NetNews Usenet Archive 1993 #1/NN_1993_1.iso/spool/bit/listserv/csgl/2126

Text File | 1993-01-12 | 9.2 KB | 200 lines

Comments: Gated by [e-mail]
Path: sparky!uunet!noc.near.net!saturn.caps.maine.edu!maine.maine.edu!cunyvm!psuvm!auvm!BEN.DCIEM.DND.CA!MMT
Return-Path: <@VMD.CSO.UIUC.EDU:[e-mail]>
Message-ID: <[e-mail]>
Newsgroups: bit.listserv.csg-l
Date:         Tue, 12 Jan 1993 14:43:39 EST
Sender:       "Control Systems Group Network (CSGnet)" <[e-mail]>
From:         [e-mail]
Subject: Re: Psychophysics , S-R "Part" of PCT models
Lines: 188

[Martin Taylor 930112 14:00]
(Rick Marken 930110.1030)

Well, I've mulled over Rick's posting on psychophysical experiments a bit,
and I'm still puzzled.  So I think it is time to stop puzzling and ask
again.  My question is twofold: (1) How would one do an experiment in
psychophysics differently from a PCT viewpoint, and (2) can one use the
results of classical psychophysics within PCT and if not why not?

I described how a psychophysical experiment is done, and Rick answered:

>I agree with this description of the typical psychophysical experiment.
>You go on to say:
>
>>There's no opportunity
>>for the subject to reduce the error of a perception.
>
>What error of perception? Error is inside the subject (if it exists).
>This statement implies that you think of "error" as an objective
>phenomenon; but it is not (for you or the subject). Error (for the subject)
>will exist if the subject is controlling a perception that cannot be
>made to equal the subject's reference specification for it.

I was considering the error within some hypothetical ECS that might in
normal (non-laboratory) circumstances be controlling a perceptual signal
derived from the signal presented by the experimenter.  I agree with
Rick's definition of error and his description of its properties.  That's
not an issue.

> What
>perception(s) is the subject controlling in a psychophysical experiment?
>Well,that's the question -- it can only be answered by research -- testing
>for the controlled variable.

I'm not sure that IS the question. What the experimenter is after is a
determination of limitations on the subject's ability to perceive, which
is one segment of what under normal circumstances is a control loop, but
under lab conditions is not.  Of course, I grant that the subject IS
controlling many perceptions during the experiment, including:

>The subject is apparently trying to
>hear the word "correct" said ALL THE TIME -- perhaps as an indication that
>the  relationship variable is being controlled to the experimenter's
>satisfaction (another controlled variable -- "being a good subject").

But with respect to influencing what perceptions will occur of the kind
normally (in the real non-laboratory world) controlled through the signals
now provided by the experimenter, the loop is broken.  None of the subject's
responses have any influence on whether the tone will come in the first
interval or the second on the next trial.

With a physical control system, one way of finding how it is constructed
is to see what it does when one or more of its feedback connections is
broken.  When a control system is actively controlling, it is very hard
to know what is inside it, because a lot of variation in its attributes
have very little effect on anything detectable from the outside.  That's
what is meant by control.

If the subject in the real (non-laboratory) world really wanted to know about
some object or happening, and the signals were as difficult as the ones
now provided by the experimenter, the subject would probably control them by
doing whatever was necessary to hear better, such as moving or asking
someone to "Speak up".  In the laboratory the experimenter prevents the
subject from doing anything to hear better, and there is conflict.  The
experimenter uses force that is overwhelming if the subject is to control
the higher-level perception of seeing the experimenter think of him/her as
a "good subject."

>> The only feedback that can do anything
>>is an indication as to whether the action was right or wrong,
>
>Feedback doesn't really DO anything -- it is always there.

Well, if you count zero effect of actions on the controlled perception as
feedback, I have to agree with you.  But it isn't very helpful if the
loop gain is zero.  The word "do" was perhaps poorly chosen.  "Be useful"
might have been better.

