What's the most specific safety rule you know of? Not the broadest. Not the one everybody's heard of. The most specific.
Most people would say something about seatbelts or sterile fields.
Right. And they'd be wrong. Because there is a federal regulation that tells pilots, in writing, that they may not talk about their cars during taxi. It exists because of one crash, on one morning, in one city. And Daniel wants to know why it took seven years to write down.
Daniel's question, yeah. He asked what brought the sterile cockpit rule into existence, and then — the part I actually think is the meat of it — what it teaches us about how something as trivial as small talk affects attention in safety-critical work. Beyond the flight deck. He said that directly. Beyond the flight deck.
So the whole episode is that. The rule, the crash, and then the uncomfortable part, which is that the same lesson keeps showing up in operating rooms and in cars and nobody has internalized it.
And the rule itself is narrow. Which is what makes it interesting. It's not a moral position about chatter. It's a clock.
A clock.
Fourteen CFR, Part 121, section 542, and its twin in Part 135, section 100. Three parts. One, no operator may require, and no crew member may perform, any duty during a critical phase of flight except what's needed for the safe operation of the aircraft. Two, no crew member may engage in or permit any activity that could distract from those duties. And it names them. Eating meals. Non-essential conversation inside the cockpit. Non-essential communication between the cabin and the cockpit. Reading anything not related to conducting the flight.
Reading. They had to say reading.
And then part three, which is the whole thing: critical phase of flight means all ground operations involving taxi, takeoff and landing, and all flight below ten thousand feet. Except cruise flight.
Except cruise. So at thirty-five thousand feet over Kansas you can discuss your divorce. At nine thousand on approach you cannot.
That's exactly the line. And it applies only to Part 121 scheduled carriers and Part 135 commercial operators. Not Part 91. Your uncle in his Bonanza is not covered.
So the rule is not "be quiet." The rule is "be quiet now."
That distinction is the episode. Sumwalt — we'll get to him — said it outright years later. It is unrealistic to expect a crew to fly together for several days and never discuss anything except flying the aircraft. Crews need to talk. The rule defines when to stop.
Fine.
That's not a rebuttal.
I know. It's agreement. I just don't like how much I agree with it.
The rule was enacted January nineteenth, 1981. Forty-six Federal Register, page fifty-five hundred. More than six years after the crash that made it inevitable.
Six years. In aviation terms that's a geological era.
Yeah. And the crash is September eleventh, 1974. Eastern Air Lines Flight 212. A DC-9-31, tail number N8984E. Charleston to Charlotte to Chicago. It crashed three point three statute miles short of runway thirty-six at Douglas Municipal, in patchy dense ground fog. Eighty-two people aboard. Seventy-two died.
Seventy-two out of eighty-two.
Yeah.
And what were they talking about.
Politics and used cars.
Of course they were.
The cockpit voice recorder picked it up. Watergate. The 1973 oil crisis. Both crew members expressing strong views, and mild aggravation, is the NTSB's phrase. They talked until about two minutes and thirty seconds before impact. The Board wrote that the conversations were distractive and reflected a casual mood and lax cockpit atmosphere.
Lax. That's the word that got a regulation written.
Sort of. The probable cause is narrower than people remember. The flight crew's lack of altitude awareness at critical points during the approach, due to poor cockpit discipline, in that the crew did not follow prescribed procedure.
So the NTSB didn't say "they crashed because they talked about Watergate."
It's not that clean. Captain James Reeves, forty-nine years old, eighty-eight hundred and seventy-six hours, was the non-flying pilot. He was required to make altitude callouts and didn't. First Officer James Daniels Junior, thirty-six, three thousand and sixteen hours, was flying the aircraft. He believed he was at about eighteen hundred feet. The altimeter read six hundred and seventy.
That's a thousand-foot error.
Likely a misread of a drum-type altimeter. Those instruments require you to actually read the drum. And the terrain warning did sound, at a thousand feet. But it was silenced, because on every descent into Charlotte that alarm went off routinely. It had become noise.
So you have two failures stacked. Nobody was watching the altitude because the conversation had the attention, and the alarm that should have caught it had been trained out of them.
Normalization of a warning. That's its own whole failure mode.
And you're going somewhere with the Colbert thing.
