2 Literature Review
4.4 Focusing the analysis
4.4.1 Categorising data
Categorising data is sometimes referred to as coding, indexing, reducing or breaking down data (Attride-Stirling, 2001). There have been challenges to the centrality of coding as a way of getting to know data. Coffey et al (1996) have argued that coding is overemphasized, given the fact that a large part of the qualitative researcher's work consists of interpretation. The other criticism levelled against categorising data is that it leads to decontextualization of data. Data becomes decontextualized when “they are separated into units of meaning through coding and sorting” (Ayres, 2003:872). However, Seers (2012) counter argues that data “are
re-contextualized as they are reintegrated into themes that combine units similar in meaning taken from the accounts of multiple research respondents” thus making it possible to make interpretation of patterns across collected data. While acknowledging Coffey et al’s reservations I argue that creating categories is important as it involves unearthing and identifying themes or patterns salient in the transcribed textual data (Attride-Sterling, 2001) and these categories eventually become the key variables used to explain similarities and variations within the data (Dudwick et al, 2006:36). However for this to happen there has to be clarity about what constitutes a theme.
For purposes of categorising data an important question to address is what counts as a theme or pattern, or what “size” does a theme need to be (Braun and Clarke, 2006)? Seers (2012), argues that themes are abstract concepts, reflecting the researcher’s interpretation of patterns across data. Braun and Clarke (2006:10) provide a more comprehensive view of theme when they argue that a theme captures something important about the data in relation to the research question, and represents some level of patterned response or meaning within the data set. Ideally there will be prevalence of the theme throughout the data but Braun and Clarke warn that mere prevalence may not be a pointer to the importance of the theme or pattern. The importance of a theme depends on whether it captures something important in relation to the overall research question. Therefore researcher judgement is critical and guides the categorising process as well as the identification o f all data that relate to the already classified patterns (Aronson, 1994).
Attride-Stirling (2001:391) advises that the most efficient way to identify data that relates to identified categories is to use a categorising framework based on specific topics, recurrent issues in the text, or on a set of theoretical constructs that are to be explored systematically. The following themes emerged from data categorising:
• the influence of language background on language use in primary classrooms, • influence of attitudes towards language policy,
• features of teacher and pupil talk,
• influence of attitudes towards code switching on patterns of language use, and • the influence of testing practices
The five broad themes identified here cover the four research questions this study sought to answer. After identifying the categories the next step was to label or name the categories, define them and provide a description of how to know when a category occurs. Attride- Stirling (2001:392) warns that “the codes in the coding framework should have quite explicit boundaries (definitions), so that they are not interchangeable or redundant; and they should also be limited in scope and focus explicitly on the object of analysis” so that every sentence in the texts does not have to be coded. Table 5 shows the themes that emerged and the sources of the data for the themes while Table 6 shows how the themes were labelled, defined and described. The themes for teachers and children’s interviews were the same with the only difference being that in the section on attitudes towards language policy children were not asked to express opinions about national language-in-education policy but were asked for their views on school language policies.
Table 5: Themes emerging from data categorising
Theme Source of evidential data Influence of differences between home and school
language
Teachers, learners and lesson observation
Effect of parents’ attitudes towards English on language use in classrooms
Teachers and learners
Support for national and school language policies Teachers and learners Teachers’ and learners’ preferred language for
communication inside classrooms
Teachers, learners and lesson observation
Using the mother tongue to teach and learn Teachers, learners and lesson observation
Teachers’ and learners’ attitudes towards code switching in class
Teachers, learners and lesson observation
Influence of testing practices on language use in classrooms
Table 6: Defining themes for data analysis
Theme Definition Description Perceptions of home
and school language practices
Awareness of the influence of the language normally spoken by teachers and learners at home or outside school and the language they use inside school and during learning
This category contains data that show the influence of language practices outside school on language use within the classroom and the school in general. It also includes data that show the influence of parental language for learning preferences on teachers’ and learners’ attitudes towards English and language use patterns found in classrooms.