>This sentence
>suggests that you buy into the notion of feedback adopted in most
>psychological studies -- as something that is GIVEN to people (here,
>have some feedback, dear, it's good for you). That is not the meaning
>of feedback in control engineering OR PCT. The feedback you are talking
>about is just an independent "stimulus" event -- it is a DISTURBANCE
>(or a potential disturbance; it is a disturbance is if it influences
>the perceptual variable that the subject is controlling).

Yes.  Sloppy wording.  The so-called "feedback" -- "correct" or "wrong"
is not feedback relating to the control of the perception based on the
signal, though in other (non-laboratory) circumstances it could be.

I dispute, though, the claim that it is simply a disturbance, inasmuch as
it is an effect of the subject's actions just as much as is the movement
or non-movement of a rock the subject tries to push.

> If people are organized as control systems,
>then they are always controlling

But they aren't always controlling every percept that they have the ability
to control, or even a tiny fraction of what they could control.  The degrees
of freedom argument assures us of that, even if it were not subjectively
evident.

>-- and S-R relationships are just
>evidence of disturbance compensation.

What is "disturbance compensation" if it is not output based on the
amplification or transformation of error within an ECS that has a constant
reference signal?

>The more reliable the S-R relationship,
>the clearer the evidence of a controlled variable because the subject is
>reliably compensating for disturbances to that variable.

That makes sense, given my immediately previous comment.

>That's why I think
>that the variable controlled in magnitude estimation is the relaitonship
>between S (as perceived) and R (in terms of its imagined representation of
>magnitude -- if it's a number)

(Actually, I wasn't thinking of magnitude estimation experiments, but I
guess the same problems arise, more or less)

But R has to be perceived and compared with something.  Yes, it's imagined,
so imagining a magnitude realted to it is possible.  But what is imagined?
It surely isn't the response itself.  It must be a perception that has some
association with the response, such as a number that can be labelled.
Something about the number is perceived as matching the S perception in some
way.  There could be a control loop that uses the value of the S perception as
a reference value and controls an imagined perception that would go along
with the resonse finally emitted, but this seems rather baroque to be
taken for granted as the way things happen.

>-- it's because the quality of the relation-
>ship between S and R is so high.

Isn't there a chance of a circular argument here?  By assumption, no S-R
organization can give high correlation between S and R, but all control
organizations give high correlations between disturbance and action.
Therefore an observed high correlation between S and R proves that they
aren't S and R after all, but disturbance and action in a control system.
This proves that there is a control system involved always when correlations
are high.

>> and that
>>feedback can only contribute to reorganization, not to the actions or
>>perceptions on the next trial.
>
>Not at all. Since the feedback is a potential disturbance it might be
>related to the responses that the subject is using to control whatever
>perception is being controlled; and indeed it does influence the subject's
>responses, and rather systematically.

Only in a statistical sense.  The probability that the subject will give
a correct response tends to be higher if the experimenter provides information
as to whether each one is correct.  There is nothing the subject can do to
generate a perception that will permit a correct response on the next trial,
as might be possible for similar signals in a real-world situation.

>>Why should taking a PCT approach make any difference to the way psychophysical
>>experiments are done or interpreted?
>
>I hope I have given you some clues.

I'm afraid I have to admit that I can't find the clues.  No Sherlock here.

>Psychophysical experiments are based
>on the assumption that responses are caused by stimuli -- rather than the
>assumption that people control sensory input.

They may have been, but I don't see (still) where a shift in that assumption
would affect either the manner or the interpretation of the experiments.

My answer to my own pair of questions is, at present,
(1) I would not do a psychophysical experiment differently, now that I know
about PCT, though I might choose different problems to address--for example
to look at whether differentiation on the perceptual side or integration on
the output side or both were involved in tight control (to mention a recent
CSG-L question).

(2) I would use the results of classical psychophysical experiments as
guides to the possibilities of control, by determining the limits of
resolution for different perceptions.

I leave it open to you to alter these answers.

Martin