strengths of repertory grid techniquePersonal constructs |
CHAPTER 25 |
This chapter introduces personal constructs and discusses:
strengths of repertory grid technique
working with personal constructs
grid analysis
some examples of the use of repertory grid in educational research
difficulties in the use of repertory grid technique in research
resources
Personal constructs are the basic units of analysis in a complete and formally stated theory of personality proposed by George Kelly in his book The Psychology of Personal Constructs (1955). Kelly’s own clinical experiences led him to the view that there is no objective, absolute truth and that events are meaningful only in relation to the ways that are construed by individuals. Kelly’s primary focus is on the way individuals perceive their environment, the way they interpret what they perceive in terms of their existing mental structure, and the way in which, as a consequence, they behave towards it.
In The Psychology of Personal Constructs, Kelly proposes a view of people actively engaged in making sense of and extending their experience of the world. Personal constructs are the dimensions that we use to conceptualize aspects of our day-to-day world, and, as Kelly writes, people differ from each other in their construction of events. The constructs that we create are used by us to forecast events and rehearse situations before their actual occurrence, and are sometimes organized into groups which embody subordinate and super-ordinate relationships. According to Kelly, we take on the role of scientist seeking to predict and control the course of events in which we are caught up. For Kelly, the ultimate explanation of human behaviour ‘lies in scanning man’s [sic] undertakings, the questions he asks, the lines of inquiry he initiates and the strategies he employs’ (Kelly, 1969). Education, in Kelly’s view, is necessarily experimental. Its ultimate goal is individual fulfilment and the maximizing of individual potential, capitalizing on the need of each individual to question and explore.
Kelly’s theory was very formally constructed, with a ‘fundamental postulate’ and 11 corollaries that followed from this. Later, even in volume 2 of this work, but more obviously even later (Kelly, 1969), he moved away from this very formal statement. Butt (2008) provides a good introduction to this broader view of Kelly and his theory. Nevertheless the formal statement provides the model for the repertory grid. The fundamental postulate is:
A person’s processes are psychologically channelized by the ways in which he or she anticipates events.
One key component in the above statement is the ways in which; for Kelly, these ways are called constructs. The person’s repertoire of constructs are the basis by which the person construes or understands his/her world and makes predictions about the future. What is to be construed or understood are the events. What are these events? It is rather an ambiguous term. In one sense they really are events such as ‘going to a party’ or ‘teaching a class’, but Kelly uses the term much more broadly to encompass all psychological objects such as ‘ideal self’ or ‘teacher I looked up to’. In the repertory grid technique, these objects are termed elements.
A repertory grid then is simply a representation of the relationship between these elements and constructs. As such it provides information that we can use to understand how ‘a person’s processes are psychologically channelized by the ways in which he or she anticipates events’. Figure 25.1 shows a simple grid layout for collecting data and Figure 25.2 shows the grid with data.
Kelly’s theory then has a number of corollaries to this fundamental postulate that relate to the constructs. An important one from a repertory grid perspective, is that constructs are essentially bipolar, that is capable of being defined in terms of polar adjectives (good–bad) or polar phrases (makes me feel happy–makes me feel sad). Other corollaries can also affect the ways in which we can use the repertory grid technique, as we shall see later. In addition to the corollaries there are other formal aspects of the theory that play roles in relating constructs to behaviour. Fransella (2003: 455–7) provides a concise summary of all components of the theory.
A number of different forms of repertory grid technique have been developed since Kelly’s first formulation. All have the two essential characteristics in common that we have already identified, that is, constructs – the dimensions used by a person in conceptualizing aspects of his or her world – and elements – the stimulus objects that a person evaluates in terms of the constructs she employs.
Since Kelly’s (1955) original account of what he called ‘The Role Construct Repertory Grid Test’, several variations of repertory grid have been developed and used in many different areas of research. In a chapter entitled ‘Some uses to which grids have been put’ Fransella et al. (2004) provide an annotated bibliography of a wide range of areas including ‘working with children’, ‘teachers and teaching’, ‘construing of professionals’ and ‘those with learning difficulties’.