Attitudes towards language policy
Language policy at national level is defined as the official stipulation of how languages are to be used in education and at the local level o f the school it is defined as language use preferences adopted and enforced by the school with or without backing from education authorities.
Category contains data that show teachers’ awareness o f language policy at macro and local levels and teachers’ attitudes towards policy stipulations. For learners this category contains data that show their awareness of and response to language policy at school level and how this impacts on how they use language in class.
Features of teacher and learner talk
Teacher-leamer talk is defined as the language used in and out of class between teachers and learners and among learners themselves whether for learning purposes or general interaction.
Category contains data that show the language the teacher permits learners to use to talk to him/her as well as the language pupils are required to use when interacting among themselves. This category excludes the language learners use to talk to peers outside the school but includes the way teachers interact with learners outside the school.
Attitudes towards code switching
Code switching is defined as the inter-sentential and intra- sentential alternate use of English and Shona/Ndebele to promote comprehension during a learning episode
Category contains data on teachers and learners’ responses to the alternate use of Shona/Ndebele and English as medium of instruction as well as reasons given for adopting code switching as a communication strategy during teaching- learning encounters.
Influence of testing practices
Testing practices are defined as the use of a particular language to assess pupil progress during or at the end of a course of instruction
Category captures data that show how teachers perceive testing practices as influencing the way they use language when teaching.
Identification of the themes was followed by the data categorising process. Taylor-Powell and Renner (2003) argue that this is the crux of qualitative analysis and involves reading and re reading transcripts in order to identify coherent categorises and collapse data from across cases into these categories. This process is based on the presumption that the origin of each unit of meaning is less important than its membership in a group of like units. “Inevitably, some of the original context in which each unit of meaning occurred is stripped away as the data are reduced” (Ayres, 2003) but such context stripping is consistent with the goal of making comparisons across respondents. To aid the categorising process in this study I utilised the list of questions below suggested by de Hoyos and Bames (2006):
• What is this incident about?
• What category does this incident indicate?
• What property of what category does this incident define? • What is the ‘main concern’ of the participants?
Adopting these questions as a basis for categorising data meant that I accepted that research questions for the study are not necessarily the same as those that guide the categorising process. In fact Braun and Clarke (2006:15) go so far as to suggest that “it is often desirable that there is a disjuncture between them”. I became aware o f the importance of this observation when I started categorising data for this study. During the constmction of the interview schedules I had grouped questions into theme areas and initially used these theme areas as thematic categories. I then sought to group participants’ responses to questions into these thematic categories. However, during the coding process it became clear that I was creating artificial categories and therefore took Braun and Clarke’s advice that research
questions should not be equated with thematic categories. I decided to search across interviews for repeated patterns of explicitly stated meanings. Explicit meanings are specified here because during transcription I had decided not to transcribe non-verbal aspects of interview participants’ speech and including latent meanings would not be justified as I had in a way stripped the data of some of the latent meanings during transcription.
The argument for not tying categories to interview questions can be illustrated by the experiences I had as soon as I started the coding process. For example, in response to the question “do you speak the same language as the learners you teach, Mrs Mguni commented:
We notice here at our school that they like English better than their mother tongue. And usually they fin d expressing themselves better in English because at home there is no proper Shona even among the parents themselves
In her response Mrs Mguni does not address the question of whether she speaks the same language as the learners she teaches. If I had used questions as the basis for categorising teachers’ responses I would have missed the fact that the response focuses on whether the language spoken outside the school is suitable for use as medium of instruction. Also in many instances although teachers were responding to a particular question their responses overlapped with other categories. This was evident in questions to do with language use in classes. In most instances teachers justified their choice of language code or code-switching practices by referring to language background and language policy. As a result the coding had to be based on patterns of themes emerging from the data rather than questions contained in the interview schedule.