Br. 1. med. Psycho/. (1977). 50, 341-348

Printed in Great Britain

34 1

Serial change in group psychotherapy D. A. Winter and C. J. Trippett Bergin (1971) has clarified the problems of demonstrating the effectiveness of psychotherapy by spelling out what is implied if we acknowledge the irreducible uniqueness of our clients: therapeutic methods and goals must be specific to the personality and problem of the client, as must the degree and direction of change expected and the dimensions of outcome by which most sensitively to register such change. Repertory grids have provided an operational definition to the personal construct theorist’s conceptualization of individual uniqueness and Slater has made possible the study of change by the analysis of grids individually, in pairs and in groups using the INGRID, DELTA and SERIES programmes respectively (Slater, 1972; Chetwynd, 1973). Several researchers have illustrated how far these analyses of grids may deepen our understanding of our clients and their relating, particularly if linked with a psychodynamic framework, and how they may enable investigation of empathy and change (Rowe, 1971; Ryle & Breen. 1972; Smail. 1972; Watson, 1972). Ryle & Lunghi (1%9) have shown predicted changes in a grid to be confirmed in the course of therapy but without using Slater’s (l%9) suggestion of control elements and constructs built into the grid; grids have been used nomothetically to distinguish neurotic from normal grid structures in a student population (Ryle & Breen, 1972). Certain features indicative of the probability of therapeutic change have been explored, e.g. relative sizes of the first two principal components (Crisp & Fransella, 1972) and the client’s construing of the implications of their complaint (Rowe, 1971). Watson (1970, 1972) and Fransella & Joyston-Bechal (1971) studied aspects of process and outcome in group therapy chiefly by the use of several group mean measures rather than individual measures of change in the repeated administration of supplied grids. Caplan, Rohde, Shapiro & Watson (1975) explored correspondences between change in standard grids and observed features of group behaviour. Fielding (1975) related self-reported symptom changes with predicted changes in individualized grids. The need remains for further exploration of the ideographic potential of grids in registering change in conjunction with their use as a standard indicator of group processes. The present study constitutes a step towards achieving this. Method Subjects in this study were members of a psychotherapy group which met weekly for 75 min sessions over a period of two years. The group consisted at the outset of two therapists, a consultant psychiatrist of Kleinian orientation and a psychologist, together with seven out-patients who had presented with symptoms not considered to reflect psychotic or organic processes. Two of these patients dropped out of the group within the first six months, to be replaced by two others, one of whom only attended for a few sessions. The final Composition of the group was therefore: six patients, four male and two female, and two therapists, both male. Repertory grids were administered to members of the group about four months after commencement of treatment and at six-month intervals thereafter, the fourth and final assessment being after the last meeting of the group. The elements in the grid were the members attending the group at the time of a particular assessment, and these were given ratings on a 0 to 99 scale on 16 constructs selected by the authors and the therapists as being relevant to group processes. Emergent poles of the constructs used were: ( I ) anxious; (3) like me in character; (4) blames others; (5) like my mother; (6) angry; (7)as I would like to be; (8) depressed; (9) clinging; (10) like my father; (11) able to accept oneself; (12) caring; (13) masculine; (14) feminine; (15) threatening; (16) how I should be; and, as a control construct (Slater, 1%9), (2) intelligent. Implicit poles were not supplied. All members attending the group at the time of an assessment session completed the grid simultaneously, except for both therapists at the third assessment and one of the therapists at the fourth assessment. Each grid was analysed at the MRC unit using Slater’s (1972) INGRID programme and comparisons

342

D.A. Winter and C. J. Trippett

between the first and last grid of each patient and of the senior therapist (omitting non-identical elements) were made using the DELTA programme (Slater, 1968). In addition SERIES analyses (Chetwynd, 1973) were carried out at each assessment of all the grids completed by the patients at that time, and DELTA comparisons were made between the resulting consensus grids and the corresponding therapist grids. At the termination of the group, the senior therapist rated on a seven-point scale (from 1 = much worse to 7 = much improved) the changes in each patient in the areas of social, marital, vocational, and intrapersonal functioning, and manifestation of chief symptoms. Results

