At the conclusion of Phase 3 of the Subject Session, the rating procedure has generated a matrix summarizing the ratings made by the subject. The matrix for each subject may be submitted to a variety of multivariate statistical analyses, including factor analysis, multidimensional scaling, and cluster analysis. When the primary concern is with the relationships among individual entities (persons, situations, events) rather than their constituent features, the preferred technique is hierarchical cluster analysis, which groups the targets together based on similarity of descriptors (Anderberg, 1973; Baker & Hubert, 1975; Blashfield, 1976; Everitt, 1974, 1979; Hubert, 1974; Johnson, 1967; Kuiper & Fisher, 1975). Cluster analysis begins by considering each target as a separate cluster, and then groups clusters together according to their similarity in terms of the attribute ratings. A cluster is added to an existing cluster only when it is more similar to all members of the cluster than it is to all members of any other available cluster. The resulting solution is hierarchical in that it produces clusters at various levels. At the lowest level, each target forms its own cluster; at the highest level, there is only one cluster -- the entire batch of targets generated by the subject. Most interest focuses on clusters at the middle level, which group relatively many targets together with relatively little loss of homogeneity.
It is important to underscore that the dimensions, factors, and clusters uncovered by these multivariate analysis reflect the way that the subjects perceive themselves and the social world around them. Especially when the cluster analysis is accompanied by a list of the features common to and characteristic of the entities (persons, situations, or selves) in that cluster, it can provide a rich body of information concerning the person's conscious mental representations of self and others. At the simplest level, a content-analysis program can count the number of times a particular attribute appears in the subject's lists. Items with high frequencies of use are good candidates for personal constructs, while the range of such frequencies, and the patterns of co-occurrence among attributes, may be good indications of the person's level of cognitive complexity.
When the final target-by-attribute matrix is submitted to cluster analysis, grouping targets together on the basis of similarity of features, one obtains a graphic display of how the individual organizes his or her social world. A sample is given below.
To demonstrate the PERSPACE program, we present an imaginary protocol. Files created in the process of running this protocol are provided in the subdirectory TEST_A on the distribution disk. The data generated in this exercise can be inspected and printed by using the VIEW DATA and PRINT DATA utilities described above. The entire VIEW.DAT file is provided as Appendix 3.
A specimen BMDP program to analyze the data with Program 2M (Cluster Analysis by Cases), and the output of this program, are found in the BMDPDEMO subdirectory, also on the distribution disk. The BMDP control language, and the entire output, are provided in Appendix 4.
In this application, PERSPACE was programmed to request free listings of both Targets and Descriptors. Descriptive information on the subject, input by the operator, is held in the SUBJECT.INF file on the subject's individual subdirectory.
In the first phase, our example subject was asked to generate a list of some important people in his life (last names were omitted to protect the innocent, but two last initials were necessary to distinguish between two targets with the same first name). This prompt, customized in the Target Prompts Selection Menu, is stored in the TARGET.INP file.
This exercise produced the following list of Target names, and their corresponding response latencies, stored in the TARGETS.SUB and VIEW.DAT files (the latter displayed below). The mean inter-response latency was 1.85 seconds, with a standard deviation of 0.57 seconds.
"Some Important People in Your Life"
2.3s. Betsy
2.2s. Beverly
1.2s. Bob
1.5s. Bill
2.4s. Carol
1.8s. Doug
1.3s. Ernie
2.1s. George
1.5s. Heather
1.8s. Irene
2.0s. Jeanne
2.6s. Jennifer D.
4.1s. Jennifer E.
1.4s. Judy
1.9s. Larry
1.6s. Leanne
1.9s. Lori
1.5s. Margie
2.1s. Martha
1.9s. Michael
1.2s. Pat
1.5s. Paula
2.2s. Rebecca
2.0s. Shelagh
1.3s. Stan
1.9s. Stevens
1.4s. Susan
1.4s. Terry
1.8s. Victor
The Response Latency Mean is: 1.85s.
The Response Latency Standard Deviation is: 0.57s.
In the second phase, our example subject was asked to describe each of the targets. His descriptive vocabulary consisted of only 11 different terms. Precisely identical terms input during this phase were automatically edited out of the DESCRIPT.ED and DESCRIPT.SUB files. These items were stored in the DESCRIP.* files. Had the items in DESCRIPT.SUB required any editing for clarification or to eliminate redundancy, the edited list would have been stored in the DESCRIPT.ED and DESCRIPT.RAT files.