James William Colbert Junior was on that flight. And two of his sons. He was the father of Stephen Colbert, who was ten years old at the time.
I didn't know that.
Most people don't. It's the detail that makes it a family rather than a statistic.
So the NTSB issued its recommendations on October eighth, 1974.
Six days after the report. And the FAA did nothing for seven years.
Why. That's the part I actually want.
Partly it's just the machinery. Notice of proposed rulemaking, comment period, industry pushback. Partly it's that the recommendation was about cockpit discipline generally, and turning "crews should be more professional" into a regulation is hard. What do you write down? How do you enforce it? And partly — this is the honest part — the industry position was that this is a training and culture problem, not a rule problem. Which sounds reasonable and is exactly what people say when they don't want a rule.
And then seven years later they wrote the rule anyway.
And then the crashes kept coming, which is really the argument.
Go on.
Air Florida, January 1982. Crashed into the Potomac. The NTSB found a probable cause was the crew's failure to enforce a sterile cockpit during the final preflight checklist. One year after the rule.
One year.
Delta 1141, 1988. Crashed after takeoff out of Dallas because the flaps weren't set. Fourteen dead, twenty-six seriously injured. The pilots were in conversation with a flight attendant while taxiing. The Board said the flap discrepancy might have been discovered if the captain had stopped the non-pertinent conversation.
Might have been discovered. That's the whole industry in four words.
USAir 1016, July 1994, Charlotte again, microburst, thirty-seven dead. The NTSB cited failure to comply with the sterile cockpit rule, and said the crew did not adhere to procedures during what they treated as a routine flight. There's a line from that trial — a Northwest 747 captain named Patrick Clyne was asked whether enforcing the sterile cockpit was nitpicking. He said, if it's nitpicking, it's nitpicking for safety, and nitpicking he was proud of.
That's a good line.
Corporate Airlines 5966, 2004, Kirksville, Missouri. Struck trees on approach. Thirteen dead. The Board blamed unprofessional behavior, and said the captain's joking established an inappropriate cockpit orientation.
Inappropriate cockpit orientation. There's a phrase for you.
And Comair 5191 in 2006 is the one people know. Lexington, Kentucky. They lined up on the wrong runway. The correct one was seven thousand and three feet. The one they used was thirty-five hundred.
Half.
Forty-nine of fifty people aboard died. And the CVR has them joking while taxiing. About their jobs. About their kids. One of them about a dog.
So you've got a rule that's been re-proven by crashes in eighty-two, eighty-eight, ninety-four, two thousand four, two thousand six. And it's still getting broken.
That's the paradox. That's where the attention science comes in.
Because every one of those crews knew the rule.
Every one. And that's the puzzle.
Robert Sumwalt.
Nineteen ninety-three, and at the time he was an analyst with the Aviation Safety Reporting System, before he became an NTSB board member. He went through sixty-three reports of sterile cockpit violations in the ASRS database. Forty-eight percent of them involved altitude deviations. Fourteen percent course deviations. Fourteen percent runway transgressions. Eight percent takeoffs or landings without clearance. Two percent near mid-air collisions.
Two percent near mid-air. That's how you find out you were about to die.
The four most common violations he found were extraneous conversation between crew members, distractions from flight attendants — that was nearly a quarter of the reports — non-pertinent radio calls and PA announcements, and sightseeing.
Sightseeing.
People flying around looking at things instead of their instruments.
In the critical phase of flight.
Yeah.
I keep waiting for the part where the violators are the bad pilots.
They aren't. That's what Sumwalt found and what later analysis confirmed. The violators are experienced, qualified, current crews. And there's a law review piece from 2007, Rapoport and Weisberg, which makes the argument that pilots simply don't believe small talk is dangerous. They know the rule. They don't believe the rule describes reality. And the thing that changes their behavior isn't conviction. It's the perception of being monitored.
So it's not a knowledge problem.
It's not a knowledge problem.
Then a rule on paper is doing almost nothing.
It's doing something, but it's doing less than people assume. And Rapoport and Weisberg make one more point that's brutal. Nobody actually knows the violation rate, because there is no effective auditing. A cockpit voice recorder only keeps the last thirty minutes or so. And violations are almost always discovered after an accident. So the data set is crashes. Which is a terrible way to measure compliance.
