Theories are a central part of RS. Perhaps a simplfied way of thinking about their role may be that MECHANISMS address the 'How' question, whereas THEORY provides the 'why', in particular explaining the relationship between CMOs.
WHICH, WHERE, HOW etc.
Whilst it might be that in primary realist research (realist evaluation) identifying relevant theories can prove to be a challenge, in realist synthesis (RS), the opposite seems to be true.
The proposition was put forward that there is no such thing as a new theory, just variations on a theme and some of you may know the 'joke' about there being only three types of interventions - carrots, sticks and sermons. The more serious point is that theories do exist, but often need 'refinement' to make them useful. So, whilst a theory may relate to a specific mechanism, our job is to work out and explain what happens to this specific mechanism in the intervention type we are interested in.
By having 'distance' or an overview from the data, there may in fact be too many theories and knowing which one to choose, why and at what level of abstraction can be a daunting task. The point made was that we can look for theory from a whole range of disciplines and what we are trying to explain may be related to the type of question that we are trying to answer. Perhaps a number of pointers may help on this front:
- it may seen obvious but it maybe worth remembering what the point of the theory is- to help us make sense of the patterns we can see in the data (well at least some of these). So the first reason for choosing any theory would be because we suspect that it maybe able to help us make sense of the data?
(There is an interesting discussion about this intial 'leap' (or retroduction) in the thread 'Interim summary - how much should we impugn...')
- it may not matter which theory we start off with, hence some of you will recall the term 'candidate theory'. The point is we have to start somewhere BUT then have to test our 'chosen' theory. There may be many false starts and the frustration maybe more with how many 'candidates' we have to reject before we get the right one :-(
LEVEL OF ABSTRACTTION
Explaining patterns in the data may require iterative searching for theory which may lie at different levels of abstraction. So in some cases a set of theories you have found may be 'subsumed' by a theory that is more abstract or even a different theory.
There is a potential tension here between wanting to explain the patterns in the data which may require more and more abstract theory and the need to be able to test that theory (in effect the theory needs to be middle-range) [There is a thread on middle-range theory though that asks what exactly does it mean for a theory to be middle-range].
Perhaps the overall message with theory is that the process is iterative and the search continues until theory that is the 'current best available that is testable within the time and resources we have' is found? Whatever we come up with will always potentially be superseded and the theory just needs to be coherent and plausible enough for now?
The above may just nicely tie into the process that keeps us reviewers 'honest' :-)
Through argument, we have to defend our the coherent and plausible theories that we believe explain (at least some of) the patterns in our data. This may happen (for example) within a review team and with other academics. Though as it was pointed out, there are 'good' arguments and 'bad' ones. Why the latter may happen is given an airing in Friedman's designed blindness paper (see Interim summary - designed blidness).