In a message dated 4/22/99 12:56:33 PM, kurtzer@BINAH.CC.BRANDEIS.EDU writes:
<< I'll make this brief. First, please take all religious controversies
off-line. It is likely important but not appropriate. That goes for
believers, non-believers, and especially smug quotables. >>
i.:
Deal. Amen. Ooops, I mean OK. Glad to see you posting more and advancing
the cause of understanding how behavior works in us pitiful self-serving
people.
Kenny
i.kurtzer (990422.1230)
I'll make this brief. First, please take all religious controversies
off-line. It is likely important but not appropriate. That goes for
believers, non-believers, and especially smug quotables.
Second, instead of explaining how sad it is the PCT gets a raw deal and
how its me against the world why
don't people give examples of how they apply PCT to their work.
M.Lazare's post was a good example of this. His approach to crisis
intervention shows that MOL is one PCT-inspried approach. I would like to
hear him reiterate his approach.
On the side, Rick, optic-flow predicts alpha will be zero. Alpha was
nowhere near zero but clustered around the displacement with one outlier
who may be following optic flow. I would have liked lots of stuff more
from the data recording, but these facts remain: the hypotheses were
chosen depending on which variable the person controlled. If they chose
one then the distrubance would require a certain path, while the other
possible controlled variable was not physically linked to the
disturbance.--and would predict no spontaneous deviation.
Since the path did deviate and cluster around the predicted deviation..not
perfectly, not even enough that the authors were fully satisfied, but
better than the "optic flow" alternative...then that hypothesis is
accpeted until someone gets off their bum to make a better prediction. To
say that optic flow was not refuted because it was not recorded it a point
more than well taken, but its prediction, recorded or not, is completely
off-base. To reintroduce it would require ad hoc assments of whther the
person was changing their goals throught the experiments. This is
generally not a popular practice. I will ask them about if such a measure
was recorded.
If anyone is interested in seeing what the hub-bub is about then I could
send you a copy.
i.
[From Rick Marken (990426.1350)]
i.kurtzer (990422.1230)
First, please take all religious controversies off-line. It is
likely important but not appropriate.
Why? Since PCT is about human nature, it seems to me that discussion
of anything -- arts, humanities, philosophy, religion, etc -- in
terms of PCT is a good idea. No one else is doing it!
Rick, optic-flow predicts alpha will be zero.
Rushton et al's version of the optic flow model predicts that
alpha (FoE) will remain _constant_ at zero _during the walk_; the
predicted behavior of alpha is shown in Figure 2(b) of Rushton et al.
The PCT version of the optic flow model (that FoE is a controlled
variable) predicts that FoE will be controlled at _some_ reference
value; whether or not it is controlled at a particular value, such
as "on target", depends on the subject. The subject could control
for FoE being 16 degrees to the side of the target. If subjects
did control for FoE being 16 degrees to the side of the target
then alpha is _not_ predicted to be zero; it's predicted to
be closer to what Rushton et al observed: 16 degrees.
Alpha was nowhere near zero but clustered around the displacement
with one outlier who may be following optic flow.
See above: the fact that alpha is "nowhere near zero" doesn't say
anything about whether or not subjects are controlling FoE
(remember, behavior is the control of perceptual _variables_).
Also, subjects (like your "outlier") don't "follow optic flow";
if anything, they _control_ it!! (Looks like Brandeis is sucking
you into the behaviorist vortex. Better sign up for the cleansing
waters of the CSG meeting soon!!)
these facts remain: the hypotheses were chosen depending on which
variable the person controlled.
This is not true. The hypotheses were chosen on the basis of which
variable was assumed to be _in control of_ (to guide) locomotion.
If Rushton et al understood the difference between a controlling
and a controlled variable they could have done the research properly.
As it is, their research tells us very little about the variable(s)
controlled when walking to a target.
