[From Rick Marken (930428.1200)
Bill Powers (930428.0700)--
The term "IV-DV" threatens to degenerate on this net into a
stereotype of an approach to human behavior. All that this phrase
means is that one variable is taken to depend on another and the
degree and form of the dependence is investigated experimentally.
This is a perfectly respectable scientific procedure.
I agree.
The basic problem with the IV-DV approach as used in the bulk of
the behavioral sciences is that it is badly used; that bad or
inconclusive measures of IV-DV relationships are not discarded,
but are published.
I think there is another problem with the IV-DV approach when it
is applied to living control systems (as it is when it is used in
the behavioral sciences). I have been assuming that the IV-DV
approach is being used correctly -- in the sense that we are
dealing with good, strong and conclusive measures of IV-DV
relationships. When you get such relationships (as you do in
many operant conditioning studies, for example) it is because
(according to PCT) your IV is the sole disturbance to a controlled
variable and the DV is the only means the subject is using to
compensate for this disturbance. It is nice (and very useful) to
get these reliable IV-DV relationships but without PCT the explan-
ations of these results always miss the controlled variable. So you
see perfect DV=f(IV) relationships and attribute f() to the organism
instead of to the feedback connection from output to controlled
variable.
The best way to determine that the relationship DV=f(IV) exists
because a variable is under control is to do another version
of the IV-DV study where you manipulate an IV and look for LACK
OF EFFECT on the DV. Of course, in this case, the DV is a measure
of the hypothetical controlled variable itself. This is still
an IV-DV experiment but now the results would NOT be viewed
as good, strong, or conclusive if there WERE a very strong relation-
ship between IV and DV; all this would mean (to a PCT researcher) is
that s/he had guessed incorrectly about the controlled variable.
A conventional researcher would find such a result "highly significant"
when, in fact, it reveals little about the subject(s) (other than the
fact that they are NOT controlling the DV, but the conventional
researcher would never reach THAT conclusion).
I agree that manipulating an IV and watching for an effect on a
DV is the way to do research. What I object to about the use of
the IV-DV approach in the behavioral sciences (besides the
problems that you mentioned -- accepting lousy data and taking
average results as a true reflection of individual results) is
that it is based on a causal (input-output) model. Because of
this, researchers cannot see that an IV-DV correlation close
to 0.0 could be their most significant finding of the year; nor
can they see that an IV-DV correlation of .999999 could reveal
absolutely nothing about the nature of the behaving system.
So, instead of criticizing IV-DV research per se, perhaps I should
make it clear that what I am criticizing is the underlying model
of the organism on which this research is based. The Test can
then be seen as IV-DV research applied to closed loop systems.
Bill Powers (930428.0815)--
Your discussion of PCT and cognitive psych was beautiful; a gem.
Thank you.
Best
Rick