Repertory Grid
Repertory Grid (actually Role Construct Repertory Grid ) German role construct repertoire test is a method of differential psychological diagnosis based on the theory of personal constructs according to George A. Kelly .
overview
Based on the psychology of personal constructs, George A. Kelly developed the Role Construct Repertory Grid (also called Kelly Grid or construct grid in Germany). The term Repertory Grid is often used as a short form. Role Construct Repertory Grid processes work with a repertoire ("repertory") of significant elements from a person's experience, such as roles (e.g. colleagues), groups (e.g. departments), but also situations (e.g. Rituals), objects (e.g. products) and abstracts (e.g. brands). With the help of dichotomous description dimensions, the so-called constructs (e.g. good vs. bad or innovative vs. traditional), these elements are assigned individual properties by the respondent. In addition, there is a quantitative evaluation, mostly Likert- scaled, so that at the end a grid (German: grid, matrix) with numerical values is created. If all elements have been assessed on the basis of a construct pair, new elements are compared with one another in order to form a further construct pair. This process is repeated until the interviewee can no longer think of any new dimensions of distinction, i.e. H. his repertoire of constructs for the given elements is exhausted.
Due to the systematic comparison and the complexity arising in the course of the interview, a targeted, deliberate influence on the result is practically impossible. The RepGrid is used to determine and evaluate subjective associations of meaning. In the sense of Kelly, this should enable an insight into the construct system of the individual. Humans describe their reality with conceptual abstractions (constructs) that have been shaped by their individual experiences. He scales them in a given matrix with regard to suitable elements that represent the scope of the investigation. The respondent thus depicts his or her individual, semantic and psychological space in the form of a grid filled with numbers. The repertory grid is thus a method with the support of which qualitative interviews with the corresponding advantages for the investigation of subjective perceptions and cognitive processes can be conducted, and thus enables a better understanding of the subjective meanings of the respondent: " It is an attempt to stand in others" shoes, to see their world as they see it, to understand their situation, their concerns. "On the other hand, it also offers the possibility of inter-individual comparison of the results:" The use of repertory grid technology today stands within this range - and tension - of idiographic investigation of the particular and the nomothetic naming of the regular. "
Planning and implementation of repertory grids
At the beginning of every repertory grid application there is the desire to learn something, mostly a question about a specific problem. It has already been mentioned that the application possibilities for a grid are basically given in all areas of life in which “construction of reality” takes place.
The selection of the elements requires the greatest care, because these essentially determine the quality and informative value of the results. The question that has to be dealt with is how the object of investigation can be meaningfully converted into elements or which elements are suitable for adequately depicting the object of investigation. The strategy for determining suitable elements for a problem is called "substituting". The substitution should on the one hand help the client to explore his problem and to find the most adequate verbalization for it, on the other hand it should also help the user to reconstruct the client's question or problem as far as is necessary to determine appropriate elements. In any case, the elements should be chosen in such a way that, on the basis of them, the most comprehensive and yet profound clarification of a certain issue can take place, or the realities of the question can be represented in a meaningful way when investigating people and groups. Elements can, for example, be determined from previously conducted, guideline-based interviews with a selection of participants. The participants can also independently work out the elements that apply to them in group discussions.
Grid studies, in which an inter-individual comparison of the results is sought not only in terms of content, interpretative, but also statistically comparative, are only meaningful to evaluate if elements and / or constructs agree for all parties involved. The procedure is standardized when both elements and constructs are specified. In this case the grid resembles a multiple semantic differential or polarity profile, but the specifics of the personal construct approach are lost. The number of underlying elements basically results from the problem from the temporal and demographic conditions.
Since the 1950s, the range of interview types has varied greatly. Since the survey of constructs is not always a usual interview, but other forms are used in grid technology, such as free association or storytelling, evocation (awakening, causing) is also spoken of. For example, while colleagues from England are increasingly promoting the qualitative side of interviewing in therapy, art and organizational consulting, the tetralemma field rating format was developed in Germany.
