IV/DV

[Bill_Powers1]

[From Bill Powers (980211.1630 MST)]

Bruce Abbott (980210.2140 EST)--

Tim Carey asked

How can you study the intensity of a CV unless you first test to find out
what the CV actually _is_?

Let us say that I conduct a study in which the participant is given control
over tone intensity and asked to maintain that intensity at a given value.
Are you saying that I don't know what the CV actually is? How about in a
tracking study? The participant is asked to keep the cursor over the
target, and can control the position of the cursor by using the computer's
mouse. I don't know what the CV actually is? So much for the Test, huh?

Yes, I would say that in the experiment you describe, you don't know what
the CV actually is. I would guess that it is some function of tone pitch
and tone intensity. You could easily show that tone intensity is not the CV
by varying the pitch -- the threshold intensity of the tone would be
_changed_ by the participant instead of being maintained constant.

Just presenting something to the participant does not establish it as a CV.
You must actually do the Test. Tracking experiments are designed so the
Test is done every time -- in fact, continuously. The cursor position is
disturbed in all the ways possible, and it is established that the handle
movements vary so as to cancel the effects of a disturbance on the
relationship of the cursor to the target. In some experiments disturbances
are applied simultaneously to target and cursor positions. The data we get
from these experiments are a direct measure of how well the proposed CV
passes the Test.

Assume that my psychophysical study reveals that the participant cannot hear
a tone below a certain intensity level.

In your setup, you show that the participant responds to tones below a
certain intensity by increasing their intensity. You can't say anything
about whether the participant can _hear_ those tones. Tim Carey caught
this, too.

(In the study, the participant has
no control over the tone's intensity, but can only signal whether she does
or does not hear the tone.) This finding leads to the prediction that the
person's ability to control the intensity to a particular reference level
will fail when the reference level is set well below the participant's
threshold for tone intensity.

Do you mean that the participant will not be able to maintain the intensity
at a level too faint to hear? The lowest you can set the reference level is
zero. Any heard sound will then lead to behavior that restores the sound
level to zero. I don't see why you say control is lost. I think it would
work just as well as ever.

  Unless my participant is trying to dupe me

(and ve have vays of finding out), it would seem that the psychophysical
study has indeed discovered something important about the participant's
perceptual input function for tone intensity, important enough to determine
whether or not the participant can or cannot establish control over the CV.

You can establish the participant's control over the perception of tone
intensity very easily using any level as the reference condition. I don't
see what's special about zero. The intensity can't be less than zero, so
zero is just another point on the scale of perceptual magnitudes. I don't
think the threshold value is important at all.

Best,

Bill P.

[Bruce_Abbott]

[From Bruce Abbott (980211.2005 EST)]

At last a free moment . . .

Rick Marken (980211.2310) --

Me:

But if the relationship between s and o deteriorates (and those
fammously high correlations become low correlations) when the
tone drops below a certain intensity, it is almost certainly
the case that control is failing because the tone can no longer
be perceived (p = 0, although i is > 0).

Now, here is a puzzle for you. As f(i) appears nowhere in the
equation relating o to s, how is it that this experiment tells
us something about f(i)?

Rick:

I don't understand why you think this is a puzzle. In this
experiment you have apparently determined (using the Test) that
i is a controlled variable. Since what is actually controlled
(according to PCT) is p, which is a function of i [f(i)], we
learn about f(i) by building a model containing an estimated
perceptual function [f'(i)] that controls like the subject.

But that isn't the puzzle. It's right above this paragraph of yours I
quoted. You might try reading it and see whether you can answer. I asked
how the experiment I described tells us something about f(i), not how some
other procedure does, such as model-building and testing. You have asserted
(with NO proof) that psychophysical procedures tell us little or nothing
about the organism, pointing to the "behavioral illusion" as the source of
your belief. I have shown that the procedure I described _does_ provide
information about f(i) _despite_ the "behavioral illusion." So I'm asking
you how it can do this if, as you assert, the behavioral illusion prevents
us from gaining such information about the organism in these psychophysical
studies.

But continuing with your line of thinking above, where does your estimated
perceptual function come from? Wouldn't the search for such a function
benefit from the knowledge that the participant cannot perceive the input
when it is below a given intensity?

We don't learn about f(i) by looking at the relationship between
o and s;

The question is not whether we do, but whether we _can_.

we learn about f(i) by modeling the subject's ability
to control i;. If the model controls i (and loses control of i)
as the subject does, then we can assume that we have a pretty
good estimate of f(i).

You are like the builder of a bridge who determines whether the girders are
strong enough for the load by building the bridge and observing whether or
not it collapses when several trucks attempt to drive over it. A structural
engineer would test the girders for their ability to withstand various loads
and would then develop a stress analysis of the bridge structure, showing
what loads each girder would be subjected to. The psychophysical procedure
is equivalent to the stress-testing of individual components, and the
modeling to the stress analysis of the structure.

In most psychophysical experiments, stimuli (like tone bursts)
are presented and the subject responds to indicate something
about the stimulus (is it there?, is it louder than another?, etc).
In such an experiment, we don't know what perception is being
controlled. We don't know the perception that is influenced by
the stimuli (disturbances); this was Bill's point to Martin.
All we know is that there is some relationship between our
stimuli and the subject's responses. This relationship, I argue,
tells you nothing about what the subject is perceiving (if the
subject is a control system). You have to Test to determine what
perception is being protected from variations in what is typically
though of as the stimulus.

Actually, you don't argue, you assert (over and over). I have noted already
that traditional psychophysical methods such as the method of limits, the
method of constant stimuli, and the method of adjustment give essentially
the same results when applied to the same individual under the same
conditions. It seems to me that this entirely disproves your claim that the
only experiments that provide valid results are those in which the
participant controls the aspect of the stimulus being tested (the method of
adjustment). How can one method "tell you nothing about what the subject is
perceiving" and the other provide that information when the results of the
two methods are comparable? You are speaking nonsense.

In particular, I challenge you to explain in a satisfactory manner why
saying "present" when you hear a tone and "absent" when you don't, as the
experimenter varies the intensity of the tone, should give different results
than pressing a button when you do not hear a tone and releasing it when you
do, wherein pressing and releasing put the direction of change in tone
intensity in the hands of the participant.

Regards,

Bruce

[R_Marken]

[From Rick Marken (980211.2120)]

Bruce Abbott (980211.2005 EST)

But that isn't the puzzle. It's right above this paragraph of
yours I quoted. You might try reading it and see whether you
can answer.

I'm lost. Why not start over again. Could you describe the puzzle
from scratch.

I have shown that the procedure I described _does_ provide
information about f(i) _despite_ the "behavioral illusion."

You're obviously convinced that the behavioral illusion creates
no problem for conventional psychophysics or for psychological
research in general. I believe that the behavioral illusion is
one of the most profound revelations we get from PCT and that it
knocks the foundations out from under conventional behavioral
science.

Basically, the behavioral illusion shows (rather dramatically)
that you can't learn what people are doing (or how they do it) by
just observing what they do (their visible behavior). Bill Powers
(980211.0316 MST) gave some examples of this (the fielder looking
like he is _predicting_ the position of the fly ball and the person
looking like he is trying to _select a route_ through a crowd) in
his reply to Dan Miller. The behavioral illusion shows that you
have to find out what a control system is trying to _perceive_ in
order to undertand what it is doing -- and how it is doing it.

But if you don't find the demos and simulations persuasive, then
you don't.

Best

Rick

···

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

[Bruce_Abbott]

[From Bruce Abbott (980212.1005 EST)]

Bill Powers (980211.1528 MST) --

Bruce Abbott (980210.1540 EST)

But to satisfy my curiosity, in your view, Bill, how would the mere use of
words invalidate the results of psychophysical studies?

No.

No? Methinks ye have misread ye question.

I can say "yes" and
"no" as easily as I can press or release a button, and thereby report the
same states of a particular perception.

Hard to establish _what_ perception, since you can't observe anyone else's
perceptual signals.

Perhaps, but not at all difficult to find a physical correlate of that
perception (e.g., tone intensity) in the sensory studies I've been discussing.

And in a tracking study, I can just
as easily report whether the cursor is above or below target as I can push a
key to make the cursor, now above the target, move down (thereby signifying
that I perceive the cursor to be above the target) or release the key to
make the cursor, now below the target, move up (thereby signifying that I
perceive the cursor to be below the target).

All true. However, all irrelevant to the point that we can't observe anyone
else's perceptual signals. We can see the cursor going up and down (since
that is our own perception), but we can't see the perceptions of the other
person. Your guesses about what the other person's movements "signify" are
only that: guesses.