The repertory grid technique draws its strength from two particular features. The obvious one is that it is an individualized technique where the respondent provides the framework as well as the responses. For example, Suto and Nádas (2009) used grids with two principal examiners to identify features of examination questions that differed in the difficulty of marking. The other, perhaps more important, strength derives from the two-way nature of the data where elements are related to constructs. This enables relationships between elements to be assessed, since there is information about each element provided by the set of constructs. Conversely the relationships between constructs can be examined through the information provided for each construct by the set of elements. Even if the individuality of the grid is restricted by the use of provided constructs or elements (as discussed below) the two-way data allows for within-respondent analyses to be carried out.
Instructions: consider how the qualities in each row apply to each of the six figures. To the extent that the quality on the left applies more to the person, give a rating closer to 1. And to the extent that the quality on the right applies more to the person, give a rating closer to 5.
The key issue in choosing elements, is that they should be a homogeneous set to ensure that the constructs elicited from some of them, will also be relevant to other elements. Yorke (1978) and Wright and Lam (2002) draw attention to problems that can arise when this requirement is not met.
In Kelly’s original technique, elements were usually chosen by the client to fit ‘role titles’ provided by the clinician. Some role titles allowed no choice: ‘me as I am now’; some allowed a possible choice: ‘your mother (or the person who filled that role in your life)’; and some allowed wide choice: ‘a teacher you admired’. Some recent research (Bell et al., 2002; Haritos et al., 2004) suggests that value-laden role titles such as the teacher role above or ‘a girl you did not like’ tend to polarize the grid by making constructs subsequently elicited more similar to one another than when role titles are neutral such as ‘a significant person in your life’.
Kelly originally suggested six ways in which constructs could be elicited from elements, the most familiar being to choose three elements and to ask the participant to specify some important way in which two of them are alike and thereby different from the third. The way in which two were alike formed one pole (the similarity pole); the way in which the third differed formed the other pole. Another way involves asking the participant for the opposite to the similarity pole. Another way suggested by Kelly was to always have the ‘as I am now’ element included. This is more widely used in psychotherapy settings since it ensures that all constructs are relevant to the self.
The task of triadic comparison is cognitively demanding, and it has been found that simply using two elements is better for children (e.g. Salmon, 1976) or those with learning difficulties (Barton et al., 1976).
One form of repertory grid technique now in common use represents a significant departure from Kelly’s original procedure in that they provide constructs to subjects rather than elicit constructs from them. Eliciting constructs from individuals follows from Kelly’s individuality corollary: persons differ from each other in their construction of events. Supplying constructs contravenes this. However Kelly also posited a commonality corollary: to the extent that one person employs a construction of experience which is similar to that employed by another, his or her psychological processes are similar to those of the other person. This suggests that there will be constructs which are common to a number of individuals.
Can the practice of providing constructs to subjects be reconciled with the individuality corollary assumptions? Despite much research, the answer is still unclear and of course is further clouded by the more recently discovered effect of value-laden role titles in elicited constructs. As Fransella et al. (2004: 48) point out however, that ‘Constructs have to be supplied in a group context if group data is required.’ Bell (2000) has shown how the commonality corollary may be simply tested by examining each supplied construct in turn for unidimensionality of the element ratings.
But the issue of supplied or elicited constructs is not necessarily an all-or-none situation. Bannister and Mair (1968) support the use of supplied constructs in experiments where hypotheses have been formulated and in those involving group comparisons. The use of elicited constructs alongside supplied ones can serve as a useful check on the meaningfulness of those that are provided; substantially lower inter-correlations between elicited and supplied constructs suggesting, perhaps, the lack of relevance of those provided by the researcher.
In Kelly’s original technique, participants were allowed to classify as many or as few elements at the similarity or the contrast pole, giving a very lopsided construct. Originally this was seen as a problem since measures of association between constructs could be affected by this (e.g. Bannister and Mair, 1968: 59) and strategies were proposed to overcome this. These strategies had problems of their own, in that they forced the participant to allocate elements to constructs in a fixed fashion, as removed from Kelly’s individual focus as are supplied constructs.
