Individuals who are interested in examining this software (which
runs on an IBM PC with 286 microprocessor and color monitor)
should send an address label and an IBM-formatted HD diskette to
me at the address given at the end of this message. In return, I will
send you the software and a brief version of the manual.
I will be interested in your comments, and in any bugs you may
find. However, I warn you that due to financial constraints the
software is unsupported at present.
John Kihlstrom
*****
PERSPACE
As part of a project supported by the Program on Conscious
and Unconscious Mental Processes of the John D. and Catherine
T. MacArthur Foundation (M.J. Horowitz, Principal Investigator),
we have developed computer software that may prove useful in the
idiographic assessment of the context-specific self-concept. The
program, known as PERSPACE (now in Version 3.5), is inspired by
Kelly's (1955) Role Construct Repertory Test for the assessment of
personal constructs. In Kelly's "Rep Test", the subject (usually a
client in counseling or a patient in psychotherapy) was asked to
name persons who exemplified each of a number of social roles.
Then the subject was asked to generate a construct on which two
of these people differed from a third; the process was repeated
with new combinations of targets, until a substantial number of
constructs had been generated. Finally, the subject was asked to
rate every construct on every target. An informal factor analysis
was then used to determine the relations among constructs and the
complexity of the individual's personal construct system. In this
way, Kelly hoped to enter into the subjective social world of his
subjects, and to understand how they compared, contrasted, and
categorized important people in their lives.
More recently, Rosenberg (1977, 1986, 1988) and Pervin
(1976, 1977) have capitalized on the power of the high-speed
computer to collect idiographic personal construct data on large
sets of persons and situations (for a review, see Kihlstrom &
Cunningham, 1991). PERSPACE, which runs on IBM-compatible
personal computers, is explicitly based on the work of Kelly,
Rosenberg, and Pervin, but it contains a large menu of options that
permits use of the program for a wide variety of purposes.
In the default configuration of PERSPACE, the subject begins
by typing in a list of targets. The free-response probes are
intentionally open-ended:
Please list all the important people in your life;
Please list all the important situations in your life; or
Please list all the important events in your life.
The size of this list is constrained only by the limitations of memory
storage. In actual practice, however, subjects are asked to
generate only about 25 targets.
For subjects (or experimenters!) who need more concrete cues,
PERSPACE also provides a menu of cued-response probes -- for
example, the categories of persons used in Kelly's original Rep
Test. The free-response cue can also be customized. The
program records the order in which targets are generated, and the
interresponse latencies. Redundant targets are automatically
deleted from the list; there is also a facility for hand-editing the
target list.
After the targets have been collected, they are presented
individually in random order, and the subject is asked to list
features associated with each one. (When operated in "person"
mode, the program automatically inserts Higgins' three
"self-guides" the actual self, the ideal self, and the ought self). In
default mode, subjects are asked to describe the person, situation,
or event in question. Again, the size of this list is unlimited, but
practical considerations limit subjects to three to five descriptors
per target. In the present context, three free-response cues are of
particular interest:
Describe yourself when you are with this person;
Describe yourself when you are in this situation; and
Describe yourself when this event occurred.
These provide the basis for assessing context-specific
self-concepts, where "context" is defined in terms of the presence
of particular people, particular situations, or particular events.
Again, precisely redundant descriptors are automatically deleted,
and there is a hand-editing facility. The program preserves
information about the order of target presentation, as well as the
interresponse latencies associated with each descriptor.
In the final phase, the attributes listed for each target merged,
and then collated with the list of targets to form a rather large
target x descriptor matrix. The cells of this matrix are then
randomly presented to subject, who rates the degree to which the
descriptor is characteristic of the target. Thus, for example, if the
subject responded "feel nervous" when asked to describe himself
in the presence of his grandfather, he would be asked to what
extent he feels nervous in the presence of all the other targets on
his list. A variety of rating scales, ranging from 2 to 10 points, are
available for this purpose. A variety of labels, as well as a custom
facility, are available for the endpoints and midpoints of the scale.
Again, the program preserves information about the order of
presentation, as well as the interresponse latencies associated
with each rating.
