Closed-loop models of "open-loop" behavior

CSGnet Archive CSG2011
1

[From Rick Marken (2011.07.09.0930)]

My “It is the cause” thread seems to have moved to a slightly different topic (the Schouten detection task) so I’ve changed the subject line because I want to get back to what I thought was the original topic: closed loop (control) vs open loop (causal) models of behavior in experiments.

Bill Powers (2011.07.04. 1145 MDT)–

BP: OK, you have a velocity-control system as the lowest level of control. But you can adjust the dynamics of an open-loop model so they exactly mimic the dynamics of a control system model. When the reference
signal is switched from the upper to the lower position, the cursor moves in some particular way – probably a 1 - exp(-kt) curve --to the new position. You can put a filter into the SR system that will do the same thing that your lower two levels of control do, and in the same way. It is always possible to replace a closed-loop model with an open-loop model having suitable dynamics, as long as the independent variables and properties of the environment and output function remain constant. This is why we need the disturbances, because open-loop models
must have a predictable environment free of “unmodeled dynamics” to work. Negative feedback control systems don’t need that kind of coddling.

RM: I just tried filtering the causal model output, using a linear and the exponential filter, and it barely improves things at all f(the R^2 for the causal model goes from .47 to .49, still less than the
R^2 of .62 for the control model). I’ll try some other filtering mechanisms when I get back from celebrating the birth (if not the maturing) of our nation; maybe I’m not filtering it correctly. I’ll give
that ol’ causal model the best deal I can afford;-)

OK, when I use the correct filtering I can get the R^2 for the open-loop model to match that for the closed-loop model. You are absolutely right; it is probably impossible to show the superiority of the closed-loop model of the apparently open-loop behavior in a reaction time task without using a disturbance. I will have to figure out how to do this.

But I think this exercise (in modeling behavior in a reaction time task) was not a complete waste of time. For one thing, it shows that the apparently open-loop behavior in a reaction time task can be modeled as closed-loop control. This is something that I have had a great deal of trouble explaining to reviewers of my papers. What I have done in the reaction time studies is take a task where the behavior seems “obviously” open loop – where the stimulus (the color of the cursor) is not in any way affected by the response (cursor movement) – and shown that the behavior can be modeled as closed loop. This is done by considering that the subject in this task is controlling the perceived relationship between the color of the cursor (the “stimulus”) and the direction of movement of the cursor (the “response”). When the task is “easy” both models account for the behavior in the task (mouse movements ) equally well. But even though there is no basis for selecting one model over the other, the simple fact that it is possible to model the behavior in this task as closed loop should convince conventional psychologists that it is at least worth considering the possibility that the “obviously” open loop behavior they study in their experiments may actually be closed loop.

A second thing about the research that makes it worthwhile (I think) is that it shows that the mechanical use of the causal (general linear) model to analyze the results of experiments can produce poor results because it doesn’t take the dynamics of behavior into account. When the reaction time task was very difficult, the causal model accounts for far less of the predictable variance than the control model because the control model “automatically” takes the dynamics of the behavior into account. The causal model can be made to match the performance of the control model but only after post hoc filtering.

The first point is probably the most important and, I think, makes the effort worth it so far. What do you think? Also, any ideas about how to show that the subject in a reaction time task is indeed controlling the relationship between S and R would be most welcome.

Best

Rick

···

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

2

[From Bill Powers (2011.07.09.1147 MDT)]

Rick Marken (2011.07.09.0930) --

RM: OK, when I use the correct filtering I can get the R^2 for the open-loop model to match that for the closed-loop model. You are absolutely right; it is probably impossible to show the superiority of the closed-loop model of the apparently open-loop behavior in a reaction time task without using a disturbance. I will have to figure out how to do this.

BP: Yes, that is true, and it's the only reason the open-loop model still persists, as far as I can see.

RM: But I think this exercise (in modeling behavior in a reaction time task) was not a complete waste of time. For one thing, it shows that the apparently open-loop behavior in a reaction time task can be modeled as closed-loop control. This is something that I have had a great deal of trouble explaining to reviewers of my papers. What I have done in the reaction time studies is take a task where the behavior seems "obviously" open loop -- where the stimulus (the color of the cursor) is not in any way affected by the response (cursor movement) -- and shown that the behavior can be modeled as closed loop. This is done by considering that the subject in this task is controlling the perceived _relationship_ between the color of the cursor (the "stimulus") and the direction of movement of the cursor (the "response"). When the task is "easy" both models account for the behavior in the task (mouse movements ) equally well. But even though there is no basis for selecting one model over the other, the simple fact that it is possible to model the behavior in this task as closed loop should convince conventional psychologists that it is at least worth considering the possibility that the "obviously" open loop behavior they study in their experiments may actually be closed loop.

BP: Yes, and you can cinch it by introducing a disturbance such that the same movements of the mouse produce a different cursor movement. You can do this by adding a disturbance to the cursor in addition to the mouse's effect, or you might try that reversal thing we did, or just changing the gain of the environmental feedback function. No open-loop model can do that.

In conventional experiments nobody adds disturbances to effects of the response variable so there is never a test to see if the subject could go on producing the same result. When, rarely, it is done, control is seen, not reaction to a stimulus. That is probably why nobody ever suspected there could be something wrong with the cause-effect idea. I think when you write about this, it would be wise to explain why the closed-loop nature of behavior hasn't already been discovered by tens of thousands of experimental psychologists. And to point out that this discovery could have been made at any time.

RM: A second thing about the research that makes it worthwhile (I think) is that it shows that the mechanical use of the causal (general linear) model to analyze the results of experiments can produce poor results because it doesn't take the dynamics of behavior into account. When the reaction time task was very difficult, the causal model accounts for far less of the predictable variance than the control model because the control model "automatically" takes the dynamics of the behavior into account. The causal model can be made to match the performance of the control model but only after post hoc filtering.

BP: That's another good point to bring out. And there's still another that would work in some experiments: measuring the reference level. The reference level is defined as that value of an input quantity at which the output just goes to zero. Clearly, open-loop systems have reference levels, too, although they can't detect errors and therefore can't correct them. That's related to the reason they can't compensate for disturbance of the cursor.

RM: The first point is probably the most important and, I think, makes the effort worth it so far. What do you think? Also, any ideas about how to show that the subject in a reaction time task is indeed controlling the relationship between S and R would be most welcome.

BP: As Martin T. has pointed out, the Schouten experiment is not a reaction-time experiment. But in a true reaction-time experiment, again all you have to do is introduce a disturbance of the output quantity, so something has to be different about the output to get the same relationship to the response button. In a light-button test, you can make pressing the button cause the light to go out. When the person is used to that, you can put in a delay before the light goes out, and see if the person pushes twice trying to make the button work. Turning the light off has nothing to do with the task -- the button has been pressed and the reaction time recorded. You could subsitute a beep for that, or anything else the person can perceive that happens when the button is pushed.

