Media training for scientists is a translation problem, not a confidence problem. In a full day with twelve technical experts at a federal agency in Washington DC, the work was eight recorded practice rounds each, converting expert answers into plain language, and giving every person a rule for questions outside their own program.
Almost every media training I run gets booked because something happened. A story ran badly. A leader got clipped. A reporter is circling. This one got booked because of an org chart.
In July 2026 I spent two days in Washington DC with a federal agency. Day one was twelve technical experts, people whose working lives are spent inside genuinely complicated material. Nothing had gone wrong. Nobody was in trouble. What had changed was a reorganization.
One of the senior scientists in the room explained it to me, and I am paraphrasing him closely. For decades, he said, we knew we were not going to be the spokespeople. We are scientists. We do the work at staff level and somebody else talks about it. Then the reorganization moved a lot of us up, and the odds that one of us gets asked for an on-camera interview went up sharply.
Why does a reorganization create a media training problem?
Because reporters and producers do not call people. They call titles.
I spent 13 years producing at ABC, NBC, and Fox. When a booker needs an expert on a technical public-health risk by four o'clock, nobody is reading anyone's publication history. They are looking at a staff directory or a press office page, and they are looking for whoever now has the senior-sounding job.
Which means the moment you are promoted into a role, you inherit the call list attached to it. Nothing about you changed. Your risk profile did.
The agency's communications leadership had spotted this, which is why we were there. If your organization has restructured in the last year, that is the audit worth running: not who is a good communicator, but whose title now sits where a reporter will look.
What is actually hard for a scientist on camera?
The expectation, in almost every article about training technical people, is that scientists are terrified and awkward and need coaxing out of their shells.
That is not what I found. This room was composed. Several of them had testified, taught, or presented to hostile rooms of peers, which is arguably harder than a local news hit. Fear was not the bottleneck.
The bottleneck was translation, and it failed in two specific ways.
The precision reflex. A scientist qualifies a statement because qualifying is correct. It varies. It depends. Under certain conditions. In a peer setting, the qualifier is the honest part. In a 90-second segment, the qualifier is very often the only part that survives the edit, and audiences hear it as hedging or as an official who will not give a straight answer.
Definition stacking. To explain A, you define B. To define B, you need C. By the third nested definition you have used your whole answer and the interviewer has moved on. Nobody in the room was being obscure on purpose. They were being complete, and complete is the enemy of clear.
What I saw was not underconfidence. It was over-responsibility. These are people who feel personally accountable for whether the public understands something correctly, which is admirable, and which is also exactly the instinct that produces an unusable answer.
How do you turn technical material into plain language?
This is the technique we spent most of the day on, and it is simple enough to run yourself. It has three passes and it needs a recording device, which for our purposes is a phone.
- Pass one: answer it the way you would answer a colleague. Record it. Do not edit while you speak, and do not try to be good. In this group, first passes typically ran 45 to 60 seconds.
- Pass two: play it back with a pen. Circle every term that would require a definition for a general audience. Also circle any phrase in the family of "it depends," "in general," and "under certain conditions." A first pass usually produced five to eight circles.
- Pass three: answer again under two hard rules. No circled term survives, and nothing gets defined. A term that has to go is replaced with a comparison to something the listener already owns. Target 20 seconds.
This works because the third pass forbids the two moves that were doing the damage. You cannot stack definitions if you are not allowed to define anything, and you cannot hedge into unusability with the qualifiers circled in front of you.
The strongest version I heard that day came from an expert who had been trying to explain a numeric threshold along with the statistical reasoning behind it. First pass: 41 seconds, six circles, and by the end of it I could not have told you whether the number was good news or bad news. Third pass: 14 seconds, one physical comparison anyone could picture in their own kitchen, and the answer to "should I be worried" was unmistakable.
He was not less accurate. He had changed the unit of accuracy from the number to the conclusion.
Several people in the room pushed back on comparisons, and they were right to. A comparison that is easy to picture can be directionally misleading, and a scientist who oversimplifies in public pays for it with peers. We did not fully resolve that tension in the room, and I would not claim we did. What we landed on was a working test: if the comparison would make you wince in front of a peer reviewer, it is too loose, so build a different one.
What happens when the question is not about your program?
The most instructive moment of the day was a failure, and it was mine to spot rather than his to fix.
One expert, mid-exercise, was asked about a story that was in the news that week. It had nothing to do with his program. He stopped. Four seconds, maybe five, which on camera is a very long time. From the outside it looked like a man who did not know the answer.
He knew plenty. When we played it back and I asked him what had happened in that pause, he described solving two problems at once. Was this question even his to answer? And if he answered it, would he be speaking somewhere he was not authorized to speak?
That second question is not nerves. It is a real institutional constraint, and in a federal agency it is a serious one. It is also completely invisible to a viewer, who reads the pause as either ignorance or concealment. Neither is what happened.
The fix is to stop making two decisions in the same second. Before any interview, decide the boundary of what you are cleared to speak to, out loud, with whoever gets to decide it. Then in the moment you have exactly one live decision: is this inside the boundary or outside it. A boundary decided in advance turns a five-second freeze into a one-second turn.
From there the move is short. Answer the part you genuinely own out loud. Name the part you do not. Route the rest by role rather than by naming a colleague, because naming a person hands the reporter their next call and it is rarely yours to give. The full system for out-of-lane questions goes further, and we spend real time on it in the room, but the pre-decided boundary does the most work on its own. If off-topic questions are your main worry, I go deeper in how to answer tough media interview questions.
Do six recorded rounds actually change anything?
Each person went roughly six rounds on camera across the day. That number is not arbitrary, and the arc was consistent enough that I can describe it.
| Round | What usually happened | What it teaches |
|---|---|---|
| 1 | The answer they would give a colleague | What you actually sound like |
| 2 | Shorter and worse, because people overcorrect | Brevity alone is not the goal |
| 3 | First genuinely usable answer | The structure holds when calm |
| 4 | Same answer, harder question | It survives pressure |
| 5 | Same answer, colder interviewer | It does not need encouragement |
| 6 | Same answer, interrupted mid-sentence | It can be re-entered, not restarted |
One recorded round tells you what you look like. That is worth having, and it is where most one-hour sessions stop. Five or six rounds tell you what you do when the conditions change, and that is the difference between having had an experience and owning a skill.
Round two is the one people are surprised by. Told to be brief, a careful person will strip out the wrong things first, usually the concrete detail, and keep the abstractions. It looks like a regression. It is actually the moment the real work starts.
What should a technical organization take from this?
- Promotions create press exposure. After any reorganization, audit who now holds the titles a reporter will search for.
- The barrier is translation, not nerves. Do not build a session around confidence for people who are not short of it.
- Ban definitions to kill definition stacking. The rule does more than any amount of advice about being clearer.
- Pre-decide the boundary. Authorization questions have to be settled before the interview, not during the pause.
- Count the rounds. One is a demonstration. Five or six is training.
If your organization is technical, the pattern here repeats almost exactly with clinicians and researchers, which I wrote up in media training for doctors. If you are on the government side, our overview of media training for government agencies covers how these sessions are usually structured, and building key messages for media interviews is the natural next step once the translation work is done.
If a reorganization has quietly changed who at your organization is likely to get the call, tell us about it here or call (646) 233-1424. Speak with Jess about what a day with your experts would actually cover.
Client details in this case study have been generalized at the client's request. The agency, its technical domain, and all participants are unnamed, and the paraphrased comment from a senior scientist is used with permission and without attribution. Dates and session structure are accurate.