Attach the questions as you will ask them, treat any recording as identifiable, describe the whole path a recording travels and where it ends, and write the confidentiality limits truthfully: a group cannot be promised secrecy, and legal reporting duties can override your intentions.
Why is a talking study read so closely?
Because almost nothing about it is fixed on paper unless you fix it. A survey exists as a file before anyone answers; an interview exists only as a plan until it happens. The protection you promise has to hold in a conversation nobody else will witness, so the documents must be specific enough to check afterwards. Which is why "semi-structured" is no defence for a vague guide: structure that stays flexible in the room can still be complete on the page.
Recording sharpens all of it. Voice prints appear among the identifiers HIPAA lists at 164.514(b)(2), and the Common Rule's test at 46.102(e)(5) asks whether identity may readily be ascertained. An audio file of a colleague discussing her own unit satisfies both, before any transcript exists. So a qualitative file rarely rests on the claim that nothing identifiable was recorded; it rests on what becomes of the identifiable thing you deliberately made.
What are the four documents, and in what order?
-
The guide, written as it will be asked
List the opening framing, every planned question in the words you will use, the probes you expect to reach for, and how you close. If a question came from a published instrument or another project's guide, say so and secure written permission to use or adapt it. Then check: does every question belong to the aim you stated? A guide wandering into disciplinary history when the stated aim is handover has changed the project without saying so.
-
The recruitment text
Whatever a participant reads or hears first belongs in the file, the message forwarded to a distribution list and the words said in a corridor included. It must describe the topic accurately, promise no benefit, and make declining ordinary. Where those invited are colleagues, the pressures described in the piece on surveying staff apply here with more force, because an interview cannot be answered privately at home.
-
The consent
Built from the elements at 46.116(b), opening with the "key information" someone needs in order to weigh the ask, as 46.116(a)(5)(i) requires. The elements that carry a talking study: what happens and how long it lasts (b)(1); foreseeable risks (b)(2); how far confidentiality will hold, put as an extent rather than a promise (b)(5); whom to contact (b)(7); voluntariness, no penalty for refusing or stopping (b)(8); and what becomes of identifiable information afterwards (b)(9). Exculpatory language is barred outright by 46.116(a)(6).
-
The data path
Not a sentence about secure storage. A sequence: the device that records, the drive the file moves to, who else touches it, what a transcript strips out, what a pseudonym replaces, where any linking list lives, and the event after which the audio no longer exists. Boards approve strict and generous plans alike. What they return is a plan too vague to check.
Can you skip the signature?
Sometimes you should. Two situations allow a waiver under 46.117(c)(1). In the first, the consent form is itself the one record tying someone to the project and a confidentiality breach is the chief risk; each person is then asked whether they want documentation, and their preference governs. In the second, the work carries minimal risk and involves nothing ordinarily signed for outside research. For a telephone interview about workplace practice, that signature sheet may be the only document linking a named person to a sensitive conversation. Request the waiver explicitly, attaching the verbal script and information sheet. A short-form route exists at 46.117(b)(2) when elements are delivered aloud; it needs a witness and an approved summary.