Chapter 5: Implementation of the research
5.6 Data analysis
5.6.3 Some data analysis issues
There are numerous issues worth noting during a phenomenographic data analysis (Åkerlind, 2005d; Åkerlind et al., 2005; Bowden, 2005; Lin & Tsai, 2008; Marton, 1986; Marton et al., 1993; Sharma, 1997; Walsh, 2000), all of which need to be resolved to guarantee the quality of the analysis, although different researchers may propose diverse solutions. Thus, as well as discovering and assigning referential and structural (internal and external horizons) aspects, identifying categories of description, and establishing the outcome space, these issues need to be clarified in the data analysis.
5.6.3.1 People-phenomenon relationship
It is inappropriate for phenomenographers to construct the structural relationship of categories of description parallel to determining the categories because “there is potential to distort the categories by including the relation of the researcher to the phenomenon in addition to the true focus of study, the relation between the subjects and the phenomenon” (Bowden, 2005, p.16). Phenomenographers are expected to understand “the way a group of individuals perceive the target phenomenon and not the phenomenon per se (which would represent the first-order perspective)” (Paakkari, 2012, p.24). Essentially, the objective of a phenomenographic study lies in the relationship between the subject and the phenomenon investigated. Although other relationships, such as the one between the researcher and the participants, and between the researcher and the phenomenon (Figure 5.1), inevitably exist in the study,
they may distract the focus, as well as the outcome, and should thus be bracketed as much as possible.
Figure 5.1 Phenomenographic relationality Source: Bowden (2005, p.13)
Without the awareness of bracketing, researchers might add or adjust “categories where this is not supported by the data” (Walsh, 2000, p.23), and they could also impose “a logical framework on the data where this is not justified” (Walsh, 2000, p.23). As a result, they analyse the data “from the researcher’s or content expert’s framework, so that the interpretation of the data is skewed toward an accepted or expert view of the phenomenon” (Walsh, 2000, p.23). In this sense, the interpretation is not based on the collected data, but on the researcher’s framework.
The concept of ‘bracketing’ is crucial for both the collection and analysis of the data in this research. Bracketing in the interviews enabled me to keep an open mind to the interviewees’ responses and facilitated the acquisition of unbiased information. During the analysis, bracketing could help to discern people’s experience and conceptions as faithfully as possible. According to Walsh (2000, p.15), the most effective way to achieve this is “to base all analysis on the transcripts: if it is not in the transcript, then it is not evidence”.
5.6.3.2 Pool of meaning or whole transcripts
There are various approaches to deal with transcripts. The ‘pool of meaning’ is basically a ‘de-contextualised collection of fragments’ of the subjects’ statements and the starting point for the data analysis (Reed, 2006). According to Marton’s (1986) method to deal with transcripts, researchers first pick up some extracts related to the research question from the whole transcripts and then place them together as a ‘pool of meaning’. Therefore, he concludes that there are two contexts to interpret the meaning of quotes, namely, the original transcripts from which they are elicited and the ‘pool of meaning’. Marton prefers the ‘pool of meaning’ approach, while Bowden (1996, 2000) uses the holistic interview transcripts for the analysis. Bowden (1996, p.61) contends that “such de-contextualisation makes the task more difficult and is a methodological variant which is at odds with the underlying relational nature of phenomenography”.
The fact that both of these approaches have their advocates is very interesting. As Reed (2006) observes, the Swedish researchers favour the ‘pool of meaning’ approach, while the Australians prefer to analyse all the transcripts. Both of these approaches have their drawbacks. Using the ‘pool of meaning’ approach, selected quotes and excerpts may not be faithfully and accurately interpreted in the de-contextualised context rather than the original transcripts from which they were elicited. Conversely, when utilising the whole transcripts approach, researchers are inclined to immerse themselves in individuals’ statements, rather than analysing them at a collective level. Some proponents of the ‘pool of meaning’ approach note that the whole transcript approach could make it difficult to identify the key aspects of experience; for instance, Åkerlind (2005d, p.327) contends that “taking a whole transcript approach to analysis may reduce the clarity of the key aspects of meaning that researchers search for, because the meaning a phenomenon holds for an individual may vary during the course of an interview”.
meaning’ approach. The essence of phenomenographic analysis lies in comparing and contrasting between different individual transcripts so as to identify the meaning and structural aspects and further generate the categories of description. While the whole transcript approach places much attention on individuals and can identify key aspects of experience within a transcript, comparison that ought to be made between different interviewees is weakened and marginalised. Thus the approach might violate the collective-level analysis that will be discussed below. Yet while using the ‘pool of meaning’ approach and removing the utterances from their context, I did not downplay the importance of the original transcripts. In practice, I often re-visited and consulted the original context to justify and clarify the meaning of utterances, especially when the statements were vague and hard to understand.
