Writing a book from interviews
A biography, an oral history, a company history, a family book, a piece of long-form reporting — all run on the same raw material. Somebody talks, and you write it down.
The writing down is the easy part. What decides whether the book is any good happens on either side of it: what you knew beforehand, what you agreed, what you asked, and what you did with the recording afterward.
This guide covers method. For putting identifiable people into a book — composites, changed names, and where a lawyer's territory begins — see Writing about real people. All of it applies here, and none of it is repeated below.
Before you sit down
An hour of someone's attention is the scarcest thing in this kind of book, and you get it once, maybe twice. Do not spend it on facts you could have looked up. Dates, job titles, the order of events, what has already been published — find that first. An interviewee who spends twenty minutes correcting your chronology has learned that you did not prepare, and people are markedly less generous with someone they have decided is unprepared. Preparation also earns better answers: "You left in 1987" gets a shrug, while "You left three months after the Dunbar contract fell through — was that connected?" gets the story.
Write twenty questions and expect to ask nine. The list is insurance, not a script: it stops a silence from panicking you into a bad question, and it makes sure the three things the book genuinely needs from this person get asked. Mark those three; the rest is optional. Then be willing to throw the order away. The best material in almost every interview comes from the thing the person got animated about, which is never the thing you planned to ask. Follow that instead — you can return to the list, but you cannot manufacture that energy twice.
Agree the ground rules first
Explicitly, out loud, in advance:
- Whether you are recording. Ask, get a yes, and then — the part people skip — get the yes in the recording. Start the device, state the date and who is present, and ask again on tape. Fifteen seconds, and it is the only version of the agreement that still exists in five years.
- On the record or not. Decide what "off the record" means to both of you before either of you uses the phrase. Not for publication at all, or usable without attribution? People mean both.
- Whether they see quotes before publication. Some interviewees will only talk on this condition; some journalists will never agree to it. Either answer is defensible. An unstated assumption is not.
- What happens if they later want something removed. The most avoided conversation and the one that causes the trouble. Settle it while everyone is comfortable.
What any of that obliges you to legally varies by country and by the kind of publication, and we will not tell you where your line is. What is universal is that the agreement be specific, mutual, and written down somewhere other than your memory.
In the room
Ask short questions. A long question is you thinking aloud, and it hands the interviewee a menu to pick the easiest item from. Ten words is plenty.
Let the silence sit. When they stop talking, wait. Three or four seconds is agonizing and it is where the good material lives — people fill a pause with the thing they were deciding whether to say. An interviewer who jumps in never hears it.
Ask concrete, not abstract. "What was the culture like?" produces a summary anyone could have written. "What did the room smell like?" produces the scene. Ask what was on the desk, what time they got home, what they ate.
Do not argue. If an answer contradicts what you know, note it and keep going. Check it later, on your own time.
In the tools. A book built from interviews has to show its sources. Back matter is built from fields, so a note on method, a list of interviews and dates, and acknowledgments to the people who talked to you take their place in the book without hand-formatting, and carry into the EPUB from the same source. Start your book → · about the tools
Transcribe, and keep the timestamps
Transcribe it, or pay to have it transcribed, but do not work from the audio alone. Six hours of recording is roughly sixty thousand words, and finding one sentence in that by scrubbing back and forth is miserable — you will need to find it when drafting, again when fact-checking, and again when someone queries the quote. Keep timestamps against every speaker turn, and put the person and the date in the filename. Treat the audio as the master and the transcript as an index into it, and back up both, per Backing up a manuscript, properly.
Tidying a quote, and the line you do not cross
Nobody speaks in sentences. A verbatim transcript of a fluent, intelligent person reads as barely literate: um, half-starts, doubled words, clauses that never land.
So quotes are tidied, and every reputable publication does it. Cutting filler, false starts and repetition is normal and expected, as is trimming the middle of a long answer where you mark the cut. Nobody is misled by that.
What you do not do is change what the person meant. Splicing two answers into one continuous quote, moving a clause onto a different subject, cutting the qualifier that made the claim defensible, tightening a hedge into a certainty — these produce a quote the person never said, and the fact that every word in it is theirs makes it worse, not better.
The test is the ethical core of this work: would the speaker, hearing the tidied version read back, recognize it as what they said? If you would be reluctant to play them the tape alongside it, you have crossed over.
Memory is confidently wrong
People misremember dates, sequences, who was present and who said what, and they do it with complete conviction. Confidence in an interview tells you nothing about accuracy. Check what you are told against something outside the room: documents, contemporaneous records, another interviewee. Where two accounts conflict and you cannot resolve them, say so in the text — a sentence naming the disagreement is more honest, and more interesting, than quietly picking one. See Fact-checking your own book.
Keep everything
Keep the recordings, the transcripts and the consent, and keep them after publication rather than until it. They are the only evidence of what was actually said. Store them with your other publishing records — see The records a self-publisher needs to keep — and decide deliberately how long you hold them, and whether anything was promised about that.
Make your book
The tools are free to use: name your book, bring your manuscript or your artwork, and work as long as you like. You only pay when you want the files without the preview watermark.
Next: Fact-checking your own book · More: Writing about real people · Quoting other people's work in your book · Writing by dictation