Non-Technical Skills: From Attitudes to Observable Behaviour

A pilot hesitates in a flight simulator during a stormy approach, observed by an instructor in a control room.

A candidate leaves the simulator with the note “lack of assertiveness” in their file. They challenge it. The instructor recalls a discomfort, a silence at the wrong moment, a hand that didn’t go to the thrust lever – but nothing they can reproduce for a third party. The note won’t be kept, not because it’s false, but because it was never more than an impression. Moving to behavioral markers isn’t an exercise in vocabulary: it’s a way to avoid writing what a camera wouldn’t have recorded.

What you actually write when you note an attitude

Between the fact and the note, there’s a staircase climbed without realizing it. At the bottom, the raw observation: at a given moment, in a given configuration, the pilot did this, said that, said nothing. One step up, the description, which already chooses its words. Above that, the interpretation: that silence becomes a hesitation. At the top, the judgment: the candidate “lacks self-confidence.” Each step makes writing easier and the trail thinner. At the last step, nothing is left to show.

The flaw isn’t moral, and it isn’t a competence flaw. An experienced evaluator quickly infers, and often correctly: it’s even what makes them a valuable instructor. The problem is that an inference written in a box replaces the fact and erases it. Six months later, no one – including the person who wrote it – can reconstruct what happened. What remains is an outdated opinion about a person, in a file that will follow them.

The inference ladder, from observed fact to personal judgmentA four-step ladder rises from left to right: what is seen, what is described, what is inferred, what is judged. A vertical dotted line separates the first two steps, which are verifiable observations, from the last two, which are interpretations. A horizontal arrow indicates that the inference grows to the right, while the opposable trail disappears.1. what is seen2. what is described3. what is inferred4. what is judgedgrowing inferencecan be provencan be interpreted
Each step gains in ease of writing and loses in traceability. The first two can be reproduced for a third party; the last two can only be defended by the authority of the person who wrote them. A usable file stays at the level of the first two.

When you write a note about someone’s behavior, you start with what actually happened (like a video recording). Then you describe it, interpret it, and finally judge it. The more you move from fact to judgment, the harder it is to prove what you wrote. That’s why it’s better to stick to describing what you saw, not what you think about the person.

During a simulator session, a pilot fails to announce a mode change on the autopilot. Instead of writing “poor communication,” the instructor notes: “Did not announce FMA change from NAV to APP at 10,000 feet during the approach briefing.” This specific observation can be verified on the replay, while the judgment “poor communication” cannot.

Three practical consequences. First: the note can’t be challenged on the facts, so it can’t be defended either. A dispute over a check is resolved on facts, and a personality assessment isn’t; the evaluator who wrote it ends up having to prove a character trait. Second: it isn’t trainable. No one trains to “be less self-confident”; they train to state a fallback plan before reaching a criterion, and that can be repeated, measured, and validated. Third: it can’t be aggregated. One hundred files marked “good crew spirit” tell a safety officer nothing, while one hundred occurrences of the same behavioral deviation point to a procedure, documentation, or initial training problem.

Anatomy of a behavioral marker

A behavioral marker is a phrase that describes a visible or audible action, attributable to someone, situated in time and in a configuration, and that a third party not present could, with a recording, say whether it happened or not. Four properties, and each one eliminates an entire family of formulations.

  • An action verb, not a state adjective. Announce, ask, reconfigure, sort the information, redistribute the tasks – never “is attentive,” “is rigorous,” “is comfortable.”
  • An observable content, not a mental content. Situational awareness isn’t visible. What is visible is the statement that manifests it: pointing out a state change before being asked, revisiting a hypothesis when a parameter contradicts it.
  • An observation condition. The same gesture doesn’t have the same value in stabilized cruise and in an missed approach with a failure in progress. A marker without context is just a slogan on the briefing room wall.
  • Refutability. If no conceivable situation would allow saying “this didn’t happen,” the phrase isn’t a marker, it’s a disguised assessment.