If they chose one then the distrubance would require a certain
path, while the other possible controlled variable was not
physically linked to the disturbance.--and would predict no
spontaneous deviation. Since the path did deviate and cluster
>around the predicted deviation...
Prisms are not a _disturbance_ to FoE _or_ target image position.
The prisms produce a change in the _feedback function_ connecting
actions (locomotion) to the results of those actions (such as FoE
and target image position).
So there is no _disturbance_ applied to one hypothetical controlled
variable (target image position) and not to another (FoE) in this
experiment. There is only a change in the feedback function that
relates actions to _both_ potential controlled variables (FoE
and target image position).
The results of the Rushton et al experiment don't rule out control
of FoE _or_ target image position. Indeed, the results are what
would be expected if subjects control _both_ variables. Of course,
the results are so noisy that it is impossible to conclude that
the subjects are controlling _any_ variable when the prisms are
on. But let's assume that the results were as predicted; that
alpha was constant and fixed at 16 degrees (as per the prediction
in Figure 2b). That doesn't mean that subjects are _not_ controlling
FoE; it just means they are not controlling FoE at the reference
value "on target". If the subject foveates target position then
FoE is also being kept fixed on some point 16 degrees to the side
of the target. FoE may be at this fixed position because it is
being _controlled_ or because it is at that position as an
uncontrolled _side effect_ of controlling target image position.
So the result of the Rushton et al study (even if they were not
uselessly noisy) would tell us only that subjects _may be_
controlling FoE and/or target image position. But we knew this
before the experiment started, didn't we;-)
The Rushton et al study is not a test for controlled variables;
it leaves the question of what a person controls when walking
to a target as unanswered as it was before Rushton et al strapped
prisms on their subjects;. The Rushton et al study is, however,
an excellent example of how lousy research on control can be when
the researchers know nothing about control.
Best
Rick
---
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken
i.kurtzer (990426.2100)
Richard Marken wrote:
[From Rick Marken (990426.1350)]
i.kurtzer (990422.1230)
>> First, please take all religious controversies off-line. It is
>> likely important but not appropriate.>Why? Since PCT is about human nature, it seems to me that discussion
>of anything -- arts, humanities, philosophy, religion, etc -- in
>terms of PCT is a good idea. No one else is doing it!
This is intended to be a scientific forum. Whether anyone else is doing
anything is irrelevant.
>> Rick, optic-flow predicts alpha will be zero.
>Rushton et al's version of the optic flow model predicts that
>alpha (FoE) will remain _constant_ at zero _during the walk_; the
>predicted behavior of alpha is shown in Figure 2(b) of Rushton et al.
As I understand it, 2b describes the prediction of alpha for the
perceived midline.To quote "Target-locomotor direction error, alpha,
between the instanteneous target and direction of locomotion (the targent
to the curve) which remains constant throughout the trajectory."Since the
target is described in veridical terms rather than subject-centered,
alpha is predicted to be equal to the degree of deflection by the
perceived midline hypothesis.
>The PCT version of the optic flow model
when was this introduced? This is the first I have heard of it.The
prediction they give is based on the idea of a iterative process of error
correction. To quote:"first, walk forward; second, rotate the gaze to
fixate the target; third, rotate the body in a direction that should
reduce the angle between the gaze and the midline; and forth, evaluate
the difference between the angle of the gaze and the orientation of the
body and reiterate the loop". It is the TOTE idea, not PCT, but it
allows a prediction based on an error-correcting mechanism.
>> Alpha was nowhere near zero but clustered around the displacement
>> with one outlier who may be following optic flow.>See above: the fact that alpha is "nowhere near zero" doesn't say
>anything about whether or not subjects are controlling FoE
>(remember, behavior is the control of perceptual _variables_).
>Also, subjects (like your "outlier") don't "follow optic flow";
>if anything, they _control_ it!!
Optic flow was not affect by the preparation, and yet subjects, except
for the outlier, made veering courses. Therefore, the controlled
variable is something else. It might be related to percieved mid-line.
