Contribution
A review by the leading CRM research group of how airline CRM training changed between 1979 and the late 1990s, what is known of its effects, and why it should be re-founded on error management. It reports two findings from the group’s own work as warnings: that acceptance of CRM concepts slips over time, and that CRM designed in the United States fits poorly in high power-distance cultures. It reports no new data; the evidence is cited, mostly from the authors’ earlier studies.
Key insight
CRM’s purpose was always to manage error, and as the training spread and was proceduralized that purpose was lost. “It seems that in the process of teaching people how to work together we may have lost sight of why working together well is important” (p. 5). Hence the fifth generation: treat error as “ubiquitous and inevitable”, and CRM behaviours as “error countermeasures with three lines of defense”: avoid, trap, mitigate (p. 7).
Assumptions
- Attitudes measured by questionnaire indicate the cognitive acceptance of CRM, and those who reject CRM in attitude are “unlikely to follow its precepts behaviorally” (p. 4).
- Line audits under non-jeopardy conditions show practice better than evaluation in the simulator, where crews know they are judged (p. 4).
- Hofstede’s dimensions of national culture (power distance, individualism, uncertainty avoidance) predict how pilots receive CRM.
- Reason’s account of human error frames the fifth generation.
Key results
- Origins. The roots “are usually traced back to a workshop, Resource Management on the Flightdeck sponsored by the National Aeronautics and Space Administration in 1979” (p. 1). “The first comprehensive U.S. CRM program was initiated by United Airlines in 1981”, modeled on Blake and Mouton’s Managerial Grid (p. 2).
- United 173 as the motive. The NTSB “had singled out the captain’s failure to accept input from junior crewmembers (a characteristic sometimes referred to as the ‘Wrong Stuff’) and a lack of assertiveness by the flight engineer as causal factors in a United Airlines crash in 1978” (p. 2). First-generation courses targeted “a lack of assertiveness by juniors and authoritarian behavior by captains”.
- Five generations. First: individual managerial style, psychological testing, games. Second (from 1986): “Crew” not “Cockpit”, team building, briefings, situation awareness. Third (early 1990s): organizational culture, automation, other employee groups, “specialized CRM training for new captains to focus on the leadership role that accompanies command” (p. 3). Fourth: FAA’s Advanced Qualification Program (1990) integrates CRM into technical training and proceduralizes it in checklists. Fifth: error management.
- Validation. “The most obvious validation criterion, the accident rate per million flights, cannot be used” (p. 4). Surrogates: line-audit behaviour and attitudes, which change “in a positive direction”. But “the fact that crews can demonstrate effective crew coordination while being assessed under jeopardy conditions does not mean that they practice these concepts during normal line operations” (p. 4).
- Rejection and non-transfer. A small subset of pilots rejects CRM in every airline; remedial training “have not proved particularly effective”. Automation precepts are often not followed on the line.
- Decay. “Acceptance of basic concepts may decay over time.” The group found “a slippage in acceptance of basic concepts, even with recurrent training” (p. 5), shown for two airlines in Figure 1. Causes are speculative: lack of management support, no reinforcement by check airmen, dilution as training spread, and proceduralization that “might also obscure the purpose of the behavior”.
- Culture. In high power-distance cultures “Exhortations to junior crewmembers to be more assertive in questioning their captains may fall on deaf ears” (p. 6).
- Error management. Organizations “must communicate their formal understanding that errors will occur, and should adopt a non-punitive approach to error” (p. 7), with confidential reporting such as the FAA’s Aviation Safety Action Programs. Its effect “remains to be evaluated empirically” (p. 8).
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | CRM training in U.S. airlines evolved through five identifiable generations since the 1979 NASA workshop | moderate as history | the authors’ account, citing program descriptions (pp. 1–3, 7) |
| C2 | CRM’s safety effect cannot be validated by accident rates; surrogate measures show changes in behaviour and attitudes | moderate | cited line audits and attitude surveys, mostly the authors’ own (p. 4) |
| C3 | Acceptance of CRM concepts decays over the years after training, even with recurrent training | weak to moderate | surveys at two airlines (Figure 1), cited from Helmreich & Taggart 1995; causes speculative (p. 5) |
| C4 | CRM imported from the U.S. had less effect, and assertiveness training conflicts with high power-distance national cultures | moderate | cross-cultural survey work cited (Merritt 1996 and others) (pp. 5–6) |
| C5 | Framing CRM as error management (avoid, trap, mitigate) will make it more acceptable across cultures and to its rejecters | weak: a proposal | argued; “remains to be evaluated empirically” (pp. 7–8) |
Concepts
- Generation (of CRM): a period’s dominant content and focus of training; the authors note a generation is “closer to that of the Drosophila than the human”.