There has to be something about the effect of the action that you can disturb.

···

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

Alice is coming to the meeting after all so her payment will apply to that. We still need to give Allie her free ride, and I hope vote her an honorarium for organizing the meeting, which has used many hours of her time and frequent trips in her car especially while looking for the best venue but also later in working on arrangements with the hotel management.

Autumn Winter has sent a belated notice that he is coming too -- his check is in the mail, he says. I don't know if Francesca is coming.

I'm proposing you as one CSG member of the Saturday panel on PCT meets Neuroscience, and Henry Yin and I as the other two. Randy O'Reilly and two graduate students will represent CU.

Best,

Bill

3

[From Rick Marken (2011.07.10.0800)]

Bill Powers (2011.07.09.1147 MDT)–

Rick Marken (2011.07.09.0930) –

RM: The first point is probably the most important and, I think, makes the effort worth it so far. What do you think? Also, any ideas about how to show that the subject in a reaction time task is indeed controlling the relationship between S and R would be most welcome.

BP: As Martin T. has pointed out, the Schouten experiment is not a reaction-time experiment.

I think that’s just a name-game thing. Perhaps I should just call it an S-R experiment. The essence of my experiment is the same as Schouten’s: a “stimulus” is presented and the subject is to make a “response” based on that stimulus. In both experiments, the response has no effect on the stimulus so, per Martin T., the response to the stimulus is open loop. Here’s what Martin had to say about it:

Martin Taylor (2011.07.05.13.19)–

MT:…the Schouten experiment required a subject
to push a button that corresponded to which of two lights went on.
Nothing about the choice of button push could affect anything about
the light, but the choice of button push was influenced by which
light went on. The link between light selection and button choice
was clearly one-way only (i.e. “open-loop”). Obfuscations about the
mechanism whereby the choice might have been made have no bearing on
the one-way nature of the relation between light selection and
button-choice.

Replace "button push" with "mouse movement" and "lights" with "cursor colors" and you've got a description of my "reaction time" experiment. So according to Martin (and virtually all the reviewers of my papers) the relationship between stimulus (independent variable) and response (dependent variable) in experiments like this, where the subject's response could affect anything about

the stimulus – that is, the behavior in virtually all psychology experiments, is “clearly” open loop. The control model I used to account for the behavior in my “reaction time” task shows that this ain’t necessarily so. Now what I have to do is develop a demonstration to show that this ain’t so, period. Thanks to you and your damn digital filter that’s going to be a bit harder to do now. It can’t be done by just adding a disturbance to the effect of the mouse on the cursor because the open-loopers already accept the fact that the “motor control” systems may be closed loop.

Best

Rick

···

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

4

[Martin Taylor 2011.07.10.11.11]

[From Rick Marken (2011.07.10.0800)]

      Bill Powers

(2011.07.09.1147 MDT)–

      Rick Marken (2011.07.09.0930) --

        RM: The first point is probably the most important and, I

think, makes the effort worth it so far. What do you think?
Also, any ideas about how to show that the subject in a
reaction time task is indeed controlling the relationship
between S and R would be most welcome.

      BP: As Martin T. has pointed out, the Schouten experiment is

not a reaction-time experiment.

      I think that's just a name-game thing.

Why?

      Perhaps I should just call it an S-R experiment. The

essence of my experiment is the same as Schouten’s: a
“stimulus” is presented and the subject is to make a
“response” based on that stimulus. In both experiments, the
response has no effect on the stimulus so, per Martin T., the
response to the stimulus is open loop.

You confound two quite different concepts, whether the response

behaviour is open-loop (which it demonstrably is not), and whether
there is a closed loop by which the action of the subject affects
the perception that is to be reported (which there demonstrably is
not).

Here’s what Martin had to say about it:

Martin Taylor (2011.07.05.13.19)–

MT:…the Schouten experiment
required a subject to push a button that corresponded to
which of two lights went on. Nothing about the choice of
button push could affect anything about the light, but the
choice of button push was influenced by which light went on.
The link between light selection and button choice was
clearly one-way only (i.e. “open-loop”). Obfuscations about
the mechanism whereby the choice might have been made have
no bearing on the one-way nature of the relation between
light selection and button-choice.

        Replace "button push" with "mouse movement" and "lights"

with “cursor colors” and you’ve got a description of my
“reaction time” experiment.

I think not. "Reaction time" experiments are quite different from

signal detection and discrimination experiments, because the
reaction time depends strongly (almost entirely?) on the internal
mechanisms of the control complex, whereas signal detection and
discrimination experiments depend on the control complex only
insofar as it works well enough to allow the subject to report
reasonably accurately what the perception seemed to have been. The
Schouten study was a discrimination experiment, not a reaction time
experiment.

Consider ANY standard tracking study.

Let's take an ordinary pursuit tracking situation. A target moves

according to some disturbance source process over which the subject
has no control. The subject controls a cursor, over which the
disturbance process has no control. The subject perceives something
about the relationship between the target and the cursor. By means
of controlling the cursor, the subject controls this relationship
perception.

There's a pathway from the disturbance source to the movement of the

target and through the subject’s perceptual apparatus to some point
where the target position can be compared with the cursor position.
Up to that point, the subject’s actions have no effect on anything
related to the target and the various perceptions related to the
target. That pathway, from disturbance source to the first place in
which the target perception becomes involved in a control process,
is simple input-output. S-R, if you like, except that there is no
“response” involved.

It's true for every instance of control, that the pathway from the

disturbance source to the point where the disturbance input meets
the input resulting from the control output is a simple connector
with an input at the disturbance source and an output at the
confluence with the control system’s action effect. Why make a
specific fuss when some experiment tries to examine some property of
this pathway?

A reaction-time experiment does not try to examine a property of

this pathway, because the reaction time involves the operations of
the control system that produces the “response”. This is why it is
not a “name-game thing” to point out that the Schouten study is not
a reaction-time experiment.

Why also do you equate psychophysical experiment looking at the

disturbance input pathway with “the behaviour in virtually all
psychological experiments”? I would say that outside psychophysics,
the results of most psychological experiments do depend on the
operations of a complex of control loops. In a psychophysical
experiment, what depends on the actions of the control complex is
that the subject is presumed to be trying to report accurately what
is perceived, and is reasonably able to do so. Failure in this
regard can only reduce the mutual information between report and the
experimenter-controlled disturbance. It can’t increase it, but if
control is good, the reduction will be minimal.