You only find out about the violations that killed people.
Yeah.
Herman, I want to push on the reframe here, because I think it's the actual point.
Go.
Sumwalt says crews need to talk. He's not being permissive. He's saying that if you take a rule that says "don't talk" and apply it literally, you break the crew. The two people in that cockpit have to coordinate on a level that requires conversation. If you silence them entirely, you get a different failure. So the rule isn't about whether; it's about when. And that means the question isn't "how do we get people to stop talking." It's "what tells a person, in the moment, that now is the moment to stop."
And that's the hard part. Because the trigger is workload, and workload is highest exactly when the cues to stop talking are weakest.
Say that again.
When you're on approach, the sterile cockpit is in effect and you're busy. The busier you are, the less spare capacity you have to notice that you should stop. That's the trap. It's not that people are careless. It's that the moment the rule is for is the moment the rule is hardest to feel.
Fine. That's the honest answer to Daniel's "how do trivial things affect attention."
It's not that small talk is trivial. It's that small talk is expensive, and it's cheapest when you need it most. The problem is that the brain doesn't bill you at the moment. It bills you later.
And the automobile version of this is beautiful.
Strayer and Drews. Dylan Strayer at Utah has been doing this for decades. The finding that mattered is that a conversation with a passenger in the car is far less impairing than a conversation on a phone. And the reason is almost embarrassingly simple.
The passenger can see the road.
When traffic thickens, the passenger notices, and they stop talking. Or they shorten. The conversation naturally thins out exactly when the driving gets hard. Whereas the person on the other end of the phone has no idea you just entered a construction zone with a merging truck, and they keep going.
So the passenger is doing the sterile cockpit for you.
The passenger is a self-imposed sterile cockpit. Which tells you the rule isn't artificial. It's encoding something that humans do naturally when they can see the stakes.
And the phone breaks that, because the other person can't see.
And that's why hands-free doesn't fix it. It was never the hands. Strayer's work found that even a hands-free conversation degrades your memory for what you looked at. You still look. You just don't remember seeing it.
There's something almost cruel about that. You did the thing and got nothing for it.
Yeah.
So take that and put it in a room with a patient in it.
The American College of Surgeons wrote a statement on distractions in the operating room in September 2016, and it imports the concept by name. It explicitly calls it the sterile cockpit. It recommends structured protocols at predictable moments — clipping an intracranial aneurysm, starting cardiopulmonary bypass — and restricting smartphones, noise, irrelevant conversation.
And the difference from aviation.
The difference is in their own words. In the operating room, critical events can and do occur throughout the operation. There is no takeoff-and-landing window. There's no ten-thousand-foot line. The entire case is potentially the critical phase.
Which means aviation's model doesn't transfer cleanly. In the cockpit the rule is bounded. In the OR it can't be.
It can't be, and the data says the cost of that is real. A 2021 study in the Australian and New Zealand Journal of Surgery recorded a door opening or another distracting event fifty-six times per hour of intraoperative time.
Per hour.
Fifty-six. And the most common and most impactful category was communication unrelated to the case.
A door opening every minute.
And a 2015 study in Ergonomics found a mean of twenty-six interruptions per case, with about eighty percent of them affecting more than one member of the team. And the average prolongation of operating time was eighteen point four six minutes per case.
Nearly twenty minutes.
Per case. There's also work from 2021 finding distractions present in somewhere around forty to forty-seven percent of surgical counts.
And a count is the moment you're literally counting instruments so you don't leave one inside a person.
Yeah.
I want the number that puts this in perspective. What's the human-error figure in aviation?
The FAA has said human error is a contributing factor in sixty to eighty percent of aviation incidents and accidents.
So the rule isn't about eliminating error. It's about shaving the top off a number that will never be zero.
That's the honest framing. You can't remove the human. You can change the conditions the human is working in, at the specific moments where the margin is thinnest.
And small talk is one of the conditions.
Small talk is the condition nobody counts. That's the point. You count fuel. You count altitude. Nobody counts the conversation about the dog.
Herman.
Yeah.
I think there's a version of this episode that's just "don't chat in the cockpit" and it would be wrong, and I want to say so plainly.
Say it.
