THE WARNING THAT WAS HEARD

What if one of the greatest weaknesses in modern safety systems is not that people remain silent, but that their warnings fail to change anything?

14:37 — Someone Speaks

Imagine a production floor on an ordinary afternoon. A machine is running, production targets are demanding, and everything appears to be under control. Then an experienced operator notices something unusual. The machine sounds slightly different. There is no alarm, no visible damage, no obvious reason to stop, but something has changed, and he knows this machine well enough to hear it.

He walks over to his supervisor.

"This doesn't sound right. It wasn't doing this yesterday."

The supervisor listens and replies, "Thanks for telling me. We'll keep an eye on it."

The operator returns to his station, production continues, and nothing happens.

Perhaps the machine is perfectly safe. Perhaps the sound is the first sign of an emerging mechanical failure. We do not know, and that uncertainty is precisely the point.

The worker noticed a difference, judged that it might matter, overcame the natural hesitation to interrupt his supervisor, and spoke. The supervisor acknowledged him. Yet the organization has still not established what the signal means, or whether its current course of action remains safe.

The worker has done what we so often ask people to do: notice, question and speak.

But has the safety system done what it owes to the worker?

This is where our next excavation begins.

The Accident That Almost Happened

On July 7, 2017, Air Canada Flight 759 approached San Francisco International Airport, cleared to land on runway 28R.

But instead of lining up with the runway, the aircraft was aligned with a parallel taxiway on which four passenger aircraft were waiting for takeoff.

During the approach, the crew noticed unexpected lights and asked air traffic control to confirm that the runway was clear.

The controller confirmed the landing clearance, yet the contradiction between what the pilots were seeing and what everyone expected to be true remained unresolved.

As the aircraft continued to descend, a pilot aboard one of the waiting aircraft questioned where it was going and warned that it was lined up with the taxiway. The crew of another aircraft switched on their lights.

The Air Canada pilots initiated a go-around, and the controller then instructed them to go around as well. The investigation established that the pilots had already begun the manoeuvre before receiving that instruction.

The aircraft passed dangerously close to those waiting below.

Fortunately, no one was injured.

The National Transportation Safety Board investigation identified several contributing factors, including expectation bias, fatigue, ineffective review of runway closure information and breakdowns in crew resource management.

This was a complex event, not simply a case of someone failing to listen, and we should be careful not to reduce it to one. The NTSB could not establish which particular cue finally triggered the pilots' decision to abandon the approach.

But one aspect deserves our attention.

An early signal that something might be wrong had already entered the communication system.

The crew had questioned whether the runway was clear. The controller had responded. Yet the exchange did not resolve the discrepancy between the aircraft's actual position and the situation everyone believed they were managing.

Other warnings and visual cues followed, and the pilots initiated the go-around.

The distinction matters.

The danger was not resolved merely because somebody spoke. Avoiding the collision required the developing situation to be recognised as unsafe and corrective action to follow.

Which raises a question that reaches far beyond aviation.

How often do organizations receive information that should challenge their assumptions, only to process it in a way that preserves those assumptions?

We Have Studied the Courage to Speak. But What About the Capacity to Listen?

Over the last three decades, organizational psychology has devoted considerable attention to employee voice.

Amy Edmondson's influential 1999 study introduced psychological safety as a shared belief that a team is safe for interpersonal risk-taking, and her research across 51 work teams in a manufacturing company found an association between psychological safety and team learning behaviour.

The implication was important: people need conditions in which they can ask questions, admit uncertainty, report mistakes and challenge assumptions without unnecessary interpersonal fear.

Later research narrowed the lens to safety itself.

In 2019, Mark Noort, Tom Reader and Alex Gillespie published a systematic review of the safety voice literature, examining speaking up about safety to prevent accidents and physical harm, and integrating findings from 48 articles.