I'd call them inferences, as they would be based on information garnered in
the study. The notion of a guess sounds more like uninformed speculation.

I know, Bill, that we can't observe anyone else's perceptions. Whether you
are doing psychophysics or tracking studies you are faced with the same
limitation, and in both cases you can only infer as best you can the nature
of the "other's" perceptions, based on the information available to you. If
you insist on observing the "other's" perceptions before accepting any
findings, both psychophysics and PCT are out the window for you.

Bill Powers (980211.1630 MST) --

Bruce Abbott (980210.2140 EST)

Let us say that I conduct a study in which the participant is given control
over tone intensity and asked to maintain that intensity at a given value.
Are you saying that I don't know what the CV actually is? How about in a
tracking study? The participant is asked to keep the cursor over the
target, and can control the position of the cursor by using the computer's
mouse. I don't know what the CV actually is? So much for the Test, huh?

Yes, I would say that in the experiment you describe, you don't know what
the CV actually is. I would guess that it is some function of tone pitch
and tone intensity. You could easily show that tone intensity is not the CV
by varying the pitch -- the threshold intensity of the tone would be
_changed_ by the participant instead of being maintained constant.

Your inference fails here, because the person is still maintaining the tone
intensity at threshold (the reference value the person has been asked to
use), but the threshold intensity (reference value) is changing with tone
frequency.

Just presenting something to the participant does not establish it as a CV.

True.

You must actually do the Test. Tracking experiments are designed so the
Test is done every time -- in fact, continuously. The cursor position is
disturbed in all the ways possible, and it is established that the handle
movements vary so as to cancel the effects of a disturbance on the
relationship of the cursor to the target. In some experiments disturbances
are applied simultaneously to target and cursor positions. The data we get
from these experiments are a direct measure of how well the proposed CV
passes the Test.

True. So now you have something that is a physical correlate of the aspect
of perception the person is controlling. You still don't know what
perception the person is controlling. Your guesses about what the person's
movements signify are only that: guesses.

Assume that my psychophysical study reveals that the participant cannot hear
a tone below a certain intensity level.

In your setup, you show that the participant responds to tones below a
certain intensity by increasing their intensity. You can't say anything
about whether the participant can _hear_ those tones. Tim Carey caught
this, too.

I described a couple of setups, but didn't provide enough detail about one
of them. In that hypothetical study, the center intensity of the tone would
be set well below threshold and the participant's ability to vary the tone's
intensity would be limited to a relatively small range that did not include
values above threshold. In another phase, the same range of variation would
be under the person's control but the center intensity would be set above
threshold. I predict that the person will be unable to control the tone's
intensity in the first case and able to do so in the second.

As both you and Tim Carey note, it is possible for the participant to feign
not being able to hear (or control) the tone. I am not interested in
assessing the psychophysical functions of uncooperative participants, any
more than you are interested in assessing the tracking abilities of such
individuals. Have you had a lot of trouble in this regard? That is, in
your experience, are those whom you have tested in tracking studies been, as
a rule, uncooperative? Have any? Furthermore, internal checks can be made
to detect "cheating," for example the insertion of so-called "catch" trials
in which no signal is presented, and examination of the psychophysical
function, which would be extremely difficult for "cheaters" to duplicate.
Then there is the ability to compare functions across individuals; in
persons with normal hearing, vision, and so on these will be very similar.

(In the study, the participant has
no control over the tone's intensity, but can only signal whether she does
or does not hear the tone.) This finding leads to the prediction that the
person's ability to control the intensity to a particular reference level
will fail when the reference level is set well below the participant's
threshold for tone intensity.

Do you mean that the participant will not be able to maintain the intensity
at a level too faint to hear? The lowest you can set the reference level is
zero. Any heard sound will then lead to behavior that restores the sound
level to zero. I don't see why you say control is lost. I think it would
work just as well as ever.

I made this distinction earlier, but apparently need to repeat it. I am
talking about the loss of correlation between disturbance and output when
the tone intensity falls below threshold. If you ask the participant to
control the tone intensity at zero, but the tone for this person disappears
at some above-zero value, then what you will find is that all variation in
intensity below the person's threshold will produce no compensatory change
in output. The person goes on controlling her _perception_ of the tone's
intensity, but her control over its physical correlate, the tone's physical
intensity, is lost.

