Accounting for

[From Rick Marken (950317.2100)]

Bruce Abbott (950317.2020 EST) --

Stimulus control is hardly in the same class as "writing your name,"
even if both involve the effect of systematic disturbances to a control
system. It is a major explanatory principle of behavior analysis

Stimulus control is, indeed, important; and it is a major explanatory
principle. But it is exactly the same _kind_ of phenomenon as "writing
your name" in response to a disturbance to a controlled variable. That
is, stimulus control (like name writing in the rubber band demo) is
something that an observer notices about behavior; it is not relevant to
the subject's controlling. If people did not write their name or if their
responses did not correlate with the stimuli they woudn't know or
care; they are just keeping their perceptions (knot position, logical
variable) in reference states; the name and the stimulus control are
irrelevant to the controller, no matter how interesting they might be to
the observer.

I think it's important to show that PCT can explain the appearances
[of stimulus control], because it is a step that is absolutely required if
PCT is to be given any serious consideration by mainstream
behavioral psychology.

It is a step we have taken over and over again; it's what our PCT demos
are all about. Conventional psychologists HATE these demos because they
show that there is something wrong with their concept of the phenomenon
(behavior) they are studying. Conventional psychologists want PCT to
be an alternative theory of behavior; PCT starts, however, by proposing
an alternative concept of behavior itself: behavior IS control.

I will set up the program I described to you (I can't do it this weekend;I'my
going on a little vacation with my honey). The demo will show that "stimulus
control" is just one possible side effect of controlling a logical variable.
When the disturbance is different, there is no more appearance of stimulus
control. There is no need to get into theory to show this; PCT shows that
the appearance of "stimulus control" is just one of several possible side
effects of controlling a logical variable.

Remember, the main contribution of PCT to psychology is factual, not
theoretical. Powers' main contribution to the world is to have noticed
that organisms _control_; they don't respond to stimuli, they don't
generate outputs, they are not driven by genetics. Forget the theory of
control for a while and try to focus on the FACT that what we call
"behavior" is either a controlled (intended) result of aciton or an
accidental side effect of that control (an observable -- but irrelevant --
consequence of the means used to achieve control).

The notion that stimulus control is "just a side-effect of control"
suggests that we already understand perfectly well how to model it,

Not true. What it suggests is that we have a pretty good idea what
variable is controlled (a logical relationship) and that a correlation
between stimuli and responses is just one of many possible patterns
that we might see the organism use (depending on disturbance) to
achieve this control. I will demonstrate these two facts about behavior
when I write the "stimulus control as side-effect" demo (though I hope
you beat me to it). We will see that a logical variable is kept under
pretty good control (I'd measure this as the proportion of total run time
that the variable is "true") and that, in some circumstances, we see
"stimulus control" (the correlation between stimuli and responses we
see in SDTEST ) and in others we don't (when we move the cursor to
the appropriate target on some proportion of the trials) .

We will be able to see these results but we won't necessarily know how
to model them in detail (though I think it really won't be that hard to
model the gross features of the data). But I think you will see that
"stimulus control" can be demonstrated to be a side effect of controlling
a logical variable even though we don't know how to model control of
that variable.

I think this "stimulus control as side effect" demo would really be
worth building. I'm sure I won't be able to start on it until next
weeknend -- and I'll probably do it in HyperCard. So I encourage you to
build a Turbo Pascal version of this demo for the PC folks; I think it
would really help people understand how PCT relates to conventional



<[Bill Leach 950318.19:08 EST(EDT)]

[Rick Marken (950317.2100)]

If I am understanding at all correctly what Bruce is trying to do, I find
myself on 'his side.'

EAB researchers are probably the one group that is capable of
incorporating PCT principles into their research without having to
necessarily openly admit to having been on the wrong track.

They already have the penchant for detailed work and repeatable results.

"Stimulus control" is an observed phenomenon (right, wrong or
indifferent). Explaining such using a methodology that duplicates the
experiemental conditions already employed, appears completely familiar to
existing researchers, provides the same observed behaviour (including
deviations) and then further goes on to explain both the behaviour and
the deviations has, I think a small chance of influencing such

Just the fact that PCT says that their current work is irrelevent is
enough to expect strong control action in opposition. OTOH, if the
demonstrations show at the same time HOW to correct what is making the
work incorrect the 'resistance' should be a bit lower.

It is a step we have taken over and over again; it's what our PCT demos
are all about. Conventional psychologists HATE these demos because they
show that there is something wrong with their concept of the ...

There is a difference here I think. In the first place, EAB researcher
deal with 'controlled environments' and repeatable data. As a general
rule they are 'uncomfortable' with having to 'explain way' any
observations. In the second place it seems that these particular
researcher could do basic PCT research without having to 'start over' by
bulldozing their existing physical facility.

Those of us that already have some understanding of PCT already KNOW that
the rubber band demo fully demonstrates the essential concept -- control.
However, the sort of thing that you guys are working on now should
demonstrate not only the principle but how to apply that principle to a
specific kind of research -- and that I believe is an important
difference. The SDTEST in fact might be totally useless to the 'average
joe' in trying to present the concept of control but to someone that has
spent literally years trying to understand such observed behaviour --
well, hopefully we'll see.