Triads and dyads
In principle, all differentiations follow two different types of evocation of constructs. In the first, the so-called dyadic variant, two elements are offered and evaluated with regard to their similarity or difference. This form of element presentation is best suited for rather narrow survey contexts, as can be the case, for example, with quick-and-dirty market research campaigns; So if the survey has to go very quickly due to survey circumstances (street survey, trade fair, etc.). The dyadic form is also better chosen if the respondent group can be classified intellectually or in terms of development as rather low (e.g. small children or clinically abnormal people, or people who have impaired perception due to their age). In addition, there may also be specific procedural aspects that justify a dyadic approach. The user may want a certain element (e.g. Me Today) to be compared with all other elements in the set. A triadic approach would be far too time-consuming here, as the many possibilities of compilation would probably exceed the time frame of a survey.
In all of these cases, a more complicated procedure, such as the triadic method below, can overwhelm the imagination.
In the second, the triadic variant, three elements are specified, two of which are to be assessed as similar and at the same time different from the third element. This approach corresponds most closely to the theoretical assumptions of the Constructs Theory of Personal Constructs, which requires the assessment of an object of experience both in its similarity and its difference to other objects of experience. The typical constructive question for this is: “ In what important way are two of them alike but different from the third? "
Opposition or differentiation
The presentation (triad / dyad) is only of peripheral importance for the instruction with which the respondents are presented with elements for differentiation, hereinafter also referred to as the “construct question”. In terms of content, it is more interesting and also more relevant to decide whether the opposite pole of a construct should be formed via the description of an element or as a counter-term to a construct. We therefore distinguish the differentiation method from the opposition method.
The "differentiation method" asks about the difference between the elements presented. If two elements are assessed and described as similar, a description is asked for the third element. Example: If the interviewee decides in the triad Angela Merkel, Günter Grass and Dieter Bohlen for the similarity of Angela Merkel and Günter Grass, the associated construct could be: “Seriousness” and the description for Dieter Bohlen perhaps “reaches many people”. One can criticize this form of questioning with the argument that one possibly obtains a construct whose poles evoke differences, but the constructs obtained do not exclude one another. With the opposition method, which first asks about the similarity of two elements and then about the contrast to the mentioned property (construct), this problem is avoided more reliably. If the opposite of the construct “seriousness” were asked in the example mentioned, the answer could be “ridiculous” and thus be perceived more as an opposite of “seriousness”.
In summary, one can say that the differentiation method is well suited for settings in which the discrete description of the elements is in the foreground. It is very useful in market research contexts to characterize brands or products. The differentiation method can also lead to demanding results in individual therapy or in coaching settings , with regard to assessing one's own person in contrast to the partner or the ideal self. The opposition method is particularly suitable when it comes to depicting the respondent's world of experience without referring to specific elements.
Rating and scaling
From an investigation perspective, the creation of a grid is an assessment task. Assessment objects (elements) should be assessed with regard to several assessment dimensions (constructs). Various methods are in use, e.g. B. nominal scaling, ranking method or multi-level rating scaling. The simplest form of comparison is based on the nominal scaling, which operates with the values 0 and 1 and is usually formulated with the assessment of the construct “applies” or “does not apply”. With this simple form of scaling, however, “skewed distributions” can arise that impair further evaluations. It is also conceivable that the elements are ranked for each construct (with 15 elements = gradation 1–15). The Stephenson Q-sorting procedure is also quite common. Here the elements on the constructs are grouped in gradations in such a way that an approximately normal distribution is created. Most often, however, all elements for each construct are classified independently of one another on a multi-level scale ( rating grid ). The opposing dimension “open” versus “closed” could therefore - similar to a semantic differential - be presented as a graded rating scale (e.g. 1 to 6 as rating) for each element. It is also better suited for mathematical evaluations of the grid. The number of gradations on a scale also depends on the objective of the investigation and may have to adhere to the specifications of the evaluation programs. In addition to the conventional scaling methods, Menzel, Rosenberger and Buve have developed the so-called "Tetralemma field".
