Rushton et al.

CSGnet Archive CSG1999
1

[From Bruce Abbott (990419.1305)]

Rick Marken (990419.1000) --

All this leads me to wonder if it really matters whether one
conceives of perceptual variables as _controlled by_ or _in
control of_ behavior.

We just had a conversation about this last week, and already you've
forgotten -- or at least enclosed it in a different compartment.

Isaac claims that Rushton et al were able to
show that FoE is not a controlled variable (a variable _controlled
by_ behavior) even though they conceived of FoE as a controlling
variable (a variable that is _in control of_ behavior). So they
were able (according to Isaac) to do _the test_ in the context of
a cause-effect model of behavior. So my question is "Does it matter
(in terms of studying behavior) whether one thinks of perception
as controlled by behavior (as in PCT) or in control of behavior
(as in conventional psychology)? And if it does matter, then why
does it matter?

Bill is right that the perception is what is actually under control, but
unfortunately we don't always have access to what other people (or animals)
actually perceive. Furthermore, what matters (for the welfare of the
individual) is not what the person perceives but what actually _is_, out
there in the external world. What makes the whole thing work is that the
perceptions usually correspond extremely well to the physical variables (or
combinations of them) that must be controlled. When you duck and see the
ball whiz by, just to the left of your face, it is usually the case that the
ball actually (by some objective exterior measure) did just miss your face,
and would have hit it if you had not ducked. Our perceptions (which we
often do control) are just proxies for the variables that actually need to
be controlled, and things can go disasterously wrong when the correspondence
between them breaks down.

Rushton et al. suggest that either optic flow or retinal position "guides"
(sets the reference for?) the person's movement toward the target. By
displacing the retinal position of the target _and_ optic flow, these
investigators were able to demonstrate that their participants were _not_
centering the target over the center of optic flow produced by their own
movements, but rather appeared to be attempting to move directly toward the
target, which would tend to foveate the target on the retina if the head and
body axis were kept aligned perpendicular to the axis of gaze or if the
participant could factor in any misalignment. Use of optic flow alone seems
to be ruled out, but it's possible that the observed performances resulted
from some combination of the use of flow and apparent target location.

Bruce

2

[From Bruce Abbott (990419.1555 EST)]

Rick Marken (990419.1300)

Bruce Abbott (990419.1305)

What makes the whole thing work is that the perceptions usually
correspond extremely well to the physical variables (or
combinations of them) that must be controlled.

I think someone should write an essay about everything that is
wrong with that statement. Actually, I think someone already has!
See W. T. Powers, "Applied Epistemology" in LCS I.

Sorry, I don't have access to LCS I, so you'll have to explain it to me. By
the way, just because you (or Bill) thinks there is something "wrong" with
that statement doesn't make it "wrong." Let's just say that you disagree
with it.

from your (vague) criticism, it is hard to tell what you find to disagree
with in the paragraph (only part of which was quoted). So I'll have to ask.
Is it that, in your opinion, when you see a truck bearing down on you at
high speed as you are attempting to cross the sidewalk, the only really
important thing is whether you can get your _perception_ of the truck to
change (so that you no longer see it bearing down on you)? In that case,
why bother to run? Why not just close your eyes and "make it go away"?

In my view the only thing that makes perceptions worth controlling is that
doing so also controls those environmental variables of which the
perceptions are a function. I may withdraw my hand from a stream of hot
water in order to control my perception of pain (Ow! That's HOT!), but the
reason my system is organized in this way is that it prevents real heat from
damaging my real hand, and I'm more likely to survive in the long run if I
can minimize the damage to myself. Such control works well only when my
perceptual systems maintain a good correspondence between the perception of
pain and and the actual temperature of the water.

Changing "guides" to "sets the reference for" doesn't change
the fact that Rushton et al view perception (optic flow or
retinal position) as the _cause_ of behavior.

Do they? Reading your words with the same (lack of) generosity, I take you
to mean that behavior is the cause of perception. Yet I know full well that
both you and Rushton et al. recognize that it's a closed loop, so that
neither description is adequate. In my view you're making a big deal out of
nothing.

Bruce

3

[From Rick Marken (990419.1300)]

Bruce Abbott (990419.1305)--

unfortunately we don't always have access to what other people
(or animals) actually perceive.

Do we _ever_ have "access" to what other people (or animals)
perceive?

What makes the whole thing work is that the perceptions usually
correspond extremely well to the physical variables (or
combinations of them) that must be controlled.

