Notes on Craft: Learning to swim in the realist ocean
- rabrams0
- Aug 6
- 3 min read
August 2026
First of all, a huge thank you to the lovely feedback we’ve received so far on this initiative and our first blog. It was perfectly described by a colleague the other day as, ‘all the bits that you wouldn’t see in a paper’. This is exactly what we wanted to convey- it’s those clarifying back and forths, and we’re really pleased to hear it’s landed well.
Now, to our August conversation. This month has taken us through the murky realm of contexts-mechanisms-outcomes and the initial differences between these concepts. This is a conversation Alankrita and I had prior to any data extraction or coding of the literature. We needed to get on the same page about what we were looking for, how and why (!). It's also a conversation we will return to again during our data extraction phrase, but for now...
The following are the key points that helped keep us focused:
Outcomes are usually the easiest elements to identify, as studies tend to describe them explicitly.
Start by identifying stated outcomes in the evidence, then work backwards to consider what contexts and mechanisms may plausibly explain those outcomes.
It's ok to abstract the outcomes, but only if this abstraction is an accurate reflection of several papers that can be clustered together, with similar or the same outcomes.
Contexts are conditions that are necessary for the identified outcomes to materialise, and these would also be somewhat concrete and identifiable, like outcomes.
Mechanisms* are rarely, if ever, explicit in the data and therefore require a degree of qualitative inference/ retroduction. This is the really creative component of realist work, whilst also staying true to the evidence.
Inferring mechanisms involves looking for latent meaning, subtext, and underlying processes rather than surface‑level descriptions. It happens iteratively, as you move through the data. This process is analogous to qualitative coding, where mechanisms develop through interpretation rather than direct extraction.
Maintain a running document/ log alongside literature review and data extraction to record emerging ideas, patterns, early thoughts about contexts, mechanisms and configurations.
Use the log for thinking aloud, capturing inspirations and analytic hunches that arise while reading papers or extracting data.
Clearly note which paper or papers triggered each idea to allow later cross‑referencing.
Treat this log as part of the analytic process rather than a polished output; entries may be provisional and revisited or not used later.
Retain original terminology where appropriate because sometimes, simplification risks misrepresenting the study.
Expect variation across studies; do not force uniformity where it does not fit the evidence.
Alankrita's reflections
Sitting down with a large batch of included papers for this realist review, just ahead of this conversation, my honest thought was: this feels like a huge ocean, and I don't know how to swim. The outcomes-first approach simplified the first step for me, which often feels daunting. However, what most relieved the anxiety was the explicit permission to be tentative and unsure; to just keep logging my thoughts and reflections. The log was meant to be a tool of analysis rather than a polished output. The pressure to be precise or "right" about what I was seeing in a paper was taken away.
*Some realists consider mechanisms as the articulation of resources and responses to that resource (see Dalkin et al., 2015 here for further information). Whilst this is a legitimate positioning, it is not how we position our understanding of mechanisms, which instead is more aligned to how Mukumbang and Wong, 2025 describe them).
If citing this, please do so as: Abrams, R and Singh A. (2026) Notes on Craft: Learning to swim in the realist ocean. August 2026.
Comments