For convenience, these can be broadly considered in relation to treatment outcome and process, although this distinction may be thought to be rather artificial. INGRID analyses Outcome: ( a )Individual predictions. On the basis of inspection of the first grid completed, one of the authors wrote a description of the construct system of each patient at commencement of treatment and made predictions as to changes in construing expected to be conducive to less disturbed inter- and intrapersonal functioning and therefore to result from successful therapy. Although an attempt was made to derive these predictions, which are presented below for all the patients who remained in the group, solely from the grid data, they were inevitably biased by the fact that both authors observed some of the group meetings. The following conventions will be used to indicate the result for each prediction: J = change in the predicted direction x = change in the opposite to predicted direction

[J,J

= prediction

partially confirmed

* = 0.05P significance ** = 0.01P significance *** = 0401P significance These results are based on one-tailed tests of the significance of the difference between Z scores from the first and last assessments. Mrs A: It is predicted that: (i) since she sees herself as close to her ideal, a prerequisite for change is that there should be an initial decrease in the correlation between constructs ‘like me ’ and ‘as I would like to be’ and between ‘like me’ and ‘how I should be’, these correlations increasing again after changes in other areas of her construing [ J ] ;(ii) as splitting mechanisms appear to operate in her perception of her parents, there will be an increase in the correlation between constructs ‘like mother’ and ‘like father’ [ J ] ;(iii) there will be a decrease in the high correlation between constructs ‘like father’ and ‘threatening’ [J***];(iv) there will be a decrease in the correlation between ‘as I would like to be’ and ‘depressed’ [ J ] . Mr B: It is predicted that: (i) in view of his low self-esteem, there will be an increase in the correlation between ‘like me’ and ‘as I would like to be’ [ J * * * ] and between ‘like me’ and ‘how I should be’ [,/*I; (ii) there will be an increase in the correlation between constructs ‘like me ’ and ‘masculine’ [ J ] ;(iii) there will be an increase in the correlation between constructs ‘like mother’ and ‘as I would like to be’ [ J * * * ] and between ‘as I would like to be’ and ‘like father’ [ J ] ;(iv) his perception of his parents will become more differentiated with a decrease in the correlation between ‘like mother’ and ‘like father’ [ J ] ;(v) there will be an increase in the correlation between constructs ‘feminine ’ and ‘caring’ [ J ] . Mr C: It is predicted that: (i) in view of his low self-esteem, there will be an increase in the correlation between constructs ‘like me’ and ‘as I would like to be’ [J***] and between ‘like me’ and ‘how I should be’ [,/***I; (ii) he will see himself as more similar to the other group members, with a consequent decrease in the sum of squares accounted for by the self element [ J ] ;(iii) there will be an increase in the correlation between constructs ‘like me’ and ‘caring’

[JI.

Serial change in group psychotherapy

343

Mr D: It is predicted that: (i) there will be an increase in meaningfulness of the self element, reflected in an increase in the sum of squares accounted for by it [ J ] ;(ii) greater introspection will result in an initial decrease, followed by an increase, in the correlation between constructs ‘like me’ and ‘as I would like to be’ [./,*I; (iii) there will be an increase in the correlation between constructs ‘like me’ and ‘masculine’ [XI; (iv) the correlation between constructs ‘like mother’ and ‘like father’ will increase [ J ] . Mrs E: It is predicted that: (i) change will necessitate a decrease in the correlation between constructs ‘like me’ and ‘as I would like to be’ [ J ] ;(ii) there will be an increase in the initially low correlation between constructs ‘like mother’ and ‘as I would like to be’ [ J ] . Mr F: It is predicted that: (i) in view of his low self-esteem, there will be an increase in the correlation between constructs ‘like me’ and ‘as I would like to be’ [ J ] ;(ii) there will be an increase in the correlation between ‘like me’ and ‘like mother’ [ J ] and between ‘like me’ and ‘like father’ [ J ] ;(iii) the correlation between ‘like me’ and ‘anxious’ will decrease [XI; (iv) the correlation between ‘like mother’ and ‘as I would like to be’ will increase [,/*I. In all, 27 individual predictions were made and of these 24 were definitely, and one partially, confirmed, eight at a statistically significant level. As a control prediction, it was expected that the correlation between constructs ‘as I would like to be ’ and ‘intelligent’ would change less than the other grid measures considered, but this proved not to be the case with all patients.