More recently, a meta-analysis by Archana Manapragada Tedone and colleagues, first published online in 2025, synthesised 49 independent studies involving 50,856 participants and found that safety voice is more prevalent where work environments, leaders and colleagues are safety-supportive.

These findings matter, and encouraging people to speak remains essential.

But speaking is only one part of communication, and communication contributes to prevention when relevant information influences understanding and action.

This brings us to a concept that deserves far more attention than it receives:

Safety Listening.

In a conceptual review first published online in 2024, Alyssa Pandolfo, Tom Reader and Alex Gillespie examined how organizations respond when people raise safety concerns.

Their work challenged a familiar assumption, namely that listening is mainly about showing interest, being approachable or making people feel heard.

Instead, they proposed understanding safety listening through observable behaviour: what the recipient actually does with the information.

Do they seek clarification? Do they reconsider their understanding of the situation, investigate, intervene where necessary? Do their actions shape what happens next?

Responding to warnings, they argued, is the necessary counterpart of speaking up.

This connects directly with our earlier excavations, in which we asked whether safety systems unintentionally consume the very human capacities they depend upon: attention, judgment, voice, adaptability.

Now another possibility emerges.

Safety may be weakened not only when the worker loses the capacity to speak, but also when the organization loses the capacity to respond meaningfully.

When Listening Becomes a Performance

Consider two supervisors who receive the same observation from an employee:

"Something has changed with this machine."

The first replies, "Thank you. I'll note it."

The second asks, "What exactly changed? When did you first notice it? Has anyone else observed it? Do we need to secure the equipment while we investigate?"

Both may be polite, and both may genuinely care about safety, but their responses create very different opportunities to understand the risk.

The first acknowledges information.

The second begins to examine it.

Investigation, of course, does not mean that every concern must stop production, trigger the replacement of equipment or be accepted at face value.

People misinterpret signals, warnings can be mistaken, and sometimes a concern is investigated and the existing controls are confirmed to be adequate.

Effective listening does not require automatic agreement.

It requires a response proportionate to the potential harm and the uncertainty involved, and a decision whose reasoning can be explained.

This matters because no organization can manage every uncertainty by stopping everything indefinitely.

But neither can it treat uncertainty as resolved simply because someone in authority feels reassured.

Reassurance is not verification.

It may be the most important sentence of this excavation.

The Research Takes Us One Layer Deeper

In 2025, Pandolfo, Reader and Gillespie published a second study in Risk Analysis.

This time they analysed 45 transcripts of flightdeck communication preceding crashes and near misses, looking not at what people believed about themselves as listeners, but at what they actually said and did when safety concerns arose.

Their findings suggest that effective safety listening engages with concerns through action and collaborative sense-making, while ineffective responses either dismiss the concern or engage with it only superficially.

Such engagement can help teams develop a shared and accurate understanding of emerging risks.

The study has limits.

These were selected aviation events, the analysis was qualitative, and we should not assume that the same patterns operate identically in every factory, hospital, construction site or warehouse.

But the underlying question is highly relevant across these environments:

What happens to safety information after it reaches someone with greater authority?

Follow the journey of a single warning.

An operator notices something unusual and decides that it matters. He communicates it. The supervisor interprets the information, the organization decides whether action is needed, and someone must determine whether the response was appropriate and effective.

Every transition in that journey is an opportunity for learning, and every transition is also an opportunity for the information to be weakened, misunderstood, delayed or lost.

A reporting system can successfully transmit a warning without successfully converting that warning into protection.

A Working Name: The Conversion Gap

I have hesitated to name what this excavation is uncovering, and I still believe we should resist naming a discovery before we understand it.

But a phenomenon without a name is hard to observe, measure or discuss, so I offer a working name, a tool for looking rather than a conclusion.

I would call it the Conversion Gap: the distance between a warning received and a justified, verified response that demonstrates whether the risk is adequately controlled.

It is not the same as silence, which concerns whether information enters the system at all, and it is not the same as poor listening, which concerns the response to that information.