The rule doesn't work because it forbids talking. It works, when it works, because it tells the crew that the moment has a name. That's the actual mechanism. It's not silence. It's a shared understanding that this stretch of flight is different. Which is why the violation is not really about noise — it's about a crew that has stopped noticing the moment.
That's why Comair bothers me. They weren't reckless. They were on a runway they'd used before, having a pleasant conversation, and the moment never announced itself.
And the runway is three thousand five hundred feet instead of seven thousand.
Yeah.
Let me sit with that for a second.
Okay.
So you've got sixty-five minutes of altitude deviations and a rule that nobody can audit, and the enforcement mechanism is essentially "somebody might listen to your final thirty minutes."
And pilots know that. They're not stupid. They know the tape keeps thirty minutes. They know nobody's listening unless something happens.
So the real compliance driver isn't the rule. It's the culture.
It's whether the crew believes the captain cares. That's the thing that's never in the regulation.
Which is exactly the lesson that doesn't transfer by writing a rule.
Right. And I'll say the thing I think the OR people get right that aviation got right late, which is that they name the specific moment. Clip the aneurysm. Start bypass. They stopped pretending the whole operation is uniformly critical, because a rule that covers everything is a rule nobody follows.
Because if everything is critical, nothing is.
Yeah.
Okay. Ten thousand feet.
What about it.
I want to hear the flight levels argument. Because I know it's coming.
The line is ten thousand feet, and the regulation says feet without ever saying which kind.
Which kind. Mean sea level versus above ground level.
In practice it's read as mean sea level in the US. Which means if you're at an airport that's already sitting at five thousand feet, you have a lot less than ten thousand feet of clearance before you're out of the critical phase. And if you're over the ocean, mean sea level and above ground level start to diverge dramatically.
And there must be somebody in this building who has an opinion about that.
There is.
Let's hear it.
It's mean sea level. It's always been mean sea level. Denver sits at fifty-four hundred feet. If you read that line as above ground level you'd have to keep the sterile cockpit in effect until you were at fifteen thousand four hundred feet, which is above the altitude most of those flights ever reach.
The reason it matters is that everybody says "ten thousand feet" like it's a magic line in the sky. It isn't. It's a number that shifts under you depending on where you took off from. I had an argument about this at a dinner four years ago, and it ended with me drawing it on a napkin.
You drew an altimeter on a napkin.
I drew a plane and an arrow going down to the ground, and a dashed line across, and the dashed line was the line, and underneath the dashed line it said sterile. The other person kept saying "well obviously it's above ground," and I kept saying it doesn't say that anywhere, and eventually his wife took the napkin.
She took it.
She folded it and put it in her bag. I have never decided whether that was a compliment.
Do you still have the diagram.
I don't have the diagram. I have the version I redrew at home, which is arguably better, because I colored it in.
What color.
Blue for the sky, brown for the ground, and the dashed line in red. It's about the size of a coaster.
And where did this expertise come from.
I used to sit in a compliance training review for a regional airline. Reviewing the videos. They had one on the sterile cockpit and they asked me for notes.
And I gave them one note. The pilot in the video was wearing first officer epaulettes. Two stripes. And he was playing the captain, who has four. And they had already paid for the shoot, so my note went down extremely badly.
They fired you over epaulettes.
They didn't renew the arrangement.
Over two stripes.
They said it was about scope. It was about the stripes. And the reason it was wrong is that they shot it on a Tuesday. The whole crew in that video was wrong for a Tuesday. I have a theory about Tuesday flights. I'm not going to get into it here.
No, you're going to have to say one thing about the Tuesday theory.
No.
Hilbert.
The theory is that the composition of a crew on a Tuesday is different from the composition on a Friday, and that the training video had been shot on a Tuesday and cast as though it were a Friday. That's as far as I'm going.
That's further than I expected.
It's a long theory. It has a chart.
I believe you.
The point about the rule is that the crews hate it more than you two are saying. Not the flying below ten thousand — everybody accepts that. The taxi. Sitting on a taxiway at O'Hare for forty minutes with the other pilot is the hardest part of the job for some of them. You cannot talk about anything except the airplane. And you can't leave. And you have to be ready.
That's the part I hadn't weighted. Taxi can be most of an hour.