(6) General predictions. Predictions were also made with regard to changes expected to occur in the grids of all the patients over the course of therapy. These were largely based on Ryle & Breen’s (1972) descriptions of features characteristic of the grids of neurotic students, it being expected that there would be a breakdown of these neurotic structures during therapy, and that they would be more evident in the grids of patients than in those of the therapists. It was therefore predicted that in all patients there would be an increase in the correlation between constructs ‘like me’ and ‘as I would like to be’;and between ‘like me’ and ‘like father’; a decrease in the mean angular distance of construct ‘like me’ from ‘like father’ and ‘like me’ from ‘like mother’; an increase in the value of the distance between ‘like me’ and ‘like mother’ minus the distance between ‘like me’ and ‘like father’; a decrease in the extremity of loading of the self element on the first two principal components; a decrease in the variance accounted for by the first two components; and a decrease in the number of elements at a distance of one or more from the self element. Of 54 general predictions, 24 were confirmed and Wilcoxon’s test revealed that for none of the measures concerned was there an overall statistically significant change. (c) Comparison of individual and general predictions. The percentages of individual predictions confirmed and of general predictions confirmed were computed for each patient. Using the Wilcoxon test to compare these two sets of percentages, it was found that significantly more individual than general predictions were confirmed (T= 0, P = 0-025; one-tailed). Process: As in Watson’s (1972) study, the lowest construct means for the patients at the initial assessment were for ‘like my mother’ and ‘like my father’ and it was predicted that the means of these two constructs would increase during therapy. This tended to be the case but there was also an increase in the patients’ average construct mean for each of the other constructs, and similarly a decrease in average variation about the construct mean for each construct. It was also expected that, in line with the Fransella & Joyston-Bechal (1971) study, changes in the distribution of variance amongst the principal components would be unidirectional and simultaneous in all group members but this was not the case.

DELTA analyses The changes observed in the results of the INGRID analyses were, of course, reflected in the DELTA comparisons between the first and last grids of the group members. The self element

344 D. A . Winter and C. J. Trippett tended to be the one that showed the most change. That the grids of the patients changed more than those of the senior therapist can be seen from inspection of Table 1, which gives the general degree of correlation between the first and last grids of each group member. Table 1. General correlation between first and last grid

Correlation

MrsA

MrB

Mr C

Mr D

MrsE

MrF

TI

0.58

0.44

0.33

0.5 1

0.47

0.49

0.74

TI =Therapist.

SERIES analyses and DELTA comparisons between consensus grids Outcome: The consensus grids, presented in the Appendix, provide a picture of change in the construct system of the ‘typical patient’, i.e. the hypothetical average patient in the group, for whom the general predictions concerning relationships between self, ideal self, and parent constructs were confirmed. The primary dimension of construing for this ‘consensus’ individual appeared to be basically evaluative in that throughout the treatment period it involved the proximity of elements and constructs to the ideal self, and contrasted the therapists, who were close to the ideal, with the patients, who were not. The major shift during therapy was in the constructs ‘like me’ and ‘like my mother’ which, from being opposed to the ideal self construct on the first principal component of the consensus grid, came to be relatively independent of this component and to define the second component, along with ‘like my father’. So, a dimension of construing arose which concerned similarity to self and parents and was independent of the main, evaluative dimension of similarity to ideal self. Another major change was in the construct ‘anxious’, which moved closer to the ideal during therapy, suggesting that anxiety became more acceptable to, and more easily tolerated by, the patients. Process: Consistently high construct means throughout treatment were found for ‘intelligent’, ‘caring’, and to a lesser extent, ‘anxious’, ‘able to accept self’, ‘masculine’, and ‘depressed’, so that these constructs were seen by the patients to be most descriptive of the group members. Lowest construct means, and therefore least applicability to the group members, were for ‘like mother’, ‘like father’, ‘threatening’, and ‘clinging’. Variations about the general means due to individual grids can be thought to give an indication of consensus in the meaning of particular constructs, and greatest agreement in meaning throughout treatment was found for ‘intelligent’ and ‘caring’, least agreement for ‘like mother ’, ‘like father ’, ‘threatening’ and ‘clinging’. Agreement in the meaning of ‘depressed’ was initially low but increased considerably, while the reverse was true for ‘angry ’. Variation about the general means due to the elements was greatest for the constructs ‘angry’, ‘able to accept self ’, ‘masculine’, and ‘feminine’ and in the later grids ‘depressed’ (these therefore being the constructs which the patients found most useful in differentiating the group members); and least for the parent constructs, ‘anxious’ and ‘threatening’. Total variation about general construct means due to individual grids and that due to elements changed in a cyclical fashion, increasing between the first and second assessments, remaining stable between the second and third, and decreasing between the third and fourth. DELTA Comparisons between patients’ consensus grids and therapists’ grids Throughout treatment, least agreement between patients and therapists was in the meanings of self and parent constructs, and ‘anxious’, while greatest agreement was in the meanings of ‘caring’, ‘able to accept self ’, ‘feminine’, ‘how I should be’ and ‘as I would like to be’. The similarity of patient and therapist grids increased somewhat during treatment, and so there was convergence in their use of the constructs.