- Error troika: avoid error, trap incipient errors, mitigate the consequences of untrapped errors (Figure 2).
- Proceduralizing CRM: “formally mandating the practice of CRM precepts”, for instance as checklist items.
- Jeopardy / non-jeopardy: assessment that can affect a pilot’s qualification, against observation that cannot.
- Drongos: the authors’ name, among others, for pilots who reject CRM.
Connections
It cites the NTSB report on United 173 (LIT-1371), though its reference list gives the aircraft as a “Douglas DC-8-54”; the report says DC-8-61. It summarizes Helmreich & Foushee (1993) on earlier programs and Helmreich & Wilhelm (1991) on outcomes, and borrows its error framework from Reason (1990, 1997). ASRS is not mentioned; its confidential reporting example is the FAA’s 1997 Aviation Safety Action Programs.
Bearing on the record
- CASE-127 and LIT-1386’s “Why this is here”. The record cites this paper for CRM as “authority without abolishing command”, keeping “avenues of challenge” open while “preserving the capacity for decisive action”. Neither phrase is in the paper, and it makes no such argument. A captain’s command is mentioned once in the text, in the third generation’s training for new captains on “the leadership role that accompanies command”, and once as an attitude scale in Figure 1; “authority” appears once, in “the absolute authority of leaders” of high power-distance cultures. The paper’s subject is error management, the evolution of training and its decay. The idea the exchange attributes to it is better sourced from the NTSB’s recommendation X-79-17 (“participative management for captains and assertiveness training for other cockpit crewmembers”, NOTE-1086) and from this paper’s first-generation description (“a lack of assertiveness by juniors and authoritarian behavior by captains”). The LIT summary’s last clause, “without abolishing the captain’s command”, should be removed or marked as the exchange’s.
- CASE-127, “What it does not show”. Supported: the authors say accident rates cannot validate CRM.
- CLAIM-422 (act, procedure, maintained capacity). The strongest bearing. The paper reports that the capacity CRM trains decays even while the procedure (recurrent training) continues, and suspects that proceduralization itself can “obscure the purpose of the behavior”. It also separates the procedure from the capacity in measurement: performance in an evaluated simulator session is not practice on the line. That is the claim’s distinction, made by the field’s own researchers, with data (weak: two airlines, attitudes only).
- CLAIM-416 (success under a test as evidence). The jeopardy point is an instance: coordination shown when assessed is evidence relative to that regime, not of line practice.
- CLAIM-424. The cultural finding bears on it: a reform that opens routes of challenge depends on whether the culture lets them be used, and in high power-distance cultures the same training “may fall on deaf ears”. The capacity to revise is not given by the procedure that provides for it.
- CLAIM-349. First-generation CRM targeted exactly the gap between a junior’s information and its authority over the captain’s action.
- THEORY-145 (espoused theory and theory in use). Decay of accepted concepts, and pilots who accept CRM in the classroom and do not follow it on the line, are the training-level form of the gap.
- No THEORY filed. The decay finding is substantive and the record does not hold it, but this paper only reports it from earlier work in one figure, for attitudes at two airlines; it does not establish it. Helmreich & Taggart (1995) would be the source to read.
- No instruction for machine-learning practice.
Limitations
- A review with no new data; most cited evidence is from the authors’ group.
- The decay evidence is attitudes at two airlines, with causes speculative.
- The fifth generation is a proposal whose effect “remains to be evaluated”.
- The account is of U.S. commercial aviation; military CRM is excluded.
- The manuscript was read, not the typeset article.
Open questions
- Does decay in attitudes predict decay in line behaviour? The paper assumes so and does not show it.
- Does the error-management framing slow the decay? Continental’s program was to be the test.