2. LITERATURE REVIEW
3.4 DATA ANALYSIS
3.4.3 Coding the Data
Within the process of coding, while some researchers adopt codes which have been determined prior to the coding procedure, most researchers ‗take the po- sition that the codes emerge from the data via a process of reading and think- ing about the text material‘ (Lichtman, 2013, p.248). Specifically in content analysis, ‗a theory or prior research is used to guide the analysis in the initial coding‘ (Lichtman, 2013, p.259), a view supported by Miles, et al., (2014, p.81) who acknowledged that ‗one method of creating codes is developing a provisional ―start list‖‘, hence deductive coding, while inductive codes are likely to ‗emerge progressively during the data collection‘. In this enquiry, a coding framework was initially employed for guidance purposes, thus promot- ing a deductive approach, while a range of other discrete codes emerged while studying the transcripts and finding that the pre-determined codes did not ac- commodate all of the data, thus promoting an inductive approach.
Although feedback and feed-forward have been found to affect students‘ self- perception, self-esteem, self-efficacy, motivation, and self-regulation (Harlen, 2006b), it has proved difficult to measure these attributes effectively owing to their subjective nature (Blumer, 1969). Nevertheless, various scales and inven- tories have been developed for this purpose, several of which were initially
employed to develop a provisional ‗start list‘ to guide the coding and catego- rising of data within this study, although these scales and inventories were not actually employed to measure these issues. The General Perceived Self- Efficacy Scale, designed by Schwarzer and Jerusalem (1995), together with the new General Self-Efficacy Scale, developed by Chen, Gully and Eden (2001) were used as a guide for coding elements relating to the subjects‘ per- ceived efficacy beliefs. Although the Intrinsic Motivation Inventory (Universi- ty of Rochester, 2011) was designed to measure subjective experiences relat-
ing to activities in laboratory experiments, the statements employed within this inventory were utilised for guidance when categorising and coding students’ perceived levels of motivation, relating to the feedback and feed-forward re- ceived within their studies.
Within individual students‘ piano practice, the ability to self-regulate is im- portant, especially when utilising the feedback and feed-forward received from their teachers. In order to analyse students‘ abilities to focus on the issues identified for improvement within their practice sessions, a number of scales relating to self-regulation were employed to develop codes and analyse the da- ta generated. The scales employed included the ‗Self-Regulation Scale‘ devel- oped by Schwarzer, Diehl and Schmitz (1999), together with the Scales for Goal Orientation of Practice, and External Action Distraction constructed by Kuhl and Beckman (1994, cited in Harnischmacher, 1997). Finally, the ‗form- ative feedback guidelines‘ presented by Shute (2008, p.177-181) as part of her literature review were employed for the purpose of categorising and coding different types of feedback and feed-forward.
As ‗naturalistic qualitative inquiry is concerned with description and explana- tion of phenomena as they occur in routine, ordinary natural environments‘ (Hitchcock and Hughes, 1995, p.296), the approach to data analysis was in- ductive in nature, where ‗patterns, themes, consistencies and exceptions to the rule‘ were formally identified. In relation to this issue, Taylor-Powell and Renner (2003) acknowledge the need for discipline, and the adoption of a sys- tematic approach to coding, analysing and interpreting qualitative data.
When coding the data, which Saldaña (2009, p.4) points out is not a ‗precise science‘, rather an ‗interpretive act‘, Seidel (1998) has observed a distinction between objectivist and heuristic codes. Objectivist codes provide a ‗con- densed representation of the facts described in the data‘ (Seidel and Kelle, 1995, cited in Seidel, 1998, p.14), while heuristic codes assist in collecting is- sues observed within the data so they can be subjected to further analysis. Heuristic codes assist in reorganising data and establishing different views or interpretations (Seidel, 1998). However, it has been acknowledged that in qualitative data analysis it is traditional to ‗treat code words as heuristic tools rather than objective representations of facts‘ (Seidel, 1998, p.13), and when coding data using ‗a heuristic approach, code words are primarily flags or signposts that point to things in the data‘ (ibid., p.14).