Serial change in group psychotherapy 345 Therapists ’post-treatment ratings in relation to grid results The senior therapist considered that, while all patients had shown overall improvement, least change had occurred in Mrs A and Mrs E, who were possibly at a disadvantage by virtue of the sex imbalance in the group (as also suggested by the negative evaluation of femininity, as compared to that of masculinity, in the consensus grids) and the fact that both therapists were male. These are also the patients for whom the grids revealed highest initial self-esteem and a predicted decrease in self-esteem took place during therapy. [Full tabulated data on the individualized and general predictions and on therapists’ ratings are available from either author on request.] Discussion The primary concern of this study has not been to assess the effectiveness of group psychotherapy but rather to examine the utility of the repertory grid as an instrument for measuring such effectiveness and monitoring the therapeutic process. This latter distinction between treatment outcome and process will be maintained in considering the major findings pertinent to these two broad areas. Outcome: Significantly more individualized than general predictions of breakdown of ‘neurotic construing’were confirmed, from which, if it can be assumed that such changes did indeed occur during therapy, it follows that the most productive use of the repertory grid in outcome research would necessitate the tailoring of criteria of improvement to each patient rather than the employment of general indices of neurosis. It is, of course, possible that global indices might be devised which are more appropriate to the general population than those derived from the work of Ryle & Breen with students, but even so a fundamental problem of research using general predictions would remain: that, as Yalom (1970) suggests, ‘positive outcome is not unidirectional for all patients’. Thus in the present study, for certain patients a decrease and for other patients an increase in self-esteem was predicted as consistent with therapeutic change. The ideographic application of the repertory grid suggested here is very much in the spirit of personal construct theory, but another valuable approach would seem to be to make use of the opportunity provided by the consensus grid to examine the progress of the typical patient in the group. It seems that for the typical patient in the present study the ideal self was a more meaningful, well-defined concept (and, incidentally, one which implied much the same attributes to all group members) than the actual self (cf. Mair, 1%7), and that during therapy there was elaboration of the actual self as a concept independent of the ideal self. Such increasing freedom of the self-concept from the constraints of the ideal self might be thought to reflect the development of a more ‘healthy’, flexible construct system. Also the realignment of the self, as well as the parent constructs, along a dimension orthogonal to that defined by the ideal self suggests a more stable and fundamental reconstruing than mere slot-rattling of the self construct along the latter dimension. Process: It should be remembered that, in view of the level of abstraction of the consensus grid, a considerable amount of information regarding individual differences is inevitably lost in its derivation. Thus, in some patients, such as Mr C, the change in construing appears to have been in the nature of a direct progression, while in others it was largely cyclical, and in the case of Mrs E termination of therapy may have been premature in that it seems to have occurred when she was in the ‘trough’of a cycle. Kelly (1955) felt that therapy involves a series of cycles of construction, and such a process is suggested by a number of the changes in the individual grids as well as by those in total variation about construct means in the consensus grids. In view of the convergence of the patients’ and therapists’ grids and the fact that there was relatively little change from first to last assessment in the grids of the senior therapist compared to those of the patients, it might be assumed that the patients to a certain extent ‘learnt the language’ of the therapists. The SERIES analyses suggest that the patients came to share a