More recently Bell (2004a) has shown that lopsidedness is associated with the super- and sub-ordinate relationships implied by Kelly’s organizational corollary. The common method now of allotting elements is the ‘rating form’. Here, the subject is required to judge each element on a multi-point scale, where one extreme (say 7) is aligned with one pole (‘notices when I am having problems’) and the other extreme (1) is aligned with the other pole (‘doesn’t notice when I am having problems’). As with most rating scale formats, questions arise as to the meaning of a midpoint rating (where an odd number of rating points are specified).
Another of Kelly’s corollaries was the range corollary: a construct is convenient for the anticipation of a finite range of events only. In a rated grid then, does a midpoint rating mean neither or both poles are relevant? Or more generally, does a grid allow missing data? Kelly’s range corollary would suggest ‘yes’ but the computation of summary measures to represent grids would usually say ‘no’. In practice this problem can be ameliorated if not overcome by the careful nomination of a homogeneous set of elements. The midpoint issue remains intriguing. Winter et al. (2010) found in a psychotherapy study, constructs, where either ‘self now’ or ‘ideal self’ was located at the midpoint, were associated with more complex cognitive structures in the grid. The rating form is the third example illustrated in Figure 25.2.
The technique known as laddering arises out of Hinkle’s (1965) linking of the notion of implication with the organization corollary (Each person characteristically evolves for his or her convenience in anticipating events, a construction system embracing ordinal relationships between constructs). Hinkle’s innovation was to replace ‘anticipation’ with ‘implication’. The linking of implication with ordinal relationships enables logical relationships to be specified between poles of different constructs.
‘Laddering’ is an exploratory technique using implication to move from a pole of a given construct to a pole of an as yet unelicited construct. It usually proceeds by asking the participant to indicate which pole of the given construct is preferred. (This is linked to Kelly’s choice corollary: a person chooses that alternative ... through which he anticipates the greater possibility for the extension and definition of his system.) Having identified the preferred pole, the participant is then asked ‘Why?’ The response to this forms one preferred pole of the higher order or implied superordinate construct. The construct can then be completed by asking the participant for the contrasting pole of the new construct. In turn the participant can then be asked for the preferred pole of this new construct and again asked ‘Why?’ to produce the first pole of the next higher order construct.
Although he never published his development of the technique of laddering, it has been used widely in many fields, particularly research into consumer perceptions (Reynolds and Gutman (1988) provide practical advice in using this technique in the context of laddering from product properties to consumer values).
Laddering is a technique for eliciting constructs in terms of a single element, the self (‘Which pole do you prefer and why?’). Pyramiding, a somewhat similar procedure developed by Landfield (1971), also uses a single element. Respondents are asked to think of a particular ‘element’, a person, and then to specify an attribute which is characteristic of that person. Then the respondent is asked to identify what kind of person would not have that characteristic. The researcher then returns to the first characteristic and asks ‘What more can you tell me about a person who has that characteristic?’ and again ‘What is the opposite of that characteristic?’ The enquiry is then repeated similarly for the opposite pole of the first characteristic. According to Landfield, the enquiry then proceeds to similarly enquire of the four construct poles thus elicited. Land-field termed this a ‘pyramid’ since it starts from one element to produce two construct poles which in turn produce four construct poles and finally eight construct poles. This kind of enquiry asks for elaborations (What more can you say?) rather than implications (Why?) and thus does not identify super-ordinate relationships between constructs as does Hinkle’s procedure.
Landfield saw his technique as purely qualitative and informing the psychotherapeutic process. Hinkle (1965) however went on to develop an Implication Grid or Impgrid, in which the subject is required to compare each of his/her constructs with every other to see which implies the other. Table 25.1 illustrates a laddering conversation.