The test is based on a question, asked before writing: a colleague who wasn’t there, with the replay, would they write the same phrase? If the answer depends on their mood, fatigue, or opinion about the candidate, the phrase should go down a step on the ladder. It’s not an exercise in style: it restores what was actually in front of them.

Common formulations and observable facts they derive from
What is written in the file The fact that probably produced it Competence to link it to
“Lack of authority” Didn’t redistribute tasks when the pilot monitoring was overwhelmed by the failure handling Leadership and teamwork
“Too self-confident” Continued the approach after exceeding a stabilization criterion without stating a fallback plan Problem-solving and decision-making
“Good situational awareness” Announced the probable runway change and rebriefed before ATC imposed it Situational awareness and information management
“Poor communication” Selected three successive mode changes without announcing them Flight path management, automation
Worn leather with visible stitching, captured in close-up under raking light
What the grid should retain isn’t the pilot’s supposed firmness, but the action performed, its timing, and the announcement that accompanied it.

The middle column is longer than the left column, and that’s precisely the objection heard: writing facts takes time. It’s true in the room, and false in the long run. An attitude file is written in a few seconds and never reused; a factual file takes a few minutes more and feeds the debriefing, the next session, the operator’s analysis, and the file in case of litigation.

From the reference to the grid: three levels, never one

The references in force, those of ICAO as well as the European framework, are all organized the same way: a deliberately limited number of competencies – around eight to ten depending on the version – each broken down into observable behaviors, in turn read under described conditions. There you find the application of procedures and compliance with regulations, communication, flight path management by automation and by manual control, leadership and teamwork, problem-solving and decision-making, situational awareness and information management, workload management.

Three levels, and the classic mistake is to flatten one. At the top, the competency: a category of analysis, never something directly observed. In the middle, the observable behavior: what is noted. At the bottom, the context: what gives value to the noted behavior. A grid that skips the middle level asks the evaluator to assess a competency directly, i.e., an abstraction – and assesses it with what they have on hand, their impression.

The reverse mistake also exists. The list of markers is extended to cover all scenarios, and you get a multi-page grid that’s no longer usable in real time. The evaluator then returns to the overall impression and completes it after the session: the markers are there, but they serve as a justification vocabulary, not as an observation tool. A useful grid fits in the hand while observing, and leaves space to write a free phrase – that’s the one that counts, the box only classifies it.

The result isn’t proof of the behavior

The most frequent contamination doesn’t come from the candidate’s sympathy or reputation: it comes from the sequence’s result. The approach was stabilized, therefore “workload management was good.” The runway excursion was narrowly avoided, therefore “situational awareness was weak.” In both cases, the behavior was read through the result, when the session exists precisely to separate them.

Matrix crossing the sequence result with the observed behaviorFour boxes cross a result achieved or degraded with a behavior conforming to the marker or deviating from it. An achieved result doesn’t erase the deviation that preceded it, and a degraded result doesn’t prove the existence of a deviation: each box indicates what the evaluator is entitled to write.what the box allows writingresult achievedresult degradedconforming to the markerdeviation from the markercompetence demonstratedin this contextthe result masks the deviationnote the deviation anywaylook for the causeoutside the behaviourdeviation and consequenceestablish the link, don't assume it
The two dangerous boxes are on the ascending diagonal. In the top right, success absolves a deviation that will repeat; in the bottom left, a scapegoat is sought in the crew when the cause is elsewhere. These are the two boxes that a well-maintained grid forces to address.

The two boxes on the ascending diagonal are the most costly. In the top right, the deviation goes unnoticed because nothing serious happened: a sloppy approach briefing, an alert acknowledged without a callout, an unannounced mode change. The flight ended well, the file is clean, and the behavior will repeat in a less forgiving configuration. In the bottom left, the crew did what was necessary and the sequence degraded anyway: the cause is in the scenario, in an ambiguous procedure, in documentation that doesn’t say what it should. Marking down the crew here means losing the most useful part of the session.

