3. METHODOLOGY
3.3 Data analysis and interpretation
3.3.1 The methodology of discourse analysis
The concept of discourse has been introduced in the previous chapter, while this section will discuss the methodology of carrying out a discourse analysis. There are many branches of discourse analysis, but the one deemed most appropriate for geographical studies of this nature is poststructuralist discourse analysis, often associated with Foucault. For Foucault, discourse refers more broadly to the totality of utterances, actions and events which constitute a given field or topic (Wylie, 2006).
He refers to them as the rules and structures that underpin and govern statements that are produced, that these statements have common elements which has a unifying effect, and that they indeed have a real effect on the world (Waitt, 2010). This may be related to the concept of ‘policy discourses’ and the hegemonic depth and extent of a discourse, discussed in the previous chapter. Dittmer (2010) describes poststructuralist discourse analysis as using a more normatively neutral perspective of ideologies, this when compared to other branches such as Marxist discourse analysis, and I deemed this as being important in an academic research project of this type.
Using discourse analysis, interpreting data is a two-stage process, often described as double hermeneutic, as the researcher must interpret an informant’s interpretation of reality (Aase and Fossåskaret, 2007). If we agree that all data is produced or constructed, then it is vital also when analysing and interpreting interviews and texts to be aware of our positionality. One does not, therefore, provide a ‘truth’ at the end of the research, but rather a situated reading of life’s phenomena (Dittmer, 2010)
Foucault feared that a methodological template would become too formulaic and reductionist, and so discourse analysis is often thought to be better carried out somewhat intuitively (Waitt, 2010). There are, however, guidelines developed in the field of social science, and geography specifically, which I have chosen to use in this study. Fairclough (2003) describes three prongs of discourse analysis as, firstly, the text itself, related to the rhetoric used. Secondly, the ‘discursive practice’ including the immediate context in which the discourse is employed, or its production, distribution and consumption, and thirdly ‘social practice’ including larger ideologies in which the social context is located.
When analysing the text itself, there are many tools within the fields of rhetoric and semiotics which are of help. Aase and Fossåskaret (2007) provide a detailed toolbox for carrying out text analysis including the analysis of symbols with attached connotations which are used to describe things, and the use and analysis of structural metaphors. An example of the use of these tools is the following excerpt, from an article in the Financial Times newspaper about Norwegian petroleum aid.
‘But the Norwegians have little to show for their efforts around the world so far. São Tomé, which sought Norway's help in 2005, has been criticized for its shaky and opaque start to awarding its oil fields and Nigeria's situation has worsened’ (Hoyos, 2006).
This may be compared to Erik Solheim’s introduction in the annual Oil for Development report from 2010.
‘The programme takes a broad approach to petroleum sector management, covering the management of resources as well as of revenues and the environment. All our efforts are informed by principles of good governance such as transparency and accountability.’ (Norad, 2011).
Taking the word ‘opaque’ as a symbol with a negative connotation and as the dichotomy of ‘transparent’, and similarly ‘shaky’ as a symbol with a negative connotation which may be viewed in contrast to the positive connotation of the symbol, ‘accountable’, one can begin to analyse the texts and gain more understanding of the different discourses and narratives used about Norway’s petroleum aid.
As part of ‘discourse practice’, identifying the status of the author or respondent (stakeholder analysis) is an important part of the process and the situating of the data according to their status (Aase and Fossåskaret, 2007). This also includes an analysis of the intended use of the discourse, its audience and its circulation. In the example given above, one could comment on the statuses of the authors respectively, as female and male, journalist and politician, British and Norwegian, and how this affects the rhetoric in their texts or responses.
When moving from text analysis to broader social context, the challenge is to connect the data set to the broader realm of geographical practice (Dittmer, 2010). Aase and Fossåskaret (2007) suggest that the context an individual places a phenomenon in can also be described as which cognitive categories they place an observation in, and how they relate these to other categories. They state that data is an observation plus a category or concept and by overlooking this contextual knowledge, which can be both historical and cultural, the meaning of an individual’s comment or text can be distorted or, at worst, absent. Through the analysis, I have attempted to gain knowledge of such local contexts or categories through the texts and interviews.
The concepts and theories discussed in Chapter 2 also provide tools as to how to make the ‘jump’ from the text and discourse practice to social practice, this being the third branch of Fairclough’s system. Using a political ecology perspective, this includes how discourses may become sedimented and reproduced, becoming ‘policy discourses’ and how these reflect larger ideologies. By including the wider discussions surrounding the resource curse, development aid, and environmental management, it is possible to locate the social practice at different scales within which the discourses are located.
3.3.2 Transcribing and translation of the interviews
The interviews were transcribed in full from the audio recordings, including pauses, emphasized words and phrases, laughter and other behaviour which was taken note of, and could be used in identifying and describing a discourse. The transcribing process was useful as I became aware of certain details that were not noted during the interviews. The transcriptions were then translated from Norwegian to English. I carried out the translations myself, being fluent in both languages and, again, the process was useful in becoming familiar with the texts. Translation was challenging at times due to a number of Norwegian idioms used in the interviews. In the text excerpts chosen for the analysis, however, I am satisfied that the texts have not lost any significant detail from their original meaning through the translation process. In addition, during the interviews, I used or repeated some of the key words and phrases from the OfD programme such as ‘capacity building’ and ‘good governance’ in English so as to be certain of the intended meaning.
3.3.3 Coding
Having transcribed and translated the interviews, I had become very familiar with the texts, but used a system of coding, which is the process of identifying and organising themes in qualitative data (Cope, 2010). This aided in finding and categorising excerpts that related to different aspects of the research. There were some areas that I had previously deemed important, but which were later regarded as less relevant having analysed the extent of the material. This included some of the more technical environmental monitoring data. Other aspects emerged as very central to the research questions, often through the recurrence of certain codes, some of these not having been previously considered. The text was marked with the chosen codes in order to aid in analysis and sorting. These codes, firstly, represented a particular theme such as
‘environmental impact of oil’ or ‘corruption’ and, secondly, frequently repeated expressions such as ‘personal opinion’ or ‘assistance’, or often repeated topics, in order to comment on trends within the texts. Inconsistencies were also coded.
3.3.4 Choosing texts and excerpts
A diversity of texts were chosen for the study because they contributed to identifying the different discourses, this being purposive sampling. Text excerpts that contributed to identifying and describing the motives behind these different discourses were also highlighted, in addition to excerpts that related to a specific theme deemed relevant to the research. These texts are from both primary and secondary sources, since primary data production was limited to Norway and I was reliant on Internet sources to identify further discourses.
The study has included several different types of texts and genres based on the concept of intertextuality, the assumption that meanings are produced as a series of relationships between texts, rather than residing within the text itself (Waitt, 2010).
Intertextuality includes the presence of elements of other texts within a text, or quotations (Fairclough, 2003). This has provided the opportunity of interpreting discourses which were not obtained through primary data production, since the texts which were analysed contained aspects of other texts. It is important to be aware of the genre of a text using this method, since the producer of each genre is addressing a particular audience and the background of the author, text and place must be researched to form an interpretation (Waitt, 2010).