[From Bruce Abbott (950622.1920 EST)]
Rick Marken (950620.1030) --
I think it is significant that reinforcement theorists must develop a new
version of their theory each time a new behavioral situation is described.
For example, in a simple operant conditioning experiment, the occurance of
food following a bar press is considered a reinforcement. I have never heard
reinforcement described as the change from "no food" to "food". So, when I
developed the E. coli demo it seemed reasonable to assume that the direction
of movement following a press was the reinforcement; like the food in the
basic operant conditioning experiment, direction of movement is a direct
result of a response. But, in order to make the reinforcement model of E.
coli behavior work, it was necessary to redefine reinforcement as _change_ in
the direction of movement after a press; change toward the target was the
reinforcement.It is also sometimes necessary to change what was considered a non-reinforcer
or punisher in certain experiments. This is what happened in the first
version of my R. coli demo. In that demo, a press leading to a change from
non-target to target is considered a reinforcer but a change from one non-
target to another could not be considered a punishment because, if it were,
the reinforcement model wouldn't work.
What nonsense!
It's like saying PCT can't be right because you have to write a different
model to account for behavior in the e. coli task and standard pursuit
tracking. To paraphase, "I think it is significant that PCT theorists must
develop a new version of their theory each time a new behavioral situation
is described." The definition of the controlled variable keeps
changing--here it was nutrient level and there it was difference between
target and cursor position. What changes is not the theory but the model
which applies the theory to the situation in question.
I have never heard
reinforcement described as the change from "no food" to "food".
Well, your understanding of reinforcement theory is about 25 years out of
date. For example, see Baum, W. (1973). The correlation-based law of
effect. _JEAB_, _20_, 137-153.
But, in order to make the reinforcement model of E.
coli behavior work, it was necessary to redefine reinforcement as _change_ in
the direction of movement after a press; change toward the target was the
reinforcement.
This sounds like a recitation of history but actually it is simply a product
of your imagination.
How can you have an "event" without change? Reinforcement _always_ involves
a change of condition. Reinforcement was not "redefined," it's the same as
it's always been.
In that demo, a press leading to a change from
non-target to target is considered a reinforcer but a change from one non-
target to another could not be considered a punishment because, if it were,
the reinforcement model wouldn't work.
No, not only must there be change, but to be reinforcing the change must
lead to improvement with respect to the desired condition. Responses that
lead to irrelevant beeps and clicks aren't punished, they simply go
unreinforced. In contrast, having the rate of increase in nutrient
deteriorate following a tumble definitely represents a worsening of
conditions and thus would be expected to punish the tumble.
The events that are considered reinforcement, punishments, etc must be
changed to account for behaviors in different situations because the behavior
of the reinforcement model is determined by the situation. The model will
only work (account for the actual behavior observed) if the modeller can pick
out aspects of the environment that will produce the right behavior. This, of
course, means that the reinforcement model cannot fail as long as one can
find appropriate environmental inputs (appropriate consequences of
responding) for the model.
This just restates your erroneous thesis. But you do bring up a point. ANY
model based on theory involves speculation about what the relevant variables
are in a given situation. In developing a control model, for example, you
must first decide what the controlled variable is. Your guess might be
wrong, and you will then have to try another. Does this mean that control
theory is wrong? Of course not. Same here: In applying reinforcement
theory to a specific situation, I am speculating at to what variables may
involved, and naturally I'm going to pick variables that allow the model to
perform correctly (don't you?). Whether we're talking reinforcement theory
or PCT, the crucial step is to TEST the assumptions of the model, i.e.,
perform tests to determine what the actual reinforcer is, or the actual
controlled variable. Our little modeling efforts have not included that
critical step, so the models remain speculative. The fact that we have not
gone beyond speculation is not a criticism of the theory on which a model is
based, reinforcement OR PCT.
Of course, the power of reinforcement theory exposes its basic flaw. The
power of reinforcement theory is achieved by looking for aspects of the
environment that can account for the behavior we see. But this means that the
same aspect of the environment must be treated as a reinforcer in some
circumstances and a non-reinforcer or punisher in others. Since the theory
provide no independent means of measuring the reinforcingness or
punishingness of the environment, we are free to ascribe these changing
characteritics to the environemnt as necessary to explain behavior. This
is not science; it is tautology.
Based as it is on a misconception of how reinforcers and punishers are
defined, and of what we are doing with these models, this conclusion is
moot. How about "since the theory (PCT) provides no independent means of
measuring the controlled variable, we are free to ascribe these changing
characteristics to the environment as necessary to explain behavior." Your
statement regarding reinforcement and punishment is as true as my
paraphrase, and for the same reason. It sure sounds good.
Regards,
Bruce