Had we skipped the second phase, and chosen to impose on our subject a set of ratings of our own choosing, these experimenter-supplied descriptors, taken from the lists available in the PERSPACE\DESCRIPT subdirectory, would have been stored in DESCRIP.RAT.
In some cases, the three construals of self -- actual, ideal, and ought -- are automatically added to the list for description.
In any event, the contents of the DESCRIP.SUB file, including the corresponding response latencies, are available for inspection in the VIEW.DAT file (displayed below). The average response latency was 3.22 seconds, with a standard deviation of 1.97 seconds.
"Please Describe This Person"
Betsy
6.9s. manxome
2.0s. uffish
Beverly
4.3s. slithy
*****
Victor
3.5s. uffish
1.8s. tolgey
The Response Latency Mean is: 3.22s.
The Response Latency Standard Deviation is: 1.97s.
After completing his list, our subject realized that he had omitted an important name, "Will". Accordingly, the Operator used the Edit Targets utility to add "Will" to the list. This edited Target list is stored in the TARGETS.RAT file, which is used in the third, ratings, phase.
In the third phase, the system created a matrix consisting of the edited Target (columns) and Descriptor (rows) lists, and filled the corresponding 330 cells (30 columns, representing targets, by 11 rows, representing descriptors) with a random sequence of numbers indicating the order of presentation to the subject. Each combination of Target and Descriptor was then presented to the subject for a rating. In this case, the rating was on a simple 0-1 scale, with the numbers representing "No" or "Yes", respectively. It took our subject approximately 18 minutes to make these 330 ratings, at a rate of about 3.3 seconds per rating.
The following is the descriptor X target matrix of ratings given by the subject:
T A R G E T S
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
1 0 0 0 0 1 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0
2 0 1 1 1 0 0 0 0 1 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0
3 0 0 0 0 1 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0
4 0 0 0 0 0 0 1 0 0 1 1 0 0 0 0 1 0 0 0 0 1 1 0 0 0 1 0 0 0
5 1 0 0 0 0 1 0 0 0 0 0 1 0 0 1 0 1 0 1 1 0 0 0 1 0 0 1 1 1
6 0 1 1 1 0 0 0 0 1 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0
7 0 0 0 0 0 0 1 0 0 1 1 0 0 0 0 1 0 0 0 0 1 1 0 0 0 1 0 0 0
8 0 1 1 1 0 0 0 0 1 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0
9 1 0 0 0 0 1 0 0 0 0 0 1 0 0 1 0 1 0 1 1 0 0 0 1 0 0 1 1 1
10 1 0 0 0 0 1 0 0 0 0 0 1 0 0 1 0 1 0 1 1 0 0 0 1 0 0 1 1 1
11 0 0 0 0 0 0 1 0 0 1 1 0 0 0 0 1 0 0 0 0 1 1 0 0 0 1 0 0 0
In addition, PERSPACE computes the marginal rating and response latencies for each target (over all descriptors) and each descriptor (over all targets), and stores these for inspection in the VIEW.DAT file (displayed below).
Data analysis employed the DATA.RAW file, which is the same as the RATINGS.DAT file, except it is unformatted. Moreover, the matrix has been rotated, with Targets as rows and Descriptors a columns, to make it compatible with the input format expected by BMDP and other conventional statistical packages. The analysis below is limited to the ratings matrix; the latency matrix was deleted.
For convenience in interpreting the output, case labels representing the targets' first names were added to the 24x11 matrix. This expanded matrix was then submitted to Program 2M (Cluster Analysis by Cases) of the BMDP statistical package (BMDP Statistical Software, 1988 Release). This proprietary software is not provided in the PERSPACE distribution disk! However, the matrix is compatible with the multivariate-analysis routines in all known commercial statistical packages.
The defaults (displayed below) for the BMDP 2M cluster analysis -- Euclidean distance measure, data standardized to z-scores, and single linkage algorithm -- should suffice for most assessment purposes.
Note that BMDP can handle a matrix of no more than 7,396 cells (or a square matrix of 86 targets by 86 descriptors; descriptors can be incremented by each target subtracted, so that the product does not exceed 7,396). This matrix is actually larger than the limit of PERSPACE -- which, as noted earlier, is a matrix of 8,575 cells. The maximum matrix size for BMDP corresponds to 7 unique descriptors for each of 32 targets, or 6 unique descriptors for each of 35 targets. Another good reason for keeping the number of targets and/or descriptors generated by subjects under control!
Sample BMDP Input
/problem title = 'PERSPACE Demonstration'.