Exchange grids are procedures developed to enhance the quality of conversational exchanges. Basically, one person’s construing provides the format for an empty grid which is offered to another person for completion. The empty grid consists of the first person’s verbal descriptions from which his/her ratings have been deleted. The second person is then invited to test his/her comprehending of the first person’s point of view by filling in the grid as s/he believes the other has already completed it. Various computer programs (‘Pairs’, ‘Cores’ and ‘Difference’) are available to assist analysis of the processes of negotiation elicited in exchange grids.
In the ‘Pairs’ analysis, all constructs in one grid are compared with all constructs in the other grid and a measure of commonality in construing is determined. ‘Pairs’ analysis leads on to ‘Sociogrids’ in which the pattern of relationships between the grids of one group can be identified. In turn, ‘Sociogrids’ can provide a mode grid for the whole group or a number of mode grids identifying cliques. ‘Socionets’ which reveal the pattern of shared construing can also be derived.
The way in which a grid is administered depends in part on the purpose and nature of the research. Where the researcher wants detailed information from few respondents, then administration is best carried out with one-on-one interviews. This has the advantage of allowing the researcher to monitor the constructs elicited for duplication or difficult to understand pole labels. Such administration usually begins with some discussion of the area to which the grid will relate and a simple trial run of two or three constructs elicited from half a dozen elements not related to the main task. Clarifications can be sought, and the preferred pole identified as the constructs are elicited.
Grids can also be collected in group testing. This usually requires a somewhat smaller grid (the study of Haritos et al. (2004) used this approach) and a proforma form that allows spaces for element and construct labels as well as grid ratings, and uses shading or some other method of identifying which elements are to form the triads (see Figure 25.1). Here overheads or PowerPoint are needed to illustrate how the grid is to be completed.
Where grids are to be collected with supplied elements and constructs (perhaps following on from some individually elicited grids, as in Reid and Holley’s (1972) study of choice of university) the grid data can be collected in a simpler questionnaire-type format. Should the ratings be collected construct by construct, rating each element in turn, or element by element, rating each construct in turn? Evidence suggests (Bell et al, 2002; Neimeyer and Hagans, 2002) that it does not matter.
In this century it is to be expected that computer administration is to be a major way in which grid data is collected. A web version can be found at http://gigi.cpsc.ucalgary.ca:2000/ for the Gaines and Shaw program Webgrid 5 that allows relatively simple grids to be elicited. Idiogrid (at the website www.idiogrid.com/) is a freeware program that allows quite complex grid elicitations to be structured and subsequently used to collect data from respondents. Both of these resources provide for data analysis of the grids collected.
Before we analyse the grid data, we need to be aware that the orientation of the data in the grid will be a function of the way in which the constructs were elicited. If the two elements that are alike are both positive figures (e.g. ‘ideal self’ and ‘best friend’) then the pole corresponding to their similarity will reflect this (e.g. generous). If the two figures are negative ones (such as ‘person I dislike’ and ‘worst teacher’) then the pole corresponding to their similarity (e.g. stingy) will – and the correlations with other constructs would be opposite.
A preliminary step in analysing a grid should often be to make all the constructs similarly aligned. This can be done by asking the respondent to indicate their ‘preferred’ pole. (This is also done in a related personal construct technique, called laddering.) Another way of doing this is when the grid contains the element ‘ideal self’. Whichever pole of each construct is aligned with the ‘ideal self’ is the preferred pole. Of course, this may not be of any use when the ‘ideal self’ is located at or near the midpoint. There is also an automatic way of doing this. We can analyse the correlations between the constructs and identify those constructs which generally correlate negatively with other constructs, and reverse these constructs.
How do we extract information from a repertory grid? There are a number of ways in which we can look at the numerical information in a grid, and these are discussed below.
Even in an individual grid, there is replicated information for elements (across constructs) and constructs (across elements). We can use this to make comparisons between constructs or between elements, or to create indices to represent these comparisons. One of the oldest of these is Bieri’s (1955) index of cognitive complexity/simplicity. This was originally a matching coefficient calculated for each pair of constructs and summed for the whole grid.
These days it is usually based on the average correlation among constructs. If this is a large value, then it means all constructs are highly correlated and relate to the elements in much the same way. The person might thus be said to be construing their world in a simple way. If the average correlation is low, then the constructs are differentiated and the person might be said to be construing in a complex fashion.
