It seems they either got lost and thought they were crossing the correct runway* or they had been expecting 31L for take off and the instructions for 4L never sunk in. The pilot on the radio is saying all the right things but the pilot steering isn’t going the right way.
*I would have thought a B777 crew flying JFK to LHR would be operating from their home base so they should know it well, but maybe not.
I just took a look at this YouTube with the real-time audio:
Wow. That was a cluster.
The audio and diagram positions in that video started while the conflict event was just about to begin, so some of the set-up is missing, but is played later out of order. I tried some other vids to try to put the positions and radio traffic into proper time and place sequence but could not find a better one. If anyone does, please share.
Essentially the AA 777 was approaching the crotch of the two runways and had been cleared to cross one of them but instead crossed the other. Oops. No issues with radios jamming, or mistaken readbacks or bad radio technique. The 777 was told to do something which was sensible for that phase of taxi, read back what they were told, then did something else instead. The “why” will be interesting from a human factors perspective, but the “what” is real clear.
And thus do 400+ people get real close to a bad situation.
It only takes a moment’s confusion or inattention.
There but for the grace of …
Know it well … yeah. The direction they did taxi and the runway they did cross was consistent with one of the other common configurations for JFK evening departures. Doubtless one they’d done countless times before. Perhaps it was simply too much habit and not enough in-the-moment.
A goof is a goof is a goof. There’s always a reason but never an excuse.
Turning back to the Delta 737, they stopped short enough that they could not have been going very fast at the moment they started the reject manuever. As such the brake energy numbers would almost certainly have been reasonable and taxiing to the terminal once things had settled down would have been safe and sensible.
Yeesh, I’ll say. 
If I had to guess the female voice was the FO doing the radio calls (correctly) and the male voice was the captain who was taxing based on what he expected. he heard “cleared to cross runway” and assumed incorrectly.
It’s human nature to hear what we expect to hear and act erroneously on it. The benefit and hope of a 2 crew cockpit is that one of them catches such a mistake.
A little bit of light relief (insofar as that can ever be the case for an aviation accident, and with apologies to the injured party) from a location near me, courtesy of the AAIB. A sign became detached from a gate due to helicopter downdraft, and cut a waiting passenger’s leg. According to the accident report, factors contributing to the sign not being properly secured included:
“During the site set-up cows escaped on to the site from an adjacent field and needed to be
removed and secured. The site set-up was then running late, and passengers were arriving
and parking in incorrect locations.”
The nature of the sign in question? It was a passenger safety notice.
Source (pdf): https://assets.publishing.service.gov.uk/media/63905d8cd3bf7f3281ab1a76/Bell_206B_G-TOYZ_01-23.pdf
With apologies for the double post, but I also wanted to share this one - someone briefly touched down their Cessna 182 on top of a truck after they approached the threshold slightly too low. From the report:
“The driver of the articulated vehicle informed the AAIB of the event and provided photographs
of tyre marks on the top of the articulated vehicle. A maintenance inspection after the event
revealed no damage to the aircraft.”
Cool stunt if you can pull it off, shame no-one filmed it! Most importantly, no-one was injured, which I guess may not have been the outcome had they been lower by a few inches.
Source (pdf): https://assets.publishing.service.gov.uk/media/63905e3ed3bf7f3281ab1a77/Cessna_182Q_G-BRRK_01-23.pdf
That’s almost certainly the nub of it.
Having now taken the time to watch both vids, I think @Richard_Pearse picked a better one than I did. Note that either vid has the unreality that ATC seems silent except when talking to these airplanes. In reality there’d be a LOT more unrelated radio traffic to other airplanes. The vid producer filtered all that out so laymen could make some sense of it all. Likewise they filtered out all but the two airplanes on the taxi diagram; in reality there was probably a sea of them out there going every which way.
