[From Rick Marken (960402.1230)]
Bill Powers (960402.0700 MST) --
I still disagree with you that these [conventioal] data are "useless."
I think "potentially misleading" might have been a better way to describe
conventional behavioral data.
I know that Bruce is not collecting data in the conventional way; so I am
confident that his _painstakingly_ collected data will be VERY useful for a
control systems analysis of operant behavior. My comments about the
"uselessness" of conventional data were based on Bruce's April Fool's post
where a conventional analysis of behavior (with no test to determine
controlled variables) was used to reject a control model in favor of a
reinforcement "model". I thought Bruce's post was a nifty (and sly) way
of showing how easy it is to come to the wrong conclusions about behavior
(like the conclusion that behavior is controlled by its consequences) when
we look at behavior through the "wrong pair of glasses" (conventional rather
than PCT).
Progress Report:
I have written a HyperCard replication of the Robertson/Glines learning
experiment and Rich Thurman has written an ever better one (also in
HyperCard). This is good news (to me) because it means that I'll actually
have something to present at the meeting and maybe Rich and I can present it
as a duet.
While working on this project I realized that there is another thing about
the study of learning (besides the fact that it hurts) that I don't like; you
can't be your own subject -- at least, not after you've mastered the task
you want learned. This is annoying because it makes it difficult to "tune up"
the tool you're using to study learning so that you can get clear results.
Right now, I think my implementation of the Robertson/Glines task (originally
conceived by Bill Powers, by the way) works pretty well (on me) but I have a
feeling that I'm going to have all kinds of surprises when I start testing
_real_ people.
Best
Rick