I think someone should write an essay about everything that is
wrong with that statement. Actually, I think someone already has!
See W. T. Powers, "Applied Epistemology" in LCS I.

Rushton et al. suggest that either optic flow or retinal
position "guides" (sets the reference for?) the person's
movement toward the target.

Changing "guides" to "sets the reference for" doesn't change
the fact that Rushton et al view perception (optic flow or
retinal position) as the _cause_ of behavior.

Best

Rick

···

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

4

[From Bruce Gregory (990419.1615 EDT)]

Rick Marken (990419.1300)

Bruce A.

> What makes the whole thing work is that the perceptions usually
> correspond extremely well to the physical variables (or
> combinations of them) that must be controlled.

Rick

I think someone should write an essay about everything that is
wrong with that statement. Actually, I think someone already has!
See W. T. Powers, "Applied Epistemology" in LCS I.

Can we then assume that the perceptions correspond poorly, if at all,
with the physical variables to be controlled? This claim certainly
_sounds_ extraordinary. Don't get me wrong, some of my best friends are
solipsists...

> Rushton et al. suggest that either optic flow or retinal
> position "guides" (sets the reference for?) the person's
> movement toward the target.

Changing "guides" to "sets the reference for" doesn't change
the fact that Rushton et al view perception (optic flow or
retinal position) as the _cause_ of behavior.

You really resist the notion that causes are attributes of models, don't
you?

Bruce Gregory

5

[From Bill Powers (990419.1446 MDT)]

Bruce Abbott (990419.1305)--

Bill is right that the perception is what is actually under control, but
unfortunately we don't always have access to what other people (or animals)
actually perceive. Furthermore, what matters (for the welfare of the
individual) is not what the person perceives but what actually _is_, out
there in the external world. What makes the whole thing work is that the
perceptions usually correspond extremely well to the physical variables (or
combinations of them) that must be controlled.

They do not, however, correspond to the actions by which we control them.
That is, if a certain action produces a particular value of the controlled
variable at one time, in a real environment a _different_ action would
probably be needed to create it a second time. It's only in the laboratory
that a given behavior always has the same effect.

Best,

Bill P.

6

[From Bill Powers (990419.1455 MDT)]

Bruce Gregory (990419.1615 EDT)--

Can we then assume that the perceptions correspond poorly, if at all,
with the physical variables to be controlled? This claim certainly
_sounds_ extraordinary. Don't get me wrong, some of my best friends are
solipsists...

We can assume that perceptions correspond very poorly with physical
measurements of the environment, and in many cases not at all. For example,
when we perceive one light as being twice as bright as another,
psychophysics tells us that the actual brightness ratio is very different,
and moreover changes with the absolute brightness. When we see one picture
as pretty and another as ugly, there is nothing at all in the environment
corresponding to those perceptions (yet an artist can easily control them).

This has nothing whatsoever to do with solipsism.

Best,

Bill P.

7

[From Rick Marken (990419.1410)]

Bruce Gregory (990419.1615 EDT)--

Can we then assume that the perceptions correspond poorly, if
at all, with the physical variables to be controlled?

You can assume whatever you like. But PCT assumes that perceptual
variables are a _function of_ variables in the physics model:
physical variables. If x, y and z are physical variables and
p = k1x+(k2y/k3z)**2 then p is what is controlled. Given this
definition of perception it makes no sense (it seems to me) to
talk about how accurately or poorly perceptions "correspond" to
physical variables.

Me:

Changing "guides" to "sets the reference for" doesn't change
the fact that Rushton et al view perception (optic flow or
retinal position) as the _cause_ of behavior.

Ye:

You really resist the notion that causes are attributes of
models, don't you?

Why do you think that? I agree that cause-effect relationships are
an essential component of any system model -- even a control
model. My point is that, in a closed loop control model, the
perceptual variable doesn't _cause_ behavior; it is controlled
(via the operation of the causal loop) by behavior. In an
"open-loop" model of behavior the perceptual variable _does_
cause behavior.

Best

Rick

···

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

8

[From Bruce Gregory 9990419.1713 EDT)]

Bill Powers (990419.1455 MDT)

We can assume that perceptions correspond very poorly with physical
measurements of the environment, and in many cases not at
all.

We can assume that _some_ perceptions correspond very poorly with
physical measurements. To suggest that most do seems to me at least to
be a perilous extrapolation.

