Chapter 3 Methodology
3.4 Data analysis
3.4.4 Categorisation and theory building
This explanation of how I used CAQDAS to facilitate my analysis might give the impression that the thematic categories I mention were arrived at after an unreflective process of examining data for indicators of thematic connections and coding them without due regard to the problematic aspects of categorisation. It might also be read as an attempt to convince the reader that
all of the thematic categories emerged exclusively from the data. However, it is important to note that I used a number of extraneous concepts later on to develop these initial categories into a thematic framework. The final section of this chapter attempts to do justice to this process of “give-and-take” between data and theory by reflecting on the nature of categories and the implications of using categorisation as an interpretive technique.
There is some debate over the use of categories in social science research and at least some of the disagreement seems to be focussed on what researchers deem categories to be. The main question is whether categories can be clearly defined and viewed as isolated variables, or whether categorisation is seen as problematic and complex. In keeping with my social constructivist perspective, I would like to reflect on the process of categorisation since my own experience of creating and using categories in this research project was not straightforward.
“In the classical model, categories are indeed categorical and express a clear and complete conceptualization of phenomena in terms of common features. A well-defined category will have attributes that are jointly sufficient and singly necessary to identify the category. Only members of the category will possess all these attributes, and the members of the category will possess each one of them”. (Dey, 2007: 169)
The neatness of this definition of categorisation accords with Glaser‟s (1978) concept-indicator model, which describes how “indicators are used not to substantiate a category empirically through description but rather to elaborate the category through exploring its different dimensions” (Glaser, 1978: 43, cited in Dey, 2007: 168). Although Glaser discourages the use of data as description, that is, to “substantiate” categories, nevertheless, he does infer that the data can be used as empirical indicators of categories. There is an inference here that the construction of categories is unproblematic since the data will naturally fall into separate categories.
However, the clarity of this conceptualisation of categories seems to be in contradiction to the practice of constant comparison (Glaser and Strauss, 1967), which involves a continuous review of categories according to their correspondence with the data. The process of categorisation is therefore on- going and open-ended since constant comparison will require the parameters of
categories to be continually revised. Furthermore, research in psychology and linguistics has found that categorisation is a far more problematic process than the “classical” model suggests.
Dey (2007) lists a number of theory-based accounts of categorisation which have emerged since the work of Rosch (1978) on prototypes and McNeil and Freiberger (1994) on fuzzy logic. According to these accounts, “categories and categorization depend on our conceptual understandings of the world, rather than on similarity between characteristics” (Dey, 2007: 170). It is this acknowledgement of the role of theory and experience which provides justification for the use of “sensitizing categories” or “sensitizing theories” in Grounded Theory, since it shows that, far from emerging directly from the data, categories are extracted by informed researchers according to their previous experience of sociological models. This is not to deny the use of similarity of features as an important element in categorisation, but it does recognise the importance of theory in the process.
“The recognition that categories are theoretically informed (or motivated) creates a conceptual space for the sensitizing role of categories that is recognized in grounded theory but that is otherwise hard to find in the classic concept–indicator model” (Dey, 2007:170).
Dey also points out that categorisation is used not just for descriptive purposes, but to explain or make inferences. Since the purpose of categorisation in Grounded Theory is primarily the identification of conceptual elements which can be used to construct theory, its inferential purpose must be acknowledged from the start. For example, during the process of analysis, I was conscious that some of my initial categories emerged from the data, but as I re-coded the data using more and more sub-categories, themes came through which prompted me to read further and find out how they were dealt with in academic literature. An example of this was the category “metacognitive skills” (see Section 4.2.3 of the Findings chapter). This theme emerged early on in the interviews as students talked about the skills they gained on Active Learning modules. Reading around the topic of metacognitive knowledge, I came across the work of Schraw and Moshman (1995), who propose three different kinds of metacognitive knowledge: tacit, informal and formal, which I used as convenient
sub-categories to code a number of comments by my interviewees. I then used a further classification proposed by these authors to characterise the source of each type of metacognitive knowledge. Using this typology to classify various kinds of metacognitive knowledge allowed me to look for similarities across interviewees and to reflect on how this level of analysis constituted a challenge to some of the large cultural explanations of learner experience which I had seen in my literature review.
I conclude this reflection on the use of categories in data analysis with a brief discussion of the criteria according to which certain categories can be judged to be more salient than others during the analytical process, either in terms of their ability to link large numbers of variables, or in their usefulness as explanatory constructs. Holton (2007) considers the identification of these “core categories” as essential to the process of theory building since this enables the researcher to limit subsequent data collection and coding to themes which are relevant to the emerging conceptual framework. For this reason, it is important to reflect on the criteria by which core categories can be recognised. Holton‟s (2007) criteria are very practical for this purpose:
“The criteria for establishing the core variable (category) within a grounded theory are that it is central, that it relates to as many other categories and their properties as possible, and that it accounts for a large proportion in the variation of a pattern of behaviour. The core variable reoccurs frequently in the data and comes to be seen as a stable pattern that is increasingly related to other variables.” (Holton: 2007: 281) In addition to providing helpful criteria, this quotation eloquently expresses the way core categories are not just “out there” waiting to be discovered by the researcher, but emerge gradually and influence the data collection process itself. In my analysis of the pre-pilot interviews, I identified a number of categories (see Table 4) which seemed to correspond to the wide range of experiences related by my participants. However, not all of these (e.g. living in the UK; motivation; work for the family business; racial abuse from local residents) either accounted for the wide variation of experiences given or related easily to other categories. On the other hand, I identified four core categories (language, relationships, skills and pedagogic differences) which were related to most of the aspects of experience discussed by my participants
in the pre-pilot group interviews, and which also seemed to hold the potential for theorisation along lines which avoided the reductionism and determinism of some of the work I refer to in my literature review. Focussing on these core categories in the main interviews enabled me to collect further data which could be used to give further support or modify my emerging conceptual framework.
In my data analysis, categorisation therefore included both comparison, a broadly objectivist process in line with Glaser‟s concept-indicator model, and concept-building, a thoroughly interpretivist process of building theory grounded in the data, but incorporating external elements such as sociological theories, common sense, personal experience etc. There is a sense therefore in which my thematic framework, which recognises the importance of language, relationships, skills and group work processes in students‟ experience of Active Learning pedagogies, is grounded in the data. However, it is also important to acknowledge the role of external elements in this emergence, including my own knowledge of social theories and my personal experience of the give-and-take between data and theorisation during the act of analysis.