There is a checklist after engine start; it is never the first or last line that is skipped, always a line in the middle, and almost always the same one. The observation repeats itself between operators, in different aircraft, with competent and trained crews. A checklist is not cut short because it was memorized incorrectly, but because it was written without considering how people read, memorize, and get distracted. Its length, order, and timing determine, before any training takes place, which part will actually be executed.
A checklist does not tell you what to do; it confirms that it has been done.
The most expensive mistake in documentation design is when the checklist takes over the work of the flow. The flow, the memorized scanning done without support, brings the aircraft to the desired state; the checklist then confirms that the achieved state is the correct one. As soon as an element exists without prior flow, the checklist becomes an execution manual: it grows, distracts the reader’s attention instead of allowing them to focus on the instruments, and stops being a safety net to become the main means of configuration.
This distinction governs everything else. A verification checklist can remain short because it only retains elements for which there is no other omission detector. An execution checklist must necessarily be long, as it must contain every action. And length determines the skipping. You do not just choose a document format; you choose, through this format, the complete execution rate you will achieve in six months, on a delayed afternoon, on the line.
Three modes coexist in most documentation, and each guarantees something different. Mixing them means believing that an executed checklist is a verified one.
| Mode | What is guaranteed | What it costs | Where it is applied |
|---|---|---|---|
| Only memorized flow | Speed, without support, the gaze remains outside | No trace, no omission detector | Configuration, dynamic phases |
| Checklist read and then executed | That the action was performed, in the written order | A pilot fully absorbed by the document | Abnormal situations, without time pressure |
| Checklist read and answered | Two looks at the same state and a verbal trace | Time and a response discipline | After a flow, in phase transitions |
| Checklist silently verified | Nothing that the opposite seat can verify | Complete loss of cross-verification | Nowhere in a manned aircraft |
The length is paid in attention, not in seconds
The usual objection is simple: a longer checklist would only take a little more time. That is wrong, and predictable. A short checklist is read; a long one is scanned. Beyond a handful of elements, the reader stops treating each line as a question and begins to treat the checklist as a block to be completed. The ends resist: you start the checklist with attention, end it with the feeling of having completed it. The middle, however, has neither the protection of the beginning nor the end.
This middle drop is not an individual weakness; it is a property of working memory, and it worsens with length: the longer the checklist, the wider the unprotected zone, and the more likely it is that an omission will go unnoticed, since the “checklist complete” call was actually made.
When checklists get too long, people stop reading every item carefully. They scan the middle part quickly, which is why mistakes often happen there. Breaking long checklists into smaller, named sections helps because each section gets the attention it needs.
Imagine a pre-takeoff checklist with 18 items. Divided into three groups of six, each tied to a natural pause in the flow, it is more likely to be completed in full, because each group feels manageable. Without this structure, the items in the middle are the ones most often skipped.
The solution is not to shorten blindly, but to create ends. A checklist with 18 elements, divided into three named sections, each completed by its own announcement, does not contain 18 elements, but three sets of six, with six protected positions instead of two. But the division must fall on a boundary that the crew already knows: a panel change, a role change, a phase transition. An arbitrary division adds load instead of removing it.
Sedimentation. Every notable event adds a line to the checklist; almost nothing removes one. Over time, a checklist after takeoff can double in length without an explicit decision to extend it. The result is a checklist that no one fully executes and that everyone certifies as executed. An entry rule without an exit rule mechanically produces an unusable document.
Order: follow the hand, not the system flow diagram
An element is not checked abstractly. It is checked with a glance at a specific location on the flight deck, and often with a hand that reaches it. The order of the elements is therefore a physical path before it is a documentary logic, and a path costs more the more it deviates: every step back requires a new visual search, and a visual search is exactly when you grab the adjacent control, the one that looks like the right one, two rows below.
Checklists written by systems are flawless on paper and mediocre on the flight deck: the fuel, electrical, and hydraulic systems do not occupy contiguous zones. On the other hand, a geometric order, a regular scanning, always in the same direction, panel by panel, is memorized as a gesture and controlled from the opposite seat, because the other pilot knows where the gaze should be at that moment. It also makes anomalies visible: a crew member looking elsewhere than expected signals, without saying anything, that they have lost the line.
Time: Anchor the checklist to an event, never to an intention
“When you are ready” is not a trigger. It is an intention: it lives in someone’s head and shifts quietly as soon as that person is busy. A usable trigger has three properties: it is external to the person who announces it, it is observable from both seats, and it is binary. The last engine stabilized, the door closed, the landing gear retracted and locked, the aircraft aligned: these events are not negotiable and cannot be postponed without the opposite seat noticing.
The second choice concerns where the checklist is placed in the workload profile of the phase. By design, a checklist almost always falls at the end of a phase, i.e., when the workload increases: exactly when available time decreases and skipping becomes rational from the perspective of the person committing it.
Checklists should start when something specific happens (like “last engine stabilized”), not when someone feels ready. This makes sure both pilots know exactly when to begin. Starting checklists too late, when things get busy, means pilots might skip important steps.
On a Boeing 737, the Before Takeoff checklist is anchored to a clear event, such as the clearance to line up. If it were triggered by « when ready », a busy crew might push it to the last moment on the runway, with less time to complete it calmly. Anchoring it to a fixed event gives the crew a predictable moment to run it.
Shifting a checklist a few dozen seconds earlier costs almost nothing, but it presupposes that an earlier trigger exists. Here, design pays off: you must find the observable event that occurs in the low workload zone, instead of anchoring the checklist to the last event of the phase, which is also the most loaded.