The tetralemma field ("four corners" in the sense of four positions or points of view) is a structure from traditional Indian logic for the categorization of attitudes and points of view, which the decision pattern "either" ("looking for") - "or" ("clear Direction ”) by questioning the opposing positions that initially appear to be incompatible and thus adding possible decision options. These four alternatives form the outer positions of a scale space in which the respondent can freely position all elements. It should be noted that scales without a middle position (2, 4, 6 etc.) prevent a “neutral” evaluation, so the respondent has to decide.
Evaluation algorithms
At the end of the investigation, there is always a completed matrix that represents the individual element assessment space. The completed grid can be evaluated in various ways, both with content analysis and with descriptive statistical methods. In the course of the development of the repertory grid, however, not only has the range of variation in the type of interview widened, the possibilities for evaluation have also increased significantly in terms of diversity and complexity. From the sole interpretation of the constructed constructs and their relationship to the elements and the elements to one another to complex, mostly computer-aided, mathematical processes, there is a wide range of evaluation approaches. With the multitude of evaluation forms, it should be taken into account that the quantitative data of the grid matrix are phenotypically mathematical numbers and the relationships between elements and constructs can be calculated to one another, but only the qualitative information gives the quantitative values a meaning. "The repgrids are located in the transition area from qualitative to quantitative methodology, because personal constructs, like other statements and messages from the APN ('informant', the author), must be understood and interpreted by the investigators (Us)".
The evoked constructs form the starting point for the evaluation of grids. They are sorted according to the meaning of their content so that similar construct or contrast poles lie next to one another. "This grouped list forms the basis for a deeper understanding of the construct world". Riemann formulates the approach as follows: “First of all, the degree of similarity between all pairs of elements or constructs should be determined”. In principle, the evaluation of a grid is about the recording of interrelationships and mutual relationships between elements and constructs to one another and to one another. A number of coefficients are available for the calculation, e.g. B. Jaccard coefficient , Phi coefficient or Spearman's rank correlation coefficient are available. According to Riemann, however, there is no such thing as an “optimal” coefficient. "The selection of a coefficient depends on how the similarity of two constructs is defined psychologically and what assumptions the investigator has about the emergence of ratings".
Hierarchical cluster analysis (Bertin)
One method is the so-called cluster analysis . Here the assessed elements and constructs are related to one another in a matrix based on their similarity. A cluster analysis in its simplest form is the mathematical sorting of the rows and columns according to their similarity with regard to the ratings in the cells. The cluster matrix is interpreted by comparing the similarity of the constructs and / or elements used. The graphic opposite shows an example of a sorted cluster matrix based on the method of the cartographer Jacques Bertin (1982).
The aim of cluster analysis is to identify entities, i.e. things such as elements or constructs, in terms of their similarity in a set. In the hierarchical cluster analysis, those constructs and elements are searched separately that were rated similarly in the interview; It does not matter whether you are in the above Tetralemma field or on the available scale. Elements and constructs are put together separately in groups. Distances between the found elements and / or constructs and the clusters are then used to iteratively build up the various similarity groupings until an optimal solution (element similarity to construct similarity and the sub-groupings over distances) has been found. Various possibilities are discussed in the literature as to which cluster analyzes are useful for repertory grids. Riemann offers a good introduction. Silke Wertz provides a solid mathematical representation in her diploma thesis. Christian Fischer, student of Arne Raeithel, also provides a nice insight into the cluster analysis for grid matrices.
Principal component analysis
Another common analysis method is principal component analysis; in English Principal Component Analysis (PCA). With this procedure, the numbers in the matrix are converted in such a way that we get coordinates on so-called main axes for elements and construct poles. If elements and constructs are shown simultaneously in one image, their mutual relationship can be viewed both geographically (distances) and idiographically (semantic directions through the constructs). This form of illustration is called the biplot method because constructs and elements can be read in their mutual relationship within a representation. As the evaluation of these biplots requires certain experience, a few introductory remarks are made at this point.