You can establish the participant's control over the perception of tone
intensity very easily using any level as the reference condition. I don't
see what's special about zero. The intensity can't be less than zero, so
zero is just another point on the scale of perceptual magnitudes. I don't
think the threshold value is important at all.

See above.

Regards,

Bruce

[Bill_Powers1]

[From Bill Powers (980212.0834 MST)]

Bruce Abbott (980211.2005 EST)--

In particular, I challenge you to explain in a satisfactory manner why
saying "present" when you hear a tone and "absent" when you don't, as the
experimenter varies the intensity of the tone, should give different results
than pressing a button when you do not hear a tone and releasing it when you
do, wherein pressing and releasing put the direction of change in tone
intensity in the hands of the participant.

Seems to me that the first question to ask is "Does it make a difference?"
Why try to settle a question of fact by logical reasoning?

Best,

Bill P.

[Bill_Powers1]

[From Bill Powers (980212.0914 MST)]

Bruce Abbott (980212.1005 EST)--

But to satisfy my curiosity, in your view, Bill, how would the mere use of
words invalidate the results of psychophysical studies?

No

No? Methinks ye have misread ye question.

The problem is not the use of words (what use did you have in mind?). The
problem is that in psychophysics, just as in PCT, you need a model of
perception in order to draw any conclusions about the relation between the
external stimulus and the internal experience. In PCT we can adopt an
arbitrary scaling factor by normalizing the perceptual input function to a
multiplier of 1; then the parameters in the comparator and on the output
side become what they must be to account for the closed-loop behavior. Of
course the input function and output function might have compensating
nonlinearities, and the actual scaling might be anything at all, but this
gives us a model that predicts, which is what we're after. Much more
detailed neurological data will be needed to go beyond that.

I can say "yes" and
"no" as easily as I can press or release a button, and thereby report the
same states of a particular perception.

Hard to establish _what_ perception, since you can't observe anyone else's
perceptual signals.

Perhaps, but not at all difficult to find a physical correlate of that
perception (e.g., tone intensity) in the sensory studies I've been

discussing.

I think you'e missing the point here. If you take the viewpoint of the
participant, you have only the inner experience to exasmine; if you take
that of an objective observer, you can see only the external physical
stimulus. Nobody can do both. But to establish the relation between the
external stimulus and the inner experience, you would _have_ to do both.

Your guesses about what the other person's movements "signify" are
only that: guesses.

I'd call them inferences, as they would be based on information garnered in
the study. The notion of a guess sounds more like uninformed speculation.

You can use words to put a spin on what you mean, but the fact remains that
you have no basis for assuming any particular perceptual model, so you're
truly just guessing: i.e., speculating in an uninformed way.

I know, Bill, that we can't observe anyone else's perceptions. Whether you
are doing psychophysics or tracking studies you are faced with the same
limitation, and in both cases you can only infer as best you can the nature
of the "other's" perceptions, based on the information available to you. If
you insist on observing the "other's" perceptions before accepting any
findings, both psychophysics and PCT are out the window for you.

No, they're not. I insist on ackknowledging that we are guessing when it
comes to "inferring" the forms of perceptual functions in any objective
way. That goes for PCT and every other approach to human perception. I've
said this repeatedly and everyone -- even Rick -- seems to have gone deaf.

···

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

Yes, I would say that in the experiment you describe, you don't know what
the CV actually is. I would guess that it is some function of tone pitch
and tone intensity. You could easily show that tone intensity is not the CV
by varying the pitch -- the threshold intensity of the tone would be
_changed_ by the participant instead of being maintained constant.

Your inference fails here, because the person is still maintaining the tone
intensity at threshold (the reference value the person has been asked to
use), but the threshold intensity (reference value) is changing with tone
frequency.

But it's a different intensity threshold, in terms of the physical
stimulus. What the participant senses as an loudness just above zero
corresponds to different objective sound energy levels at different
frequencies. This is a well-known phenomenon, and is the reason for
plotting sound loudness rather than sound intensity as the subjective
variable. Perceived loudness is a function of both sound intensity and
frequency. The actual perceptual variable experienced as the loudness of a
sound corresponds to neither intensity alone nor frequency alone.

