The repertory grid is a powerful instrument not only for the study of the individual in idiographic detail, but also for the comparison of different respondents or of the same respondent on different occasions. A person's construction processes can be followed closely, thus allowing for an assessment of therapeutic, educational, marketing and temporal effects. Changes in the structure and content of their constructions can be thereby monitored (see Feixas, 1991; Winter, 1985b).
The comparison of grids completed by different individuals can best be done with designs that employ the same constructs. This usually means that the investigator must provide the constructs to be used with (the design for the analysis of the family construct system proposed by Procter, 1985, would be an exception to this). Nevertheless, when eliciting the constructs and elements, the scoring system and the size of the grid are kept constant for a group of subjects so that the cognitive measures derived from their grids can be compared. This allows for comparisons to be made, even though the constructs and elements may be different.
In this manual, we will illustrate the comparison of two grids elicited from one person at two different points in time. In the case of Daniel, the therapy process that takes place between the grid administrations is the focus of the evaluation. As a result, the second time the grid is administered, the same elements and constructs appearing in the first grid are reused with the subject simply rating them again. A common problem in such analyses, however, is that the change brought about by therapy is so significant that the subject can no longer work with the "old" constructs, since he or she no longer sees things from that point of view. Also, instructing the subject to score the same constructs again means that new constructs (and perhaps new elements, i.e., new significant others) that are an essential part of the therapeutic process cannot be included. The alternative is to elicit new constructs and elements altogether, yet the opportunity to make detailed comparisons within the consistent set of figures and constructs represented in the old system would be lost. In order to appreciate fully the comparison potential of two grids, we have decided to retain the constructs and elements already elicited in Daniel's first grid. At the same time, we are aware that the opportunity is lost to evaluate any new constructs that may have resulted from therapy.
A first-level comparison of the Daniel-1 (pre-therapy) and Daniel-2 (post-therapy) grids can be obtained by analysing the second grid using the same procedure followed with the first grid and making a qualitative comparison between the two.
An initial analysis is done by comparing the first axis of each grid. In Daniel's case, the first axis from the second administration (Figure 9) has essentially the same structure as the first axis of the previous administration. The extremes, however, appear inverted in the dual diagram due to the mathematical logic in representing these new axes. In effect, the elements IDEAL and PREVIOUS THERAPIST appear on one extreme of the axis, the top one in this second grid, while they were at the bottom in Daniel-1. However, they are at a greater distance from each other, which suggests that the previous therapist is not as idealised as before. On the other extreme of this axis are the elements NON-GRATA and GRANDFATHER associated with the construct "irresponsible," "intolerant" and "nervous." Nevertheless, there is an important change: the SELF element has moved to the centre of the axis, away from the extreme where it was located before therapy. Thus, two major changes emerge from this comparison: a slight modification of the grid's structure and a substantial reconstruction of the self in relation to this axis. This same type of qualitative comparison can be carried out with the rest of the axes.
Comparisons of the first and second grid element clusters are indicative of changes in the way individuals perceive themselves and others. Several differences on Daniel's post-therapy cluster of elements (Figure 10) are noteworthy with respect to his pre-therapy grid (Figure 5). In the post-therapy cluster, SELF is no longer associated with NON-GRATA but with FATHER, FRIEND and previous THERAPIST. This change suggests that Daniel sees himself better situated among other positive male figures following therapy. Similarly, the IDEAL element is no longer associated with previous THERAPIST but is an independent branch not linked to any other element.
The distance matrices and correlations can also be compared in terms of the questions asked in the first grid (see Table 14). For example, there is an observable reduction in the distance between the self and the parents. As far as the elements nearest to the ideal are concerned, the PREVIOUS THERAPIST is still in first place but is closely followed by the SELF. As will be discussed, this indicates a fundamental positive change in the way the self is construed.
As mentioned previously, one of the main advantages of extracting cognitive measures from the grid is that they can be easily compared to other subjects in the same group or to the same subject on different occasions. On the other hand, working with quantitative index values is useful in idiographic as well as nomothetic research.
Generally, the grid measures show that there have been considerable changes in the way the self is construed. This is coherent with the positive social and symptomatic outcome of therapy. Results in table 14 show that Daniel sees himself more similar both to father and mother after therapy --this shift is especially noticeable in the father's case. The previous therapist is seen in a less idealized way at the end of therapy. But perhaps the most remarkable change has been a considerably higher self-esteem. The SELF-IDEAL correlation has changed from being high and negative (r = -0.84) to being moderately positive (r = 0.40), suggesting a greater sense of adjustment. At the same time, the SELF-OTHERS correlation has made a considerable turn from moderately negative (r = -0.26) to moderately positive (r = 0.30). The IDEAL-OTHERS correlation also suffered a slight change from r = -0.18 to r = -0.32, indicating that others are no longer so much a part of Daniel's values; perhaps therapy has made him more demanding of others.
As far as structural features of the construct system are concerned, there are some changes worth noting. Although there are no remarkable changes with Bieri and Intensity scores, there is a slight increase in the degree of differentiation of the system, as can be inferred from the almost seven-point reduction in the PVAFF. This variation suggests an improvement in the complexity of Daniel's system, achieved by increasing the degree of differentiation without losing integration. This potential structural difference is supported by the significant reduction in the use of extreme scores which, in turn, indicates a considerable reduction in the degree of polarised construing.