Science Primer

[From Rick Marken (950702.0930)]

Bruce Abbott (950702.1025 EST)--

No, reinforcement theorists can't WIN.

You see, if they don't come up with a way to explain the behavior shown in
your test, then reinforcement theory is false.

But if they do come up with a way to explain the behavior shown in your
test, then, as you say, you will take this as evidence that reinforcement
theory is untestable, which is even worse than being false.

Either way, the reinforcement theorist looses. It's a perfect Catch-22.

If my perception is wrong on this, then please explain what outcome you
would take as _supporting_ reinforcement theory as opposed to damaging it.

I am prepared to assume that ALL the current evidence supports reinforcement
theory. In science, we don't look for support for a theory, we look for tests
that will REJECT it. If the results of the test do NOT reject the theory,
then our confidence in the value of the theory INCREASES. What I am asking
is that you propose a test -- one that has not yet been done -- that COULD
reject reinforcement theory if the results came out in a particular way. If
such a test does NOT reject reinforcement theory, then our confidence in
the theory increases.

Every experiment that tests PCT can reject it. That's why the results of the
little conflict study I mentioned earlier were taken so seriously. They
appeared to reject the basic PCT model; in fact, they didn't, not because
we "shored up" the model to make it fit the results but because, in evaluating
the results we (I) neglected how an existing component of the model --
transport lag -- would affect the results.

Frankly, I don't believe that reinforcement theory can even handle the
existing data that you say it can handle. But there seems to be no
convincing you that this is the case -- and I think it's really unnecessary.
Let's just assume that reinforcement theory has been a great success and
press on from there.

All I'm asking for is a little skepticism on your part; the same skepticism
we bring to PCT. When I do a PCT experiment, I typically see, first, what
the PCT model does in the experiment. When I do the experiment, I am saying
"OK, nature, if my PCT model is right, then a living system should behave
in the same way as the model". If the living system does NOT behave like the
model, then we have to take a careful look at the model and (if the
deviation in substantial) be willing to reject it. So far, living systems
have behaved exactly like the PCT model in every test. But EVERY subsequenct
test is still a possible rejection of the PCT model.

I would like to see the same thing done with reinforcement theory. Derive
a prediction from the theory -- based on computer modelling or whatever.
Once you know how the reinforcement model behaves in a particular
situation, then test living systems in that situation and see if they behave
in the same way.

What I am asking from you is a prediction from reinforcement theory that
differs from a prediction of PCT. Then we can see which model predicts
the behavior of the real living system.

This is called "doing science". There is no Catch-22, unless of course
you feel some personal investment in the theory that might be rejected.
I feel that my proposal is quite fair; design an experiment where
reinforcement theory and PCT make quite different predictions; then see
which of the predictions is correct.

Science. I know it's a new and frightening experience for a psychologist.
But try it. You might like it.