This is exactly the sort of situation the Test was designed for. If you ask
a person to hold loudness at a specific level (any level), you can see how
pitch and frequency must vary together to maintain the same perceived
loudness. You could manipulated pitch as a disturbance and let the person
control by varying intensity, or use intensity as a disturbance and let the
person control by varying pitch. Or you could manipulate both pitch and
intensity, varying the mix until you minimize the participant's control
actions (which could affect both pitch and intensity equally). When you can
present a large number of different combinations of pitch and intensity
that lead to NO action by the participant, you have found a curve of equal
loudness in pitch-intensity space.

I suspect that something like this must have been done to get those
loudness curves.

I described a couple of setups, but didn't provide enough detail about one
of them. In that hypothetical study, the center intensity of the tone would
be set well below threshold and the participant's ability to vary the tone's
intensity would be limited to a relatively small range that did not include
values above threshold. In another phase, the same range of variation would
be under the person's control but the center intensity would be set above
threshold. I predict that the person will be unable to control the tone's
intensity in the first case and able to do so in the second.

I agree. But this still says nothing about what the actual controlled
variable is. You have the same problem we have with PCT and the Test. You
can show that you're measuring something related to the actual perceptual
variable, but you still don't know what the actual variable is.

As both you and Tim Carey note, it is possible for the participant to feign
not being able to hear (or control) the tone.

Feigning is not the point. It's possible that the auditory nerves begin
generating signals at very low levels of sound intensity, but that higher
systems require larger signals before they can recognize that there is a
sound being perceived -- before the intensity signals can be classified as
"loudness," perhaps. This would fit with what Martin Taylor is saying about
there being no true threshold of audition. There may be multiple
thresholds, depending on what kind of information you're trying to detect.
I suspect that if you were asking the participant to distinguish between a
sound and a touch, both at the low end of the scale, the threshold would be
higher than for either one alone. If a person is waiting for a very tiny
touch, and you sound a faint tone, the person will probably push the button
or say "yes." At the intensity level, all perceptions are alike. If you
want to distinguish G from G#, I expect the intensity threshold will be
higher than it is just for distinguishing sound from no sound -- it
probably takes a bigger signal to create a sensation perception (and that
IS a guess).

I made this distinction earlier, but apparently need to repeat it. I am
talking about the loss of correlation between disturbance and output when
the tone intensity falls below threshold.

Fine. That's a purely empirical fact. But it tells us nothing about the
perceptual input function. You have to assume a model before you can say
anything about the perceptual signal. For example, a zero input signal
might produce 1000 impulses per second of perceptual signal, with the
frequency declining as the input intensity rises. The person calls 1000
impulses per second "no sound", and 0 impulses per second "the loudest
possible sound."
You can set up a control model to keep the input intensity at zero by
keeping the perceptual signal at 1000 impulses per second; no problem.

You say you don't believe that model? Then you must have one you do
believe. And that's the model you use when you think you're "inferring" the
perceptual signal. The model above gives a completely different picture of
the perceptual signal, yet it would work for any circumstances you want to
try, including your elaborate two-stage experiment for determining the
threshold.

Best,

Bill P.

[Tim_Carey1]

[From Tim Carey (980213.0700)]

[From Bruce Abbott (980212.1005 EST)]

As both you and Tim Carey note, it is possible for the participant to

feign

not being able to hear (or control) the tone. I am not interested in
assessing the psychophysical functions of uncooperative participants, any
more than you are interested in assessing the tracking abilities of such
individuals.

Bruce, how do you know whether you've got an uncooperative participant or
not, unless you do the Test? It's pretty easy to tell during a tracking
task whether or not the person is keeping the cursor aligned with the
target. How can you tell in the psychophysical studies _that are currently
being done_? Isn't the first assumption of the study that the subject _is_
cooperating?

a rule, uncooperative? Have any? Furthermore, internal checks can be

made

to detect "cheating," for example the insertion of so-called "catch"

trials

in which no signal is presented, and examination of the psychophysical
function, which would be extremely difficult for "cheaters" to duplicate.
Then there is the ability to compare functions across individuals; in
persons with normal hearing, vision, and so on these will be very

similar.

Again, Bruce, if we're comparing conventional research with PCT research I
don't think it's fruitful to talk about what _could_ be done. How often are
internal checks made to detect cheating in psychophysical studies? Are
"catch" trials a standard feature of the research you describe? And how
often are functions compared across individuals?