Perceived midline accounted for the data better. Until a better story
comes along, percieved-midline wins.As for outliers, there is no reason
that he/she couldn't be controlling "optic flow". I hope more tests are
done.
>(Looks like Brandeis is sucking
>you into the behaviorist vortex. Better sign up for the cleansing
>waters of the CSG meeting soon!!)
No comment, except this.
>> If they chose one then the distrubance would require a certain
>> path, while the other possible controlled variable was not
>> physically linked to the disturbance.--and would predict no
>> spontaneous deviation. Since the path did deviate and cluster
>>around the predicted deviation...>Prisms are not a _disturbance_ to FoE _or_ target image position.
>The prisms produce a change in the _feedback function_ connecting
>actions (locomotion) to the results of those actions (such as FoE
>and target image position).
Let me think about this. But tentatively I don't think they are changing
the feedback function per se, but only adding a tonic component to the
target image position. If you have a reason to argue for the feedback
function I'd be glad to listen. Please leave the vortex comments to
yourself.
i.
[From Rick Marken (990426.2110)]
Me:
Since PCT is about human nature, it seems to me that discussion
of anything -- arts, humanities, philosophy, religion, etc -- in
terms of PCT is a good idea. No one else is doing it!
i.kurtzer (990426.2100)
This is intended to be a scientific forum.
So? Are certain human activities (art, humanities, philosophy,
religion) off limits to scientific study? I thought PCT is supposed
to help us understand _everything_ people do -- from the sublime
(art) to the ridiculous (religion). What's wrong with testing one's
understanding of a scientific model of behavior (PCT) by seeing
how it might explain behaviors that are both interesting and
important to many people?
As I understand it, 2b describes the prediction of alpha for the
perceived midline.
Figure 2b describes the prediction if the subject controls visual
target location, keeping the target at the midline (fovea); in
this case, alpha will be constant at 16 degrees. What Rushton et al
don't get is that even if the subject does control visual target
location this variable can be controlled at _any_ reference value
the subject pleases. So the subject could be controlling target
location even if alpha is constant at a value other than 16 degrees;
the subject could even be controlling visual target angle if alpha
is constant at _zero_ degrees.
Rushton et al are also wrong to conclude that the subject is
definitely _not_ controlling FoE if alpha is constant at 16 degrees
(rather than zero). Alpha could be constant at 16 degrees if the
subject keeps FoE at a reference point that is 16 degrees to the side
of the target. Indeed, the constancy of alpha at _any_ value tells
us _nothing_ about what variable the subject is controlling.
We don't know what the subject is controlling because Rushton et al
did not apply disturbances to the hypothetical controlled variable
and look for lack of effect. Prisms are not a disturbance because
there is no action (direction of locomotion) the subject can take
(other than removing the glasses) that will counter the effect
of the prism on the possible controlled variables. No matter
what the subject does, the prisms displace visual target location
by 16 degrees. They don't displace the FoE relative to the point
in the visual field toward which you are walking; but the subject
must walk at an angle relative to the visual FoE in order to keep
it on a target that is straight ahead; so the prisms do affect the
feedback function if the output is direction of locomotion relative
to the visual location of the FoE.
Me:
The PCT version of the optic flow model
Issac:
when was this introduced?
I introduced it in my last post. In the PCT explanation, perceptual
_variables_ are controlled relative to reference specifications
that _may vary_.
The prediction they [Rushton et al] give is based on the idea
of a iterative process of error...It is the TOTE idea, not PCT,
but it allows a prediction based on an error-correcting mechanism.
What it doesn't "allow" is recognition that FoE and visual target
location are perceptual _variables_ and if these variables are
controlled they are controlled relative to the subject's "secularly
adjustable" references. Rushton et al assume that FoE must be at
the target (alpha = zero) if it "controls" the subject's locomotion;
PCT tells us something different: the FoE can be at any value the
subject _wants_ it at. This is a good lesson in why people
who talk about "error correction" and TOTEs and other algorithms
that seem PCT-like are just S-R theorists in drag;-)
Optic flow was not affect by the preparation, and yet subjects,
except for the outlier, made veering courses. Therefore, the
controlled variable is something else.
