Research Methods
Three clusters of coded extracts, four candidate statements each. One names the subject matter, one makes an analytic claim, one is a useful staging post, and one asks the data to carry something it cannot.
Fictional — the same invented extracts used across this module's qualitative tools
By the end you should be able to tell a topic summary from a developed theme, and name the central organising concept a theme needs.
About 20 minutes. Nothing you do here is saved or sent anywhere.
Early in an analysis, grouping extracts under "asking for help" is exactly the right thing to do. It becomes a problem only when it appears in a results section as though it were a finding.
The cluster stays at the top of the stage while you work, so you can check each candidate against the codes and quotes it is supposed to organise.
A theme has a central organising concept: one idea that explains why these extracts belong together and what they show.
A student's results section has a section headed "Communication", containing extracts about unread emails, template replies and information passed between flatmates.
"Asking for help" tells a reader what the extracts are about. "Visibility is the price: the first question converts a private gap into a public one" tells them what the analysis found. The second can be disagreed with, which is the test — a statement nobody could argue with is usually not yet a claim.
A theme is built around one idea, and that idea is what explains why these particular extracts belong together. If you can move an extract out of a theme without the theme changing, the theme is probably a bucket. If several extracts each show a different face of one concept, you have something.
Grouping extracts by what they mention is how most analyses start, and it is how you find out what you have. The failure is publishing the staging post: a results section whose headings are nouns is usually a report of what was talked about rather than an analysis of it.
Three candidates in this tool fail for reasons that have nothing to do with abstraction. One asserts a cause that an interview study cannot establish. One attributes an intention to an institution nobody interviewed. One imports a psychological construct and a between-person comparison that qualitative data are not built to deliver. Reaching further is not the same as reaching outside.