"The Fallacy of Behaviorism" (Powers, 1989)

[From Chris Cherpas (951107.1734 PT)]

[re: >"The Fallacy of Behaviorism" (Powers, 1989, pp 233-240)]

I just got a bunch of PCT literature to read and am getting very
interested in understanding it and, hopefully, applying it to
computer-based education systems. However, since I've been
heavily schooled in the Skinnerian tradition, I can't help
but make some comment on something I've read. Perhaps this has
been covered already, so please refer me to where it has.


We know that such a computer would be unaffected by the
reinforcements. What, then, is the difference between
the patterns of behavior when the organism and the computer
are "responding?" It is significant that this test has never
(to my knowledge) been done. As far as I know, nobody has
ever compared cumulative records obtained when (a) an organism
is operating the bar or key, and (b) an artificial device
programmed to behave arbitrarily is operating the bar or key.

Not true. In the lab I worked in, we routinely ran programs
to operate the devices used in operant conditioning experiments
to test the equipment. Your statement sounds as if someone from
the EAB camp would really be caught with their pants down on this
one -- that you thinks it's amazing that nobody would ever think of doing
this. Think again. The fact is that it would be uninteresting in
many circumstances. Of course, feedback functions (e.g., what effect
does a given rate of key-pecking have on rate of reinforcement) have
been around a long time and are useful for testing, yes, models.


I venture that many of the "findings" from operant-conditioning
experiments would remain unchanged if the bar or key were being
pressed arbitrarily, even at random. The "matching law," for
example, would continue to be supported as before, and would tend
to fail as before -- or so I claim.

I see a lot of "nobody understands me" coming from CSG, but it's
hard to take this kind of comment seriously. Would you expect to
see scallops on FIs or post-reinforcement pauses on FRs with
random responding? Duh. Random responding on Concurrent VIs
would not produce matching, my friend. It would produce the worse
case of undermatching that anybody had ever seen.


Until a behaviorist troubles to do this experiment, we have no
reason to believe that the results of operant-conditioning
experiments reveal anything of importance beyond the properties
of the appar atus.

Wow. Those behaviorists must really be biased not to have thought
of this crucial, all-important experiment. Frankly, if it's
so important, why don't YOU do the experiment? That would seem
to be the quickest way of dispelling any illusions about the
value of the EAB. So, I challenge anybody on this list to get
matching by responding randomly on concurrent VIs. Show me the
stimulus generalization gradients, the matching-to-sample performances,
the controlled interresponse-time distributions, and a host of other
clear demonstrations of control by responding arbitrarily.

In general, I'm really getting interested in PCT because it does
offer the promise of something more. But don't BS me, OK? If it's
EAB models you want to see, a somewhat recent source is JEAB,
Volume 57, Number 3, May, 1992: "Special Issue on Behavior Dynamics."
I can't say I'm satisfied with what you'll find, otherwise I wouldn't
be so interested in PCT, but at least the models of the organism
are getting more explicit.

Meanwhile, the "establishing operation" (e.g., Michael, 1982, JEAB),
might be worth checking out, if only because it is a concept that
is aimed toward better defining environmental events which dynamically
set reference levels. You heard me.