By the way, I don't get this last check you mention .... are you assuming
here (with the comparison across functions) that someone with normal
hearing will also have normal vision?

Cheers,

Tim

[Martin_Taylor4]

[Martin Taylor 980213 0:30]

Tim Carey (980213.0700) to Bruce Abbott (980212.1005 EST)]

Are
"catch" trials a standard feature of the research you describe?

It is impossible to get a measure of discrimination without them. So yes.

And how
often are functions compared across individuals?

Psychophysics is one area in which the individual is studied rather than
the population. What is usually of interest is a whole pattern of results
when different parameters are varied. And what leads an experimenter to
believe that the subjects are consistently trying, and to be able to see
when they aren't, is a whole pattern of consistency when parameters are
varied, such as phase angles, harmonic content, onset and offset rates...
(quite apart from the fact that one can discover from the correlations
in the pattern of correctness of successive responses what is the minimum
sensitivity the subject must have had to achieve that pattern, even when
the overt percentage correct is much worse).

When one has found a pattern of results for one subject across parameter
variations, _then_ the results can be, and are usually, compared across
subjects. Frequently the study only has a few, perhaps as few as four,
subjects. It's a lot of work training a subject and a lot of work getting
the data. In one study I did in the late 60's, a single subject provided
over 1.5 million responses. You don't want to do that too often:-)

Martin

[Bruce_Abbott]

[From Bruce Abbott (980213.1145 EST)]

Bill Powers (980212.0914 MST) --

I've been puzzling over your reply, Bill, trying to understand it. Mind
you, I have no trouble following your logic and the points you have been
advancing. (In fact I agree with most of them.) I just don't see what they
have to do with the argument I have advanced, and which I thought was the
topic of discussion. The latter seems to have gotten lost somehow.

Quite some time ago now, I provided a thought experiment which shows that
one can learn _something_ about a participant's perceptual input function
based on standard psychophysical tests. Rick Marken was at the time
asserting that "little or nothing" about that function could be determined
from such tests. To defend his assertion, Rick kept pointing to the
"behavioral illusion." Apparently, Rick sees the stimulus being presented
on a given trial (e.g., a tone of a given intensity) as a disturbance to
some controlled variable, and the response ("present" or "absent") as the
behavioral output. The function relating disturbance to output, when a
variable is under control, yields no information about the organism, so Rick
concludes that psychophysical tests reveal nothing about the organism's
input function with respect to the sensory quantity being examined. While
granting the significance of the behavioral illusion, I have argued that
this reasoning is mistaken. As stimulus intensity is lowered, there comes a
point beyond which the participant has increasing difficulty perceiving that
the stimulus is present. The change in the participant's output from saying
"present" on every trial to saying "absent" on most trials is not the result
of disturbing the participant's perception of stimulus intensity, so the
"behavioral illusion" does not apply. Instead, it derives from the failure
of the stimulus to produce a discriminable change in perception. The change
in the participant's output reflects a characteristic of the participant's
sensory mechanism, not the inverse of an environment function relating
verbal output to a disturbance of stimulus intensity (stimulus intensity is
not the CV).

The point you are making is quite different and only tangentially related to
mine. For some reason, you want to focus on the nature of the receptor, and
your point is that we cannot tell what the actual effective stimulus is
without conducting the appropriate tests. This is rather elementary and I
rather doubt that anyone here would disagree. This issue is not the one
being debated. As Martin notes, whether certain receptors on the tongue are
activated by NaCl, electron flow, or something else has little or no
relevance for the question of how low a concentration of NaCl a person can
detect under given conditions.

When you concluded that psychophysics has little to offer PCT because CVs
are normally well within the range of detection, Martin objected that
knowing how small a change in CV can be discriminated would certainly be
important when evaluating the potential to control the CV within particular
limits. What you seem to be missing is that psychophysics is concerned with
how much change in stimulus is required before one can detect the change
(difference thresholds) as well as how much stimulus is required before one
can detect its presence (absolute thresholds).

All in all, I continue to be amazed at view being defended here that 150
years of research on the relationship between stimulation and perception has
no relevance for a view which holds that behavior is the control of perception.

By the way, I really don't care whether you wish to view my postion as
"defending conventional research." If a particular line of "conventional
research" helps us to understand how the sensory systems work, I think it
would be idiotic to ignore it just because it wasn't specifically designed
as a control study. I have been trying to get someone -- you, Rick -- to
show me how a PCT analysis reveals that such results are invalid, but so
far, all I have heard are irrelevant arguments about other matters.

