Abstract
Crew resource management was developed in response to a specific diagnostic finding: that flight crews possessing all information necessary to avoid an accident nonetheless crashed because that information could not travel upward across the authority gradient separating first officer from captain. Nearly five decades after the discipline’s founding, the asymmetry it was designed to correct remains measurable in the accident record. This paper argues that the persistence of this failure reflects a structural defect in the design of the monitoring pilot’s role rather than a deficiency in training delivery. The monitoring pilot is granted voice without authority, is asked to exercise judgment under severe time compression against a statutory hierarchy, and faces an incentive structure that penalizes correct intervention in the ordinary case while vindicating it only when intervention has become too late to matter. The paper surveys the available remedies — procedural, regulatory, technological, and organizational — and argues that each works by removing the need for interpersonal intervention rather than by improving it. This constitutes an implicit admission that the mechanism at the discipline’s foundation does not scale to the cases that matter most. The paper concludes with observations on the attribution of blame to monitoring pilots in public commentary and on the occupational consequences that follow such attribution independently of any formal finding of fault.
1. Introduction
The discipline now called crew resource management originated in a 1979 workshop convened by NASA under the title Resource Management on the Flightdeck. It responded to a pattern that had become undeniable across the preceding decade: sophisticated aircraft operated by qualified crews were being lost in circumstances where the information required to prevent the loss was present in the cockpit and audible on the recorder, but had failed to produce corrective action.
Two accidents defined the pattern. At Tenerife in March 1977, the first officer of a KLM 747 voiced doubt about takeoff clearance and was overridden by a captain who was the most senior pilot in the airline and the head of its training department. In December 1978, a United Airlines DC-8 exhausted its fuel and crashed near Portland while the captain worked a landing gear indication problem and the flight engineer’s progressively more urgent fuel reports failed to alter the captain’s course. In both cases the necessary information existed. In both cases it could not move upward.
Crew resource management was the institutional answer. Its early formulations concentrated on assertiveness training for junior crewmembers, on the legitimacy of the challenge, and on the captain’s obligation to receive it. Successive generations broadened the curriculum to encompass workload management, situational awareness, threat and error management, and — in its more recent formulations — the explicit designation of pilot monitoring as a defined operational role rather than an ancillary one.
The discipline has produced real gains. The class of accident in which a crew loses situational awareness while nobody speaks has become markedly less common. What has not changed, across forty-seven years, is the directional asymmetry of monitoring failure.
2. The Documented Asymmetry
The National Transportation Safety Board’s 1994 safety study of flight-crew-involved major accidents in scheduled air transportation found that the captain was the pilot flying in approximately four-fifths of the accidents examined. The interpretation the Board offered, and which subsequent work has broadly sustained, is that monitoring failure is not randomly distributed between seats. First officers fail to challenge captains substantially more often than the reverse.
Two features of this finding warrant emphasis.
First, it has proven durable. It was not a snapshot of pre-CRM practice that subsequent training resolved. It describes a condition that persisted through the introduction of CRM, through its successive generational revisions, through the mandating of CRM training for Part 121 operators, and through the formal elevation of pilot monitoring to a named role with its own procedures.
Second, the durability is itself diagnostic. A training deficiency that survives five decades of training reform is unlikely to be a training deficiency. The more parsimonious explanation is that the training is being applied to a role whose structure defeats it.
3. Voice Without Authority: The Structural Defect
The monitoring pilot possesses a diagnostic function and a communicative function. He does not possess an authority function.
3.1 The escalation architecture and its terminal rung
Most carriers’ operations manuals prescribe some version of a graduated challenge: state the observation, restate it with greater directness, articulate the consequence, and — where the manual goes this far — assume control of the aircraft. The first three rungs are verbal. Only the fourth constitutes action.
The fourth rung is precisely where operations manuals become vague or fall silent. This is not accidental. Specifying it requires an air carrier to set down in writing the conditions under which a first officer may take an aircraft away from the pilot in command, and to commit in advance to protecting him afterward. Few carriers have been willing to do either.
