CHAPTER IV RESEARCH METHODOLOGY
4.5 Data analysis
No single model or reference was used to devise the approach to data analysis adopted in this study but a combination of many: Spradley’s domain
analysis (Spradley, 1979); Strauss’ open, axial, and selective coding (Strauss, 1987); Rubin and Rubin’s stages for coding interview data (Rubin and Rubin, 1995); Kvale’s steps of interview analysis and approaches to the analysis of meaning (Kvale, 1996); Boyatzis’ ideas for the formation of theme and category clusters (Boyatzis, 1998); Creswell’s description of holistic, embedded, within- case, and cross-case analyses (Creswell , 2007); and Yin’s analytic techniques (Yin, 2009).
A distinction should be drawn between the analysis of the data gathered through the grammaticality judgement task and of those collected through the rest of the methods. In the case of the former, the teachers’ corrections were examined in relation to the four tasks they had been asked to complete: 1. identify the grammar mistakes, 2. provide the correct grammar version of each faulty part, 3. name the grammar aspect in each correction, and 4. explain the grammatical rule which they thought had been broken. The results obtained were then categorised. In the case of Emma, for instance, the data were arranged into the following emerging categories and sub-categories: corrections (mistakes identified and corrected, mistakes not identified, mistakes identified but not corrected, unnecessary correction, wrong correction, lack of confidence), grammatical terminology (mentioned, not mentioned, wrong), and explanations (complete, no or incomplete, wrong) (see sample in Appendix 3). In other words, the analysis of the data generated through the grammaticality judgement task consisted of two stages: examination of the teachers’ corrections and the categorisation of the results.
In the case of the rest of the methods, the material was first analysed thematically and then codified. Although this represents a departure from some widely used procedures (e.g., grounded theory), where themes emerge at a late stage in the analysis following open and axial coding, it allowed due weight to be given to the categorisation which informed the research questions and offered a way of dealing with the complexity and interrelatedness of teacher cognition.
A chart was created with three columns: 1. raw data, 2. theme(s), and 3. codes, comments, and insights (see sample in Appendix 4). First, the data were linked to pre-conceived themes which were derived from the research questions of the study [salient grammar teaching features (research question 1); PLLEs, KAG, and GRPK (research question 2); and CFs (research question 3)], though some other topics emerged such as teachers’ knowledge of learners, syllabus,
and context; teachers’ beliefs and perceptions; teachers’ personal and professional history; information about the school; and the effect of the study on the teachers and their teaching practices. This thematic analysis was not an easy task, especially because teachers’ cognitive bases cannot always be seen as separate and discrete. As Nilsson points out, “a limitation with analysis based on ‘separate knowledge bases’ […] is that, given the complexity and intermingling of participants’ knowledge, thoughts, ideas, and practice in the process of teaching, although analysis assumes that the knowledge bases are completely separate and distinct, in reality they are combined in a complex web” [sic] (2008: 1288). Second, the data chunks associated with specific themes were codified. This process generated codes, comments, and insights which further specified the nature of the themes in the second column.
Once all the data had been analysed thematically and codified, the codes, comments and insights associated with similar themes (e.g., GRPK) were grouped in a separate document. Within each theme, these codes, comments and insights were examined in detail and arranged into categories. When categories had been derived within each theme, the material was further analysed within and across the categories. This resulted in a new categorisation of the data, in which several previous categories had merged into one, some others had been eliminated, and some new ones had emerged. Rubin and Rubin claim that “examining the material in individual categories allows you to refine what a concept means, compare examples of a theme, or piece together the separate events in a narrative”, and add that “comparing material across categories allows you to figure out which themes seem to go together or contradict each other” (1995: 241). Subsequently, a detailed description and interpretation of the data in each category were provided. During this process, the categories were again subject to some modifications and a final set of categories was produced. The description was often illustrated with quotes from the participants and with a brief reference to their context of occurrence. This required that I examine the raw data again and ensure that all evidence was grounded in the data. In short, this approach to data analysis involved four definite processes: thematic analysis, codification, categorisation, and description and interpretation of data. These processes were iterative in nature, with themes, codes, and categories being constantly relabelled, redefined, eliminated, reconsidered, regrouped, etc.
Data analysis and description were carried out first within each case
(within-case analysis). Then, a comparison of categories across the two cases
(cross-case analysis) was made, which produced an answer to research issue A. Finally, there was an interpretive phase in which I engaged in a discussion of the relationships among PLLEs, KAG, GRPK, and CFs in relation to the teachers’ salient grammar teaching features (research issue B).
4.6 Conclusion
This chapter has provided a comprehensive discussion of the paradigmatic and methodological considerations underlying this study. I now move on to describe the research findings.