Rubber band demo

CSGnet Archive CSG1993
1

from Ed Ford (930127:1018)

Gary, Chuck, and all you PCT demonstrators -

I have a different way of demonstrating PCT with a rubber band.
Take two rubber bands (I prefer the big ones) and knot them
together. Ask the participant to hold the ends of the rubber
bands, one in each hand, facing you. With her hands outstretched,
the knot will be directly in front of both of you. Then, point
your finger at the knot and ask her to keep the knot directly in
front of the tip of your finger. Begin moving your finger and she
will automatically look at the relationship of the knot to the tip
of your finger. Next, ask her to look at her left hand and watch
it's actions, and try to achieve the same goal of keeping the knot
at the tip of your finger. Obviously she can't. In fact, there is
a strong internal urge to take a look at the knot-tip of finger
relationship. Thus she will perceive the need for feedback and the
inability to achieve goals by watching behaviors. I've found this
is the best way to lay to rest the fact that we control our
actions.

An alternative is to get two people to participate. Begin with the
two rubber bands knotted together. She holds the end of one rubber
band, and he holds the end of the other rubber band, with the knot
between them. Again, you point your finger with the tip being
right at the knot. Then, move your finger around, and they have to
keep the knot right at the tip of your finer. Now ask them to
achieve the same goal by watching their actions and to concentrate
on how they move their hands as they attempt to reach their goal.
Or, tell them to watch each other's actions. Or, ask one of them
to close their eyes and the task for the other becomes more
difficult. Now move your finger about. Again, the internal desire
on their part to look at the knot shows the need for feedback to
achieve internal goals and the inability to control by
concentrating on the actions. I find when you ask people to switch
from watching the knot and its relationship to the tip of your
finger to watching their hands move, it becomes so obvious how we
control for input, not output. The nice thing about these
demonstrations is that you don't need a chalk board, you can
demonstrate this anywhere (with the exception of a phone booth),
and you still maintain control over the disturbance.

What do you think?

Ed Ford ATEDF@ASUVM.INRE.ASU.EDU
10209 N. 56th St., Scottsdale, Arizona 85253 Ph.602 991-4860

2

[From Rick Marken (930128.0800)]

Ed Ford (930127:1018) --

What do you think? [Of the rubber band demos]

Very nice!

Regards

Rick

3

[From Bruce Abbott (951111.1550 EST)]

Gary Cziko 951111.1722 GMT --

Since I suspect that perhaps
Bruce and certainly Chris are not familiar with this awesome demonstration
of the behavioral illusion and the limitatin of the IV-DV approach to
understanding behavior, I am going to my best to explain it here (again?)
and request that all those who have not done it or seen it before give it a
try.

Been there, did it, showed it to my students. Perhaps you do not recall the
two of us sitting at your kitchen table and doing this together, but I
certainly do! [Hmmm. Perhaps you believe that I didn't understand what I saw?]

Again, what is wrong is not the "IV-DV" (push-observe) approach but the
lineal model used to interpret the results. If "push-observe" can't tell
you anything useful about control loops, then you might as well stop yanking
away at the opposite end of that rubber band. Without a proper model, one
is stuck with identifying a series of empirical relationships which
themselves vary with other variables. For example, you might measure the
thickness of the rubber bands and use that variable to modulate the
functional relationship between E's hand position and S's. Changing the
thickness of the band would be equivalent, in an operant study, to changing
the force-requirement on the pigeon-key or perhaps the ratio requirement.
But a complex empirical relationship may turn out to be the consequence of a
rather simple internal organization, as in the case of the rubben band demo.
It is indeed a great demo, both for illustrating the limitations of a purely
functional account of behavior and for teaching how control systems work.
What it decidedly does NOT do is show that "IV-DV" _methods_ are useless in
the study of behavior; quite the opposite. The assumption of lineal
causation by which the results of these methods are interpreted is the real
problem, not the methods themselves.

Regards,

Bruce

4

[from Gary Cziko 951111.2300 GMT responding to From Bruce Abbott
(951111.1550 EST) about the string enhanced rubberband demos]

Been there, did it, showed it to my students. Perhaps you do not recall the
two of us sitting at your kitchen table and doing this together, but I
certainly do!

We did so much fun stuff that weekend, I can't remember if we did this as well.

[Hmmm. Perhaps you believe that I didn't understand what I saw?]

No, just couldn't recall for sure if was you or Dag Forssell sitting at my
kitchen table.

Again, what is wrong is not the "IV-DV" (push-observe) approach but the
lineal model used to interpret the results. If "push-observe" can't tell
you anything useful about control loops, then you might as well stop yanking
away at the opposite end of that rubber band.