Not true. As I said, even if these veering courses were _exactly_
those expected under the assumption that visual target location
is controlled, they could also be observed if the subject were
controlling FoE at a reference point 16 degrees to the side of
the target. The "veering path" tells us nothing about what variable
(FoE or visual target location, both or neither) is being controlled
when a subject wlaks to a target. The problem is those pesky
reference signals. And, of couse, the fact that Rushton et al
didn't do the test for controlled variables.
If you have a reason to argue for the feedback function I'd be
glad to listen.
I stated my arguments above. The basic idea is that there is
no action (direction of locomotion) that will cancel the effect
of the prisms on the possible controlled variables. No matter
what the subject does (in what direction the subject walks),
the prisms are displacing FoE and visual target location by
16 degrees.
Please leave the vortex comments to yourself.
Sorry. Maybe chaotic attractor would have been more appropriate;-)
Best
Rick
---
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken/
i.kurtzer (990427.0045)
Rick Marken wrote:
[From Rick Marken (990426.2110)]
Me:
> Since PCT is about human nature, it seems to me that discussion
> of anything -- arts, humanities, philosophy, religion, etc -- in
> terms of PCT is a good idea. No one else is doing it!i.kurtzer (990426.2100)
> This is intended to be a scientific forum.
So? Are certain human activities (art, humanities, philosophy,
religion) off limits to scientific study? I thought PCT is supposed
to help us understand _everything_ people do -- from the sublime
(art) to the ridiculous (religion). What's wrong with testing one's
understanding of a scientific model of behavior (PCT) by seeing
how it might explain behaviors that are both interesting and
important to many people?
I have seen no rational testing of the relevance of PCT to art and
religion. The fact that you contrast the sublime with the ridiculous
using art and religion as examples displays an immaturity unbefitting a
man that is past fifty, and a confusion of one's personal opinion with
science.
> As I understand it, 2b describes the prediction of alpha for the
> perceived midline.Figure 2b describes the prediction if the subject controls visual
target location, keeping the target at the midline (fovea); in
this case, alpha will be constant at 16 degrees. What Rushton et al
don't get is that even if the subject does control visual target
location this variable can be controlled at _any_ reference value
the subject pleases.
This is "extra" to the test at hand, and does not criticise the work. I
agree with you and have indicated this to the author. I will relay his
response
Rushton et al are also wrong to conclude that the subject is
definitely _not_ controlling FoE if alpha is constant at 16 degrees
(rather than zero). Alpha could be constant at 16 degrees if the
subject keeps FoE at a reference point that is 16 degrees to the side
of the target.
That is ad hoc though an interesting point that should be followed up.
Indeed, the constancy of alpha at _any_ value tells
us _nothing_ about what variable the subject is controlling.
This seems strange since the physical connection of the world shows that
the projection is a function of the subject's orientation and the
introduced displacement. These two have been found to be negatively
correlated within a range about the known displacement so that the
displacement effects have been compensated if perceived-midline is the
controlled variable. This is assuming a stable reference rather than
your suggested ad hoc varying reference. This assumption is used in any
experiment though it should be checked.
Me:
>The PCT version of the optic flow model
Issac:
> when was this introduced?
I introduced it in my last post. In the PCT explanation, perceptual
_variables_ are controlled relative to reference specifications
that _may vary_.
I think i was just suprised at how it became The PCT model. And a
variable referece is not the sine qua non of PCT.
> The prediction they [Rushton et al] give is based on the idea
> of a iterative process of error...It is the TOTE idea, not PCT,
> but it allows a prediction based on an error-correcting mechanism.What it doesn't "allow" is recognition that FoE and visual target
location are perceptual _variables_
Was it accidental that the authors say perceived midline?
and if these variables are
controlled they are controlled relative to the subject's "secularly
adjustable" references.
which is another interesting question.