5.6.3.3 Mixed conceptions in responses
In all the cases, the students did not simply express their sole conception of learning; for example, one response may have contained mixed conceptions across distinctive categories. A number of researchers have encountered a similar situation (Lin & Tsai, 2008; Marton et al., 1993). As Chiou, Liang and Tsai (2012) observe, the developmental and experiential components of conceptions of learning can exist simultaneously. Unsurprisingly, individuals can have numerous conceptions of learning, even if they have formed more advanced conceptions. It is suggested that the most dominant and significant category should be interpreted and elicited by the researcher in an attempt to make the analysis direct and clear, since the interviewees could have proposed some explanatory conceptions to arrive at the dominant one (Sharma, 1997); in other words, the researcher is expected to be able to identify the true meaning and major purpose of a participant’s response. However, Lin and Tsai (2008, p.564) disagree with this position and state that “dominant and minor categories coexisting simultaneously may provide potential indications toward the conceptions of learning. The ‘whole picture’ of the learners’ conceptions of learning needs to be entirely and truly represented”.
In my research, I attempted to combine the merits of both solutions, but I followed neither of them strictly. Sharma’s (1997) recommendation reminded me that for some interviewees there might be some ways of experiencing learning that seemed to be more important than others. If the students compared two conceptions and deliberately chose their preferred one, I only took account of that one, rather than both of them. But I was also aware that Sharma’s suggestion could lead to results which might not always be faithful to the interviewees because of the researchers’ intervention while discriminating the data. The most important value of Lin and Tsai’s (2008) solution was that it advised me to keep an open mind to all possible conceptions during the data analysis. Moreover I did not exclude the possibility that the interviewees may have treated some conceptions as equally important. However, I also noticed the weakness of this proposal that it was inappropriate to aimlessly list and value all the conceptions as equal regardless of their significance.
5.6.3.4 The collective level
The transcripts need to be analysed at a collective level, as Collier-Reed and Ingerman (2013, p.244) state “it is important to recognise that the outcome of an analysis is firmly located at the level of the collective, and that attributing it to an individual student is methodologically inappropriate”.
Experience can be sensitively influenced by the context, and the participants may have expressed distinctive meanings in different circumstances. The range of variation of all the participants was likely to have been involved in the range across each participant. Thus, the whole set of transcripts was able to represent a picture of the ways in which the students experienced a particular phenomenon at a specific time and in a specific context (Åkerlind et al., 2005). This is an important basis on which researchers claim to make a collective-level interpretation.
Phenomenographic studies often investigate a range of meanings in a particular group of people rather than revealing the different meanings expressed by one interviewee. Each transcript is explained based on the similarities and dissimilarities within the holistic set of transcripts, and none of them can be interpreted independently of the others (Åkerlind, 2005b). The essence of phenomenographic analysis lies in comparing and contrasting between different individual transcripts so as to identify the meaning and structural aspects and further generate the categories of description.
In addition, a collective-level analysis is related to the ultimate aim of phenomenographic research, which, as Åkerlind, Bowden and Green (2005, p.76) claim;
[…] is not to capture any particular individual’s understanding, but rather to capture the range of understandings across a particular group. In other words, the analysis goes across and between all of the interview transcripts so that the categories of description that are yielded reflect not individual meanings or conceptions, but rather conceptions from a pool of meanings. The interpretation is, thus, based on the interviews as a holistic group, not as a series of individual interviews.
It is the crucial aspects of the collective experience, rather than the details of individuals’ experience, that should be highlighted during the analysis (Åkerlind et al., 2005).
In practice, I stopped myself from indulging too much in analysing individual transcripts. The ‘pool of meanings’ approach ensured that the analysis was on the basis of contrasting between various transcripts. If some individual’s utterance seemed to be different and special, I would first ensure the meaning of it was faithfully understood and interpreted. Then I placed it in the ‘pool of meanings’ and compared it with others, rather than focusing on a personal story.
In addition to the data collection, Säljö’s (1997, p.177) criticism, that it is problematic for phenomenographic researchers to choose to consider the “utterances from individuals made in specific situations and with varying motives” as indicative of conceptions, could be further minimised here following the collective-level analysis. Adawi et al. (2001, pp.19-20) contend that
[Säljö's criticism] seems to confuse the individual and the collective levels, which leads to an understanding that a phenomenographic analysis is an analysis of individual pieces of data, where it is in fact an analysis of a set of pieces of data at the collective level. It is the whole of the data material, generally interviews, that goes to make up the pool of meaning with which the researcher engages to analyse structure and meaning, […] not as a set of individuals but as a deliberately varied and holistic sample of the population of interest.
As stated, the emphasis of phenomenography lies in the collective mind. Phenomenographic researchers must not indulge in an individual’s world too much, rather it is the collective level that analysis should be carried out. Sandberg (1997, p.206) notes that a conception “cannot be seen in its entirety in data obtained from a single individual, but only in data obtained from several individuals”, and each individual can only “express some important aspect of the particular conception”. Even though some interviewees might not be able to articulate their ways of experiencing, the ultimate aim is the variation of conceptions among the group of students (Cope, 2000; Smith, 2010).