The term ‘template analysis’ refers to a particular way of thematically analysing qualitative data. The data involved is usually in the form of interview transcripts, however this may be any kind of textual data, including diary entries, text from electronic interviews or open-ended question responses on a written questionnaire.
Template analysis involves the development of a coding ‘template’, which summarises themes identified by the researcher(s) as important in a data set, and organises them in a meaningful and useful manner. Hierarchical coding is emphasised, using broad themes such as ‘responses to illness’, encompassing successively narrower, more specific ones including ‘changed relationships’ and ‘changed relationships with health professionals’.
Analysis often starts with some a priori codes, which identify themes strongly expected to be relevant to the analysis. However, these codes may be modified or dispensed with if they do not prove to be useful or appropriate to the actual data examined. Once any a priori themes are defined, the first step of the analysis is to begin reading through the data, marking in some way any segments that appear to tell the researcher something of relevance to the research question(s). Where such segments correspond to a priori themes, they are coded as such. Otherwise, new themes are defined to include the relevant material and organised into an initial template, which is normally undertaken after initial coding of a sub-set of the data, for example, after reading through and coding the first three of 15 transcripts in a study. This initial template is then applied to the whole data set and modified in the light of careful consideration of each transcript. Once a final version is defined, and all transcripts have been coded to it, the template serves as the basis for the researcher’s interpretation or illumination of the data set and the writing up of findings.
Before carrying out any research project, the researcher should have a clear idea of the philosophical position the work is coming from. In most quantitative research, however, the assumptions and prescriptions of the hypotetico-deductive approach are taken for granted, as are their implications for research design and methods of data analysis. A psychologist writing up an experimental study is unlikely to feel the need to philosophically justify the use of a control group or the appropriateness of using confirmatory statistical tests to make judgements about causality. In contrast, qualitative research encompasses a range of philosophical positions that have differing implications for how research should be carried out and how data should be analysed. It is therefore important for qualitative researchers to carefully consider the philosophical stance of their work.
In order to make clear why template analysis should or should not be used in particular ways, below is presented a short discussion on three key terms: epistemology, methodology and method.
Epistemology refers to the assumptions we make about what it is possible for us to know and how we can obtain this knowledge. A key area of epistemological debate in qualitative research is the extent to which any method can provide access to the personal world of the research participant.
Methodology refers to the general approach taken to carrying out a piece of research. Analytical induction (Johnson, 1998), Interpretive Phenomenological Analysis (Smith, 1996) and discursive psychology (Willig, 2001) may all be seen as methodologies.
Method refers to the particular techniques used to collect and analyse data. Narrative interviews, semi-structured interviews, repertory grids, participant observation and the use of pre-existing texts can all be used as qualitative methods of data collection.
In terms of how these concepts relate to each other, we can see a hierarchy whereby epistemological assumptions shape methodological approaches, which in turn inform the choice of method. The relationships are not wholly deterministic, especially not that between methodology and method. For example, a relativist epistemological position underlies several methodological approaches including discursive psychology, Foucauldian discourse analysis and Q-methodology, which might use a number of different methods of data collection, ie interviews, recordings of ‘natural’ conversations and analysis of pre-existing texts.
The method choice of data analysis needs to be guided by the methodological position of a piece of research and its underlying epistemological assumptions. Some analytical methods are directly linked to particular methodologies, thus if you are using classical grounded theory you will necessarily employ the sequence of open and axial coding it specifies. Other methods may be used within a variety of methodologies, but precisely how they are used will differ from one to another. Template analysis falls into this latter category, it can be employed to analyse any form of textual data from many but not all methodological and epistemological positions. It can be used in research that takes a similar realist position to mainstream quantitative psychology, in seeking to ‘discover’ the underlying causes of particular human phenomena. In contrast it can also be used within a ‘contextual constructivist’ position (Madill et al, 2000), which assumes there are always multiple interpretations to be made of any phenomenon, and that these depend upon the position of the researcher and the specific social context of the research.
If you were using template analysis to analyse data from a realist-oriented study, you would need to be concerned with the reliability of the coding process, visit the technique: quality checks and reflexivity for further information. Also, it is likely that you would include quite a few a priori codes, reflecting areas highlighted in advance as important to addressing your research question, visit the technique: themes and codes section for further information. If you had taken a contextual constructivist position, you would want to find ways of examining issues of reflexivity and the nature of the researcher-participant relationship. You would also need to use the technique in a flexible way, that helped you to consider multiple interpretations of the data, and discouraged you from closing down to one ‘best’ reading too early.