In this enquiry, through the adoption of a ‗constant comparative method‘, as outlined by Thomas (2009, p.198), which involved reading, and re-reading the transcripts in order to ensure that codes were applied with consistency, objec- tivist codes were employed where the transcripts provided ‗unambiguous‘ in- stances of specific issues such as the feedback provided to flag an error, for example playing an incorrect note, or feedback indicating the location of a
mistake, using the musical score to illustrate the error. Heuristic codes were applied to raise attention to issues within the transcripts that required further exploration, and the use of memos within NVivo 10 were employed to pro- mote ‗continued reflection‘ (Saldaña, 2009, p.37).
The availability of the original video and audio data enabled ‗repeated scruti- ny‘ of both talk, and in the case of the video data, non-verbal communication ‗at extraordinary levels of detail‘ (Luff and Heath, 2012, p.256). It has also been observed that when coding video data, or a transcription of a video re- cording, in some instances coding may be enhanced by viewing the videos in slow motion, a process which also promotes consistency and the verification and confirmation of the codes applied (Mondada, 2012) (see Appendix 4 for an example of video coding). In relation to this point, on occasions, for pur- poses of clarification or verification, while coding transcripts, the actual video recordings of lessons, and audio recordings of interviews, were revisited. With regard to the coding procedure, Krippendorff (2013, p.42) has observed that ‗deductive and inductive inferences are not central to content analysis‘ and some inferences may be abductive in nature, such as observing whether respondents in interviews are telling lies from their ‗non-verbal (facial) behav- iour‘. This proved pertinent in this enquiry, as some students would indicate in interviews that they had put a lot of time or effort into their piano practice while, on occasions, progress observed in lesson observations did not always support their comments.
It has been observed that the process of generating codes using CAQDAS can be exhaustive in nature, so it is important to ensure that this process is kept manageable (Gibson, 2010). The frequency counts represented the number of
times specific codes were applied in each case study. In relation to this point, Saldaña (2009, p.62) acknowledged the legitimacy of ‗simultaneous coding‘, or the application of ‗two or more different codes within a single qualitative datum‘, particularly as Glesne (2006, cited in Saldaña, 2009, p.62) observed that ‗social interaction does not occur in neat, isolated units‘. Within this en- quiry, there were occasions when codes were applied simultaneously to a spe- cific datum, although this occurred rarely.
While CAQDAS packages, such as NVivo 10, do not analyse data, they facili- tate the process of coding (Liamputtong, 2009) which is actually embedded within the data analysis procedure (Miles, et al., 2014). Within the case study folders in NVivo 10, transcripts were organized into sub-folders relating to each week of the study, totalling nine sub-folders for each case study. The documents were coded, initially guided by the range of scales and inventories discussed above, but other codes relating to the research questions were devel- oped through a process of refinement and revision, or relating to specific is- sues involved in studying the piano, such as ‗feedback about how to practise‘. Other codes were subsequently amalgamated, such as ‗highlighting errors without providing the correct response‘ and ‗incorrect, teacher acknowledges incorrect response with no additional information‘.
3.5 Conclusion
Within this chapter, the paradigmatic position, design frame, research meth- ods, research instruments, and the data coding and analysis procedures for this enquiry have been explained and discussed, together with issues relating to the ethics of conducting research, both with young students and adults.
A qualitative approach, with additional frequency counts was utilised within this research (Greene, 2008; Denscombe, 2008; Leech et al., 2010; and Tashakkori, 2009), focusing upon four case studies, which was considered an appropriate number of cases within a multiple case study design (Stake, 2006). The actual research methods and research instruments employed within the case studies were designed to gain a deep understanding of perceptions of the efficacy of feed-up, feedback and feed-forward in this particular context by observing piano lessons consistently over a prolonged period of time. The va- lidity of these methods has been considered, together with issues pertaining to the reliability of the data generated. The findings for each discrete case study are presented in the following chapter, and subsequently compared as part of the Discussion Chapter.