IV/DV, Do you believe in Demos?

[From Rick Marken (980204.0930)]

Bruce Nevin (980203.2049 PST)--

What's the distinction between controlled input and controlled
perception?

The input is the environmental correlate of the controlled
perception.

Tim Carey (980204.1810)

Is what you're saying this: when we see someone repsond to a
stimulus in a psychophysical experiment, all we are seeing are
actions opposing disturbances?...

Basically, yes. What I am saying is simply that a psychophysical
experiment is the same as any other conventional psychology
experiment; a stimulus (disturbance) variable is manipulated and
a response (output) variable is measured. If the subject of the
experiment is a control system then any observed relationship
between stimulus and response reflects characteristics of the
environment, not the subject. I think I give a reasonably clear
description (and demonstration) of what is going on in such
experiments (psychophysical experiments and all other conventional
experiments) in my "Behavioral Illusion" demo at

http://home.earthlink.net/~rmarken/demos.html

Once one understands the _principle_ illustrated by this demo --
and sees why it works (I describe the same demo, using rubber
bands instead of computer programs, in the "Dancer..." paper) --
one will see why, _in principle_ the results of psychophysical
experiments tell us very little about the subject in the
experiment but a great deal about the subject's environment.

This discussion has led me to a (possibly false) realization
about why some people are able to grasp the revolutionary
significance of PCT while others are not. Of course, much of it
has to do with protecting existing beliefs from the disturbance
of PCT. But I think some of it has to do with a difference in the
way people look at what we call "demos".

When I programmed my first tracking demo and saw that there was
no correlation between cursor movements (stimulus) and handle
movements (response) I realized that, if this held up (and it did,
as shown in my "S-R vs control" demo), then the foundations of
experimental psychology were shattered. I knew that this simple
tracking task illustrated a _general principle_ and that the
principle held whether the controlled input was a cursor in a
tracking task of a drug in an addiction situation; if inputs
don't cause the outputs that keep the inputs under control,
this is true for all the inputs and outputs that are involved in
control.

I feel the same way about the "Behavioral illusion" demo. This
demo illustrates the behavioral illusion in a simple situation
where we know what variable the subject is controlling (shape or
angle) and we know (well, I, the programmer, knows) the feedback
function relating output to input. But I see the demo as illustrating
a general principle -- one that applies to all experiments where
a stimulus is manipulated and a response is measured with no
awareness that both may have an influence on some controlled
perceptual variable -- the principle that we call "the behavioral
illusion".

I look at the PCT demos as illustrations of important general
principles in simple situations where these principles can be
seen clearly. I see the PCT demos of control phenomena as
analogous to Galileo's "demos" of physical phenomena. When I
hear people say "sure, the behavioral illusion occurs in that
simple demo but not in real, psychophysical experiments" I realize
that these people see these demos differently than I do. It's like
hearing someone say to Galileo "sure, balls accelerate downward in
your demo but that doesn't mean that really interesting objects,
like people, bridges, etc.accelerate downward; they're not balls
and they're not rolling on planes".

So there's really nothing I can say in reply to Bruce Abbott
(980204.0505 EST), who apparently sees the PCT demos differently
than I do. Bruce says:

So you see, Rick, when I perform psychophysical studies I am
learning a good seal about the characteristics of the organism,
contrary your assertion that I am only learning about the
characteristics of the environment. This is no "behavioral
illusion."

Maybe all I can do is paraphrase my hero: "And yet, it is an
illusion":wink:

Best

Ricardo Galilei

···

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

[From Bruce Gregory (980204.1250 EST)]

Rick Marken (980204.0930)]

I look at the PCT demos as illustrations of important general
principles in simple situations where these principles can be
seen clearly. I see the PCT demos of control phenomena as
analogous to Galileo's "demos" of physical phenomena. When I
hear people say "sure, the behavioral illusion occurs in that
simple demo but not in real, psychophysical experiments" I realize
that these people see these demos differently than I do. It's like
hearing someone say to Galileo "sure, balls accelerate downward in
your demo but that doesn't mean that really interesting objects,
like people, bridges, etc.accelerate downward; they're not balls
and they're not rolling on planes".

A powerful insight. Congratulations!

