Participant-Led Construct Discovery
Let participants define the phenomenon before you build the measurement.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 93%
Participant-Led Construct Discovery begins before the questionnaire. Researchers first ask people how the phenomenon should be investigated and invite free descriptions of differences, locations, intensities, contexts, or associated feelings. They then look for recurring dimensions while deliberately withholding inherited labels that could force responses into old theories. Those dimensions become neutral questionnaire items, and analysis determines whether responses actually cluster into distinguishable types. The output is not simply a list of quotes; it is a testable measurement model grounded in participants' experience. The framework is especially valuable when experts disagree about basic categories, because it separates discovery from confirmation and makes the assumptions embedded in labels visible.
Origin
King's team began by asking women what scientists should ask about orgasm. Participants described different experiences using dimensions such as location, intensity, and associated feelings. The team then built a questionnaire while avoiding the historically loaded clitoral-versus-vaginal split and analyzed the responses for emergent categories.
Core principles
- 01Participants may experience distinctions that established categories erase.
- 02Open-ended discovery should precede forced-choice measurement.
- 03Loaded labels can reproduce the history and assumptions a study is meant to test.
- 04Categories should emerge from patterns in descriptions, not from researcher preference.
How to run it
- 1
Open the Question
Ask participants what distinctions matter and what researchers routinely miss. Use prompts broad enough to surface unexpected categories.
Pro tip Invite participants to correct the premise of the question itself.
Watch out Do not begin by asking respondents to choose among the categories you hope to prove.
- 2
Collect Experience Dimensions
Code descriptions for recurring dimensions such as location, intensity, sequence, context, and associated feelings. Preserve minority descriptions instead of averaging them away too early.
Pro tip Separate dimensions from labels so several experiences can share one feature.
Watch out Frequency alone does not determine whether a distinction is meaningful.
- 3
Audit Loaded Language
Trace proposed labels back to the theories, history, or status judgments they carry. Replace them with neutral descriptions of observable or reportable features.
Pro tip Ask what answer each label quietly encourages.
Watch out A familiar term may smuggle a disputed mechanism into the measurement.
- 4
Build the Instrument
Turn the discovered dimensions into questionnaire items or interview prompts. Keep wording descriptive and allow combinations that the original categories would have excluded.
Pro tip Pilot the instrument with participants who were not involved in the discovery interviews.
Watch out Do not collapse dimensions into types before collecting enough structured responses.
- 5
Test Emergent Categories
Analyze whether response patterns form distinct, interpretable clusters. Treat the categories as provisional and validate them against new samples or independent measures.
Pro tip Name categories only after their defining dimensions are clear.
Watch out Statistical separation does not automatically establish a biological or causal difference.
In the wild
King's team first invited women to say what scientists should ask and heard that the singular category of orgasm missed meaningful variation. Participants described experiences by location, intensity, and associated feelings. The researchers avoided the inherited clitoral-versus-vaginal split, built a more neutral questionnaire, and found two broad response patterns described as surface and deep.
→ The measurement reflected participant-reported distinctions instead of forcing experiences into a historical binary.
A subscription company stops asking customers to choose only between price, product, and support. It first asks former customers what the survey should have asked, codes recurring dimensions, removes internal jargon, and builds a neutral instrument around the discovered reasons. This is an illustrative application of the episode's method.
→ The team uncovers a previously hidden category: customers left because the product created coordination work for colleagues.
Common mistakes
Starting With the Inherited Binary
A disputed historical split can predetermine what respondents are able to report.
Naming Clusters Too Early
Premature labels encourage researchers to interpret ambiguous responses as confirmation.
Is it for you?
Best for
It is best for poorly understood experiences, emerging markets, and domains burdened by contested terminology.
Not ideal for
It is not ideal when the construct and measurement instrument are already mature, validated, and uncontroversial.
From the transcript
“we we just sort of really um just opened it up uh to ask women what you what should scientists be asking them”
“they come up with various kinds of descriptors of of location intensity uh and associated feelings”
“we're very very careful to stay away from those things for a whole bunch of reasons”
From the episode
Why Does The Female Orgasm Exist? - Dr Robert King - #977
Dr Robert King