The PCA is based on an existing data matrix made up of elements and constructs. By means of a factor analysis, the charges of the variables are represented as points in the factor space. The result of the factor analysis are mutually independent factors that explain the relationships between the constructs bundled in them. Factor analysis is a data-reducing and hypothesis-generating process, suitable for checking the dimensionality of complex structures. The aim of a principal component analysis is to reduce the number of constructs specified in the repertory grid to as few independent components as possible and at the same time to achieve maximum variance explanation.
If one considers elements as variables and thus as vectors in the n-dimensional human space, the cosine of the angle between two vectors corresponds to the correlation between the variables [cos (0 °) = 1; cos (90 °) = 0; cos (180 °) = −1). For z-standardized variables (i.e. vector lengths of 1), the same value is obtained by projecting one vector onto the other.
If you now have several variables, the first factor to be extracted is the resultant of all variable vectors. The factor loadings (= the correlations of the variables with the factor) correspond to the cosines (Cosini?) Of the angles of the variable vectors to the factor (or the distances resulting from the projection onto the factor). The next factor is placed in the person space at right angles to the first, and so on, until a coordinate system is created from q factors in which all variable vectors can be represented in q -dimensional space (see figure on the right).
.png)
So it makes no sense to calculate factor analyzes when you have fewer people than variables: The n people span an n -dimensional space. If one now has more variables than people, these variables inevitably correlate with one another, since only n mutually orthogonal vectors can be accommodated in an n -dimensional space .
The whole thing can also be imagined with people in the variable space. The coordinate system spanned by the variables is then rotated until the projections of the person vectors on one of the dimensions are at a maximum; this dimension is then the first factor (which then explains most of the variance). Then the whole thing - while maintaining the position of the first factor - is rotated until the next dimension explains as much as possible of the remaining variance. This continues until all variance is explained. In fact, principal component analysis first extracts as many factors as there are variables, which explains all of the variance in the variables.
"These new axes can be understood - as is usual in factor analysis - as basic dimensions of the 'cognitive similarity space' or as mathematical aids, which initially have no independent meaning, for generating a picture of the mutual connection of the judgments." The principal component analysis is suitable for the calculation of single grids as well as for multigrid analysis. However, multigrid analyzes only work if at least the elements and / or the constructs for a calculation project are identical. In the resulting main component space (single or multiple analysis), elements and constructs can be viewed in their reciprocity.
Quality notes
The repertory grid method is a methodical procedure based on a constructivist understanding of science. It enables what few other methods are capable of, namely the systematic recording of a repertoire of subjective ideas about certain facts and their inter-individual comparison. The main advantages of the procedure are that it offers the respondent free opportunities to express themselves, but at the same time proceeds in a structured manner so that a quantitative evaluation is made possible and the “flood of data” of qualitative procedures can be avoided. Another advantage is that the basic method can be adapted to the specific application, e.g. B. in terms of the degree of standardization or the scope of the survey. Other variations of the method are also possible, e.g. B. language-free construction, comparative grids ("shared grids"), ABC method and playful grids. This means that the procedure can be used almost universally when it comes to diagnosing the subjective conceptions of reality of respondent groups (e.g. individual and group coaching, personal diagnostics, market research, corporate culture analysis, and much more).
The repertory grid technique is a qualitative and particularly in-depth process. All methods, tests and procedures in research are measured against the traditional quality criteria of validity , reliability and objectivity , regardless of the respective theoretical context . According to Fromm, however, tests of the validity of grid interviews make no sense. "Whether z. For example, a grid interview actually records the relevant distinctions that a person applies to a certain area of experience, can at best be assessed by the interviewed person himself in terms of content validation. Since the procedure is not intended to capture any specific characteristics (such as 'fear' or 'intelligence'), validation on an external criterion does not make sense because it is entirely open which criterion it could be. "Slater formulates in a similar way: " The reason is that the theory from which psychometric methods for measuring reliability and significance are derived assumes that samples can be drawn at random from an objectively defined population. The assumption can be satisfied by the nomothetic data in the table of test scores, but not by the idiographic data in a grid . "
With regard to the requirement for construct validation, the question to be answered is whether what was collected can be accepted as a personal construct in accordance with the theoretical requirements. This question cannot be related to the results, but at best to the basic procedure of the survey. On the basis of Kelly's theory, constructs are understood as preverbal distinctions that are often difficult to conceptualize. On the one hand, the often rapid processing of differentiation tasks can be mistrusted because it may only capture what is easy to formulate. On the other hand, this procedure, which allows the person examined enough time to formulate, increases the probability that the method not only collects existing constructs, but also those that the person examined only develops during the survey. According to Fromm, on the basis of the Personal Construct Theory, the second procedure mentioned would be preferable to the first as being more valid.