We can also compute such averages for each construct to determine which constructs are like the others and which are different. Table 25.2 shows the average (root-mean-squared) correlations for each construct in the same grid. This output was obtained from the freeware DOS program GRIDSTAT available at www.repgrid.unimelb.edu.au/grids.htm. It can be seen that the construct ‘sociable – aloof’ is less related to other constructs, while overall there is a consistent and substantial similarity in the ways these constructs are applied.
Of course the construct correlations that form the basis of this index can also be analysed in other ways, such as with principal components. Table 25.2 also shows that one principal component accounted for nearly 66 per cent of the variance in construct correlations. The component loadings also show constructs where the poles are reversed with respect to the orientation of the other constructs. All construct loadings have negative signs except for ‘sociable – aloof and ‘rejects ideas – accepts’. While the latter construct poles can easily be seen to be unaligned with the others (where the left-hand pole is the positive quality), sociable rather than aloof would normally be seen as positive. For this grid however aloof is perceived as the more positive quality in a teacher.
Cluster analysis is another way of depicting relationships in a matrix of measures of association among elements or constructs. Figure 25.3 shows element Euclidean distances and a hierarchical clustering of these. Good teacher and ideal teacher (teacher I would like to be) are similar, while teacher I learned a lot from is also similar to these two. Me as a teacher now is weakly associated with an ineffective teacher and somewhat less associated with teacher I did not learn from.
Particularly useful elements where grids involve the self as an element, are the elements self now and ideal self. The discrepancy between these two can be taken as a measure of self-esteem or used to generate a self-identity plot (as in Figure 25.4, taken from the Idiogrid program) where self and ideal are reference axes against which are plotted the other elements.
TABLE 25.2 GRID SUMMARY MEASURES
Average [Root-Mean-Squared (RMS)] Correlations |
|
|
Mean |
S. Dev. |
|
0.59 |
0.47 |
1 quiet – loud |
0.29 |
0.34 |
2 sociable – aloof |
0.45 |
0.43 |
3 open – private |
0.66 |
0.57 |
4 creative – follows set plans |
0.68 |
0.51 |
5 independent – dependent |
0.70 |
0.56 |
6 listens – doesn’t listen |
0.68 |
0.55 |
7 rejects ideas – accepts |
0.58 |
0.55 |
8 strict – lax |
0.58 |
0.50 |
Average of Statistic |
0.13 |
0.07 |
St. Dev. of Statistic |
Construct Component Loadings |
|
|
quiet – loud |
-0.82 |
|
sociable – aloof |
0.36 |
|
open – private |
-0.62 |
|
creative – follows set plans |
-0.90 |
|
independent – dependent |
-0.93 |
|
listens – doesn’t listen |
-0.96 |
|
rejects ideas – accepts |
0.93 |
|
strict – lax |
-0.80 |
|
Percentage Variance |
65.98 |
|
Dendrogram using average linkage (between groups)
Rescaled distance cluster combine
In the 1960s Patrick Slater (Slater, 1964) introduced a technique (that he called ‘principal components’) we now know as singular-value-decomposition which enables both elements and constructs to be represented together. There have been a number of different ways of representing the constructs and elements in these maps. Elements always appear as points in the map. Constructs however, are shown in different ways. The construct data are like a principal component solution with two columns of coordinates which define a point in the space. However representations differ. Often the point is reflected back through the origin to make a line symmetric about the origin, the two ends of which represent the construct poles. Other representations (such as that originally used by Slater) show the construct poles as points on a circle encompassing the elements.
Figure 25.5 shows a spatial representation of our sample grid using the program Idiogrid with elements shown as points and constructs shown as vectors (lines) symmetric about the origin. The longer the line, the more important the construct. The horizontal dimension separates better and poorer teachers, with constructs similarly aligned, while the vertical axis is aligned with the isolated construct aloof – sociable distinguishing principally between teacher I didn’t learn well with and me as a teacher now.