The trap: the note that follows the result. When the sequence ends well, the behavioral deviation that preceded it disappears from the file. When it ends badly, the evaluator reconstructs a posteriori the markers that “explain” the result, and always finds them. Both errors have the same root: what is noted is what was understood at the end, not what was seen at the moment. It’s also the fastest way to turn a correct decision with an uncomfortable result – a discontinued approach that costs fuel and punctuality – into a poor assessment.

What makes two evaluators diverge

Have several instructors trained in the same place assess the same sequence: it’s rare for the notes to match. Part of the divergence is irreducible and harmless. The rest has a name, and can be treated. The halo effect, first: an exceptional competence drags three more up, poor manual flying drags them all down. The recency effect next: the last sequence weighs more than the four previous ones, simply because it’s fresh. Then each one’s severity or leniency, stable over time and generally invisible to the person carrying it. Finally the central tendency, which consists in never using the extremes of the scale, and which makes the grid blind where it would be most useful.

Dispersion of notes attributed to the same session by several evaluatorsTwo schematic distribution curves. The dotted curve, obtained with an attitude scale, is wide and low: the evaluators are spread over almost the entire scale. The solid curve, obtained with levels anchored in observable behaviors, is narrow and high: the evaluators converge around the same level.evaluatorsanchored levelsattitude scalelowhighattributed level
It’s not the evaluators’ severity that varies the most, it’s what they note. As long as the level is read from an impression, each one places the cursor on their internal scale; anchored in observables, the gap shrinks to a disagreement over grading, which is treated in standardization sessions.

The remedy isn’t to demand more rigor, it’s to remove from the evaluator decisions they can’t make alone. A level must be described by what is seen, not by an adverb: “conforming,” “above,” “below” anchor nothing until it’s said to what. Standardization sessions that work all start with the same material: a common replay, individual, silent grading, then an exchange – and what is discussed there is never the candidate, it’s the divergences between evaluators. Tracking each one’s severity over time completes the device, provided it’s presented for what it is: a calibration tool, not a grading of instructors.

The debriefing that produces evidence

A facilitated debriefing isn’t a pedagogical courtesy. It has an evidential function: it’s the moment when the facts are stated to the person concerned, where they can correct, complete, or challenge them, and where the link between the fact and the competence is built together. A note whose existence is discovered by the candidate when reading the file hasn’t passed this stage.

The rule of conduct is simple to state and hard to follow: the evaluator brings the facts, the candidate brings the analysis. “After the turn, the speed fell below the selected speed for approximately ten seconds; what were you focused on at that moment?” doesn’t resemble “you weren’t in the loop.” The first formulation allows a verifiable answer, the second only opens to a justification or silence. And when the candidate themselves makes the link with the competence, you’ve gained the two things sought: a defensible file and a learning point.

Detail preserved of a fact depending on the moment it’s writtenA decreasing curve represents the detail still available at the time of writing, depending on the delay since the fact. Three dotted vertical markers indicate the end of the sequence, the end of the session, and the next day. The drop is steeper between the end of the sequence and the end of the session.detail still availableend of sequenceend of sessionthe next daydelay before writing
The quality of a file is played out during the session, not in front of the input screen. What disappears first isn’t the event, it’s its context: the configuration, the order of the announcements, what was said in the briefing. And that is precisely what makes the difference between a marker and an adjective.

The right reflex: write the fact before choosing the box. During the session, note only a raw phrase, in the past, with its temporal and configuration reference: what was said, done, or omitted, and in what state the aircraft was. The link with a competence and the choice of the level come later, with hindsight, from this phrase. The reverse order – marking first, justifying later – produces exactly the files that can’t be defended, because the justification is fabricated to fit the already marked box.