There are at least a few of what we call “threats” here which are opportunities to get it right or get it wrong. None of which are excuses, but all of which contribute incrementally to aligning the proverbial Swiss cheese so a mistake leaks all the way through to actually occurring. Our goal is to “trap the threat” by anticipating it might happen, detecting it promptly if it does, and reacting to it promptly to beat it back.
JFK was departing on both the runway Delta was using and AA was supposed to us, and also the other runway that AA ultimately taxied towards. The ATIS would have mentioned both runways and may or may not have indicated which would be used for which departure routings. Sometimes your ATC route clearance will give a big hint as to which runway to expect, and sometimes it won’t. In any case, before they pushed back they will have decided, based on the information available, which runway to expect, and they will have set up the computers & cockpit knobs accordingly and discussed the taxi plan to that expected runway. Were they expecting 31L or were they expecting 4L? That will come out in the investigation eventually, but we don’t know now. Having to perform all this setup based on an expectation is a threat. And if the set-up does prove wrong, another threat is having to fix all that set-up & briefing and mental alignment while out on the taxiways moving along with the other traffic.
Now, maybe 20 minutes after all the briefing and 40 minutes after that computer set-up, they finally push back, start, and taxi up to the edge of the ramp, and call ground control for taxi clearance. Ground control tells them in standard ATC sequence but also in standard shorthand: “Start by going this way immediately, you’re eventually taxiing to runway X, but along the way stop at point Y and await further instructions”. The female voice (99.99% certainly the FO) reads back, also in correct shorthand, “Start by going this way immediately, then along the way stop at point Y and await further instructions.” Perhaps critically she did not read back the part about “you’re eventually taxiing to runway X”. Maybe that was a simple mistake, maybe that was her expecting that runway as the destination so felt that that part of the readback was unnecessary, or maybe she heard it but didn’t process it or read it back because it bounced off her expectation to taxi instead to 31L. Or was the radio insanely busy and she was simply trying to be extra brief, generally a good impulse, but one to be applied only with great care and finesse? No way for us to know. But that’s a threat she created, or at least failed to trap.
Typically the other pilot (Captain in this case) will be listening to both the clearance, and the readback and form his (in this case he/his) own opinion of what they should be doing. And as well he will read back his understanding to her and/or raise an objection if he thinks the three of them, ATC, FO, and Captain, don’t all agree on their understanding. Absent the CVR recording, which is long gone, nobody will ever know for sure what they did or didn’t say to each other.
A long-haul flight like that will carry a second FO. Who will be riding in the cockpit and should be fully participating mentally in the flight from arrival at the aircraft up through shortly after they reach cruise altitude. During cruise each of the three pilots in turn will be napping in the back while the other two drive across the ocean. An hour-ish before landing the last napper awakens and rejoins the cockpit and all three fully participate from then until the jet is parked at the gate. Was this person helpful here? Or were they mentally tuned out, physically present but functionally absent? Or worse yet, paying just enough attention to be in the way, adding confusion, not dispelling it? We don’t know.
As I mentioned, the actual ATC radio will have been pretty continuous yak-yak to other planes in addition to the few radio calls we hear. All of which you listen to, trying to form a mental picture of who’s where, who might interfere with your path, and who you’ll be following. The ideal is to have a mental model of everyone around you whose movements might be relevant to your own. This process invades your thinking a bit and you need to guard against assuming “airplane X got this clearance, so I will too”. Expect is one thing, but assume is a mental bridge too far. All the radio chatter also interferes a bit with communicating with your own crewmates. You need to talk about what you’re doing, but also don’t want to talk on top of ATC chatter. Unless the yakyak is so dense it’s just impossible not to talk over it at least sometimes. How many planes ahead of them rightly went the way they mistakenly did? Threat.