For example,

when we perceive one light as being twice as bright as another,
psychophysics tells us that the actual brightness ratio is
very different,
and moreover changes with the absolute brightness.

Quite so, but this does not imply that the perceptions do not correspond
with physical measurements of the environment. It only tells us that the
function specifying this correspondence is somewhat more complicated
than we might have at first thought.

When we
see one picture
as pretty and another as ugly, there is nothing at all in the
environment
corresponding to those perceptions (yet an artist can easily
control them).

An artist can control his or her perceptions of something they _call_
beauty. But without the Test we have little idea of what perceptions
they are actually controlling.

Bruce Gregory

9

[From Bruce Gregory (990419.1735 EDT)]

Rick Marken (990419.1410)

Bruce Gregory (990419.1615 EDT)--

> Can we then assume that the perceptions correspond poorly, if
> at all, with the physical variables to be controlled?

You can assume whatever you like.

That's most generous of you.

But PCT assumes that perceptual
variables are a _function of_ variables in the physics model:
physical variables. If x, y and z are physical variables and
p = k1x+(k2y/k3z)**2 then p is what is controlled. Given this
definition of perception it makes no sense (it seems to me) to
talk about how accurately or poorly perceptions "correspond" to
physical variables.

For some of us "being a function of" is a form of corresponding to. You
apparently prefer the former. I'm not sure that we have enough evidence
of the manner in which perceptions are constructed to be as certain as
you are that they are mathematical functions.

I agree that cause-effect relationships are
an essential component of any system model -- even a control
model. My point is that, in a closed loop control model, the
perceptual variable doesn't _cause_ behavior; it is controlled
(via the operation of the causal loop) by behavior. In an
"open-loop" model of behavior the perceptual variable _does_
cause behavior.

But we agree, do we not, that there is no such thing as an "open-loop"
model of behavior?

Bruce Gregory

10

[From Rick Marken (990419.1500)]

Bruce Abbott (990419.1555 EST)--

Is it that, in your opinion, when you see a truck bearing down
on you at high speed as you are attempting to cross the sidewalk,
the only really important thing is whether you can get your
_perception_ of the truck to change

Yes.

In that case, why bother to run? Why not just close your eyes
and "make it go away"?

Because visual perceptions are not the only ones people control.
And they are certainly not the most important perceptions to
control when a truck is bearing down on you. But I do think that
it is important that you be able to control these other perceptions
(of being bashed, crushed, and maimed) by controlling your visual
perception of the truck.

Bruce Gregory (990419.1735 EDT) --

But we agree, do we not, that there is no such thing as an
"open-loop" model of behavior?

Not at all. There are many open loop models of behavior; the
open loop model is the basic model of behavior used in the
behavioral sciences today. There certainly is such a thing as
an open loop model of behavior. What I think we would agree on
is that open loop models of behavior are wrong because they
don't account for the data.

Best

Rick

···

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

11

[From Bill Powers (990419.1908 MDT)]

Bruce Abbott (990419.1555 EST)--

Is it that, in your opinion, when you see a truck bearing down on you at
high speed as you are attempting to cross the sidewalk, the only really
important thing is whether you can get your _perception_ of the truck to
change (so that you no longer see it bearing down on you)? In that case,
why bother to run? Why not just close your eyes and "make it go away"?

Do you understand the epistemological problem here? Sure, experience and
logic tell you that the perception of a truck bearing down on you is
probably one to avoid. But you still can't see what it is in reality to
which this perception corresponds. You can make predictions about
relationships among perceptions -- between, for example, a perception of
holding your hand over a candle flame and a perception of heat, then pain.
But that's as far as you can go. The external counterparts of your
experiences remain hypothetical.

And hardware-wise, the ONLY thing your brain can know about is neural
impulses.

Changing "guides" to "sets the reference for" doesn't change
the fact that Rushton et al view perception (optic flow or
retinal position) as the _cause_ of behavior.

Do they? Reading your words with the same (lack of) generosity, I take you
to mean that behavior is the cause of perception. Yet I know full well that
both you and Rushton et al. recognize that it's a closed loop, so that
neither description is adequate. In my view you're making a big deal out of
nothing.

As long as it seems like nothing, you will remain a behaviorists.

Best,

Bill P.

12

[From Bill Powers (990419.1916 MDT)]

Bruce Gregory 9990419.1713 EDT)--

We can assume that _some_ perceptions correspond very poorly with
physical measurements. To suggest that most do seems to me at least to
be a perilous extrapolation.