/input variables = 13.
format = '(11(f1,1x),2A4)'.
/variable name = beamish, brillig, frabjous, frumious, manxome,
mimsy, outgrabe, slithy, tolgey, uffish, vorpal,
name1, name2.
label = name1, name2.
/end
0 0 0 0 1 0 0 0 1 1 0 Betsy
0 1 0 0 0 1 0 1 0 0 0 Beverly
0 1 0 0 0 1 0 1 0 0 0 Bob
0 1 0 0 0 1 0 1 0 0 0 Bill
1 0 1 0 0 0 0 0 0 0 0 Carol
0 0 0 0 1 0 0 0 1 1 0 Doug
0 0 0 1 0 0 1 0 0 0 1 Ernie
1 0 1 0 0 0 0 0 0 0 0 George
0 1 0 0 0 1 0 1 0 0 0 Heather
0 0 0 1 0 0 1 0 0 0 1 Irene
0 0 0 1 0 0 1 0 0 0 1 Jeanne
0 0 0 0 1 0 0 0 1 1 0 JnnfrD
1 0 1 0 0 0 0 0 0 0 0 JnnfrE
0 1 0 0 0 1 0 1 0 0 0 Judy
0 0 0 0 1 0 0 0 1 1 0 Larry
0 0 0 1 0 0 1 0 0 0 1 Leanne
0 0 0 0 1 0 0 0 1 1 0 Lori
0 1 0 0 0 1 0 1 0 0 0 Margie
0 0 0 0 1 0 0 0 1 1 0 Martha
0 0 0 0 1 0 0 0 1 1 0 Michael
0 0 0 1 0 0 1 0 0 0 1 Pat
0 0 0 1 0 0 1 0 0 0 1 Paula
1 0 1 0 0 0 0 0 0 0 0 Rebecca
0 0 0 0 1 0 0 0 1 1 0 Shelagh
0 1 0 0 0 1 0 1 0 0 0 Stan
0 0 0 1 0 0 1 0 0 0 1 Stevens
0 0 0 0 1 0 0 0 1 1 0 Susan
0 0 0 0 1 0 0 0 1 1 0 Terry
0 0 0 0 1 0 0 0 1 1 0 Victor
0 0 0 1 0 0 1 0 0 0 1 Will
/end
The output of BMDP Program 2M (an edited version of which is displayed below) yielded four distinct clusters of Targets.
Our subject's "Favorite People" are grouped into four distinct clusters, each of whose members are perceived to be more similar to each other than they are to the members of other clusters.
 |
Cluster 1 contains eleven individuals (Betsy, Doug, JenniferD, Larry, Lori Martha, Sheila, Susan, Terry, and Victor), all described as manxome, uffish, and tolgey. |
|
Cluster 2 contains eight individuals (Ernie, Irene, Jeanne, Leanne, Pat, Paula, Stevens, and Will), all described as outgrabe, frumious, and vorpal. |
|
Cluster 3 contains seven individuals (Beverly, Bill, Bob, Heather, Judy, Margie, and Stan), all described as brillig, slithy, and mimsy. |
|
Cluster 4 contains four individuals (Carol, George, JenniferE, and Rebecca), all described as frabjous and beamish. |
PAGE 1 BMDP2M
BMDP2M - CLUSTER ANALYSIS OF CASES
Copyright 1977, 1979, 1981, 1982, 1983, 1985, 1987, 1988
BMDP Statistical Software, Inc.
Version: 1988 (IBM PC/DOS) No Math Coprocessor Required.
11/22/92 AT 10:19:25
PROGRAM INSTRUCTIONS
PROBLEM TITLE IS
PERSPACE Demonstration
NUMBER OF VARIABLES TO READ IN. . . . . . . . . 13
NUMBER OF VARIABLES ADDED BY TRANSFORMATIONS. . 0
TOTAL NUMBER OF VARIABLES . . . . . . . . . . . 13
CASE FREQUENCY VARIABLE . . . . . . . . . . . .
CASE WEIGHT VARIABLE. . . . . . . . . . . . . .
CASE LABELING VARIABLES . . . . . . . . . . . .name1 name2
NUMBER OF CASES TO READ IN. . . . . . . . . . . TO END
MISSING VALUES CHECKED BEFORE OR AFTER TRANS. . NEITHER
BLANKS ARE. . . . . . . . . . . . . . . . . . . MISSING
NUMBER OF WORDS OF DYNAMIC STORAGE. . . . . . . 16298
NUMBER OF CASES DESCRIBED BY INPUT FORMAT . . . 1
VARIABLES TO BE USED
1 beamish 2 brillig 3 frabjous 4 frumious 5 manxome
6 mimsy 7 outgrabe 8 slithy 9 tolgey 10 uffish
11 vorpal
INPUT FORMAT IS
(11(f1,1x),2A4)
MAXIMUM LENGTH DATA RECORD IS 30 CHARACTERS.
