The Mystery of Colgan 3407

By now we have grown accustomed to the banality of the causes of most accidents. Rather than highly technical failures of incomprehensible electronic or aerodynamic systems, many of them involve extremely basic errors of judgment or airmanship on the part of flight crews. The crashes of a Singapore Airlines 747 in 2000 and of a regional jet at Lexington, Kentucky, in 2006 were due to pilots trying to take off from the wrong runway, a blunder so fundamental and seemingly obvious that no technical means are in place to guard against it. The immediate cause of the Colgan Air 3407 crash was an error equally elementary. When the stick shaker warned of an impending stall, the captain did exactly the wrong thing: He pulled the control yoke back and kept holding it back all the way to the ground.

The airplane, a Bombardier Q400, was maneuvering to intercept the localizer on a night ILS approach. The engines were at flight idle. There was some ice buildup on the airframe, but not a lot. The autopilot was on. The airspeed had been around 175 knots during the descent but began to bleed off as the airplane was dirtied up, and the captain corrected for a slight downward altitude excursion without increasing power. At an indicated airspeed of 130 knots, the stick shaker triggered and the autopilot automatically disconnected, as required for a pilot-initiated stall recovery.

Most likely, the stall warning and autopilot disconnect took the captain, who was flying, by surprise. They were premature, occurring 20 knots above the normal stick-shaker speed, because the airplane was equipped with a switch that allowed the crew to select an elevated set of reference speeds in icing conditions, and the higher speeds had, in fact, been selected. The board suspected, however, that the captain had lost awareness of that fact. He had also apparently lost awareness of the airspeed indicator and of the rising red bar along its right-hand edge that provided early warning of the impending, but still remote, stall.

At this point the airplane still had an airspeed margin of perhaps 25 knots above the stall. But this is where things got strange. The moment the autopilot disconnected, the captain reacted by pulling back on the control yoke. Perhaps he remembered the reference speed reset and was expecting the stick pusher to kick in momentarily. But he did not merely maintain the current pitch attitude. In just four seconds, the pitch attitude increased to 25 degrees and the speed dropped below 120 knots. At the top of the flight director, a red chevron appeared, pointing downward — an urgent signal to get the nose down. The airplane had rolled to the left, and the captain responded with near-maximum power, a right-roll command and right rudder. But he kept holding the yoke back. Things were happening rapidly — too rapidly, perhaps, for thought or training to override whatever primitive and misguided impulse had taken possession of him. After two more seconds, the pitch attitude was 30 degrees, the speed was 100 knots, and the stick pusher kicked in, to no avail. The airplane gyrated from side to side; it had gained 200 feet and its airspeed had dropped below 80 knots before the nose finally fell through the horizon.

At this point the bank angle was more than 90 degrees, and the first officer, perhaps reverting to her days as an instructor in small airplanes, had raised the flaps. This was a mistake, but, given the actions of the captain, it had little influence upon the outcome. She then asked whether she should raise the gear. The captain, still struggling to right the airplane with aileron and rudder and obstinately battling the stick pusher with a pull that would eventually reach 160 pounds, replied, “Gear up.” At almost the same moment the big turboprop made its final post-stall gyration, rolling right and slicing into a vertical dive. The flight director’s red chevron now lay upon a solid brown background, pointing forlornly toward an unseen sky. But it was too late.

It’s the job of the NTSB to try to make sense out of strange events like this one, and it tried mightily to do so this time. It discussed the possible role of fatigue but concluded that the evidence was circumstantial and not very persuasive. Indeed, the cockpit voice recorder tape, in which the captain and first officer chatted garrulously, relating their life stories and pausing only briefly during “sterile cockpit” periods, hardly suggested two people surrendering to torpor.

Another one of the usual suspects was the lack of realism of flight simulators, especially at the edges of the flight envelope, and the misleading quality of flight training that emphasizes the approach to the stall, but not the stall itself or post-stall gyrations. Indeed, pilots who are trained from the outset in simulators and in turbine aircraft have very little experience with, or sense of, the natural evolution of a stall. To further mislead those pilots, one criterion of a proper recovery from an impending stall is minimum loss of altitude, whereas, as light-aircraft pilots who fully stall their airplanes know, the essential considerations are angle of attack and airspeed. A significant altitude loss is willingly accepted in the name of avoiding a dangerous secondary stall.

A third area of inquiry was the captain’s aptitude for his job. Many pilots who had flown with him gave him a good report, but he had failed a number of flight tests. He had reported only one of these on his application for employment with Colgan Air, but the company had disregarded it because it had occurred in 1991, so early in his flying career that it no longer seemed relevant. He had also received disapprovals on single-engine and multi-engine commercial check rides but had neglected to mention those. Later, while employed by Colgan Air, he had received unsatisfactory grades on three proficiency checks. In general, the items that he had failed had been ones outside the normal run of the mill — single-engine landings, rejected takeoffs, that sort of thing. He had always passed on a recheck, but the pattern of failures was suggestive of a person who functioned well under controlled circumstances but might be prone to break down under pressure.

Strangely, however, the captain was not the only one who had ever responded to a stick pusher or shaker by pulling back. In 2004, an empty Bombardier regional jet was on a repositioning flight when its crew decided to let the autopilot take it up to 41,000 feet. While the two pilots joked around, the airplane got slower and slower as the autopilot attempted to maintain the programmed climb rate. Finally, the airplane stalled, and the crew reacted to five stick-shaker and four stick-pusher activations by pulling harder on the control column each time.

In 1996, a DC-8 with an inoperative stall-protection system crashed when it stalled at 14,000 feet during an evaluation flight and the crew applied power but did not lower the nose. But what may be the most egregious case on record involved an MD-82 in Venezuela. It stalled, for reasons that the NTSB does not make clear, at 31,000 feet, and the crew held nose-up elevator, in spite of a “Stall! Stall!” aural warning and continuous stick shaker, for a minute and 46 seconds before impact.

Ultimately, the NTSB concluded simply that what happened on the Colgan Air flight just happened. “The accident captain’s history of training failures,” the board wrote, “showed that he had demonstrated weaknesses throughout his career with instrument flying skills and had relied heavily on the autopilot to help him stabilize the airplane.” Surprise. The lack of external visual cues and the knowledge that the airplane was close to the ground made the situation worse. “The captain’s response to stick-shaker activation should have been automatic,” the board continued, “but his improper flight control inputs were inconsistent with his training and were instead consistent with startle and confusion.”

The NTSB's animated reconstruction of the accident, complete with instrument displays and a moving control yoke, can be found on YouTube.

It is fascinating to observe the innocuousness of events preceding the stall and the blinding rapidity with which they unfold after it. It is also interesting to see that the captain's use of power, ailerons and rudder was timely and appropriate. He was, as Hamlet says of himself, "but mad north-northwest." He was doing everything right except one thing — and that was the thing that mattered most.

This article is based on the NTSB's report of the accident and is intended to bring the issues raised to our readers' attention. It is not intended to judge or to reach any definitive conclusions about the ability or capacity of any person, living or dead, or any aircraft or accessory.