Regards,

Bruce

[Bill_Powers1]

[From Bill Powers (980213.1400 MST)]

Bruce Abbott (980213.1145 EST)--

To defend his assertion, Rick kept pointing to the
"behavioral illusion." Apparently, Rick sees the stimulus being presented
on a given trial (e.g., a tone of a given intensity) as a disturbance to
some controlled variable, and the response ("present" or "absent") as the
behavioral output. The function relating disturbance to output, when a
variable is under control, yields no information about the organism, so Rick
concludes that psychophysical tests reveal nothing about the organism's
input function with respect to the sensory quantity being examined.

The problem here, which Martin seems to have too, is that you're taking the
physical stimulus that the experimenter decides to apply as the criterion
defining the stimulus. If you apply a salt solution, then the stimulus must
be salt concentration. The point Rick and I are trying to make, each in his
own way, is that the _experimenter's_ definition of the stimulus is
irrelevant; there is no guarantee that what the experimenter defines as the
stimulus is what the subject's perceptual signals represent.

Suppose you define the stimulus as the velocity of air blowing on someone's
skin. You can measure the threshold of detection in the same way you would
measure thresholds of sound intensity or light intensity. But we know that
there are no air-velocity detectors per se; the primary effect that is
perceived is cooling of the skin, with secondary sensations at higher
velocities from the disturbance of hairs. If the experimenter thinks that
he is measuring the threshold of perception of air velocity, he is simply
mistaken.

The most direct way to discover what is actually being perceived is the
Test. That's the point of the argument on my side and Rick's. Maybe there
are other ways to get the same information, but the Test is a simple and
clear-cut way of doing it. Some psychophysical experiments use methods that
amount to the same thing as the Test. We can probably accept those results,
or use the results to apply the Test ourselves. But if the experimenter
simply manipulated some physical variable and had the subject saying
"yes/no", it's almost certain that no effort was made to determine the
variable actually corresponding to the perception, and I think we're
perfectly justified in saying that that kind of experiment simply has to be
redone -- we can't trust its results.

Best,

Bill P.

  While

granting the significance of the behavioral illusion, I have argued that
this reasoning is mistaken. As stimulus intensity is lowered, there comes a
point beyond which the participant has increasing difficulty perceiving that
the stimulus is present. The change in the participant's output from saying
"present" on every trial to saying "absent" on most trials is not the result
of disturbing the participant's perception of stimulus intensity, so the
"behavioral illusion" does not apply. Instead, it derives from the failure
of the stimulus to produce a discriminable change in perception. The change
in the participant's output reflects a characteristic of the participant's
sensory mechanism, not the inverse of an environment function relating
verbal output to a disturbance of stimulus intensity (stimulus intensity is
not the CV).

The point you are making is quite different and only tangentially related to
mine. For some reason, you want to focus on the nature of the receptor, and
your point is that we cannot tell what the actual effective stimulus is
without conducting the appropriate tests. This is rather elementary and I
rather doubt that anyone here would disagree. This issue is not the one
being debated. As Martin notes, whether certain receptors on the tongue are
activated by NaCl, electron flow, or something else has little or no
relevance for the question of how low a concentration of NaCl a person can
detect under given conditions.

When you concluded that psychophysics has little to offer PCT because CVs
are normally well within the range of detection, Martin objected that
knowing how small a change in CV can be discriminated would certainly be
important when evaluating the potential to control the CV within particular
limits. What you seem to be missing is that psychophysics is concerned with
how much change in stimulus is required before one can detect the change
(difference thresholds) as well as how much stimulus is required before one
can detect its presence (absolute thresholds).

All in all, I continue to be amazed at view being defended here that 150
years of research on the relationship between stimulation and perception has
no relevance for a view which holds that behavior is the control of

perception.

···

By the way, I really don't care whether you wish to view my postion as
"defending conventional research." If a particular line of "conventional
research" helps us to understand how the sensory systems work, I think it
would be idiotic to ignore it just because it wasn't specifically designed
as a control study. I have been trying to get someone -- you, Rick -- to
show me how a PCT analysis reveals that such results are invalid, but so
far, all I have heard are irrelevant arguments about other matters.

Regards,

Bruce