Two views of behavior Two views of behavior

[From Rick Marken (970317.2230 PST)]

Bruce Abbott (970317.1720 EST) --

The Dutch astronomer Tycho Brahe collected some of the most
accurate data on the apparent motions of the planets then

Actually, Tycho Brahe was a _Danish_ astronomer and his observations
were, indeed, extremely useful to subsequent modelers (like Kepler
and Newton). You must be thinking of the observations made by Tycho's
Dutch cousin, Donkee Brahe, who made extremely precise measurements
of the the extent to which patterns of stars matched the profiles
of famous models of the day. Donkee was collecting these measurements
on the basis of his "model-based" theory of the heavens. Donkee's
observations are, indeed, analogous to those made by conventional
behavioral scientists.

PCT will have a better chance of being accepted in the near future
if it is offered as a theory that promises to organize and explain
the many existing empirical "laws" of behavior as opposed to being
offered as a view that promises to dump all existing psychological
knowledge and ongoing research into the trash.

I don't see how these two options differ. PCT does offer a theory
that promises to explain existing empirical data; what we (you!)
are finding is that the data itself and the knowledge based on it
make it impossible for PCT to keep this promise. Indeed, I can't
think of one single piece of "existing empirical data" that we (you)
have been able to explain using PCT. Even Bill Powers' PCT model of the
shock avoidance data is suspect given what we now know about how rats
press bars.

The fact that so many of the proponents of PCT wish to pursue
the latter strategy is, I fear, not only misguided but a genuine

The tragedy is your inability (unwillingness, really) to see why
existing psychological knowledge and on-going research tells us
nearly nothing about control. Maybe (but I doubt it) I can make the
point with the Behavioral Illusion Java demo. Try controlling the
orientation of line in Experiment 1. The result is a very nice, linear
relationship between IV (height ofrectangle) and DV (mouse position).
Here we have an empirical law; increases in the height
of a rectangle lead to proportional movements of the mouse.

Now, what does a control theorist do with this "fact"? The best the
control theorist can do is say "it's possible that this relationship
exists because the subject is controlling some variable". But what
variable? Nothing about shape or line orientation is included in the
report of the data. All you know is this IV-DV relationship and,
since you know PCT, you _suspect_ that a variable related to the
IV and DV _might_ be under control.

Even if the experimental report included enough information to let
you _guess_ that the subject might have been controlling shape or
line orientation you still would have no way to tell which it was.
And even if you _knew_ that the subject was controlling line orientation
when a particular IV-DV relationship was observed, you would be wrong to
conclude that this particular IV-DV relationship -- this "existing
empirical law" -- indicates that a person is controlling line
orientation. As the results of Experiment 2 in
the demo show, exactly the same linear relationship between IV and DV
can be observed when the subject controls shape.

PCT says that observed relationships between IV and DV in experiments on
living organisms are likely to be side effects of the controlling done
by these organisms. This, it seems to me, is as far as PCT can
go in terms of "organizing and explaining the many existing empirical
laws of behavior". The "existing empirical laws" might suggest what
variables an organisms _might_ be controlling. But that's where the PCT
research must begins.

I still think it is possible, in principle, to explain some "existing
empirical laws" using PCT. It would be possible, for example, to fit the
IV-DV data from the Behavioral Illusion demo with the PCT model. But to
do this, the modeler would have to come up with pretty
accurate guesses about the variables under control (shape and line
orientation). And even if the modeler lucked out and guessed the correct
controlled variables it would be necessary to observe
measures of these variables during the experiment in order to check the
correctness of the model. This kind of data is generally not
available from conventional research studies; but it is essential.
You can see why if you think about how easy it would be to fit the
IV-DV data from the Behavioral Illusion demo using shape as the
controlled variable when the controlled variable is actually line
orientation. The model would fit, but it would be wrong.

I'll tell you what, Bruce. You show me one example of using the
PCT model to quantitatively fit some "existing empirical data"
from conventional psychology and I'll stop kvetching about your
obsession with salvaging the refuse of a moribund discipline.



[From Bruce Gregory (970318.1025 EST)]

Rick Marken (970317.2230 PST)

What a terrific motto: "Salvaging the Refuse of a Moribund
Discipline since 1973...."

Bruce Gregory