3.2 The statutory obstacle
Under 14 C.F.R. § 91.3, the pilot in command has final authority over the operation of the aircraft. Part 121 reinforces this allocation throughout. A first officer who assumes control from a captain is acting against the person in whom federal regulation vests operational authority, and no provision of the regulations affords him protection for having done so. Whatever protection exists is retrospective, discretionary, and contingent on outcome.
3.3 The incentive inversion
The decisive structural feature has received insufficient attention in the CRM literature, and it may be stated as a proposition: intervention is penalized when it succeeds and vindicated only when it has become too late to matter.
Consider the ordinary case. A monitoring pilot judges an approach unsafe and assumes control at two hundred feet. The subsequent landing is uneventful. Nothing in the record demonstrates that the landing would have been otherwise had he not acted. He has committed an act of insubordination against the pilot in command, disrupted the crew relationship, generated a report, and produced no evidence of necessity. The counterfactual is unobservable, and he bears the cost of an intervention whose justification cannot be shown.
Consider now the case in which he does not intervene and the aircraft is lost. His judgment is retrospectively vindicated by the wreckage. The vindication is worthless.
The monitoring pilot therefore operates under a decision structure in which the expected cost of intervening exceeds the expected cost of not intervening across the great majority of cases in which intervention would have been correct. This is not a failure of courage. It is a rational response to an inverted incentive gradient, and no quantity of assertiveness training alters it, because assertiveness training addresses the willingness to speak rather than the consequences of acting.
3.4 Time compression
The final approach and landing phase compresses the entire escalation sequence into seconds. A monitoring pilot who identifies an unstable approach at five hundred feet has perhaps twenty seconds before the outcome is determined. Within that interval he must diagnose, articulate, observe the absence of response, escalate, observe the absence of response again, and decide whether to act against the pilot in command. The graduated architecture presumes a timescale the landing phase does not supply.
4. The Inverted Gradient
A circumstance deserving separate treatment arises when the authority gradient and the experience gradient run in opposite directions: a captain recently qualified on type paired with a first officer holding substantially more experience in the same aircraft.
This configuration is not unusual. Every newly upgraded captain spends an initial period paired with first officers who know the aircraft better than he does. The industry has long recognized the risk and has developed two mitigations.
The regulatory mitigation is 14 C.F.R. § 121.652, which raises landing minimums by one hundred feet and one-half mile for a pilot in command with fewer than one hundred hours as PIC in type. The carrier-level mitigation is the pairing restriction — commonly termed a green-on-green prohibition — barring the scheduling of a low-experience captain with a low-experience first officer.
Neither mitigation addresses the interpersonal consequence of the inversion, which is this: the first officer’s superior knowledge of the aircraft strengthens his diagnostic position while weakening his standing to act on it. He knows he is correct. He also knows that saying so crosses two boundaries simultaneously — the authority boundary and the implicit affront of instructing a senior pilot in an aircraft the senior pilot has flown for a shorter time. The inversion makes the diagnosis easier and the intervention harder.
A further consideration concerns career shape rather than hours. A first officer who has previously held command, whether at a regional carrier or in military service, carries an internalized model of decisional authority into the right seat. A first officer who has occupied only that seat possesses no such model. He has never been the person who decides. The difference does not appear in a logbook, and it is not captured by total time, but it bears directly on the probability that a monitoring pilot will convert diagnosis into action.
5. When the Verbal Channel Functions and the Outcome Does Not Change
The conventional account of monitoring failure holds that the monitoring pilot fell silent. A distinct and analytically more troubling category exists: cases in which the monitoring pilot diagnosed correctly, communicated repeatedly, and produced no corrective response from the pilot flying.
The distinction matters because the remedies differ entirely. Assertiveness training addresses silence. It has nothing to offer against the second case, which is a failure of reception rather than of transmission.