346 D. A. Winter and C. J. Trippett

common meaning of the construct ‘depressed’, while their understanding of ‘angry ’ diverged. It seems that the latter construct may have initially been a rather impermeable one for the patients and that during therapy its permeability, and therefore openness to differences in application, increased, with more behavioural elements being subsumed within its range of convenience: at termination of therapy, anger was perhaps indicated by silence as well as by a raised voice, by tightly folded arms as well as by a clenched fist. As treatment progressed, the patients found all constructs to be more applicable to their fellows, and also used them in a less extreme fashion to discriminate amongst the latter. This seeming recognition by the patients of greater similarities amongst the group members may reflect the supposed curative factor of universality (Yalom, 1970) but is open to other interpretations, as will be seen below. Validity: As in most of the research on psychotherapy, questions of validity must be raised in drawing conclusions from the results of the present study. Although the use of predictions partly helps to reduce the problem of validity, as the only independent measures of change employed apart from the grid were the therapists’ post-treatment ratings, which clearly could not be free of bias, it is not possible to assert that the changes observed in the grids necessarily reflected successful therapy. Indeed, some of the changes may have merely been, at least in part, a result of factors associated with completion of the grid, such as familiarity with the constructs employed. Thus, there was an increase in the patients’ average construct mean from first to last assessment for each of the constructs, and so the increase in their average construct mean for the parent constructs, which it was predicted would occur as a result of such features of the therapeutic process as the consideration of transference phenomena, may have been less a function of the latter than of the serial assessments. Similarly, the decrease in average variation about the construct mean for each of the constructs may have been purely artifactual. In future research along these lines, it would therefore be desirable to include a no-treatment control group of patients also completing grids at six-monthly intervals; and a more sophisticated independent measure of change than the therapist’s rating scale should be used. A clear rationale for making individualized predictions is also desirable as in the present study they inevitably contained a subjective element. The construing of the person who predicts is as relevant and accessible to analysis as that of the therapist or patient. For example, the control prediction that there would be little change in the correlation between ‘as I would like to be’ and ‘intelligent’was made on the assumption that ‘intelligent’was a uniform and universal ideal. In some patients an increase in this correlation occurred which was larger than changes predicted in other construct correlations and is possibly attributable to idealization of the therapists. Thus the use of control constructs built into the grid, as advocated by Slater, was not found to be helpful in this study. While the use of supplied constructs has facilitated comparisons between group members, more personally significant individualized predictions could be made if at least some of the constructs employed were elicited. The present study gives some indication of which of the supplied constructs were least useful in providing additional information and in allowing the patients to discriminate amongst their fellows, and which could therefore be profitably replaced by elicited constructs in future research. Constructs ‘threatening’ and ‘clinging’ appear to have been the least useful and ‘angry ’, ‘depressed’, ‘masculine’, ‘feminine’ and ‘able to accept oneself’ the most useful. If each patient were using his personal elicited constructs with a uniform set of supplied elements, Slater’s recently developed PREFAN programme would allow a comparable analysis to that provided by SERIES in the present study. Summary and conclusions A standard repertory grid was administered on four occasions at six-month intervals to the members of a psychotherapy group. Both individualized and standard general predictions were

Serial change in group psychotherapy 347 made as to changes in the grids judged to indicate therapeutic improvement. Significantly more of the individualized predictions were confirmed. The results therefore support the use of individualized measures in outcome studies and illustrate the potential value of prediction, although they cannot unequivocally be attributed to successful therapy. The typical, i.e. hypothetical average patient in the group appears to have initially been limited in his construing by his ideal self-concept and to have come during therapy to develop an actual self-concept independent of the latter. There were also changes in construing which would seem to reflect the therapeutic process. The repertory grid appears to be a useful instrument for further research on therapeutic process and outcome, particularly if combined with independent measures of change. Acknowledgements We are grateful to Dr W.Brough for his help in providing the patients for the study and to Dr R. Garside for suggestions regarding the analysis. We wish to thank Dr Slater and the MRC Unit, and Mr P. Clarke at Newcastle Regional Health Authority for carrying out the many computer analyses involved.