I N P U T V A R I A B L E S . . . . .
BASED ON INPUT FORMAT SUPPLIED 1 RECORDS READ PER CASE.
NUMBER OF CASES READ. . . . . . . . . . . . . . 30
PRINT DISTANCE MATRIX . . . . . . . . . . . . . NO
TYPE OF TREE PRINTED. . . . . . . . . . . . . . VERTICAL
CALCULATING PROCEDURE . . . . . . . . . . . . . SUM-SQR
STANDARDIZATION OF INPUT DATA . . . . . . . . . YES
AMALGAMATION RULE . . . . . . . . . . . . . . . SINGLE
NUMBER OF NEIGHBORS USED FOR DISTANCE CALC. . . 1
PAGE 2 BMDP2M PERSPACE Demonstration
C N
O 1 1 1 1 2 2 2 2 2 3 2 2 2 1 1 1 2 1 1 2 1
A . 1 6 2 5 7 9 0 4 7 8 9 0 6 2 1 6 1 0 7 5 8 4 9 4 3 2 3 3 8 5
S L B D J L L M M S S T V W S P P L J I E S M J H B B B R J G C
A e o n a o a i h u e i i t a a e e r r t a u e i o e e n e a
E B t u n r r r c e s r c l e u t a a e n a r d a l b v b n o r
E s g f r i t h l a r t l v l n n n i n g y t l e e f r o
L y r y h a a n y o e a n n e e i h r c r g l
D a e g r n e e e e l c E e
l h s r y a
AMALG.
DISTANCE * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
0.000 -+- I I I I I I I I I I I I I I I I I I I I I I I I I I I I
0.000 -+-- I I I I I I I I I I I I I I I I I I I I I I I I I I I
0.000 -+--- I I I I I I I I I I I I I I I I I I I I I I I I I I
0.000 -+---- I I I I I I I I I I I I I I I I I I I I I I I I I
0.000 -+----- I I I I I I I I I I I I I I I I I I I I I I I I
0.000 -+------ I I I I I I I I I I I I I I I I I I I I I I I
0.000 -+------- I I I I I I I I I I I I I I I I I I I I I I
0.000 -+-------- I I I I I I I I I I I I I I I I I I I I I
0.000 -+--------- I I I I I I I I I I I I I I I I I I I I
0.000 -+---------- I I I I I I I I I I I I I I I I I I I
0.000 I I I I I I I I I I I I I I -+- I I I I
0.000 I I I I I I I I I I I I I --+- I I I I
0.000 I I I I I I I I I I I I ---+- I I I I
0.000 I I I I I I I I I I I ----+- I I I I
0.000 I I I I I I I I I I -----+- I I I I
0.000 I I I I I I I I I ------+- I I I I
0.000 I I I I I I I I I I I I -+-
0.000 I I I I I I I I I I I --+-
0.000 I I I I I I I I I I ---+-
0.000 I I I I I I I -+- I I
0.000 I I I I I I --+- I I
0.000 I I I I I ---+- I I
0.000 I I I I ----+- I I
0.000 I I I -----+- I I
0.000 I I ------+- I I
0.000 I -------+- I I
5.227 --------+----------- I I
5.357 -------+------------------- I
5.405 ----+--------------------------
PAGE 3 BMDP2M PERSPACE Demonstration
DISTANCES BETWEEN CASES REPRESENTED IN SHADED FORM.
HEAVY SHADING INDICATES SMALL DISTANCES.
