Comair 5191 - Confirmation bias and framing

The cockpit voice recorder (CVR) transcript of ill-fated Comair flight 5191 crash upon takeoff at Lexington this last August was just released (January 17th), and one of the interpretations of it supports cognitive issues related to "framing" and "confirmation bias". The CVR transcript is the first one on the list of links here.

Below is a copy of a post I made to a different forum on the subject. "FO" is First Officer, the pilot in the right-hand seat who took over flying once they were taxied to and on the [wrong] runway by the left-hand seat Captain. As noted earlier, the aircraft in question only has nosewheel steering for the left-hand seat, so responsibilities had to be divided this way, even though the FO was going to fly as Pilot in Command (PIC) to Atlanta once taxiing was done.

The focus of this discussion is not on failings of the crew members, but on everything else, the "system factors" that contributed to this situation, the factors that provided the metahorical gun and the ammunition, loaded the gun, cocked the hammer, and handed it to the crewmembers who, at the end of that casual chain, pulled the trigger.

The human errors can be dealt with somewhat by training individuals or crews, but the system factors need to be dealt with by changes to the infrastructure.

And, not a single word of discussion in that 30 minutes on the CVR that the taxiway or runway lights were out - from which I'd guess they had already discussed this and the FO had shared his observation from Friday evening that the Northeast end of runway 22 had no lights working at all. The only time the subject comes up again is mid-roll, after crossing the real 22, when the FO says "dat wierd with no lights" and the Captain says "yeah." That's not proof, but it's the first point at which reality differed from his mental model, and it's the first point a comment is made.

Another framing issue was their delight that they had a very simple clearance that couldn't get "any easier that that", lessening vigilence more.

But there is still the factor of the wrong airport diagram, which is still wrong. ( http://www.naco.faa.gov/d-tpp/0701/00697AD.PDF for those who don't have charts.) The taxiway they should have been using, just west of Alpha-7, is not shown. (it's visible in the photo on wikipedia, at
http://en.wikipedia.org/wiki/Comair_Flight_5191 )

Here's one possible scenario for a different type of framing. The crew understood that A7 was going to be blocked, and got the news that the next best taxiway was in use. Whoever told them that (suppose) meant the new one, the one not shown. They looked at their charts and figured that must mean Alpha-6. They expect an intersection with no lights, no pavement markings, no signage, and four possible ways to go. They expect to see across from them 2 concrete taxiways and, to their extreme left, runway 22, which they expect to be 150 feet wide and have no lights. There is no other spot on that diagram with two concrete taxiways across from them.

But there is a place in reality with two concrete taxiways and a 150 foot wide unlit runway to their extreme left, which is where they really were.
They didn't know about the new taxiway, and they may not have realized that runway 22 was actually 150 feet wide (with only 75 usable).

So, again, nothing seemed out of place, and the decision was "really easy" - taxi to the unlit runway intersection and take the extreme left 150 foot wide runway - "impossible" to make a mistake.

It would be valuable to know what combination of barricades and signage was visible from their gate to the taxiway they took.

This discussion doesn't argue that the flight crew didn't commit errors, which they clearly did. The purpose of a safety review of this kind is not to assess legal liability, but to look for intervention points where this same kind of error could be prevented in the future.

The scenario I describe above is a perfect case where the worst possible error can occur even in a situation that the participants view as on in which an error would be impossible. (This is why operating rooms have "time outs" before surgery begins now.)

In this scenario (even if something different happened in Lexington), the tower or ground crew or briefers are working, literally, on a different map of the world than the pilots. Phrases could occur, such as "the taxiway is closed, use the next one" are rich in undetected ambiguity. Is taxiway A-7 closed its entire length, or just half way down it? Is "the next one" A-6, as shown on the flight crew's official FAA airport diagram, or is it the brand new taxiway not shown on any diagram? This phrase is not in the transcript - but we are missing the conversation where this was discussed prior to the CVR's 30 minute recording.

It could have happened, is the point. More precisely, things LIKE IT could happen in the future in different circumstances: conversational partners have different mental maps and don't realize it, and carry on what looks like an unambiguous conversation to both parties, but with totally different meaning. Framed in that context, everything that follows makes perfect sense, and even satisfies multiple cross-checks for being correct. Those in that frame say "It's impossible to get this wrong." The tower, seeing only one runway lit, could say "It's impossible to get this wrong." Vigilance is never triggered by the unfolding events on either side. The mental frame is so strong that the First Officer brought the Captain into his world as well. People are focusing on data that support their model, not looking for anything that might challenge the model - at least, not until midway through the takeoff roll.

This is, in this case, an error that following the standard protocols and procedures correctly would have detected. On the other hand, the hundreds of other check list items distracted from this one, so adding more procedures and protocols is not a guaranteed good thing. The First Officer was head down, working in the cockpit, not attending to taxiing because he was working all the other preflight checklist items.

Oh, and one more "system" factor that also would have changed the outcome entirely. The airport has a slight hill in the middle of it, so that the far end of the runways are not visible from the near end, but are out of sight over the hillcrest. The visibility is reported by ATIS ALPHA to be 8 miles. If it's correct that the crew believed the lights were just out at the near end, but turned on at the far end of the runway they sought, and the airport had been entirely flat, they would have seen at a glance that their model didn't fit. In fact, if they had seen the end of the runway even with lights on, they would have seen at a glance that it was not the longer runway.
Reviewing system factors, any one of which might have changed the outcome:

* the airport had a hill in the middle of it.
* Runway 26 was 150 feet wide, but shown on the airport diagram as 75 feet wide.
* The lights were out at the takeoff end of runway 22 for something like 30 minutes, the same window of time as when the First Officer arrived friday night and noted the situation.
* The airport diagram did not correspond to reality, lacking the extra runway. The airport diagrams that are current are still not updated. The small versions of the airport diagrams on the instrument procedure plates make it look like there is a closed runway that comes all the way down to runway 26, marked by an "X", even though the larger diagram shows there is a gap (from the one marked with several "x" flags.)
* There is at least one possible confusion of route to the takeoff point that the diagram discrepancy allows, in which a sharp left turn onto a 150-foot wide runway at the end of the taxiing made perfect and unambiguous sense.
* Only the pilot in the left hand seat could steer the nosewheel steering, which separated the pilot in command (the first officer) from the activity that was done incorrectly by the senior Captain in the left-hand seat.
* Unquestioned cultural convention demands that the higher ranking officer sit on the left side, even though it would make more sense in this aircraft to have the pilot in command sit on that side.
* The tower was understaffed, and the lone occupant was (correctly) busy with other traffic at the crucial few seconds when he otherwise might have idly watched flight 5191 taxi into position and noted the error. It wasn't the tower responsibility to do that, but it could have occured and caught the problem.
* The only person who noticed the error, apparently, a ramp worker, had no way to communicate by radio to the aircraft and his attempt to run to the runway and wave down the plane did not succeed.
* The aircraft was not equipped with the $18,000 piece of equipment that would have automatically detected the runway error and alerted the crew, possibly because the airline was in bankruptcy proceedings.
* The crew seemed to behave as if operating in violation of FAA regulations was something they were routinely expected to just do and shut up about - judging from the fact that they continued to attempt a takeoff from an unlit runway, even though the first officer sighed when he commented that the lights were out all over the place.

All of the above, while not "causal" in some senses of the word, are also factors that, if they were changed, would have changed the outcome or very likely changed the outcome of this flight. They may not alter the legal assignment of liability and "blame" for the outcome, but they should illuminate intervention points for preventing similar events in the future.

technorati tags:Comair, 5191, Comair5191, CVR, safety, framing, confirmation, NTSB, accident