Another way of representing the whole grid is by a technique known in the repertory grid world as ‘focusing’. This means the rows (constructs) and columns (elements) are reordered and the grid data and labels shown in this reordering. This approach was originally devised by Shaw and Thomas (1978) and employs single-linkage hierarchical clustering to align constructs and elements. Other clustering such as k-means can be used and Table 25.3 shows such a rearrangement of the sample grid from the GRIDSTAT program referred to earlier. The focused grid is shown in two forms. On the left is shown the grid with two misaligned constructs, ‘sociable – aloof’ and ‘rejects ideas – accepts’ realigned as ‘aloof- sociable’ and ‘accepts – rejects ideas’ while on the right, the grid as elicited is shown. The original grid has four groups of constructs but in the aligned form there are only two.
Jones (1999) used repertory grids alongside interviews and participant observation to elicit head teachers’ views of their roles and agenda in changing times. While the study found an increase in their management activities (one construct), it also found that not only did their changing role not lead to their deprofessionalization but also their core values were rooted in their values in, and views of, education (a second construct). The super-ordinate constructs for the primary head teachers were child-centred and management, in that order, i.e. the management systems were there to serve the child-centred values and vision. Constructs elicited included, for example: child-centred problem-solving, implementation policy, evaluation, involving other agencies, problem-solving and paperwork.
Bezzi (1999) used repertory grids to explore the perceptions of the images of the geosciences held by a university geology lecturer and five undergraduates at the beginning and end of the academic year. Participants were provided with six names of science subjects (e.g. physics, geography, geology) as elements and 15 constructs were elicited by Kelly’s triadic method separately at the beginning and the end of the year. Construct labels were treated as qualitative data and were classified into five categories by the author: (i) nature of science (e.g. objective/subjective); (ii) aspects of investigation (e.g. use of maps or charts/no such use); (iii) application of science and its professional aspects (e.g. more employment opportunities/less, modifies environment/preserves environment); (iv) affective aspects (e.g. like/dislike, difficult/easy); and (v) characteristics of the courses (with lab/without lab).
Bezzi (1999) found constructs to be predominantly of the first two kinds, that is dealing with the scientific essence of the disciplines. He used principal component representations of both elements and constructs (as in Figure 25.5) to identify constructs associated with the elements ‘geology’ and ‘geography’ both before and after the course of instruction for each of the five students and the lecturer to show how their perceptions of these two subjects had shifted over the year. Bezzi (1999) was able to conclude that simply studying the content of science did not lead to a greater understanding of the role of science in society or how to make good public decisions about scientific issues confronting society.
Lui and Lee (2005) showed how repertory grids could be incorporated into a learning programme in an examination of conceptual understanding in computer-mediated peer discourse. Twelve graduate students rated each of six database design methodologies on 11 supplied database design concepts. The course instructor also completed this task. Use of supplied elements and constructs enabled the researchers to feedback information to students about conflicts between their perspectives and for the students to take these differences into account during a week of online unstructured discourse in which students were required to reach a consensus perspective. Following this, students again completed the grids, and it was found that student concepts became significantly more aligned with the instructor’s perspective. Lui and Lee (2005) conclude by suggesting that the methodology could be extended to using student-derived constructs although this would require more complex procedures.
Yeung and Watkins (2000) employed the repertory grid technique to investigate how student teachers in Hong Kong developed a personal sense of teaching efficacy. A pilot study was used to generate elements such as ‘self-efficacy’, ‘teaching practice’, ‘teaching practice supervisors’, ‘pupils’ and ‘lessons’. Constructs were then individually elicited from 27 students using the triadic procedure with cards. Yeung and Watkins (2000) matched constructs between student teachers to identify core constructs and create networks of similarity among student teachers using the REPGRID software of Shaw and Gaines (REPGRID 2, 1993). They found that third-year students’ perceptions were more homogeneous than those of first-year students.
Suto and Nádas (2009) used repertory grids to investigate why some GCSE examination questions in maths and physics were harder to mark accurately than others. Two highly experienced principal examiners generated constructs of question features for triads of questions (elements) and rated each question (i.e. the elements) on these constructs. The study examined the constructs generated in detail and related this to marking accuracy.
