have to explain that, and explain why my earlier analysis is wrong in this respect. Why does the length of time matter between when a subject decides which interval contained a signal and when the subject pushes the button to tell the experimenter what he decided?

Their interpretations, because of being uninformed by control theory, are probably incorrect most of the time -- for example, every time they describe an action intended to change a perception or defend it against disturbance as a "response." Organisms do not "respond." It's OK to use that term as shorthand for an intentional action, but only if you know it's shorthand and doesn't imply an output that is simply caused by an input.

Although what you say is probably true, it's really irrelevant to the validity and usefulness of the results obtained.

Psychophysicists, aside from you, don't know that even a simple time delay is made up of many time delays in different parts of the intervening circuitry concerned with variables as different as sensations and relationships. There's no such thing as "the" reaction time, and any conclusions based on such a concept are simply wrong.

You are wrong about psychophysicists not knowing this. "Reaction time" is often used as a measure of differential effects, but the problems with it (at least some problems with it) are well known, and experimenters try to set up the experiments so that the effects of those problems are minimized. A lot of effort has gone into trying to tease out the different stages of the process over at least the last 45 years (When I was on sabbatical at the University of Keele in the UK in 1966-7, Martin Regan was doing very ingenious experiments on just this topic.

I'm sorry to say this, but over the years, I have observed that you very often say "people don't know" of "people have ignored" or similar kinds of statements, as a generic slam against whole classes of people, many of whom may well have been working hard to resolve exactly the issues they are accused of ignoring. That their efforts are hampered or invalidated by not understanding PCT does not mean that they are unaware of the problems at issue.

When you say psychophysicists verify that the subject is controlling the same variable the experimenter wants them to control, I'm sure (or was sure) you don't mean that literally. I don't believe any psychophysists do an actual test for a controlled variable, which in its full technical form is just a test of a control-system model. I don't see any evidence that they have tested models at all.

Are you saying that my description of the questions asked in "The Test" were wrong? In what way? Or did I miss a crucial question?

When you couched your descriptions in control-theory terms, such as saying that experimenters make sure the subject is controlling against the correct disturbances, I thought you should write to say that this is what the experimenters were actually doing as opposed to what they thought they were doing.

Did I not say this quite explicitly? I thought I made it clear that I believed most experimenters to be doing the right thing unknowingly.

Most of the rest of that section was about such PCT reinterpretations (rather generous ones, but I thought this would be a good way to suggest more useful interpretations to these people). I didn't realize that you were trying to tell me that psychophysicists were already doing everything a PCT experimenter would do. If you meant that, I completely reject that idea. They were not. They missed the boat.

That's purely ideological statement, I think. If I can return to you your own comment: "I think this is a reference signal, not a perception."

I said right up front in that message [Martin Taylor 2010.07.05.00.05] : "I'm beginning to see that a good part of what is taught as good experimental practice in psychophysics actually is performing the test for the controlled variable, whether the experimenters know it or not (and most don't)." and "Of course, most practitioners will not realize that they are doing "The Test", any more than most people realize they are always controlling their inputs, but it seems to me that good practice nevertheless usually, if not always, results in the experimenter performing the Test for the controlled variable."

I'm not clear why you didn't realize I said that.

[Stuff about Holst]

The past is past. Why defend it as a matter of principle?

Why is a technical analysis "defence of the past as a matter of principle"?

As I said, there was a period -- quite a long period -- when I might well have accepted the party line that most studies done without a proper understanding of PCT would be worthless. It has been only since I became more deeply aware of what PCT implies, and have developed a better understanding of its analytic underpinnings, that I have come to realize that the party line has to be tempered by an examination of what the experiments actually do tell us. And in coming to that realization, I realize that most psychophysical experiments are valid, regardless of what the experimenters thought they were doing (and I don't mean magnitude estimation :-).

