CHAPTER 5: RESEARCH METHOD CONTENT ANALYSIS
5.3 Approaches to content analysis
Beck et al. (2010) classified approaches to content analysis under two broad categories: mechanistic (or traditional) and interpretative. The mechanistic approach is structure orientated. It tends to evaluate data by identifying the presence or frequencies or description length of some predetermined words or concepts. This approach basically involves routine word counts (Deegan and Rankin, 1996; Campbell, 2003), sentence counts (Milne and Adler, 1999; Patten and Crampton, 2004), page proportion counts (Unerman, 2000) or frequency of disclosure (Ness and Mirza, 1991). Each has its own advantages and limitations. The mechanistic approach has been the dominant approach in most of the prior environmental research as its pure quantification character is argued to provide more valid and reliable results. However, it is also argued that this approach fails to satisfy any explanatory query within a hermeneutic-interpretative study, where the main focus is to identify the underlying meaning or message that a text is intended to convey (Beck et al., 2010). Instead, the approach is appropriate when the research question is more about the amount (high/low) of specific disclosure and tends to examine relationships among different variables with the disclosure pattern.
The interpretative approach, by contrast, is similar to the hermeneutic-classificatory content analysis discussed in Bos and Tarnai (1999). This is more meaning-orientated and focuses on the underlying theme of the text under investigation (Beck et al., 2010). This approach requires an understanding of both denotative (manifest) and connotative (latent) meanings of the text under study. Denotative meaning refers to the obvious meaning of a text or fact; whereas, connotative meaning refers to the inner meaning of a text as a whole through combining the understanding of its individual elements (Ahuvia, 2001). While traditionally,
analysis of the connotative meaning or latent content refers to the interpretative approach, Ahuvia (2001) stressed that because an act of interpretation means assigning meaning to abstract symbols, denotative meaning would also constitute interpretation. In his words:
Denotative interpretations are so highly conventional and frequently practiced that we often create them without being aware that we are performing an interpretative act. This can create the illusion that the denotative meanings we perceive are parts of the physical text itself not interpretations (Ahuvia 2001, p. 142).
In general, the interpretative approach tends to capture the meaning of the text by first creating categories that share a commonality based on specific themes and then explaining the text contents under the relevant category (Graneheim and Lundman, 2004). The connotative interpretative approach specifically requires the collaborative conduct of coding by researchers who have high levels of expertise in terms of ‘theoretical sensitivity’and an awareness of the ‘subtleties of meaning of the data’ (Ahuvia, 2001, p. 144). Under this approach, researchers would define criteria before analysing texts, but would not create a fixed set of coding rules to operationalise that definition. Instead, they would examine a text within the context of its preparation and perception to decide whether it could be categorised according to the definition.
The interpretative approach is criticised for its fundamental subjectivity as it largely depends on the researchers’ preference to categorise the theme and to elicit the underlying meaning of the narrative. As the approach does not employ the use of rating or mutually exclusive categories, it leaves the data analysis with insufficient rigour and reduces the validity and the reliability inherent in the method’s design (Boettger and Palmer, 2010).
However, such criticism was countered by Feldman (1995) who argued that the meaning of every text is context-dependent. For example, to interpret the meaning of the text: ‘What do you mean?’ it is required to know in which context the text is produced. Based on the related context, the text can be perceived to be an innocent query; alternatively, it could be an angry response. As Feldman (1995, p. 11) stated:
Because every context is unique and contexts are constantly emerging, there cannot be a set of pre-existing rules that are waiting to be followed.
Therefore, the fundamental requirement of having a specific set of coding rules in other approaches of content analysis (such as the mechanistic approach) is strictly contra- indicated in the connotative interpretative approach. As it assumes that while the mechanist approach attempts to fit the texts under study against some pre-determined rules to increase the inter-rater reliability, it may do so at the expense of losing the appropriate meaning of a particular text. In connotative content analysis, inter-rater reliability is substituted by ‘public justifiability’. As is explained by Ahuvia (2001, p. 146):
To achieve public justifiability, the researchers include the focal texts, their codings, and, if necessary, a justification of their codings along with the manuscript when submitting it for publication. In this way, the quality of their coding can be directly assessed by the reviewers.
There is no simple way to decide which approach is the best to conduct content analysis; rather, it depends on the complexity of the tasks associated with the research in question. Where the research questions seek to find out the quantity or pattern of specific disclosures based on straightforward interpretation and where the research materials involve large amounts of secondary data, the quantitative approach would be the best option to follow. On the other hand, where the objective is to gain greater understanding of what is communicated through a text and how or where the research involves qualitative interview data, the interpretative approach would be the best option.