I don't know where your threshold is for defining "most." Among the
perceptual variables that have no known physical counterparts are
sensations (tart), configurations (amoeboid), transitions (with a few
possible exceptions loike velocity and acceleration), events (tick-tock,
tick-tock, or is it tock-tick, tock-tick?), relationships (distance),
categories (grapes), programs (demo1), principles (safety first), and
system concepts (democracy). All of those, I maintain, are creations of
human perceptual systems. About the only perceptions with any direct
correspondence to physical variables are intensities.

Or try it the other way around. What perceptions correspond to entropy,
valence, quantum spin, density, 6400 Angstroms, X-rays, atmospheric
pressure, work, energy, et. et. et. cetera?

I think you're confusing faith that our perceptions must have SOME sort of
origins in the reality outside us with knowledge of just what those origins
are.

For example,

when we perceive one light as being twice as bright as another,
psychophysics tells us that the actual brightness ratio is
very different,
and moreover changes with the absolute brightness.

Quite so, but this does not imply that the perceptions do not correspond
with physical measurements of the environment. It only tells us that the
function specifying this correspondence is somewhat more complicated
than we might have at first thought.

All right, how do you find out what it is in reality that corresponds to
our physical measurements? For example, when you apply the probes of a
voltmeter to the terminals of a battery, the meter shows a reading that you
can perceive -- a digital number or a deflection of a needle. To what,
inside the battery, does the reading correspond? Can you perceive that,
too? Scientific instruments present one version of reality to our senses
and awareness; our senses, without the instruments, present another version
to awareness. But no instruments and no senses can tell us what the
instruments or the senses are responding to. That is a matter for
hypothesis and theory.

When we
see one picture
as pretty and another as ugly, there is nothing at all in the
environment
corresponding to those perceptions (yet an artist can easily
control them).

An artist can control his or her perceptions of something they _call_
beauty. But without the Test we have little idea of what perceptions
they are actually controlling.

In this case the Test will tell you nothing about reality. All it will tell
you is that when the artist controls his perception of beauty, you also
perceive something being controlled that you are willing to call beauty (or
else you don't, and conclude that the artist isn't controlling anything you
can see). There is no objective counterpart of beauty or ugliness. The Test
does not establish a correspondence of perception with reality. It
establishes a provisional correspondence of the controlling system's
hypothetical perceptions with the observer's perceptions.

Best,

Bill P.

13

[From Bill Powers (990419.1954 MDT)]

But we agree, do we not, that there is no such thing as an "open-loop"
model of behavior?

Open-loop model of behavior:

food-powder blown into mouth ---> miracle ----> salivation.

Perhaps you mean "no _valid_ open-loop model of behavior."

Best,

Bill P.

14

[From Bruce Gregory (990420.1045 EDT)]

Bill Powers (990419.1954 MDT)

>But we agree, do we not, that there is no such thing as an
"open-loop"
>model of behavior?

Open-loop model of behavior:

food-powder blown into mouth ---> miracle ----> salivation.

Perhaps you mean "no _valid_ open-loop model of behavior."

What appear to be open-loop models are, on closer inspection either not
open-loop or fail to account for the phenomenon. As I recall, you made
these points yourself.

Bruce Gregory

15

[From Bruce Gregory (990420.1052 EDT)]

Bill Powers (990419.1908 MDT)

Do you understand the epistemological problem here? Sure,
experience and
logic tell you that the perception of a truck bearing down on you is
probably one to avoid. But you still can't see what it is in
reality to
which this perception corresponds. You can make predictions about
relationships among perceptions -- between, for example, a
perception of
holding your hand over a candle flame and a perception of
heat, then pain.
But that's as far as you can go. The external counterparts of your
experiences remain hypothetical.

Sounds like solipsism to me...

And hardware-wise, the ONLY thing your brain can know about is neural
impulses.

This is a somewhat idiosyncratic use of the word "know".

>Do they? Reading your words with the same (lack of)
generosity, I take you
>to mean that behavior is the cause of perception. Yet I
know full well that
>both you and Rushton et al. recognize that it's a closed
loop, so that
>neither description is adequate. In my view you're making a
big deal out of
>nothing.

As long as it seems like nothing, you will remain a behaviorists.

A theological dispute that is as unlikely to be resolved as all
theological disputes. The only important question is what models are
used and what they predict. If they predict the same things arguing
about them is theology. If they predict different things an experiment
is called for. One of the few valid uses of a theory is to serve as a
license to conduct an experiment.