In a similar vein Johnson (2008) used the repertory grid procedure with assessors of vocationally related portfolios to elicit constructs of differentiation among portfolios. Six assessors generated 131 constructs over six assessment objectives. There was generally agreement between the assessors about the qualities of the commonly identified constructs, but Johnson (2008) did identify some potentially problematic linguistic issues, usually between the notions of quality and quantity.
Madill and Latchford (2005) explored identity change in four medical students over their first year of medical training, particularly in relation to their experience of human dissection. Each participant completed two repertory grids (one oriented towards their identity construction, the other drawing out their experience of human dissection) at two time points, early in term 1 and towards the end of term 3. The identity constructs elicited involved three common themes: dedication, competence and responsibility as well as negative reactions, such as feeling driven and stressed. Three major themes were apparent in their experience of human dissection: involvement, emotional coping and ability. Complex patterns of relationships between the grids and between occasions led the authors to see a development of a vulnerable sense of professionalism alongside a frustration of losing out potentially on wider aspects of personal development.
Lown et al. (2009) similarly investigated medical students’ perceptions of their personal and professional development. However they used an inventory adapted from a form of the repertory grid devised by Button (1994) to measure self-esteem that focused on the ideal and actual self. This adaptation of a grid was necessary because the large sample size (n = 339) required a more constrained version of the grid than is possible in individual-based studies and is a good example of a ‘repertory grid influenced’ study. All students generated five qualities that distinguished between an ‘ideal medical student’ and a ‘less than ideal medical student’ and rated themselves on this quality. The 1,531 attributes generated were reduced to a pool of 100 through the elimination of redundancies, etc., and 49 of these were selected for a freesort grouping by a subset of the students. Multidimensional scaling was then used to identify clusters from the co-occurrences produced by the freesort grouping. Seven clusters were identified and the original 1,531 terms classified under these. Most common categories were Personal and Professional Conduct, Committed Work Ethic and Time and Self Management.
There are also some examples of the use of laddering in research. Crudge and Johnson (2007) used the repertory grid to elicit a set of constructs relevant to web search engines (such as Google) from ten information science undergraduates. They then used laddering to determine the reasons for a construct’s importance within the user’s mental model. Using standard qualitative techniques they identified three hierarchical strata that conveyed the interrelations between basic system description, evaluative description and the key evaluations of ease, efficiency, effort and effectiveness. Two additional layers related to the perceived process and the experience of emotion. They concluded that their model of key evaluations with the conjunctions of procedural elements provided a framework for further research to evaluate search engines from the user perspective.
In another use of laddering Voss et al. (2007) used two laddering techniques (personal interviews and laddering questionnaires) to identify desired qualities of lecturers in a sample of 82 business management students. They found that personal interviewing led to more complex ladder structures with more components. Among the substantive findings they found that students’ academic interests motivated them less than the vocational aspects of their studies.1
The major overarching difficulty in repertory in research is the tension between individuality and commonality. A grid which is elicited wholly from the respondent is the most valid representation of that person’s construing. However research often demands replications across subjects, so that for some purposes there needs to be commonality across respondents. Where the researcher’s interests are qualitative, then individual grids form a useful way of collecting qualitative data.
However when the structure of the grid data is of interest, the quantitative aspects of the grid become important. When these are specific to an individual because of the individualized specification of elements and elicitation of constructs, the quantitative component of an individual’s grid cannot be related to that of others. In such situations the only way these individualized grids can be compared is through the use of grid summary measures, such as ‘cognitive complexity’. When grids have some aspects in common, for example role-specified elements or supplied constructs, it is possible to analyse the common aspects of such grids (see Fransella et al., 2004: 98–101).