The Conversion Gap concerns the whole journey that follows: whether a signal from the people closest to the work is turned into understanding, into a decision, into an appropriate and verified response, and finally into an answer that returns to the person who spoke.

That response may lead to stronger safeguards, a change in working conditions, further investigation or confirmation that existing controls are sufficient.

Not every warning requires a physical change.

But every credible warning deserves a response appropriate to the potential consequences and the uncertainty involved.

Seen this way, the gap can open at any of those transitions, and it can open quietly.

The concern is acknowledged but not examined.

It is examined but not decided.

It is decided but not implemented.

It is implemented but never verified.

Or everything is done well, but the person who raised it never learns what happened, and so draws their own conclusions about whether speaking was worthwhile.

A form can be closed at every one of these points, and on paper the system will look as though it worked.

If this idea holds, it implies a different way of reading our own safety indicators.

Most organizations count how many warnings enter the system: near-miss reports, hazard observations, safety suggestions.

But the number of reports alone cannot tell us how effectively those warnings have been investigated, understood and addressed.

Perhaps we should also examine the quality of the response, the time taken to address significant risks, the evidence supporting decisions, the verification of controls and the feedback provided to those who raised concerns.

Of course, these measures would have to account for the severity, credibility and complexity of each reported situation. A minor observation and a warning about a potentially fatal hazard cannot be treated as equivalent simply because both appear in the same database.

Nor should we assume, without research, that a single conversion rate would be a better indicator of safety performance.

For now, the Conversion Gap is a question worth investigating, not a validated performance measure.

What matters is not simply how much information a system receives, but what the system becomes capable of doing with it.

The Invisible Cost of Being Heard but Not Taken Seriously

In our previous excavation we explored the human capacity that organizations develop, protect or unintentionally consume.

Now consider what happens to someone who repeatedly raises legitimate concerns and sees little meaningful response.

They may keep reporting.

Or they may gradually conclude that reporting changes nothing.

This does not have to happen through fear.

Sometimes silence grows out of disappointment, out of fatigue, or out of the quiet calculation that speaking again will cost energy without making any difference.

This is not a new observation, and it is worth acknowledging that the research has been here before us.

Pinder and Harlos described a form of acquiescent silence, rooted in resignation rather than fear. And in their exploratory study of employee silence, Frances Milliken, Elizabeth Morrison and Patricia Hewlin examined why employees choose not to communicate concerns to those above them, including beliefs about the consequences and usefulness of speaking up.

Morrison's 2023 review of the voice and silence literature shows how complex these motives remain.

What the Conversion Gap may add is a way of investigating where that resignation develops: not necessarily in the moment of speaking, but in everything that happens, or fails to happen, afterwards.

This suggests a troubling feedback loop.

An organization receives warnings but responds ineffectively.

Employees gradually lose confidence that raising concerns is worthwhile.

Fewer concerns reach decision-makers.

Management sees fewer reports, and the apparent reduction in problems may be mistaken for improved safety.

This is a plausible mechanism, not a universal law, and it needs to be investigated rather than assumed.

But it exposes a weakness in how we interpret safety indicators: a low number of reported concerns does not, by itself, prove that risks are well controlled, and a high number of reports does not, by itself, prove that an organization is learning.

There is something deeply human hidden in this process.

When someone raises a concern, they are not simply transferring information. They are placing a small part of their professional judgment, their experience and their trust into someone else's hands.

How the organization handles that trust may shape what they choose to do next time.

And the next time may matter far more than the first.

A Safety System That Can Be Changed by What It Learns

Most organizations already have extensive reporting processes: near-miss forms, hazard observation cards, safety committees, corrective action databases, management reviews.

These are valuable instruments that create structure, traceability and accountability.

The problem is not the existence of procedures.

The problem appears when completing the procedure becomes a substitute for resolving the underlying uncertainty.

A warning is received, a form is completed, an action is assigned, and the action is closed.