Bruce Gregory

16

[From Rick Marken (990420.0830)]

Bruce Gregory (990420.1045 EDT)--

What appear to be open-loop models are, on closer inspection
either not open-loop or fail to account for the phenomenon. As
I recall, you made these points yourself.

I think it's pretty easy to tell when a model is open loop;
an open loop model of behavior is described in every
introduction to behavioral research text: it's the _general
linear model_ and it is the basis for the design and analysis
of all conventional behavioral. The general linear model
says that behavior (y) is a linear function of input variables
(x1, x2...xn) plus an additive error component (e):

y = k1x1+k2x2+...kix1*x2+e

There is no question that this is an open loop model: output
(y) is a function of input (x1, x2...) and that's that; there
is no loop.

What we find "on closer inspection" is that this model does not
represent the _observed_ relationship between output and input.
That is, assuming that the x's are sensory variables (which they
are not; another problem with conventional models) the general
linear (open-loop) model of behavior fails to account for the
fact that x1 = f1(y), x2 = f2(y) ... etc. A closed loop model
more accurately represents that actual behavioral situation
by using simultaneous equations to represent the _simultaneous_
effect of input on output and of output on input.

Best

Rick

···

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

17

[From Bill Powers (990420.1048 MDT)]

Bruce Gregory (990420.1045 EDT)--

Perhaps you mean "no _valid_ open-loop model of behavior."

What appear to be open-loop models are, on closer inspection either not
open-loop or fail to account for the phenomenon. As I recall, you made
these points yourself.

Yes. But there are plenty of invalid models of behavior around. You are
right, though, in saying that many supposedly open-loop models actually
require the services of someone who knows what result the model is supposed
to produce, perceives the result it does produce, and continually or
periodically adjusts the model until the error is gone.

Best,

Bill P.

18

[From Bill Powers (990420.1052 MDT)]

Bruce Gregory (990420.1052 EDT)--
ME:

And hardware-wise, the ONLY thing your brain can know about is neural
impulses.

BRUCE:

This is a somewhat idiosyncratic use of the word "know".

Oh, gosh, I'd better change what I mean by the word "know," then. We
wouldn't want to get out of step, would we?

Best,

Bill P.
'

19

[From Bruce Gregory (990420.1315 EDT)]

Bill Powers (990420.1052 MDT)

Bruce Gregory (990420.1052 EDT)--
ME:
>> And hardware-wise, the ONLY thing your brain can know
about is neural
>> impulses.
>
BRUCE:
>This is a somewhat idiosyncratic use of the word "know".

Oh, gosh, I'd better change what I mean by the word "know," then. We
wouldn't want to get out of step, would we?

I would hope not. In any case, the statement that "the ONLY thing your
brain can know about is neural impulses" is a model, not an observation.
I happen to agree that is a perfectly reasonable model, but no less a
model for that. The word "know" in this case is idiosyncratic because
brains are not normally described as knowing anything. "I'm not sure
about this, but my brain knows it," sounds peculiar. It might be more
understandable to say that as far as we know the only inputs to and
outputs from the brain consist of neural impulses. This observation
suggests strongly that we concentrate on understanding the role of these
neural impulses in constructing our view of the world.

Bruce Gregory

20

[From Rick Marken (990420.1030)]

Me:

Changing "guides" to "sets the reference for" doesn't change
the fact that Rushton et al view perception (optic flow or
retinal position) as the _cause_ of behavior.

Bruce Abbott (990419.1555 EST) --

Do they?

Unquestionably. That's why they didn't know how to test for
a controlled variable. They think perceptual variables cause,
control or guide behavior; they have no idea that perceptual
variables are controlled _by_ behavior.

Reading your words with the same (lack of) generosity, I take
you to mean that behavior is the cause of perception.

One of the causes of perceptual variation, yes.

Yet I know full well that both you and Rushton et al. recognize
that it's a closed loop, so that neither description is adequate.

Rushton et al might know (as you do) that moving to a target is
a closed loop process. What they (and you) don't know is how
this loop works. They (like you) don't know that closed negative
feedback loops are organized around the control of perceptual
variables. That is, they (like you) don't know what a controlled
variable _is_. That's why they (like you) have no idea how to
test for controlled variables.

In my view you're making a big deal out of nothing.

Rest assured that your view (behaviorism) is (by far) the majority
view in the behavioral sciences. It just happens to be the _wrong_
view.

Best

Rick

···

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