Once they begin taxiing, a couple or few minutes after the last clearance but still about 3/4ths of a mile from their cleared waiting point near the intended runway crossing point, they’re told to join another taxiway and cross a particular runway: 31L. Which implicitly also means “proceed beyond the previous waiting point according to what I just said instead”. No reiteration of which runway they’re going to, namely 4L, the one they’re not supposed to be crossing. Nor was the reiteration required by standard phraseology. But it’s commonly given. The FO accurately read back what was said and did not add any confirmatory statement about which runway they were going to. Nor was that required. But again is often done. Two opportunities there for confusion to be dispelled, or for confusion to be revealed. Neither were taken. Threat & threat.
Further, when a clearance is given that can’t be acted on for considerable distance and only after a long passage of time it becomes mentally “stale”. Where expectation or habit may unconsciously overcome what’s actually supposed to be remembered. A very good technique is to write this stuff down immediately. Easier for the FO with 2 free hands than the Captain with about 1/2 of a free hand while taxiing. As well easier to see your writing in the daytime, not after dark. Yes there are lights for your clipboard & paper, but you try to keep them off or very dim as much as possible to better see outside. Threat.
From the wait point Y they’d been given, crossing either runway and going to the other is a completely plausible standard JFK maneuver. That’s a threat too, one driven by airfield design. “Cross A; go to B” and “cross B; go to A” can be interchanged for each other pretty easily. Whether that’s ATC saying the wrong thing, pilots hearing the wrong thing, or hearing one then remembering and doing the other. Any of those are plausible mistakes that can and do happen. Threat. Ideally you’d like waiting to be at places where there’s really only one plausible next move; not two equally plausible next moves, both of which immediately involve entering different runways. But the taxiways are where they are. ATC does what they can to mitigate the threat, but they’re stuck with the layout too.
After an unknown-to-us amount of waiting and after a bunch of airplanes have probably departed on 31L before 4L, a few have crossed 4L to go to the far end of 31L, and also a bunch of airplanes have probably crossed 31L between those two points heading to 4L, they finally get up to the point where they were originally told to wait and the clearance was later amended so they should turn right 90 degrees and cross 31L.
Instead they go straight ahead and cross 4L = the wrong runway. Having gotten the “turn and cross” clearance 5 or 10 minutes ago, and watched a half-dozen airplanes take each branch at that point up ahead, what was in their mind when they finally got there themselves? Did they simply forget what they were supposed to do? Did they have such a deep-seated expectation for using 31L that they still didn’t process what they heard and read back 10 minutes prior? We don’t know.
Right there the taxiways jog and twist a bit and change names in a confusing 5-way intersection. “Going straight” isn’t really straight, it takes a roughly 30 degree turn to the left starting around what’s a sweeping 90 degree turn, then take a branch off to the right, turning about 30 degrees back to the right in the course of that transition to the branch. Immediately followed by entering the runway. Despite being well away from any buildings or obstacles, it’s tight close quarters maneuvering in a 777-sized jet. Did the micro concern of negotiating that jog correctly without putting landing gear in the grass override the macro concern of where they were versus where they were supposed to be and whether what they were doing was what they were supposed to be doing? We don’t know.
A few hundred jets do this right every day. These pilots did it right a hundred times before. Why not that night?
As I said previously, a goof is a goof is a goof. There is a reason, but there’s never an excuse. Good thing they’re all alive to think about it.
I didn’t know that. A simple and easy mistake to make and it’s missed by 3 people. Ouch.
And my thought was that they have to keep the instructions to a minimum so that it doesn’t tie up communications at a busy airport. The only time I was in BOS as a loadmaster I listened to the ground frequency and the ground clearance person talked non-stop for a couple of minutes. It was like listening to an auctioneer. I was flustered just listening to it and the pilots were looking at each other with a WTH expression.
That’s life in the Bigs. Ground control is nutso much of every day at every hub in the industry. BOS, JFK, LGA, EWR, PHL, ATL, ORD, DEN, PHX, LAX, SFO, etc. Sometimes it’s calm and mellow. But that’s not the way to bet, and especially not at peak times, like 6-9pm at JFK.