The likelihood that the control systems are not absolutely perfect

is the reason careful psychophysicists regard their experimental
results as being lower bounds on the subject’s perceptual ability.
And that’s why the control mechanisms involved in creating the
response, while interesting in themselves, are of little concern
when the object of study is the input pathway.

··· 
In respect of your "reaction time" study, some months ago I was

doing something similar by using a square disturbance in the
TrackAnalyze program (I mean that all movements of the target happen
at random times and are abrupt from “old level” to “new level”).
Bill had suggested (as I think do you) that a two-level control
system with no loop delay could be as good a match to the human as a
one-level system with a loop delay. This seems to be reasonably true
with a smooth disturbance, but it is far from true with the square
disturbance. A one-level control loop with a fixed delay provides a
much better fit. It would be interesting to know if this is true for
your experiment, in which the cursor colour signals the change of
target level.

Martin
5

[From Bill Powers (2011.07.10.1116 MDTY)]

Rick Marken (2011.07.10.0800) –

BP: As Martin T. has pointed out, the Schouten experiment is not a
reaction-time experiment.

I think that’s just a name-game
thing. Perhaps I should just call it an S-R experiment.

BP: It’s nothing simple. It’s actually a very complex,
hard-to-model experiment, at least as I recall it (I can’t find my
references to it).
As I remember it, when the light goes on the subject is supposed to wait
for a beep and then make a judgment as to whether one of two lights is on
or off, even if not sure yet, and press one of two buttons to indicate
which light. So this involves perception of light, perception of pressing
a button, perception of a beep in relation to each of two lights, a
relationship of each button to a light, and the relationship of the
chosen button press to the correct light. The variable investigated is
the way the percentage of correct indications changes as the delay
between onset of the light and the time when the beep occurs is
manipulated. If I am remembering incorrectly, Martin can correct
me.
Martin is emphasizing the obvious fact that pressing the button has no
effect on the status of the light. There is no path from the button to
the light-bulb, so clearly there is no loop of that kind. However, he is
also saying that there is no effect of pressing the button on any
perception, and that isn’t correct. Pressing a button affects the
perception of the relationship between the press and the lights and many
other perceptions of higher order than light intensity. The elements to
be considered are two lights, two buttons, a beep, and visual and
proprioceptive aspects of the pressing action. The perceptions involved
of relationships, sequences, and logical relationship among the elements
are numerous and occur at quite a few levels. There’s no simple way to
know which levels are involved in determining the final result.

RM: The essence of my experiment
is the same as Schouten’s: a “stimulus” is presented and the
subject is to make a “response” based on that stimulus. In both
experiments, the response has no effect on the stimulus so, per Martin
T., the response to the stimulus is open loop. Here’s what Martin
had to say about it:

Martin Taylor (2011.07.05.13.19)–

MT:…the Schouten experiment required a subject to push a
button that corresponded to which of two lights went on. Nothing about
the choice of button push could affect anything about the light, but the
choice of button push was influenced by which light went on. The link
between light selection and button choice was clearly one-way only (i.e.
“open-loop”). Obfuscations about the mechanism whereby the
choice might have been made have no bearing on the one-way nature of the
relation between light selection and button-choice.

BP: Yes, you are both oversimplifying the experiment to fit your
conception of what it is about. That makes it easier to analyze, of
course, but since the analysis in both cases ignores complexities in the
experiment, the analysis is more or less irrelevant. It applies to an
imaginary experiment in which ONLY the aspects you choose to describe
exist. The other aspects, in both your cases, are dismissed as
unimportant or, as Martin sometimes puts it, outside your range of
interests. I’m sure you can both justify doing that, especially since
that’s done to some extent in analyzing all experiments. But everything
important you choose to ignore is quicksand into which any conclusion can
sink without a trace.

RM: Replace “button
push” with “mouse movement” and “lights” with
“cursor colors” and you’ve got a description of my
“reaction time” experiment.

BP: What about “number of lights” and “number of
buttons” and “timing of beep” and “waiting for the
beep” and so on? I guess if you put those things aside, your
experiment is exactly like Schouten’s.

RM: So according to Martin (and
virtually all the reviewers of my papers) the relationship between
stimulus (independent variable) and response (dependent variable) in
experiments like this, where the subject’s response could affect anything
about the stimulus – that is, the behavior in virtually all psychology
experiments, is “clearly” open loop.

BP: I think you meant to write “Could not affect …” That is
perfectly true. Nothing about the response can affect the stimulus,
because the stimulus, a physical variable outside the person, can’t be
affected by the response, a visible consequence of an output from the
nervous system, that occurs after the stimulus happens. The light bulb
stays lit no matter how hard or fast you hit the button. There are,
however, many perceptions involved that include the response as one
perceived element, and those are all affected by changes in the
response. Perhaps you failed to get across the idea that you were talking
about perceptions inside the organism, not events and variables in the
environment. That is important to explain when the audience seldom talks
about perceptions inside the organism (and hates the idea that there can
be any difference between perception and reality in a “trained
observer”).

RM: The control model I used to
account for the behavior in my “reaction time” task shows that
this ain’t necessarily so. Now what I have to do is develop a
demonstration to show that this ain’t so, period. Thanks to you and your
damn digital filter that’s going to be a bit harder to do now. It can’t
be done by just adding a disturbance to the effect of the mouse on the
cursor because the open-loopers already accept the fact that the
“motor control” systems may be closed loop.

BP: Why don’t you just leave out the digital filter and the other things
I said, then, and use what’s left to prove the point you want to make? I
didn’t mean to inconvenience you. You know what I would say about the
result, of course, but you haven’t even tried offering a bribe yet. Maybe
you could persuade me to keep this just between you and me. Of course
that leaves the possibility open that Martin might offer more, but what
the heck, guys, business is business.

On the other hand, maybe before you can persuade the “motor
controllers,” you will have to introduce the idea of levels of
perception and control and convince them that they really exist. That
would be a useful accomplishment and might soften up the
resistance.

Best,

Bill P.

6

[Martin Taylor 2011.07.10.17.35]

[From Bill Powers (2011.07.10.1116 MDTY)]

  Rick Marken (2011.07.10.0800) --

        BP: As Martin T. has pointed out, the Schouten experiment

is not a
reaction-time experiment.

    I think that's just a

name-game
thing. Perhaps I should just call it an S-R experiment.

  BP:       It's nothing simple. It's actually a very complex,

hard-to-model experiment, at least as I recall it (I can’t find my
references to it).

  As I remember it, when the light goes on the subject is supposed

to wait
for a beep and then make a judgment as to whether one of two
lights is on
or off, even if not sure yet, and press one of two buttons to
indicate
which light. So this involves perception of light, perception of
pressing
a button, perception of a beep in relation to each of two lights,
a
relationship of each button to a light, and the relationship of
the
chosen button press to the correct light. The variable
investigated is
the way the percentage of correct indications changes as the delay
between onset of the light and the time when the beep occurs is
manipulated. If I am remembering incorrectly, Martin can correct
me.