Rushton et al assume that FoE must be at
the target (alpha = zero) if it "controls" the subject's locomotion;
This is what the field has argued, the authors take that claim at face
value and introduce a preparation that does not change that and should
lead to no change in the path.
PCT tells us something different: the FoE can be at any value the
subject _wants_ it at.
Why does PCT argue for FoE, why not perceived midline? Why is this PCT
telling us rather than your model?
This is a good lesson in why people
who talk about "error correction" and TOTEs and other algorithms
that seem PCT-like are just S-R theorists in drag;-)
And how the bridge-burning of the net has not changed. Notice I have not
at any time said they were "down " with PCT. I just brought it up as
something we can use. When this is done I would like to hear your
comments about the other paper I sent.
> Optic flow was not affect by the preparation, and yet subjects,
> except for the outlier, made veering courses. Therefore, the
> controlled variable is something else.Not true. As I said, even if these veering courses were _exactly_
those expected under the assumption that visual target location
is controlled, they could also be observed if the subject were
controlling FoE at a reference point 16 degrees to the side of
the target.
Again, this is ad hoc.
The "veering path" tells us nothing about what variable
(FoE or visual target location, both or neither) is being controlled
when a subject wlaks to a target. The problem is those pesky
reference signals. And, of couse, the fact that Rushton et al
didn't do the test for controlled variables.> If you have a reason to argue for the feedback function I'd be
> glad to listen.I stated my arguments above. The basic idea is that there is
no action (direction of locomotion) that will cancel the effect
of the prisms on the possible controlled variables. No matter
what the subject does (in what direction the subject walks),
the prisms are displacing FoE and visual target location by
16 degrees.
That's right . There is a constant effect of the prisms, but how is
that different from a constant disturbance on a cursor? The placement of
the image relative to midline is a function of the actor's orientation
and the lens. The actors act such that the displacement's effect are
minimized. What they minimize most is the disparity of the image with
the perceived midline; i..e until a better experiment comes along that is
the best CV we got.
i.
[From Rick Marken (990427.0930)]
i.kurtzer (990427.0045)--
I have seen no rational testing of the relevance of PCT to art
and religion.
What would the use of a model be if you could only apply it
to situations in which it has been "rationally tested"? If
you followed this rule to the letter you would have to be
against using PCT to understand the behavior of disruptive
students, teachers who deal with disruptive students, etc. Art
and religion are purposeful behaviors; PCT explains purposeful
behavior. Once one knows the basic science of perceptual
control I think it's a good exercise to see how the PCT model
might explain _any_ interesting purposeful behavior.
Me:
What Rushton et al don't get is that even if the subject does
control visual target location this variable can be controlled
at _any_ reference value the subject pleases.
Ye:
This is "extra" to the test at hand, and does not criticise the
work.
This is not an "extra" to the test; it's a real problem. Any test
for controlled variables _must_ take into consideration the
possibility that the reference for a controlled variable may
change; this is why a _continuously varying disturbance_ should
be applied to the hypothetical controlled variable _while_ you
monitor the state of that variable.
Me:
Rushton et al are also wrong to conclude that the subject is
definitely _not_ controlling FoE if alpha is constant at 16
degrees (rather than zero). Alpha could be constant at 16 degrees
if the subject keeps FoE at a reference point that is 16 degrees
to the side of the target.
Ye:
That is ad hoc though an interesting point that should be
followed up.
How is it ad hoc? Actually, I'm not sure it's correct unless
the "reference point" is a retinal location since the point
in the visual field that is at the FoE will be changing. But
my basic point still holds; you can't tell what variable is
under control simply by looking at the behavior of a variable
(like alpha) that is not subjected to variable disturbance.