Now imagine all those international flights with crews who speak English as a 2nd language.
This is going to be the next big thing, making communication a better process.
Agree completely.
Most of these places are more patient with the foreigners than with Americans. ORD in particular has gotten a LOT better when dealing with the Asians than they used to be. JFK has a lot less patience and make a lot less allowance than ORD now does. But they still make some allowance.
Another problem comes up when they assume a US carrier or a pilot with an American accent is highly familiar with that airport and its latest changes. Some are, many aren’t. And everybody who goes there has to be a noob at some time.
The central challenge is the only way to depressurize the radios is to split the airport into more sections with separate frequencies and separate controllers for each section. FAA is already short-handed and especially at these high-cost-of-living very high stress locations. Ultimately the ATC system is a taxpayer-supported government service and like any other such service, we don’t get the quality / quantity the customers want, we get the quality / quantity Congress is willing to pay for.
The fact that most of these airports are old and small and have doubled their terminal size and their average airplane size since they were designed, but have grown exactly zero more acreage doesn’t help. JFK at rush hour looks a lot like Manhattan at rush hour. Gridlock, physical gridlock, is only a couple of goofs away all night every night. There just isn’t room to leave the large margins that would depressurize everything.
It doesn’t even have to be ESL pilots, just pilots with a different accent might have trouble understanding instructions from an American controller (and vice versa, it’s a general language thing, nothing to do with one country being more or less intelligible than another.) It’s even difficult for someone with the same accent to pick up radio instructions until they’ve had some practice at it, as I’m sure you would remember from your student pilot days.
One thing that makes it easier to understand radio calls is to know in advance what is likely to be said, but this leads to expecting something to be said then hearing what you expect rather than what is actually said.
Most here can probably relate to the situation where someone is talking to you and you’re not paying attention. Then they pull you up on it, “hey you aren’t listening!” and you are then able to parrot back word for word exactly what they were saying, “ha! see I was listening”. But the truth is you wren’t actually listening and until you played the words back in your head you hadn’t fully processed them and hadn’t comprehended what they meant.
It’s pretty easy to do that on the radio too. The controller gives you an instruction, and you read it back perfectly but it never goes any deeper into your brain and you don’t comprehend what the instruction actually means. Writing an instruction down is good for engaging another part of the brain which helps it to go into the comprehension bits, it also serves as a record of what was said, or at least what you thought was said.
Edit: Lots of guys will type the gist of a taxi instruction into the MCDU (flight management computer interface).
A few decades ago Quebec decided that air traffic would have to communicate in French because of their language laws. That was a disaster (not literally), but one of the amusing things was that pilots who spoke native French had to ask the controllers to speak English anyway because A) Quebec French accents are different, and B) You learn to hear aviation jargon in English, and eventually you learn the cadence and patter and it makes it easier to understand. The same instructions in a different language, even if you speak that language, turn out to be not nearly as easy to understand unless you use it all the time.
I think they eventually abandoned it due to obvious safety concerns.
I’m not sure if it has been discussed on the SDMB but an A320 very nearly hit the ground a mile short of the runway on a dark and stormy night because the French controller gave the wrong altimeter setting in English to the English speaking crews. The QNH was 1001 but the flight crew were told 1011 by the controller. They did not pick up that it was 10 hectopascal different from the ATIS and subsequently flew an approach about 300 feet lower than their altimeters told them they were. They flew the missed approach when they didn’t get visual with the runway. Flight data showed they got within 6 feet of the ground during the go-around.
There were lots of things that went wrong, it wasn’t just a language issue but it was largely down to communication or a lack of it.
Ouch!.
I (and my several thousand compatriots from various companies) deal routinely now with multiple flavors of Spanish, Brazilian Portuguese, and French-, Dutch-, and British-colonial English. The controllers are often speaking the local native language to the local native airplanes in the local native accent and meanwhile aviation English to the foreigners such as myself.