So far, so good.

  Martin is emphasizing the obvious fact that pressing the button

has no
effect on the status of the light. There is no path from the
button to
the light-bulb, so clearly there is no loop of that kind. However,
he is
also saying that there is no effect of pressing the button on any
perception, and that isn’t correct.

I didn't expect you would get that interpretation from anything I

wrote. I thought I emphasized that the subject’s action no effect on
any perception * ** prior to where the perceptual signals
participate in the lowest level control loop*** . That’s
pretty much a tautology, isn’t it? The perception that the light is
on or off is not a perception the subject can control (other than by
not looking). The perception of whether this implies an answer “1”
or “2” is a perception the subject can control.

After that, of course the subject's actions influence perceptions.

Those higher-level perceptions are, or may be, controlled. The
experimental data have no bearing on them. The only effect control
of those perceptions can have on the data is to reduce the mutual
information between response and experimenter-generated event.

You have to study control system variables using control system

methods.

Martin
···

On 2011/07/10 2:30 PM, Bill Powers wrote:

7

[From Rick Marken (2011.07.10.1520)]

Martin Taylor (2011.07.10.11.11)–

Rick Marken (2011.07.10.0800)]

      RM: Perhaps I should just call it an S-R experiment. The

essence of my experiment is the same as Schouten’s: a
“stimulus” is presented and the subject is to make a
“response” based on that stimulus. In both experiments, the
response has no effect on the stimulus so, per Martin T., the
response to the stimulus is open loop.

MT: You confound two quite different concepts, whether the response

behaviour is open-loop (which it demonstrably is not), and whether
there is a closed loop by which the action of the subject affects
the perception that is to be reported (which there demonstrably is
not).

Actually, I’m saying that there is a closed loop by which the action of the subject affects the perception that is to be reported. You’re model is this:

S → P → R

Mine is this:

S -->P → R
^ |
>___|

See the difference.

Best

Rick

···

–
Richard S. Marken PhD
rsmarken@gmail.com

www.mindreadings.com

8

[From Rick Marken (2011.07.10.1540)]

Bill Powers (2011.07.10.1116 MDTY)]

Rick Marken (2011.07.10.0800) –

BP: Martin is emphasizing the obvious fact that pressing the button has no
effect on the status of the light. There is no path from the button to
the light-bulb, so clearly there is no loop of that kind.

I know.

However, he is
also saying that there is no effect of pressing the button on any
perception, and that isn’t correct.

Exactly. A pretty important “not correct” if you ask me. It’s the “not correct” analysis that I hear from every reviewer as well. It’s the “not correct” that allows one to continue studying control systems as though they were S-R systems.

BP: Yes, you are both oversimplifying the experiment to fit your
conception of what it is about. That makes it easier to analyze, of
course, but since the analysis in both cases ignores complexities in the
experiment, the analysis is more or less irrelevant. It applies to an
imaginary experiment in which ONLY the aspects you choose to describe
exist. The other aspects, in both your cases, are dismissed as
unimportant or, as Martin sometimes puts it, outside your range of
interests. I’m sure you can both justify doing that, especially since
that’s done to some extent in analyzing all experiments. But everything
important you choose to ignore is quicksand into which any conclusion can
sink without a trace.

I’m not trying to analyze the Schouten experiment. I’m simply showing that an apparently open-loop behavior (like making a discriminative response to a stimulus – which is what subjects did in my unfortunately named “reaction time” experiment) can be modeled as closed loop.

RM: So according to Martin (and
virtually all the reviewers of my papers) the relationship between
stimulus (independent variable) and response (dependent variable) in
experiments like this, where the subject’s response could affect anything
about the stimulus – that is, the behavior in virtually all psychology
experiments, is “clearly” open loop.

BP: I think you meant to write “Could not affect …” That is
perfectly true.

Correct.

RM: The control model I used to
account for the behavior in my “reaction time” task shows that
this ain’t necessarily so. Now what I have to do is develop a
demonstration to show that this ain’t so, period. Thanks to you and your
damn digital filter that’s going to be a bit harder to do now. It can’t
be done by just adding a disturbance to the effect of the mouse on the
cursor because the open-loopers already accept the fact that the
“motor control” systems may be closed loop.

BP: Why don’t you just leave out the digital filter and the other things
I said, then, and use what’s left to prove the point you want to make?

I don’t need to. I already explained the points I’m making and they are made just fine even if I include the findings with the digital filter.

I
didn’t mean to inconvenience you.

Oh my god. You thought I was actually mad at you for having pointed out the equivalence of the outputs by an open and closed loop system when there are no disturbances. The “damn digital filter” comment was meant as a joke. I’m thrilled that you pointed it out to me; thrilled that I was able to implement it successfully; and thrilled that I have more work to do to prove that the closed loop model is, in fact, superior to the open loop model of the behavior in this task (if it is;-).

I’ll guess I’ll have to work on my humor skills as well;-)

Best

Rick

···

You know what I would say about the
result, of course, but you haven’t even tried offering a bribe yet. Maybe
you could persuade me to keep this just between you and me. Of course
that leaves the possibility open that Martin might offer more, but what
the heck, guys, business is business.

On the other hand, maybe before you can persuade the “motor
controllers,” you will have to introduce the idea of levels of
perception and control and convince them that they really exist. That
would be a useful accomplishment and might soften up the
resistance.

Best,

Bill P.

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

9

[From Bill Powers (2011.07.10.1941 MDT)]

Martin Taylor 2011.07.10.17.35 –

BP earlier: Martin is
emphasizing the obvious fact that pressing the button has no effect on
the status of the light. There is no path from the button to the
light-bulb, so clearly there is no loop of that kind. However, he is also
saying that there is no effect of pressing the button on any
perception, and that isn’t correct.

MT: I didn’t expect you would get that interpretation from anything I
wrote. I thought I emphasized that the subject’s action no effect on any
perception ** prior to where the perceptual signals participate in the
lowest level control loop**. That’s pretty much a tautology, isn’t
it?

BP: Is there any perception prior to the signals in the lowest level?
Anyway, the lowest level may not contain the perception that the person
is controlling, which has been my point all along. You can’t tell from
the data. If, instead of controlling the perception of light intensity
(which is impossible without shutting the eyes or looking away as you
say), the person might be controlling some relationship of which the
light intensity and the perceptions of button pressing are components.
Either control process could involve pressing a button when a light goes
on. However, the relationship perception can be affected by both the
state of the light and the state of the press. If that is the controlled
variable, then the observed discrimination effects are not due to light
intensity alone, vitiating the analysis based on the assumption that they
are. I don’t know what the alternate analysis would have to be, but it
would be different because there are two variables where only one was
present before.

If you assume that the subject is attending only to light intensity
without the state of the pressing being involved at all, then all you say
about the results is true. But we can’t see what the subject is focused
on and nothing was done by Schouten that would enlighten us.

Everything I am saying about your analysis applies equally to the model I
proposed, in which I assumed that there was a perception of light
intensity that rose with time after the light turned on, and a noise
level such that more mistakes were made immediately after the light
turned on than some time later. That was only my attempt to find an
explanation given that intensity was the perception of importance. But I
wouldn’t offer that as the correct explanation, simply because of the
considerations above: I have no way of knowing what the controlled
variable really was and therefore what the subject was actually trying to
do.

In my analyses of tracking data, I always check the apparent reference
level to see if the subject understood the instructions concerning the
object of the task. That gives me some confidence that the subject
interpreted the instructions as I intended.

MT: The perception that the
light is on or off is not a perception the subject can control (other
than by not looking). The perception of whether this implies an answer
“1” or “2” is a perception the subject can
control.

BP: True, assuming that the perception that the light is on or off is the
whole story. I don’t know if it is. I know that if I were doing the task,
I would be very conscious of the relationship of the buttons to the
lights, and my aim would be to press the right button that goes with the
light that was on when the beep occurred. That’s what I would be
controlling for: not just a button press, but the right button
press. The states of the buttons and the beep would be in awareness all
the time. It’s a very complex experiment.

MT: After that, of course the
subject’s actions influence perceptions. Those higher-level perceptions
are, or may be, controlled. The experimental data have no bearing on
them. The only effect control of those perceptions can have on the data
is to reduce the mutual information between response and
experimenter-generated event.

BP: Maybe that’s the origin of the observed effect and determines its
observed size. I don’t think we will ever know unless we redesign the
experiment and do it all over again.

You have to study control system
variables using control system methods.

Gosh. Really?

Bill

10

[Martin Taylor 2011.07.10.22.37]

[From Rick Marken (2011.07.10.1520)]

        Martin Taylor

(2011.07.10.11.11)–

Rick Marken (2011.07.10.0800)]

                RM: Perhaps I should just call it an S-R

experiment. The essence of my experiment is the same
as Schouten’s: a “stimulus” is presented and the
subject is to make a “response” based on that
stimulus. In both experiments, the response has no
effect on the stimulus so, per Martin T., the
response to the stimulus is open loop.

        MT: You confound two quite different concepts, whether the

response behaviour is open-loop (which it demonstrably is
not), and whether there is a closed loop by which the action
of the subject affects the perception that is to be reported
(which there demonstrably is not).

      Actually, I'm saying that there _is_ a closed loop by which

the action of the subject affects the perception that is to be
reported. You’re model is this:

      S --> P --> R



      Mine is this:



      S -->P --> R

               ^       |

               >___|



       See the difference.

I see the difference in the picture, but I don't think either of

them actually represents what either of us model. I can create lots
of different pictures and ask if you can see the difference. The
fact that I can see the difference in the pictures says something
about my eyesight, not about the models either of us would bring to
the party.

In your model, what perception do you think is controlled? And why

would you think I would model a straight-through link from the
perception of the event to the output of the response? What relation
does the model you ascribe to me have to anything I have written on
the topic, recently or in the past?

···

On 2011/07/10 6:21 PM, Richard Marken wrote:

11

[From Bill Powers (2011.07.10.2020 MDT)]

Rick Marken (2011.07.10.1520) --

Actually, I'm saying that there _is_ a closed loop by which the action of the subject affects the perception that is to be reported. You're model is this:

S --> P --> R

Mine is this:

S -->P --> R
         ^ |
         >___|

See the difference.

BP: That is what I see on my screen. When you do ASCII diagrams, you have to specify a monospaced font such as Courier in order to preserve positions.

Actually, Martin's model is the same as yours, but he's assuming that P is an intensity perception, bottom level, and that R has no effect on it, which it doesn't. My model is

S --> P1 ----> P2 ---> ... Pn --> R
                            ^ |
                            >______|
                          n levels of
                       kinesthetic perceptions

... where it is Pn rather than P1 that the subject tries to discriminate when the beep occurs.

As I said to Martin, we will never know unless we redesign the experiment.

PCT AXIOM: No organized behavior ever occurs except for the purpose of controlling a perception.

To view that diagram, make sure you use the Courier font to display it.

Best,

Bill P.

12

[Martin Taylor 2011.07/10/22.42]

[From Bill Powers (2011.07.10.1941 MDT)]

  Martin Taylor 2011.07.10.17.35 --

      BP earlier: Martin

is
emphasizing the obvious fact that pressing the button has no
effect on
the status of the light. There is no path from the button to
the
light-bulb, so clearly there is no loop of that kind. However,
he is also
saying that there is no effect of pressing the button on any
perception, and that isn’t correct.

    MT: I didn't expect you would get that interpretation from

anything I
wrote. I thought I emphasized that the subject’s action no
effect on any
perception ** * prior to where the perceptual signals
participate in the
lowest level control loop*** . That’s pretty much a
tautology, isn’t
it?

  BP: Is there any perception prior to the signals in the lowest

level?
Anyway, the lowest level may not contain the perception that the
person
is controlling, which has been my point all along.

Well, it's nice to know that your point and my point are the same.

It really reduced an argument to just a bit of fun, doesn’t it?

  You can't tell from

the data. If, instead of controlling the perception of light
intensity
(which is impossible without shutting the eyes or looking away as
you
say), the person might be controlling some relationship of which
the
light intensity and the perceptions of button pressing are
components.

So we both have assumed, yes.

  In my analyses of tracking data, I always check the apparent

reference
level to see if the subject understood the instructions concerning
the
object of the task. That gives me some confidence that the subject
interpreted the instructions as I intended.

As I pointed out many months ago, all practiced psychophysicists

perform “The Test for the controlled variable”, whether they know it
or not. I explained how they did it, and you agreed this was so.

But even if they did not, a consistently imperfect level of

correspondence between the experimenter-controlled event and the
subject’s response is a very difficult thing to fake (I’ve tried),
and to produce a level of correspondence better than one’s real
perceptual ability is impossible to do unless there is some kind of
“Clever Hans” effect that the experimenter has missed.

All in all, I really can't see a disagreement between what you say

in order to prove me wrong, and what I have said. It looks like
disagreement for the sake of disagreement, not disagreement over any
technical point.