Crew resource management contains an unexamined premise at exactly this point: that a properly trained captain will respond appropriately to a properly delivered challenge. The discipline invests heavily in the quality of the challenge and comparatively little in what follows when a correctly delivered challenge fails. Yet the monitoring pilot’s entire remaining toolkit past that point consists of repetition — and repetition is precisely what the escalation architecture prescribes, up to a terminal rung most manuals decline to specify.
Where an accident record shows a monitoring pilot who identified a developing problem early, stated it, and continued stating it without producing a response, the finding is not that CRM training was inadequately delivered. The finding is that CRM’s design contains no provision for its own failure.
6. Remedies and Their Difficulty
Six categories of remedy merit examination. A pattern emerges across them.
6.1 The mandatory go-around gate without pilot-in-command discretion
Mechanism. A rule specifying that an approach not meeting stabilization criteria at a defined gate must be discontinued, with no discretionary override, enforced by mandatory flight data review of every exceedance.
Effect on the gradient. Decisive. The monitoring pilot’s task converts from persuasion to citation. He need not win an argument; he invokes a rule both pilots know will be reviewed on the ground regardless of outcome. The interpersonal channel is bypassed.
Difficulty. Substantial and multi-sourced. Such a rule arguably intrudes on § 91.3 authority, which the FAA has been consistently unwilling to disturb. Enforcement requires flight data monitoring, and FOQA programs are voluntary and protected from enforcement use under 14 C.F.R. Part 193 — a protection that constitutes the bargain securing operator and union participation. Compelling the data would destabilize that bargain. Additionally, a rigid gate carries its own hazard: rushed go-arounds initiated from low-energy states at low altitude constitute a recognized loss-of-control category. The remedy substitutes one accident class for another, and the net effect is contested.
6.2 The formalized two-challenge rule with explicit transfer-of-control authority
Mechanism. Operations manual language providing that where a monitoring pilot has issued two challenges without adequate response, he is authorized and required to assume control, with an explicit commitment that the carrier will not treat such action adversely.
Effect on the gradient. Considerable, in that it converts an act of insubordination into an act of compliance. The intervention is no longer against procedure; it is procedure.
Difficulty. Chiefly one of drafting. The language must survive arbitration, must define “adequate response” with sufficient precision to be applicable in seconds, and must not create liability for a carrier whose first officer takes an aircraft unnecessarily. The protective commitment is the hardest element: it requires the carrier to bind itself in advance regarding conduct it cannot foresee. Some carriers have adopted such provisions. Many have not, and the FAA has not required them.
6.3 Automated alerting
Mechanism. Runway overrun awareness and alerting systems, which compute stopping distance in real time and alert independently of crew communication. Comparable logic exists for approach energy management.
Effect on the gradient. Complete for the specific hazard addressed. The aircraft issues the warning, not the first officer. The authority gradient becomes irrelevant because no human is challenging any other human.
Difficulty. Economic. European regulators have moved toward requiring such capability on newly certificated large aeroplane types; the FAA has not. The obstacle is retrofit cost across an existing fleet, and it falls hardest on precisely the operators where the hazard concentrates: cargo carriers flying converted airframes of considerable age. The cost-benefit methodology, which estimates benefits from historical accident frequency and monetizes them through the lives exposed aboard the aircraft, systematically undervalues such mandates in all-cargo operations. The remedy that most cleanly solves the gradient problem is the one least likely to be mandated where it is most needed.
6.4 Pairing restrictions
Mechanism. Prohibiting the scheduling of a low-experience captain with a low-experience first officer, whether defined by hours in type, total time, or time since upgrade.
Effect on the gradient. Indirect but genuine. It does not improve intervention; it reduces the frequency of the configuration in which intervention is hardest.
