Use open probes rather than leading paraphrases
Aliases: open probe · leading paraphrase · moderator neutrality
What it is
An open probe asks participants to supply meaning, reasons, or examples. A leading paraphrase adds the moderator's evaluation or causal account while appearing to confirm understanding—for example, converting “I paused” into “so the button was confusing.” Paraphrase is not inherently wrong; the risk is inserting an explanation that the participant did not provide.
Why it happens
Participants often treat a moderator's summary as a cue to the expected interpretation. Directional paraphrases narrow the response space, organize later talk around the researcher's hypothesis, and can turn a polite “yes” into apparent evidence. Clarification first, followed by a tentative summary in the participant's language and an explicit chance to correct it, leaves more interpretive authority with the speaker.
Studying it
Code question type, concepts newly introduced by the moderator, participant corrections, and concrete evidence added after probes in pilot and study recordings. Trace important claims to who first introduced them; themes appearing only after directional prompts deserve weaker evidential weight. Comparisons among moderators should also account for guide position and participant composition.
Where it stops holding
Open grammar is not automatically neutral: “Why didn't you…” presupposes an action, while a broad “anything else?” may produce little detail. Closed confirmation can be necessary for safety events, timelines, or terminology, provided it confirms an already stated fact rather than supplies a motive. Power differences may still inhibit correction of a paraphrase.
Applying it
- Ask what happened and what a participant means by a term before proposing an explanation.
- Reuse participant vocabulary and frame summaries as explicitly correctable.
- Put hypotheses in moderator notes rather than questions, then seek support and counterexamples.
- Audit the prompt that produced each major claim and downgrade evidence elicited by leading prompts.