Martin
13

[From Bill Powers (2011.06.11.0245 MDT)]

Martin Taylor 2011.07/10/22.42

BP: Is there any perception
prior to the signals in the lowest level? Anyway, the lowest level may
not contain the perception that the person is controlling, which has been
my point all along.

MT: Well, it’s nice to know that your point and my point are the same. It
really reduced an argument to just a bit of fun, doesn’t
it?

BP: Wait a minute, how can that be true? You have been emphasizing that
the response of selecting the right button and pressing it can’t affect
the light bulb’s being on or off, and is very unlikely to affect the
intensity-level perception of the intensity. I have agreed with that. But
that does not hold true for a higher-level perception which is derived
from both the light intensity and the perceptions involved in pressing
the button.

I said

BP earlier: You can’t tell from
the data. If, instead of controlling the perception of light intensity
(which is impossible without shutting the eyes or looking away as you
say), the person might be controlling some relationship of which the
light intensity and the perceptions of button pressing are
components.

And you said, to my astonishment,

MT: So we both have assumed,
yes.

BP: In my view, that is exactly the point on which we have disagreed the
most directly. I have been maintaining that the discrimination could be
of a relationship; you have been maintaining that it is of an
intensity.

The requirements of the task suggest that subject takes the signal to
indicate when to review short-term memory to try to see what the
relationship of the light to the press was at the time of the beep.
Seeing a memory of a light on and the right press under way, or no light
on and no press under way, would match the reference condition and not
lead to initiating a new press. Seeing the light on and no press or the
wrong press under way is an error, which is corrected if possible by
initiating the right press (if necessary, after aborting the wrong
one).

A mistake could indicate that either the memory of the state of the press
was in error, or the memory of the state of the light, or both. It could
also indicate that the correct relationship was perceived, but the wrong
press was initiated. If all the experimenter knows is that a relationship
between two perceptions has been disturbed, there is no way to know which
one of them changed. And if all he knows is whether the right or the
wrong response occurred, he has no idea of where in the process the
mistake originated on any one trial. We have to know the nature of the
perceptual input functions to decide that, and we don’t.

MT: As I pointed out many months
ago, all practiced psychophysicists perform “The Test for the
controlled variable”, whether they know it or not. I explained how
they did it, and you agreed this was so.

BP: I may have agreed to a specific positive instance of your claim, but
I can’t imagine having agreed to the generalization. You are saying
“they” when you must mean “some of them”, and
implying “all the time” when you must mean
“sometimes.” I certainly don’t agree with it now. Testing for
the controlled variable is highly unlikely to occur by accident, though
that’s not inconceivable as a rare occurrance. It requires a particular
strategy of forming hypotheses and then applying disturbances (or waiting
for them to occur naturally) to test them. That’s hard to do
unintentionally.

MT: But even if they did not, a
consistently imperfect level of correspondence between the
experimenter-controlled event and the subject’s response is a very
difficult thing to fake (I’ve tried), and to produce a level of
correspondence better than one’s real perceptual ability is impossible to
do unless there is some kind of “Clever Hans” effect that the
experimenter has missed.

BP: I’m not looking for a fake response, only for an honest report of an
apparent relationship that is ambiguous, or merely a mistaken action. The
person could correctly perceive the state of the light every time but
emit the wrong response when required to select it prematurely, as when
one tries to type too fast. Does a typo indicate that you intended the
wrong letter and typed it, or the right letter and mistyped it? You can’t
tell from looking at what was typed.

MT: All in all, I really can’t
see a disagreement between what you say in order to prove me wrong, and
what I have said. It looks like disagreement for the sake of
disagreement, not disagreement over any technical
point.

BP: And from over here, it looks as if you might be defending a bad
experiment in which you happen to have a personal interest in the form of
a published paper analyzing it.

It would be better if neither of us were guilty of the respective
accusations of nefarious motives. It’s not like you to descend to the ad
hominem level.

Best,

Bill P.

14

[Martin Taylor 2011.07.11.09.02]

[From Bill Powers (2011.06.11.0245 MDT)]

Martin Taylor 2011.07/10/22.42

BP: Is there any perception prior to the signals in the lowest level? Anyway, the lowest level may not contain the perception that the person is controlling, which has been my point all along.

MT: Well, it's nice to know that your point and my point are the same. It really reduced an argument to just a bit of fun, doesn't it?

BP: Wait a minute, how can that be true? You have been emphasizing that the response of selecting the right button and pressing it can't affect the light bulb's being on or off, and is very unlikely to affect the intensity-level perception of the intensity. I have agreed with that. But that does not hold true for a higher-level perception which is derived from both the light intensity and the perceptions involved in pressing the button.

True, and long agreed, except that I would not be so dogmatic as to assume that the perceptions involved in pushing the button are the first controlled perceptions in which the disturbance signal consequent on the event are involved.

I have left unspecified what I think the first controlled perception might be, because to specify it is to start building the model of the control processes that produce the response. To build that model is a worthy endeavour, an endeavour that occupied a long thread a while back, but an endeavour irrelevant to the analysis of the psychophysical experiment, which requires only that the control processes function reasonably reliably.

I said

BP earlier: You can't tell from the data. If, instead of controlling the perception of light intensity (which is impossible without shutting the eyes or looking away as you say), the person might be controlling some relationship of which the light intensity and the perceptions of button pressing are components.

And you said, to my astonishment,

MT: So we both have assumed, yes.

BP: In my view, that is exactly the point on which we have disagreed the most directly.

Then I guess you haven't been reading my recent writings, and have not remembered our previous discussion, in which we started from an agreement that the first controlled perception was likely to be a relationship perception. Our disagreements in that thread were centred on just what the relationship might be, not on whether the controlled perception was a relationship between the perceptual input from the event and some other perception.

I have been maintaining that the discrimination could be of a relationship; you have been maintaining that it is of an intensity.

Can you find anywhere I said that? It is directly contrary to my understanding.

The requirements of the task suggest that subject takes the signal to indicate when to review short-term memory to try to see what the relationship of the light to the press was at the time of the beep. Seeing a memory of a light on and the right press under way, or no light on and no press under way, would match the reference condition and not lead to initiating a new press. Seeing the light on and no press or the wrong press under way is an error, which is corrected if possible by initiating the right press (if necessary, after aborting the wrong one).

We both seem to have the same hypothesis, yes. But remember that it is a hypothesis, or perhaps better, a speculation. Nevertheless, it is one on which we agree.

MT: As I pointed out many months ago, all practiced psychophysicists perform "The Test for the controlled variable", whether they know it or not. I explained how they did it, and you agreed this was so.

BP: I may have agreed to a specific positive instance of your claim, but I can't imagine having agreed to the generalization. You are saying "they" when you must mean "some of them", and implying "all the time" when you must mean "sometimes." I certainly don't agree with it now.