You can call the prism displacement a "disturbance" if you
like (though their effect on the possible controlled variables
is _not_ independent of the effect of actions on those variables);
but this "disturbance" doesn't change during the course of the
walk to the target. So we don't know whether the stability
(or lack thereof) of any variable (like alpha) is the result
of its being controlled relative to a fixed or variable reference
is an irrelevant side effect of control of some other variable.
Me:
Indeed, the constancy of alpha at _any_ value tells us _nothing_
about what variable the subject is controlling.
Ye:
This seems strange since the physical connection of the world
shows that the projection is a function of the subject's
orientation and the introduced displacement.
The point is that alpha could be constant because the reference
for FoE is changing; it could also be constant because the retinal
location of FoE is controlled at a fixed reference. It could also
be constant because target position is controlled at a reference
other than "at midline" (so alpha could be constant at something
other than 16 degrees -- it could even be zero degrees -- if the
subject is controlling target position). All these possibilities
can be tested only by applying changing disturbances to these
variables _while_ the subject is walking toward the target.
By the way, if the subject controlled only target position then
how does s/he get to the target? The target can be kept at the
midline if the subject is walking toward the target in a particular
trajectory but it can also be kept at the midline if the subject
walks backward or stands still. I have a feeling the subjects
must be controlling something _like_ FoE in order to be able
to walk _toward_ the target.
I think i was just suprised at how it became The PCT model.
And a variable referece is not the sine qua non of PCT.
It's sure _one_ sine qua non! And it makes a _big_ difference
in how one goes about testing for controlled variables. It's
the main thing that distinguishes Rushton's cause effect
view of moving to a target (where FoE or target position are
assumed to _cause_ locomotion) from the control view (where
FoE or target position are assumed to be controlled relative
to a possibly varying reference). Rushton's concept doesn't
allow for the possibility that the target may actually be
staying slightly to the side of the midline (if it does) because
it's the subject's _purpose_ that it be there. Such a result
would indicate (to Rushton et al) that target position is
_not_ the cause of behavior; target position wouldn't _cause_
people to move in the "wrong" direction.
Me:
What it doesn't "allow" is recognition that FoE and visual target
location are perceptual _variables_
Ye:
Was it accidental that the authors say perceived midline?
No. What I was emphasizing in my statement was the term
"variables". Rushton et al didn't seem to realize that these
perceptions are _variable_ and that they can, therefore, be
maintained at any of many different values selected by the
subject.
Me:
PCT tells us something different: the FoE can be at any value the
subject _wants_ it at.
Ye:
Why does PCT argue for FoE, why not perceived midline?
It doesn't. PCT just says that _if_ FoE is controlled it can be
controlled at many possible values.
until a better experiment comes along that is the best CV
we got.
But we've got no CV at all. The plots of alpha should be
straight; they are all curved; this result is consistent
with FoE being controlled or target position being controlled
at a variable reference. This experiment really tells almost
nothing about what the subjects are controlling. As I said,
it could serve as a source of hypotheses for a properly done
test for controlled variables; but one could say that about
just about _any_ conventional experiment.
I guess my point is that there is nothing especially PCT-like
about the Rushton et al study; it's what would be expected from
researchers who have no idea what a controlled variable _is_.
I'll leave the evaluation of the other paper to Bill P. (in
case you sent it to him and he cares to do it). It looks like
it's too complex for me (the Rushton et al study is pushing my
envelope). It's tough to evaluate conventional research from a
PCT perspective because they do a lot of things that _look like_
aspects of the test. But their procedures and assumptions (like the
assumption that alpha = zero means that FoE, not target position,
is controlled) have strange little kinks that are hard to
catch.
I understand your interest in "building bridges" between PCT
and conventional research. It just doesn't seem to work very
well. The only bridge I see between conventional and PCT
research is that conventional research can be a _starting point_
(inasmuch as it provides hypotheses about possible controlled
variables) for appropriate tests for controlled variables.
Best
Rick
--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken