Grounded theory was developed by Barney Glaser and Anselm Strauss in 1967, partly as a reaction against ‘Grand Theory’ – the very abstract, conceptual theories used in sociology at the time. There are technically three different versions of the theory as Glaser and Strauss later diverged in their views about forcing or emerging theory; the original 1967, a 1978 and a 1992 version. However the essence is still the same – a theory grounded in the behaviour, words and actions of those under study.
Grounded theory research enters the worlds of respondents to observe the environment, interactions and interpretations that people make. It’s a systematic and rigorous approach to collecting and analysing qualitative data to enhance our understanding of a social or psychological phenomena. It is meant to be explanatory rather than descriptive.
The theory that emerges is a set of relationships that offer a plausible explanation of the phenomenon under study. (Strauss & Corbin, 1994). The theory is developed by constructing alternative explanations until a best fit is developed – best fit being the simplest model that links as many of the diverse findings as possible in a useful and pragmatic way. Grounded theory is best used for topics of interest where relatively little is already known, as it enables the researcher to build their own theory from the ground up.
The Process of discovering the theory
Raising Generative Questions before the fieldwork?
Grounded theory is sometimes presented as requiring a blank slate to start with, but according to Goulding (1999) this is a misconception. Familiarity with previous theories helps with pattern recognition and conceptual leverage – however if the subject area has been extensively researched and there is already a large body of knowledge it may be hard to use Grounded theory as the researcher will be full of preconceptions and may end up testing existing work rather than developing new ideas. Strauss and Corbin (1990) argue it is best not to conduct a literature review beforehand but that the researcher should analyse their own feelings and prejudices.
If generative questions are raised during the setting up of the project they remain open to change.
Multi-method research – a generous definition of data
Grounded theory may be based on single or multiple sources of data – interviews, observations, focus groups, case studies, experiential research but the methods need to remain relatively unstructured. The more structured the method the more the preconceptions of the researcher can influence the data. (However respondents need some structure to know what to talk about). ’All is data’ – newspaper articles, conversations, TV shows, can also be included as the subject of study if they are relevant.
The different directions of Glaser and Strauss
Glaser’s current approach, which can be found at www.groundedtheory.com, emphasises no pre research literature review, no taping and no talk about the findings during the writing up. The idea is to aid inductive thinking and the emergence of theory as much as possible using preconscious processing. In this case the data that is analysed consists of the researcher’s own notes (memos). Glaser also emphasises theoretical sensitivity, which is the researchers ability to understand the meaning and sensitivity of data. Glaser does use 18 theoretical coding families, while Strauss (now dead but whose ideas live on in Strauss and Corbin) is described as forcing theory through the rigorous coding process described below.
Continuous interplay between collection and analysis. You analyse during data collection.
In grounded theory analysis is iterative – it starts as soon as there is some data and it may be that you choose to collect other cases or data from different people as the analysis develops. For example if the study is about encouraging children to do homework and parents and teachers are first interviewed, it may emerge that older siblings are influential and may need to be added to the sample. This is called ‘theoretical sampling’, where the sample is driven by the emerging theory. Data is collected until the point of saturation – where no more is being learnt.
Different types of coding (ways of thinking to help you develop the theory)
Note that these are not just coding as labeling – they are interpretative processes.
- Open coding to develop categories (being OPEN to ideas)
- Axial coding to connect the categories in new ways
- Selective coding – to build the story
Example is from Graham R Gibbs, University of Huddersfield on YouTube
Open coding breaks down the data into units of meaning – identifying the concepts the data is related to. For example “I’m not very good at maths” might be coded as “neg self-attribution /maths”. The idea is to begin developing concepts at the first level of abstraction. Open coding should avoid restating what respondents said and go up to motivations and intentions.
According to O’Callaghan (1996) the key questions to be asking during open coding are:
- What is happening in this data?
- What is the basic (socio-psychological) problem? (It started with sociology but can be applied to many other types of problems)
- What accounts for it?
- What patterns are occurring here?
Or simply: who (are the relevant actors), what (does it mean to the people involved), why (what are the motivations), where (how does the environment affect what happens) when, how? Strauss and Corbin also suggest using ‘what if’ analogy questions? What if the weight lifter was a violinist? What would they need to know or do? What does the weight lifter do that a violinist also needs to do? These are examples of questions that help theoretical sensitivity, help pull out the implications of the data, and there are many others.
Codes are refined through constant comparison. Take a situation and the codes applied to it, and compare with a similar situation. Most likely the codes will have to change or evolve in order to include both the situations. Then retrieve all the text has the same codes, and refine the codes further (or create new ones) to fit that text. At some point there will be saturation – no further changes, variations, illuminations can be made, and then it’s time to stop. It’s a very thorough process and is designed to increase validity.
At the same time the researcher does memoing (notes to self, a ‘thought dump’) to capture thoughts and insights that arise during the coding process. (E.g. Idea: does perception of being good at a subject influence willingness to do homework?)
There may well be some visual display in order to see how codes are related to each other, because closely related codes become a category. The codes should relate to the same phenomenon, and then the category needs to be named. This stage is quite likely to bring in theoretical ideas, as you will be referring to social or psychological phenomena. Strauss & Corbin suggest it’s useful to use words or concepts from the participants in the naming.
Codes and categories can be dimensionalised – they have a range of properties. These are explored by the researcher, to see if they can be applied elsewhere. For example the concept of Attention, might include the properties of focus vs. distraction, short versus long spans, selective versus general, watching, listening, etc. (These might be called sub-codes).
Axial coding is about the relationships between codes/concepts and is the basis of the theory. Examples are: how is code A related to Code B? Does one maybe cause another? Are there correlations or intervening variables? What influences the central phenomena – if you have one. What kind of strategies are being used, what consequences may there be? The coding paradigm can be a visual representation of the model as it is being built and often includes actions, interactions and consequences.
Selective coding is aiming to identify a single category as the Central Phenomenon in order to build a story. This can be very hard, as there may be lots of important categories, especially as the core category (or at the most 2 or 3) should account for a large proportion of the behaviour observed and be based on recurring themes from the data. This is then put into a narrative that forms the report of the research.
Later versions of grounded theory suggest it’s important to have an Integrative diagram and/or to work in a group session where different members of the research team are able to interact and share ideas to increase insight. The one at left shows using the process of grounded theory in research on how the Internet has affected the correspondence art network.
Some also suggest there should be verification. One of the drawbacks of grounded theory is that it can result in theories with a high level of internal logic and validity, but it is not clear to what extent and for how long, the theory can be applied to other situations. According to Glaser and Strauss, how good a theory is can be judged by
- Closeness of fit between theory and data
- How understandable it is (by a lay person working in the field)
- How it can generalise to diverse situations
- Whether it can allow some control or change of the situations
Cohen, L, Manion, L, Morrison, K, (2007) Research Methods in Education, Routledge.
Glaser, B., & Strauss, A. (1967) The Discovery of Grounded Theory: strategies for qualitative research Chicago: Aldine
Goulding, C, (1999) Working Paper Series: Some reflections on paradigm, theory and misconceptions. Wolverhampton Business School Management Research Centre ISSN 1363-6839
honoria madelyn starbuck dissertation: Effects of the Internet on the Correspondence Art Network, Chapter 2: Methods of collage of data using grounded theory http://www.mailartist.com/honoria/research2003/methods.htm
Schreiber, J, Asner-Self, K, (2011) Educational Research, John Wiley & Sons.
Strauss, A. & Corbin, J. (1990) Basics of Qualitative Research: grounded theory procedures and techniques London: Sage.