No, I realize you are controlling for perceiving me to disagree with you. I don't know why you find this important, but apparently you do. I wish it were not so, but I know not what to do about it.

An experimenter who did not test that the subject could and would "follow instructions" (i.e. succeed in producing the response expected by the experimenter under conditions in which the discrimination should be clear to the subject) would have failed some very early course in lab technique. The same goes for tests of the apparatus, etc. etc.

In Schouten's case, part of the test was to ensure that the subject would time the button press reasonably well when presented with beeps much delayed after the light onset so that the light choice would have been clear at beep-time. He did measure when the actual responses were made, and the data are based on those measurements, not on the beep timing.

Testing for the controlled variable is highly unlikely to occur by accident, though that's not inconceivable as a rare occurrance. It requires a particular strategy of forming hypotheses and then applying disturbances (or waiting for them to occur naturally) to test them. That's hard to do unintentionally.

I'm not sure where "unintentionally" gets into the act. We do it all the time in everyday interaction. "The Test" is simply a formalized version of how we go about finding what people want and why they do thus-and-so. In experimental technique, it was long ago formalized.

MT: But even if they did not, a consistently imperfect level of correspondence between the experimenter-controlled event and the subject's response is a very difficult thing to fake (I've tried), and to produce a level of correspondence better than one's real perceptual ability is impossible to do unless there is some kind of "Clever Hans" effect that the experimenter has missed.

BP: I'm not looking for a fake response, only for an honest report of an apparent relationship that is ambiguous, or merely a mistaken action. The person could correctly perceive the state of the light every time but emit the wrong response when required to select it prematurely, as when one tries to type too fast. Does a typo indicate that you intended the wrong letter and typed it, or the right letter and mistyped it? You can't tell from looking at what was typed.

Again we agree. Most psychophysical experimenters accept that even practiced subjects have some level of "typo" error, typically in the range of 1% to 2%. This is the primary reason why high d' values are usually taken with a very large grain of salt, and why the Schouten experiment is important.

MT: All in all, I really can't see a disagreement between what you say in order to prove me wrong, and what I have said. It looks like disagreement for the sake of disagreement, not disagreement over any technical point.

BP: And from over here, it looks as if you might be defending a bad experiment in which you happen to have a personal interest in the form of a published paper analyzing it.

It would be better if neither of us were guilty of the respective accusations of nefarious motives. It's not like you to descend to the ad hominem level.

Nor like you. The preceding two paragraphs seem to conflict with each other.

···

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

We DO agree on all the technical detail that I can see. Where we do NOT agree is on how the agreed technical detail affects the interpretation of experiments.

Your position seems to be that whatever the technical facts of the case, if an experiment was not conceived with full knowledge of PCT, it is fatally flawed, and for this reason alone, Schouten's experiment is a "bad experiment" and the only conceivable reason I defend it is that I published a reanalysis of his data, and I must be controlling a self-image of never having made a mistake.

My position is that if one looks at an experiment through PCT spectacles, and the analysis shows that the results are meaningful, then the data must be regarded as meaningful.

In the case of psychophysical experiments, there are only two components in the pathway between experimenter-controlled events and subject-controlled responses: (1) whatever perceptual events occur to produce a signal that directly disturbs a controlled perception, and (2) whatever control processes produce an output that counters that disturbance. The analysis of the results of a psychophysical process reflect only the processes in stage (1), provided that stage (2) works well. Other kinds of study are required in order to test models of how stage (2) may function.

Martin

15

[From Bill Powers (2011.06.11.0822 MDT)]

Martin Taylor 2011.07.11.09.02 --

BP earlier: Wait a minute, how can that be true? You have been emphasizing that the response of selecting the right button and pressing it can't affect the light bulb's being on or off, and is very unlikely to affect the intensity-level perception of the intensity. I have agreed with that. But that does not hold true for a higher-level perception which is derived from both the light intensity and the perceptions involved in pressing the button.

MT: True, and long agreed, except that I would not be so dogmatic as to assume that the perceptions involved in pushing the button are the first controlled perceptions in which the disturbance signal consequent on the event are involved.

BP: But you are (or someone was) dogmatic in assuming that the light's onset was the event deemed to have been perceived at the time the beep sounded. Post hoc, ergo propter hoc?

MT: I have left unspecified what I think the first controlled perception might be, because to specify it is to start building the model of the control processes that produce the response. To build that model is a worthy endeavour, an endeavour that occupied a long thread a while back, but an endeavour irrelevant to the analysis of the psychophysical experiment, which requires only that the control processes function reasonably reliably.

BP: Which control processes? It also requires that you know what the subject was perceiving and what the intent was in acting. When you present a stimulus to a subject, you have no knowledge of what perceptions are being derived from the physical inputs to the subject's nervous system. Of course you have an informal idea and your own experience of it to go on, and those are useful in helping you form hypotheses to test. But when the time comes to test them, you have to ignore all those informal things and try to make an airtight scientific case for what you claim is going on. Yes, it seems that the subject is responding open-loop to the onset of the light. That's obvious to you and I can understand why -- but when we discount that obviousness and look for actual proof, the problems begin to appear. The necessary questions weren't asked and the necessary hypotheses were not tested. Too many unsupported premises were introduced. The result is a well-defined measure of the behavior of almost undefined variables.

I think the missing details matter. You evidently don't. So clearly, we're not going to agree on the significance of the Schouten experiment or your analysis of it until that disagreement ends. At this moment I don't see any path by which we can reach agreement, so I recommend that we simply shelve this whole discussion until such time as one of us or someone else sees something neither of us has seen.

Best,

Bill P.

16

[From Rick Marken (2011.07.11.1140)]

Bill Powers (2011.07.10.2020 MDT)–

BP: Actually, Martin’s model is the same as yours, but he’s assuming that P is an intensity perception, bottom level, and that R has no effect on it, which it doesn’t.

Exactly. Martin’s model is the same as mine except that it’s different.

My model is

S → P1 ----> P2 —> … Pn → R

                       ^      |

                       >______|

                     n levels of

                  kinesthetic perceptions

… where it is Pn rather than P1 that the subject tries to discriminate when the beep occurs.

Right. Your model is completely different than mine, except that it’s the same. Since you seem to be doing a nice job of explaining the difference between your model and Martin’s (and mine, since mine is the same as Martin’s, only different) I’ll just duck out here. I was just confusing things by explaining the difference between my model and Martin’s since I was actually explaining the difference between my model and mine(which is the same as Martin’s). I can hardly wait for the meeting and similarly enlightening discussions;-)

Best

Rick (doing it in Courier) Marken

···

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

17

[From Bill Powers (2011.07.12.1551 MDT)]

Rick Marken (2011.07.12.1130) --

I'm trying to convince behavioral researchers that the point you made in your 1978 Psych Review article does not just apply to the obviously closed-loop behavior in a tracking task (the task you used to demonstrate the analysis in the Psych Review paper) but to all behavior (which was your point in that article), including the behavior in psychological experiments, which appears to be open loop because "the subject's output can have not effect on the input" (which is what I am constantly told by reviewers).

