Analyzing experiment; reorganization

[From Bill Powers (930115.1700)]

Martin Taylor (930115.1430) --

RE: experiment

I'm not sure of the implication arrow. If this is actually an
implication, then the experimenter's condition, as stated to the
subject is (t1 -> b1) AND (t2 -> b2) ...

I intended it to be an implication. "It is not the case that t1
is TRUE and b1 is FALSE." The instructions as you described them
didn't say that the subject was NOT to press the button if he did
NOT hear the tone in the interval. Also, my OR connective says
that EITHER OR BOTH implications, if true, will make the
perceptual signal true; your AND says that they must BOTH be
true. Either of these propositions could have been the controlled
perception, because I presume you never tested the other
possibilities by having a tone in BOTH intervals, or by arranging
for the subject to experience an indication of a button-press
without intending that it be pressed. I dons't recall, for that
matter, any instruction not to press both buttons. According to
my proposal about the controlled perception, if a tone occurred
in both intervals, it would suffice to press either button. Under
your proposal, it would be necessary to press BOTH buttons.
Regardless of how you state the instructions, you can't simply
assume that the subject is controlling for the same logical
condition you read into the instructions -- unless you list all
the possible minterms and specifically state their presence or
absence -- and verify that the subject understands the
specification by actually testing for the controlled variable.

This is a trap, by the way, of digital design. We state the
logical condition we want to be true, without realizing that this
leaves intact many other combinations that might also give a
value of TRUE unless they're specifically ruled out.

The main point I was trying to make was that what you see simply
as a "response" must be treated, in PCT, as a controlled
perception, under the premise that no action ever takes place but
to control some perception. The pressing of a button is just as
much a perception as the tone is. The subject, therefore, is
asked to control a perception of a logical condition made up of
heard tones in one interval or the other, and the felt sensations
of (or some kind of feedback from) pressing one button or the
other. The reference level for the perception as I defined it
would be "TRUE." I suspect that an actual experiment in which all
the perceptions could be disturbed would show that the subject
was actually controlling for some condition other than the one I
proposed. And given the ambiguous instructions, I doubt that all
subjects would be controlling for the same condition. The onlyt
way to find out is to find out.

It would be best to conduct the experiment first with the tones
unmasked and easily audible, disturbing not only which time=slot
the tone appears in, but also adding externally-caused presses of
the buttons to disturb the other two perceptual variables. In
this way you could establish what the controlled condition
actually is. Then, when the tones are masked, you will be able to
tell whether a particular "error" is actually an error relative
to the subject's controlled variable, or only a difference
between what you consider the instructions to mean and what the
subject considers them to mean.

By the most conservative estimate, as I pointed out, the
subjects were completely mistuned at least 5% of the time.

Was there any way to be partly mistuned, with only an indication
of "right" or "wrong?" I think that speaking of mistuning is
reading more into the data than they could deliver.

I don't suppose you happened to put in disturbances of the
button positions, did you?

We did nothing to make the linkage between perception of tone-
in-interval and selection of button difficult.

To test for controlled variables, however, that is just what you
have to do. You must apply disturbances and see whether they are

I hope that my comments have answered your question as to how PCT
might lead to doing this experiment differently, and might call
for different interpretations of the results.


RE: reorganization

Noting the caveats, is this still your viewpoint? As I read the
above, it denies the possibility of "learning to learn," or of
"developing creativity."

It doesn't rule out learning to learn; it just says that what is
learned is not the ability to reorganize, but some systematic
strategy or algorithm. Systematic strategies and algorithms are
carried out by the logical levels of the hierarchy, not by the
reorganizing system. As to developing creativity, I suppose that
one can learn a strategy that includes flipping a coin. I doubt
that reorganization can be learned -- what's to learn?

More subtly, one can perhaps learn how to set up situations in
which reorganization is likely to happen, and one can learn not
to resist or try to escape from the disorganization that results.
But as to learning how to reorganize, I don't see that as
"falling within the scope of the theory." Of course, maybe the
theory's wrong.

Bill P.