Name the checklist after its trigger. “After the stabilization of the last engine” is verifiable; “after engine start” is debatable. If the title of the checklist contains the event that opens it, the omission of the trigger becomes audible to the opposite seat, and the question “Where are we?” disappears from the flight deck.
Interruption is the dominant error mode
No checklist is executed in one go on the line. It is interrupted by a frequency, a signal, a third party entering. The breaking point is not the interruption itself, often legitimate and unavoidable: it is the resumption. The crew resumes work at the item they believe they had reached. Yet the memory of the position in a checklist is exactly what the interruption first destroys, and it is not replaced by doubt: it is replaced by certainty. That is why an omission after an interruption is almost never detected; no one feels that they are skipping something.
Two arrangements make resumption reliable, and neither depends on memory. The first is physical: mark the position before handling the interruption, never after, the finger on the line, the cursor in place, the card with the current element on top on the kneeboard. The second is a resumption rule written in the manual, rather than left to momentary judgment.
Resume one element earlier. After every interruption, resumption occurs at the element preceding the one you believed you had reached. The cost is a few seconds and some redundancy; the benefit covers the most common case where the item was called but not verified. If the position is lost, the entire section is repeated, and a short section makes this rule bearable, which directly relates to the choice of length.
Which elements deserve to be included?
The selection criterion does not depend on the importance of the affected system, but on two independent factors: the severity of the consequences of the omission and the probability that the omission will reveal itself before the occurrence of its effects. A control whose incorrect position activates a warning, turns on a light, or changes the behavior of the aircraft in a minute does not need a line in the checklist: it already has its detector, and this detector is better than the document.
Not every control needs to be on a checklist. If a mistake would be caught by a warning light or the plane’s behavior, it doesn’t need a checklist line. Only things that could cause serious problems without any warning should be included.
On an A330, the crossfeed valve position doesn’t need a checklist line if a misplaced valve is quickly flagged by the fuel system alerts. But the flaps position does need a line, because a wrong setting might not be detected in time, with catastrophic results. This is why checklists focus on critical failures that would otherwise stay silent.
The field that justifies a line is therefore narrow. It is that of so-called critical elements, those whose absence only becomes apparent when it is too late to act. Everything else falls under the flow, a delayed verification, or nothing. This discipline is difficult because it forces the removal of lines that someone added for a good reason. But a line kept out of caution is not neutral: it consumes the attention that is missing three lines later, and this attention is taken from a critical element.
Writing an element: the expected state, not the past action
A well-written element sets an expected state, not a performed action. The difference is seen in the answer it allows. “Verified” is an answer available without looking; a value read aloud is not. The most dangerous phrase in a checklist remains “as required”: it transfers the choice of the value to the reader, precisely when the checklist should ensure that two people converge on the same value.
| Formulation of the element | What is actually verified | Error mode |
|---|---|---|
| Fuel circuit—verified | That someone has looked at something | The answer is available without looking: nothing distinguishes the verification from the reflex |
| Crossfeed—closed | The position of a named, unique, and visible control | Marginal, provided only one control on the flight deck bears this name |
| Flaps—as required | Nothing: the expected value is left to the reader | Two pilots confirm two different values, without knowing it |
| Flaps—set, value announced | The displayed value, compared to the briefing value | Marginal: the discrepancy becomes audible as soon as the answer comes |
- Good: “Crossfeed—closed” (specific state)
- Bad: “Fuel circuit—verified” (vague)
- Worst: “Flaps—as required” (no standard)
- Best: “Flaps—set, value announced” (exact verification)
The challenge-and-response rule plays the same role: the person reading the challenge names the expected state, the person who answers describes the observed state. If the answer is a fixed word, no information circulates, and the ritual replaces the verification. If it is a value, the discrepancy is immediately audible, even for the third crew member on the flight deck, who was not asked.
The answer that precedes the gaze. In a memorized checklist, the answer comes before the eye reaches the control. It is honest: the pilot is not cheating; they remember the state in which they left the aircraft. Through this mechanism, a checklist executed as a formality confirms a state that does not exist. An answer containing a variable value, a number, a position, a quantity, remains the most effective brake because it cannot be produced from memory without being heard.
What can be measured in your own checklist
None of the above can be verified on paper. Three observations, conducted on the line or in the simulator without announcing what is being observed, are sufficient to classify a checklist. The first: where the execution is accelerated. The point where the rate clearly increases is where the checklist is no longer read. The second: the delay between the challenge and the response. A zero delay, repeated, indicates an answer produced from memory. The third: the gap between the written checklist and the checklist actually executed, not the omissions, but the elements that crews spontaneously shifted, grouped, or rewrote. This gap is a diagnosis of the design, not a problem of discipline: it shows where the document contradicts the flight deck.

In your own documentation, three questions deserve to be asked. Is the trigger of each checklist an observable event or an intention? On what boundaries do the sections fall, and how many elements separate two announcements? And above all: what is the procedure for removing a line? Most documentation systems can add a line in a few days and have no established path to remove it. It is the gray zone that remains: elements derived from a restriction of the flight manual are not discussed, but the vast majority of the lines in an operator’s checklist do not fall under it, and their retention is justified nowhere.
For those preparing a type rating, a revalidation, or a renewal, the consequence is direct: it is no longer a matter of being able to recite the checklist. What is now verified is the ability to explain why a particular element is there, which event opens the checklist, and what exactly happens when it is interrupted in the middle. It is also the reading that prioritizes training modules because it separates the operator from the reader.
The applicable requirements, the aircraft flight manual, and the operator’s approved documentation take precedence over this text in all circumstances: it describes design mechanisms, does not replace a published procedure, and does not authorize a change to an operational checklist on its own.

Leave a Reply
You must be logged in to post a comment.