And of course there’s hectopascals and sometimes metric visibilities mixed with traditional US inches mercury and feet/miles of vis. At least we don’t deal with metric altitudes like the folks flying to China and (formerly) Russia.
The opportunities to foul that up are legion. But after a few years, it’s not a big deal.
I am uniformly impressed by how hard the ground folks are trying to make this work. At the same time, by and large they are not stuffing quite so many airplanes into quite such a small space as the US and the EU / UK try to do. I’m told that Heathrow is the absolute limit case of this, and it routinely humbles New Yorkers on their first pass through there.
It’s eye-opening to be reminded just how much an LNAV/VNAV approach depends on the correct altimeter setting. Foul that up and centered needles are driving you along a totally different path through space to an uncertain fate. The RA is your only hope of detecting the discrepancy. Which is only useful over very flat or very well-known terrain. A set-up to screw up with little opportunity for an independent backstop.
The same altimeter setting error on an ILS results in the right path through space although you may intersect the ground unexpectedly at the bottom or along the way to the FAF. Or, more benignly, fail to break out at the bottom and go around unnecessarily.
A big, nay, huge, difference. And one not as appreciated as it ought to be.
It was originally my understanding – apparently mistaken – that English being the universal language of aviation, it was always spoken at all major international airports, though smaller regional airports could use the local language. But the former is apparently not the case, as I specifically remember hearing French being spoken between an Air France airliner and ATC at CDG in Paris. This strikes me as a huge safety problem in that non-French pilots might miss crucial information about what’s going on around them, or might struggle to understand it. I don’t know whether this use of French is cultural/political or just an innocent matter of convenience, but it sure doesn’t seem safe. But the aborted attempt to mandate French at major Quebec airports was definitely a political spinoff of the idiotic language laws there, proposed by morons with apparently no understanding of aviation conventions or aviation safety.
The French are pretty stubborn about speaking French in France. Understandable I suppose, we’d be pretty keen to speak English in English speaking countries if French was the international standard. It is a risk though, as you note. In the incident I referenced above, the French controller was giving the correct altimeter setting in French but the incorrect setting in English. If you could understand all of what you were hearing on the radio you’d quickly note that you were being given different information from other aircraft.
ETAA: as prefigured by @Richard_Pearse just above …
ETA: @wolfpup
There is definitely a tradeoff.
Lots of non-US controllers speak very nice aviation English. But scratch their surface and it’s clear they are only saying the magic phrases out of the big book of standard jargon. They cannot understand a non-standard query nor respond to it in English. If you have a complex problem or need, you’re mostly screwed unless you can express it in the local language.
The local pilots for whom English is not their native language have the semi-opposite problem. If they are forced to speak only English they can make standard requests and understand standard repsonses. But they’re hosed if something unusual comes up.
Consider e.g. Mexico, where about half the jet flying is done by Mexican flagged carriers flown by Mexican aviators, and another material fraction is flown by fellow Latin Americans whose native language is some flavor of Spanish. All those pilots and all those controllers could choose to communicate through two language barriers just so the Gringos, Brits, and Canadians can follow along. Or they can communicate in Spanish and be sure they’re understood by each other. That seems a no-brainer to me.
Bottom line: it’s a two-edged sword.
I can’t speak Spanish on the ground too much beyond ordering a beer, a bimbo, and a burrito. But I can now follow ATC Spanish pretty well. I could never make a significant request in Spanish, nor read back a materially complicated clearance delivered to me in Spanish. But I can sure understand enough of what the controller is saying in Spanish to Spanish speaking pilots in flight that I can maintain my own situational awareness of what’s going on around me. As long as the controller is kind enough to use English when addressing me, I’ll be fine.
Given that English is not nearly as universal as white European folk assumed it was back in 1945 when all the treaties were signed, the current mélange is about as good as it can get.