In the event that the investigator prefers to analyze a particular
set of rating scales instead of allowing subjects to generate their
own, the descriptor phase can be skipped entirely, and the
investigator can supply the subject with any of a wide variety of
conventional schemes:
Interpersonal traits from...
Benjamin's Interpersonal Circle, and
Wiggins' circumplex;
Emotion terms from...
Ekman & Friesen,
Plutchik,
Russell,
Shaver et al., and
Watson & Tellegen; and
Trait adjectives from...
Goldberg's "1710" list,
McCrae & Costa's "Big Five"
Norman's "Big Five",
Peabody's "Big Five", and
Wiggins's "Big Five".
Obviously, the rating phase can be rather protracted. Consider,
for example, a subject who generated five unique descriptors for
each of 25 targets, resulting in a set of 125 descriptors. The
resulting target x descriptor matrix would then have 3,125 cells. At
the rate of one rating per second, it would take a subject almost an
hour to fill in the entire matrix. Accordingly, the rating session can
be interrupted and the computer turned off. When PERSPACE is
restarted, it will begin the session where the subject left off.
After the subject sessions have been completed, the analysis
begins. For both ratings and interresponse latencies, PERSPACE
computes the mean and standard deviations associated with each
target (across descriptors) and each descriptor (across targets).
More important, the target x descriptor matrix is put into a format
compatible with the requirements of major data-analysis packages,
such as SPSS or BMDP (BMDP can manage a square matrix of 86
targets and 86 descriptors, or its rectangular equivalent, up to a
total of 7400 cells). Thus the data can be submitted to a variety of
multivariate statistical analyses, including factor analysis,
multidimensional scaling, and cluster analysis. Our preferred
mode is cluster analysis, which groups targets together according
to similarity of descriptors -- or, in the present instance, similarity of
self-descriptors.
Let's illustrate PERSPACE with a concrete example. Suppose
we had a subject, let's call him John, who is asked to list some of
his favorite people, and who generates the following list: Dick,
George, Charles, Hazel, Linda, Daniel, Sheila, Robyn, Fred, Rick,
and David. This is the first phase of the PERSPACE procedure. In
the second phase, he is asked to describe what he is like (what he
thinks, feels, wants, and does) when he is in the company of each
of these individuals, in response to which he generates the
following list: Dick--manxome, George--uffish, Charles--tolgey,
Hazel--outgrabe, Linda--frumious, Daniel--vorpal, Sheila--frabjous,
Robyn--beamish, Fred--brillig, Rick--slithy, and David--mimsy. In
the final phase, he rates every target in terms of every descriptor,
using a simple two-point (yes/no) scale. When the resulting 10
(targets) x 10 (descriptors) matrix was submitted to a cluster
analysis, we might observe the following groupings of targets:
Dick and Linda,
in whose presence John perceives himself
to be manxome and frumious;
Charles, David, and Robyn,
in whose presence John perceives himself
to be tolgey, mimsy, and beamish;
Dan, Hazel, and Rick,
in whose presence John perceives himself
to be vorpal, outgrabe, and slithy; and
Fred, George, and Sheila,
in whose presence John perceives himself
to be brillig, uffish, and frabjous.
The demonstration is, of course, contrived, but it illustrates the
potential of PERSPACE for revealing the conceptual self in
context. John has not one self-concept but four, quite different
from each other, and each tied to the presence of specific people
in his social environment. Real data, of course, will certainly be
more complex than this. There are problems: where to partition the
solution so as to produce a balance between the number of
clusters and their homogeneity (and thus, ascertaining the basic
level of the self-concept; the reliability of the solution, in terms of
both internal consistency and test-retest stability; and the validity of
the subject's self-perceptions. Still, the program has great
promise, and we look forward to exploring it in both clinical and
experimental contexts.
Reference
Kihlstrom, J.F., & Olsen, D. (1992). User manual for the
PERSPACE software system, Version 3.5. Unpublished
manuscript, University of Arizona. Available from John F.
Kihlstrom, Department of Psychology, Yale University, P.O.
Box 208205, New Haven, Connecticut 06520-8205.
John F. Kihlstrom, Professor
Department of Psychology, Yale University
P.O. Box 208205, New Haven, Connecticut 06520-8205
Telephone (203) 432-2596 Facsimile (203) 432-7172
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%