Difficulty. Operational and economic. Pairing restrictions constrain crew scheduling, and they bind most severely at carriers experiencing rapid growth or high attrition — which is to say, at carriers where the underlying risk is greatest. A carrier that has doubled its fleet in fifteen months, or that has experienced attrition sufficient to force premature upgrades, may find a green-on-green prohibition operationally impossible to satisfy. The restriction is therefore weakest precisely where it is most needed, and this is a structural rather than an incidental defect.
6.5 Line-oriented flight training incorporating captain error
Mechanism. Simulator scenarios in which the captain is the erring party and the first officer must escalate to intervention, with the intervention rehearsed rather than merely discussed.
Effect on the gradient. Moderate and likely underrated. Rehearsal converts an unprecedented act into a practiced one, and the barrier to intervention is partly a barrier of novelty.
Difficulty. Cultural more than technical. Such scenarios require captains to be trained through their own failure in the presence of subordinates, which is uncomfortable and which some training departments resist. They consume simulator time, which is capacity-constrained at growing carriers. And the scenario must avoid becoming perfunctory, which is the customary fate of training exercises whose outcome is known in advance.
6.6 Flight data monitoring as organizational surveillance
Mechanism. Systematic review of unstable-approach continuation rates, disaggregated by crew pairing, route, airport, and time of day, with intervention at the training level where patterns emerge.
Effect on the gradient. None on any individual approach. Considerable over time, in that it makes visible to management a behavior that is otherwise invisible.
Difficulty. The data is voluntary and protected. Carriers without FOQA programs generate no such visibility. Where programs exist, the protections that secure participation also constrain the uses to which findings may be put. And the underlying industry finding — that the overwhelming majority of unstable approaches are continued to landing rather than discontinued, on the order of nineteen in twenty — indicates a base rate so high that identifying individual outliers is analytically difficult.
6.7 The pattern across remedies
Every effective remedy operates by removing the need for interpersonal intervention rather than by improving the capacity for it. The mandatory gate substitutes a rule for an argument. The two-challenge provision substitutes procedure for insubordination. Automated alerting substitutes a machine voice for a human one. Pairing restrictions reduce the frequency of the hard configuration. Line-oriented training rehearses the act so that it need not be invented under pressure.
None of them makes a first officer better at persuading a captain. This constitutes the discipline’s implicit admission that the interpersonal mechanism at its foundation does not scale to the steep-gradient, time-compressed case — which is the case that produces fatalities.
7. Attribution and Its Occupational Consequences
A final matter falls outside the technical analysis but bears directly on whether any remedy can succeed.
Public commentary following a runway excursion or approach-and-landing accident characteristically distributes blame between the two pilots on the basis of partial early information, frequently before the identity of the pilot flying has been established. Where a monitoring pilot is identified as having raised concerns without producing a corrective response, a recurring characterization holds that he was ineffective in his role — that a more capable monitoring pilot would have prevailed.
This characterization misstates the structure of the role. The monitoring pilot’s prescribed function is to identify, articulate, and escalate. Where operations manual language stops short of authorizing transfer of control, articulation and escalation exhaust the role as designed. To describe a pilot who performed the entirety of his defined function as ineffective is to hold him accountable for authority the position does not confer.
The consequences of such attribution are not confined to reputation. A first officer at an early career stage who is involved in a fatal accident carries that involvement in his employment record, discoverable through the Pilot Records Database, which every Part 121 carrier must query before hiring. The database contains employment and training records; it does not contain the Board’s probable cause statement. The flag therefore surfaces without the exculpatory context, and the burden of explanation falls on the airman at every subsequent application.
Hiring at major carriers proceeds by committee applying risk-averse heuristics to a large applicant pool. A candidate carrying an accident in his file competes against candidates with comparable qualifications and clean files. No committee need conclude that he was at fault. It need only prefer the uncomplicated file. The resulting exclusion is produced by an accumulation of individually defensible decisions, none of which constitutes a finding, and against which no appeal lies because no adverse determination was ever made.