What makes a grade defensible

A note based on markers doesn’t draw its strength from its accuracy – no one can guarantee it – but from the chain that produces it. Each link is visible, and each one can be discussed separately.

Chain that makes a note opposableFour boxes connected by arrows: the observed fact dated and located, the concerned competence, the deviation from the expected marker in this context, then the attributed level with its written trail. A dotted branch goes from the observed fact to an isolated box, the personal judgment, marked as a path without trail and without possible action.observed factdated and locatedcompetenceconcerneddeviation from theexpected markerassigned leveland its recordpersonal judgementno record, no possible action
A note is defended by the path that leads to it, never by itself. Each arrow is a legitimate point of challenge, and that’s precisely what makes it strong: the candidate can discuss the fact, its link, the deviation, or the grade, without ever having to discuss their person.

This chain also sets the burden of discussion. A candidate who challenges no longer challenges “the note” as a whole: they challenge a fact, or its link with a competence, or the magnitude of the deviation, or the grade. The evaluator responds on the same ground. It’s tedious, and that’s precisely what’s missing from the files that escalate. Conversely, a file marked “inappropriate attitude” only leaves the candidate one possible response, blame the evaluator, and they use it.

Two notation systems applied to the same session
Objective Attitude notation Marker-based notation
What is graded The person and their supposed dispositions An action placed in time, in a described context
Source The overall impression, consolidated at the end of the session The facts noted during the session, as the sequences unfold
Verification by a third party Impossible: the evidence faded with memory Possible as long as the replay and notes exist
Challenges Word against word, on moral ground Focuses first on the facts, then on the grade
Follow-up in training General advice, with nothing to verify A behavior to install, observable in the next session
Use by flight safety Adjectives that no analysis can add up Countable deviations that guide procedures and training
Deviation between two evaluators Wide, dominated by each one’s severity Reduced as soon as each level is described by what is seen

One last point, often overlooked: grading a training session and grading a check are not the same thing, even when using the same grid. In an evidence-based training scheme, the evaluation phase, the manoeuvres training phase and the scenario-based training phase don’t pursue the same objective, and mixing their grades empties the exercise of its meaning. What the operator does with each grade must be written somewhere, and known by those who give them as well as by those who receive them.

What remains to check in your own organisation

Four questions should be put to the documentation, not the instructor. What competency framework is actually in force in the training manual, and is the grid in use its exact breakdown? Is each level of the scale described by observables, or by adverbs? How long are the raw notes – the phrases taken during the session, not just the boxes – kept, and does the candidate have access to them? Is there a written challenge procedure, with a deadline and a third party? On these four points, the answer is often “we do it this way,” which isn’t an answer.

Two gray areas remain, and it’s better to name them than to treat them in silence. First, the evaluator is often the only witness: outside the simulator, there’s no replay, and the fact only survives through the note taken in the heat of the moment – hence the disproportionate importance of the writing moment. Then, behavior that is compliant on the line can be a deviation in a session, and vice versa: the economy of announcements that works in a seasoned crew isn’t transferable to a verification where you have to make visible what you do. A reference doesn’t say it; the briefing must.

For those preparing for an instructor or examiner qualification, the change is deeper than a change of form. The test no longer requires recognizing a good pilot, it requires producing an observation, linking it, noting it, and defending it to the person concerned. It’s a different craft from piloting, it’s learned, and it’s lost through lack of maintenance – standardization sessions also exist for that. Grids and case exchanges between practitioners on forums.fl330.com often reveal more than a reference, and the training offered on formations.fl330.com starts from the same diagnosis: the difficulty isn’t judging, it’s writing.

This article describes a mechanism, not an applicable standard. The regulation in force, the means of compliance chosen by the authority, and the operator’s approved documentation – reference, grid, scale, challenge procedure – prevail over all the above, including the formulations given here as examples.

Commentaires

Leave a Reply

Découvrez l’écosystème FL330

FL330, c’est toute l’aviation professionnelle — explorez nos autres espaces.