When asked about reliability, i.e. H. the reproducibility of the results, one can ask the counter question from the grid perspective, which results are meant, the number of constructs, the formulations, ratings or the relationships between constructs, elements, etc. There are many different ways of calculating reliability coefficients for the grid. The theory behind the method postulates that personal constructs are neither stable nor that they can be formulated identically at different times. “ Man, in Kelly terms, is 'a form of motion' not a statistic object that is occasionally kicked into movement .” “[…] High values are only to be expected in certain cases, e. B. when it comes to central constructs or structures of the construct system. "
Regarding the question of objectivity, Fromm explains that objectivity is guaranteed both in the implementation and in the evaluation. However, he restricts this statement to the quantitative evaluation, in which he guarantees a maximum of objectivity with the help of computer programs. In summary, it can be said that the traditional quality criteria can certainly be applied to the repertory grid method. If, as discussed above, the grid is used in a suitable manner, objectivity and reproducibility are largely guaranteed and respectable reliability coefficients can be demonstrated.
The time required for the implementation of approx. 1 to 1.5 hours per person and grid is not very high. If special evaluation software is used, the results are available relatively simultaneously with the completion of the survey. Since the repertory grid technique offers the respondents the opportunity to describe their constructs themselves and thus offers a reflective view of their own dimensions for describing phenomena in their living environment, acceptance is usually very high.
literature
- Bertin, J. (1982). Graphic representations and graphic processing of information. Berlin, De Gruyter.
- Fischer, C. (1989). Exploratory analysis of Kelly matrices. Hamburg, University of Hamburg.
- Fransella, F. and D. Bannister (1977). A Manual for Repertory Grid Technique. London.
- Fransella, F., R. Bell and D. Bannister (2004). A Manual for Repertory Grid Technique. Wiley & Sons, Weinheim.
- Fromm, M. (1995). Repertory Grid methodology. A textbook. Weinheim, German Studies Publishing House.
- Gilberto, M., C. Dell'Aversano, and F. Velogna, Eds. (2012). PCP and Constructivism: Ways of Working, Learning and Living. Firneze, Libri Liberi.
- Peter R. Hofstätter : Differential Psychology (= Kröner's pocket edition . Volume 403). Kröner, Stuttgart 1971, ISBN 3-520-40301-3 .
- Kelly, GA (1955). The Psychology of Personal Constructs. New York, Norton.
- Kibéd, MV v. and I. Sparrer (2005). On the contrary, Heidelberg, Carl Auer.
- Lewin, M. (1986). Psychological research in outline. Berlin, Springer.
- Menzel, F., M. Rosenberger and J. Buve (2007). "Understand emotional, intuitive and rational constructs." Personnel management 2007 (4): 90–99.
- Meyer, M. and K. Lundt-Verschaeren (1998). Development of a grid to record the job description of full-time supervisors. Diploma thesis in the psychology course, University of Bremen.
- Osgood, CE, G. Suci and P. Tannenbaum (1976). The logic of semantic differentiation. Psycholinguistics. H. Halbe. Darmstadt: 232-267.
- Raeithel, A. (1990). Work on the methodology of psychology and the Kelly matrix method. Habilitation thesis. Hamburg, University of Hamburg.
- Raeithel, A. (1998). Self-organization, cooperation, drawing process. Wiesbaden, West German publishing house.
- Riemann, R. (1991). Repertory Grid Technique - Manual. Göttingen, Hogrefe.