Bruce

[From Bruce Nevin (980204.1141 PST)]

(Rick Marken (980204.0930))--

Bruce Nevin (980203.2049 PST)--

What's the distinction between controlled input and controlled
perception?

The input is the environmental correlate of the controlled
perception.

It seems to me that psychophysics is about inferring the relationship
between controlled input (as observed and measured by the experimenter) and
controlled perception (as reported by the participant). Have I got that right?

If the controlled input is at any higher level than first-order
perceptions, then one is also involved with inferring the structure of
input functions and output functions of second-order and higher-order
control systems. Have I got that right?

It seems to me that the latter cannot be done without testing for probably
many controlled perceptions. Have I got that right?

I think that these last two questions touch on the bone of contention.

"Controlled input" confused me. It might not be a good term without
clarification. Of course, "stimulus" is worse. It is the observer's
(relevant) perception, right?

Rick Marken (980203.1950)--

in a psychophysical experiment, you are controlling
a perception (p) that depends on both environmental stimuli (s) and
you own outputs (o):

(1) p = f(g(o) + h(s))

p is the controlled perception
f() is the perceptual function relating controlled input to controlled
perception
g() is the "feedback function" relating o to the controlled input (in
environment)
h() is the function relating the s to the controlled input (in environment)
o is the behavioral output, the organism's means of controlling p.
s is the environmental stimulus, the observer's means of inferring
control of p,
       which you termed the "controlled input".

  Bruce Nevin

[From Bruce Abbott (980204.2000 EST)]

Rick Marken (980204.0930) --

So there's really nothing I can say in reply to Bruce Abbott
(980204.0505 EST), who apparently sees the PCT demos differently
than I do.

Actually, you could do considerably more in reply: you could deal with the
argument I raised. Talking about how convinced you were by the equations
demonstrating the behavioral illusion is not proof that my argument is
incorrect. I am convinced by those equations, too. Your argument seems to
be that if I don't see your point, then I must not understand or accept the
behavioral illusion as real. It is quite apparent to me that you have not
been paying attention to what I actually have been saying, and take me to be
defending some other proposition.

What you are missing is that the experiment determines how the relationship
between o and s (the behavioral illusion) _changes_ as i decreases and p
ceases to vary with i, because the perceptual input mechanism is insensitive
to i below a certain intensity value.

Think, Rick. How would the relationship between s and o change if for every
value of i, the PIF returned zero? I assert that while p would continue to
be controlled (as always), o would no longer vary with disturbances to i.

I look at the PCT demos as illustrations of important general
principles in simple situations where these principles can be
seen clearly. I see the PCT demos of control phenomena as
analogous to Galileo's "demos" of physical phenomena. When I
hear people say "sure, the behavioral illusion occurs in that
simple demo but not in real, psychophysical experiments" I realize
that these people see these demos differently than I do.

When you hear people say "sure, the behavioral illusion occurs in that
simple demo but not in real, psychophysical experiments," you are hearing
what you want to believe they are saying rather than what they are actually
saying. Read my lips, er, words: THE BEHAVIORAL ILLUSION OCCURS IN
PSYCHOPHYSICAL EXPERIMENTS, JUST AS IN THE DEMOS. My argument is not that
the behavioral illusion fails to apply to psychophysical experiments, but
that the information yielded in such experiments nevertheless does tell you
something about the person's perceptual input function. Put _that_ in your
snipe and poke it!

Regards,

Bruce

[From Bruce Gregory (980204.2150 EST)]

Bruce Abbott (980204.2000 EST)

When you hear people say "sure, the behavioral illusion occurs in that
simple demo but not in real, psychophysical experiments," you are hearing
what you want to believe they are saying rather than what they are actually
saying. Read my lips, er, words: THE BEHAVIORAL ILLUSION OCCURS IN
PSYCHOPHYSICAL EXPERIMENTS, JUST AS IN THE DEMOS. My argument is not that
the behavioral illusion fails to apply to psychophysical experiments, but
that the information yielded in such experiments nevertheless does tell you
something about the person's perceptual input function.

Sort of like the information about the disturbance in the perceptual signal,
right?

Put _that_ in your snipe and poke it!

I'm not sure that is tactful.

Bruce

[Martin Taylor 980206 10:45]

Bruce Nevin (980204.1141 PST)

It seems to me that psychophysics is about inferring the relationship
between controlled input (as observed and measured by the experimenter) and
controlled perception (as reported by the participant). Have I got that right?