I realize you and Rick are controlling with a rather high gain a perception of the value of non-PCT-guided research, with a reference level of "NIL", and that my comments since February have been disturbances to that controlled perception, disturbances that you have resisted with a bewildering variety of different kinds of objections. But could you go up a level, and ask if there is anything you might be controlling that generates this "NIL" (or any other) reference value for that perception? And is there any way of controlling some other perception so as to maintain this higher-level one near its reference value, while at the same time allowing for rational discussion of the perception "value of some non-PCT-guided research"?

Your succession of comments do disturb a perception that I control for: "This discussion is rational and scientific".

Martin

[From Bill Powers (2010.07.19.1528 MDT)]

Martin Taylor 2010/07.29.16.52 --

Horning in on your conversation with Rick:

MMT: Let's try to clear up these misunderstandings, then. I'll cite what you wrote and relate it to my understanding. All references are to [Martin Taylor 2010.07.14.22.22] in response to and with quotes from [From Rick Marken (2010.07.14.1812)].

Point 1: The subject can use control in imagination to generate the answer he intends to convey to the experimenter (using the example of a four-alternative forced choice detection study).

[RM] A person can do the forced choice detection task in imagination just fine.

[MT Now] That sounds to me exactly like what I have been trying to get across. How else should I interpret it than that the subject can do the forced choice detection task in imagination just fine?

BP: Just that way: in imagination, where the experimenter can't see what is going on. The subject could do the entire series of button presses in relation to the light onsets without ever actually pressing the button. Of course the experimenter would soon stop the experiment and complain that the subject wasn't doing it right. Behave, dammit!

[RM] What Rick actually says is this: only one value of the imagined "response" component of the controlled perception will bring the controlled perception to it's reference. So if the controlled perception is a match of perceived tone interval with response interval, then if the subject hears the tone in interval 3 then he can keep this perception under control (in imagination) only by imagining himself responding "interval 3".

[MT Now] That seems to emphasise and make more precise the claim that the subject generates the choice of answer in imagination. But it contradicts [From Rick Marken (2010.07.15.0910)]: "in my model of an experiment the response is what corrects the error." and "in a control system controlling a relationship between S and R the error correction occurs _when_ R occurs."

BP: That looks contradictory because you're switching from discussing the imagined perception and the error it corrects to discussing the external counterpart of the real perception, which corresponds, we can accept, to the experimenter's perception of the error and its (actual) correction. When Rick says "the response is what corrects the error", I believe he means both the imagined response invisible to the experimenter and the occasional real response which both subject and experimenter can perceive, and the error as either subject alone or both of them experience it.

If the subject's relationship level issues the reference signal to the system that was selected to produce the sensations and sight of pressing the button, and for any reason the result does not create the intended relationship to the light (for example, the subject's finger slips off the button), the subject might immediately try to correct the mistake. If that happens, we can conclude that the subject was perceiving, at the time of the overt "response," the relationship between the light and the actual button press rather than the imagined one. We could arrange for the button to move sideways just as the light comes on.

MMT: Point 2: The overt response is part of the dialogue with the experimenter, and is not a part of the control of the relationship between "stimulus" and "choice of answer" in the 4AFC experiment.

[RM immediately following the first quote above] ....But if the subject does the experiment entirely in imagination it will be impossible for the _experimenter_ to determine which of the intervals the subject selected as the one containing the tone.

[MT Now] How else should I interpret this than as saying that only when the subject makes the overt response can the experimenter know which choice the subject made as to which interval contained the tone?

BP: That is exactly how you should interpret it, I believe.

MMT: And how else should that be interpreted in control-theory terms than that the overt response is part of the set of dialogue loops that include both the subject and the experimenter?

BP: Rick said "entirely in imagination," meaning that the subject would control the perceived relationship without ever doing anything the experimenter could see. You may classify the actual response as "part of the set of dialogue loops that include both the subject and the experimenter," but that affects only your perceptions and has nothing to do with what goes on inside the subject. And even in your own perceptions, the response must actually occur for the experimenter to see, to fit your classification.

