[From Rick Marken (980315.1650)]
Re: The Robertson, Goldstein, Mermel, & Musgrave paper on
testing the self as a control system, Jeff Vancouver
(980313.1500 EST) says "I cannot believe Rick would like such
a design". At least one other person has expressed private
dismay at my approbation for the paper so I thought I would
explain what I liked (and think could be improved) about the
Robertson et al research.
Unquestionably, the best part of the paper is the wonderful
comment made in the discussion of the First design. After
"factoring out" self knowledge, Robertson et al found some
statistical support for their hypothesis; subject's with
higher self knowledge ("High Discrepancy" group) corrected
nearly all false attributions and accepted hardly any (I think
the table is mislabeled). Then I read the following comment:
"A chi-sqaure test comparing first column against the second two
lumpe together was significant at the 5% level, so we might have
submitted for publication. Instead, however, we made the mistake
of attempting to replicate."
I nearly fell on the floor laughing. What a great way to introduce
the fact that the results of the replication were the exact
opposite of what was found in the original study.
I thought the second design was a good approach to getting an idea
of what a person's qi (controlled "self" perception, in this case)
might be. Unfortunately, the results were broken out by "sample"
rather than by subject but I infer (from the sample sizes) that all
subjects (except one in sample 2 and one in sample 3) corrected
disturbances to their own self- descriptions. I would have liked
to have seen more data collected from the individual subjects, but
this was done in design 3.
Design 3 shows that every individual subject (except 7) reacted to
imagined contradictions of statements containing self-descriptive
attributions as expected. The results for statements containing
neutral attributions are less clear; two subjects (4 and 5)
corrected these statements and nearly all subjects made some
modifications. This suggest that the "neutral" attributions were
not really that neutral.
Anyway, what I liked about both designs 2 and 3 was 1) they were
done on a one-individual-at-a-time basis and 2) they were done
with the goal of getting perfect results (and in the case of some
subjects this goal was acheived).
But it's the attitude behind these studies that I like best; and
that attitude is captured by the comment I quoted above about
making the "mistake of attempting to replicate" Robertson, et al
were not going to settle for "statistically significant" results.
They went for results that were true of at least one individual
_all the time_.
What I would like to see in future studies like this is more
attention to determining qi itself. For example, we know from
the data of design 3 that subject 6 produced perfect results,
correcting or modifying every disturbance to a "relevant"
statement and ignoring ("accepting") every disturbance to neutral
statements. At this point the researcher should be able to come up
with some idea (perception) of what "self image" (qi) subject 6 is
trying to control. The researcher can see which adjectives are and
which are not disturbances to this "self image". So the reseacher
should be able to formulate a description (perception) of subject
6's self image (for example, "self-sufficient but humble").
Given this hypothesis about qi, the researcher himself should be
able to pick some new adjectives that would and that would not be
disturbances to it. The next step in the experiment would be to
see whether statements containing these new adjectives are
"corrected", "modified" or "acceped", as experted. If there are
unexpected corrections, modifications or acceptances, the hypothesis
about the controlled "self" perception could be revised appropriately,
so that any new hypothesis is consistent with all previous
"corrections", "modifications" and "acceptances" ("modifications"
would be particularly useful for revising hypotheses about qi
since the subject is basically telling you why the disturbing
adjective is or is not "disturbing").
The goal of this research would be the same as the goal of Robertson,
et al, viz, perfect results, which in this case would mean that
the experimenter could predict _exactly_ how the subject will
respond to _every_ new adjective proposed as a "self description").
It will almost certainly be necessary to "twiddle" with the
methodology (as Robertson et al did -- another thing I like about
their study) in order to get to the point where you are getting
perfect results. For example, you may find that it's better to use
adjectives with modifiers (eg. "fairly helpful" rather than just
"helpful") as the disturbances.
We don't know how to test to determine a controlled "self"
perception right now any more than people knew how to measure
the gravitational constant right after Newton proposed it's
existence. We just have to try stuff to see what works (and we
know it _doesn't_ work to the extent that the results are not
perfect).
Robertson et al tried to measure a controlled "self" perception
in several different ways. They showed how to _approach_ the
study of controlled perception; you approach it with an attitude
to _expecting perfection_ and fiddling with the design until you
get it (or a close approximation).
Replication is the _essense_ of the PCT approach to research; you
repeat what you did, with slight variations, _until_ you start
getting clear results. This is what Robertson et al did. What's
best about the Robertson et al paper is summerized in their own
description of their approach to the research:
"We made the mistake of attempting to replicate".
They replicated their way to something that's a lot closer
to perfection than most of what I've seen in the "conventional"
psychology literature.
Best
Rick
···
--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken/