[Hans Blom, 941121]
(Bill Powers (941118.0830 MST))
What Thorndyke actually observed was that the cats varied their actions
until the puzzle box opened and they could escape. The opening of the
puzzle box was caused by the cats' behavior. If the behavior had to vary
from one test to another in order to cause the same outcome, the cat
varied its behavior. So it was the cat which was selecting its behavior
to create a desired consequence, not a consequence that was selecting a
desired behavior of the cat.
Some clarity on this issue might be gained by discriminating 1) control,
using a control system with fixed parameters, and 2) learning, changing
the parameters of the control system.
I have no doubt that an SR researcher would agree with the text above, if
the context were ONE experiment, an analogy of which would be one run of a
control system using a fixed set of parameters. But an SR researcher would
disagree if the above statement were applied to a SEQUENCE of experiments.
He would point out that the behavior in subsequent runs is not the same;
that actions that led to success (escape) would occur with increasing pro-
bability and/or with an ever shorter delay, until it appeared that suc-
cessful behavior appeared immediately after placing a cat in its cage if
the offered escape method remained fixed.
Learning is, I think, not the same as control; a better analogy of
learning is the adaptation of the parameters of the control system in such
a way that an "optimal" control system results.
Things get slightly more complex when parameters are adjusted continuous-
ly, and not in a batch-like fashion, but the principle is the same. There-
fore, in my opinion, conditioning ought not be compared with control, but
with reorganization. In the process of reorganization, the consequences DO
select the resulting behavior.
Greetings,
Hans Blom