From Tom Bourbon [950801.1652]
I've lost the mail in which Bruce Abbott reminded me that Timberlake modeled
"behaviors" as though they could be "allocated," or "in competition." (My
mail service is screwed up again -- I'm posting in open loop mode.) He also
reminded me that, in a procedure different from Timberlake's, Staddon tried
to model the behavior of rats on various schedules of reinforcement as
though each rat had a "set point" (reference signal) for a particular rate
of behavior (responding -- pressing the bar). Bruce's comments sent me
scurrying to locate some theses by a few of my students, back when I was a
professor. Those students had tried, with primitive apparatus, to replicate
or extend the work of a few behaviorists who _came close_ to studying
control by experimental animals -- people like Marwine, Collier and associates.)
The theses were run in the late 1980s. In them, the students (i.e., we)
cited and discussed some data and modeling from JER Staddon (1983),
_Adaptive Behavior and Learning_, Cambridge: Cambridge University Press.
Staddon was aware of some of the problems that can arise when one (e.g.,
Timberlake, modeling the data of Marwine and Collier, among others) tried to
model "behaviors" as though they were pitted against one another. Staddon
applied a control theory model to some of his earlier data (JER Staddon,
1979, Operant behavior as adaptation to constraint, _Journal of Experimental
Psychology: General_, _108_, 49-67. Staddon even cited Powers, on Control
Systems Theory, as PCT was known back then.
Back then, Staddon tried to model animals that were working under different
schedules of reinforcement as though each animal had a set point (reference
signal) for "eating rate," which I take as being different, in some way(s),
from "reinforcement rate." Staddon was trying to model the same kind of
"inverted U" function (showing lever presses per session vs dippers per
session -- water was the "reinforcement") that has been the object of
modeling in the recent exchanges on CSG-L. Back in the 80s, my students and
I talked all around the problems we saw with Staddon's work, but we _never_
recognized what Bruce did: the animals really did behave as though they had
a "set point" for responding at the maximum rate, _no matter what_ the
experimenter's imposed schedule of "reinforcement."
Great insight, Bruce. You have cut the foundation out from under a huge
portion of the literature on operant performance.
Bill Powers and Rick Marken have already acknowledged your contribution,
Bruce -- I saw that much before I lost the recent mail and stopped receiving
additional mail. Let me add my congratulations.
I also saw that Bill Powers "unloaded" on Rick Marken, for what Bill
characterized as Rick's negative role in the discussions about using PCT to
model Staddon's work. I missed most of the early discussion, but in the
part I saw, I thought Rick was being his usual rigorous PCT self. If ever
there were a person whose "personality" and intent did not come across well
in the medium of e-mail, it is Rick. I thought Rick was simply making the
case that it would be simply _great_ if someone were to try to directly
study the phenomenon of control by experimental rats and pigeons. I tried
to say the same thing in my come-lately remarks on this thread. I guess that
makes be a bad guy, as well.
I don't think Rick was being a horses rear in this discussion, and I think
Bruce has shown himself perfectly capable of taking care of himself.
Good work, Bruce. Now, _run some real data on control by experimental animals_!