References BERGIN.A. E. (1971). The evaluation of therapeutic outcomes. In A. E. Bergin & S. L. Garfield (eds), Handbook of Psychotherapy and Behaviour Change. New York: Wiley. CAPLAN, H. L., ROHDE, P. O., SHAPIRO, D. A. & WATSON,J. P. (1975). Some correlates of repertory grid measures used to study a psychotherapeutic group. Br. J . med. Psychol. 48, 2 17-226. CHETWYND, S. J . (1973). Outlinebf the analyses available with GAP, the Grid Analysis Package. Dept. of Psychiatry, St George’s Hospital, London SW17. CRISP,A. H. & FRANSELLA, F. (1972). Conceptual changes during recovery from anorexia nervosa. Br. J. med. Psycho/. 45, 395405. FIELDING, J. M. (1975). A technique for measuring outcome in group psychotherapy. Br. 1. med. Psychol. 48, 189-198. FRANSELLA, F. & JOYSTON-BECHAL, M . P. (1971). An investigation of conceptual process and pattern change in a psychotherapy group over one year. Br. J. Psychiat. 119, 199-206. KELLY,G. A. (1955). The Psychology of Personal Constructs. New York: Norton. MAIR, J. M. M. (1%7). Some problems of repertory grid measurement. 2: The use of whole figure constructs. Br. J. Psycho/. 58, 271-282.

ROWE,D. (1971). Poor prognosis in a case of depression as predicted by the repertory grid. Br. J. Psychiat. 118, 297-300. RYLE, A. & BREEN,D. (1972). Some differences in the personal constructs of neurotic and normal subjects. Br. 1. Psychiat. 120, 483489. RYLE,A. & LUNGHI, M. E. (1%9). The measurement of relevant change after psychotherapy: use of repertory grid testing. Br. J. Psychiat. 115, 1297-1304. SLATER, P. (1968). Summary of the output from DELTA. Dept. of Psychiatry, St George’s Hospital, London SW17. SLATER, P. (1%9). Theory and technique of the repertory grid. Br. J. Psychiat. 115, 1287-12%. SLATER.P. (1972). Notes on INGRID 72. Dept. of Psychiatry, St George’s Hospital, London SW17. SMAIL,D. J. (1972). A grid measure of empathy in a therapeutic group. Br. J. med. Psychol. 45, 165- 169. WATSON,J. P. (1970). A repertory grid method of studying groups. Br. J. Psychiat. 117, 309-318. WATSON,J. P. (1972). Possible measures of change during group psychotherapy. Br. J. rned. Psychol. 45, 71-77. YALOM, I. D. (1970). Theory and Practice of Group Psychotherapy. New York: Basic Books.

Received 27 April 1976; revised version received 5 August 1976 Requests for reprints should be addressed to D. A. Winter, Senior Clinical Psychologist, Chace Wing, Enfield District Hospital, The Ridgeway, Enfield, Middlesex. C. J. Trippett is at St Nicholas Hospital, Gosforth, Newcastle-upon-Tyne. This work was carried out while the authors were at the Department of Clinical and Educational Psychology, University of Newcastle-upon-Tyne.

348 D. A . Winter and C. J. Trippett Appendix The repertory grids: Plots of elements in construct space The consensus grids ~

Grid I masculinet anxious

Cpt I1 (2 1.1 % variance) threateningt

blames others

caring

angry* Mr C X

clinging

how I should be* TI X

Mr B intelligent* able to accept self*

X

depressed* X

X

Cpt I

T2

like me* like mother*

Mr D

(50.9% variance) X

Mrs E

as I’d like to be* X

Mrs A femininet

like father

_ _ _ _ _ ~

~

Cpt I1

Grid IV

[ 16.4% variance)

like mothert anxious

like met like fathert Mr D X

blames others depressed*

masculine* how I should be intelligent*

clinging* Mr B

as I’d like to be*

xT1

angry*

X X

X

Mrs A

7-2 caring*

Cpt I (59.7% variance)

X

Mrs E able to accept self*

X

Mr C

feminine threatening

* High loading on Cpt I. t High loading on Cpt 11.

Serial change in group psychotherapy.

Br. 1. med. Psycho/. (1977). 50, 341-348 Printed in Great Britain 34 1 Serial change in group psychotherapy D. A. Winter and C. J. Trippett Bergin...
576KB Sizes 0 Downloads 0 Views