But was the risk understood?

Was the underlying condition changed, if change was necessary?

Was the effectiveness of the response verified?

Did the person who raised the concern receive a meaningful answer?

And did the organization learn something that could influence its future decisions?

The UK Health and Safety Executive describes effective worker consultation as a two-way process in which employees can raise concerns and influence decisions about how risks are managed.

That distinction between communication and influence brings us close to a principle:

A mature safety system is not merely one in which people can speak, but one in which relevant evidence, wherever it originates, can change decisions.

That does not mean eliminating hierarchy, replacing professional judgment with consensus, or weakening engineering safeguards, procedures and regulatory requirements.

Those protections remain essential.

It means keeping the organization's decision-making structure responsive to credible information from the people closest to the work, especially when that information contradicts what the organization expects to see.

The quality of a safety culture is revealed not only by what people are willing to say, but also by what an organization is willing to reconsider.

A Small Experiment: The Warning Trace

Before claiming that we have found a new safety principle, we should test it.

I propose a simple exercise, which I call the Warning Trace.

Select a recently reported safety concern, not necessarily a serious one: an unusual vibration, an ergonomic difficulty, unexpected equipment behaviour, a recurring deviation.

Then follow the information through the organization, asking four questions.

1. What did the person actually observe?

Not which category was ticked on the reporting form, but what changed in the real working environment.

2. What happened when the observation reached someone with authority?

Was the concern clarified, investigated or dismissed, and what information shaped the response?

3. What decision followed?

Did the organization change a control, verify that existing safeguards were adequate, or leave the situation unresolved? Was the decision justified by evidence, and was the response proportionate to the potential harm?

4. What returned to the person who spoke?

Did they understand the decision and its reasoning? Was the outcome checked? And did the experience make future communication more or less likely?

These questions are not yet a validated scientific instrument.

They are a practical starting point for investigation.

But traced across several concerns, they may help make the Conversion Gap visible, revealing something conventional reporting statistics often hide: the distance between information received and a justified, verified response to the risk.

The exercise may also reveal cases in which the system responded appropriately without changing any physical conditions, because investigation demonstrated that existing safeguards were adequate.

That distinction matters if we want to learn rather than simply generate more corrective actions.

And perhaps, over time, this kind of investigation could help us develop a better understanding of how organizations transform warnings into prevention.

What Have We Found Beneath This Layer?

Our excavation began with the person arriving at work.

We then explored how safety systems shape the person that worker becomes, and whether organizations protect or consume the human capacities on which safety depends.

Now we have uncovered another relationship.

The worker's capacity to notice and speak is only part of the picture.

The organization must also possess the capacity to receive, interpret, decide and respond, and to close the loop with the person who spoke.

The research gives us good reasons to examine this relationship, even if it does not yet establish a complete new paradigm.

For now, I would preserve one finding:

The value of a warning does not lie in the fact that it was reported, but in what the organization is able to learn and do because it was reported.

Safety does not exist only in documents, meetings or intentions.

It exists in the conditions people encounter when they return to work.

If a concern leads to better understanding, stronger controls or a verified decision that existing protections are sufficient, communication has contributed to prevention.

If it leads only to acknowledgement and administrative closure, we may have documented the warning without truly learning from it.

And perhaps the Conversion Gap offers us a way to investigate that difference.

Not as another box to tick.

Not as another number to report.

But as an invitation to look more closely at the journey between a human observation and an organizational decision.

The Next Question

If two organizations receive exactly the same warning, from equally competent workers, under otherwise similar conditions, why does one change course while the other continues?

Is the difference leadership, decision-making authority, time pressure, organizational incentives?

Is it the distance between the people who experience risk and those who make decisions about it?

Or something more fundamental about how responsibility is distributed?

We should investigate.

Because the next layer of safety may not be about giving people a voice at all, but about understanding who has the power to turn that voice into protection, and whether the organization has been designed to let that happen.

The excavation continues.