I understand the "IV-DV" approach to be more than "push-observe." It also
tells you WHAT to observe, that is, some variable that changes along with
the "IV." In Bill Power's three-rubberband two-knot demo, it will lead you
at best to the wrong knot, the one NOT being controlled. "IV-DV" does not
say "vary the IV and observe what doesn't change." But that's indeed how
we find out what is being controlled (along, of course, with some
additional requirements, such as the variable is indeed perceived by the
subject, and control is lost if the subject is eliminated).

What it decidedly does NOT do is show that "IV-DV" _methods_ are useless in
the study of behavior; quite the opposite. The assumption of lineal
causation by which the results of these methods are interpreted is the real
problem, not the methods themselves.

We have no argument if you agree that the "IV-DV" method can at best be
only a first step in understanding behavior that is controlling a
perception. One then needs to follow it up with The Test of the controlled
variable to understand what is really going on. Do you agree?

But I do have some difficulty separating methods and the interpretation of
results generated by a method, since how one interprets seems to me to be
an essential part of a method. I have a bookshelf loaded with books of
IV-DV research methods. Included in each of these books is discussion of
how to interpret results, for example, a correlation of .80 means that 64%
of the variance of Y is accounted for by variance in Y; a significant
t-value in a well-controlled experimental study means that the IV is the
cause of the group difference in the DV. None of these books goes on to
describe The Test. They all assume that DVs are caused by IVs, provide
methods to see if the relationship is "significant," and then provide
guidelines for interpreting the strength and importance of the IV-DV
relationship.

A dowsing rod may be a fine way to locate water if it followed up by a
geological study and other reliable methods. But then why bother with the
dowsing rod?--Gary

5

[From Bruce Abbott (951112.1015 EST)]

Gary Cziko 951111.2300 GMT --

Bruce Abbott(951111.1550 EST)

[Hmmm. Perhaps you believe that I didn't understand what I saw?]

No, just couldn't recall for sure if was you or Dag Forssell sitting at my
kitchen table.

I see what you mean . . .

I understand the "IV-DV" approach to be more than "push-observe." It also
tells you WHAT to observe, that is, some variable that changes along with
the "IV." In Bill Power's three-rubberband two-knot demo, it will lead you
at best to the wrong knot, the one NOT being controlled. "IV-DV" does not
say "vary the IV and observe what doesn't change." But that's indeed how
we find out what is being controlled (along, of course, with some
additional requirements, such as the variable is indeed perceived by the
subject, and control is lost if the subject is eliminated).

Yes, Gary, but in ordinary experiments what the experimenter is doing
usually functions as a disturbance of some sort, and what is observed is the
effect of this disturbance on some behavior. If your disturbance produces
actions, you have disturbed a controlled variable, no? So this method IS
telling you something valuable.

We have no argument if you agree that the "IV-DV" method can at best be
only a first step in understanding behavior that is controlling a
perception. One then needs to follow it up with The Test of the controlled
variable to understand what is really going on. Do you agree?

Yes, that has been my position all along, although I believe that much more
information can be extracted from those kinds of observations than you
probably do. But the bottom line is that the Test is necessary, yes.

But I do have some difficulty separating methods and the interpretation of
results generated by a method, since how one interprets seems to me to be
an essential part of a method. I have a bookshelf loaded with books of
IV-DV research methods. Included in each of these books is discussion of
how to interpret results, for example, a correlation of .80 means that 64%
of the variance of Y is accounted for by variance in Y; a significant
t-value in a well-controlled experimental study means that the IV is the
cause of the group difference in the DV. None of these books goes on to
describe The Test. They all assume that DVs are caused by IVs, provide
methods to see if the relationship is "significant," and then provide
guidelines for interpreting the strength and importance of the IV-DV
relationship.

I'd say it's a good bet that 100% of the variance in Y is accounted for by
variance in Y. (;->

I think you have to keep in mind that "cause" does not mean "direct cause."
The proper conclusion is that, under the conditions studied, the changes in
Y result from changes in X; the changes in Y are not due to the action of
some third variable that influences both X and Y, nor is the relationship
actually the reverse, with Y influencing X rather than X influencing Y.
None of those books describe the Test, because the Test is a special
application of the experimental method designed to answer a specific
theoretical question.

Secondly, I am not advocating those statistical methods for studying control
in individuals. I agree that the single-subject approach is required. Even
so, I do believe that some of the findings obtained by these methods are
worth paying attention to, although I don't want to open THAT can of worms
here. But one has to be very cautious to use such information only as the
starting-point for the single-subject investigation.

A dowsing rod may be a fine way to locate water if it followed up by a
geological study and other reliable methods. But then why bother with the
dowsing rod?--Gary

Gary, Gary, Gary. I was with you right up to there, and then you had to go
and ruin it. After agreeing that standard experimental methods are useful
at some stage in the investigation, what do you offer as your stand-in for
these standard methods? Pseudoscience. I would restate your example this way:

A geological study may be a fine way to to locate places where water is
likely to be found, but to prove that there's water there, you have to drill.

Regards,

Bruce