I knew a captain who made it a game. He'd let the other pilot get three sentences into something and then hold up his hand, and the rule was that the clock restarted. Six years of flying together and the first officer never got past three.
Did you fly with him.
I was on the ground crew. I heard about it constantly.
That's a strange thing to make into a game.
It's a strange rule to have to make into a game. That's the part you left out. What you said about the moment having a name is right, and the Comair thing is right, but the thing you didn't say is that the name has to be said out loud by somebody, every time, or it doesn't exist. That captain said it out loud with his hand. That's what compliance actually looks like.
So it isn't the regulation.
The regulation is the napkin. The hand is the rule.
I'll take that further. It means the failure in every one of those crashes wasn't a crew that forgot the rule. It was a crew where nobody said the name.
Yeah. That's where I'd put it. And the one thing I'd correct from earlier — the epaulettes were a symptom. The whole video had a Friday crew on a Tuesday shoot. You can't train the moment if the people in the video never had to feel it.
Herman, take that and go somewhere with it.
I will. Because that reframes the OR problem entirely. The surgeons don't have a hand to hold up. They don't have a captain who can end a conversation with a gesture. The ACS statement is an attempt to build the hand — to say, out loud, before we clip this aneurysm, everybody stops. That's a spoken sterile cockpit. And the fifty-six distractions an hour number tells you how rarely anyone says it.
And I'd like to go back to where Daniel's question actually lands.
He asked what the rule teaches us about small talk in safety-critical contexts beyond the flight deck. And the answer is not "small talk is dangerous." It's that small talk is one of the few hazards that never shows up on an instrument, and the only defense is a person deciding, out loud, in the moment, that this is the part where we stop.
And the rule is a piece of paper that helps that person decide.
And the paper has been sitting there since nineteen eighty-one and Comair still happened in two thousand six.
It did.
So the paper is necessary and nowhere near sufficient.
Yeah. That's what Rapoport and Weisberg are saying too, in a harsher way. The perception of monitoring does more work than the rule. Which is a depressing finding for anybody who has ever written a policy and assumed the writing was the hard part.
And the reason it keeps coming back is that the operator knows the rule. They're not uninformed. They're experienced. That's the part that gets left out of the coverage. The story is always "the pilot was distracted." The story is never "the pilot was qualified, current, and had done this approach two hundred times, and the moment stopped feeling different."
Familiarity erodes the signal.
Every safety-critical field is fighting that, and none of them can write a rule that fixes it.
I'll say one more thing. Human error is in sixty to eighty percent of these accidents. That number gets read as "people are the problem." I read it as "people are always in the loop, and the loop has to be designed for them."
The fact that the rule is about timing and not silence is the design.
Yeah.
Charlotte remembers this. There's a memorial at the airport. Charlotte Douglas International, at the overlook, dedicated on the fifty-first anniversary of the crash. September of last year.
The Charlotte Observer did a documentary for the fiftieth anniversary. Five parts. The Untold Story of Charlotte's Deadliest Plane Crash.
The city has held onto it. And the rule it produced is still being broken in cockpits and, at a much higher rate, in operating rooms.
Fifty-six distractions an hour.
Which raises the one question I don't have an answer for. If the rule is this obvious, this specific, this thoroughly re-proven, and the violation rate is unknowable because the only way anyone finds out is a crash — then the rule isn't the fix. It's a marker that the problem exists. And we keep treating the marker as the fix.
The actual answer is that it works when somebody says the moment out loud.
Out loud, every time. The hand on the taxiway. And the discipline of knowing when to stop talking is, in the end, just the discipline of noticing.
Which is the hardest thing anybody asks a person to do in a job they've done a thousand times.
Well. That's what Daniel asked. That's the answer.
Thanks to Hilbert Flumingtop, our producer, for keeping the napkin version of the regulation on file.
If this was your kind of episode, go back for episode fifty-seven, From Lawyers in Limousines to Developers in Their PJs; episode ten, How ASR Went From Frustration To ... Whisper Magic; and episode five, Fine-Tuning ASR For Maximal Usability. This has been My Weird Prompts, the human-AI collaboration podcast. Everything's on my weird prompts dot com. If you want to send us your own prompt, you can do that on Telegram at t dot me slash MWP listener bot.
We'll be back soon.
See you then.