CASE CASE
NO. LABEL
1 Betsy X0*
6 Doug XXOO**
12 JnnfrD XXXOOO***
15 Larry XXXXOOOO****
17 Lori XXXXXOOOOO*****
19 Martha XXXXXXOOOOOO******
20 Michael XXXXXXXOOOOOOO*******
24 Shelagh XXXXXXXXOOOOOOOO********
27 Susan XXXXXXXXXOOOOOOOOO*********
28 Terry XXXXXXXXXXOOOOOOOOOO**********
29 Victor XXXXXXXXXXXOOOOOOOOOOO***********
30 Will XO*
26 Stevens XXOO**
22 Paula XXXOOO***
21 Pat XXXXOOOO****
16 Leanne XXXXXOOOOO*****
11 Jeanne XXXXXXOOOOOO******
10 Irene XXXXXXXOOOOOOO*******
7 Ernie XXXXXXXXOOOOOOOO********
25 Stan XO*
18 Margie XXOO**
14 Judy XXXOOO***
9 Heather XXXXOOOO****
4 Bill XXXXXOOOOO*****
3 Bob XXXXXXOOOOOO******
2 Beverly XXXXXXXOOOOOOO*******
23 Rebecca XO*
13 JnnfrE XX00**
8 George XXOOO***
5 Carol XXXXOOOO****
PERSPACE is intended to be used for assessment purposes. In principle, any psychometric instrument should have four properties: a standardized format and method of administration and scoring; norms from a sample representative of the population in which the instrument is to be applied; reliability, in terms of internal consistency, interjudge agreement, or test-retest stability; and validity, in terms of empirical relations with an external criterion of the attribute ostensibly measured by the instrument. These are not easy standards to meet in an instrument designed for idiographic use, but it is possible to indicate how each of these issues might be treated during further program development.
In some sense, standardization is inherent in the technique, as the whole assessment enterprise is completely under computer control. However, for purposes of nomothetic assessment -- comparing the spaces found in different (types of) patients, for example -- more standardization might be in order. For this reason, PERSPACE includes an option through which the assessor can provide either the targets, or the features, or both to the subjects -- rather than letting them generate both freely. Under these circumstances, subjects can be compared with each other (or aggregated groups of subjects could be compared with other aggregated groups) with respect to the manner in which they organize a standard set of targets. This option would permit complete standardization, although of course it would effectively destroy the technique as a idiographic clinical assessment device.
Norms really are irrelevant to idiographic assessment. In any event the PERSPACE procedure is arduous (and expensive) enough to effectively prevent us from collecting normative data on a large, representative sample of the (presumably nonpatient) population. However, it would be a relatively simple matter to determine the most frequently listed targets and features within each broad category (e.g., persons, situations, or events).
Interjudge agreement is clearly irrelevant, as no judges are involved in this procedure. However, a related issue is presented by the nature of the solutions generated by cluster analysis programs. These solutions are hierarchical: at one level, maximizing homogeneity, there are as many clusters as there are entities. At another level, minimizing the number of clusters, there is only a single cluster including the entire set of entities. By analogy, a factor analysis yields at one level, as many factors as there are items, and at another a single general factor running through the entire item set. A successful cluster analysis yields an intermediate number of clusters, partitioning the solution at some middle level, that groups a relatively large number of targets together with relatively little loss of homogeneity. Unfortunately, there are no algorithms (such as excluding factors with eigenvalues less than 1.0) available for determining precisely where the best partition level lies. This problem of tradeoff remains a judgmental matter, introducing the problem of the reliability with which different observers would assign the same partition level to a given solution. This is, of course, an empirical matter that could be studied using simulated cluster-analysis solutions.
The internal consistency of the subject's responses is also a matter of some concern, because the commonest use for the technique will involve a single assessment. If the subject's responses are unreliable, then any clustering solution derived from them must be meaningless. The standard way of assessing internal consistency is some variant on Cronbach's coefficient alpha. In the present context, probably the best approach is through a variant on split-half reliability. After the entire target-by-feature matrix has been constructed the entire set of targets or features is randomly divided into halves that each represent an unbiased sample of the subject's ratings, and cluster analysis is applied to each half separately. If the structures are reliable, essentially the same hierarchical solutions should be obtained in the halves as were obtained in the whole set.
The standard strategy for assessing test-retest stability is to have the person complete a procedure on two separate occasions. There are several different ways of applying this strategy. (a) Reliability of listing a particular target: If a subject includes "Father" on his first list of "People I Know", will he also do so on his second? (b) Reliability of listing a particular feature: If a subject lists "Loving" somewhere on her first list, will that attribute also appear on the second? (c) Reliability of listing a particular feature for a target: If a subject freely describes her "Father" as "Handsome" on an initial test, will she do so again on the retest? (d) Reliability of assigning a particular rating: If a subject gives "Father" a rating of 1 ("Somewhat Applicable") on "Loving" during the test, will he do so again on the retest? (e) Reliability of solution: Will the hierarchy extracted in the cluster analysis of the initial test resemble that extracted in the retest? These forms of reliability can be assessed with standard contingency and correlation statistics. The problem, again, is that the full procedure is necessarily arduous and expensive. Therefore, for purposes of reliability studies (especially the last two) we might want to work with an abbreviated form of the procedure -- e.g., one in which subjects list only 5 features for each of 20 central targets.
