Re: introductions

Jack Gerber (
Thu, 13 Oct 1994 19:51:07 -0400 (EDT)

On 13 Oct 1994, A. J. Zolten wrote:

> Welcome Jack,
> I am curious as to how you became interested in personal Construct
> theory. Could you elaborate your experience?
> A.J. Zolten, Ph.D.
> University of Central Arkansas

I am happy to oblige. You might find it interesting. My primary field
of interest is the Rorschach. Some years ago I came up with the
question, "Why does one person look at an inkblot and see a bat while
another looks at the same inkblot and sees a butterfly. Why is this?"
One possible answer might be that the "butterfly" has some connotative,
special meaning to that person.

To investigate this, I looked for a way of measuring the connotative
meaning of a Rorschach percept. I discovered the semantic differential.
It seemed to fit the bill so I did an investigation using the SD to
measure the connotative meaning of subject's own responses (something
which has never been done in a published study). Those of you with long
experience with the repertory grid can no doubt anticipate what happened.

I discovered that the test-retest reliability of the SD was extremely
poor. On further investigation, it turned out that the scales I was
using (which were quite fascinating to ME) were utterly meaningless to
the subjects. They were quite obliging but they were answering what, to
them, were nonsense questions. How could a bat be sharp?

I went back to work and looked for a better way of measuring connotative
meaning. I then came across the repertory grid which seemed to answer my
needs. Since the scales had to be uniform across subjects, I did a
preliminary investigation using the standard rep grid techniques to find
out what dimensions might be relevent to typical Rorschach percepts. I
limited my investigation to only animals and humans which, fortunately,
make up the vast bulk of all percepts.

I came up with a series of relevent dimensions and, using them,
constructed a questionaire that asked subjects to rate the percepts along
a series of unipolar scales. There were actually scales representing
opposite pairs but presented in a unipolar manner to check internal

Unfortunately, once again the test-retest reliability proved poor. I
decided to try a new technique of "anchoring" because I discovered some
of my subjects were losing sight of what they were answering. I began by
asking what, TO THEM, would be a symbol of happiness, fear, danger,
whatever. The I asked them, "In comparison to [the symbol they came up
with] is a bat, dangerous/happy/etc." I saw this as being closer to the
original technique.

In a series of sample tests, using typical percepts, I found an excellent
degree of reliability. I thought I finally had my instrument.

As a check, however, I gave the identical questionaire to subjects
without the anchor. Unfortunately, I got the same high level of
test-retest reliability as with the anchor. There was no difference.

That's where things stand now. I have to find out why I'm getting good
reliabilty with sample data but poor reliability with subjects exploring
their own percepts.

There are a variety of possibilities. Since this is a very informal
account, I would like to add my observation that in gathering the sample
dimensions, I realized that the dimensions subjects choose to rate the
target items seem to be more revealing than the ratings themselves. I
haven't figured out how I might apply this observation, however.

Comments are welcome.

Jack J. Gerber, Ph.D.