The asymmetry with a senior airman is severe. A first officer with substantial total time and a long clean record presents an accident as an anomaly within a large sample. A first officer near the regulatory floor presents it as a substantial fraction of the whole record, at precisely the career stage requiring an employer willing to extend opportunity.
This produces an outcome of some importance to the safety system’s own stated aims. A monitoring pilot who speaks up, is not heeded, and survives may find his career terminated by mechanisms that never adjudicate his conduct. The system’s operative answer to such an airman is that it will believe him and decline to employ him. Whatever is claimed about valuing crewmembers who raise concerns, the observable consequence runs in the opposite direction, and it does so without any actor having chosen it.
8. Conclusion
Crew resource management was founded on the diagnosis that safety-critical information failed to travel upward across the cockpit authority gradient. Forty-seven years of curricular development have improved the transmission of that information without altering the structural conditions governing its reception or the monitoring pilot’s capacity to act when reception fails.
The monitoring pilot occupies a role granting voice without authority, exercised under severe time compression, against a statutory hierarchy, within an incentive structure that penalizes correct intervention in every case short of catastrophe. That pilots in this position frequently decline to intervene is not a training failure. It is the predictable output of the role as designed.
The available remedies are known and their difficulties are well understood. Each works by removing the interpersonal channel rather than repairing it. The discipline’s honest position is that it has addressed the cases in which nobody spoke and has made limited progress on the cases in which someone spoke and nothing changed.
Where an investigation establishes that a monitoring pilot performed the whole of his defined role, the appropriate finding concerns the design of the role rather than the performance of the individual occupying it. Whether such a finding produces institutional change, or joins the several prior findings of essentially the same character, remains the open question — and the answer to it will be determined less by the quality of the finding than by whether any actor possesses both the authority and the incentive to act on it.
Notes
Sourcing and verification. This paper draws on the general CRM literature and on the regulatory framework as understood by the author. Several load-bearing items should be verified against primary sources before publication or circulation:
- The NTSB’s 1994 safety study (A Review of Flightcrew-Involved Major Accidents of U.S. Air Carriers, 1978 through 1990, NTSB/SS-94/01) — the approximate four-in-five figure for captain-as-pilot-flying is cited from recollection and the exact proportion and sample definition should be confirmed.
- The Flight Safety Foundation’s go-around research — the finding that roughly nineteen in twenty unstable approaches are continued to landing derives from the Foundation’s Go-Around Decision-Making and Execution Project (Blajev and Curtis). The precise figure, methodology, and sample should be checked.
- The status of EASA rulemaking on runway overrun awareness and alerting systems is described here in general terms. The current requirement, its applicability date, and whether it extends beyond newly certificated types require direct verification against EASA documents.
- Section 121.652 high-minimums provisions are stated from recollection. The current text, applicability thresholds, and available exceptions should be confirmed against the current CFR.
- Pilot Records Database requirements (the 2021 final rule implementing Section 203 of the Airline Safety and Federal Aviation Administration Extension Act of 2010) — the scope of records included, and specifically whether NTSB findings are excluded, should be confirmed. The paper’s argument in Section 7 depends on that exclusion.
On the Tenerife and Portland accidents. Both are described here in the compressed form customary in CRM literature. Readers requiring precision should consult the Spanish authority’s report on Tenerife and NTSB/AAR-79/07 on United 173.
On scope. This paper addresses the authority gradient specifically and does not treat the broader CRM curriculum, threat and error management, or the substantial literature on national and organizational culture in multi-crew operations — including the cross-cultural work on power distance, which bears on the same problem from a different direction and which a fuller treatment would incorporate.
A methodological caution. Section 5 describes a category of monitoring failure in which the verbal channel functions and the outcome does not change. Where a reader has a particular accident in view, the identity of the pilot flying and the pilot monitoring is a threshold fact, and early investigative releases frequently do not establish it. The analysis in Sections 5 and 7 applies only where that attribution has been made. Applying it before attribution is established repeats the error the paper criticizes.
