[From Rick Marken (950212.1710)]
Bruce Abbott (950212.1740 EST) --
Welcome back. Glad your mom-in-law is OK.
In the few spare moments I've had I've managed to get a PCT version
of the inverted-t illusion working, with some interesting results.
Great. Could you send the Turbo code? I'd like to try it out.
The standard model with integrating output function tends to yield a
good correlation but terrible RMS error because of the effect of the
illusion.
Rats. I listen to Ed Ford and go off and spend some quality time with
my wife and Bill Powers (950212.1640 MST) beats me to the punch;-)
Bill says:
the main reason your model didn't fit the data almost perfectly is that
you need to build the illusion into the model.
Yes, indeed. Remember, modellers: the perceptual function is THE
most important component of the PCT model. The model won't behave
like the person unless it is controlling nearly the same perception as the
person is controlling. Bill mentioned the most obvious possibilities for
controlled perceptual variables in this experiment: p = h-a*v and p =
a*(v/h) where h and v are horizontal and vertical length, respectively
and a is the anisotropy factor.
Try them both, Bruce, and see which gives the best fit to the subject's
behavior. When you do this you will have done, as Bill says,
the Test for the Controlled Variable in a nice refined quantitative
form.
I suggest that this research could be the basis for your first research
publication based on PCT. The results could be quite interesting. If you
test several subjects, for example, you may find that some subjects
control h-a*v while others control a*(v/h). You may also have to select
disturbances judiciously to get the data to be able to resolve the
difference between the various perceptual models (see the last
section of the "Behavior in the first degree" paper in my "Mind
Readings" book, if you have it, to see why).
Keep us posted on the progress of this research; I find it VERY
interesting.
Best
Rick