I think the strategy has to be to admit right away that what they are going to say is true: the response has no effect on the input. However, it has an effect on something else that is also affected by the response.

Also, to bolster that idea, you can outline how you see if that hypothesis stands up to testing -- what you would take as confirmation and what as discomfirmation. It's a model, of course, but not just a guess.

I guess we just keep on keepin' on.

Best,

Bill P.

18

So now it's "the first controlled perception." I have no idea what you mean by that, and as usual you don't explain it. I don't know what you mean by a "disturbance signal" or by "consequent on the event" or "are involved", either. Reading these posts makes me feel aphasic.

I've been looking over some of the old posts from 2009 in which the arguments were no better than they are now, and in which there were no conclusions reached that we agreed on. It's just hopeless. There is so little actual communication between us in this whole series, going back as far as I have traced it, that there is no way to make sense of the discussion at all. I give up. Whatever has gone wrong here doesn't look fixable.

Bill P.

···

At 07:13 PM 7/14/2011 -0400, you wrote:

[Martin Taylor 2011.07.14.18.51]

[From Bill Powers (2011.06.11.0822 MDT)]

Martin Taylor 2011.07.11.09.02 --

BP earlier: Wait a minute, how can that be true? You have been emphasizing that the response of selecting the right button and pressing it can't affect the light bulb's being on or off, and is very unlikely to affect the intensity-level perception of the intensity. I have agreed with that. But that does not hold true for a higher-level perception which is derived from both the light intensity and the perceptions involved in pressing the button.

MT: True, and long agreed, except that I would not be so dogmatic as to assume that the perceptions involved in pushing the button are the first controlled perceptions in which the disturbance signal consequent on the event are involved.

19

From Bill Powers (2011.07.11.1455 MDT)]

Rick Marken (2011.07.11.1140) –

Bill Powers (2011.07.10.2020 MDT)–

BP: Actually, Martin’s model is the same as yours, but he’s assuming
that P is an intensity perception, bottom level, and that R has no effect
on it, which it doesn’t.

Exactly. Martin’s model is the same as mine except that it’s
different.

BP: Hmm. I didn’t say that very well, did I? What I was trying to say was
that Martin’s model was described as if the perception being
time-discriminated was light intensity, which means that even if there
were a feedback connection from the response to that perception, the
feedback effect would come too late to affect the judgment of the state
of the light. But in fact the apparatus does not connect back to the
light and thus it doesn’t allow the response to affect the perception of
intensity at all.

My model proposes that there is feedback from the response at a higher
level (relationships, for example) and no feedback at the intensity
level. So what the experimenter thinks is a discrimination of the on-off
state of the light might instead pertain to the relationship of sensed
button press to light. That would appear to be a response to the onset of
the light, when it was actually a response to a change in a derived
perception of higher order that includes the state of the response in its
definition. So part of the “mutual information” would be
information about the state of the response used in calculating the
mutual information. As in all closed-loop arrangements, the apparent
relationships are not the actual ones.

The reason I proposed a model at all was simply to show that there was an
alternative to what the experimenter assumed. And even farther in the
background was the simple fact that we have no way other than surgical
procedures to verify any proposal concerning what perceptual signals are
doing inside the brain. There is no way to check directly the assumptions
about what is going on inside the organism, and without a model Schouten
couldn’t even use the indirect model-based method we use.

I know this kind of experiment is common, but I can’t take it seriously.
We’d be better off trying to understand the circuitry and the signals in
it, and only then, if there are questions remaining, finding new ways of
characterizing the behavior of the actual rather than the hypothetical
signals. This verges on being a matter of taste rather than science, so I
suppose I should just nod and say “very interesting” and let it
go.

Best,

Bill P.

20

[From Rick Marken (2011.07.12.1130)]

Bill Powers (2011.07.11.1455 MDT)–

Rick Marken (2011.07.11.1140) –

Bill Powers (2011.07.10.2020 MDT)–

BP: Actually, Martin’s model is the same as yours, but he’s assuming
that P is an intensity perception, bottom level, and that R has no effect
on it, which it doesn’t.

RM: Exactly. Martin’s model is the same as mine except that it’s
different.

BP: Hmm. I didn’t say that very well, did I? What I was trying to say was
that Martin’s model was described as if the perception being
time-discriminated was light intensity, which means that even if there
were a feedback connection from the response to that perception, the
feedback effect would come too late to affect the judgment of the state
of the light. But in fact the apparatus does not connect back to the
light and thus it doesn’t allow the response to affect the perception of
intensity at all.

My model proposes that there is feedback from the response at a higher
level (relationships, for example) and no feedback at the intensity
level.

Mine too.

So what the experimenter thinks is a discrimination of the on-off
state of the light might instead pertain to the relationship of sensed
button press to light. That would appear to be a response to the onset of
the light, when it was actually a response to a change in a derived
perception of higher order that includes the state of the response in its
definition. So part of the “mutual information” would be
information about the state of the response used in calculating the
mutual information. As in all closed-loop arrangements, the apparent
relationships are not the actual ones.

Yes, much like in our model of magnitude estimation (http://www.mindreadings.com/BehavioralIllusion.pdf).

The reason I proposed a model at all was simply to show that there was an
alternative to what the experimenter assumed. And even farther in the
background was the simple fact that we have no way other than surgical
procedures to verify any proposal concerning what perceptual signals are
doing inside the brain. There is no way to check directly the assumptions
about what is going on inside the organism, and without a model Schouten
couldn’t even use the indirect model-based method we use.

These are good points relative to the psychophysical aims of the Schouten experiment. My goal is somewhat more general. I’m trying to show that in all psychological experiments where the behavior appears to be open-loop it is actually closed loop; behavior is always aimed at controlling some perception. In psychological experiments with humans at least one of those controlled perceptions is what the subject is instructed to control; in most experiments it’s a perception of a relationship between the IV and DV.

I’m trying to convince behavioral researchers that the point you made in your 1978 Psych Review article does not just apply to the obviously closed-loop behavior in a tracking task (the task you used to demonstrate the analysis in the Psych Review paper) but to all behavior (which was your point in that article), including the behavior in psychological experiments, which appears to be open loop because “the subject’s output can have not effect on the input” (which is what I am constantly told by reviewers).

Best

Rick

···

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