MMT: Since these are the main points in the post on which I thought you agreed with me, but about which you say: "At every point in your post where you see me agreeing with something you said, I see continual misunderstanding on your part." would you mind explaining in words of one syllable wherein my continual misunderstanding lies?

BP:I have tried to show you, above. I think you are concealing the real problem by hiding it behind the phrase "the overt response is part of the set of dialogue loops that include both the subject and the experimenter." That imagined set exists only in your mind, and gives you a needed reason to ignore the fact that the subject has to control the relationship of button-press to light either in imagination (to plan it) or in reality (to make it happen overtly). You assume that if the subject has planned the right perception, then all he has to do is issue the reference signal to the right lower system and it will be infallibly carried out. Because it will be carried out reliably, you assume, the subject does not need to perceive the actual result.

That enables you to say that the subject's action does not affect the controlled variable. And to say that you also have to switch the definition of the controlled variable from the relationship between light and button press to just the light. The button press does not affect the light, but it does affect the controlled variable, which is the state of the relationship between button and light.

I can't say it any plainer than that. Your objective appears to be to justify a certain interpretation of a specific experiment. I won't enquire as to why this justification is wanted; that's none of my business. But to do this, you have to switch definitions in mid-sentence and use slippery phrases like the one about the imaginary set. I don't think you ordinarily do this, so I suppose there must be some fairly strong reason. Whatever it is, I don't accept the results.

Best,

Bill P.

[From Bill Powers (2010.07.19.1845 MDT)]

Martin Taylor 2010.07.19.17.16 --

I think we can let this one cool off for a while. I don't want to find myself telling you you don't have a right to pursue anything that interests you.

Best,

Bill P. Still your friend, I trust.

[From Rick Marken (2010.07.19.1900)]

�Martin Taylor (2010.07.19.17.16)--

I realize you and Rick are controlling with a rather high gain a perception
of the value of non-PCT-guided research, with a reference level of "NIL",

This is not quite correct. I am controlling with rather high gain for
the _possibility_ that _all_ behavior, including the behavior in
psychophysical experiments, is _control_ and that the correct model of
this behavior is PCT. This is a guess, of course. It has to be
tested. But it sure looks like subjects in the experiments we have
been taking about -- reaction time and other psychophysical
experiments -- are controlling the variables they are asked to
control, which have to do with what they are supposed to _do_ (R) in
relationship to what is presented to them (S). That is, they seem to
be controlling some kind of relationship between S and R. But this is
just my guess has to be tested.

You, ont he other hand, are proposing that the variables I consider to
be controlled -- the relationship between S and R mainly -- in
psychological experiments are not controlled. This is your guess and
your guess is as good as mine. Now we have to devise a test to see
which guess, if either, is correct. No need for further argument; let
nature decide. For example, we should be trying to think of a way to
test whether the XOR relationship between S and R in a reaction time
task is controlled (as Bill and I think it is) or not (as per your
theory). If it turns out that this variable -- or a similar
relationship between S and R -- is under control, then, indeed, the
information you get by looking at the relationship between S and R
variables in this study is misleading. If, however, you are correct
and the relationship between S and R is not under control then there
is no problem; psychologists have been using the correct methods all
along.

But could you go up a
level, and ask if there is anything you might be controlling that generates
this "NIL" (or any other) reference value for that perception?

Again, that's not quite what I'm controlling for. I'm controlling for
the fact that the value of conventional experiments is essentially
"NIL" IF the behavior under study is closed-loop control. If it's not,
then the results of the experiment are very useful.

And is there
any way of controlling some other perception so as to maintain this
higher-level one near its reference value, while at the same time allowing
for rational discussion of the perception "value of some non-PCT-guided
research"?

Yes, as I said, I think it can be very easily done. Just take your
experiments out in the sun and hold them down on Highway 61...er.. do
the test for the controlled variable to see if the system is
controlling a relationship between S and R. If not, then there is
plenty of value to the non-PCT guided research.

Best

Rick

···

--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com