More between input that is controlled (by a machine) on behalf of the
experimenter and perception that is _not_ controlled by the subject
as reported by the subject.

If the perception were controlled by the subject, there would not be
much relationship between it and the input controlled by the experimenter's
machine, would there? Remember that most of our perceptual world consists
of _un_controlled perceptions. We control only a few at any one time.

If the controlled input is at any higher level than first-order
perceptions, then one is also involved with inferring the structure of
input functions and output functions of second-order and higher-order
control systems. Have I got that right?

I don't understand this question. If the controlled input is a choice
between Mozart and Hendrix, how does that relate to inferring the structure
of any low-level input functions?

It seems to me that the latter cannot be done without testing for probably
many controlled perceptions. Have I got that right?

I suppose that before you ask someone to judge whether a musical snippet
is Mozart or Hendrix, you might want to check whether s/he is deaf. Is
that what you mean? If so, you are right. You might also want to check
whether s/he has ever heard of the two composers. So you would be right
that several levels might have to be checked. It's rather like the
question posed to Rick (and not answered) about the effect of presenting
a variation in level of a sound at an intensity too soft to be heard.
If the subject can't hear the sound at all, s/he is not likely to control
its level very accurately. If s/he has never heard of either Mozart or
Hendrix, s/he isn't likely to identify accurately which snippets belong
to which composer.

I think that these last two questions touch on the bone of contention.

Which bone of contention? Whether it is the subject or the environment that
can tell whether a piece of music is by Mozart, or whether a drop placed
on the tongue is salty? I don't think that your questions touch on that
issue, which seems to me to be the only bone of contention associated
with this topic.

Martin

[Bruce Nevin (980204.1141 PST)]

(Martin Taylor 980206 10:45)--

What I am saying is that to recognize
  Sound
  s
  as music
  of Mozart or Hendrix
involves a number of levels of the perceptual hierarchy.

If the aim is to determine the form of a perceptual input function at the
highest of the levels that are involved in this task--call it category
level, say--one must first determine the effects of perceptual input
functions at the lower levels and take them into account.*

This does not seem to be what is being done. I doubt it is feasible.

I conclude that the aim of psychophysical research cannot be to determine
the form of a perceptual input function except for investigations at the
lowest level of the hierarchy, intensity, e.g. for hearing threshold.

What is the aim of psychophysical research at any but the lowest level of
the hierarchy, intensity? More specifically, what does it contribute to PCT?

I am not denying its relevance to PCT. I don't know enough to make that
judgment one way or the other. I am asking you to explain how
psychophysical research is relevant to PCT.

I will be out of touch for about a week.

  Bruce Nevin

* Otherwise any attempt to deduce perceptual input properties bearing on
this judgment will be as though there were one single PIF for a music genre
detector, or a Mozart vs. Hendrix detector, agglomerating all the levels of
perception that are involved into one input function, an absurdity.

[Bruce Nevin (980206.2155 PST)]

(Martin Taylor 980206 10:45)--

One more note: Yes, the participant is not controlling the perception of
the stimulus that the experimenter's machine is producing. But the
participant is controlling a perception of the relationship between that
perception and some reference value--"audible" or "Hendrix". There are
other kinds of control involved in formulating and delivering a report to
the investigator, perhaps as simple as pressing and releasing a button.
What you are measuring is the participant's ability to control those
perceptions. A part of this is the participant's ability to perceive the
intended input and whatever is involved in becoming aware of that
perception, two distinct things.

I remember a presentation of a language-related study at BBN, some of the
participants were from one of the Eastern European countries, and the
investigator remarked how intensely competitive they were. Their response
times were considerably faster than anyone else's. I forget what he did,
discount them, or fudge the data somehow to get consistent results.

This is groping around the inputs and outputs of a black box. As you know,
the only way we know to open a black box is to model it such that the model
performs like the organism, and the structure of the model is known. Then
you can determine what the form of an nth-level PIF must be for the model,
or the organism, to behave as it does. But from outside the black box, you
get an agglomerated result--or at least you can't tell what you've got.

So we can't get the form of a PIF from psychophysical research. What does
it give us?

  Bruce Nevin