- Rosenberger, M. (2014). Vademecum repgrid. A guide to the professional application of the repertory grid technique. Volume 1: Legitimation, Theory, Methodology and Methodology. Norderstedt. bod.
- Rosenberger, M. (2015). Vademecum repgrid. A guide to the professional application of the repertory grid technique. Volume 2: Experts, practice, successful application examples. Norderstedt. bod.
- Rosenberger, M. (2006). Social control of virtual companies - Optimization of social relationships using the Repertory Grid Technique. Taunusstein, Driesen.
- Rosenberger, M. and M. Freitag (2009). The repertory grid technique. Handbook Methods of Organizational Research: Quantitative and Qualitative Methods. Cool / Strodtholz / Taffertshofer. Wiesbaden, VS Verlag.
- Scheer, JW and A. Catina (1993). Psychology of Personal Constructs and Repertory Grid Technique. Introduction to Repertory Grid Technique. Basics and methods. JW Scheer and A. Catina. Munich, Hans Huber: 8-10.
- Slater, P. (1977). Dimensions of Interpersonal Space. London, Wiley.
- Wertz, S. (2006). Repertory Grid - Investigation of a data analysis process. Konstanz, University of Konstanz.
- Willutzki, U. and A. Raeithel (1993). Repertory grids software. Introduction to Repertory Grid Technique. Basics and methods (Volume 1). JW Scheer and A. Catina. Munich, Hans Huber: 68-79.
Description 1
Description 2
Description 3
Description 4
Web links
- Homepage with a lot of information about PCP and Repertory Grid
- Freeware Repertory Grid
- Open source software for analyzing grids
- Homepage of the George Kelly Society
- Homepage repgrid
Individual evidence
- ↑ Kelly GA (1955)
- ↑ Kelly, GA (1955, p. 270)
- ↑ Rosenberger (2014), p. 109
- ↑ Rosenberger (2014)
- ↑ Fransella / Bannister 1977, p. 5
- ↑ See also Raeithel 1993, p. 42
- ↑ Fransella / Bell / Bannister 2004, p. 167
- ↑ Rosenberger 2015
- ↑ Rosenberger 2006, p. 201 ff.
- ↑ Meyer / Lundt-Verschaeren 1998, p. 104 ff.
- ↑ Willutzki / Raeithel 1993, pp. 70 ff.
- ↑ Osgood, Suci et al. 1976; Hofstätter 1971
- ↑ Gilberto, Dell'Aversano et al. 2012
- ↑ Menzel, Rosenberger et al. 2007
- ↑ Frans Ella / Bannister 1977
- ↑ Kelly 1955, p. 222
- ↑ cf. a. Scheer 1993
- ↑ Stephenson CR (1935)
- ↑ Osgood / Suci / Tannenbaum 1976, p. 256 ff.
- ↑ Menzel, Rosenberger and Buve, 2007, p. 95
- ↑ Varga from Kibéd / Sparrer 2005
- ↑ Rosenberger / Friday 2009
- ↑ Frans Ella / Bell / Bannister 2004
- ↑ Raeithel 1993, p. 42 ff.
- ↑ Raeithel 1990
- ↑ Raeithel 1993, p. 42
- ↑ Riemann 1991, p. 26
- ↑ Riemann 1991, p. 28
- ↑ Riemann, R. (1991)
- ↑ Wertz, S. (2006)
- ↑ Fischer, C. 1989
- ↑ Raeithel 1993, p. 54
- ↑ Raeithel 1993, p. 53
- ↑ Slater 1977, pp. 143 ff.
- ↑ Fransella, Bell et al. 2004, p. 98
- ↑ cf. Fromm 1995, 205 ff.
- ^ Lewin 1986: 77 ff.
- ↑ Fromm 1995, p. 203
- ↑ Slater 1977, p. 127
- ↑ Fromm 1995, pp. 203 f.
- ↑ Fransella / Bannister 1977, p. 82
- ↑ cf. a. Fransella et al. 2004, p. 133
- ↑ Fromm 1995, p. 204
- ↑ Fromm 1995, pp. 205f.
- ↑ Rosenberger 2014, p. 187