This is, perhaps, the toughest nut to crack. In some sense the procedure attempts to determine how the subject perceives the social world, and there is not really any way to check (or contradict) the data that flows from it. Perhaps alternative cognitive tasks can provide convergent validity of the structures obtained from cluster analysis. In addition, the possibility of a validity check is suggested by Bruner's old dictum that "The purpose of perception is action". That is, the person ought to behave similarly toward targets that are clustered together in subjective space. For example, consider form of the mapping technique intended to identify context-specific selves. The subjects would list the current situations in their lives, describe themselves in each of these situations, and then rate themselves in each situation in terms of each descriptor. Suppose the cluster analysis for a particular subject indicated that Self in Situation A was very similar to Self in Situation B, but very different from Self in Situation C. If we could observe the subject in each of the situations, or obtain personality ratings of the subject by judges who have had the opportunity to observe him or her in one situation but not the others, we would expect similar behaviors or ratings in A and B, different ones in C. Magnusson and Mischel, among others, have done studies of this broad type. They are expensive, but positive results would have considerable theoretical as well as practical importance.
Despite concern with its psychometric properties -- standards that have been developed in the context of nomothetic assessment -- PERSPACE is intended for idiographic assessments, especially in clinical contexts. It is intended to enable clinicians to enter the subjective worlds of their clients, and for clients to articulate what might otherwise be a rather inchoate mass of impressions and reactions, and to reflect on themselves and their personal relationships.
As we envision its clinical use, PERSPACE will be employed early in the therapeutic cycle, as part of routine intake assessment. Thus, at the same time as the client is receiving the standard battery of psychological tests, he or she will also be completing a version of PERSPACE. (Many clients must be placed on a waiting list before they can be seen: PERSPACE, which is designed to be completed by a subject with minimal involvement from the therapist or technician, would seem to be a perfect way to occupy their time). Thus, near the outset of treatment, the therapist will have available a graphic representation of the important people, places, or events in the client's life (as seen by the client) and how they are perceived (again, by the client) to be related. But unlike the results of other psychological tests, we do not intend that the PERSPACE map be held in pectore by the therapist. Rather, we believe that the results of the assessment should be shared with the client, and that clients should be actively encouraged to reflect on their significance.
Since the focus of psychodynamic therapy is on social relationships and personal experiences, rather than symptomatic behaviors, and the goal of therapy is to change these relationships, or at least the client's perspective on them, we also suggest that the PERSPACE procedure be repeated at the point of discharge, as a way of gauging what has been accomplished. Some economies may be injected into the followup assessment by eliminating the first two segments of the procedure -- retaining the original (edited) sets of targets and descriptors, and simply asking the client to provide a new set ratings, resulting in a second target x descriptor matrix for comparison with the first. If anything has changed over the course of treatment, we should expect the second PERSPACE map to differ from the first, and in particular ways dictated by the goal of treatment.
Certain research uses are also suggested by the technique. For example, our laboratory has long been interested in the notion of context-specific selves -- that is, in the idea that one's mental representation of oneself is not monolithic, but rather includes a number of rather different self-concepts, each specific to a particular class of social situations (Kihlstrom & Cantor, 1984; Kihlstrom et al., 1988). In ongoing research, we ask people to generate a list of the important situations in their lives, and then ask them to describe themselves in each of these situations. In principle, the resulting clusters represent context-specific selves. Observations of the subject in these different situations, or ratings of the subject made by the people that he or she encounters in them, should reveal significant differences corresponding to the different self-concepts.
Similarly, subjects might be asked to list the important people in their lives, and then describe themselves in relation to them. Again, the resulting clusters represent context-specific selves, with persons rather than situations serving to define the different contexts. If two people grouped closely together have radically different impressions of the person, or if the person displays quite different patterns of behavior in their presence, this might indicate a clinically significant discrepancy between self-perception and reality. These kinds of self-rating procedures are not so arduous as they sound -- in fact, in our experience of pilot studies, college student subjects find it quite interesting; there is no reason to think that other psychologically minded persons shouldn't as well. With the advent of powerful, high-speed microcomputers, and sophisticated statistical analysis packages to run on them, the assessment technology proposed herein is within reach of even modest laboratories and clinics.
This page last revised 04/08/10 02:58:48 PM.