There are also some important issues to consider concerning aspects of the grids themselves, particularly relating to the issue of bipolarity in the grid. When only one pole of the construct is used, unwarranted inferences about constructs’ polar opposites may be made. Yorke’s (1978) illustration of the possibility of the researcher obtaining ‘bent’ constructs suggests the usefulness of the opposite method (Epting et al., 1971) in ensuring the bipolarity of elicited constructs. There is also the previously mentioned uncertainty about the meaning attached to midpoint ratings. Value-laden element role titles (such as ‘A teacher I disliked’) can affect the structure of the grid (see Haritos et al., 2004), as can the orientation of construct poles. A number of practical problems experienced in collecting grid data (although it should be said these problems are not confined to grid methodology) are identified by Yorke (1978):
Variable perception of elements of low personal relevance.
Varying the context in which the elements are perceived during the administration of the grid.
Halo effect intruding into the ratings where the subject sees the grid matrix building up.
Accidental reversal of the rating scale (mentally switching from 5 =high to 1 =high, perhaps because ‘5 points’ and ‘first’ are both ways of describing high quality). This can happen both within and between constructs, and is particularly likely where a negative or implicitly negative property is ascribed to the pair during triadic elicitation.
The size of the grid can also present problems. Even the simple grid in Figure 25.2 with only six elements and eight constructs contains 16 pieces of qualitative data and 48 pieces of quantitative data. It is unlikely that such a small grid could represent the ways in which a person sees a complex aspect of their world such as a person’s view of their teaching world. But a substantially sized grid (in the psychotherapy world, a 15 element 15 construct grid is common and has 225 pieces of quantitative data) takes a substantial amount of time to collect. Laddering has its own problems, both in process (Butt, 1995) and in the hierarchical implications that follow (van Rekom and Wierenga, 2007).
Another problem is the continuing tension between theory and method in the repertory grid. Major works devoted to the repertory grid technique (Bannister and Mair, 1968; Fransella and Bannister, 1977; Jankowicz, 2003; Fransella et al., 2004) have all been written from a personal construct theory perspective and emphasize the importance and relevance of Kelly’s theory to the usage of the technique. Yet most research with the repertory grid is carried out with the grid being used in a purely methodological and atheoretical fashion that is content with a passing reference to Kelly as the originator of the grid. As this chapter demonstrates, the theory can be used to understand what is happening in the grid, but of course it is not essential to its use. The real drawback to the personal construct theory background of writers in this area is often with the jargon employed, particularly for indices or measures developed to summarize structures in the grid. For example, the average correlation used to summarize relationships between constructs was originally termed ‘intensity’ by Bannister and is often referred to by that name.
It might be thought that the repertory grid is unchanging. However, there has been continuing research to inform the technique itself (some details of such developments have been mentioned in this chapter) as well as new measures that can be derived from grid data (for example Bell (2004b) has devised a new index of inconsistency in ratings in the grid) that could have applications in educational research. Most, if not all, of these developments occur through the work of researchers identified with personal construct psychology. Accordingly a number of resources available on the Web are listed as follows:
A very general and comprehensive site with many resources and links:
another comprehensive site:
Personal Construct Psychology searchable database:
www.uow.edu.au/health/psyc/research/pcp/database/index.html
Online journal Personal Construct Theory and Practice:
Europe:
North America:
Australia:
Webgrid 5, the current web-based software of Shaw and Gaines:
http://tiger.cpsc.ucalgary.ca/
Idiogrid, a comprehensive windows-based program:
Gridstat, the author’s comprehensive DOS grid program (Italian version available). Also a comprehensive document on using SPSS syntax with grids is available here: www.repgrid.unimelb.edu.au/grids.htm
Gridsuite, a grid program available in German or English:
Gridcor, a grid program available in Spanish or English:
www.terapiacognitiva.net/record/gridcor.htm
Rep 5, the Shaw and Gaines comprehensive program:
Scivesco, analyses multiple grids with sophisticated graphics:
http://elementsandconstructs.de/en/home
Companion WebsiteThe companion website to the book includes PowerPoint slides for this chapter, which list the structure of the chapter and then provide a summary of the key points in each of its sections. In addition there is further information on repertory grids and triadic elicitation. These resources can be found online at www.routledge.com/textbooks/cohen7e.
