Doing PCT Research

[From Rick Marken (941011.1100)]

Bruce Abbott (941011.1020) --

How about presenting an example of an important fact from that old
time psychology?

Ye:

The Weber-Fechner law, the serial-position effect in list learning, Pavlovian
conditioning phenomena (the phenomena, not Pavlov's interpretation of them),
Thorndike's law of effect, discrimination learning phenomena, stimulus
control of behavior, extinction, short-term memory phenomena, contrast
effects, attachment (imprinting), overlearning, closure, perceptual
illusions,

These are just words; I would like a quantitative description of the behavior
that constitutes one of these "facts". Describe the fact in detail
(quantitiive description of the variables involved) and then explain its
relevance to PCT, i.e. what you think PCT is supposed to explain!! I'd be
especially interested in a fact that can, as you say,

help to fill in the missing details in the PCT model.

You say:

My personal ambition is to discover how to apply PCT (model and methods) so
as to correctly explain a large number of phenomena within my own field
(learning and behavior).

Your success at achieving this goal depends on what you mean by "explain". If
what you want are qualitative, verbal explanations of these phenomena, then
you should have no problems -- no more problems than Carver, Scheier, et al.
If, however, you want a quantitive match between a working PCT model and
actual behavioral data, then you might have a tougher time, but I would
certainly salute that goal.

However, if I buy into your thesis that these well-known and replicable
phenomena are irrelevant nonsense from a PCT perspective, then my goal is
misguided--there is nothing to explain (or at least nothing worth
explaining). So naturally, perceiving your position as a serious disturbance
to my goal, I have responded to counter it.

Apparently you are familiar with Bill's analysis of the Verhave rat
experiment. I would be thrilled if you could do this kind of an analysis on
other conventional data. I would like to see, for example, how you can get a
control model to produce Weberean discimination data, serial position curves,
Pavolvean conditioning data, etc. I think you might have a problem in many of
these cases, simply the appropriate data is just not available or because the
data is not good enough to model -- too noisy. But let's assume that you
can model it. That would be nice to see. Other than Bill's work and that of
a few others, I haven't seen much quantitative fitting of the behavior of
working PCT models to existing data [besides the Vehave data, Bill also has
an unpublished model that fits all the ratio scheduling operant data
summerized in a paper by Staddon; the "little man" program mimics data
obtained by Bizzi (I believe); the CROWD program can be viewed as an attempt
to account for Clark McPhail's crowd behavior data; I have modelled some
results of coordination studies described by Fowler and Turvey and by Kelso].
So don't let me dissuade you from your efforts to explain existing results
with PCT; let me see how you do it; or if you want suggestions about how to
do it, just post some of the detailed data that you want to explain.
When you do that, I think the problems of the relevance of PCT to
conventional data will become clear.

If you are familiar with Bill's rat experiment analysis, then perhaps you can
get a sense of the real source of my exasperation about the failure of PCT
researchers to turn away from "old time" research and toward PCT research.
Bill found in his analysis, for example, that rats were controlling
something more like the "probability of" than the "interval between" shocks.
He also found that the data was not good enough to allow very precise
estimates of the rat's sensitivity to shock (gain). In other words, this
analysis should have been the start of experimental studies of the variables
rats control. The first thing a curious operant researcher might have done
after reading Bill's paper would have been to redesign this kind of operant
experiment to give the rat continuous control over a presumed controlled
variable. Instead of deliving shock or no shock a random times,
for example, why not have the voltage across the grid varying continuously
(and slowly) and let the rat compensate for these variations by varying the
force on a bar? Then we could test to see whether the rat controls
perceived current and get precise measures of the parameters of control.

My problem is that nobody follows up on the implications of the PCT analyses
of conventional data that have been done; it's just business as usual.
After seeing the results of an analysis like Bill's analysis of an
operant study, people just keep right on doing the kind of research one does
if one is trying to figure out how stimuli affect responses, how
reinforcements select behavior or how cognitions generate output. I guess my
real gripe is with what researchers do after they say "I can account for it
with PCT". What most of them DON'T do is PCT research -- aimed at determining
what is controlled, how it is controlled and why. Maybe you, Bruce, will
now start doing some of the required PCT research. It's just a matter of
doing it.

Best

Rick